Existence of Novel Branched Side Chains Containing -1,2 and -1,6 Linkages Corresponding to Antigenic Factor 9 in the Mannan of Candida guilliermondii

Isolation of β-linkage-containing side chain oligosaccharides from the mannan of Candida gilliermondii IFO 10279 strain has been conducted by acetolysis under mild conditions. A structural study of these oligosaccharides by one- and two-dimensional NMR and methylation analyses indicated the presence of extended oligosaccharide side chains with two consecutive β-1,2-linked mannose units at the nonreducing terminal of α-linked oligosaccharides. The linkage sequence present in this mannan, Manβ12Manα13Manα, has also been found in the mannan of Saccharomyces kluyveri but not in the mannan of Candida species. Furthermore, these oligosaccharides are branched at position 6 of the 3-O-substituted mannose units as follows. and The H-1 signals of the mannose units substituted by a 3,6-di-O-substituted unit showed a significant upfield shift (Δ = 0.04-0.08 ppm) due to a steric effect. The inhibition of an enzyme-linked immunosorbent assay between the mannan of C. guilliermondii and factor 9 serum with oligosaccharides obtained from several mannans indicated that only the oligosaccharides with the above structure were active, suggesting that these correspond to the epitope of antigenic factor 9.

The H-1 signals of the mannose units substituted by a 3,6-di-O-substituted unit showed a significant upfield shift (⌬␦ ‫؍‬ 0.04 -0.08 ppm) due to a steric effect. The inhibition of an enzyme-linked immunosorbent assay between the mannan of C. guilliermondii and factor 9 serum with oligosaccharides obtained from several mannans indicated that only the oligosaccharides with the above structure were active, suggesting that these correspond to the epitope of antigenic factor 9.
In an earlier paper (15), we demonstrated the presence of a third type of a ␤-1,2-linked mannose unit attaching to an ␣-1,3linked one in the cell wall mannan of Saccharomyces kluyveri and speculated on the presence of the same type of ␤-1,2-linked unit in those of C. albicans serotype A and Candida guilliermondii based on the presence of characteristic H-1-H-2-correlated cross-peaks in their two-dimensional HOHAHA 1 spectra.
There are several reports on the responsibility of ␣-linked side chains (16), ␤-linked ones (9), or complex side chains with ␣and ␤-linkages (17) of cell wall mannan for the adherence of C. albicans cells to host cells in the initial step of Candida infection. Furthermore, mannans or mannooligosaccharides of C. albicans cells are known to stimulate cytokine production (18 -23). Therefore, the identification of the third type of ␤-1,2 * 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  linkage containing mannan side chains is important for understanding the pathogenecity of C. guilliermondii and its accurate serodiagnosis.
C. guilliermondii, which is one of the causes of human candidiasis in immunocompromised hosts, has antigenic factors 1, 4, and 9 (24,25). Although the structure corresponding to antigenic factor 4 was recently identified to be the following (26), there is no report of the chemical structure corresponding to antigenic factor 9 except a study by Ataoglu et al. (27). They showed that the factor 9 serum reacts with Saccharomyces cerevisiae X2180-1A-5 (mnn2) mutant strain cells, which have a linear ␣-1,6-linked mannan corresponding to the backbone in their cell wall. Therefore, we tried to detect a novel side chain corresponding to antigenic factor 9 in the cell wall mannan of a pathogenic yeast C. guilliermondii. For fragmentation of the mannan, we applied a mild acetolysis, which selectively cleaves backbone ␣-1,6 linkages to retain ␤-1,2 and branched ␣-1,6 linkages as well as ␣-1,2 and ␣-1,3 linkages (4,15,26). Structures of the resultant oligosaccharides were determined by oneand two-dimensional NMR techniques. Consequently, we demonstrated the presence of the third type of ␤-1,2 linkage containing side chains in Candida species mannan corresponding to antigenic factor 9.
Preparation and Acetolysis of Mannan-Yeast cells were grown at 28°C in shaking liquid culture containing 0.5% yeast extract, 1% peptone, and 2% glucose. Mannan was extracted from the cells with water at 135°C for 3 h and was separated by precipitation with Fehling's solution (26). The mannan prepared from the cells of C. guilliermondii IFO 10279 strain was designated as fraction G. Acetolysis under conventional (31) and mild (4) conditions was performed as described in a preceding paper (26). Fractionation of the resultant mannooligosaccharide mixture was achieved using a column (2.5 ϫ 100 cm) of Bio-Gel P-2 (extra fine). Elution was carried out with water, and aliquots of eluates were assayed for carbohydrate content by the phenol-sulfuric acid method (32). Separation of higher oligosaccharides by HPLC was carried out with a column (10 ϫ 500 mm) of YMC-Pack PA-25. Elution was done with a 52:48 (v/v) mixture of CH 3 CN and water, and the eluates were monitored with a differential refractometer. Eluates corresponding to each peak were rechromatographed on the same column.
Nuclear Magnetic Resonance Spectroscopy-All 1 H NMR experiments were performed with a JEOL JNM-GSX 400 spectrometer at 400 MHz. The spectra were recorded using a 1% (w/v) solution of each mannan or oligosaccharide in 0.7 ml of D 2 O at 45°C. Acetone (2.217 ppm) (33) was used as the internal standard for 1 H NMR.
Inhibition Test of Enzyme-linked Immunosorbent Assay-Enzymelinked immunosorbent assay was conducted as described in a preceding paper (13). Enzyme-linked immunosorbent assay-inhibition test using a factor serum was basically conducted as described by Okawa et al. (34). A haptenic oligosaccharide solution (50 ml) was mixed with 5-folddiluted factor 9 serum (50 ml) and preincubated for 2 h at 25°C. The reaction mixture was then added to the wells of a fraction G-coated microtiter plate and incubated for 2 h at 25°C. After washing, 1000fold-diluted goat anti-rabbit IgG antibody peroxidase-conjugate was added to the wells and kept for 2 h at 25°C. Finally, a substrate solution of 0.01% o-phenylenediamine and 0.006% H 2 O 2 in 150 mM citrate buffer (pH 5.0) (100 ml) was added, and the color was measured at 492 nm after the addition of 2 M H 2 SO 4 (50 ml).
Methylation Analysis-Methylation of oligosaccharides was performed according to Ciucanu and Kerek (35). Gas chromatography of O-methyl-O-acetyl-D-mannitols was performed using a glass column (3 mm ϫ 200 cm) containing 3% OV-210 on Supelcoport (100 -200 mesh) at 185°C using N 2 as the carrier gas at a flow rate of 20 ml/min.
Other Methods-For ␣-mannosidase treatment, the mannooligosaccharide mixture (200 mg) was dissolved in 50 mM sodium acetate buffer (pH 4.6) (2 ml) containing 20 units of ␣-mannosidase. After incubation for 48 h at 37°C, the reaction mixture was boiled for 5 min to inactivate the enzyme. Total carbohydrate was determined by the phenol-sulfuric acid method of Dubois et al. (32) with D-mannose as the standard. Total FIG. 1. Elution patterns of oligosaccharides obtained from fraction G by acetolysis. A, B, elution was performed with a column (2.5 ϫ 100 cm) of Bio-Gel P-2 before (q) and after (E) ␣-mannosidase treatment. A, acetolysis was performed with (CH 3 CO) 2 O/CH 3 COOH/ H 2 SO 4 (10:10:1, v/v/v) at 40°C for 12 h (conventional conditions). B, acetolysis was performed with (CH 3 CO) 2 O/CH 3 COOH/H 2 SO 4 (100:100:1, v/v/v) at 40°C for 36 h (mild conditions). C, elution pattern of ␣-mannosidase-treated acetolysate B by HPLC with a column of YMC-Pack PA-25. AM 2 -AM 7 in panel A indicate mannobiose to mannoheptaose obtained by the conventional acetolysis. BM 6 -BM 10 in panel C indicate mannohexaose to mannodecaose obtained by the mild acetolysis followed by ␣-mannosidase treatment.
phosphate was determined by the method of Ames and Dubin (36), using KH 2 PO 4 as the standard.

RESULTS
Preparation of Mannan-The cell wall mannan prepared from the crude extract of C. guilliermondii IFO 10279 strain cells by the Fehling's solution method was designated as fraction G. Clearly, this mannan contains no phosphodiesterified oligosaccharides, which have been found in the mannan of C. albicans (1,26), judging from the lack of an H-1 signal at about 5.55 ppm on 1 H NMR corresponding to the 1-O-␣-phosphorylated mannose unit (1,13,37), unreactivity to factor 5 serum (25,38), and a negative result for phosphate analysis.
Acetolysis of Mannan- Fig. 1A shows the elution pattern of the acetolysate of fraction G from the Bio-Gel P-2 column obtained under the conventional conditions. Each oligosaccharide from biose to heptaose was rechromatographed by the same column and designated as AM 2 to AM 7 . To obtain side chain oligosaccharides with ␤-1,2-linked mannose units and ␣-1,6-linked branching ones, we also carried out the mild ace-tolysis of fraction G. By this fragmentation, oligosaccharides up to decaose were obtained (Fig. 1B). These oligosaccharides were then digested with ␣-mannosidase, and the enzyme-resistant oligosaccharides were applied on a column of Bio-Gel P-2. However, it was difficult to separate the higher oligosaccharides from each other (Fig. 1B). Therefore, these oligosaccharides were separated by HPLC and rechromatographed by the same column. The oligosaccharides from hexaose to decaose were designated as BM 6 to BM 10 .

1
H NMR Analysis of Oligosaccharides-The 1 H NMR spectra of the oligosaccharides from AM 2 to AM 4 obtained by the conventional acetolysis were the same as those obtained from the mannans of S. cerevisiae (33,39) or S. kluyveri (15,40). Namely, AM 2 and AM 3 consist only of ␣-1,2 linkages, and AM 4 contains a nonreducing terminal ␣-1,3-linked mannose unit. Because AM 5 shows signals at 5.379 and 5.042 ppm, the ␣-1,3linked mannose unit of this oligosaccharide is substituted by an ␣-1,2-linked mannose unit (40, 41) (Fig. 2). AM 6 shows an additional signal at 5.139 ppm, and the signal dimension at   7 shows signals corresponding to a ␤-1,2-linked mannose unit, 4.758 ppm, and an ␣-1,3-linked mannose unit substituted by a ␤-1,2-linked one, 5.245 ppm, both of which were found in the mannan of S. kluyveri (15). The conventional acetolysis conditions cleave ␤-1,2 and ␣-1,6 linkages; therefore, part of these oligosaccharides seems to correspond to a degradation product of the parent side chains.
On the other hand, the 1 H NMR spectra of oligosaccharides higher than pentaose obtained by the mild acetolysis commonly show signals at about 4.84 ppm corresponding to two consecutive ␤-1,2-linked mannose units (3,4,15). As shown in the preceding papers (4,15), the H-1 proton of an ␣-1,2-linked mannose unit substituted by a consecutive ␤-1,2-linked one, 5.138 ppm, appears at about 0.02 ppm upfield from that substituted by a single ␤-1,2-linked one, 5.160 ppm. Therefore, the signal at 5.236 -5.245 ppm of the H-1 proton of an ␣-1,3-linked mannose unit substituted by a single ␤-1,2-linked mannose unit (15) seems to shift to about 5.22 ppm with the addition of consecutive ␤-1,2-linked units. Because BM 6 shows a signal corresponding to an ␣-1,2-linked mannose unit substituted by an ␣-1,3-linked mannose unit, 5.033 ppm, it is reasonable to assign the signal at 5.218 ppm to the ␣-1,3-linked mannose unit substituted by consecutive ␤-1,2-linked mannose units. Therefore, we can propose that the chemical structure of BM 6 is as follows. BM 8 shows new signals at 5.223, 4.846, and 4.838 ppm in addition to those of AM 6 . This indicates that the structure of BM 8 was that of AM 6 with two ␤-1,2-linked mannose units at the nonreducing terminal as follows.
It is obvious that BM 9 and BM 10 contain one and two ␣-1,6linked mannose units, respectively, judging from the dimension of the signals at about 4.91 ppm. In the spectrum of BM 9 , about two-thirds of the signal at 5.368 ppm corresponding to Man-D seems to be shifted upfield to 5.311 ppm by the addition of an ␣-1,6-linked mannose unit to Man-E of BM 8 . Furthermore, about one-third of the signal at 5.269 ppm corresponding to Man-B is also shifted upfield to 5.231 ppm as the result of the attachment of an ␣-1,6-linked mannose unit to Man-C. Namely, BM 9 seems to be a mixture of two isomers with a difference in the branching point. Finally, the two ␣-1,6-linked mannose units of BM 10 seem to attach on Man-C and Man-E judging from the presence of two upfield-shifted signals at 5.291 and 5.226 ppm.
Determination of Branching Point-From the structural study of S. kluyveri (15) and C. albicans (26) mannans, we can speculate that the ␣-1,6 branching point of the side chain of C. guilliermondii mannan is the 3-O-substituted mannose units. Although the H-1 signal of a mannose unit does not shift by phosphorylation or glycosylation at the O-6 position, the attachment affects the chemical shift of some ring proton signals allocated around the substituted position (15,26,(42)(43)(44). Therefore, to detect the branching point of oligosaccharides, we recorded one-and two-dimensional HOHAHA spectra, and we found that the shifted ring protons correlated only with the 3-O-substituted mannose units of BM 7 , BM 9 , and BM 10 compared with those of BM 6 and BM 8 (data not shown). These findings demonstrate that the branching point of these oligosaccharides is the 3-O-substituted mannose unit. This result was confirmed by methylation analysis of the oligosaccharides as shown in Table I. Namely, BM 7 , BM 9 , and BM 10 gave 2,4di-O-methyl-1,3,5,6-tetra-O-acetyl mannitol, which corresponds to the 3,6-di-O-substituted mannose unit. On the other hand, 2,4,6-tri-O-methyl-1,3,5-tri-O-acetyl mannitol, which corresponds to the 3-O-substituted mannose unit, was obtained from BM 6 , BM 8 , and BM 9 but was not obtained from BM 7 and BM 10 .
Sequential NMR Assignment-A sequential assignment study of the H-1 and H-2 signals of these oligosaccharides was performed to confirm the structure by the method described by Hernandez et al. (42) with slight modification (8,26). The right side of the diagonal of each panel in Fig. 3 shows the relayed COSY, whereas the left side shows the rotating frame NOE spectroscopy. In this figure, cross-peaks labeled with primed letters indicate through-space interresidue H-1-H-2Ј or H-1-H-3Ј connectivities between two adjacent mannose units. On the other hand, cross-peaks labeled with unprimed letters indicate intraresidue H-1-H-2-or H-1-H-3-correlated cross-peaks caused by J-coupling. By this procedure, the H-1 and H-2 signals of BM 6 were sequentially assigned from the H-1 of the Man-A, A-AЈ-B-BЈ-C-CЈ-D-DЈ-E-EЈ-F (Fig. 3). Similarly, the H-1 and H-2 signals of BM 7 , BM 8 , BM 9 , and BM 10 were assigned sequentially (Fig. 3). Because an ␣-1,3 linkage gives a weak H-1-H-2Ј NOE cross-peak in addition to a strong H-1-H-3Ј NOE cross-peak, we can use the H-1-H-2Ј NOE cross-peak cЈ instead of the H-1-H-3Ј NOE cross-peak CЈ for the assignment or to confirm the connectivity. Especially for BM 10 , two H-1-H-2correlated cross-peaks c and e of 3-O-substituted mannose units, Man-C and Man-E, were better separated than the corresponding two H-1-H-3-correlated ones, C and E. Therefore, it was effective to use the cross-peaks c and e and H-1-H-2Ј NOE cross-peaks cЈ and eЈ.
The results summarized in Table II clearly demonstrate that the attachment of an ␣-1,6-linked mannose unit to Man-C and Man-E causes an upfield shift of the H-1 signals of Man-B and Man-D, respectively, due to a steric effect (15,26).
Two-dimensional Homonuclear Hartmann-Hahn Spectroscopy of Mannan-The molar ratio of the mannan side chains was calculated from the dimensions of the H-1 and H-2 signals of fraction G (Fig. 4A) based on the assignment result of the cross-peaks on the two-dimensional HOHAHA spectrum (Fig.  4B) followed by the method described previously (15). As shown in Table III, the cross-peaks 1 and 3 in the two-dimensional HOHAHA spectrum of fraction G were assigned to correspond   Table II. Cross-peaks 4 and 6 correspond to ␣-1,3-linked mannose units substituted with a ␤-1,2 linkage by mannose and ␤-1,2-linked mannobiose unit(s), respectively. On the other hand, cross-peaks 7 and 8 correspond to ␣-1,2-linked mannose units substituted by ␤-1,2-linked one(s) (15). The absence of these cross-peaks indicates that this mannan has no side chains corresponding to the C. albicans serotype A-specific epitope, antigenic factor 6. Cross-peak 12 indicates the presence of the 3-O-substituted ␣-1,2-linked mannose unit. Although the H-1 signal at 5.037 ppm overlaps those of crosspeaks 11 and 12, the H-2 signal at 4.213 ppm corresponds only to the cross-peak 12. Therefore, we can determine the molar ratio of the 3-O-substituted ␣-1,2-linked mannose units or that of the ␣-1,3-linked ones in mannan as shown in Table III. To distinguish the NMR spectra of the two kinds of ␣-1,6-linked mannose units, branch and backbone forming ones, the signals of the ring protons of both mannose units were compared. Fig.  4C indicates the one-dimensional HOHAHA spectrum of BM 7 recorded by the irradiation of the signal at 4.914 ppm corresponding to the branching ␣-1,6-linked mannose unit. On the other hand, Fig. 4D shows a normal 1 H NMR spectrum of a linear ␣-1,6-linked backbone mannan prepared from S. cerevisiae X2180-1A-5 (mnn2) mutant strain cells. As shown in Fig.  4C, the H-4 signal of the former mannose unit appeared at a characteristic region, 3.66 -3.68 ppm, and gave cross-peaks 18 and 19 corresponding to the H-1-H-4-and H-2-H-4-correlated ones, respectively. This finding suggests that the cross-peaks 13, 18, and 19 correspond to the branching ␣-1,6-linked mannose unit, and cross-peak 14 corresponds to the backbone ␣-1,6linked mannose unit. Although it is difficult to determine the dimension of the H-1 signal of the branched mannose unit because of the overlapping of cross-peaks 13 and 14, we can estimate it from the dimensions of the H-1 signals of cross-peaks 2 and 5. Because the H-1 signal dimensions of cross-peaks 4 and 6 are the same as those of cross-peaks 17 and 16 (half of the signal at 4.849 ppm), respectively, the H-1 signal dimension of cross-peak 5 can be determined by subtraction of the H-1 signal dimension of crosspeaks 4 and 6 from that of the signal at 5.218 ppm as shown in Table III. These results indicate that the amount of the ␣-1,6linked branching mannose units is slightly smaller than that of the total ␣-1,3-linked ones but is sufficient to attach to all of the 3-O-substituted ones in the ␤-1,2 linkage-containing side chains. Namely, BM 6 , BM 8 , and BM 9 correspond to the degradation products of BM 7 or BM 10 on mild acetolysis. From these results, we can propose the chemical structure of the cell wall mannan of C. guilliermondii IFO 10279 strain as shown in Fig. 5.

DISCUSSION
In 1988, Kogan et al. (45) reported the presence of a 2,3-di-O-substituted mannose unit in the side chain of C. albicans and C. guilliermondii mannans based on the results of methylation analysis of polysaccharides. Later, Kagaya et al. (38) suggested the presence of a branching structure (46) in the C. guilliermondii mannan from the cross-reactivity of a monoclonal antibody against factor 4. Recently, we found that the antigenic factor 4 corresponds to an ␣-1,6-branched side chains of the mannan with a comb-like structure (26).
Ataoglu et al. (27) reported that the antigenic factor 9 corresponds to a consecutive ␣-1,6-linked mannose unit from the reactivity of factor 9 serum to the cells of the S. cerevisiae X2180-1A-5 (mnn2) mutant strain, which have linear ␣-1,6linked mannan. It is true that fraction G exposes about 50% of the ␣-1,6-linked backbone mannose units (Fig. 5). Because factor 9 serum was prepared simply by absorption of the anti-C. guilliermondii whole cell serum with C. albicans serotype A cells (24), it is reasonable to expect that it contains antibodies against several epitopes, including the backbone mannose units. In this study, however, we could demonstrate the existence of novel side chains containing a third type of ␤-1,2-linked mannose unit, BM 7 and BM 10 , as the specific structure for C. guilliermondii mannan corresponding to antigenic factor 9.
In 1981, Zhang and Ballou (40) reported the presence of O-linked branching mannooligosaccharides up to octaose in S. kluyveri mannoprotein. In the study, they analyzed the structure of oligosaccharides by the methylation technique. However, because it is impossible to determine the linkage se- quence from the methylation analysis data, they proposed the structure of mannopentaose to mannooctaose based on that of the shorter ones. On the other hand, the sequential assignment method of oligosaccharides through HMBC (15,47) or NOE (8,26,42) cross-peaks between the glycosylated two mannose units was demonstrated to be suitable for assigning and determining the linkage sequence by this and the preceding studies.
The upfield shift effect of the H-1 signal of an ␣-1,2-linked mannose unit substituted by a 3,6-di-O-substituted one, the effect of which was first found on the branched side chains of the mannan of S. kluyveri (⌬␦ ϭ 0.047 ppm) (15) and later on that of C. albicans (⌬␦ ϭ 0.055 ppm) (26), was observed on BM 7 , BM 9 , and BM 10 (⌬␦ of Man-B ϭ 0.044 -0.049 ppm). In this study, we also found the same effect of the H-1 signal of an ␣-1,3-linked mannose unit, Man-D, substituted by a 3,6-di-Osubstituted one (⌬␦ of Man-D ϭ 0.062 ppm for BM 9 and 0.082 ppm for BM 10 ). The large upfield shift effect found in Man-D of BM 10 seems to be due to the attachment of two ␣-1,6-linked  M denotes a D-mannopyranose unit. The side-chain sequence is not specified. The molar ratio of the side chains in the mannan is expressed as a percentage of the total side chains. The values are calculated from the dimensions of the 1 H NMR signals in Fig. 4. mannose units to the neighboring 3-O-substituted ones of Man-D at the reducing and the nonreducing sides. The upfield shift of the H-1 signals seems to be the result of a steric effect; therefore, it is of interest to identify the conformation of these oligosaccharides.
From the results of this and the preceding (12,26) structural studies of the mannans containing ␣-1,6-branched side chains, we can speculate that the ␣-1,6-mannosyltransferase responsible for the biosynthesis of branched side chains requires oligosaccharides containing an ␣-1,3 linkage as an acceptor. This hypothesis is supported by the results of Pang et al. (48). They reported that the mannan of an S. kluyveri (mnn1) mutant strain that lacks ␣-1,3-mannosyltransferase activity also lacks the branching ␣-1,6-linked mannose unit. To understand the timing of the transfer of an ␣-1,6-linked mannose unit to the side chain, however, we need to determine the substrate specificity of the ␣-1,6-mannosyltransferase. Recently, we detected a ␤-1,2-mannosyltransferase responsible for the synthesis of the second type of ␤-1,2 linkage (antigenic factor 6) (49). Now we are interested in an ␣-1,6-mannosyltransferase and a ␤-1,2mannosyltransferase responsible for the synthesis of the branch and the third type of ␤-1,2 linkage (antigenic factor 9), respectively.