Recombinant human bile salt–stimulated lipase (hBSSL) was expressed in and secreted by Pichia pastoris, an organism exploited for the large-scale production of recombinant (glyco)proteins by bioprocessing technology. The 76.3-kDa glycoprotein was associated with 75–80 Man and a small amount of GlcNAc. hBSSL has one N-glycosylation site at Asn187, which was 38–40% occupied with a Man10GlcNAc2 structure defined previously in Pichia as the oligosaccharide-lipid form of Man9GlcNAc2 trimmed of the middle-arm terminal α1,2-Man and elongated with Manα1,2Manα1,6-disaccharide attached to the lower-arm core α1,3-Man (Trimble et al. , J. Biol. Chem., 266, 22807–22817). The C-terminal 192 residues of hBSSL contain 16 Pro-rich 11-amino-acid repeats, which include 32 Ser/Thr residues as potential O-glycosylation sites. Using hBSSL as a platform to study Pichia's O-glycosylation capabilities, we found that nearly all of these sites were occupied by mannose-containing O-glycans, whose structures, after β-elimination and purification, were assigned by 1H NMR and, in some cases, by linkage-specific exoglycosidases and methylation analysis. The most abundant O-glycan was α1,2-mannobiitol (55%), followed by α1,2-mannotriitol (16%) and mannitol (10%) and a lesser amount was α1,2-mannotetraitol. Unexpectedly, Man5 and Man6 O-glycans were present, which had the structure Manβ1,2Manβ1,2Manα1,2(Manα1,2)1,2mannitol. Also a small amount of a phosphorylated Man6 O-glycan was characterized by MALDI-TOF MS postsource decay analysis as having the reducing-end mannitol disubstituted with a glycosidically linked phosphorylated Man and an unbranched Man4 polymer elongated from a different mannitol carbon. This is the first report of the synthesis of β-Man- and phosphate-containing O-linked constituents on glycoproteins synthesized by P. pastoris.
Pancreatic human bile salt–stimulated lipase (hBSSL) works in concert with colipase-dependent pancreatic lipase to digest and assimilate triglycerides in adults of many species. Both enzymes are present in low levels in newborn humans, which leads to malabsorption of fats in some formula-fed neonates. Human milk triglycerides, which initially may account for 50% of a breast-fed infant's dietary energy supply, are efficiently utilized, even in preterm infants, due to the presence of hBSSL in the mother's milk (reviewed in Hernell and Bläckberg, 1994).
The hBSSL gene was cloned from a mammary gland cDNA library (Baba et al., 1991; Nilsson et al., 1990), and sequencing revealed the human form to be 742 aa in length, with the first 20 residues removed in vivo as a cleavable signal sequence. The catalytic domain, consisting of the first 530 residues, is highly conserved in nature, with an N-glycosylation site at Asn187. The C-terminal tail region in hBSSL, consisting of residues 531–722, contains 16 repeats of a proline-rich 11-residue sequence within residues 538–713. Different species have different numbers of the mucin-like C-tail repeat units. hBSSL is heavily glycosylated, and a number of studies have characterized both the N- and O-glycoforms of the human enzyme (Landberg et al., 1997; Mechref et al., 1999; Wang et al., 1995), including natural variants (Strömqvist et al., 1997) as well as changes in glycoforms during lactation (Landberg et al., 2000).
Recombinant DNA constructs of human BSSL reveal that neither elimination of the N-glycosylation consensus sequence at Asn187 by an N187Q substitution nor removal of the heavily O-glycosylated tail region's 16 proline-rich 11-residue repeats negatively effects the expression, activation by bile salts, catalytic activity, or substrate specificity of the enzyme (Bläckberg et al., 1995). The dispensability of either N- or O-glycosylation is confirmed by the expression of active full-length and C-terminal tail-free constructs in Escherichia coli (Hansson et al., 1993). Thus the role of the extensive O-glycosylation is unproven, but the large C-terminal tail should be very hydrophilic and accessible (Wang et al., 1995), factors that may relate to its retention in the intestinal track in vivo.
Efficient production of a recombinant form of hBSSL by Pichia pastoris with the enzymatic properties of the native protein for potential use as an infant formula supplement has been described (Sahasrabudhe et al., 1998). Pichia pastoris has been used extensively by biopharma for the large-scale production of glycopharmaceuticals, but little is known about this organism's ability to synthesize other than short α1,2-linked mannose O-glycans (reviewed by Bretthauer and Castellino, 1999). Thus the abundance of this recombinant full-length hBSSL provided a platform to explore the capacity of P. pastoris to form O-glycans with potentially novel linkage arrangements. Like many other yeast, Pichia synthesizes α1,2Man1–4 O-linked glycans. Unexpectedly, however, a portion of the α1,2-Man3 and α1,2-Man4 chains were found to be capped at the nonreducing end with a Manβ1,2Manβ1,2-disaccharide, and a small amount of a branched Man6-phosphorylated O-glycan also was identified.
Results and Discussion
Characterization of purified hBSSL secreted by P. pastoris
The full-length hBSSL gene was cloned into vector pHILD4 to form pARC5799, which was used to transform P. pastoris for high-level expression and secretion of hBSSL (Sahasrubudhe et al., 1998). The hBSSL protein present in the fermentor broth was purified as described (Sahasrubudhe et al., 1998) and served as the starting material for the current studies on hBSSL glycosylation by Pichia.
Figure 1 shows the sodium dodecyl sulfate (SDS) 8% polyacrylamide gel electrophoresis (PAGE) pattern of purified hBSSL stained with Coomassie blue. The full-length protein without its cleavable signal sequence has a predicted mass of 76.3 kDa (ENTREZ accession number P19835), and the protein migrates at about 102 kDa. Previous studies have shown that the Pichia-expressed hBSSL binds the lectin conacanavalin A, indicating the presence of Man. Acid hydrolysis and Dionex high-performance anion exchange chromatography (HPAEC) revealed Man as the only hexose present, with a small amount of GlcNAc associated with hBSSL, and phenol sulfuric acid analysis using Man as a standard revealed 75–80 Man/peptide (data not shown).
N-linked glycosylation of Pichia-secreted hBSSL
There is one potential N-glycosylation sequon at Asn187 in the hBSSL sequence (Baba et al., 1991), and the presence of GlcNAc by Dionex HPAEC suggested that the site was at least partially occupied. However, endo β-N-acetylglucosaminidase H (Endo H) treatment of the hBSSL had little effect on the glycoprotein's apparent migration (Figure 1). Nevertheless, solvent precipitation and methanol extraction (Verostek et al., 2000) of the Endo H–digested sample provided an N-glycan fraction, over 80% of which eluted as Man10GlcNAc on a calibrated Bio-Gel P-4 column (Figure 2). This assignment was confirmed by the single matrix-assisted laser desorptin/ionization time-of-flight mass spectrometry (MALDI-TOF MS) signal at 1865 Da for the sodium adduct of Hex10GlcNAc in the pooled fractions 127–133 (Figure 2, inset). Recovery of the Endo H–released N-glycans from three separate trials with this batch of purified hBSSL indicated an Asn187 site occupancy of 38–40%. Small amounts of Man8,9,11,12GlcNAc were also present in the profile (Figure 2).
Interestingly, in a previous study on recombinant Saccharomyces cerevisiae invertase (Suc2p) expressed in Pichia (Trimble et al., 1991), the same range of Man8–12 GlcNAc N-glycan sizes was found to be present, but the distribution comprised nearly equal amounts of Man8,9GlcNAc, representing 75% of the N-glycans present, with Man10GlcNAc constituting only 18% of the total. To assess whether the Man10GlcNAc present on hBSSL had the same structure as that on Suc2p, we analyzed the pooled Man10GlcNAc by high-field 1H nuclear magnetic resonance (NMR) (data not shown). The hBSSL structure (Scheme 1), which includes residue numbers and their nominal anomeric proton chemical shifts (δ, ppm), is identical to the Suc2p Man10GlcNAc pool characterized earlier (Trimble et al., 1991). This isomer is the Man8GlcNAc core derived by trimming in the endoplasmic reticulum of three Glc and the middle-arm α1,2-Man from the Glc3Man9GlcNAc2 precursor initially transferred to nascent proteins, followed by outer-chain mannan initiation with α1,6-Man residue 12 added by Och1p and capped with α1,2Man residue 13 in the Golgi apparatus (reviewed in Gemmill and Trimble, 1999a).
O-glycosylation of hBSSL synthesized by Pichia
After Endo H digestion to remove N-glycans, the hBSSL O-glycans were β-eliminated in the presence of sodium borohydride, which converts the reducing-end Man to Man-ol. The pool of O-glycans, prepared as described in Materials and methods, was chromatographed on the calibrated Bio-Gel P-4 column with an included internal marker of 14C-Man-ol eluting in peak 1 (Figure 3). The phenol–sulfuric acid assay profile revealed five additional peaks, which were collected as pools 2–6 by combining the fractions indicated by the bars over the peaks in Figure 3. These represented ManMan-ol through Man5Man-ol, whose masses were confirmed by MALDI-TOF MS (Table I). The Man-ol in pool 1 was quantitated by ratio analysis of the 14C in the profile with the recovery of Man-ol by Dionex HPAEC on an MA1 analytical column using an authentic Man-ol standard for calibration.
|Pool no.a||Glycan sizeb||Measuredc||Calculated||No. Chains/hBSSLd||No. Man/hBSSL|
|Σ = 27–30||Σ = 66–78|
|Pool no.a||Glycan sizeb||Measuredc||Calculated||No. Chains/hBSSLd||No. Man/hBSSL|
|Σ = 27–30||Σ = 66–78|
Pools 1–6 as identified in P-4 profile in Figure 3.
Man-ol quantitated by HPAEC, others by phenol–sulfuric acid assay.
By MALDI-TOF MS in the positive ion mode.
Average of three hBSSL preparations analyzed.
MALDI-TOF MS and Dionex HPAEC analysis of hBSSL O-glycans
Preliminary MALDI-TOF MS analysis of pools 2–6 revealed some cross-contamination in pools 4 and 5, so these were rechromatographed on Bio-Gel P-4 and the central 85% of the phenol–sulfuric acid assay absorbance for each peak was pooled (data not shown). The final pools 2–6 were analyzed by MALDI-TOF MS, and all but pool 4 revealed a single molecular weight species representative of ManxMan-ol, as summarized in Table I. The additional glycan in pool 4 is a phosphorylated Hex5Hex-ol, which must have a compact solution structure to elute with a Hex4Hex-ol species on gel permeation chromatography.
Each of the pools was analyzed by Dionex HPAEC on a PA1 analytical column; the resultant profiles are shown as a montage in Figure 4. Each pool has one main component with traces of species coeluting with the main components in adjacent peaks of the Bio-Gel P-4 column (Figure 3). The absence of a novel peak in the Dionex trace for pool 4 (Figure 4) suggests that the phosphorylated glycan in this pool (Table I) is a minor component.
On the basis of phenol–sulfuric acid assay absorbance values for three separate β-elimination experiments with this batch of hBSSL, the recovery of O-linked glycans in a typical Bio-Gel P-4 profile (Figure 3) was 74–92% of the starting material, depending on the amount of hBSSL treated, column pooling, and assay losses. With the recoveries normalized, Table I summarizes the number of Man1–5 Man-ol chains present on an “average” hBSSL molecule. Also included is the monosaccharide value assayed as Man-ol. Note that the distribution of chain sizes (Table I) is in agreement with the starting calculation of about 75 O-linked Man/hBSSL. The C-terminal proline-rich 11-amino-acid repeats in residues 538–713 have a total of 32 Ser and Thr residues as potential O-glycosylation sites. The assignment of 27–30 O-glycans/hBSSL (Table I) means that 85–95% of these residues are glycosylated.
Postsource decay (PSD) MALDI-TOF MS analysis of pool 4 phosphorylated hexasaccharide alditol
The unusual phosphorylated O-glycan in pool 4 was present in such a small amount that no effort was made to isolate it from the major Man3Man-ol component for 1H NMR analysis. However, some structural information could be deduced from the PSD MALDI-TOF MS spectrum of the 1095-Da ion shown in Figure 5.
The relationship of parent/daughter ions is consistent with the structure included in Figure 5. Table II summarizes the measured and calculated masses of the fragments and their structures. A second, minor Man5(H2PO3Man)Man-ol isomer appears to be present, in which the PO4 may be on the penultimate Man of the compound; this would account for the small signal at 428 Da for H2PO3-Man2. Note that this fragment cannot be readily obtained from the major isomer shown in Figure 5, even by multiple fragmentations.
|m/z observed||m/z calculated||Identity|
|m/z observed||m/z calculated||Identity|
Assigned to a minor isomer.
The major Man4(H2PO3Man)Man-ol isomer would be expected to be composed solely of α/β1,2- and α1,6-linkages on the basis of the O-glycan structures in pools 2–6 determined by 1H NMR, previous studies on Pichia glycans (Kobayashi et al., 1986), and this organism's resident glycosyltransferases (Verostek and Trimble, 1995). It is interesting to note that the secretion leader sequence of the S. cerevisiae BAR1 protease gene product is heavily O-glycosylated and that many of the O-mannosyl chains bear terminal phosphate residues (Jars et al., 1995). This is in contrast to the Man phosphate in diester form found on S. cerevisiae N-glycans (Hernández et al., 1989).
Structure of the neutral O-glycans in pools 2–6
ManMan-ol and Man2Man-ol. The most abundant O-linked oligosaccharides on recombinant hBSSL are the di- and trisaccharides, which represent nearly 60% and 20% of the total, respectively (Table I). Both the 1D and 2D double quantum filtered (DQF)–correlation spectroscopy (COSY) spectra (not shown) of pool 2 and 3 constituents were identical to spectra recently published for Manα1,2Man-ol and Manα1,2Manα1,2Man-ol β-eliminated and characterized from Schizosaccharomyces pombe glycoproteins (Gemmill and Trimble, 1999b). These O-glycans are also common to Hansenula polymorpha (Cohen and Ballou, 1980), S. cerevisiae (Ballou, 1990), and Saccharomyces kluyveri (Zhang and Ballou, 1981) glycoproteins. Both glycans were hydrolyzed to free Man and Man-ol in the expected ratios by Aspergillus satoi α1,2-mannosidase (not shown), in agreement with their proposed structures.
A 2D relayed rotating-frame Overhauser spectroscopy (ROESY) experiment (Cipollo et al., 2000) on the pool 4 O-glycan (Figure 6A) revealed three anomeric protons at chemical shifts of 5.057, 5.243, and 5.335 ppm. These correlated with C2-Hs at 4.091, 4.007, and 4.132 ppm, respectively, with J1,2 coupling constants of ∼1.5 Hz characteristic of α1,2-linked Man. The experiment (Figure 6A) shows the linkage of residues d → c → b through cross-peaks between d's H1 (5.057 ppm) and c's H1 (5.335 ppm) and b's H1 (5.243 ppm), thus providing a proof of the (Manα1,2)3Man-ol assignment. Note the similarity of the anomeric proton chemical shifts for residues d and c in the pool 4 O-glycan with the α1,2-linked lower arm Man residues 11 and 8 on the N-linked Man10GlcNAc in Scheme 1. A reduced α1,2-linked Man tetrasaccharide released from Candida parapsilosis cell wall mannan by acetolysis and chemically defined by methylation analysis provides an anomeric proton signal match (Funayama et al., 1983) to the chemical shift values recorded here for pool 4.
Interestingly, conversion of the reducing-end Man to Man-ol shifts the anomeric proton of the Man 2-O substituting the Man-ol (residue b) upfield by nearly 0.1 ppm (Funayama et al., 1983). A similar observation has been made by Shibata et al. (1995) on comparing reduced and nonreduced α1,2-linked mannan fragments isolated from C. albicans cell walls. The shift appears to result from increased shielding of this residue's anomeric center by the loss of the rigid ring structure of the reducing-end Man, on formation of the alditol.
The initial 1H NMR spectrum of the pool 5 O-glycan included, in addition to nominal levels of α/β1,2-Man anomeric protons, a large signal at 4.904 ppm, characteristic of α1,6-linked Man. However, the C1-H/C2-H cross-peak and coupling constants to the putative ring C2-Hs for α1,6Man were inconsistent with these being glycan protons (Vliegenthart et al., 1983). Concerned that the sample was contaminated with a residual volatile organic compound from the column chromatography buffer, we rechromatographed the pool 5 glycan on Bio-Gel P-4, lyophilized the material and again exchanged it with D2O for NMR spectroscopy. Because of the small amount of Hex5-ol remaining (see Figure 3), 1H NMR analysis was performed using a Nalorac microsample probe, which provided the 1D spectrum in Figure 6B. Note in this well-shimmed data set, the ∼1.5 Hz J1,2 splitting of the anomeric protons at ∼5.1 and ∼5.4 ppm, indicative of α-mannosides, and the lack of splitting of the anomeric protons at ∼4.85 and ∼4.9 ppm, characteristic of β-mannosides (Vliegenthart et al., 1983). C2-H chemical shifts were determined from a separate 2D DQF COSY experiment (data not shown).
Thus the pool 5 O-glycan appears to be composed of two α- and two β-mannosides with the fifth residue a mannitol. Methylation followed by MS indicated that all residues were glycosidically 1,2-linked (data not shown). Studies on acetolysis fragments from P. pastoris strain 1FO 0948 cell-wall mannan (Kobayashi et al., 1986) provided a minor Man5 compound with anomeric 1H signals for a glycan containing β1,2-linked mannoses that was identical to those recorded for the pool 5 O-glycan reported here. This in concert with the more comprehensive data generated for the more abundant pool 6 O-glycan suggests the pool 5 O-glycan has following structure:
The minor signals in Figure 6B (*) at 5.21, 5.15, and 4.88 ppm are from a residual amount of the pool 6 O-glycan in the sample and will be identified shortly.
The 1D 1H NMR spectrum of the pool 6 O-glycan revealed five anomeric protons at 5.386, 5.225, 5.148, 4.883, and 4.850 ppm (Figure 6C). The first three resonances reveal J1,2 coupling constants of ∼1.2–1.5 Hz expected for α-mannosides and are reminiscent of the (Manα1,2)3Man-ol anomeric protons seen in the pool 4 O-glycan (Figure 6A). The two upfield resonances show no splitting and duplicate the β-mannoside resonances seen in the pool 5 O-glycan (Figure 6B). A 2D DQF COSY spectrum provided the C1-H/C2-H cross-peaks for the pool 6 O-glycan listed later. Interestingly, Kobayashi and co-workers (1986) identified β1,2-Man-containing structures in acetolysis fragments from pathogenic C. albicans cell wall mannan; of particular interest among these was a pentasaccharide, Manβ1,2Manα1, 2Manα1,2-Manα1,2Manαβ (Kobayashi et al., 1989), which was used in more recent experiments as an acceptor in C. albicans extracts to assay for a nonreducing-end β1,2-mannosyltransferase (Suzuki et al., 1995). The product formed was a hexasaccharide shown by 1H NMR at 45°C to be:
The pool 6 O-glycan reveals nearly identical J1,2 cross-peaks, except that, due to residue a's reduction to an alditol, residue b's C1-H moves upfield to 5.225 ppm and residue c's C1-H moves downfield to 5.386 ppm:
These representative shifts in both b's and c's anomerics on the reduction of Man to Man-ol have been documented previously (Funayama et al., 1983; Shibata et al., 1995). In keeping with the structure shown, β-mannosidase released two Man residues and provided a Man3Man-ol tetrasaccharide alditol on Dionex PA1 chromatography. The intact hexositol was resistant to jackbean and A. satoi α-mannosidases (data not shown).
To date, only Man and α1,2-linked Man2–4 polymers have been shown to constitute the O-linked glycans β-eliminated either from Pichia cell wall mannans or from recombinant kringle 1–4 domain (Duman et al., 1998) and mouse gelatinase B produced in Pichia (reviewed in Bretthauer and Castellino, 1999). The current study on recombinant hBSSL glycosylation have revealed the capability of P. pastoris to form a wider variety of O-linked glycan structures than previously known. Expectedly, the most abundant saccharides were α1,2-Man polymers of two and three residues, with a small percentage elongated to the α1,2-linked tetrasaccharide. Surprisingly, a portion of the tri- and tetrasaccharide were elongated further to pentaose and hexaose sizes by addition of a Manβ1,2Manβ1,2-cap (pools 5 and 6). Although these structures have been reported as minor components among the N-glycan side chains present in insoluble cell wall mannan from both Candida and Pichia sp. (Funayama et al., 1983; Kobayashi et al., 1986, 1989; Shibata et al., 1995), they have never been shown to be O-glycan constituents on Pichia sp. or any other fungal glycoproteins that we can document.
Extracts of C. albicans contain multiple β1,2-mannosyltransferases (Suzuki et al., 1995). Pathogenic Candida strains reveal few if any cell wall mannoproteins with only one β1,2-linked Man cap on the N-glycan side chains (Suzuki, 1997). In pathogenic C. tropicalis strains, ∼15% of the side chains have two, whereas 5% have three nonreducing terminal β1,2-linked Man residues. This suggests that addition of the first nonreducing-end β1,2-linked Man by β1,2-mannosyltransferase I is rate limiting (Suzuki et al., 1995). Pichia does not appear to elongate the O-linked glycan β1,2-Man cap beyond two residues. It is unknown whether Candida sp. synthesize comparable α1,2-Man-containing O-glycans.
MALDI-TOF MS experiments revealed the presence—although in too small an amount for the glycosidic linkages to be established by NMR methods—of a phosphorylated O-glycan in pool 4 with a mass of 1095 Da. This glycan fragmented by PSD in a manner consistent with a H2PO3Man attached in diester linkage to the reducing-end Man-ol that was also substituted by a tetrasaccharide (Figure 5). Given the linkage structures of O-glycans defined here and elsewhere (Duman et al., 1998) and given the complement of mannosyltransferases present in Pichia (Verostek and Trimble, 1995), we speculate that the extended polymer is an α/β1,2-linked Man with PO4Manα1,6-linked to the reducing-end Man-ol. ManPO4 in diester linkage has been found α1,6-linked in N-glycan cores in S. cerevisiae mannan (Hernández et al., 1989) and in O-linked mannooligosaccharides previously characterized on the Bar1p protease found in this organism (Jars et al., 1995).
β1,2-Man-linked mannose is immunogenic in vivo, and cell wall structures with β1,2-mannobiose caps described in this article constitute antigenic factor 6 in C. albicans and antigenic factor 9 in Candida quilliermondii, both of which are pathogenic in humans (reviewed in Suzuki, 1997). The capacity of P. pastoris to synthesize these structures, even in small amounts, should be of interest to academic researchers, but, more importantly, will have to be considered in the context of the future application of this organism in industrial bioprocess design.
Materials and methods
Approximately 250 mg recombinant full-length hBSSL was purified from the culture broth of P. pastoris strain GS115 transformed with the BglII-digested plasmid pARC5799 (Sahasrabudhe et al., 1998). Bio-Gel P-4 (extra-fine mesh) was from Bio-Rad Laboratories (Hercules, CA) and NaBH4, 99.8% D2O, and 2,5-dihydroxybenzoic acid were from Sigma Chemical (St. Louis, MO). CIL (Cambridge, MA) supplied the 99.996% D2O. NMR tubes (535 pp) were supplied by Wilmad Glass Works (Buena, NJ). A. satoi α1,2-mannosidase was from Glyco (formerly a product of Oxford Glycosystems) and Endo H (Trumbly et al., 1985) and jack bean α-mannosidase (Trimble et al., 1978) were prepared in our laboratory. Helix pomatia β-mannosidase was purchased from Seikagaku America (Ijamsville, MD). All solvents, sulfuric acid, and phenol were ACS reagent grade or better from commercial suppliers. Monosaccharide and monosaccharide alditol standards were obtained from Dionex (Sunnyvale, CA). [U-14C]Man (279 µCi/umol) was purchased from Amersham Pharmacia Biotek (Piscataway, NJ). Dowex 50 H+ (50–100 mesh) and Amberlite MB-3 mixed bed resin were from Fisher Scientific (Rochester, NY). SepPak C18 cartridges were supplied by Millipore (Bedford, MA).
N-linked oligosaccharides were hydrolyzed from SDS-denatured hBSSL with Endo H at 50 mU/ml overnight at 30°C (Trimble and Maley, 1984), and the released N-glycans were isolated by acetone precipitation followed by solubilization in 60% CH3OH (Verostek et al., 2000). O-linked oligosaccharides were β-eliminated from Endo H–treated hBSSL by solubilization of the precipitated protein in 0.1 M NaOH in the presence of 1 M NaBH4 (Gemmill and Trimble, 1999b). After overnight incubation at 45°C, residual NaBH4 was destroyed by dropwise addition of glacial acetic acid. The mixture was passed over a 40-ml column of Dowex 50 H+, and borate was removed from the eluate as methylborate by rotary evaporation three times from 1% acetic acid in methanol. The reduced O-glycan pool was solubilized in dH2O; residual peptide and/or SDS was removed by passage through a 1 × 3 cm column of Amberlite MB-3 mixed bed resin prior to polishing by passage through a disposable C18 SepPak cartridge.
N- and O-linked glycans were concentrated by rotary evaporation to 1.5–2 ml and chromatographed on a 1.6 × 95 cm Bio-Gel P-4 (extra-fine mesh, lot 44671A) in 0.1 N acetic acid/1% butanol at 8.4 ml/h and room temperature. Fractions of 0.74–0.77 ml were collected and analyzed for neutral hexose by a modification (Byrd et al., 1982) of the phenol–sulfuric acid assay (Dubois et al., 1956).
MS of glycans was performed with a Bruker Reflex MALDI-TOF instrument. Sample spots of 25–50 pmol glycan were prepared in 2,5-dihydroxybenzoic acid matrix. Molecular weight data were obtained from the resultant average of 50–200 3-ns pulse shots from a 337 nm laser. A PSD spectrum of the pool 4 phosphorylated O-glycan was generated by isolation of a roughly 20 m/z window about the parent ion mass and recording the PSD product ions observed through 10 reflectron voltage reduction step spectra. These 10 spectra were then pasted together to form the final PSD spectrum.
The Endo H–released N-glycans and O-glycans in pools 2–6 were prepared for 1H NMR spectroscopy in 99.996% D2O at 1–2 mM final concentration in 0.5 cm NMR tubes as described (Ziegler et al., 1999). 1H NMR 1D and 2D DQF COSY experiments were conducted on a Bruker Avance 500 MHz DRX spectrometer at 296 K as described (Cipollo and Trimble, 2000), and 2D relayed ROESY experiments to selectively detect 1,2-glycosidic linkages have been detailed (Cipollo et al., 2000). Some NMR spectra were analyzed using NMRPipe/NMRDraw software (Delaglio et al., 1995; available online at http://spin.niddk.nih.gov/bax/software/nmrpipe). After rechromatography of the pool 5 glycan (see Discussion), the remaining sample, exchanged with 99.996% D2O, was examined in a Nalorac microprobe on the Varian 500 MHz spectrometer at the University of Colorado Health Sciences Center, and a portion was analyzed by methylation and MS fragmentation at the Boston University NIH-supported Mass Spectrometry Resource Center.
Analytical HPAEC of reduced O-glycan pools (Figure 4) was performed using a Dionex GP50 gradient pump system and ED-50 electrochemical detector on an 0.4 × 25 cm PA1 column employing PeakNet 6 software. The column was eluted isocratically with 16 mM NaOH at a flow rate of 1 ml/min. Man-ol was quantitated using an 0.4 × 25 cm MA1 column eluted at 0.4 ml/min with 0.26 M NaOH. At 29 min the NaOH was reduced to 0.15 M for 5 min prior to reequilibration in 0.26 M NaOH for 36 min.
[14C]Man-ol was prepared by reducing [U-14C]Man (279 µCi/µmol) with 1 M NaBH4 in 0.1 M NaKB3O4, pH 9.4, overnight at room temperature. Cations were removed after acidification with glacial CH3COOH by passage through a Dowex 50 H+ column. Borate was removed by rotary evaporation as described, and the [14C]Man-ol was further purified by chromatoraphy on the Bio-Gel P-4 column described. Exoglycosidase digestions with jack bean α-mannosidase were conducted in 50 mM sodium acetate buffer, pH 4.5, 50 mM NaCl, and 0.1 mM zinc acetate; α1,2-mannosidase digestions were performed in 100 mM sodium acetate buffer, pH 5.0. β-Mannosidase assays were conducted in 10 mM sodium citrate/phosphate buffer, pH 5.0. After overnight hydrolysis at 30°C, ice-cold methanol was added to 60% final concentration, and after 1 h at −20°C, precipitated protein was removed by centrifugation. Supernatant fractions were flash-evaporated to dryness in the tip of a 5-ml screw-capped glass conical test tube. Samples were dissolved in 30 µl dH2O, and a portion of each was subjected to Dionex PA1 chromatography as described. SDS–PAGE on 8% resolving gels was performed as described by Laemmli (1970).
We thank Dr. G. Das and his research team for the gift of recombinant hBSSL expressed in P. pastoris used in this study (Sahasrabudhe et al., cited in the references). We thank the Wadsworth Center Photography and Illustrations Group for aid in preparing figures and the care and patience of Tracy Godfrey in preparing the manuscript. Portions of this work were performed in the Wadsworth Center Biological Mass Spectrometry and NMR Structural Biology Cores. We greatly appreciate discussions with Dr. Brad Bendiak at the University of Colorado Health Sciences Center and the 1D NMR spectrum of the pool 5 O-glycan provided by him (Figure 6B). Dr. Cathy Costello kindly provided the methylation/MS analysis of the pool 5 O-glycan at the Boston University School of Medicine NIH-supported Mass Spectrometry Resource Center. Funding of this work, in part, by USPHS NIH grant GM23900 (to R.B.T.) is gratefully acknowledged.
2Wadsworth Center C-547, New York State Department of Health, PO Box 509, Albany, NY 12201-0509; and 3Astra Zeneca Research Foundation India, Bellary Road, Hebbal, Bangalore 560 024, India