N - and O -Glycosylation and Phosphorylation of the Bar Secretion Leader Derived from the Barrier Protease of Saccharomyces cerevisiae *

A secretion leader derived from a domain of the extra- cellular Barrier protease of the yeast Saccharomyces cerevisiae has been expressed in wild-type and in mnn1 , mnn9 , and mnn1 mnn9 glycosylation mutant strains of S. cerevisiae . Structural comparison of the extracellular leader by mass spectrometry, peptide mapping, and elementary analysis proved that all strains produced a heterogeneous, heavily glycosylated polypeptide of 161 amino acids with both N - and O -glycosylation and phos- phorylation. All three potential Asn N -linked sites were glycosylated to some extent with the expected struc- tures. Neither the different growth media used nor the glycosylation mutations had significant effect on O -gly- cosylation with respect to both site selectivity and size of the carbohydrate structures. All 33 Ser and 21 Thr residues in the polypeptide were glycosylated at least partially, with an average of more than 2 mannoses/site. Although the mnn1 mutation blocks addition of (cid:97) 1,3-linked mannose, the bar secretion domain expressed in the mnn1 and mnn1 mnn9 transformants unexpectedly contained some O -linked structures with at least 4 man- noses/chain. These O -linked structures were as large as when the leader was expressed in the mnn9 and wild-type strains. The bar secretion domain also had a previously undocumented phosphorylated O -linked structure. A leader sequence fusion secretory path-way,

A leader sequence secures secretion of a protein of interest (1). As the fusion protein proceeds through the secretory pathway, it may undergo extensive post-translational modifications (e.g. glycosylation of specific asparagine (Asn) and serine or threonine (Ser/Thr) residues (2,3)). These modifications may be important for efficient secretion of the protein (3). However, very little is known specifically about how post-translational modifications (especially O-glycosylation) affect the ability of leaders to export proteins. Therefore, the purpose of this work has been to determine if and how cell physiology and growth conditions influence post-translational modifications of a secretion leader derived from the signal peptide and primarily the third domain of the BAR1 gene product of the yeast Saccharomyces cerevisiae. This secretion leader has been used to export heterologous proteins (e.g. platelet-derived growth factor and insulin precursors (4)) and will be referred to as the bar secre-tion domain (BSD). 1 Previous studies have indicated that the BSD contains substantial levels of N-and probably also O-glycosylation (5).
The most common post-translational modifications in S. cerevisiae are N-and O-glycosylation. Oligosaccharides consisting of Glc 3 Man 9 GlcNAc 2 may be transferred as a unit to specific asparagine residues (Asn-Xaa-Ser/Thr, where Xaa can be any amino acid except Pro; Ref. 6) in nascent peptides in the lumen of the endoplasmic reticulum. These oligosaccharides are trimmed in the endoplasmic reticulum to generate a Man 8 GlcNAc 2 intermediate, which in yeast is usually elongated in the Golgi to yield species of Man 9 -14 GlcNAc 2 (6,7). On secreted glycoproteins from yeast, some of these core oligosaccharides become further elongated by addition of an ␣1,6linked mannose and its extension to form an ␣1,6-linked backbone and by addition of short ␣1,2-linked side chains, which in turn may get the addition of ␣1,3-linked side chains to form an outer chain of Ն50 mannoses (hyperglycosylation) (6). Additional mannose residues in phosphodiester linkages may also be present in the core and in the outer chain (8,9).
Ballou (10) has isolated mnn mutants that are defective in mannan biosynthesis. The mnn1 mutation results in the synthesis of N-and O-linked structures (N-and O-structures) that lack terminal ␣1,3-linked mannose attached to mannose in ␣1,2-linkage, and the mnn9 mutants lack the ability to elongate the ␣1,6-linked backbone of the N-structures (10). Due to these mutations, the mnn9 strain primarily makes N-structures of Man [11][12][13] GlcNAc 2 , and the mnn1 mnn9 mutants make N-structures of mainly Man 10 GlcNAc 2 units along with a small amount of Man 11 GlcNAc 2 units (10). The MNN1 gene encodes the ␣1,3-mannosyltransferase (11,12), and the MNN9 gene is believed to encode a protein involved in maintaining a Golgi apparatus with functioning mannosyltransferases (13).
No consensus sequence for O-glycosylation has been found (2,14). The first sugar transfer of mannosyl residues to Ser or Thr is catalyzed in the endoplasmic reticulum and requires dolichol-phosphate-mannose as an intermediate (15). The additional mannosyl residues (maximally 3-4) are transferred from GDP-Man (15). It is uncertain whether the second ␣1,2linked mannose is added in the endoplasmic reticulum or Golgi (15,16), but the third ␣1,2-linked mannose is added in the Golgi (17). An additional one to two ␣1,3-linked mannoses may then be added to the three ␣1,2-linked mannoses in the Golgi. O-Structures with one to five mannoses are expected on pro-teins expressed in the wild-type and the mnn9 strains (2,10,18). Addition of the first and probably the second ␣1,3-linked mannose requires the MNN1 gene (10). Therefore, the expected O-structures on proteins expressed in the mnn1 and mnn1 mnn9 strains have one to three ␣1,2-linked mannoses (10).
The DNA sequence of the secretion leader used in this work has three potential N-linked (Asn) and 54 potential O-linked (Ser/Thr) glycosylation sites (see Fig. 1) and has been expressed in wild-type and in mnn1, mnn9, and mnn1 mnn9 glycosylation mutant strains of S. cerevisiae. Post-translational modifications of the BSD have been analyzed with respect to site selectivity by peptide mapping, sequencing, and size of O-structures by electrospray (ES) liquid chromatography mass spectrometry (LC/MS). An unusual phosphorylated post-translational modification was investigated by ES tandem mass spectrometry (MS/MS) and elementary analysis.
The transformants were grown in fermentation medium (FM) 1 (7.5 g/liter yeast extract (Difco), 14 g/liter KI, 0.1 ml of concentrated HCl, and 2.0% (w/v) D-glucose in tap water). In fermentations of the wild-type, mnn1 mnn9, and mnn9 strains, 0.75 M D-sorbitol was added to the medium. Strains carrying the mnn9 mutation require the presence of D-sorbitol to osmotically stabilize the cells during growth (10). Tanks of 60 liters of medium were inoculated with 3% (v/v) wild-type or mnn1 mnn9 transformants, and the cells were fermented at 30°C for 48 h. D-Glucose was fed continuously during the fermentation at a rate of 1.5 g/liter/h, and pH was kept at 4.5 with ammonia. The mnn9 and mnn1 transformants were grown at 30°C for 48 h in Fernbach flasks of 10 ϫ 1200 ml of FM1 ϩ D-sorbitol and 6 ϫ 1200 ml of FM1, respectively. Additional D-glucose was added to the cultures to a final concentration of 2% (w/v) after 24 h. Wild-type transformants were also grown in 1200 ml of FM2 containing 40 g/liter yeast extract (Difco), 20 g/liter Bacto-peptone (Difco), 0.1 ml/liter pluronic acid (100% v/v), 18 g/liter D-glucose, pH 6.5 at 30°C. At 90 h, 5% (v/v) glacial acetic acid was added to the medium, and the fermentation was stopped after 96 h.
Purification-BSD from a wild-type strain will be designated BSDwt, a similar nomenclature is used for the other strains. The number 2 in the name (e.g. BSDwt2-N) indicates that the strain has been grown in FM2, and the letter N indicates that the N-structures have been enzymatically removed. For BSDwt and BSDwt-N, the culture supernatant of 60 lilters was filtered through a 0.45 m 2 polypropylene Gelman filter, concentrated to 1.5 liters on a hollow fiber system (UFP-5-C-55, cut off 5000 Da, A/G Technology Corporations), precipitated by ethanol (20), and resuspended in 640 ml of 5 mM EDTA. A volume of 40 ml of the resuspended precipitate (BSDwt) containing 10 mM dithiothreitol was fractionated on a Sephacryl column (4.0 ϫ 86 cm S-200, HR; Pharmacia) equilibrated in 0.15 M NaCl and 5 mM EDTA. After addition of 16 ml of 0.5 M sodium acetate, pH 5.2, 80 ml of resuspended ethanol precipitate was deglycosylated by incubation at 37°C for 2 days with 3 units of endo-␤-N-acetylglucosaminidase H (Boehringer Mannheim). This batch (BSDwt-N) was then fractionated as described for BSDwt. The two batches were individually desalted on a reversed phase HPLC C-4 column (2.2 ϫ 25 cm, Vydac) by elution with a gradient of 5-35% (v/v) acetonitrile (ACN) in H 2 O and 0.1% (v/v) trifluoroacetic acid. The samples were lyophilized and stored at Ϫ20°C. BSDwt2-N was purified from 500 ml of broth after 96 h of fermentation as described for the BSDwt-N, except that BSDwt2-N was concentrated on an Amicon cell (cut off 50 kDa, Amicon) instead of the hollow fiber system.
BSDmnn1 mnn9 was purified as the BSDwt sample, but the broth was concentrated on a 1.5-liter Amberchrom CG71-md column (Toso-Haas). After loading the sample, the column was washed with 20 liters of 0.1 M sodium acetate, 2 mM EDTA, and NaCl (to 15 millisiemens/cm), pH 4.5, followed by 8 liters of 2 mM EDTA. The sample was eluted with 1050 ml of 40% (v/v) ethanol in 2 mM EDTA and ethanol precipitated (20). BSDmnn1, BSDmnn1-N, and BSDmnn9 were purified as described for BSDwt-N, but after the HPLC C-4 column, the samples were individually separated on a reversed phase HPLC C-18 column (0.46 ϫ 25 cm, Vydac) eluted with a gradient from 2-41% (v/v) ACN in H 2 O and 0.05% (v/v) trifluoroacetic acid. BSDmnn9 was further separated on a divinyl benzene column (0.46 ϫ 50 cm, PLRP-S 1000 Å, Polymer Laboratories) with a similar gradient. BSDmnn1-N and BSDmnn9 were then individually purified by gel filtration on a TSK SW3000 column (0.75 ϫ 30 cm, Hewlett Packard) by elution with 0.15 M NaCl and 5 mM dithiothreitol, concentrated on an HPLC C-18 column, vacuum rotor evaporated, and stored at Ϫ20°C. The BSD batches were analyzed by SDS-polyacrylamide gel electrophoresis (21,22) and immunoblotted with antibodies against substance P (19,20,23).
Amino Acid Sequence Analysis-N-terminal sequences were determined by automated Edman degradation for the BSD batches and several peptides from peptide mapping of the BSD using an Applied Biosystems model 475 or model 476A (24).
Amino Acid Analysis-All samples, except BSDmnn1, BSDmnn1-N, and BSDmnn9, were analyzed by amino acid analysis through hydrolysis of known volumes containing 0.5-0.8 nmol of protein with 6 M HCl/2% phenol for 22 h at 112°C (25); results were not corrected for loss during hydrolysis.
Peptide Mapping-Approximately 0.2 amidase unit of lysyl endopeptidase from Achromobacter protease I (lysyl endopeptidase, Wako Chemicals U. S. A., Inc.) was added per 1.0 mg of BSD glycoprotein dissolved in 0.9 ml of 50 mM Tris-HCl and 6.5 mM dithiothreitol, pH 8.9. The sample was left at 4°C for 92-144 h, and then the digestion was stopped with 50 l of glacial acetic acid/ml sample. The glycopeptides were chromatographed on reversed phase HPLC C-18 columns (0.46 ϫ 25 cm or 0.21 ϫ 25 cm, Vydac) eluted in gradients from 2-80% (v/v) ACN in H 2 O and 0.05% (v/v) trifluoroacetic acid.
ES MS-The analysis was performed using an API III liquid chromatography triple quadrupole mass spectrometer fitted with an articulated ion spray plenum and an atmospheric pressure ionization source (Perkin Elmer Sciex) and a scanning range of m/z 0 -2400 Da (26). The instrument was tuned and calibrated with the singly charged ammonium adduct ions of polypropylene glycols under unit resolution. The orifice value was usually held at 65 V to avoid O-glycosidic cleavage of the glycopeptides. In ES MS analysis, the samples were introduced at a flow rate of 5 l/min and analyzed in the mass range of m/z 600-2400 Da at a step size of 0.1-0.2 atomic mass unit and a dwell time of 0.5-0.75 ms. In comparative peptide mapping by ES LC/MS, 40 -200 g of BSD digestion mixtures were separated on a reversed phase HPLC C-18 column (0.21 ϫ 25 cm, Vydac), typically with a flow rate of 0.2 ml/min (split to 30 l/min flow and infused onto the ES MS). The following gradient was used: In ES MS/MS analysis, the fractions from a peptide digest of the BSDwt-N were introduced at a flow rate of 5 l/min and analyzed in the mass range of m/z 0 -1800 Da at a step size of 0.1-0.2 atomic mass unit and a dwell time of 0.5-0.75 ms. The collision gas was a mixture of argon and nitrogen (90:10).
Matrix-assisted Laser Desorption (MALDI) MS-Intact BSDmnn1 mnn9 and BSDwt-N as well as glycopeptides of BSDwt-N collected from a peptide digest were analyzed by Charles Evans & Associates on a reflector Vision 2000 MALDI MS time of flight mass analyzer (Finnigan) (27). The mass range was 500 -500,000 Da. A total of 5 pmol BSDmnn1 mnn9, dissolved in 0.1% (v/v) trifluoroacetic acid to l0 pmol/l and mixed 1:1 with the 2,5-dihydroxybenzoic acid matrix, was analyzed with 5 pmol of apomyoglobin as an external calibrant. BSDwt-N was analyzed the same way. Lyophilized fractions from a peptide digest of 1.0 mg of BSDwt-N were redissolved in 0.1% (v/v) trifluoroacetic acid, mixed 1:2 with the above matrix, and analyzed with renin substrate and human insulin as internal standards.
Elementary Analysis-Phosphorus and sulfur were analyzed by inductively coupled MS on an Elan 5000 A (Perkin Elmer Sciex) in collaboration with Thomas Chapin (University of Washington). The instrument was blanked against 1% (v/v) HNO 3 in H 2 O and calibrated with solutions of (NH 4 ) 2 SO 4 and KH 2 PO 4 in 1% (v/v) HNO 3 in H 2 O. The samples were analyzed in positive ion peak hop mode where the mass analyzer dwells at m/z 31 for 1000 ms and then 34 m/z for 1000 ms, etc. The samples were introduced with a continuous flow nebulizer at a flow rate of 1 ml/min (28). Intact BSDwt-N, BSDmnn1 mnn9, and fractions from a peptide digest of BSDwt-N were analyzed.
Peptide Synthesis-A peptide (LVNIQTDGSISGAK) corresponding to lysyl endopeptidase peptide 3 of the BSD was synthesized on an Applied Biosystems model 431A peptide synthesizer using Fmoc (9fluorenylmethoxycarbonyl) chemistry and standard cycles for 1-hydroxybenzotriazole activation (29 -31). The product was purified by reversed phase HPLC, and the identity was confirmed by ES MS analysis and Edman degradation.
Characterization of the BSD Batches-The proposed signal peptide of the bar secretion leader contains 24 amino acid residues (20). Amino acid sequence analysis of purified BSD confirmed that the signal peptide in all batches is processed between amino acids Ala 0 and Leu 1 (Fig. 1) as expected from von Heijne's rule (33). The polypeptide has a calculated molecular mass of 16 Site Selectivity of Post-translational Modifications by Peptide Mapping of BSDwt-N and BSDmnn1 mnn9-Theoretically, lysyl endopeptidase should cleave the BSD (referred to as Leu 1 to Met 161 ) into 8 peptides (P1-P8; Fig. 1). Peptide 2 consists of Lys 26 only; the rest were glycopeptides and will be referred to as (GP1-GP8). Separation of the digestion mixtures on a reversed phase HPLC C-18 column gave reproducible but complex UV-215 absorbance traces with broad overlapping peaks, as shown for BSDwt-N in Fig. 2. The cleavage site between Lys 105 and Pro 106 was not always cleaved. In general, GPm,n indicates that glycopeptide GPm and GPn are not cleaved, so in this case the above glycopeptide is designated as GP6,7. Additionally, an unexpected cleavage site between Leu 11 and Ser 12 was observed in BSDwt-N, giving rise to GP1ЈЈ (Figs. 1 and 2).
In Edman degradation data, the repetitive yield for each sequence was calculated based on a regression analysis of the more stable amino acids in each sequence (Ala, Glu, Gly, Ile, Leu, Lys, Phe, and Val), and the average repetitive yield was 91% for all the sequences from the BSD. The post-translationally modified amino acids were identified as missing PTHderivatives in Edman degradation. The obtained yield (Y n ) in percentage of the expected yield (Y e ) (based on the regression analysis) was calculated for every PTH-Ser, PTH-Thr, and PTH-Asn in each sequence. PTH-Ser and PTH-Thr are unstable. Additionally, the position of Ser, Thr, and Asn in a sequence and the amount of material used in Edman degradation both affect the yields of these amino acids. These effects were tested with a synthetic lysyl endopeptidase Peptide 3, where the Y n in percentages of Y e were 62, 26, and 29% of PTH-Thr 32 , PTH-Ser 35 , and PTH-Ser 37 , respectively (Fig. 1).
Each glycopeptide was analyzed several times due to overlapping peaks. A Ser or Thr was categorized as highly Oglycosylated if the Y n of that residue in all analyses was less than 10% and the increase in yield (Y n Ϫ Y n Ϫ 1 ) was less than 5% of Y e , moderately O-glycosylated if Y n Ͻ25% of Y e , and poorly O-glycosylated if 25% Ͻ Y n Ͻ 70% of Y e . Potential N-glycosylation sites were categorized similarly. The results of Edman degradation for both BSDmnn1 mnn9 and BSDwt-N are shown in Fig. 1. Because not every fraction was analyzed, the summary in the figure is an indication of the glycosylation rather than its absolute quantitation. The BSDmnn1 mnn9 had 30 highly, 13 moderately, and 10 poorly O-glycosylated sites of the total 54 Ser and Thr, respectively. Asn 19 and Asn 89 residues were moderately N-glycosylated, and Asn 124 were poorly N-glycosylated. The BSDwt-N had 35 highly, 8 moderately, and 11 poorly O-glycosylated sites, respectively. Asn 19 and Asn 124 residues seemed to be highly N-glycosylated, and Asn 89 seemed to be moderately N-glycosylated. Differences in O-glycosylation sites on BSD expressed in the mnn1 mnn9 strain and the wild-type strain were apparent on 22 sites (Fig.  1), with the most pronounced differences seen at Ser 37 , Thr 100 , and Ser 147 . However, there was no clear pattern in the extent of O-glycosylation on any of the 22 sites. If there was a pattern as to the extent of N-glycosylation, the BSDwt-N had a higher amount of Asn involved in N-glycosyl linkages than the BSDmnn1 mnn9.
ES MS analysis of different fractions from the peptide maps showed that they contained the expected heterogeneous glycopeptides. For example, fraction 27 from BSDwt-N contained GP5 with 1 N-acetylglucosamine and 29 -33 mannoses (referred to as P5ϩ1GlcNAcϩ(29 -33)Man) and GP5 with 1 Nacetylglucosamine and 28 -33 mannoses and an additional 80 mass units. This additional mass could represent a sulfate group (S) or phosphate group (P) (34) (referred to as P5ϩ1GlcNAcϩ(28 -33)ManϩP/S), as discussed in the next section. As the heterogeneity of a glycopeptide varied with respect to the number of N-acetylglucosamines, mannoses, and a phosphate group (or sulfate group), in the following section, a distribution will refer to a given glycopeptide that has a particular number of GlcNAcs and a phosphate group (or sulfate group) across a range of mannoses (e.g. P5ϩ1GlcNAcϩ(28 -33)ManϩP/S). Each of the "members" within the distribution will be referred to as a variant of that distribution (e.g. P5ϩ1GlcNAcϩ28ManϩP/S). Each glycopeptide variant eluted with decreasing mass as a function of time, as expected (35). All the glycopeptides were observed in distributions with and without an addition of 80 mass units. Some of the glycopeptides with different numbers of mannoses, N-acetylglucosamine, phosphate, and/or sulfate may have the same masses (e.g. GP5,6 and GP5) and cannot be distinguished by ES MS analysis. In this case it was determined by Edman degradation that only GP5 was present. However, Edman degradation could not solve this problem with GP4 and GP8, because they coeluted.
By using orifice values between 45-70 V, the heterogeneity observed by ES MS seemed real and not generated during the analysis by bond breakage between sugar units on the glycopeptide in the declustering region between the orifice and the first reference-only quadrupole. In agreement with the litera-ture (35), orifice values between 55-65 V seemed to be optimal to avoid fragmentation and to obtain good sensitivity. Generally, the same distributions and relative intensities of the variants were observed by MALDI MS and ES MS, showing that the heterogeneity was in fact real. However, the ES MS data gave the best accuracy.
An Unusual Post-translational Modification-Because no phosphate or sulfate was added during the purification procedures of the BSD batches and the last purification step was a reversed phase C-4 column, it was assumed that phosphorus and/or sulfur found in these samples would be covalently attached to the glycoprotein. The sulfur concentration in the BSD batches was expected to be 7 sulfurs/molecule of protein, because the DNA sequence has 3 Cys and 4 Met. In elementary analysis the sulfur concentration was 6.3 and 7.8 sulfurs/molecule in the BSDwt-N and BSDmnn1 mnn9, respectively, values which were within 2-3 times the standard deviation of the expected value. Therefore, there is no evidence for post-translational sulfate addition. In contrast, the BSDwt-N and BSDmnn1 mnn9 contained approximately 6.9 and 3.6 mol of phosphorus/mol of protein, respectively, all of which must be added post-translationally. On the basis of these results, it was assumed that the unexpected variants with an additional 80 mass units could be associated with post-translational addition of a P. Analysis of fractions containing GP1, GP3, or GP5 from the BSDwt-N enriched with an additional 80 mass units showed the presence of phosphorus. Treatment with Escherichia coli or bovine alkaline phosphatase of the BSDwt-N did not change the elution time on a reversed phase HPLC C18 column, but treatment in mild acid delayed the elution time of BSDwt-N on the column (data not shown), as expected if a phosphate group was chemically cleaved by the acid treatment (36).
The product ions generated in positive ion mode ES MS/MS analysis of the selected precursor ion m/z 1633.9, representing P3ϩ11ManϩP of BSDwt-N, is shown in (Table I). Note the abundance of ions representative of O-glycosidic cleavage to yield both reducing and nonreducing terminal fragments without any evidence of amide bond cleavage of the peptide backbone, which is typical for a glycopeptide (37). Adopting the nomenclature proposed by Domon and Costello (38), the reduc-  representing P3ϩManϩP. The next mass loss is equivalent to a mannose and a P, although no mass is present that could represent P3ϩP. At the low mass-to-charge region, three characteristic oxonium ions at m/z 242.93, 405.02, and 567.91 are observed, arising from cleavage at the glycosidic linkage with charge retention on the sugar. These represent Man-Pϩ, Man-Man-Pϩ, and ManManMan-Pϩ, respectively. These two observations lead to the conclusion that the phosphate group is attached to the sugar moiety, rather than the backbone of the protein. These oxonium ions were only observed when selecting precursor ions of variants expected to contain a P. As mentioned previously, GP4 and GP8 coeluted and may have the same masses, but based on data from ES MS/MS analysis, GP8 seemed to elute slightly earlier than GP4. Fig. 3, with the m/z values for all the masses in the range of m/z 600-2400 for each single scan. The glycopeptides can be observed in the contour plot as negatively sloping "streaks" (39) eluting in the same order as observed in Fig. 2. However, the elution times of the glycopeptides are not identical in the two figures due to the two different size columns used. As an example, the sum of all the mass spectra in the scans 327-384 of GP5 in the high mass range is shown in Fig. 4. The data can be summarized as P5ϩ1GlcNAcϩ (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)ManϩP and P5ϩ1GlcNAcϩ(24 -37)Man. A summary of the distributions of each glycopeptide eluting in the ES LC/MS analysis of BSDwt-N is shown in Table II. Extensive O-glycosylation is observed with typically more than two mannoses/O-linked site for the most intense variants of each distribution for every glycopeptide.

Comparative Peptide Mapping by ES LC/MS-The contour plot from the analysis of BSDwt-N is shown as an example in
The obtained ES LC/MS data for all the BSD batches have been summarized in Table III. Only the range of mannoses/Olinked site of each distribution is shown. Obviously, different distributions are observed in batches that have and have not been deglycosylated with endo-␤-N-acetylglucosaminidase H. Unexpectedly, GP1 and GP6,7 were observed with 1 GlcNAc in batches that had not been deglycosylated with endo-␤-N-acetylglucosaminidase H. It cannot be excluded that these distributions with 1 GlcNAc were generated in the ES LC/MS analysis. Such an artifact has previously been reported (40). GP7 gave weak signals when present due to incomplete cleavage between Lys 105 and Pro 106 . Because BSDmnn9, BSDmnn1, and BSDmnn1-N were analyzed in small amounts of 40 -80 g/ sample only, some variants were not observed in these batches (e.g. GP7 and phosphorylated distributions of GP6). In BSDmnn1 and BSDwt, the distributions of glycopeptides with hyperglycosylated N-structures could not be expected to fall inside the mass range of m/z Ͻ 2400 Da, and these distributions were not observed. The batches do not show identical distributions, but a great deal of similarity is present, and no clear pattern differences were observed. Differences were primarily noted for the less intense variants (i.e. some phosphorylated distributions of very weak intensities were observed in BSDwt-N but not in BSDwt2-N) ( Table III) Data from GP4 and GP8 were not included because the elution of these peptides was uncertain. No distributions of GP5 without GlcNAc were present in any of the batches, indicating that the Asn 89 residue was originally fully N-glycosylated.
Distributions of phosphorylated glycopeptides gave less intense signals in ES MS analysis than those without phosphate. This is as expected due to the electronegative character of the P, which decreases the likelihood of ionization. No distributions with more than one P/peptide were observed in the ES MS analysis; if present, they may not have ionized in the positive ion mode used.
The growth medium tested did not seem to affect the posttranslational modifications of the BSD expressed in wild-type S. cerevisiae. The similarities of the glycopeptide distributions between BSDwt-N (fermented in FM1) and BSDwt2-N (fermented in FM2) were high in terms of both the number of mannoses/O-linked site (Table III) and the abundance of each variant (ES MS data not shown). The only exceptions were some phosphorylated distributions of weak intensities. All this suggests fairly consistent processing in vivo.

DISCUSSION
As previously proposed (5), the BSD is a heavily glycosylated heterogeneous protein. The N-structures were influenced by the mnn1 and mnn9 mutations as expected, and hyperglycosylation was observed only on the BSD expressed in the mnn1 and wild-type strains (10). All the BSD had the same molecular mass after endo-␤-N-acetylglucosaminidase H treatment. Therefore, the difference in the peak molecular mass between the BSDwt-N (38,949 Da corresponding to 133 mannose residues/molecule) and BSDmnn1 mnn9 (44,317 Da corresponding to 162 mannose residues/molecule) of 5,368 Da (analyzed by MALDI MS) may be explained by partial occupancy of the 3 potential N-glycosylation sites residues 19, 89, and 124 in the BSDmnn1 mnn9 sample with the expected structure of Man 10 GlcNAc 2 . Carbohydrate analyses confirmed a high content of 106 -329 mannose residues/molecule in the BSD batches and the expected size of the N-structures. 2 In all batches, some N-glycosylation was found at Asn 19 and Asn 124 , as determined by Edman degradation and ES LC/MS analysis. Asn 89 seemed to be fully N-glycosylated, based on ES LC/MS experiments, but Edman degradation indicated some unglycosylated material as well. Because an Asn 84 precedes the Asn 89 in the sequence, the result from Edman degradation could be less accurate.
No clear pattern was observed regarding the site selectivity and the extent of O-glycosylation for BSDwt-N and BSDmnn1 mnn9 based on Edman degradation. Almost all the 54 Ser and Thr residues were O-glycosylated to some extent. Clustering of Ser/Thr in O-glycosylated mucins has been described (41). This was also observed in the BSD, but no clear information that may help predict the amino acid distribution around O-linked sites in S. cerevisiae was established. Strahl-Bolsinger and Tanner (42) observed that a Gly residue immediately preceding Thr (Gly-Thr) and an acidic amino acid in the vicinity of a hydroxy amino acid eliminated the mannosyl acceptor property of the hydroxy amino acid on synthetic peptides in vitro. Their data have been supported by analysis of human IGF-I (43) but not by analysis of the B-chain from platelet-derived growth factor (44), both expressed in S. cerevisiae. Their conclusions are also not supported in the present study with BSD, at Gly 5 -Thr 6 , Gly 96 -Thr 97 , Ser 21 -Glu 22 , Glu 82 -Thr 83 , and Glu 125 -Thr 126 , where prevention of the mannosylation of the hydroxy amino acid did not occur (Fig. 1). Pro enriched in positions Ϫ1 and ϩ3 did not seem to be an indicator of O-glycosylation for the 6 Pro in BSD, in contrast to O-linked N-acetylgalactosamine in mucins (41). This result could suggest, as also proposed by Strahl-Bolsinger and Tanner (42), that the rules for site selectivity of O-glycosylation are somewhat different in mammalian and yeast cells. Other in vitro studies with the yeast Candida albicans using synthetic peptides indicated that Thr was more frequently O-mannosylated than Ser and that it was unlikely that both Ser and Thr were mannosylated on the same peptide (45). None of these observations were seen in the BSD.
Surprisingly, no apparent influence on the O-glycosylation by the mnn1 mutation could be detected from the immunoblot. The M r after removal of the N-structures indicated substantial amounts of O-glycosylation. In ES LC/MS analysis there were O-structures present with at least 4 mannoses/site (chain) on both the BSD from the mnn1 and mnn1 mnn9 strains. It was confirmed by carbohydrate analysis that all strains produced O-linked chains of 1-5 mannoses, 2 although, it is unknown if these long O-linked chains are present on both Thr and Ser residues. This result was surprising because the mnn1 mnn9 and mnn1 strains were not expected to have O-structures with more than 3 mannoses/site because the mnn1 mutation prevents addition of terminal ␣1,3-linked mannoses (10). There is no obvious explanation why the O-structures on BSD expressed in the mnn1 and mnn1 mnn9 strains were longer than 3 man-noses/site, and it is unknown how the fourth and fifth mannoses were added, perhaps as ␣1,2-linked mannoses or ␣1,3linked mannoses. If the latter is true, then there must be an additional ␣1,3-mannosyltransferase present in S. cerevisiae.
The glycopeptides were observed with ranges of 0.0 -4.3 mannoses/O-linked site in ES LC/MS analysis. It was shown that the same heterogeneity in a given glycopeptide was observed by both MALDI MS and ES MS, but it was not proven that the relative intensities of the variants (peaks) observed in the mass spectra actually reflect the relative quantities of each variant. However, in comparison within (and only within) a given distribution of a glycopeptide that varies only in the number of mannoses, it is assumed that the intensities of the peaks (variants) reflect the actual concentrations of the variants. Then, based solely on the most intense variants in each distribution, the average O-linked chain length/site is greater than 2.0 in all the batches. The total number of mannoses on the 8 glycopeptides from BSDwt-N (attached to the most intense variants in distributions without phosphorylation; Table  II) is 126, equivalent to 2.3 mannoses/O-linked site on average. This hypothetical BSD structure is in agreement with the expected number of 133 mannoses corresponding to 2.5 mannose residues/O-linked site calculated from the M r peak of BSDwt-N from MALDI MS.
For BSD expressed in the wild-type strain, no significant differences in post-translational O-linked modification were observed between BSD samples produced in FM1 ϩ D-sorbitol and FM2. In contrast to literature reports (46 -48), the two media (and pH) did not differently influence the O-glycosylation pattern of the BSD, except for minor differences in a some of the phosphorylated distributions of weak intensities.
ES MS and elementary analysis showed that the BSD was post-translationally modified with phosphate. The phosphate group was attached to the O-linked sugar moiety, a linkage that has not previously been described for S. cerevisiae. A product ion of m/z 783.03, representing [P3ϩ1ManϩPϩ2H]2ϩ, was present in ES MS/MS data of P3ϩ11ManϩP, showing that the phosphate may be attached to the first mannose on a Ser or Thr. It is unknown if phosphate was attached to the sites of only Ser, only Thr, or both. Phosphates on the N-linked oligosaccharides in S. cerevisiae are in acid labile diesters with 1-2 mannoses in ␣-glycosidic linkage to the phosphate units that are esterified to position 6 of mannose units in the N-structures (8). A similar phosphorylated O-structure could explain why treatment of the BSDwt-N with mild acid but not alkaline phosphatases delayed the elution time on a reversed phase HPLC C-18 column (36).
Yeast glycoproteins with phosphorylation are primarily found in the cell wall, although the role of mannose phosphorylation in S. cerevisiae is unknown (2,6). Some prokaryotic and eucaryotic heat shock proteins show a stress-dependent phosphorylation of the polypeptide (e.g. Ser) leading to a modulation of their function (49). It is possible that the phosphorylation plays a role in the putative function of the BSD as a chaperone, particularly through the cell wall.
The contaminant Hsp150p is a heavily O-glycosylated secretory protein (32). Other heat shock proteins (e.g. Hsp70p) (50,51) have been shown to serve important functions as chaperones involved in the maintenance or change of the conformation of other proteins. The Hsp150p and the BSD both have extensive O-glycosylation; therefore, it is most likely that they are nearly fully extended in these regions and have structures resembling a semi-flexible rod, as seen for mucins (52). Perhaps the shape is important in the proposed role of the BSD as a secretion chaperone for a protein (either the Bar protease or a heterologous protein) in transit through the Golgi and/or cell wall. Also, the O-structures could be important in protecting the heterologous protein from protease attack in transit through the secretory pathway.
A secretion leader derived from a domain of the Barrier protease of S. cerevisiae has been expressed in wild-type, mnn1, mnn9, and mnn1 mnn9 glycosylation strains. These mutations and different growth conditions had limited effect on the posttranslational modification of the part of the leader sequence exported to the culture medium (except for the size of the N-structures). It is possible that post-translational modifications are important, especially O-glycosylation, for the leader sequence conformation and thereby perhaps for the proposed function of the sequence as a secretion chaperone.