Structure of oligosaccharides on Saccharomyces SUC2 invertase secreted by the methylotrophic yeast Pichia pastoris.

Saccharomyces SUC2 invertase, secreted by the methylotrophic yeast Pichia pastoris and purified to homogeneity from the growth medium by DE-52 chromatography, appeared on sodium dodecyl sulfate-polyacrylamide gel electrophoresis as a diffuse ladder of species at 85-90 kDa, while the secreted Saccharomyces form migrated as a broad band from 100 to 150 kDa. Endo-beta-N-acetylglucosaminidase H released the Pichia invertase carbohydrate generating a 60-kDa protein with residual Asn-linked GlcNAcs and oligosaccharides separated on Bio-Gel P-4 into Man8-11GlcNAc. Nearly 75% of the oligosaccharides were equally distributed between Man8,9GlcNAc, while 17% were Man10GlcNAc and 8% were Man11GlcNAc. Oligosaccharide pools were analyzed for homogeneity by high-pH anion-exchange chromatography, and structures were assigned using 500 MHz one- and two-dimensional 1H NMR spectroscopy. Pichia Man8GlcNAc was the same isomer as found in Saccharomyces, which arises by removing the alpha 1,2-linked terminal mannose from the middle arm of the lipid-oligosaccharide Man9GlcNAc (Byrd, J. C., Tarentino, A. L., Maley, F., Atkinson, P. H., and Trimble, R. B. (1982) J. Biol. Chem. 257, 14657-14666). The Man9GlcNAc pool was 5% lipid-oligosaccharide precursor and 95% Man8GlcNAc isomer with a terminal alpha 1,6-linked mannose on the lower-arm alpha 1,3-core-linked residue (Hernández, L. M., Ballou, L., Alvarado, E., Gillece-Castro, B. L., Burlingame, A. L., and Ballou, C. E. (1989) J. Biol. Chem. 264, 11849-11856). An alpha 1,2-linked mannose on the new alpha 1,6-linked branch in Man9GlcNAc provided 80% of the Man10GlcNAc, which is the structure on Saccharomyces invertase (Trimble, R. B., and Atkinson, P. H. (1986) J. Biol. Chem. 261, 9815-9824). A minor Man10GlcNAc (12%) and the principal Man11GlcNAc (82%) were the major Man9,10GlcNAc with novel alpha 1,2-linked mannoses on the preexisting alpha 1,2-linked termini. Although Pichia glycans did not have terminal alpha 1,3-linked mannoses as found on Saccharomyces core oligosaccharides, over 60% of the structures were isometric configurations unique to lower eukaryotes.

Pichia pastoris is a methylotrophic yeast, which has been exploited for the high level expression of heterologous proteins by ligation to the methanol-inducible alcohol oxidase ( A O X I ) promoter of gene sequences of interest (1,2). As part of a study to determine the characteristics of heterologous glycoprotein secretion from yeast, a vector (pGSlO2) consisting of the Saccharomyces SUC2 invertase coding sequence coupled to the AOXl promoter was constructed and used to transfect Pichia (3). On induction with methanol, invertase was secreted by the transformed cells over a period of 100 h to a level of 2-3 g/liter.
The invertase secreted by wild-type Saccharomyces strains consists of 140-kDa subunits which are in excess of 50% by weight glycan (4, 5). The carbohydrate is distributed as families of asparagine-linked Man8_14GlcNAcn and M~I I ,~~G~C N A C~ oligosaccharides attached to 13 of 14 -Am-X-Thr/Ser-sequences in the protein (6, 7). In contrast to Saccharomyces, SUC2 invertase secreted by Pichia is a more homogeneous product consisting of -85-kDa subunits, which by indirect methods appear to be associated with 8-10 oligosaccharides of the size Man8-14GlcNAc2 (3). Incorporation of [3H]Man by Pichia revealed that over 85% of the invertase label was in short oligosaccharides, nearly 70% of which were Man8,9GlcNAc2 (8).
The presence of short oligosaccharides on Pichia invertase suggested that glycan processing may be different in this organism from that in Saccharomyces. Accordingly, Pichia invertase, isolated from the medium of methanol-induced cells, was treated with endo HI; the individual Man8_,,GlcNAc oligosaccharide pools were purified by gel filtration on Bio-Gel P-4. High pH anion-exchange chromatography (HPAEC) was used to determine the number of isomers in each size pool (10, l l ) , while one-and two-dimensional 'H NMR spectroscopy a t 500 MHz was employed to assign structures. Man8,gGlcNAc and the major component of ManloGlcNAc were identical to the structures found previously on Saccharomyces invertase. The remaining ManloGlcNAc and ManllGlcNAc isomers differed from those in Saccharomyces The abbreviations used are: endo H, endo-(3-N-acetylglucosaminidase H (EC 3.2.1.96); PAGE, polyacrylamide gel electrophoresis. HPAEC, high-pH anion-exchange chromatography; PAD, pulsed amperometric detection; AMMS, anionic micromembrane suppressor; SDS, sodium dodecyl sulfate. 22807 (9), in that no terminal a1,3-linked mannose was present. Rather, the novel Pichia invertase Manlo,llGlcNAc oligosaccharides were the Saccharomyces forms of Mans,loGlcNAc elongated by addition of a1,2-linked mannoses to preexisting terminal al,2-linked residues. EXPERIMENTAL PROCEDURES Materials P. pastoris strain GS115 (his4) transformed with the SUC2 expression vector pGSIO2 was used for the production of invertase (3). Invertase was also induced in Saccharomyces strain MBY-21Aa (secl8-l,trp289,leu2-3,112), provided by Dr. R. Schekman (University of California, Berkeley), which was transformed with SUC2 multicopy plasmid pRB58, from Dr. M. Carlson (Columbia University). endo H was the cloned, protease-free enzyme described previously by this laboratory (12). Resins were obtained from the following suppliers: DE-52 microgranular cellulose, Whatman, Inc.; Sephadex G-25 and protein-A-Sepharose, Pharmacia LKB Biotechnology Inc.; and Bio-Gel P-4 (-400 mesh), Bio-Rad Laboratories. Immobilon polyvinylidene difluoride membranes for Western blotting and Cls Sep-Pak cartridges were purchased from Millipore Corp. Nylon filter membranes were from Schleicher and Schuell. Invertase antibodies were raised in rabbits using carbohydrate-free internal invertase as the immunogen (13), and the IgG fraction was purified by protein A-Sepharose chromatography as directed by the manufacturer. Alkaline phosphatase-conjugated anti-rabbit IgG was from Promega Biotech and used according to the provided protocol. NMR tubes (0.5 cm, Cat. number 535pp) were from Wilmad Glass, and 99.96% 'H20 was obtained from Sigma, while 99.996% 'HZ0 was from Cambridge Isotopes Laboratory. For HPAEC with PAD sodium hydroxide (50% solution) in water and sodium acetate were from Fisher and Fluka, respectively. SDS-PAGE chemicals were obtained from Bio-Rad.

Methods
Purification of Pichia SUC2 Znuertase-Approximately 250 ml of shake-culture broth from methanol-induced, transformed Pichia cells containing 385,000 units of activity (100 mg of invertase protein) was dialyzed against several changes of 10 mM sodium phosphate buffer, pH 6.5, and loaded onto a 2 X 20-cm column of DE-52 microgranular cellulose equilibrated in the same buffer. Invertase was eluted with a linear 800-ml gradient of 0-0.3 M NaCl in 10 mM sodium phosphate, pH 6.5, at a flow rate of 40 ml/h, and 3-ml fractions were collected a t 4 "C. Approximately 95% of the activity eluted in a sharp peak at 40 mM NaCI.
The DE-52 peak was homogeneous with respect to invertase protein (4000 IU/mg) but contained a large excess of phosphomannan not covalently associated with the enzyme. The invertase and phosphomannan were coprecipitated with two volumes of cold acetone at -20 "C for 18 h. The precipitate was centrifuged for 15 min at 5000 X g and 4 "C, and the supernatant was discarded. The pellet, containing invertase and mannan, was washed with 50 ml of cold 10% trichloracetic acid, which solubilized the mannan but not the invertase. The invertase precipitate was centrifuged as above, dissolved in a minimum volume of 50 mM NaOH (-5 ml), and immediately dialyzed at 4 "C against 1 liter of sodium acetate buffer, pH 5.5. The final recovery of invertase was 93 mg (93%) which, was covalently associated with 75-80 mannose residues/60-kDa subunit, determined with the phenol sulfuric assay using mannose as a standard (14). SDS-PACE was performed on 12%, 0.5-mm-thick slab gels, and proteins were electroblotted onto Immobilon membranes for Western analysis using a Bio-Rad Mini Protean system (15). A comparison of the Pichia and s e d 8 (37 "C) invertases by SDS-PAGE Western blot is shown in Fig. 1.
Deglycosylation of Pichia Invertase-The invertase preparation (-9.3 mg of protein/ml) was heated at 100 "C for 5 min in 10 ml of 2 mM sodium acetate/0.3% SDS, pH 5.5. After cooling to room temperature, endo H was added to 100 milliunits/ml, and deglycosylation was allowed to proceed for 20 h a t 37 "C.
The released oligosaccharides were separated at room temperature from the residual protein/SDS micelles by chromatography on a 2.6-X 200-cm column of Sephadex G-25 equilibrated and eluted with water at a flow rate of 30 ml per h. The oligosaccharide-containing fractions were pooled and reduced to dryness by rotary evaporation (Buchi RllO Rotavapor). The residue was dissolved in 2 ml of 0.1 N CH&OOH/0.5% n-butanol and resolved into separate chain sizes by three cycles of chromatography on a 1.6-X 96-cm Bio-Gel P-4 (-400 mesh) column exactly as described (9). Prior to the final chromatography step on Bio-Gel P-4, which is shown in Fig. 2, the oligosaccharide pools were passed through a disposable CIS RP Sep-Pak cartridge to remove any residual peptide material.
HPAEC of Monosacharides and Oligosacharides-Analyses were performed using HPAEC with PAD essentially as described (10). The chromatograph consisted of a Dionex GPM-I1 pump a PAD-I1 detector, an Eluent Degas Module, a post-detector anionic micro-membrane suppressor unit (AMMS), and an AutoRegen pump and cartridge. The pulse potentials for the PAD-I1 were El = 0.05 V, t, = 480 ms (range 2, position 5); E2 = 0.60 V, t2 = 120 ms (position 2); and E3 = -0.60 V, t:, = 60 ms (position 1). The time constant was set to 3 s. The Dionex Eluent Degas Module was used to sparge and pressurize the eluents with helium. The system was controlled and data was collected using Dionex A1450 software. Sample injection was with a Spectra-Physics SP8880 autosampler equipped with a 2 0 0 4 sample loop. The Rheodyne injection valve was fitted with a Tefzel rotor seal to withstand the alkalinity of the eluents.
Oligosaccharide alditols of the ManGllGlcNAc Bio-Gel P-4 fractions were prepared (9) using nonradioactive NaBH4, and oligosaccharides were separated using CarboPac PAlOO columns (4 X 250 mm). Eluent 2 was 200 mM sodium acetate, eluent 3 was water, and eluent 4 was 1 M NaOH (Eluent 1 was not used for this application). Eluent 2 was filtered through a 0.2-pm nylon membrane before use. The water for all eluents was glass-distilled using a Corning Mega Pure system and collected directly into a glass reservoir. Eluent 4 was prepared by a suitable dilution of a 50% NaOH solution (19.3 M ) with water. The flow rate was 0.8 ml per min. For the preparative isolation of oligosaccharides, the post-detector AMMS was used to remove sodium ions from the eluent. The counter-current flow (from the AutoRegan pump) was 30 ml per min. Under these conditions, up to 250 mM sodium was removed. Samples were collected using an ISCO FOXY I1 fraction collector.
Electrochemical response factors of the sized oligosaccharides were determined relative to glucose. The concentration of the oligosaccharides was based on mannose, which was determined using HPAEC with PAD after hydrolysis with 2 N trifluoroacetic acid for 4 h at 100 "C (16). Monosaccharides were analyzed with a CarboPac PAlOO column (4 X 250 mm). The eluents for monosaccharide analysis were water (Eluent 1) and 200 mM sodium hydroxide (Eluent 2). Monosaccharides were eluted isocratically at 16 mM NaOH. After 20 min the column was eluted for 10 min with eluent 2, followed by return to initial conditions in 2 min. The time between injections was kept constant at 50 min in order to minimize retention time drift ( 4 min).
'H NMR Spectroscopy-Samples were flash evaporated to dryness and exchanged three times by flash evaporation from 1 ml of 99.96% 2Hy0. They were lyophilized from 1 ml of 99.996% DzO and dissolved in 0.7 ml of 99.996% 'HZO containing equimolar acetone as an internal chemical shift marker (6 = 2.225 ppm relative to 4,4-dimethyl-4silapentane sulfonate). Samples were flame-sealed in 0.5-cm NMR tubes, and two-dimensional 500-MHz 'H NMR spectra were recorded at 296 K at the Albert Einstein College of Medicine NMR facility as described (17). The two-dimensional spectra were from phase-sensitive correlation spectroscopy acquired by the method of States et al. (18). Data were analyzed on a Varian VXR4000 work station and on a Silicon Graphics Iris using Hare Research, Inc., software. Resonance intensities were integrated by cutting out and weighing peaks from expansions of the anomeric and C2-H regions of one-dimensional spectra. The anomeric proton of the core pl,4-linked mannose (residue 3) was obscured by the residual HOZH peak. However, residue 3's C2-H resonance is found at 4.241 ppm in Man,GlcNAc, at 4.154 ppm in Man9.10GlcNAc along with the C2-H of residue 4 and isolated at 4.158 ppm in ManllGlcNAc. By integration of the C1-and C2-H peaks, residues 3 and 4 were confirmed to be present at 1 mol in all oligosaccharides studied.

RESULTS AND DISCUSSION^
P. pastoris transformed with pGS102 secreted a low-molecular-weight form of SUC2 invertase into the medium as the principal protein, which was purified to homogeneity by a single DE-52 chromatography step (see "Experimental Procedures"). Nearly 95% of the enzyme was recovered, which exhibited a specific activity of over 4,000 IU/mg protein, the expected value for the homogeneous product (13). However, further characterization of the dialyzed DE-52 pool revealed the presence of 3-4-fold more carbohydrate than would be expected for an 85-kDa glycoprotein with 20% carbohydrate by weight (3,8). Much of the carbohydrate was found to be soluble phosphomannan that could be removed from the DE-52 invertase glycoprotein after coprecipitation with acetone followed by selective solubilization with 10% trichloroacetic acid. Fig. 1 shows that the final Pichia invertase preparation appears on SDS-PAGE as a ladder of species of about the same size as the endoplasmic reticulum form of invertase made by the Saccharomyces secl8 mutant at 37 "C, which has 9-11 Man8GlcNAcn core oligosaccharides/subunit (9,19).
In order to determine the size distribution and structure of the oligosaccharides associated with Pichia invertase, the protein was deglycosylated with endo H. Fig. 1 shows that endo H treatment of both the Pichia-derived and Saccharomyces secl8 (37 "C) invertases generated the expected 60-kDa form of the protein, which retains only the asparagine-proximal GlcNAc residues (5, 7). Chromatography of the released oligosaccharides revealed that over 90% of the neutral hexose was in species of the size ManGllGlcNAc (Fig. 2), while 10% or less was found in oligosaccharides eluting in the void volume of the Bio-Gel P-4 column (not shown). Upon estimating the size of the voided glycans to be ManZBoGlcNAc (5), it was calculated that less than 3% of the oligosaccharides associated with this preparation of Pichia invertase were larger than ManllGlcNAc. Given the high level of mannan in the starting material, it is probable that the large glycans   were carried through the purification as a contaminant and were not covalently associated with the invertase. There was no phosphate associated with the Mans-llGlcNAc oligosaccharides from Pichia invertase.
The distribution of oligosaccharides on the 14 potential glycosylation sites on invertase (7) ranges from 8 to 11 oligosaccharides per subunit in wild type and sed8 (37 "C) forms. Analysis of the purified Pichia invertase provided a value of 74 mannoses per subunit (see "Experimental Procedures") associated with the ManGllGlcNAc oligosaccharides (Fig. 2). The distribution of these chain lengths on an "average" subunit was estimated by dividing the total mannose recovered in each peak (Fig. 2) by the chain length, which provides the molar ratio of species present (Table 1). Normalizing the molar ratio of lengths to 74 mannoses yields the number of mannoses in each size class, which on division by the chain length gives the average number of each species/subunit ( Table 1). The distribution on the Pichia enzyme of 6 to 10 oligosaccharides with a range of chain lengths (Mans-ll) diminishes the resolution of the individual isoforms in comparison with secl8 (37 "C) invertase which is associated with 8 to 11 Man8GlcNAc chains (Fig. 1).
To determine whether the Pichia MansllGlcNAc had the same or different structures as those found previously on Saccharomyces glycoproteins (9,20,21), the separated oligosaccharide pools (Fig. 2) were subjected to one-and twodimensional 500-MHz 'H NMR spectroscopy. Fig. 3 shows the anomeric and partial C2-H proton regions of the spectra of ManGI1GlcNAc. Tables 3-6 in the miniprint section summarize the integration of proton intensities and their apportionment to specific glycosidic linkages in Man8-llGlcNAc, respectively. See Figs. 5 and 7 for relevant parts of the twodimensional phase-sensitive COSY spectra. The distribution of oligosaccharide isomers in each pool was also assessed by HPAEC, which is capable of separating high-mannose branched isomers within a given glycan size class (10, 11). A compilation of the profiles for reduced Man8_llGlcNAc-ol is shown in Fig. 4, and Table 7 provides the retention times and electrochemical response factors for the Man8-llGlcNAc-ol species. Table 2 summarizes the structures of oligosaccharides deduced in this study. All structures contained the Mans GlcNAc core (structure I), and include resonance identification numbers and linkage assignments to aid in cross-referencing residues in the figures and tables. The mannose residues of interest in the subsequent Man9-11GlcNAc structures are boxed for clarity.
Man8GlcNAc"The spectrum (Fig. 3A) and integration of resonances (Table 3) reveal Pichiu Man8GlcNAc (Table 2, structure I) to be identical to Saccharomyces invertase (20) and whole-cell glycoprotein Man8GlcNAc specie^.^ Of the three possible isomers that can be generated by trimming a Trimble single al,2-linked mannose from the precursor lipid-oligosaccharide form of MangGlcNAc (residues 9,10, or 11 in structure IIb, Table 2) only this one, which lacks the middle-arm a1,2linked terminal residue 10, has been found in all fungal glycoproteins examined to date. It is generated in Saccharomyces by the action of an endoplasmic reticulum processing mannosidase (20), which recently has been purified and shown t o generate only this isomer of MansGlcNAc in vitro (22,23). HPAEC efficiently separates the possible ManeGlcNAc branch isomers (lo), and a single species (Fig. 4A) was found using HPAEC with PAD which eluted a t 21.4 min ( Table 7). This species coeluted with the endoplasmic reticulum a-mannosidase trimming product from Saccharomyces invertase   structure I n ) . The inset in D shows an expansion of the two-component peak that eluted at 40.9 min ( Table 7).
(data not shown) (20). Thus, Pichia would appear to utilize the same early processing pathway enzymes that are present in Saccharomyces. MangGlcNAc-The Pichia invertase MangGlcNAc (Fig. 3B  and Table 4) provides a spectrum which is also identical to that found for Saccharomyces invertase MangGlcNAc (9, 24). This compound (structure IIa, Table 2) is the MangGlcNAc structure to which an a1,G-linked mannose (residue 12) has been added to the al,3-linked mannose of the lower-arm (the al,3-branch) residue 5. This configuration was originally assigned in Saccharomyces mnn mutant oligosaccharides (24) and now has been confirmed to be the isomeric form associated with wild type Saccharomyces and Pichia glycoprotein^.^ The small signal at 5.403 ppm indicates the presence of the precursor form of MangGlcNAc (20) from which mannose residue 10 was not completely removed during processing (structure IIb, Table 2). By integration of the NMR spectrum this isomer represents about 5% of the Mang pool (Table 4, isomer IIb). HPAEC with PAD of the Pichia Mang revealed two peaks (Fig. 4B). One comprised 4% of the total electrochemical response ( Table 7), which eluted where the lipid- The response factors for all species were found to be similar (2.3-2.6 relative to glucose external standard); therefore, peak area is directly related to molar proportion. Studies with monosaccharides indicate that the 2-OH of mannose are the major ionized groups during HPAEC of high mannose oligo-Kaur, S., Townsend, R. R., Liang, W., Trimble, R. B., and Burlingame, A. L., manuscript in preparation.   saccharide alditols (25). Thus, the increased retention time (approximately 4 min) of the yeast Man9GlcNAc oligosaccharide ( Table 2, structure IIa) compared to precursor Man9GlcNAc ( Table 2, structure IIb) may be explained by the net increase of one free 2-OH. Contributions of spatial factors (26), inter-and intraresidue hydrogen bonding, and cooperative interactions (27) to HPAEC retention times also have been discussed.
MunloGlcNAc-On the basis of the spectrum (Fig. 3C) and its integration (Table  5), the principal Pichiu invertase ManloGlcNAc is the species found previously on Saccharomyces invertase (9). This is the glycoprotein form of Man9GlcNAc in which al,2-linked mannose 13 has been attached to the al,6-linked residue 12 (structure IIIa, Table  2). By integration ( Table 5, structure IIIa) this species is at least 80% of the Manlo isomers in the pool. The small residual signals at 4.931 and 4.914 ppm indicate the presence of mannose in al,6-linkage to 2-0-substituted or 2-0-unsubstituted mannose, respectively (28). The sum of resonance intensities at 5.149, 4.931, and 4.914 ppm is slightly in excess of 2 mol suggesting that all Manlo isomers have both a1,6linked residues 6 and 12, plus additional al,6-linked mannose on some species. Unlike the Man9 pool, there is no signal at 5.403 ppm in the ManloGlcNAc pool to indicate the presence of any al,2-linked substitution of residue 7. The total contribution of terminal al,2-linked mannose (residues 9, 11, 13) falls about 0.2 mol short of the 3 mol expected in ManloGlcNAc (Table 2, structure IIIa), indicating that 20% of the species have only two terminal al,2-linked residues.
All Saccharomyces Mans-14GlcNAc core oligosaccharides examined to date reveal only 1 mol of resonance intensity at 5.304 ppm, a signature of the lower arm internal al,2-linked mannose residue 8. In Pichia ManloGlcNAc this resonance integrated as 1.13 mol (Table 5) indicating additional internal cul,2-linked mannose beyond that provided by residue 8. This additional intensity would result upon terminal al,2-linked mannose substitution of a pre-existing terminal al,2-linked residue, such as to 9 or 11 in structure IIa (Table 2).
Evidence to support a new 2-0-substituted a1,2-linked residue is found in the two-dimensional projections of the Man,,GlcNAc spectrum (Fig. 5C). Additional multiplets are seen in the J1.2 cross-peaks a t C1-H/C2-H -5.304/4.113 ppm of ManloGlcNAc compared to this region in the Man9GlcNAc spectrum (Fig. 5 B ) , which indicates internal a1,Z-linked mannose beyond the 1 mol provided by residue 8. In the J2.3 crosspeak region, this partial residue is found at C2-H/C3-H -4.113/3.955 ppm (Fig. 5G) and is clearly resolved from the J2,3 of residue 8. Note that the C3-H of residue 8 in MansGlcNAc is found at 3.955 ppm (Fig. 5 E ) and shifts upfield to 3.920 ppm on addition of the al,6-linked residue 12 to form ManyGlcNAc (Fig. 5F). The appearance of a partial C3-H resonance at 3.955 in Manlo (Fig. 5G) suggests that the 2-0substituted al,2-linked mannose is spatially removed from the oligosaccharide core which could result by terminal addition of al,2-linked mannose 14 to either 11 or 9.
Previous NMR studies, which examined the three possible thyroglobulin Mans isomers (20) and several IgM isomers (28), showed that the terminal al,2-linked mannose 11 in oligosaccharides with the sequence 11 8 5 Manla2ManlaZManla+ caused an -0.013 ppm of upfield shift in 5's C1-H relative to isomers which had the sequence 8 5 Manla2Manla+.
From the work of Cohen and Ballou (28), this shift appears to result from a through-space shielding of 5's C1-H anomeric center by the conformational proximity of residue 11's ring protons.
By analogy, a similar upfield shift should occur to residue 6's C1-H if an al,2-linked residue was added terminally to the upper-arm residue 9, as this would also provide two mannoses in series-linked a1,2 (structure IIIb, Table 2). Residue 6's C1-H is normally found at 5.149 ppm, but note that Fig. 3C also shows a small peak at 5.127 ppm, which integrates to about 0.1 mol of intensity (Table 5). This upfield shift in a portion of residue 6 is equivalent in magnitude to the excess 2-0-substituted mannose (0.13 mol) over that provided by residue 8 (Table 5). Furthermore, it is also the magnitude of the terminal al,6-linked mannose signal at 4.931 ppm. Thus, on the basis of the Manlo size constraint, about 12% of the Manlo pool can be assigned the structure of the major Man9 isomer with an additional al,2-linked terminal mannose attached to residue 9. (Table 5, isomer IIIb). Evidence against a1,2-terminal substitution of residue 11 is the absence of an upfield shift in residue 8's C1-H, which would be expected from the conformational interaction of the new terminal al,2-linked mannose on mannose 8* 2 residues away. On the basis of the NMR data (Fig. 3C) and the chemical shift database (20, 28), residue 14's attachment is assigned to residue 9 rather than 11. The small residual signal at 4.914 ppm, indicative of a1,6linked mannose in polymer form or to 2-0-unsubstituted mannose, can be accommodated by addition of an al,6-linked residue 15 to extend residue 12 ( Table 2, structure IIIc and  Table 5 ) . This isomer accounts for the excess of al,g-linked mannose over the 2 mol contributed by residues 6 and 1 2 ( Table 5). Verification that the minor resonance peaks at 4.931 and 4.914 ppm in the Manlo one-dimensional spectrum (Fig. 3C) were due to al,6-linked mannose was provided by their J1,* cross-peaks in a two-dimensional expansion of the COSY spectrum (not shown).
Three peaks were found after HPAEC with PAD of the ManloGlcNAc oligosaccharide pool (Fig. 4C and Table 7). A major peak (Rr = 33.5 min) comprised 72% of the electrochemical response. A shoulder on the major peak (RT = 34.2 min) and the most retained peak (RT = 37 min) gave 18 and 10% of the response, respectively. The proportion, as well as the number of species, were in agreement with the ManloGlcNAc isomers, which were deduced from the NMR spectra (Tables 2 and 5). Based on the published retentiontime trends discussed above (11,26), structure IIIc (Table 2) was assigned to the most retained peak (RT = 37 min) in the ManloGlcNAc pool.
The ManloGlcNAc fractions from invertase of Saccharomyces were compared by HPAEC to those expressed in Pichia. Fig. 6A shows that the major oligosaccharide species from Pichia and Saccharomyces co-eluted. Fig. 6, B and C, shows that the trailing shoulder of the major peak is not present in the ManloGlcNAc fraction of Saccharomyces invertase oligosaccharides. However, the last eluting peak, assigned to the al,6-extended ManloGlcNAc oligosaccharide, co-eluted in both cases.
ManllGlcNAc-This oligosaccharide pool represented a very small fraction of total Pichia invertase carbohydrate (Table l), nevertheless the one-dimensional NMR spectrum revealed unique features (Fig. 3 0 ) not seen for Saccharomyces invertase ManllGlcNAc (9,21). The first was the presence of over 1.8 mol of intensity at 5.305 ppm (Table 6), where a1,2linked 2-0-substituted mannose appears (28). Saccharomyces ManllGlcNAc preparations studied to date (9, 21) reveal 1 mol of 2-O-substituted, al,2-linked mannose (residue 8 in structure I, Table 2). The excess intensity at 5.305 ppm in Pichia ManllGlcNAc indicates that most species have an additional al,2-linked mannose in series, which means that one or more of the al&linked terminal residues 9, 11, or 13 are 2-0-substituted.
The arguments used in assigning al,B-linked terminal residue 14 to upper-arm residue 9 in ManloGlcNAc (Table 5, isomer IIIb), also apply to the major Manl1GlcNAc species. Thus, residue 14 appears to be attached to residues 9 and/or 13 rather than 11. The 1D spectrum of ManllGlcNAc (Fig.  30) provided 1 mol of intensity at 5.150 ppm and 0.9 mol shifted upfield to 5.130 ppm for al,6-linked 2-0-substituted mannose. As discussed above (20,28), this upfield shift appears to result from a through-space interaction of residue 14's ring protons with the anomeric center of residue 6 (or 12), causing a shielding effect. Although mannoses 14 and 6 (or 12) are separated by two residues in the primary sequence ( Table 2, structures IIIb and IVa), the conformation of this trisaccharide, based on a space-filling model, confirms their proximity (not shown).
The two-dimensional COSY NMR data provide evidence that both residues 9 and 13 are substituted. The J1,2 crosspeaks in Fig. 7 show that the intensity of residue 6 at Cl-H/ Evidence against a major substitution of residue 11 by residue by 14 is provided by the absence of an apparent through-space effect on residue 8's anomeric center two mannoses away, which would he expected for this configuration as discussed above for Manlo isomers (20,28).
HPAEC of Man,,GlcNAc provided evidence for four components, the major one of which represented about 78% of the total (Fig. 4 0 and Table 7, ManllGlcNAc peaks 1 and 2). Inspection of the peaks in Fig. 4 0 reveals skewing when compared to peaks for the Manalo species, suggesting isomeric heterogeneity. Reprocessing of the integrated data for Fig.40 showed the major peak (RT = 37.2 min) to consist of two components in about a 3:2 ratio, which would correspond to the 9 and 13 substituted isomers. Assigning which is the major species will be done by mass spectrometry of the isolated isomers, as one will generate a minus 5 Hex form while the other will generate a minus 4 Hex form upon fragmentation. The minor (5%) isomer on the backside of the main peak (Fig. 4D) may represent a small amount of a third possible variation where lower arm mannose 11 is terminally 2-0substituted (  (Tables 2 and 5) with additional al,2-linked terminal mannose attached either to residue 12 or 15 (Tables 2 and 6, isomers IVb and IVc). These two isomers comprise about 11% of the total and appear to correspond to the two-component peak eluting at 40.9 min, which represents 16% of the integrated area on HPAEC (Fig. 40,  inset).
The final, minor component of the Manll pool represented about 6% of the integrated area of HPAE chromatograms and eluted at 45.3 min (Fig. 40, Table 7). This isomer appears to be ManloGlcNAc isomer IIIc with an al,Z-linked terminal residue 14 attached to upper-arm residue 9 ( Table 2, structure Man G A W c IVd). This isomer accounts for the excess al,Z-linked 2-0substituted mannose over that provided by ManllGlcNAc isomers IVa, as well as most of the al,6-linked 2-0-unsubstituted mannose whose signal appears a t 4.914 ppm (Fig. 3 0 and Table 5). The long retention time of isomer IVd is attributed to the absence of al,2-substitution of the M a n l a G M a n l a b chain on the lower arm (11).
The second notable difference in Saccharomyces and Pichia ManllGlcNAc oligosaccharides is the absence of signals for al,3-linked terminal mannose in the latter (9,21,28). Although it can be difficult to distinguish al,3-linked mannose in one-dimensional spectra because the C1-H signal at 5.144 ppm coincides with 2-0-substituted al,6-linked mannose (28), the C2-H of 3-0-substituted al,2-linked mannose shifts markedly from the envelope region at 4.085 to 4.224 ppm, where it is easily integrated (9,28). None of the Pichia oligosaccharides examined in this study reveal any C2-H signal at 4.224 ppm indicative of 3-O-substituted, al,2-linked mannose (see Fig. 3).
Absence of al,3-linkage isomers was also supported by HPAEC of ManllGlcNAc from P. pastoris and S. cereuisiae invertase. The major Saccharomyces ManllGlcNAc was found to contain one al,3-linked Man, attached to residue 11 as shown in Scheme 1 (9,21). Residues designated by upper-case letters are those added on elongation of the trimmed Man8GlcNAc core after leaving the ER (9,19). Analytical HPAEC of Saccharomyces ManllGlcNAc is shown in Fig. &4, while ManllGlcNAc from Pichia is shown in B for comparison. A 1:l admixture of ManllGlcNAc from the two sources is shown in C. The major ManllGlcNAc from Saccharomyces did not coelute with any of the four ManllGlcNAc isomers from P. pastoris (Fig. 8C). Thus, HPAEC provides a clear separation of Pichia branch isomers of ManllGlcNAc with various al,2-substitutions ( Table 2, IVa-d) from the major Saccharomyces ManllGlcNAc with terminal al,3-linked mannose (Scheme 1).
In conclusion, Pichia yeast synthesize and secrete high levels of heterologous Saccharomyces SUC2 invertase, which on gels resembles in size the endoplasmic reticulum form of the enzyme synthesized at 37 "C in s e d 8 mutants (Fig. 1). On Bio-Gel P-4 ( Fig. 2), endo H-released oligosaccharides are mostly Man8,9GlcNAc with lesser amounts of Manlo,llGlcNAc  Table 1). Structural studies using high-field NMR techniques (Figs. 3,5, and 7) confirm that the majority of Mangl0GlcNAc have structures identical to those found associated with Saccharomyces invertase (Tables 2-5 Tables 2, 5, and 6). A proposed metabolic interrelationship of the oligosaccharide structures deduced in this study is summarized in Fig. 9.
These studies confirm and extend our previous findings, which demonstrate the ability of HPAEC to separate not only linkage isomers, but also oligosaccharides which differ only in the branch position of a single residue (10). Examples of this resolving power are the separation of precursor and yeastprocessed MangGlcNAc linkage isomers ( Fig. 4B and Table   2, structures IIa and IJb), ManloGlcNAc linkage isomers IIIa and IIIb from IIIc ( Fig. 4C and Table 2), and the unique elution position of ManllGlcNAc from Saccharomyces (Scheme I), which contains an a1,3-linked Man in contrast to the same size oligosaccharides from Pichia with a1,Z-linked termini (Fig. 8). Under the conditions described, linkage isomers were separated by at least 4 min. Branch isomers (no change in substitution) were more difficult to resolve and appeared as shoulders and asymmetrical peaks (Fig. 4 0 ) , but the indicated heterogeneity was completely consistent with structural assignments made by NMR analysis.
An unexpected finding in this study was that Pichia invertase oligosaccharides do not undergo terminal addition of a1,3-linked mannose during processing of the core Mans-GlcNAc to larger species. It is not currently known whether Pichiu can add al,3-linked terminal residues to any cellular glycans, but this is under investigation. Although the MangGlcNAc pool revealed a minor contaminant of the lipidoligosaccharide form of Man9 (structure I n , Tables 2 and 4), the one-and two-dimensional NMR spectra of larger species did not reveal signals from residual glucose residues. Therefore, if glucose is initially present on Pichia lipid-oligosaccharide, as is found in other fungal species (29,30), it is efficiently removed during processing.
As a final note, it has been suggested that because Pichia glycans are shorter than Saccharomyces forms (8), they may be more "mammal-like" than "yeast-like.'' The structural studies reported here show that, although 75% of the oligosaccharides are Mans,gGlcNAc, 95% of the MangGlcNAc and all of the Manlo,llGlcNAc are isomers uniquely found on fungal glycoproteins. From a biotechnological perspective, it is not known at present the extent to which the Pichia Mang-llGlcNAc species would be antigenic in mammals, but studies addressing this question also are in progress. and from Saccharomvces invertase (9) were mixed and chromalagraphed using two Serial CarbaPac PAlW columns and a PAtW guard column. The CoIumnSwere equihbrated in 12 mM sodium acetate (E2: 6%) and tW mM NaOH (E4: Iwb). The oligosaccharides were eluted with a linear gradient up to 26 mM sodium acatate over 50 mm. Afler to min at the limn conditions. the columns were returned to initial condtions in 2 min. Panels B and C are chromatograms of approximately 250 pmols 01 Saccharomvces and Plchia invertase Man,,GicNAc's. respsctiyely.