The Accumulation of Man6GlcNAc2-PP-dolichol in theSaccharomyces cerevisiae Δalg9 Mutant Reveals a Regulatory Role for the Alg3p α1,3-Man Middle-arm Addition in Downstream Oligosaccharide-lipid and Glycoprotein Glycan Processing*

N-Glycans in nearly all eukaryotes are derived by transfer of a precursor Glc3Man9GlcNAc2 from dolichol (Dol) to consensus Asn residues in nascent proteins in the endoplasmic reticulum. The Saccharomyces cerevisiae alg(asparagine-linked glycosylation) mutants fail to synthesize oligosaccharide-lipid properly, and thealg9 mutant, accumulates Man6GlcNAc2-PP-Dol. High-field 1H NMR and methylation analyses of Man6GlcNAc2released with peptide-N-glycosidase F from invertase secreted by Δalg9 yeast showed its structure to be Manα1,2Manα1,2Manα1,3(Manα1,3Manα1,6)-Manβ1,4GlcNAcβ1,4GlcNAcα/β, confirming the addition of the α1,3-linked Man to Man5GlcNAc2-PP-Dol prior to the addition of the final upper-arm α1,6-linked Man. This Man6GlcNAc2 is the endoglycosidase H-sensitive product of the Alg3p step. The Δalg9Hex7–10GlcNAc2 elongation intermediates were released from invertase and similarly analyzed. When compared withalg3 sec18 and wild-type core mannans, Δalg9 N-glycans reveal a regulatory role for the Alg3p-dependent α1,3-linked Man in subsequent oligosaccharide-lipid and glycoprotein glycan maturation. The presence of this Man appears to provide structural information potentiating the downstream action of the endoplasmic reticulum glucosyltransferases Alg6p, Alg8p and Alg10p, glucosidases Gls1p and Gls2p, and the Golgi Och1p outerchain α1,6-Man branch-initiating mannosyltransferase.

With the exception of some protists, all eukaryotes co-translationally transfer Glc 3 Man 9 GlcNAc 2 from a Dol 1 intermediate to selected Asn residues designated by the sequon, Asn-Xaa-Ser/Thr, where Xaa is any amino acid except Pro. We are beginning to understand why eukaryotes expend so much en-ergy synthesizing this evolutionarily conserved tetradecasaccharide, only to modify it by removal of the entire glucotriose unit and one or more of the core mannose residues. It is intuitive that the biochemical intermediates formed provide information necessary for downstream oligosaccharide processing events, and recent evidence for this has been provided by studies on the Alg10p ␣1,2-glucosyltransferase (1). Early work with mammalian systems provided much of what is currently known concerning N-linked glycosylation (2,3), including a fair picture of the distribution of oligosaccharide-processing activities through subcellular fractionation, lectin binding, and immunolocalization techniques (4). However, with the exception of the Lec Chinese hamster ovary cells, the genetic approach to glycoprotein study in somatic cells has proven difficult.
Yeast have provided a more amenable system of study with the development of three classes of mutants. In Saccharomyces cerevisiae they are alg, mnn, and sec mutants. N-Glycan synthesis in S. cerevisiae proceeds much like that of higher eukaryotes up through the early ER processing events, including removal of the glucotriose moiety and the central arm ␣1,2linked mannose (5,6). Although subsequent processing differs from that found in higher eukaryotes, this yeast provides a good model system to investigate the structural information contained in the Glc 3 Man 9 GlcNAc 2 -PP-Dol precursor transferred to protein, since nearly the entire phosphorodolichol and much of the Golgi processing pathways have been dissected by the generation of the alg and mnn mutants (7)(8)(9)(10).
Genes for a number of steps in OSL biosynthesis were not readily identified in yeast, because their absence produced no apparent phenotype in the presence of wild-type OST. However, in recent work, Aebi's group has isolated ALG3, ALG6, ALG8, ALG9, and ALG10 loci by rescuing through complementation the respective synthetically lethal phenotypes occurring in conjunction with mutant subunits of OST (1,(11)(12)(13). Subsequent to their isolation, strains for each deletion have been generated in a wild-type OST background. These mutants provide an advantage for investigating Golgi-processed N-glycan structures, because gene ablation prevents N-linked glycan structures associated with leaky alleles, for example alg3-1 (14).
The alg9 mutants accumulate endo H-sensitive Man 6 GlcNAc 2 -PP-Dol, the glycan of which is shown in this work to be Man␣1,2Man␣1,2Man␣1,3(Man␣1,3Man␣1,6)-Man␤1,4GlcNAc␤1,4GlcNAc␣,␤, differing from Man 5 GlcNAc 2 -PP-Dol of alg3 mutants only by the addition of the central arm-initiating ␣1,3-linked Man (see residue 7 in Scheme I). In addition, we define downstream processing of ⌬alg9 glycans found on secreted invertase as was done previously on wildtype and alg3 sec18 invertase as a model secreted glycoprotein (15,16). This study reveals the strong regulatory influence addition of this central arm ␣1,3-linked Man plays in subsequent ER and Golgi processing events. Findings here provide implications regarding the in vivo substrate specificities of the ER glucosyltransferases and glucosidases and of the Golgi Och1p ␣1,6-Man branch-initiating enzyme.

EXPERIMENTAL PROCEDURES
Materials-S. cerevisiae haploid strain yg414 (Mat␣ ade2-101 ura3-52 his3D200 lys2-801⌬alg9::G418), was supplied by S. te Heesen (ETH, Zurich, Switzerland). The ⌬alg9 strain was transformed with the pRB58 multicopy yeast invertase plasmid for overexpression using the lithium acetate procedure (17). Bio-Gel P-4 was from Bio-Rad. Sephacryl S-300 Superfine was a product of Amersham Pharmacia Biotech, and DE52-cellulose was obtained from Whatman. Cellulose TLC plates were purchased from EM Separations Technology. The [2-3 H]Man was obtained from American Radiolabeled Chemicals, Inc. Ecolume scintillation mixture was purchased from ICN Pharmaceuticals, Inc. Sigma was the source of 99.8% and 99.96% D 2 O, while 99.996% D 2 O was from Cambridge Isotopes Laboratories. Man 3 GlcNAc[ 3 H]ol was from a previous study (16). Endo H and PN-Gase F were prepared as described previously (18,19). Silica-gel 60A TLC plates were purchased from Whatman. Castanospermine was obtained from Sigma. All solvents were American Chemical Society "reagent grade" or better.
Yeast Microsomes-Microsomal membranes were prepared as described previously (20). Briefly, 10 g of ⌬alg9 S. cerevisiae yeast cells were resuspended in 20 ml of buffer (50 mM Tris-Cl, pH 7.5, 5 mM MgCl 2 , 10 mM ␤-mercaptoethanol) and transferred to a Bead Beater TM (Biospec Products Bartlesville, OK) along with 20 ml of acid-washed Ballantoni glass beads (0.45-0.55 mm). The yeast suspension was beaten for 60 s and cooled on ice for 5 min. This was repeated six times or until the yeast cells were more than 90% disrupted as determined in the light microscope (magnification, ϫ440). The slurry was removed from the blender, rinsed with buffer, and centrifuged at 2,000 ϫ g for 10 min at 4°C in 40-ml polycarbonate tubes to remove the nuclei and cell walls. The supernatant was transferred into a 40-ml polycarbonate tube and centrifuged at 40,000 ϫ g for 60 min at 4°C. The supernatant was discarded, and the microsomal pellet without the brownish mitochondrial upper layer was gently resuspended in buffer and dispersed in a small Dounce homogenizer with a Teflon pestle. The final protein concentration was determined by the method of Bearden (21) using bovine serum albumin (Sigma) as a standard. Microsomes (30 -40 mg of protein/ml) were frozen as 30-l beads in liquid N 2 and stored at Ϫ70°C.
Glycolipid Biosynthesis in Vitro-Glycolipids were synthesized as described previously (22). Incorporation of [ 3 H]Man from GDP-[ 3 H]Man or [ 14 C]Man from GDP-[ 14 C]Man into endogenous yeast acceptors was performed in a final reaction volume of 110 l. For synthesis of preparative amounts of ⌬alg9 OSL, the reactions were incubated for 20 min at 27°C. All reactions were terminated by addition of 2 ml of CHCl 3 : CH 3 OH (2:1 v/v). The glycolipids were extracted by a modified procedure of Waechter et al. (20). The CHCl 3 /CH 3 OH (2:1) and CHCl 3 / CH 3 OH/dH 2 O (1:1:0.3) extracts contained Man-P-Dol and OSL, respectively, based on their behavior on silica-gel and cellulose TLC. For preparation of large amounts of OSL, reactions were scaled up 2-fold. To reduce OSL glucosylation, some reactions contained 10 M UDP. The resulting OSL was purified on cellulose TLC plates using Man-P-Dol and/or OSL standards. Purified OSL was scraped from TLC plates, resuspended in 4 ml of CHCl 3 /CH 3 OH/dH 2 O (1:1:0.3), the matrix removed by centrifugation, the supernatant concentrated under reduced pressure at 28°C, and the residue resuspended in a small volume of fresh CHCl 3 /CH 3 OH/dH 2 O (1:1:0.3).
Glycosidase Digestions-To isolate transferred oligosaccharides, microsomal pellets were solubilized in 1% SDS in 50 mM sodium phosphate buffer (pH 8.5) with heating. For endo H digestion, the pH was adjusted to 5.5 with 1.0 M phosphoric acid and endo H added at 50 milliunits/ml. The reactions were incubated at 37°C for 16 h, and endo H activity was verified by hydrolysis of Man 6 GlcNAc 4 -Asn-dansyl, followed by paper chromatography of the released GlcNAc-Asn-dansyl moiety (23). For PNGase F digestions, SDS was removed from the solubilized pellet protein by extraction with 80% acetone. The resulting pellets were resuspended in 50 mM sodium phosphate buffer, pH 8.5, boiled to resolubilize them, and PNGase F was added to 100 milliunits/ ml. The reaction was incubated at 30°C for 16 h, and activity was verified by hydrolysis of [ 3 H]dansyl-fetuin-pentaglycopeptide, followed by paper chromatography (24).
Purification of External Invertase and Oligosaccharide Isolation-Secreted invertase was purified from crude cell extracts by 35% ammonium sulfate and pH 4 precipitations, followed by DE-52 and Sephacryl S-300 chromatography as described previously (15). A typical purification from 225 g of cells yielded 47 mg of external invertase. The Nlinked oligosaccharides were hydrolyzed from invertase by treatment with PNGase F as described above, isolated by solvent precipitation as described previously (25), then chromatographed on a calibrated column of Bio-Gel P-4 (95 cm ϫ 16 mm) with 0.1 N acetic acid, 1% 1-butanol as the eluant at 8.8 ml/h and room temperature. Fractions of 0.73 ml were collected, and aliquots were assayed for total hexose and radioactivity, which included an internal marker of Man 3 GlcNAc-[ 3 H]ol. Mass Spectrometry-MALDI-TOF mass spectrometry was performed on a Bruker Reflex instrument. Samples of 25-50 pmol were prepared with 2,5-dihydrobenzoic acid as matrix. Data accumulated for 10 -50 3-ns pulses of the 337 nm laser were averaged for each sample. Analyses were performed in linear and reflective mode.
HPAEC Branch Isomer Analysis-Pooled aliquots were chromatographed on a Dionex HPAEC system using a voltage PAD response detector and a PA-100 column. Samples were separated using 100 mM NaOH accompanied by the following sodium acetate gradient: isocratic at 35 mM for 5 min, then 35-170 mM over 45 min. Individual runs included raffinose or known glycans as internal standards.
Invertase Assay-External invertase activity was followed through the purification scheme using a modification of the method of Goldstein and Lampen (26).
Methylation Linkage Analysis-Samples were analyzed by methylation as described (27) using the NaOH/Me 2 SO method. Briefly, the free hydroxyls of the oligosaccharides were deprotonated with NaOH/ Me 2 SO. Then CH 3 I was added to replace the free hydroxyls with methoxy groups. The methoxylated oligosaccharide was hydrolyzed in strong acid, evaporated under low pressure, and applied to Whatman Silica-gel 60A TLC plates. The plates were developed twice with CH 3 CN-CHCl 3 -CH 3 OH, 3:9:1 (v/v/v), thoroughly dried between each ascent, and rapidly dipped into a solution containing 3 g of N-(naphthyl)ethylenediamine and 50 ml of concentrated H 2 SO 4 in 1 liter of CH 3 OH. The plates were dried and placed in an oven for 10 min at 120°C. All saccharide standards were purchased from Sigma. 1 H NMR Spectroscopy-Oligosaccharides (0.2-0.5 mg) were exchanged three times by rotary evaporation from 99.8% D 2 O and twice from 99.96% D 2 O, followed by lyophilization. Lyophilized samples were dried over P 2 O 5 in vacuo for a day or more, then reconstituted in 0.5 ml of 99.996% D 2 O to a final concentration of 0.25-0.70 mM. Samples were quickly transferred to 5-mm tubes (Wilmad Co., no. 535pp, previously washed and exchanged with 99.8% D 2 O), flame-sealed, and examined at 300 K by 1D and 2D DQF-COSY phase-sensitive 1 H NMR spectroscopy at 500 MHz as described previously (28 -30). Spectral width in the 11.7-tesla field was 1502 Hz for all experiments. For acquisition of 1D data, 1024 scans were collected over 4096 data points. The limit of resolution was 0.0045 ppm based on the ratio of the width of the widest peak at half-height (2.26 Hz) to the number of Hz per ppm (500.13 Hz/ppm). For homonuclear 2D DQF-COSY, 1.5 s of low power presaturation on residual HDO at 4.79 ppm was applied. Data collection for the 2D experiments was 4096 data points in t 2 and 512 complex data points in the indirect t 1 dimension.
Glucosylation in Vivo-⌬alg9 cells were grown overnight to stationary phase and collected by centrifugation for 5 min at 3000 rpm and room temperature in a Sorvall TC6 centrifuge equipped with an H400 rotor. The yeast were washed twice in glucose-free YP medium by centrifugation, and incubated in the presence or absence of 5 mM castanospermine (Sigma) for one 1.5 h in YP ϩ 1% glucose. The yeast were again washed twice in glucose-free YP medium by centrifugation, and cells (2 ϫ 10 9 ) were resuspended in 400 l of YP ϩ 0.15% glucose containing 500 Ci of [2-3 H]Man (20 Ci/mmol) (American Radiolabeled Chemicals Inc.). After 2 min of labeling 2000-fold excess of unlabeled Man was added to reactions. Aliquots of 125 l were taken at 0, 1, and 10 min, and reactions were terminated by rapid addition to 4 ml of CHCl 3 /CH 3 OH (2:1) while vortexing.
General Methods-Neutral hexose was determined by a scaled-down version of the phenol-sulfuric assay with mannose as the standard (31). Protein was determined using either the method of Bearden (21), with bovine serum albumin as standard, or in the case of purified invertase by absorbance at 280 nm (25). Radioactivity was measured with a Beckman LS-3801 scintillation spectrometer in Ecolume (ICN) scintillation fluor.

Characterization of ⌬alg9
Invertase N-Linked Glycans-⌬alg9 yeast were transformed with pRB58 (see "Experimental Procedures"), and several transformants producing invertase activities of over 700 IU/g (wet weight) were accessioned. The best overproducer of invertase, JCY362, was chosen for external invertase purification from the ⌬alg9 background, which yielded high specific activity enzyme (79 mg, 3990 units/mg protein) with an overall recovery of 72% from 535 g of cells.
Bio-Gel P-4 size exclusion chromatography of the PNGase Freleased glycans from this invertase preparation provided four major pools, labeled A through D in Fig. 1, which eluted on the calibrated column in volumes consistent with Hex 6 GlcNAc 2 (fractions130 -136), Hex 7 GlcNAc 2 (fractions 124 -129), Hex 8.5 GlcNAc 2 (fractions 118 -123), and Hex 10 GlcNAc 2 (fractions 113-117), respectively. Each pool was rechromatographed, and the central 85% of the resulting peaks were re-pooled for further analysis (data not shown). Fig. 2A shows the MS spectrum of pool A, revealing one size isomer with masses of 1420 and 1436 Da for the sodium-and potassium-adducted species, respectively, the expected M r for a Hex 6 GlcNAc 2 oligosaccharide. The HPAEC data (Fig. 3) confirm that pool A consisted of a single Hex 6 GlcNAc 2 isomer. The inset in Fig. 2A shows that endo H at 50 milliunits/ml for16 h at 37°C, conditions that will not hydrolyze the reducing-end GlcNAc from the alg3 Man 5 GlcNAc 2 precursor (25), removed a mass equivalent to the reducing-end GlcNAc (204 Da). This result is consistent with the notion that the Hex 6 GlcNAc 2 is the first endo H-sensitive structure in the N-linked glycosylation pathway, the expected product of the Alg3p step shown in Scheme IC (32).

MALDI-TOF MS Analysis of Bio-Gel P-4 Pools A-D-
Pool B contained two glycan sizes; major ions of Hex 7 GlcNAc 2 appeared at 1581 and 1597 Da for the sodium and potassium forms, respectively, while lesser intensities were seen for ions at 1743 and 1759 Da, consistent with the same ion forms of Hex 8 GlcNAc 2 (Fig. 2B). Although MS is not wholly quantitative, signal intensities from the size neighbors in a homologous structural series should be relatively proportional. On this basis, Hex 7 GlcNAc 2 represented approximately 90% of pool B while Hex 8 GlcNAc 2 represented approximately 10%. As Fig. 2C shows, pool C contained both Hex 8 GlcNAc 2 and Hex 9 GlcNAc 2 , the former comprising approximately 20% of the pool while the latter made up the remaining 80%. Pool D contained a single glycan size consistent with Hex 10 GlcNAc 2 sodium and potassium forms at 2067/2083 Da, respectively.
HPAEC Analysis-In order to estimate the number of branch isomers present in the four oligosaccharide pools, each was analyzed by HPAEC using an analytical Dionex CarboPak PA-100 column (4 ϫ 250 mm). As indicated above, pool A gave a single peak (Fig. 3) and co-eluted with the smallest ⌬alg9 glycan from OSL (data not shown), consistent with the hypothesis that it is the core alg9 isomer unmodified by further processing after transfer to protein and transport through the secretory pathway. Quantitation of pool A from the Bio-Gel P-4 column indicated this to be the most abundant isomer among the Hex 6 -10 GlcNAc 2 structures, representing 24% of the N- glycan chains found on secreted invertase. Fig. 3 shows that pool B yielded five peaks corresponding to the presence of a minimum of five branch isomers, with the most abundant glycan species representing 53% of the profile area. The remaining four minor peaks contained 24%, 16%, 4%, and 2% of the total profile area, respectively.
Pool C provided six peaks, where the major species represented approximately 68% of the total area of the analytes (Fig.  3). The other peaks were 10%, 10%, 6%, 4%, and 2% of the profile area, respectively. Pools B and C contained one common peak eluting at 11 min, which represented 2% of the area in pool B and 10% of the area in pool C. Since pools B and C both have a Man 8 GlcNAc 2 component, this peak likely represents an isomer common to both pools. Finally, in pool D five glycan peaks were separated. The main species represented 61% of the chromatogram area. Other isomers represented 24%, 7%, 4%, and 4% of the total peak area present in the HPAEC profile.
NMR Analysis of ⌬Alg9 Core Oligosaccharides-In order to determine what structures were present in each of the Bio-Gel P-4 pools A-D, 1D 1 H NMR spectra were collected for each. These are shown as a montage in Fig. 4, and integrations of proton intensities for established reporter groups present in expansions of the spectra are summarized in Table I. Also present in Table I is the assignment of the proton intensities for the monosaccharide constituents in the Man 5 GlcNAc 2 oligosaccharide that accumulates in alg3 yeast, the precursor of the ALG3 step, whose structure is shown in Scheme IB. All structures assigned for ⌬alg9 glycans in the current work had as their core GlcNAc residues 1 and 2 and Man residues 3, 4, 5, 8, and 11 (Scheme IB). Fractional molar proton intensities in pools A-D (Table I) were assigned as additions to this Man 5 GlcNAc 2 core, which resulted in 18 structurally related members of a homologous biosynthetic series summarized in Scheme II. The isomer identifications in Scheme II relate the hexose number present and the order in which fractional proton intensities were used to deduce the structures, e.g. isomer 7a denotes the first Hex 7 GlcNAc 2 configuration assigned in pool B.
Within each pool, the number and amount of each isomer assigned agreed closely with the number and area of HPAEC peaks present (Fig. 3) and size proportions estimated by MALDI-TOF MS (Fig. 2). The subsequent paragraphs explain how each assignment was made. Some assignments required DQF-COSY experiments for validation, which are presented in Figs. 5 and 6.
Pool A: Hex 6 GlcNAc 2 -The smallest oligosaccharide found SCHEME I. Representative glycan structures and their anomeric 1 H NMR resonances (␦, ppm). A, the Glc 3 Man 9 GlcNAc 2 transferred to protein in wild-type cells (boxed area indicates the residues conserved in the core alg9 glycan); B, alg3 Man 5 GlcNAc 2 ; C, alg9 Man 6 GlcNAc 2 ; D, the alg9 glycan containing all Golgi modifications seen in this study. The Man added by Och1p is boxed (36). on secreted invertase in ⌬alg9 yeast was Hex 6 GlcNAc 2 , which was completely endo H-sensitive to removal of the reducing end GlcNAc ( Fig. 2A, inset). Given the substrate specificity of endo H, the ⌬alg9 Hex 6 GlcNAc 2 structure should be the Alg3p product, the cytosolic Man 5 GlcNAc 2 precursor with the added middle arm ␣1,3-linked Man residue 7 (Scheme I, B and C, respectively). The determination of the glycan structures in a homologous series by high field proton NMR is quite straightforward, because the addition of subsequent glycosidic linkages results in characteristic chemical shift resonances that can be compared with large libraries of related glycan proton chemical shifts available in the literature and data bases. Importantly, the 1D spectra can be accurately integrated to assign the components of complex mixtures (33).
Integration of signals in Fig. 4A ( Table I, pool A) between 4.70 and 5.40 ppm provides 6.00 mol of hexose protons, including 1 mol for residue 3 obscured by residual HDO signal centered at 4.78 ppm. Note the 1-mol resonance intensity integrated at 5.11 ppm, which is indicative of terminal ␣1,3-linked Man residue 7 (33,34). This resonance is the only detectable difference on comparing the ⌬alg9 Man 6 GlcNAc 2 NMR spectrum with that of alg3-1 Man 5 GlcNAc 2 (14) described previously and included in Scheme I. A 2D DQF-COSY experiment (data not shown) showed J 1,2 cross-peak for residue 7 at 5.11(C1-H)/4.05(C2-H) ppm, confirming this assignment (34). In addition, residue 4's C2-H shifts downfield on 3-O substitution from 3.97 to 4.12 ppm, verifying addition of residue 7 to residue 4 (Scheme IC). As will be seen, this resonance is present in all glycans Hex 7 GlcNAc 2 and larger, which means the ⌬alg9 Man 6 GlcNAc 2 is an invariant core component of all the Man Ͼ6 GlNAc 2 structures elucidated.
In one preparation of ⌬alg9 Hex 6 GlcNAc 2 , Glc 1 Man 5 GlcNAc 2 was found to be a minor species (8%). This assignment was made on the basis of the total anomeric proton integration and presence of ϳ0.08 mol of Glc anomeric resonance intensity at 5.25 ppm J 1,2 coupled (3.5 Hz) with its C2-H at 3.54 ppm as expected for an ␣-anomeric proton with an axial C2-H as found in Glc (14 -16, 25, 34, 35). This species was present in the shoulder of the P-4 Bio-Gel fractions most distant from the V o and is not considered further when assigning Hex 7-10 GlcNAc 2 structures.
Pool B: Hex 7-8 GlcNAc 2 -Pool B integrated to a total of 7.10 mol of anomeric protons (Table I), assuming 1 mol of intensity for residue 3 obscured by the HDO signal. This indicates that 90% of the pool is Hex 7 GlcNAc 2 and 10% is Hex 8 GlcNAc 2 , in good agreement with MALDI-TOF MS distribution observed in Fig. 2B. C1-H resonance intensity in excess of the core Man 6 GlcNAc 2 structure was present at chemical shifts 4.89, 5.03, 5.12, and 5.25 ppm (Table I). At 4.89 ppm, 1.20 mol of resonance intensity was present. One mol was from residue 4 of the core structure (Scheme IB), and 0.20 mol was assigned to the branch-forming unsubstituted ␣1,6-linked Man residue 12. The resonances for 4 and 12 correlate with their definitive C2-H resonances at 4.13 and 3.98 ppm, respectively (14 -16, 25, 34). Thus, 20% of Pool B is isomer 7a (Scheme II), which is the expected product of the Och1p outer chain initiating enzyme (36).
At 5.03 ppm 1.06 mol of resonance intensity was integrated. One mol of this signal was assigned to residue 11 of the core structure and the additional 0.06 mol to residue 13, the ␣1,2-Man cap on residue 12 (see Scheme ID). The predicted 2D DQF-COSY cross-peaks for these residues were observed at 5.03(C1-H)/4.08(C2-H) ppm (Fig. 5A) as expected for all terminal ␣1,2-linked mannoses. Together with the size constraint placed on the pool by total integrated C1-H resonance intensity, mass spectrometry data, and the above resonance assignments, ϳ6% of pool B contain residues 12 and 13 as their Golgi modification to the ⌬alg9 Man 6 GlcNAc 2 core to yield isomer 8c (Scheme II).
The total resonance intensity between 5.11 and 5.15 ppm was 1.74 mol. One mol is from ␣1,3-linked Man residue 7 in the core structure (Scheme IC), while the other 0.74 mol represents ␣1,3-linked Man residue 14 and the 2-O-substituted residue 12 in isomer 8c assigned above. Both of the 3-O-substituting Man's C2-Hs were observed in the 2D DQF-COSY spectrum at 4.08 ppm (Fig. 5A) (15,25,37). Subtracting 0.06 mol of residue 12 present in 8c from the 0.74 mol integrated leaves 0.68 mol of ␣1,3-linked Man C1-H resonance, which is seen as residue 14 on the basis of the intense 2D DQF-COSY cross-peak at 5.14(C1-H)/4.06(C2-H) ppm (Fig. 5A). This defines isomer 7b (Scheme II) as a major constituent of pool B. Furthermore, A very weak base-line 5.12(C1-H)/4.22(C2-H) ppm resonance was also observed, but is not apparent in the projection in Fig.  5A. This cross-peak is assigned to residue 16, which was previously shown to be due to a novel 3-O-substituting mannose in alg3 sec18 invertase oligosaccharides (15,25). As will be shown below, components of the Hex 8 -9 GlcNAc 2 and Hex 10 GlcNAc 2 pools also contain the Man␣1,3Man␣1,3-disaccharide. Furthermore, a minor component of pool B (2%) eluted at 11 min on HPAEC with a component of the Hex 8 -9 GlcNAc 2 isomers in pool C (see Fig. 3), also predicted to contain residue 16 (see Table I, Scheme II, and Fig. 5B). These two are most probably the same isomer, and its presence in pool B may have resulted from combining overlapping fractions from multiple Bio-Gel P-4 runs. Taken together, these calculations are consistent with about 60% of pool B being isomer 7b and about 4% isomer 8d (Scheme II).
The resonance intensity at 5.25 ppm was 0.10 mol (Table I) and indicates the presence of residue G 1 . This assignment is supported by the 2D DQF-COSY cross-peak at 5.25(C1-H)/ 3.54(C2-H) ppm (Fig. 5A), which is in the region expected for this axial ring proton (25,35). The resonance peak indicates that 10% of pool B is the core ⌬alg9 glycan, assigned isomer 7c, which retained the ␣1,3-Glc on transport through the Golgi after leaving the ER (Scheme II).
For the oligosaccharide structures studied here, a C2-H resonance intensity can appear at 4.22 ppm from residue 3 in the absence of the through space effect caused by residue 12's substitution of residue 5 (15,37), the 3-O substitution of residue 11 by 14 or G 1 (25), or the 3-O substitution of residue 14 with 16 (see Scheme ID). In pool B, 1.52 mol of resonance intensity was present at 4.22 ppm (Table I), which, as noted above, reveals a principle J 1,2 cross-peak in 2D DQF-COSY spectrum with residue 11's C1-H at 5.04 ppm, already assigned to the major constituent of this pool, isomer 7b. Isomers 7b and Pool C: Hex 8 -9 GlcNAc 2 -The total Man and Glc anomeric proton resonance intensity of Pool C was 8.82 mol, including 1 mol of resonance intensity for HDO-obscured residue 3, which corresponds to 82% Hex 9 GlcNAc 2 and 18% Hex 8 GlcNAc 2 (Table I). These values are in agreement with mass spectrometry profile (Fig. 3C). At 4.89 -4.91 ppm there was 1.18 mol of resonance intensity for ␣1,6-linked Man residues 4 and 12 (see Scheme ID). Subtracting 1 mol of intensity for the common core residue 4 leaves 0.18 mol by difference, indicating that 18% of the pool isomers had an unsubstituted residue 12, verified (see Fig. 5B) by the 2D DQF-COSY 4.91(C1-H)/3.97(C2-H) ppm cross-peak (15). Because the upper arm is abbreviated due to the alg9 lesion and pool C glycans are limited to 14. c Core residue 3 could not be integrated accurately due to water suppression, but must be present in all N-linked glycan cores. Hex 8 -9 GlcNAc 2 size, all isomers with unsubstituted 12, or without 12 altogether, must contain a lower arm substitution on residue 11. These can be 3-O substitutions by G 1 or 14 (15). Residue 14 can also be 3-O-substituted with 16. Representations of these structures are shown in Scheme II as isomers 8a, 8b, 8d, and 9a.
Between 5.03 and 5.05 ppm, 1.72 mol of ␣1,2-linked terminal Man anomeric protons were integrated (Table I). One mol is from the core residue 11 (Scheme IC), and the other 0.72 mol results from residue 13, which 2-O-substitutes 12. Between 5.11 and 5.14 ppm, there was 2.82 mol of proton resonance intensity. One mol is from core residue 7 (Scheme ID). Residue 12's cross-peak, when substituted with 13, is found at 5.14(C1-H)/4.02(C2-H) ppm (37) (Fig. 5B), and indicates that 72% of the pool contains structures in which the ␣1,6-Man outer chain branch initiation has an ␣1,2-Man cap. Thus, residue 7 and 2-O-substituted 12 account for 1.72 of the 2.82 mol of resonance intensity between 5.11 and 5.14 ppm, which leaves 1.10 mol to assign.
Subtracting isomer 8d's contribution (ϳ0.10 mol) to the 1.20 mol of intensity at 4.22 ppm leaves 1.10 mol, which correlates exactly with the residual 1.10 mol of 5.14 ppm resonance intensity calculated above (2.82 mol total minus 1.72 mol for 7 and 2-O-substituted 12) and assignable to terminal ␣1,3-linked SCHEME II. Interrelationship of ⌬alg9 Hex 6 -10 GlcNAc 2 species deduced in this study. The identifiers for structures are those used in the text. The regions of the secretory pathway in which the reactions occur are indicated at the top as bracketed areas denoted ER and Golgi. Glycans of interest are indicated as: A, alg3-type Glc 1 Man 5 GlcNAc 2 ; B, the family of N-glycans exhibiting persistent Glc; C, mature wild-type core-filled isomers or substrates for elongation to "mannan"; D, pathways leading to structures containing 16 (see text); E, structures leading to both normal core-filling or isomers containing 16. The asterisk appears above the glucosylated structures deduced from in vivo [2-3 H]Man labeling studies (see text). Glycans 8x and 9x were not seen but are implied by the presence of 10d. The Alg3p substrate Man 5 GlcNAc 2 from which all structures in this study are derived is indicated by Ќ.
Man. These are assigned collectively to 3-O-substituted residues 14, and/or 15, whose C1-, C2-, and C3-H cross-peaks have been identified above. Structure 9c can be assigned as the major component (68%) of pool C on the basis of its Man 9 GlcNAc 2 size, the large signal for 2-O-substituted 12 (Fig.  5B), and one ␣1,3-linked terminal Man distributed between residues 11 and 13. Accounting for all integrated protons, 2D DQF-COSY cross-peak intensities, and Hex 8,9 GlcNAc 2 size distribution of the pool, the remaining isomers in pool C are estimated to be: 8a, 6%; 8b, 2%; 8d, 10%; 9a, 10%; and 9b, 4%, which are in very good agreement with the number of separated peaks and their areas seen in pool C by analytical HPAEC (Fig. 3).
Pool D: Hex 10 GlcNAc 2 Pool-The total integrated Man and Glc anomeric proton intensity of this pool was 10.0 mol, which agrees with a single MALDI-TOF MS mass for a Hex 10 GlcNAc 2 sized glycan (Fig. 2D). Increased C1-H resonance intensity above that provided by the core Hex 6 GlcNAc 2 was observed at 5.25, 5.14, 5.04, and 4.89 ppm (Table I).
At 4.89 -4.91 ppm 1.07 mol of resonance intensity for 2-Ounsubstituted ␣1,6-linked Man was integrated. Subtracting 1 mol for core residue 4 leaves 0.07 mol. Since 12 is fully substituted by 13, in order to satisfy the Hex 10 GlcNAc 2 size constraint, the additional 0.07 mol of 4.89 -4.91 ppm resonance is likely to be present as an additional ␣1,6-linked Man on the new lower arm ␣1,6-linked branch and is assigned as residue 17 substituting residue 12. This structure is supported by the presence of a 2D DQF-COSY cross-peak of low intensity at 4.91(C1-H)/4.04(C2-H) ppm (not apparent in Fig. 5C) and assigns 7% of pool D as isomer 10d (Scheme II).
At 5.25 ppm 0.03 mol of resonance intensity was observed for G 1 , and the 2D DQF-COSY cross-peak at 5.25(C1-H)/ 3.54(C2-H) ppm confirms its presence. Thus, 3% of pool D is 10e, Glc 1 Man 9 GlcNAc 2 , which appears to retain an untrimmed glucose on exit from the ER to the Golgi (Table I).
At 5.11-5.15 ppm 3.90 mol of resonance intensity was integrated. Of this value, 1 mol is from core residue 7 and 1 mol from 2-O-substituted residue 12 documented above. This leaves 1.90 mol of resonance to be assigned to ␣1,3-linked residues 14, 15, and 16 (Scheme ID), and, indeed a strong cross-peak is seen in the 2D DQF COSY spectrum of Pool D (Fig. 5C)  ppm, which is more intense in pool D than seen in pool C (Fig.  6, compare panels B and C).
At 4.22 ppm 1.93 mol of C2-H proton resonance intensity was integrated. As already described for the other pools, this is due to the 3-O substitution of 11 with G 1 or 14, of 13 with 15, or of 14 and/or 15 with 16 (Scheme ID). As described above for pool C, cross-peaks for all of these residues were observed in pool D's 2D DQF-COSY spectra (Figs. 5C and 6C, respectively). Distribution of the integrated protons, consistent with relative crosspeak resonance intensities, and the Hex 10 GlcNAc 2 size constraint, allows assignment of the two major isomers as 10a (61%) and 10b (24%) (Scheme II). These structures account for 1.70 mol of 1.93 mol of resonance intensity at 4.22 ppm leaving 0.23 mol, and isomers 10d and 10e, assigned above, account for an additional 0.13 mol, leaving 0.10 mol to assign. Similarly, isomer 10a, 10b, 10d, and 10e account for 1.80 of the 1.90 mol of ␦ 5.13 resonance intensity calculated in the previous paragraph. Isomer 10c (5%) accounts for the remaining 0.10 mol of 3-O-substituted Man at 4.22 ppm, the 0.10 mol of ␣1,3-Man protons at 5.13-5.15 ppm, and the 0.05 mol of unsubstituted residue 11 at 5.03 ppm. These assignments are in very good agreement with the pool's HPAEC profile, both in the number of peaks found and their respective areas (Fig. 3).
In Vitro Characterization of alg9 N-Linked Oligosaccharides-Because Man 6 GlcNAc 2 was the smallest glycan released from ⌬alg9 secreted invertase, either this or its glucosylated form would be expected to be present on the ⌬alg9 OSL precursor. Fig. 7 shows comparative permethylation analysis of in vitro and in vivo-synthesized alg9 Hex 6 GlcNAc 2 . The former was released by mild acid hydrolysis from OSL synthesized in vitro under limiting glucosylation conditions (see "Experimental Procedures"), while the latter was the pool A Man 6 GlcNAc 2 .
In Vivo Glucosylation-NMR-derived structures of secreted invertase glycans in ⌬alg9 yeast indicate that ϳ7% of Hex 6 -10GlcNAc 2 isomers retained residue G 1 (Scheme II). To ascertain whether the persistence of G 1 is a product of nominal glucose addition of this residue, as in alg3 yeast (14), or a remnant of full glucosylation and processing, ⌬alg9 cells were pulse-chase labeled with [2-3 H]Man in the absence (Fig. 8, A-C) or presence (Fig. 8, D-F) of CST, and the labeled glycans were released from glycoprotein pellets by endo H and analyzed as described under "Experimental Procedures." At 0 min of chase, greater than 95% of endo H released glycans migrated as Glc 3 Man 6 GlcNAc (Fig. 8A, peak centered at 31 min) and ϳ5% migrated as Man 6 GlcNAc (Fig. 8A, peak centered at 8 min) in the absence of CST. The glycans migrated as the expected sizes, Hex 11 and Hex 8 , respectively, on Bio-Gel P-4 (data not shown). The Glc 3 Man 6 GlcNAc oligosaccharide lost the equivalent of two hexose units upon digestion with jack bean ␣-mannosidase as revealed by Bio-Gel P-4 chromatography (data not shown), all data being consistent with its assigned identity. In the presence of CST, 0 min of chase revealed that virtually all detectable glycans released from protein by endo H were Glc 3 Man 6 GlcNAc (Fig. 8D, peak centered at 31 min). The lack of Man 6 GlcNAc in the presence CST and its trace amount in the absence of CST clearly demonstrate that Glc 3 Man 6 GlcNAc 2 is the main if not the only glycan transferred to protein in the ⌬alg9 background.
At 1 min of chase in the absence of CST smaller glycans form, Aliquots from each reaction were terminated by addition to CHCl 3 / CH 3 OH (2:1) at 0, 1, and 10 min of chase. All pellets were solubilized and N-glycans released with endo H, which were characterized on the Dionex PA-100 column, previously calibrated with authentic oligosaccharide standards from this and a previous study (15). Panels A and D, 0 min of chase; panels B and E, 1 min of chase; panels C and F, 10 min of chase. Additional details are given under "Experimental Procedures." while in the presence of CST only a trace amount of these smaller forms are detectable. From these results it is evident that the smaller glycan species (Fig. 8B, peak centered at 20 min and shoulder at 16.5 min) are largely generated by glucosidases I and II in the absence of CST but cannot be formed in the presence of the glucosidase inhibitor. In the presence of CST at 1 min of chase larger glycans are formed (Fig. 8E, peak centered at 43 min), which is expected as the Glc 3 Man 6 GlcNAc present on accessible regions of the peptide backbone would be expected to be elongated by Golgi mannosyltransferases.
At 10 min of chase nearly all of the [2-3 H]Man-labeled Glc 3 Man 6 GlcNAc (Fig. 8C, peak centered at 30 min) has been processed to Man 6 GlcNAc (Fig. 8C, peak centered at 8 min) in the absence of CST, while in the presence of CST the majority of endo H-released oligosaccharides remain Glc 3 Man 6 GlcNAc (Fig. 8F, peak centered at 30 min). This result clearly demonstrates that Glc 3 Man 6 GlcNAc 2 is transferred to protein and largely processed to Man 6 GlcNAc 2 in the ⌬alg9 background.

DISCUSSION
The structure of the alg9 core oligosaccharide and its Golgimodified elongation products on secreted invertase have been defined through methylation analysis, HPAEC, MALDI-TOF MS, endo H sensitivity, pulse-chase[2-3 H]Man labeling, and high field 1 H NMR spectroscopy. The ⌬alg9 core Man 6 GlcNAc 2 N-glycan transferred to protein in vitro, via its dolichol-linked intermediate, and that present on secreted invertase, are one and the same structure. This is the first endo H-sensitive oligosaccharide formed during OSL synthesis and is the Alg3pdirected product (Scheme IC). Examining the synthesis and processing of ⌬alg9 N-linked glycans reveals an important role for the ALG3 phosphorodolichol pathway step in downstream processing events leading to mannan on secreted S. cerevisiae invertase.
Previous studies have shown that a significant proportion of the glycan chains on alg3 sec18 invertase retained one or more Glc residues transferred from OSL during glycosylation in the ER (14,15,25). Furthermore, the level of OSL glucosylation in the alg3 background was very low, with only ϳ7% of the chains transferred to protein containing the normal glucotriose unit. Thus, the Man 5 GlcNAc 2 -PP-Dol translocated into the yeast ER in alg3 is both poorly glucosylated, and the portion that is glucosylated becomes a poor substrate for glucosidases I and II once transferred to protein. Although little residual Glc survived on ⌬alg9 invertase glycans, initially making it difficult to assess the potential level of glucosylation and subsequent processing, [2-3 H]Man pulse-chase labeling demonstrated convincingly that the ⌬alg9 fully glucosylates the truncated Man 6 GlcNAc 2 -PP-Dol, and that Glc 3 Man 6 GlcNAc 2 is the major glycan transferred to protein from OSL in this mutant (Fig. 8). Furthermore, the complete absence of NMR signals for the terminal ␣1,2-linked Glc residue, G 3 , or the penultimate ␣1,3 Glc residue, G 2 , and the only small residual level of G 1 (Table  I) on the invertase oligosaccharides imply that both glucosidases I and II have nearly wild-type activity on the glucosylated alg9 oligosaccharides. It is worth noting that wild-type N-glycans retain a similar amount of G 1 as seen here in the ⌬alg9 strain (34). Thus, an important conclusion of the current work is that the middle arm ␣1,3-linked Man specified by the ALG3 locus provides structural information that potentiates the Alg6p, Alg8p, and Alg10p ER glucosyltransferases and Gls1p and Gls2p trimming glucosidases.
The NMR data show the complete absence of the upper arm Man␣1,2Man␣1,6-residues, indicating that, without the addition of the ␣1,2-linked Man 10 to residue 7 by Alg9p (Scheme IA), no further Man additions can occur on the OSL precursor by downstream Man-P-Dol dependent mannosyltransferases.
The reason such strict dependence on the Alg9p step evolved is not known with certainty, but may relate to the order of subsequent processing reactions. It is known that this disaccharide, consisting of the upper arm Man␣1,2Man␣1,6-, residues 9 and 6, respectively, is required for efficient ER ␣1,2-mannosidase (Mns1p) trimming of the middle arm ␣1,2-Man (38), and removal of this residue in conjunction with glucose trimming appears to be an integral part of protein quality control editing in the S. cerevisiae ER (39).
Mannan outer chain synthesis begins with the addition of ␣1,6-linked Man 12 to the lower arm core ␣1,3-linked Man 5 (Scheme ID) (7, 10) catalyzed by Och1p in the cis-Golgi, whose in vitro substrate specificity has been defined (36). Pyridylaminated Man 5 GlcNAc 2 -PA derived from the alg3⌬och1mnn1 triple mutant (Scheme IB), formed only 9% of the Och1p product compared with Man 8 -9 GlcNAc 2 -PA as substrate. Another glycan, a Man 7 GlcNAc 2 -PA, differing from the core alg9 structure (Scheme IC) by the presence of the upper arm ␣1,6-linked Man (residue 6, see Scheme IA), converted 60% of input to product compared with the wild-type Man 8,9 GlcNAc 2 -PA substrate in this study. Although the ⌬alg9 core Man 6 GlcNAc 2 was not tested as a substrate in this study (36), it appears from the current work that the addition of residue 7 does help potentiate the activity of Och1p. When comparing the alg3 and ⌬alg9 glycan pools that have one Golgi-type hexose addition (Hex 6 GlcNAc 2 for alg3; Hex 7 GlcNAc 2 for ⌬alg9), only 3% of the former's pool isomers have the Och1p addition (15), while 22% of the latter's have this addition (Scheme II). This clearly implies that in vivo the addition of Man residue 7 to the alg3 core Man 5 GlcNAc 2 (Scheme IC) significantly increases the proportion of glycans that are substrates for Och1p activity compared with those without this residue.
A novel series of core-filled structures was assigned containing terminal Man␣1,3Man␣1,3-disaccharides (see Scheme II). These structures have been identified previously on S. cerevisiae wild-type Man 14 GlcNAc (16) and to a greater extent on larger alg3 core glycans (15). Currently, little is known regarding the enzyme activity catalyzing this addition. Studies on the mnn1 mutant noted the absence of terminal Man␣1,3Man␣1,3disaccharide on O-linked glycans (40), suggesting that at least the penultimate ␣1,3-linked Man is added by the MNN1 encoded mannosyltransferase (40,41). Three genes in the protein sequence data base, yil014w, ygl257c, and ynr059w, closely related to MNN1, have been implicated in the terminal ␣1,3-Man addition to O-glycans. 2 Furthermore, a new family of enzymes that add ␣1,3-Man to ␣1,3-Man termini in O-linked glycans has been described recently in S. cerevisiae (42). It is possible that one of these genes, yil014, ygl257, and/or ynr059w, may be responsible for decorating existing N-linked ␣1,3-Man caps with a 3-O-substituting Man under certain metabolic conditions or in isolated genetic backgrounds.
The interrelationship of the alg9 structures deduced in this study is summarized in Scheme II. Kinetically, those glycans that are good substrates for mannan elongation will be depleted preferentially from the isomer pools. Outer chain elongation to mannan begins with the Och1p addition of ␣1,6linked Man 12 to the core ␣1,3-linked Man residue 5 (Scheme ID). The major difference between wild-type core-filled mannan and that of alg9 is the accumulation of isomer 7b (Scheme II), which escapes the Och1p ␣1,6-Man addition (residue 12), although to a much lesser extent than described earlier for alg3 glycans (15,25). While the ALG3-directed Man␣1,3-addition enhances to nearly wild-type levels OSL glucosylation and subsequent removal of those added Glc residues from glycoprotein, the Och1p activity clearly did not reach wild-type levels on the basis of residual isomers 6a, 7b, and 8d (Scheme II). This suggests that completing the upper arm Man␣1,2Man␣1,6branch (residues 9 and 6; Scheme IA) provides additional structural information recognized by the Och1p enzyme. An important next step in understanding the role of a given mannose addition on the subsequent OSL processing steps will be to analyze glycan structures in alg mutants that accumulate OSL intermediates further down the phosphorodolichol pathway. To this end, we have begun a structural analysis of the alg12 mutant (39), which accumulates the alg9 product.
While yeast are primitive eukaryotes, the conserved aspects of their N-glycosylation pathways and homology of many biosynthetic components with those of higher eukaryotes has yielded important understanding concerning errors in metabolism leading to OSL truncation in humans. This has proven to be the case in several forms of carbohydrate-deficient glycoprotein syndrome. Indeed, a lesion in the human homolog of the ALG6 defines a new form of carbohydrate-deficient glycoprotein syndrome type I (43)(44)(45). It is likely that, while diagnosis of CDGS may fall into a limited range of recognizable phenotypes, the underlying genomic lesions may be numerous, and defects in many of the ALG genes are clear candidates to be causative agents in this emerging disease.