Structure, Position, and Biosynthesis of the High Mannose and the Complex Oligosaccharide Side Chains of the Bean Storage Protein Phaseolin*

Phaseolin, the major storage protein of the common bean (Phaseolus vulgaris), is a glycopro- tein which is synthesized during seed development and accumulates in protein storage vacuoles or protein bodies. The protein has three different N-linked oligosaccharide side chains: Mane(GlcNAc)z, Man7(GlcNAc)2, and Xyl-Man,(GlcNAc), (where Xyl represents xylose). The structures of these glycans were determined by ‘H NMR spectroscopy. The Mane(GlcNAc)z glycan has the typical structure found in plant and animal glycoproteins. The structures of the two other glycans are shown below.

Phaseolin was separated by electrophoresis on denaturing gels into four size classes of polypeptides. The two abundant ones have two oligosaccharides each, whereas the less abundant ones have only one oligosaccharide each. Polypeptides with two glycans have M~~, ( G~C N A C )~ attached to AsnZs2 and Mane(GlcNAc)z attached to AsnS41. Polypeptides with only one glycan have Xyl-M~~, ( G~C N A C )~ attached to AsnZs2. Both these asparagine residues are in canonical glycosylation sites; the numbering starts with the N-terminal methionine of the signal peptide of phaseolin. The presence of the 1Man,(GlcNA~)~ and of Xyl-Man3(GlcNAc)z at the same asparagine residue (position 252) of different polypeptides seems to be controlled by the glycosylation status of When Asp341 is unoccupied, the glycan at AsnZb2 is complex. When AS^,^' is occupied, the glycan at AsnZs2 is only modified to the extent that 2 mannosyl residues are removed.
The processing of the glycans, after the removal of the glucose residues, involves enzymes in the Golgi apparatus as well as in the protein bodies. Formation of the Xyl-Man,(GlcNA~)~ glycan is a multistep process that involves the Golgi apparatus-mediated removal of 6 mannose residues and the addition of 2 N-acetylglucosamine residues and 1 xylose. The terminal N-acetylglucosamine residues are later removed in the protein bodies. The conversion of Mano(GlcNAc)z to Man7(GlcNAc)z is a late processing event which occurs in the protein bodies. Experiments in which [3H]glucosamine-labeled phaseolin obtained from the endoplasmic reticulum (i.e. precursor phaseolin) is incubated with jack bean a-mannosidase show that the high mannose glycan on AsnZ5', but not the one on is susceptible to enzyme degradation. Incubation of [,HI glucosamine-labeled phaseolin obtained from the Golgi apparatus with jack bean 8-N-acetylglucosaminidase results in the removal of the terminal N-acetylglucosamine residues from the complex chain. These observations are discussed in relation to the hypothesis that the control of glycan modification is determined largely by the accessibility of the glycan chains to the glycosidases and glycosyltransferases in the Golgi apparatus.
The asparagine-linked oligosaccharides found on plant glycoproteins, like those of other eukaryotes, fall into two general categories: high mannose and complex oligosaccharides. The high mannose oligosaccharides generally have the formula Man5_9(GlcNAc)2, with a branching pattern of mannosyl residues similar to that of the high mannose oligosaccharides of animal and yeast cells (1,2). The complex oligosaccharides have a Man3(GlcNAc):, core to which one or more of the following sugars may be attached xylose, fucose, N-acetylglucosamine, and/or galactose. The fucose is attached to the proximal N-acetylglucosamine of the chitobiose core, as in lima bean agglutinin (3) and laccase of Acer pseudoplatanus (4); whereas the xylose is attached to the p-linked mannose residue, as in stem bromelain (5), the protease inhibitor of Caesalpinia pulcherrima (6), and laccase (4). The structures of only a few N-linked glycans of plant glycoproteins have been determined. Because of our interest in the biosynthesis and transport of storage proteins and lectins in developing legume seeds (7), we have made a detailed study of the structure and biosynthesis of the oligosaccharide side chains of phaseolin, the major storage protein of the common bean. This protein accumulates during seed development in special protein storage vacuoles (protein bodies).
When phaseolin isolated from cotyledons of beans (Phaseolus vulgaris cv. Greensleeves) is subjected to SDS-PAGE,' it can be separated into four polypeptides, more accurately called size classes (8). These four polypeptide classes, referred to as A, B, C, and D, range in M , from 52,000 to 45,000. Each polypeptide band can be further resolved by two-dimensional electrophoresis into two or more polypeptides with different isoelectric points near pH 5 (9,10).
Previous reports from our and other laboratories detailed some important events during the biosynthesis and processing of phaseolin (11). When cotyledon mRNA is translated in uitro, two phaseolin polypeptide size classes of M , 48,000 ( a ) and 45,000 (p) are made (12). I n uiuo, phaseolin is synthesized on polysomes bound to the rough ER (13). The polypeptides are cotranslationally glycosylated with either one or two oligosaccharide chains, converting the a-polypeptides into polypeptides A and B, with A having two oligosaccharide chains and B having one (11). Analogously, the @-polypeptides become the glycosylated polypeptides C and D, with C having two oligosaccharide chains and D having one. Transport of these glycosylated polypeptides through the Golgi complex (14) and their ultimate deposition in the protein bodies (15) are accompanied by further processing steps.
transferases. There is a considerable amount of information concerning the processing of glycoproteins in animal cells (18,19), but very little is known about these events in plant cells. The processing of N-linked glycans in plant cells also begins with the loss of 3 glucose residues, followed by the removal of up to 7 mannose residues and the possible addition of fucose, xylose, N-acetylglucosamine, and galactose residues. The result of these processing events is a variety of glycans that range in size up to H~X~~_~~( G~C N A C )~ (where Hex represents hexose) when assayed on Bio-Gel P-4 columns (20). Developing cotyledons of leguminous seeds represent an excellent system to study the processing of N-linked glycans (7, 22).
Phaseolin is encoded by a small multigene family (23). Gene copy number analysis indicates that there are approximately seven phaseolin genes/haploid genome (24). The phaseolin gene family can be divided into two main gene types, a and @, which encode the two polypeptide size classes a and p detectable after translation of mRNA in uitro. The nine longest of the published nucleotide sequences of phaseolin cDNA clones show 98% homology between the a-and @-type genes (24). All amino acid sequences derived from these cDNA sequences have two canonical glycosylation sites. We refer to these sites as and counting amino acid residues from the initiating methionine of the signal sequence. Amz5' is located in a sequence Gly-Asn-Leu-Thr-Glu in a hydrophilic protein domain, and is in a Val-Asn-Phe-Thr-Gly sequence in a hydrophobic protein domain.
In this paper, we report the structures of the three different phaseolin oligosaccharides as determined by 'H NMR spectroscopy. We show their distribution among the various phaeolin polypeptides and the specific glycosylation sites (asparagine residues). In addition, we provide evidence for specific processing steps in the Golgi complex and the protein bodies.

MATERIALS AND METHODS AND RESULTS*
The structures of the three principal glycans present in Pronase digests of purified phaseolin were determined by 'H NMR and are shown in the Miniprint (Figs. 1-5). The principal polypeptides A and C each have Man,(GlcNAc), and Man7(GlcNAc)2 in equal proportions. Polypeptide D has mainly a small complex glycan: Xyl-Man3(GlcNAc),. We have no analytical data on polypeptide B because we were unable to purify sufficient amounts. However, indirect evidence indicates that its glycan resembles the one from polypeptide D. Both polypeptides B and D have a single glycan (ll), the polypeptides stain poorly with the Schiff stain for glycoproteins (42) (data not shown), and the glycan is endo-p-Nacetylglucosaminidase H-resistant (Fig. 2).
Identification of the Glycosylation Site for the Three Different Phaseolin Oligosaccharides-The amino acid sequences of the phaseolin polypeptides derived from the nucleotide sequences of nine different cDNAs show two possible glycosylation sites/polypeptide: and (16,46). To find out which oligosaccharide is attached to which asparagine residue, affinity-purified phaseolin from cotyledons labeled for 24 h with [3H]glucosamine was digested with trypsin. The resulting mixture of peptides and glycopeptides was fractionated by HPLC on a C1, column (Fig. 6A). Only one major radioactive peak (fraction 81) was observed (Fig. 6B). An aliquot of the peak fraction was digested with Pronase, and gel filtration of the resulting glycopeptides resolved three peaks corresponding in size to Mang(Gl~NAc)~Asn, Man7(GlcNAc)2(Gly)Asn, and Xyl-Man3(GlcNAc)2(Gly)Asn (data not shown). Next, the tryptic glycopeptides from HPLC fraction 81 were fractionated by ConA chromatography. When the radioactive material which did not bind to ConA was rerun on the reverse-phase C I S column, a major peak at fraction 81 appeared again (Fig. 6C). Amino acid sequencing of the first five amino acids of this tryptic glycopeptide (peak I) revealed the sequence Gln-Asp-Asn-Thr-Ile (QDNTI), which is identical with the amino acid sequence of the predicted tryptic glycopeptide that includes ( Fig. 7 ) . Sizing of the glycan on Bio-Gel P-4 after exhaustive Pronase diges-Fracllon Nurnbet FIG. 6. Reverse-phase HPLC separation of tryptic peptides from phaseolin. N-[3H]Acetylgluc~samine-1abeled phaseolin, prepared as described for Fig. 1, was digested with ~-1-tosylamido-2phenylethyl chloromethyl ketone-treated trypsin. A-C, tryptic peptides separated on a Vydac c 1 8 column. The column was eluted with 0.1 M sodium phosphate, pH 2.2, for 30 min, followed by a linear gradient of 0-50% acetonitrile for 120 min. D, tryptic peptides were separated on an Alltech CB column. This column was eluted with 0.1% (v/v) trifluoroacetic acid in water for 30 min, followed by a linear gradient of 0-35% acetonitrile in 0.1% trifluoroacetic acid for 150 min. A, elution profile of the tryptic peptides as measured by absorbance at 214 nm. B, elution profile as measured by radioactivity.
The radioactive peak (fraction 81) was applied to a ConA column with results similar to those shown in Fig. 3A. C, elution profile of the glycopeptide fraction which did not bind to ConA (peak I in Fig.   3A). D, elution profile on the c8 column of the glycopeptide fraction that bound tightly to ConA (peak I1 in Fig. 3A). Peaks I (C), 11, and I11 (D) were submitted to amino acid sequence analysis. tion showed it to be the complex phaseolin glycan Ph 1. When the fraction 81 glycopeptides that bound ConA were fractionated by HPLC on a C, column, we observed two major radioactive peaks (Fig. 6D, peaks I1 and 111). These glycopeptides were subjected to amino acid sequencing, and the oligosaccharides were analyzed by gel filtration after exhaustive digestion with Pronase. Analysis of peak I1 identified it as Mang(GlcNAcf2 attached to a peptide starting with the sequence Ala-Thr-Ser-Asn-Val (ATSNV). Similarly, peak I11 was identified as Man7(GlcNAc)2 attached to a peptide starting with the sequence Gln-Asp-Asn-Thr-Ile (QDNTI). Comparison of these amino acid sequences with the two predicted tryptic glycopeptides indicates that Mang(GlcNAc), is attached to and Man7(GlcNAc)2 is attached to Asn2" (Fig. 7 ) .
Biochemical Characterization of the Phaseolin Oligosaccharide Intermediates-To study the biosynthesis and processing of the phaseolin oligosaccharides, we labeled cotyledons for 3 and 24 h with [3H]glucosamine, isolated phaseolin, and separated the polypeptides by preparative SDS-PAGE. Polypeptides A, C, and D were obtained by electroelution and digested with Pronase, and the glycopeptides were fractionated on ConA (Fig. 8). The polypeptide size class B, which was not clearly separated from peptides A and C on the gels, remained as a contaminant in the size classes A and C. The glycopeptides prepared from polypeptides A and C after a 3-h labeling period all bind tightly to ConA-agarose. After 24-h labeling, 88% of this glycopeptide fraction still binds to ConA. In contrast, polypeptides of the phaseolin size class D isolated after 3-h labeling contained a mixture of glycans which either did not bind to the ConA column or were slightly retarded ( Fig. 8, lower left panel, peaks A and B, respectively) or bound tightly and were eluted with a-methylmannoside. After 24-h labeling, the proportion of the complex glycans was about 70% of the total and now eluted as one major peak from the ConA column (Fig. 8, lower right panel).
The glycopeptides resolved by ConA chromatography were analyzed by sizing on a long column (1 x 100 cm) of Bio-gel P-4 before and after treatment with various glycosidases. The results of these experiments are presented in Figs. 9-11 and can be summarized as follows. Initially (3-h labeling period), both polypeptides A and C have two Mang(GlcNAc)2 glycans. One glycan remains as Mang(GlcNAc)2, and the other is slowly processed to M~II~(G~CNAC)~. Polypeptide D has mainly a complex glycan which, in short-time labeling experiments, has terminal N-acetylglucosamine residues. After 24-h labeling, these terminal N-acetylglucosamine residues are no longer present. processing events take place in the cell, cotyledons were labeled for 1 h with [3H]glucosamine, and the organelles were fractionated on sucrose gradients. The homogenization procedure disrupts the large protein bodies, and their proteins become admixed with cytosolic proteins (28). Rough ER and the Golgi apparatus were collected according to their sucrose densities of 1.13 and 1.18 g -~m -~, respectively. Radiolabeled phaseolin, isolated from the membrane preparations with anti-phaseolin-Sepharose in the presence of Tween 20 and from the soluble fraction by the affinity method of Stockman et al. (27), was digested with Pronase; and the glycopeptides obtained were submitted to ConA affinity chromatography. The glycopeptides of total phaseolin isolated from the ER carry high mannose-type oligosaccharides exclusively as they all bind to ConA-agarose (Fig. 12, upper panel). On the other hand, about 25% of the glycopeptides of phaseolin isolated from the Golgi apparatus and the soluble fractions do not bind to ConA (Fig. 12, middle and lower panels). This is consistent with the known role of the Golgi apparatus in the conversion of high mannose to complex side chains on glycoproteins (see Ref. 19).

Localization of the Phaseolin Oligosaccharide Processing
The various glycopeptides obtained by ConA affinity chromatography were further analyzed by gel filtration (Fig. 13). In addition to the glycopeptides described above, those obtained from soluble phaseolin labeled for 3 and 24 h with N-[3H]acetylglucosamine were analyzed. Gel filtration of the ConA-binding (ConA+) glycopeptides from the ER-localized phaseolin shows that they co-migrated with Mang-(GlcNAc)zAsn. They are not distinguishable from the ConA+  (14). Fractions containing the rough ER and Golgi apparatus were diluted with an equal volume of phosphatebuffered saline containing 1% Tween, and the phaseolin was isolated using anti-phaseolin-Sepharose. Phaseolin from the soluble fraction was isolated by immunoaffinity chromatography. Each fraction of phaseolin was digested by Pronase, purified by passage through the short Bio-Gel P-4 column, and then fractionated on ConA-agarose columns as described for Fig. 1. The arrows indicate the starting point of elution with 200 mM a-methylmannoside. glycopeptides found in the Golgi apparatus or the soluble fraction after 1-h labeling. However, the ConA+ glycopeptides of soluble phaseolin labeled for 3 h can be resolved into a major peak corresponding to Mang(GlcNAc)ZAsn and a minor peak that co-migrated with Man7(GlcNAc)z(Gly)Asn. After a 24-h labeling period, the Man9(GlcNAc)pAsn and Man7(GlcNAc)z(Gly)Asn peaks are present in a ratio of 1:1, as found for the mature protein. Thus, the processing from Mang(GlcNAc)z to Man7(GlcNAc), is only apparent after a long labeling period and is therefore presumed to take place after the transport of phaseolin to the protein bodies.
Gel filtration of the Golgi apparatus-derived phaseolin glycopeptides that did not bind ConA (ConA-) yielded two peaks that eluted in positions corresponding to the complex glycopeptides obtained from 3-h-labeled polypeptide D (compare the Golgi panel in Fig. 13 to the CONTROL panel in Fig. 10). Treatment of these Golgi apparatus-derived glycopeptides with /3-N-acetylglucosaminidase generated a product which  Fig. 12 were analyzed by gel filtration as described for Fig. 9. Glycopeptides of phaseolin from the soluble fraction labeled for 3 and 24 h with [3H]glucosamine were included. Glycopeptides which were not retained by ConA-agarose columns are labeled as C o d -, and those which required a-methylmannoside elution are designed C o d + . Arrows M 9 and MI indicate the elution positions of Man9(GlcNAc)PAsn and Manl(GlcNAc)nAsn, respectively. co-migrated with Xyl-Man3(GlcNAc)z(Gly)Asn, and simultaneous treatment with /3-N-acetylglucosaminidase and a-mannosidase resulted in a product that co-migrated with Xyl-Manl(GlcNAc)z(Gly)Asn (data not shown). The complex (ConA-) glycopeptides obtained from the soluble phaseolin fraction after 1-h labeling showed the same size distribution as was found for the same in the Golgi apparatus. After a 3-h labeling period, the glycopeptide corresponding to the peak with less terminal N-acetylglucosamine residues increased in level. After 24-h labeling, all the ConA-glycopeptides comigrated with Xyl-Man3(GlcNAc)2(Gly)Asn. Thus, the processing of a high mannose glycan to a complex glycan occurs in the Golgi complex where the complex glycan obtains terminal N-acetylglucosamine residues. This is followed by the slow removal of terminal N-acetylglucosamine residues after the protein arrives in the protein bodies. Accessibility of Phaseolin High Mannose Oligosaccharides to a-Mannosidase Digestion in Vitro-Phaseolin isolated from the rough ER carries only high mannose-type oligosaccharides with 9 mannose residues (Fig. 13, upper left panel). Digestion of undenatured ER-derived phaseolin with jack bean a-mannosidase converts approximately half of the Mang(GlcNAc)* groups into an oligosaccharide shortened by 4 mannose residues (data not shown). To determine if this accessibility to a-mannosidase in vitro related to the attachment site of the oligosaccharide (AsnZ5' or we analyzed tryptic glycopeptides of the a-mannosidase-treated phaseolin. The mixture of tryptic peptides was separated by C, reversephase HPLC (see Fig. 6D). The glycopeptides were identified by measuring the radioactivity in each fraction, and the N termini of the glycopeptides were determined by amino acid sequence analysis. The oligosaccharides attached to these glycopeptides were analyzed by sizing them on a calibrated column of Bio-Gel P-4 after exhaustive digestion with Pronase (Fig. 14, middle panel). Two controls are included in Fig. 14. Tryptic peptides, obtained from phaseolin which was not incubated with a-mannosidase, were separated on the Cs column. One-half of each glycopeptide fraction was digested with Pronase, and the other half with a-mannosidase followed by Pronase. The reaction products were analyzed by sizing them on a calibrated column of Bio-Gel P-4 (Fig. 14, upper and lowerpanels, respectively). The results can be summarized as follows. The N-terminal region of the glycopeptide in HPLC fraction 118 (Fig. 6, HPLC peak 11) is H2N-Ala-Thr-Ser-Asn-Val (ATSNV) and carries Mang(GlcNAc)2. Comparison with the amino acid sequences of the two predicted glycopeptides (Fig. 7) shows that this high mannose oligosaccharide is attached to in the hydrophobic protein domain. When present in the undenatured ER-derived glycoprotein, this oligosaccharide is not accessible to jack bean FIG. 14. Gel Wltration of glycopeptides from rough ER-derived phaseolin. The procedures for labeling and isolating phaseolin from rough ER were as described for Fig. 12 with the exception that the cotyledons were labeled for 2 h. An aliquot of phaseolin (25,000 cpm) was treated for 24 h with a-mannosidase. Both the treated sample and an untreated sample were then digested with trypsin, and the tryptic peptides were separated on a C, HPLC column as described for Fig. 6. The tryptic glycopeptides so obtained were then digested with Pronase, and the reaction products were analyzed by gel filtration on the long Bio-Gel P-4 column (upper and middle panels). In addition, samples of phaseolin not treated with a-mannosidase were subjected to trypsin digestion, HPLC fractionation, and finally amannosidase treatment before Pronase digestion and gel filtration (lower panel). Arrows M9 and M , indicate the elution positions of Mang(GlcNAc)zAsn and Manl(GlcNAc),Asn, respectively. a-mannosidase. It is readily degraded by a-mannosidase in the corresponding tryptic glycopeptide; however, the main digestion product co-migrates on the Bio-Gel P-4 column with Manl(GlcNAc),Asn (fraction 48), although a significant proportion is still somehow protected by the attached peptide and only shortened by 2-4 mannose residues.
The N-terminal region of the glycopeptide in HPLC fraction 128 (Fig. 6, HPLC peak 111) is H,N-Gln-Asp-Asn-Thr-Ile (QDNTI) and also carries Mang(GlcNAc),. Comparison with the two predicted glycopeptides (Fig. 7) indicates that this oligosaccharide is attached to AsnZ5' in the hydrophilic protein domain. When ER-derived phaseolin (undenatured) is treated with jack bean a-mannosidase, this high mannose oligosaccharide is shortened by 2-4 mannose residues. Most of the corresponding tryptic glycopeptide is totally accessible to a-mannosidase, which becomes trimmed to Manl-(GlcNAc),(Gly)Asn (fraction 51). However, as seen for the HPLC peak I1 tryptic glycopeptide, some of the material is partially resistant to a-mannosidase and only shortened by 2-4 mannose residues. It is not entirely clear why glycopeptides obtained by trypsin digestion are only partially degradable by a-mannosidase, as the corresponding glycopeptides from Pronase digests are completely degraded by a-mannosidase to Man,(GlcNAc),(Gly)Asn (data not shown). We favor the explanation that the peptide portion of the glycopeptide causes steric hindrance of the enzyme. The experiments discussed here show that the oligosaccharides of ER-derived phaseolin, which are accessible to jack bean a-mannosidase in uitro, are the same ones which are normally modified Mang(GlcNAc)2 to Man,(GlcNAc);! in the phaseolin polypeptides A and C in the protein bodies and Mang(GlcNAc), to Xyl-Mans(GlcNAc)n in polypeptide D in the Golgi apparatus.
Accessibility of the Terminal N-Acetylglucosamine Residues of the Complex Phaseolin Oligosaccharide to 0-N-Acetylglucosaminidme Digestion in Vitro-Phaseolin isolated from the Golgi fraction of cotyledons labeled for 1 h with [3H]glucosamine was digested with Pronase, and the resulting glycopeptides were separated by ConA chromatography. When the glycopeptides that did not bind ConA were treated with P-Nacetylglucosaminidase, more than 90% of the label appeared as free N-acetylglucosamine (data not shown). Thus, more than 90% of the N-[3H]acetylglucosamine incorporated into the complex side chains of l-h-labeled phaseolin resides in terminal (@-N-acetylglucosaminidase-accessible) residues.
The sensitivity of the terminal N-acetylglucosamine residues of the complex oligosaccharides in native phaseolin to /3-N-acetylglucosaminidase in uitro was studied by treating N-[3H]acetylgluc~~amine-labeled phaseolin with the enzyme. The phaseolin was obtained from cotyledons which had been labeled for 1 h with [3H]glucosamine. After this short labeling time, most of the N-[3H]acetylgluc~~amine is in terminal Nacetylglucosamine residues rather than in the chitobiose core. After inactivation of P-N-acetylglucosaminidase, phaseolin was digested by Pronase, and the glycopeptides obtained were separated by ConA affinity chromatography. Analysis of the nonretarded glycopeptide fraction of the enzyme-treated and control phaseolin showed that the terminal N-acetylglucosamine residues can be removed by treatment of native phaseolin with (3-N-acetylglucosaminidase. The amount of radioactivity in N-[3H]acetylgluc~~amine in in uitro treated phaseolin was less than 10% compared to the controls. Gel filtration analysis of the same glycopeptides from in uitro treated phaseolin showed that the glycopeptides co-migrated with the mature glycopeptides obtained from mature phaseolin in the protein bodies (data not shown).

DISCUSSION
The results reported in this paper confirm and extend the limited amount of information available on the structure of N-linked glycans of plant glycoproteins in general and seed storage glycoproteins in particular. Plant glycoproteins have been shown to contain both high mannose and complex Nlinked glycans, and it is known that the latter derive from the former (1-7, 14, 23, 47, 48). We isolated three different abundant glycans from affinity-purified phaseolin and determined their structures. Two are typical high mannose oligosaccharides, and one is complex. The two high mannose oligosaccharides have 9 and 7 mannose residues and share the unique branching pattern of mannosyl residues reported for high mannose glycans of yeast and animal cells (1,2). In addition to these, phaseolin contains small amounts of Mans and Man, glycans (48). By collecting the peak fractions of the glycopeptides separated on Bio-Gel P-4, we obtained samples for 'H NMR analysis which were uncontaminated by these minor glycan species. The complex glycan of phaseolin with 3 mannose residues and 1 xylose residue linked Dl-2 to the P-linked mannose seems to be a common component of complex N-linked plant oligosaccharides (4)(5)(6). A Dl-2-linked xylose residue has been found in only one animal glycan (44). In addition to this core, most complex glycans of plant glycoproteins have an a1-3-linked fucose residue on the proximal N-acetylglucosamine, and some have N-acetylglucosamine, galactose, and fucose residues emanating from the core.
By sequencing tryptic phaseolin glycopeptides which were separated by reverse-phase HPLC and identified by their specific oligosaccharide, we have shown that Mang(GlcNAc), is always attached to whereas Man,(GlcNAc), and Xyl-Man3(GlcNAc)p are always attached to Thus, the Mang(GlcNAc), glycan is located in a hydrophobic protein domain, and the glycans which undergo processing are both in a hydrophilic part of the protein (24).
Polypeptides A and C have only high mannose glycans with Man7(GlcNAc), attached to AsnZ5' and Mang(GlcNAc), attached to Polypeptide D has only one glycan: mostly Xyl-Man3(Glc-NAc), and a smaller amount of Man7-(GlcNAc),. It is likely that polypeptide D is a mixture of two slightly different polypeptides: one with Man7(GlcNAc), and one with Xyl-Man3(GlcNAc),. Xyl-Man3(GlcNAc)z is always attached to A S P , and Man,(GlcNAc), also occupies this site on the related polypeptide, as it does on polypeptides A and C. The presence of a glycine residue at the N-terminal side of asparagine clearly identifies this glycosylation site when glycopeptides are analyzed. Whether polypeptide D lacks a second glycosylation site or has an unoccupied glycosylation site is not clear. All nine cDNAs which have been sequenced give rise to polypeptides with two glycosylation sites. This means that it is more likely that the second site is unoccupied rather than nonexistent. We interpret the presence of the two types of glycans in polypeptide D as evidence for incomplete processing when the protein passes through the Golgi apparatus. We speculate that a portion of the D polypeptides reach the protein bodies with Mang(GlcNAc), glycans and that these are slowly processed to MandGl~NAc)~ (see below). Whereas we have no data on the glycan of polypeptide B, we postulate that it is the same as Polypeptide D. The method used to purify the polypeptides (SDS-PAGE followed by electroelution) did not allow us to isolate polypeptide B away from polypeptides A and C because the three polypeptides are not clearly separated on the overloaded gels. Biochemical Characterization and Localization of the Phaseolin Oligosaccharide Intermediates-By labeling cotyledons for different intervals with [3H]glucosamine, we were able to identify early and late glycan processing events, and subcellular fractionation experiments permitted us to place the early processing events in the Golgi apparatus and late events in the protein bodies. The various processing steps are summarized in Fig. 15. In the ER, we find only the Man9(GlcNAc)2 form of glycans, indicating that the removal of terminal glucose residues occurs very quickly. In the Golgi apparatus, we find unprocessed high mannose glycans and fully processed complex glycans with terminal N-acetylglucosamine residues. This indicates that the intermediate processing steps occur rapidly in the Golgi apparatus. Transport of the glycoproteins to the protein bodies occurs with terminal N-acetylglucosamine residues attached to the complex glycans. Our findings confirm that in plants the transformation of high mannose oligosaccharides into complex glycans occurs in the Golgi complex (14). Plant glycan processing enzymes such as mannosidases I (49) and 113 have recently been characterized but not yet localized, whereas some of the glycosyltransferases such as fucosyltransferase (50) and N-acetylglucosamine transferase (51) have been identified and localized in the Golgi apparatus. Our finding that processing (removing of Nacetylglucosamine and mannose) occurs in the protein bodies is consistent with the observation that protein bodies contain a-mannosidase and @-N-acetylglucosaminidase (52).
Accessibility of Glycans to Modifying Enzymes-Experiments with glycoproteins from plants (53), yeast (54), and virus envelopes (55) show that the control of glycan modification is, in large part, determined by the accessibility of glycans to the processing enzymes in the Golgi complex. This hypothesis derives from studies in which native glycoproteins are incubated in the absence or presence of glycosidases (endo-@-N-acetylglucosaminidase H, a-mannosidase) and then run on SDS gels. Any change in M, caused by glycosidase treatment can yield an estimate of the extent of glycan accessibility to modifying enzymes. The results obtained here are in agreement with the hypothesis stated above. We observed that the Man9(GlcNAc)2 on Asn252 of ER-derived phaseolin is partially susceptible to a-mannosidase (2-4 mannose residues are re-A. Elbein, personal communication. moved), whereas the Man9(GlcNAc), on ASP1 is not altered by exposure to the enzyme. Thus, only the glycan which can be modified in the Golgi apparatus is accessible to the glycosidase in vitro. As discussed above, the glycan on Amz5' only becomes complex when the second glycosylation site is unoccupied. In addition, we found that the terminal N-acetylglucosamine residue(s) on the complex glycan of Golgi apparatus-derived phaseolin can be removed by incubation of the native protein with @-N-acetylglucosaminidase. It is of particular interest in this regard that the glycan which is accessible to glycosidases (and presumably to glycosyltransferases as well) is in a hydrophilic region of the polypeptide (Asn'"), whereas the glycan which is not accessible is in a hydrophobic region of the polypeptide (24). It may well be that high mannose chains in hydrophobic pockets are sequestered and thus remain unmodified. However, the converse is not necessarily true. When analyzing the glycans of phytohemagglutinin, we found an unmodified chain, which is not accessible to a-mannosidase in vitro, attached to an asparagine residue in a hydrophilic region (56).
Of particular interest with respect to the processing of glycans is our finding that the presence of a glycan on A s P may determine whether the glycan at is processed or not. We interpret our results as showing that the absence of a glycan from AS^^^^ results in the more extensive processing of the glycan on If glycans are not accessible, they remain unmodified. The absence of a glycan on may result in greater accessibility to the Golgi enzymes of the glycan attached to AS^^^'.
Role of Terminal N-Acetylglucosamine Residues-Our results show that the complex side chain of phaseolin polypeptide D acquires terminal N-acetylglucosamine residues in the Golgi apparatus, which are later removed in the protein bodies. Earlier (57), we made a similar observation for the complex side chain of phytohemagglutinin, which, like phaseolin, moves via the ER-Golgi apparatus-protein body transport pathway in beans. Since protein bodies are similar to lysosomes (both contain a full complement of acid hydrolases (see Ref. 7), it is tempting to speculate that these terminal N-acetylglucosamine residues may play a role in targeting in the same manner that terminal phosphate residues target lysosomal hydrolases to lysosomes (58). Such a role for the terminal N-acetylglucosamine residues seems unlikely in view of the finding that tunicamycin does not inhibit the transport of unglycosylated phytohemagglutinin to the protein bodies (59). Furthermore, many protein body proteins, such as soybean glycinin and pea legumin, are not glycoproteins.
Our laboratory has recently investigated (60) the structural requirements of oligosaccharides that can accept xylose and/ or fucose residues during the formation of complex glycans in the plant Golgi apparatus. Only glycans containing at least 1 terminal N-acetylglucosamine residue can serve as acceptors of xylose and fucose. As both phytohemagglutinin and phaseolin contain glycans with transient N-acetylglucosamine residues and these same glycans contain xylose and/or fucose, then the terminal N-acetylglucosamine residues may simply constitute recognition markers for later Golgi apparatus-mediated processing events.