Structures of the 0-glycosidically Linked Oligosaccharides of Human IgD*

In the previous communication (Mellis, S. J., and Baenziger, J. U. (1983) J. Biol. Chem. 258, 11546-11556), the structures of the oligosaccharides present at the 3 asparagine glycosylation sites of a human IgD myeloma protein were defined. In this communication, we present the structures of the O-glycosidically linked oligosaccharides located in the hinge region of IgD:WAH. Three or four threonine residues and one serine residue in the region bear O-glycosidically linked oligosaccharides. Approximately 50% of these molecules have the structure Gal beta 1 leads to 3 GalNAc which is identical with the structure of the predominant oligosaccharide in the hinge region of human IgA1 myeloma proteins (Baenziger, J. U., and Kornfeld, S. (1974) J. Biol. Chem. 249, 7270-7281). The remainder of the oligosaccharides contain 1 or 2 residues of N-acetylneuraminic acid and have the structures NeuAc alpha 2 leads to 3Gal beta 1 leads to 3GalNAc (30%), Gal beta 1 leads to (NeuAc alpha 2 leads to 6)GalNAc (12%), and NeuAc alpha 2 leads to 3Gal beta 1 leads to 3(NeuAc alpha 2 leads to 6)GalNAc (8%). The sialylated molecules have not been encountered previously on other human immunoglobulin heavy chains. These structures, however, have been described on a number of secreted and membrane glycoproteins. Examination of oligosaccharides isolated from different subregions of the IgD hinge indicated that a specific distribution of the sialylated structures among the glycosylated amino acids of the hinge region is not likely.

Only a small number of serum proteins are known to contain both 0and N-glycosidically linked oligosaccharides.
Among these are human IgA, (1, 2), human IgD (3-5), and rabbit IgG (6). In the preceding paper (7), we determined the structures of the oligosaccharides located at each of the asparagine glycosylation sites of a human IgD myeloma protein 1gD:WAH. We now present data establishing the structures of the oligosaccharides 0-glycosidically linked to serine and threonine residues present in the hinge region of 1gD:WAH.
IgD, like IgG, and IgA, contains a segment of heavy chain I Recipient of Research Career Development Award KO4 CA00671 from the National Cancer Institute, Department of Health, Education, and Welfare.
between the Fd and Fc regions termed the hinge region. In the 6 and c t l heavy chains, this region contains the only sites of 0-glycosidically linked carbohydrate among all the immunoglobulin classes of humans (8). The hinge region is believed to confer flexibility to immunoglobulin molecules and may transduce conformational signals from the antibody-combining site to biological effector domains of Fc (9). Whereas the function of the secreted form of IgD is unknown, the membrane form of this immunoglobulin is believed to mediate important immunoregulatory functions in B cells (reviewed in Refers 10-12). It has been proposed that the hinge region of IgD is of special importance in these biological processes (13). Dr. Frank Putnam, who with his co-workers has determined the complete amino acid sequence of 1gD:WAH (14), has provided us with glycopeptides encompassing specific portions of the IgD hinge. In this paper, we present the detailed structural characterization of the 0-glycosidically linked oligosaccharides of IgD as well as information regarding the distribution of different oligosaccharide structures within the hinge region.

Materials
Human Myeloma IgD: WAH-Hinge region glycopeptides were kindly provided by Dr. Frank Putnam, Department of Biology, Indiana University, Bloomington. The complete amino acid sequence of 1gD:WAH has been established (14).
Reagents-Triethylamine (Eastman) was redistilled once by flash evaporation. Glacial acetic acid was purchased from Fisher and boric acid from Sigma. The DA-X8-11 resin was purchased from Durrum Chemical Corp. The remainder of materials used have been described in the previous communication (7).
Enzymes-The purification of enzymes from jack bean meal, Clostridiumperfringins, and Diplococcuspneumoniae, have been described previously (7). Neuraminidase from Newcastle Disease Virus was the generous gift of Dr. James Paulson, Department of Biological Chemistry, UCLA School of Medicine.

Methods
The methods employed have been described in the previous paper (7) with the following additions.
Alkaline Degradation-Fifty-nmol aliquots of hinge region glycopeptides were subjected to alkaline degradation at 37 "C in 50 pl of 0.1 N NaOH containing 0.2 or 0.4 M NaB[3H]4 (347.5 mCi/mmol) for 36 or 56 h, respectively. Results were similar under both conditions. The reaction mixture was acidified to pH 4.6 with 1 M acetic acid and the products were applied to a column (0.7 X 4 cm) of AG-50W-Xl2 (H' form, 200-400 mesh). Oligosaccharides were eluted with 7 ml of Hz0 after which peptide was eluted in 7 ml of 0.5 M NH,OH. Oligosaccharides were then re-N-acetylated by successive additions of 100 pl of 5% of acetic anhydride in saturated NaHC03 three times. Na was removed by passage over AG-50W-Xl2.
Ion Suppression Amine Adsorption HPLCl-This method fraction-ates anionic complex carbohydrates on the basis of net carbohydrate content as well as linkage positions of anionic moieties (15, 16). Chromatography was performed on a Varian Model 5000 liquid chromatograph with a MicroPak AX-5 column (4 mm X 30 cm) (Varian). Flow was 1 ml/min. The mobile phase consisted of a linear gradient produced by mixing two solvents from reservoirs A and B. Reservoir A contained 3% (v:v) glacial acetic acid in an 8020 mixture of acetonitrile:H20 and was titrated to pH 5.5 with triethylamine (Buffer A). Reservoir B contained 3% (v:v) glacial acetic acid in HzO, titrated to pH 5.5 with triethylamine (Buffer B). The initial ratio of Buffer A:Buffer B was 955 and the proportion of Buffer B was increased at a rate of 1% per min. For analytical chromatograms, 3Holigosaccharides were injected in up to 300 pl of starting buffer and the eluate collected directly into scintillation vials at 0.3-min intervals. Water and 3a70 complete mixture (Research Products Inc.) were added for scintillation counting. For preparative fractionation, the eluate was collected in glass tubes and aliquots were removed for scintillation counting.
Anion Exchange HPLC on MicroPak AX-10-Anionic oligosaccharides were fractionated analytically as well as preparatively on the basis of net charge as described in Ref. 17.
Borate Anion Exchange HPLC-Monosaccharides were analyzed as their NaB[3H]4 reduced derivatives on DA-X8-11 as described by Barr and Nordin (18) except that 0.3-ml fractions were collected.
Analytical Methods-Amino acid analysis was performed following 24-h hydrolysis in 6 N HC1 using a Waters HPLC amino acid analysis system. Neutral and amino sugars were determined by gas chromatography as described previously (19). Sialic acid was measured by a modification of the thiobarbituric acid assay of Warren (20).

RESULTS
The complete amino acid sequence of 1gD:WAH has been reported and the locations of the four or five oligosaccharides 0-glycosidically linked via GalNAc to hinge region Ser/Thr residues determined (14). We determined the carbohydrate composition for a glycopeptide encompassing all of the glycosylation sites of the hinge region ( Table I). The results indicated that Gal and GalNAc were present in nearly equimolar amounts and that sialic acid was also present at approximately 50% of this amount. As the structure of the predominant 0-glycosidically linked oligosaccharide present in the hinge region of human IgA has been determined to be Galpl-t3GalNAc (I), the compositional data we obtained suggested that the same structure may be present in the hinge region of 1gD:WAH with the additional presence of sialylated derivatives.
A glycopeptide encompassing all glycosylation sites of the hinge region was subjected to alkaline degradation in 0.1 N NaOH containing 0.4 M NaB[3H]4. During the &elimination of the oligosaccharides from the glycopeptide, tritium was introduced into the reducing termini by reduction with the NaB[3H]4. This reaction was monitored by amino acid analysis before and after alkaline degradation and it was determined that all GalNAc residues were released from the glycopeptide concomitant with the destruction of 4 threonine residues ( Table 11). The expected conversion of 1 residue of serine to alanine was not observed; however, this is believed to be a technical artifact due to reasons described in a later section.
The 3H-oligosaccharides were initially analyzed by anion exchange HPLC on MicroPak AX-10 (17). The chromatographic profile obtained is displayed in Fig. 1. Three species were observed and two of these co-migrated with neutral and    Val normalized to 1, result of a single determination. Not fully resolved from GalNAc peak. 4. An aliquot of the NaB[3H]4-reduced oligosaccharides (60,000 cpm) was analyzed by anion exchange HPLC on MicroPak AX-10 as described previously (17). N and S-(1,2,3,4) indicate the elution positions of complex type oligosaccharides containing 0, 1, 2, 3, and 4 sialic acid residues, respectively.
HPLC. The converse experiment was also performed in which oligosaccharides were preparatively fractionated by ISSA-HPLC and then individually analyzed by anion exchange HPLC. It was thereby determined that ISSA-HPLC fractions H I and X corresponded to neutral species, H2 and H3 corresponded to species which co-eluted with monosialylated complex type oligosaccharides, and H4 was the species that eluted in a position intermediate between the elution positions of diand trisialylated complex type oligosaccharides on anion exchange HPLC.
In order to establish that the anionic character of species H2, H3, and H4 were due to the presence of sialic acid, the total oligosaccharide mixture was digested with neuraminidase from C. perfringins and the products were analyzed by ISSA-HPLC (Fig. 2B). This digestion caused the disappearance of species HZ, H3, and H4 from the chromatographic profile with a concomitant increase in the ratio of species H1 to X (Fig. 1B). Similar digestion of a mixture of ISSA-HPLCpurified H2, H3, and H4 resulted in the formation of only one species which co-migrated with HI. The results of these digestions indicated that H2, H3, and H4 were sialylated derivatives of the neutral oligosaccharide H1. The results also indicated that HZ and H3 were most probably linkage isomers of monosialylated H1 because these molecules co-migrated on ISSA-HPLC after removal of the sialic acid residues. Digestion of the total oligosaccharide mixture with the a-2,3specific neuraminidase from Newcastle Disease Virus (29) resulted in the disappearance of oligosaccharide species H2 and H4 from the profile (Fig. ZC). The effect of Newcastle Disease Virus neuraminidase digestion on ISSA-HPLC purified oligosaccharide H4 is shown in D. After digestion with this enzyme, essentially all of the H4 oligosaccharides were converted to a species which eluted in the position of oligosaccharide H3.
The chromatographic and enzymatic data presented above indicated the following. 1) H1 is the basic structure from which H2, H3, and H4 are derived by the addition of sialic acid residues. 2) H2 is a structural derivative of H1 with 1 sialic acid in a-2,3 linkage. 3) H3 is structural derivative of H1 containing 1 sialic acid linked in a manner other than 01-2,3. and 4) H4 is a structural derivative of H3 containing an additional residue(s) of sialic acid in a-2-3 linkage. In order to further elucidate these structures, the following studies were conducted.  (Table 11) and the finding that H1 was neutral by anion exchange HPLC, it was proposed that H1 was the NaB[3H]4-reduced derivative of Galpl+3GalNAc, the predominent structure located in the hinge region of IgA, (1).
Digestion of ISSA-HPLC-purified oligosaccharide H1 with pgalactosidase (jack bean meal) resulted in the formation of a product which co-migrated with authentic GalNAcol on both ISAA-HPLC and borate anion exchange HPLC (Fig. 3). This latter technique was employed as GalNAcol and GlcNAcol are not effectively resolved by ISAA-HPLC, whereas borate anion exchange HPLC readily separates these molecules (18). As noted in previous studies of Galpl-t3GalNAc (1, 30), digestion with this enzyme proceeds slowly and 4 days (with four additions of enzyme) were required to achieve the degree of completion observed.
The linkage position of Gal to GalNAcol in oligosaccharide H1 was determined by methylation analysis. Mass spectrometry of the intact permethylated oligosaccharide indicated that the structure of this disaccharide was Gal@l~3GalNAcol and not Gal@l+4GalNAcol or Gal@l-+6GalNAcol based on the following observations. In selected ion recording, signals were detected at m/e 88, 130, 187, 219, 276, 304, 378, 422, and 466 which were in ratios consistent with the fragmentation of a permethylated hexopyranosyl Dl-3 2-acetamido-2-deoxyhexitol as described previously (26). The absence of a strong signal at m/e 174 and the presence of significant responses at m/e 304 and 378 ruled against the 1 4 linked compound. The absence of a strong signal at m/e 174 and the presence of significant responses at m/e 378,422, and 466 ruled against the presence of the 1 4 linked compound (26). Permethylated alditol acetates generated from oligosaccharide H1 by acetolysis-acid hydrolysis (23) and methanolysis (25) were also analyzed and resulted in the detection of terminal galactose and 3-monosubstituted N-acetylgalactosaminitol, further confirming the structure assignment of this species as GalPl-, 3GalNAcol.
Analysis of Oligosaccharide H2"The digestion by Newcastle Disease Virus neuraminidase of oligosaccharide H2 to a species which co-migrated with H1 ( Fig. 2C) indicated that H2 was derived from H1 (Gal@1+3GalNAcol) by the addition of an a-2-3 linked neuraminic acid. Because the 3' hydroxyl of the GalNAcol was already occupied by galactose in a glycosidic linkage, the neuraminic acid must have been linked to the 3' hydroxyl of the galactose residue. Methylation analysis identified 3-monosubstituted galactose, 3-monosubstituted GalNAcol, and terminal N-acetylneuraminic acid. These results determined that the structure of oligosaccharide H2 was NeuAcc~2+3Gal@l+3GalNAcol. Analysis of Oligosaccharide H3"Because oligosaccharide H3 was resistant to digestion with the a-2,3-specific neuraminidase from Newcastle Disease Virus (Fig. l , C and D), it was proposed that H3 is derived from H1 (GalPl-, 3GalNAcol) by the addition of 1 sialic acid residue linked in a manner other than 2,3. Methylation analysis identified terminal galactose, terminal N-acetylneuraminic acid, and 3,6-disubstituted N-acetylgalactosaminitol. All of these results defined the structure of oligosaccharide H3 as GalPl+ 3(NeuAca2+6)GalNAcol.
Analysis of Oligosaccharide H4"The digestion by Newcastle Disease Virus neuraminidase of oligosaccharide H4 to a species which co-migrated with H3 on ISSA-HPLC (Fig. 2 0 ) indicated that H4 was derived from H3 (Gal@l~3(NeuAcaZ--, 6)GalNAcol) by the addition of a sialic acid residue linked a2,3 to the terminal galactose of H3. Methylation analysis of oligosaccharide H4 identified terminal N-acetylneuraminic acid, 3-monosubstituted galactose, and 3,6-disubstituted Nacetylgalactosaminitol. These results demonstrated that the structure of oligosaccharide H4 was NeuAca2--*3Gal@l--, 3(NeuAca2-+6)GalNAcol. It is not known why oligosaccharide H4 eluted later than a disialylated complex oligosaccharide on anion exchange HPLC (Fig. 1); however, the absence of more extensively substituted sugars in methylation analysis definitively ruled out the presence of additional sialic acid residues. Variability is often noted in the elution position of disialylated oligosaccharides on anion exchange HPLC (17). This may be related to differences in net carbohydrate composition (31) or subtle differences in the net charge of differently linked sialic acids at pH 4.0.
Analysis of Fraction X-Fraction X eluted in the position of a monosaccharide in ISSA-HPLC but did not co-migrate with any monosaccharide standard on borate anion exchange HPLC (Fig. 3). This species was not 3H-GalNAcol. Gas liquid chromatography-mass spectrometry of methylated and trimethylsilylated derivatives did not reveal the presence of any identifiable saccharide material. We therefore conclude that X is either a product of the alkaline-peeling reaction (32) or is a radioactive contaminant from the NaB[3H]4 reagent. Distribution of Oligosaccharides in the Hinge Region-Defined glycopeptides encompassing either the entire hinge region or subfragments within it were subjected to alkaline degradation and the 3H-oligosaccharide products were analyzed by ISSA-HPLC. The relative compositions of oligosaccharide species H1, H2, H3, and H4 present on the different glycopeptides are displayed in Table 111. Glycopeptides C699-162, C6106-137, and complete 1gD:WAH contained all of the hinge region glycosylation sites. Glycopeptide C6106-124 contained the glycosylated serine at C6109, and glycopeptide (36125-137 contained t>he glycosylated threonine residues at C6126, and 127 and (26131 and/or Cb132 (14). A similar distribution of oligosaccharide species was present on all glycopeptides which indicated that all structures are likely to be found at each site and that there is little specificity of the pattern of glycosylation at different glycosylation sites in the hinge region. The overall presence of sialylated species in ISAA-HPLC was consistent with the quantities of sialic acid found in carbohydrate analysis ( Table I) and indicated that sialic acid was not lost nonspecifically during the purification of the hinge subfragments or during alkaline degradation. The presence of an 0-glycosidically linked oligosaccharide at serine C6109, which had not been indicated by a conversion of 1 residue of serine to alanine after alkaline degradation (Table I), was confirmed by the detection of the full set of 3H-oligosaccharides liberated from glycopeptide C6106-125 (Table 111). We presume that the unexpectedly elevated level of serine in the amino acid analysis following alkaline degradation was due to the presence of a serine-containing contaminant.

DISCUSSION
A schematic representation of the constant region of the human 6 chain is depicted in Fig. 4. Emphasis has been placed on oligosaccharide structures and their localization. The predominant oligosaccharide species in the hinge region of 1gD:WAH is the nonsialylated disaccharide Galpl-3GalNAc. This structure comprises "50% of all oligosaccharides at each hinge region glycosylation site. The two monosialylated isomers NeuAca2+3Galpl+=3GalNAc and Galpl"+ 3(NeuAcaS-&)GalNAc comprise -40%, and the remaining -10% of the oligosaccharides have the structure NeuAca2+ 3Ga1/31-+3(NeuAca2"+6) GalNAc. Although the sialylated 0linked oligosaccharides of the IgD hinge region have not previously been described on other human immunoglobulins, all of the oligosaccharide s t~c t u r e s determined in this study have also been found on a variety of secreted and membrane associated glycoproteins. These include canine submaxillary mucin f33), bovine kininogen (34), and human erythrocyte sialoglycoprotein (35), among others (36).
In IgD:WAH, 0-glycosidically linked oligosaccharides have been localized to serine C6109 and threonines C6126, 127, 131, and 132. It is uncertain whether both, or only one of the threonines at C6131 and C6132 are glycosylated (14). The amino acid sequence of the constant region of another human IgD myeloma heavy chain (NIG-65) has recently been determined (37,38). The amino acid residues described thus far for C6:NIG-65 are identical with those of 1gD:WAH and include the hinge region sequence. It is notable that differences in the location of 0-glycosidically linked hinge region oligosaccharides have been reported (14,38). 1gD:NIG-65 is reported to contain 7 sites of 0-linked glycosylation encompassing the 5 proposed for 1gD:WAH plus SerC6110 and ThrC6113 (38). This variability in oligosaccharide locabation may be secondary to possible differences in Fd or light chain sequences resulting in differential exposure of hinge region glycosyltransferase acceptor sites during biosynthesis. Alternatively, genetic or clonal differences in the glycosyltransferase properties of the different plasmacytomas may underlie the differences which have been observed. Based on the proposed distribution of the 0-linked glycosylation sites in the hinge region of 1gD:NIG-65, Takayasu et al. (38) have proposed two rules which designate particular amino acid sequences as generalized acceptor sequences for the 0-linked glycosylation of glycoproteins. These rules propose that the specific amino acid sequences Ala-X-Ala-Ser-Ser or Ala-X-Ala-Thr-Thr (quintet rule) and Val-Pro-Thr (triplet rule) may serve to elicit the transfer of GalNAc to the serine and threonine residues of these sequences in human IgD as well as in other glycoproteins. The proposed rules are unlikely to be generally applicable for a number of reasons. 1gD:WAH contains the same hinge region amino acid sequence as 1gD:NIG-65 but is not glycosylated at all of the same hinge region positions (14). This indicates that the quintuplet and triplet sequences are not strict determinants of oligosaccharide localization. IgAl contains an extensive degree of 0-glycosylation in its hinge region; however, the amino acid sequences surrounding these oligosaccharides are predominantly a repeating pattern of Pro-Ser and none of the quintuplet or triplet sequences are present (1). There is also a considerable body of data available which indicates that in http://www.jbc.org/ Downloaded from a variety of glycoproteins which contain 0-linked oligosaccharides, the glycosylated serine and threonine residues are not immediately surrounded by any specific set of amino acid residues (reviewed in Refs. [39][40][41]. Aubert et ai. (42) examined the amino acid sequences surrounding 9 different sites of 0glycosylation and did not find any characteristic amino acid sequence associated with the 0-glycosylated residues. Aubert et al. (42), through the application of a computer program for the prediction of protein secondary structure based upon amino acid sequence data (43), also made the observation that each of the sites of 0-glycosylation examined in their study was capable of participating in a @-turn structure. The authors concluded that the 8-turn conformation served an important role in maintaining the accessibility of specific serine and threonine residues to the action of Nacetylgalactosaminyltransferase, the initial enzyme in the biosynthetic pathway of 0-linked oligosaccharides (42). This is consistent with the fact that 0-glycosylation is generally a late event in post-translational modification and occurs after proteins have attained a highly ordered structure (44). It is likely that accessibility of specific Ser and Thr residues to the N-acetylgalactosaminyltransferase as well as other higher order structural features will prove to be more critical determinants in the specific localization of 0-linked oligosaccharides than will the presence of characteristic amino acid sequences as seen in N-glycosylation.
According to Putnam et al. (45) the hinge region of the 6 chain differs from the hinges of y and (Y in four notable characteristics: 1) its extreme length (-64 residues); 2) its division into a GalNAc-rich NH2-terminal half and a highly charged COOH-terminal half; 3) its composition and unusual distribution of amino acids; and 4) the presence of a single half-cystine. The highly charged segment of the IgD hinge exhibits extreme sensitivity to proteolysis and is capable of forming an a-helical structure; however, the GalNAc-rich segment is relatively resistant to protease digestion and is believed to exist in a random conformation (14).
The randomness of peptide conformation in the GalNAcrich region of the hinge may be further amplified by the heterogeneity of oligosaccharide structure which we found to be present at different glycosylation sites. The same distribution of oligosaccharide structures present in the hinge region as a whole was found on glycopeptides which encompassed either the glycosylated serine at C6109 or the glycosylated threonine residues between CS126 and CS132 (Table  111). This heterogeneity of oligosaccharide structure can be expected to further accentuate the configurational disorder of the hinge region which has been predicted on the basis of amino acid sequence alone (14).
The absence of GalNAc alone or NeuAccu24GalNAc linked to serine or threonine provides suggestive evidence that the glycosylation pattern observed on 1gD:WAH reflects biosynthetic variability in the activities of neuraminyl transferases upon an obligate Gal@l+3GalNAc-Ser/Thr substrate in these cells. It has been noted, however, that the 1,3 8galactoside linkage in this structure is difficult to degrade under laboratory conditions (1,30). This observation leaves open the possibility that more extensively sialylated oligosaccharides were initially synthesized and that the low level of disialylated oligosaccharides observed may be the result of glycosidases present in the serum of this myeloma patient. It will be necessary to perform biosynthetic studies with IgD plasmacytoma cells in vitro in order to ultimately resolve this issue.
The heterogeneity of oligosaccharide structure found in the hinge region of 1gD:WAH as well as the ubiquitous presence of these structures on many glycoproteins suggest that the function of these glycans is not related to a particular mechanism involving the biological recognition of specific carbohydrate structures; however, this possibility has not been ruled out. The model of membrane IgD-mediated lymphocyte activation proposed by Putnam et al. (45) postulates that prior to antigen exposure, the GalNAc-rich portion of the hinge region serves to protect the high charge segment from proteolysis. Upon antigen binding, a conformational change would expose the high charge segment to enzymatic cleavage and ultimately result in blast transformation by one or more mechanisms (45). Testing of this hypothesis awaits determination of the structure of the oligosaccharides present on the membrane form of IgD and the performance of further immunological studies.