N-Glycan Structures of Pigeon IgG A MAJOR SERUM GLYCOPROTEIN CONTAINING Galα1–4Gal TERMINI

We had shown previously that all major glycoproteins of pigeon egg white contain Galα1–4Gal epitopes (Suzuki, N., Khoo, K. H., Chen, H. C., Johnson, J. R., and Lee, Y. C. (2001) J. Biol. Chem. 276, 23221–23229). We now report that Galα1–4Gal-bearing glycoproteins are also present in pigeon serum, lymphocytes, and liver, as probed by Western blot with Griffonia simplicifolia-I lectin (specific for terminal α-Gal) and anti-P1 (specific for Galα1–4Galβ1–4GlcNAcβ1–) monoclonal antibody. One of the major glycoproteins from pigeon plasma was identified as IgG (also known as IgY), which has Galα1–4Gal in its heavy chains. High pressure liquid chromatography, mass spectrometric (MS), and MS/MS analyses revealed that N-glycans of pigeon serum IgG included (i) high mannose-type (33.3%), (ii) disialylated biantennary complex-type (19.2%), and (iii) α-galactosylated complex-type N-glycans (47.5%). Bi- and tri-antennary oligosaccharides with bisecting GlcNAc and α1–6 Fuc on the Asn-linked GlcNAc were abundant among N-glycans possessing terminal Galα1–4Gal sequences. Moreover, MS/MS analysis identified Galα1–4Galβ1–4Galβ1–4GlcNAc branch terminals, which are not found in pigeon egg white glycoproteins. An additional interesting aspect is that about two-thirds of high mannose-type N-glycans from pigeon IgG were monoglucosylated. Comparison of the N-glycan structures with chicken and quail IgG indicated that the presence of high mannose-type oligosaccharides may be a characteristic of these avian IgG.

N-glycans were found to possess Gal␣1-4Gal at the non-reducing termini of tri-, tetra-, and penta-antennary structures (2). No sialylation is found on the branch that contains the Gal␣1-4Gal sequence. Because biosynthesis and glycosyl modification of major egg white glycoproteins of chicken are carried out in tubular gland cells of oviduct (3), the most likely site of fabricating Gal␣1-4Gal linkages on the PEW glycoproteins would be in the corresponding cells of pigeon oviduct.
The presence of Gal␣1-4Gal or substances similar to P 1 antigen on glycoproteins is limited even among other birds that have been studied. For example, P 1 antigenic activities is absent in the blood of chicken, gander, turkey, quail, duck, and pheasant (16,17), and no ␣-galactoside is found in the glycans of ovomucoid from chicken, quail, and duck (18 -22). On the other hand, P 1 antigenic activities have been found in the blood and/or eggs of pigeon, turtle dove (Streptopelia resoria), budgerigar, and cockatiel (16,23,24). Salivary gland mucin glycoproteins of Chinese swiftlets (genus Collocalia) contain O-linked glycans with Gal␣1-4Gal sequence (25).
The biological significance of Gal␣1-4Gal in pigeon and some other avian glycoproteins is still unknown. Whether Gal␣1-4Gal is located specifically on PEW glycoproteins or glycoproteins of other pigeon organs is still unsettled. François-Gérard and co-workers (16,17) reported the presence of P 1 antigenic activity in pigeon blood. However, they did not indicate whether the antigenicity was located on glycolipids or glycoproteins and whether the P 1 antigens were produced by pigeon themselves or derived from foreign bodies. To understand further the characteristics of species-specific oligosaccharides, we investigated the presence of glycoproteins with Gal␣1-4Gal in pigeon serum, lymphocyte, and liver. Here we report that one of the major glycoproteins containing Gal␣1-4Gal in serum is pigeon IgG, which possesses complex-type N-glycans containing Gal␣1-4Gal as well as high mannosetype oligosaccharides. The high mannose-type N-glycans are exclusively found on the CH3 domain, and the site specificity is probably a characteristic in avian IgGs. Unlike IgGs in other animals, the complex-type glycans were mainly tri-as well as bi-antennary structures. Moreover, N-glycans containing both Gal␣1-4Gal␤1-4Gal␤1-4GlcNAc and Gal␣1-4Gal␤1-4Glc-NAc branches were detected by MS/MS analysis, which were not found in PEW glycoproteins.
Buffers and Standard Procedures-Tris-buffered saline contains 50 mM Tris⅐HCl (pH 7.4) and 150 mM NaCl. TBST contains 0.1% Tween 20 in Tris-buffered saline. Guanidine⅐HCl (8 M) was made in 0.2 M Tris⅐HCl (pH 8.0). Procedures for SDS-PAGE, Griffonia simplicifolia-I lectin blotting, and N-terminal sequence analyses have been described previously (1). Sequence identity search was performed using NCBI data bases. Protein concentrations were measured by the BCA assay (26) using bovine serum albumin (BSA) as a standard.
Preparation of Extracts from Pigeon Tissues and Cells-Blood (ϳ5 ml/pigeon) drawn from six pigeons were collected in blood collection tubes (13 ϫ 75 mm) containing 0.057 ml of 15% K 3 EDTA solution (BD Biosciences). The lymphocytes were isolated with Ficoll-Paque TM Plus and washed with Dulbecco's phosphate-buffered saline twice. Pellets of lymphocytes and recovered plasma were stored at Ϫ20°C until use. The lymphocytes and a portion of pigeon liver were homogenized in 100 mM sodium cacodylate (pH 7.0) containing 1% Triton X-100 at 4°C. The supernatants separated by centrifugation were used as extracts.
Isolation of Major Glycoproteins from Pigeon Plasma-Diluted pigeon plasma (ϳ6.2 mg of protein/ml) was centrifuged to remove insoluble materials. Supernatant was filtered through a 0.45-m membrane and injected onto a Superdex 200 column (maximum 0.2 ml for an HR 10/30 column, and 1 ml for a HiLoad 26/60 column) equilibrated with 50 mM sodium phosphate (pH 7.0) containing 150 mM NaCl. The flow rates were 0.25 ml/min for the HR 10/30 column and 2 ml/min for the HiLoad 26/60 column. Proteins in the eluate were monitored by A 280nm , and individual peaks were collected manually. GS-I-positive fractions were concentrated with ultrafiltration (YM-10 membrane) and desalted by washing with 10 mM Tris⅐HCl (pH 8.0) for further purification by ion exchange chromatography. A column of DEAE-Sepharose Fast Flow (HiTrap, 1 ml) was washed with 1 M NaCl in 10 mM Tris⅐HCl (pH 8.0) and equilibrated with 10 mM Tris⅐HCl (pH 8.0). After the sample injection, the column was washed with 10 mM Tris⅐HCl (pH 8.0) for 30 min (flow rate, 1 ml/min), and then the concentration of NaCl in the elution was linearly increased up to 0.4 M within 80 min (flow rate, 0.5 ml/min). The major peak was collected and concentrated with a Centricon YM-10.
GAF Treatment of Glycoproteins-Glycoprotein samples (10 g each) in 20 l of 0.4% SDS and 100 mM 2-mercaptoethanol in 10 mM Tris⅐HCl (pH 8.0), were heat-denatured at 90°C for 3 min. After the solutions were cooled to room temperature, 1% (v/v) Nonidet P-40, and 1 unit of GAF were added to the solution of glycoproteins. The reaction mixtures were incubated at 37°C for 16 h to complete de-N-glycosylation and then heated at 100°C for 5 min to inactivate GAF.
Preparation of Glycopeptides for Peptide Sequencing-To reduce the existing disulfide bonds and to block random reformations of intramolecular disulfide bonds, 16 mol of tributylphosphine and 0.16 mmol of 4-vinylpyridine were added to 1 mg of pigeon IgG in 0.1 M Tris⅐HCl (pH 8.5) and incubated at room temperature for 2 h under nitrogen. The reaction mixture was dialyzed against water and lyophilized. One-third of the reduced and alkylated pigeon IgG was suspended with 40 l of 50 mM NH 4 HCO 3 (pH 8.4) and incubated with 12.5 g of trypsin (sequencing grade) at 37°C, overnight. A portion of the supernatant was further treated with GAF (0.5 units/10 l) at 37°C, overnight. The peptide mixtures, before and after the GAF treatment, were analyzed with reversed phase-HPLC on a Shim-pack CLC-ODS column. The mobile phases were solute A, 0.05% trifluoroacetic acid, and solute B, 90% CH 3 CN in the 0.05% trifluoroacetic acid. Elution (1 ml/min) was conducted by a linear gradient of 0 -50% of solute B in solute A developed over 100 min. Each peak monitored by A 210nm was collected and kept at 4°C. A portion of the isolated glycopeptides was treated with GAF, and the released oligosaccharides and the deglycosylated peptide were separated with HPLC.
Preparation of Oligosaccharides for Structural Analyses-Pigeon IgG (1 mg) was reduced with 180 l of 60 mM dithiothreitol in 8 M guanidine⅐HCl at room temperature for 1 h. For alkylation of thiols, 240 l of 0.18 M iodoacetamide in 8 M guanidine⅐HCl was added, and the mixture was incubated at room temperature for 30 min in the dark. The reaction mixture was dialyzed against water and lyophilized. The alkylated pigeon IgG was suspended with 200 l of 50 mM NH 4 HCO 3 (pH 8.4), digested with 10 g of trypsin at 37°C for 4 h, and followed by addition of 10 g of chymotrypsin (37°C overnight). After inactivating the enzymes at 100°C for 10 min, the digest was lyophilized. Oligosaccharides were released with GAF treatment in 50 mM NH 4 HCO 3 (pH 7.8) at 37°C overnight. After inactivating GAF at 100°C for 5 min, the digest was lyophilized. To remove inorganic cations and peptides, the mixture was loaded onto 1 ml of Dowex 50W ؋ 2 (H ϩ form, 50 -100 mesh, Sigma) packed in a 1-ml syringe. The column was washed with 5 ml of water, and the collected effluent was lyophilized. The sample was dissolved with 200 l of water, loaded onto a Carbograph tube (25 mg, Alltech), washed with 1 ml of water, and then eluted with 500 l of 25% CH 3 CN containing 0.05% trifluoroacetic acid.
Exo-glycosidase Digestion-The released N-glycan mixtures were digested sequentially with 50 milliunits of neuraminidase from Arthrobacter ureafaciens (Roche Applied Science) in 25 l of 50 mM sodium acetate buffer (pH 5.0), 0.5 units of ␣-galactosidase (from green coffee bean, Calbiochem) in 100 l of 50 mM citrate-phosphate buffer (pH 6.0), followed by 5 milliunits of ␤1-4 galactosidase (from Streptococcus pneumoniae, Calbiochem) in 50 l of 50 mM sodium acetate buffer (pH 6.0). All digestions were carried out at 37°C for 24 h. Each enzyme digestion was desalted by passing through a mixed bed ion-exchange column packed with a mixture of 1 ml each of Dowex 50W-X8 (50 -100 mesh, H ϩ form) and AG 3-X4 (100 -200 mesh, free base form, Bio-Rad) resins, prior to taking an aliquot for permethylation and MS analysis. PAglycans were desalted instead by passing through a Sep-Pak C18 cartridge (Waters), washed with water, and then eluted with 50% methanol in water.
Chemical Derivatization and GC-MS Linkage Analysis-N-Glycan and PA-derivatized N-glycan samples were permethylated using the NaOH/dimethyl sulfoxide slurry method as described by Dell et al. (27). For GC-MS linkage analysis, partially methylated alditol acetates were prepared from permethylated derivatives by hydrolysis (2 M trifluoroacetic acid, 121°C, 2 h), reduction (10 mg/ml NaBH 4 , 25°C, 2 h), and acetylation (acetic anhydride, 100°C, 1 h). GC-MS was carried out using a Hewlett-Packard Gas Chromatograph 6890 connected to a Hewlett-Packard 5973 Mass Selective Detector. Sample was dissolved in hexane prior to splitless injection into a HP-5MS fused silica capillary column (30 m ϫ 0.25 mm inner diameter). The column head pressure was maintained at around 8.2 pounds/square inch to give a constant flow rate of 1 ml/min using helium as carrier gas. Initial oven temperature was held at 60°C for 1 min and increased to 90°C in 1 min and then to 290°C in 25 min.
MS Analysis for Glycans and Glycomics-For MALDI-TOF MS glycan profiling, the permethylated derivatives in acetonitrile were mixed 1:1 with 2,5-dihydroxybenzoic acid matrix (10 mg/ml in acetonitrile), spotted on the target plate, air-dried, and recrystallized on-plate with ethanol whenever necessary. Data acquisition was performed manually on a benchtop MALDI LR system (Micromass) operated in the reflectron mode. For 2,5-dihydroxybenzoic acid matrix, the coarse laser energy control was set at high and finely adjusted using the % slider according to sample amount and spectra quality. Laser shots (5 Hz, 10 shots/spectrum) were accumulated until a satisfactory signal to noise ratio was achieved when combined and smoothed. Glycan mass profiling was also performed on a dedicated Q-TOF Ultima MALDI instrument (Micromass) in which case the permethylated samples in acetonitrile were mixed 1:1 with ␣-cyano-4-hydroxycinnamic acid matrix (in acetonitrile, 0.1% trifluoroacetic acid, 99:1, v/v) for spotting onto the target plate. The nitrogen UV laser (337 nm wavelength) was operated at a repetition rate of 10 Hz under full power (300 J/pulse). MS survey data were manually acquired, and the decision to switch over to CID MS/MS acquisition mode for a particular parent ion was made on-thefly upon examination of the summed spectra. Argon was used as the collision gas with a collision energy manually adjusted (between 50 and 200 V) to achieve optimum degree of fragmentation for the parent ions under investigation.
Off-line nanoelectrospray (nano-ESI) using the borosilicate metalcoated glass capillary option was performed on a Q-TOF Ultima API instrument (Micromass) equipped with a nanoflow source, mainly for CID MS/MS analysis of permethylated glycans samples. The capillary voltage was set at about 1.0 -1.2 kV. Either protonated and/or sodiated doubly or triply charged parent ions may be selected for MS/MS. A collision energy setting of 20 -40 eV was usually sufficient for doubly protonated species, but 60 -80 eV was needed for singly and doubly sodiated parent ions. Permethylated samples were dissolved in 50% acetonitrile, 0.2% formic acid solvent and a 2-l aliquot was typically loaded into the capillary for nano-ESI analysis.
Preparation and Isolation of PA-derivatized Oligosaccharides-Lyophilized free oligosaccharide fractions (from 2 mg of pigeon IgG) were dissolved in 40 l of 2-aminopyridine solution (1 g of 2-aminopyridine in 580 l of concentrated HCl (pH 6.8)) and heated at 90°C for 15 min with heating block. Freshly prepared NaCNBH 3 solution (7 mg/4 l water) was added into the reaction mixture and then heated at 90°C for 1 h. PA-oligosaccharides were fractionated by gel filtration on a Sephadex G-15 column (1.0 ϫ 40 cm, in 10 mM NH 4 HCO 3 ), monitored by a fluorescence (excitation, 300 nm; emission, 360 nm), and lyophilized. The mixture of PA-oligosaccharides was first separated by HPLC with a TSK gel DEAE-5PW column (7.5 ϫ 75 mm) as described previously (28), and the separated neutral, mono-sialyl, and di-sialyl fractions (monitored by fluorescence, Ex, 320 nm; Em, 400 nm) were collected separately and lyophilized. These fractions were individually dissolved in water and separated on a Shim-Pack CLC-ODS column (6.0 ϫ 150 mm) as described previously (1). Elution was performed at a flow rate of 1.0 ml/min at 55°C using eluent A (0.005% trifluoroacetic acid) and eluent B (0.5% 1-butanol in eluent A). The column was equilibrated with a mixture of eluents A/B ϭ 90:10 (v/v), and after injection of a sample, the composition of the eluents was changed linearly to A/B ϭ 30:70 in 120 min. Each peak was collected, lyophilized, and analyzed with MALDI-TOF-MS. For determination of N-glycan structures by two-dimensional mapping with HPLC, the isolated PA-oligosaccharides were analyzed using CLC-ODS and TSK gel amide-80 columns (4.6 ϫ 250 mm) as described previously (29). Elution position of each of PAoligosaccharides on CLC-ODS and amide-80 columns was expressed as glucose units (GU), based on the elution position of isomalto-oligosaccharide series. Reference PA-oligosaccharides from asialo-human IgG, bovine RNase B, and chicken serum IgG were prepared by the same method. Structures of the reference compounds are shown in Fig. 9. PA-derivatized N-glycans B-1, C-1, and D ( Fig. 9) were obtained from ␣-fucosidase and N-acetyl-␤-D-hexosaminidase-digested PA-oligosaccharides from human IgG. N-Glycans B-2 and C-2 were prepared from B-1 and C-1, respectively, with ␤-galactosidase digestion.

Detection of ␣-Galactoside on Pigeon Glycoproteins by GS-I Lectin and Anti-P 1 mAb
Approximately equal amounts of proteins from pigeon serum, liver, lymphocytes, and egg white were subjected to SDS-PAGE, transferred onto a PVDF membrane, and visualized by staining with Coomassie Brilliant Blue (CBB) (Fig. 1A, CBB staining) or probed by blotting with GS-I lectin to detect terminal ␣-Gal residues. As shown in Fig. 1A (GS-I staining), some proteins from pigeon serum, liver, and lymphocytes were visualized by this staining as were glycoproteins of PEW. Intensities of the GS-I staining did not coincide with those of the CBB staining, suggesting that proteins carry ␣-Gal residues with varied density. Immunoblotting with anti-P 1 mAb (specif-ic for P 1 blood type) (Fig. 1A) gave similar staining patterns as shown by the GS-I lectin blotting. Extract from pigeon heart also stained with GS-I and anti-P 1 mAb (data not shown). These data indicate that many kinds of glycoproteins in pigeon organs most likely contain terminal Gal␣1-4Gal as found in PEW (1).
The presence of ␣-Gal residues on proteins was probed in pigeon, chicken, bovine, and human sera. Although approximately the same amounts of respective proteins in these sera were stained with CBB, only pigeon serum proteins stained strongly by GS-I (Fig. 1B). Because twice as much proteins was used in this experiment (Fig. 1B) as the previous experiment ( Fig. 1A), more bands were clearly visualized by the GS-I staining. Bovine is known to produce Gal␣1-3Gal (30 -32), but its serum was stained with GS-I only faintly. Serum proteins of chicken and human, not known to produce Gal␣1-3Gal or other ␣-Gal epitopes, was not stained by GS-I, confirming that these serum proteins have no terminal ␣-galactosyl residues.

Isolation of Pigeon Plasma Glycoproteins
Pigeon plasma proteins (600 g total) were separated by gel filtration using a Superdex 200 HR 10/30, as shown in Fig A, expression of Gal␣1-4Gal in pigeon. Proteins (5 g each) from pigeon serum, liver, lymphocyte, and egg whites were heat-denatured with sample buffer containing 3% SDS and 5% 2-mercaptoethanol and separated by SDS-PAGE (12.5%). After electrophoresis, the proteins were transferred to PVDF membranes and stained with CBB, GS-I lectin, or anti-P 1 mAb. B, comparison of the presence of ␣-galactosyl residue on pigeon, chicken, bovine, and human sera glycoproteins from by GS-I lectin blotting. Ten g of proteins from sera were separated by SDS-PAGE, transferred to a PVDF membrane and stained with CBB or GS-I.
were negative to GS-I. The fraction 13 was the largest peak detected by A 280nm , but no protein/peptide bands were visualized with CBB or GS-I staining. One of the glycoproteins in the fraction 6, which was eluted at the same elution position as chicken IgG, showed the strongest intensity by the GS-I staining. In contrast, major proteins in fraction 8, where BSA is expected to elute, were stained with CBB strongly but not at all with GS-I. The similar elution profile was detected when 6 mg of pigeon plasma proteins was separated using a semi-preparative Superdex 200 column (2.6 ϫ 60 cm) (data not shown).
Fraction 6 from Superdex 200 was further purified with a DEAE-Sepharose column eluting with a gradient of NaCl, as shown in Fig. 2D. The major peak on the anion-exchange column was collected, and the purity of the protein was confirmed by SDS-PAGE (Fig. 2D, inset). As in the case of chicken immunoglobulins (33), two distinct bands (67 and 27 kDa) were visualized, suggesting it to be IgG. The protein after the purification with gel filtration and anion-exchange columns was estimated to be ϳ5 mg from 1 ml of pigeon serum. This value is within the range of reported concentration of IgG in adult pigeon serum (5.92 mg/ml) (34).

Identification of the Pigeon Serum Glycoprotein by
Partial Peptide Sequencing N-terminal amino acid sequences of the light (L) and heavy (H) chains were homologous to the corresponding variable regions of immunoglobulins of chicken, duck, and human (Table  I). Because avian IgG, IgA, and IgM share the same N-terminal variable domains (35), internal peptide sequencing on the Fc region would be required for the determination of the immunoglobulin classes (36 -38). The predominant serum immunoglobulin in birds is IgG, also known as IgY, because of its unique structure. It is functionally equivalent to mammalian IgG but has one additional constant region domain in its heavy (H) chains (Fig. 3) (33,39). Throughout this paper, we refer to the predominant serum immunoglobulin in birds as serum IgG for convenience. The amino acid sequences of chicken and duck immunoglobulin chains (H chains) suggest that two potential N-linked glycosylation sites are indeed well conserved between chicken and duck, located in the Fc region ( Fig. 3) (36,40). In contrast, positions of potential N-glycosylation sites on Fc regions of avian IgA and IgM are different from those of IgG. Therefore, isolation and peptide sequencing of N-glycosyl peptides were performed in order to determine the immunoglobulin class based on sequence homology. The C18-HPLC elution profiles of tryptic peptides of pigeon IgG, before and after GAF treatment, indicated that ''Peak A'' shifted its position from 46.4 to 50.1 min after GAF treatment (Fig. S1). As shown in Table I, the glycopeptide sequence of the pooled peak A is homologous to that of a CH3 domain of chicken or duck IgG. The sequence alignment showed that conservation of the potential N-glycosylation sites is extended to pigeon, chicken, and duck but not to chicken IgM and IgA. Based on these sequence data, the isolated protein was identified as pigeon serum IgG. The CH3 domains of chicken and duck IgG resemble the CH2 domain of mammalian IgG (36,39,40). This appears to be the case for pigeon IgG also, because the peptide sequence containing the N-glycosylation site was similar to the corresponding site of the CH2 domains of human (Table I) and other mammalian IgG (data not shown).

Release of ␣-Galactoside-containing Oligosaccharides from Pigeon IgG with GAF
Existence of ␣-galactoside-containing N-glycans in pigeon IgG was probed by SDS-PAGE before and after GAF digestion. Decrease in molecular size after GAF digestion (analyzed by SDS-PAGE) was apparent in the H chains of pigeon IgG, as was observed in chicken IgG (Fig. S2A). The band of L chains were not shifted by the GAF treatments. The H chain of pigeon IgG was originally stained with GS-I strongly but not stained at all after de-N-glycosylated (Fig. S2B). This indicates that ␣-Gal-bearing oligosaccharides on pigeon H chain were completely removed by GAF. The H and L chains of chicken IgG were not stained by GS-I, confirming that they do not possess ␣-galactosyl residues.

MS and MS/MS Analysis of N-Glycans from Pigeon IgG
The N-glycans released by GAF from pigeon IgG were first profiled by MALDI-TOF MS analysis of the permethylated derivatives. Substantial heterogeneity was evident (Fig. 4A), but the glycans can be conveniently categorized into the following: (i) high mannose-type, (ii) disialylated biantennary complex-type, and (iii) ␣-galactosylated complex-type structures.
High Mannose-type N-Glycans-The two most abundant peaks in Fig. 4A correspond to Hex 9 HexNAc 2 (m/z 2397) and Hex 10 HexNAc 2 (m/z 2601), which also afforded oxonium-type fragment ions at m/z 2097 and 2301, respectively, via cleavage at the chitobiose core. Smaller amounts of Hex 5 HexNAc 2 (m/z 1579) and Hex 8 HexNAc 2 (m/z 2192) were also present. Upon ␣-mannosidase treatment, only these components were digested, giving two major signals at m/z 1579 (Hex 5 HexNAc 2 ) and 1783 (Hex 6 HexNAc 2 ) in addition to a signal corresponding to Hex 1 HexNAc 2 (data not shown). The same digestion when applied to high mannose structures derived from RNase B yielded only the expected product, i.e. Man 1 GlcNAc 2 . The data are therefore consistent with the presence of a Glc residue at one of the non-reducing termini of a Man 9 GlcNAc 2 that would block the digestion of the Man residues (later to be shown to be on the ␣3-arm), allowing a removal of only 4 or 5 Man residues from the ␣6-arm. The presence of a monoglucosylated Man 9 GlcNAc 2 along with Man 9 GlcNAc 2 as the major glycoforms has been reported for chicken and quail IgG (41-43) and may be a characteristic of avian IgG. Further characterization based on two-dimensional HPLC mapping of the PA-derivatives confirmed this structure (see below) and has further resolved Hex 9 HexNAc 2 into Man 9 GlcNAc 2 (n-6, major component) and Glc 1 Man 8 GlcNAc 2 (n-7, minor component), along with Glc 1 Man 9 GlcNAc 2 (n-8), as the predominant high mannose-type structures. Interestingly, only (ϮGlc 1 )Man 9 GlcNAc 2 and (ϮGlc 1 )Man 8 GlcNAc 2 structures were found on the glycopeptide isolated from the CH3 domain (Fig. 4C).
Disialylated Biantennary N-Glycans-The sialylated components were first identified by the m/z values of the molecular ion signals and subsequently confirmed by digestion with neuraminidase (Fig. 4B). A prominent desialylated component afforded a major molecular ion at m/z 2071 corresponding to a biantennary complex-type structure that was absent before desialylation (Fig. 4A). The biantennary branching pattern was  established by MS/MS analysis (Fig. S3). In particular, the ion at m/z 1143 from MALDI-MS/MS (Fig. S3B) unambiguously defined the presence of two antennae extending from the trimannosyl core, whereas the ion pair at m/z 432 and 464 from nano-ESI-MS/MS (Fig. S3A) indicated an N-acetyllactosamine (LacNAc) unit and not a type 1 chain (Gal␤1-3GlcNAc) (27). Removal of two Hex units by ␤4-galactosidase from S. pneumoniae resulted in a shift of the corresponding molecular ion signal at m/z 2071 (Fig. 5A) to m/z 1661 (Fig. 5B) which confirmed this assignment. Finally, methylation analysis of the desialylated samples clearly demonstrated a disappearance of 6-linked Gal (data not shown) as compared with the GC-MS profile of the sample not treated by neuraminidase (Fig. S4). The molecular ion at m/z 2794 in Fig. 4A could thus be assigned as ␣2-6-disialylated biantennary complex-type N-glycan which shifted to give a molecular ion of m/z 2071 after neuraminidase digestion (Fig. 4B) and remained resistant to ␣-galactosidase digestion (Fig. 5A). No other sialylated component of high abundance was apparent from the analysis of the total mixtures, although monosialylated counterparts of several other complex-type structures could be detected as minor peaks, the detailed analysis of which was made possible after HPLC fractionation (see below).
␣-Galactosylated N-Glycans-Consistent with the earlier detection by antibody and lectin, the majority of the non-sialylated complex-type structures were shown to carry terminal ␣-Gal by two criteria. First, the m/z of most of the molecular ions detected shifted to lower values after digestion with ␣-galactosidase (Fig. 5A). Second, methylation analysis indicated a high abundance of terminal Gal and 4-linked Gal (Fig. S4), supporting the presence of Gal␣1-4Gal epitope. The mass spectrum (Fig. 4A) indicated a high degree of heterogeneity within the ␣-Gal-containing N-glycans, but their m/z signals could be assigned as having Ϯ(Fuc 1 )Hex 2-5 HexNAc 3 and Ϯ(Fuc 1 )-Hex 4 -8 HexNAc 4 on the trimannosyl core (Man 3 GlcNAc 2 ). After sequential ␣-galactosidase and ␤4-galactosidase digestion, the major products observed were (GlcNAc 4  to the trimannosyl core (data not shown). No HexNAc-HexNAc type termini were detected. The other major molecular ion signal at m/z 1661, (GlcNAc 2 )Man 3 GlcNAc 2 , corresponded to the digested product of the sialylated biantennary structure, whereas the high mannose structures at m/z 2397 and 2601 remain unchanged.
When the major molecular ion signals from samples not treated with the exoglycosidase (Fig. 4A) were subjected to MALDI CID MS/MS analysis, additional unique structural features were unveiled. Fig. 6 shows spectra of MS/MS analysis for the Hex 4 -8 HexNAc 4 CF series, where CF denotes fucosylated trimannosyl core. Interpretation of the MS/MS data for the Hex 4 -8 HexNAc 4 CF series was summarized in Fig. 7. As expected, the presence of Gal-Gal-GlcNAc terminal epitope was indicated by the sodiated b ion at m/z 690 which could be detected in all of the glycans analyzed. For the larger components, however, another b ion at m/z 894, corresponding to sodiated Hex 3 HexNAc ϩ , was also very prominent. In the case of Hex 8 HexNAc 4 CF, sequential loss of the Hex 2 HexNAc or Hex 3 HexNAc terminal units concomitant with emergence of free OH group(s) yielded the complementary set of y ions (e.g. at m/z 3292, 2420, 1548, 3088, and etc., see Fig. 7E). MS/MS data also indicated that the majority of the Hex 8 HexNAc 4 CF components is triantennary carrying a bisecting GlcNAc that is readily lost from the parent ion under the MS/MS conditions employed. Two ions at m/z 1548 (corresponds to the fucosylated trimannosyl core with a bisecting GlcNAc and three free OH groups resulting from the loss of all three antennary sequences, Fig. 7E, left) and m/z 1289 (corresponds to a further loss of the bisecting GlcNAc giving a total of four free OH groups, Fig. 7E, right), which were present in all Hex 4 -8 HexNAc 4 CF series structures, supported the presence of bisecting GlcNAc. Thus, the structure of Hex 8 HexNAc 4 CF is most likely a triantennary bisected and fucosylated trimannosyl core, extended with one Hex-Hex-HexNAc (Hex 2 HexNAc) and two Hex-Hex-Hex-Hex-NAc (Hex 3 HexNAc) sequences (Fig. 7E).

Fractionation and Structural Analysis of PA-derivatized N-Glycans from Pigeon IgG
The presence of Gal 3 GlcNAc antenna was unexpected and has resulted in greater heterogeneity, because each glycan as assigned above may consist of smaller amounts of isomers with different combinations of fully galactosylated to non-galactosylated antennae. This was found to be the case when the glycans from pigeon IgG were derivatized with 2-aminopyridine and fractionated first into neutral, monosialylated, and disialylated fractions with a DEAE column (Fig. 8A) and then further fractionated on an ODS column (Fig. 8B), followed by MS and MS/MS analyses of the collected major fractions (Tables II  and III and Supplemental Material Figs. S1 and S2). Molar ratios of pigeon IgG N-glycans calculated from the peak areas of PA-oligosaccharides were neutral 64.8% (high mannosetype, 33.3%, others, 31.5%), monosialylated 16.0%, and disialylated 19.2%. N-Glycan structures of the three categories were determined as follows.
PA-derivatized High Mannose-type N-Glycans-Mass spectrometric analysis revealed that fractions of neutral N-glycans eluted at around 4 -12 min on the ODS column (n-5 to n-10) were Hex 5, 8, 9 -11 HexNAc 2 -PA, suggesting that these are high mannose-type oligosaccharides. Their elution positions on the ODS column also confirm this assignment (29). The fractionated high mannose- type N-glycans (n-5, n-6, n-7, n-8, and n-9) were further examined by the two-dimensional HPLC mapping (29). The elution position of the PA-N-glycans on both ODS and  Fig. 7. Some fragment ions could not be labeled or assigned, because of complication arising from isomeric structures. amide-80 columns, before and after ␣-mannosidase digestion, were compared with those of the reference compounds derived from bovine RNase B and chicken serum IgG (Fig. 9). Because the elution positions on the two columns as well as mass values for the N-glycans of n-5, n-6, n-7, n-8, and n-9 were indistinguishable with those of Man 8 GlcNAc 2 -PA, Man 9 GlcNAc 2 -PA, Glc 1 Man 8 GlcNAc 2 -PA, Glc 1 Man 9 GlcNAc 2 -PA, and Man 5 -GlcNAc 2 -PA, respectively (data not shown), their structures were determined as shown in Table II. PA-derivatized Disialylated Biantennary N-Glycans-The two disialylated N-glycans (ds-6 and ds-8) were analyzed by MALDI-MS before and after permethylation to define their molecular compositions and by MALDI MS/MS and HPLC for structural assignment (Table III). N-Glycan ds-8 clearly corresponded to the disialylated biantennary structures detected in the MS analysis of the total mixtures for which further MS/MS analysis of the desialylated structures (Fig. S3) have unambiguously established its biantennary structures. In addition, after ␣-neuraminidase digestion, the elution positions of ds-8 on HPLC columns were indistinguishable from human IgG Nglycan D (Fig. S5B), confirming its structure as non-bisected biantennary complex-type without core fucosylation. MS and MS/MS analysis of N-glycans ds-6 ( Fig. S5A and Table III) led to identification of a unique NeuAc-HexNAc-HexNAc antenna on ds-6 in which the Hex to HexNAc substitution constitutes the sole difference between ds-6 and ds-8. This was supported by the fact that the elution of ds-6 (GU Amide , 6.9) on the am-ide-80 column was 0.3-0.4 GU Amide earlier than that of ds-8 (GU Amide , 7.2) (Fig. S5B), which is attributable to the difference in unit contribution values between Hex and HexNAc (29,44). Likewise, the GU Amide of both ds-6 and ds-8 was reduced by about 0.3-0.4 GU after desialylation, indicative of their ␣2-6 but not ␣2-3 sialylation (45,46). When the desialylated ds-6 was further treated with N-acetyl-␤-D-hexosaminidase, the elution positions on ODS and amide-80 columns (GU ODS , 10.3; GU Amide , 5.6) were indistinguishable from the reference Nglycan B-1 but different from C-1 (GU ODS , 7.7; GU Amide , 5.7). This suggests that desialylated ds-6 released two HexNAcs from the Man␣1-3 branch. Further digestion with ␤-D-galactosidase resulted in a product eluting at the same position as N-glycan B-2 (GU ODS , 9.5; GU Amide , 4.7) but different from C-2 (GU ODS , 7.1; GU Amide , 4.6), thus further confirming that the NeuAc-HexNAc-HexNAc antenna of ds-6 is extending from the Man␣1-3 branch. NeuAc-HexNAc-HexNAc unit is most likely corresponding to NeuAc␣2-6GalNAc␤1-4GlcNAc, although no further evidence can be obtained at present because of lack of the sufficient sample quantity.
PA-derivatized ␣-Galactosylated Neutral N-Glycans-The ODS-fractionated neutral complex-type structures (eluted over 20 -70 min), identified by MS and MS/MS, are listed in Table  S1. Each of the fractions (Fig. 8B) was first screened by MALDI-TOF MS, and the assigned composition was further confirmed by subsequent analysis of the permethylated derivatives. MALDI MS/MS was performed where the sample quan- tity permits, and the observed fragment ions collectively established the permutation of galactosylated antennae. In general, the fragmentation is similar to the underivatized samples (Fig.  7), except that each of the y fragment ions carrying the reducing terminal HexNAc would have a mass increment of 92 units due to the PA tag. For clarity, only the characteristic fragment ions essential for structural assignment are listed in Table S1. As illustrated in Fig. S6, the important features are as follows.
1) Core fucosylation was assigned if a prominent ion at m/z 566 was detected, corresponding to (OH)(Fuc)GlcNAc-PA. This is further supported by b ions resulting from cleavage at the chitobiose core.
2) Sodiated b ions at m/z 894, 690, and/or 486 identified the presence of Hex 3 HexNAc-, Hex 2 HexNAc-, and/or Hex 1 HexNAcantennae, respectively. This was corroborated by sequential loss of the antennary units from the parent ions, giving rise to a series of y ions. The presence of a bisecting GlcNAc was indicated by detection of a facile loss of terminal HexNAc from the parent and other y ions.
3) Multiple cleavages often resulted in a pair of characteristic b ions corresponding to the trimannosyl core, and the number of free OH groups contained can be used to infer the number of antennae attached originally.
Conclusions that can be drawn from more detailed MS/MS analysis of the individually purified components are in general consistent with the analyses of the total mixtures (Fig. 4A), except that ODS fractionation of the PA-derivatives separated isomeric series and also led to detection of a more complete range of minor components.
PA-derivatized ␣-Galactosylated Monosialylated N-Glycans-Although only a single disialylated biantennary structure was prominent, the presence of other minor sialylated structures was noted from MS analysis of the total mixtures (Fig. 4A). Indeed, after PA derivatization and fractionation on the ODS column, about 60 peaks could be detected for the monosialylated fraction. A select few of the monosialylated components were screened by MALDI-MS before and after permethylation to define their molecular compositions and, wherever possible, analyzed by MALDI MS/MS for structural assignment (Table S2). MS/MS analysis of the major monosialylated glycans indicated that all sialylated antennae are NeuAc-Hex-HexNAc with no additional Gal inserted ( Fig. S8 and Table S2 and Fig. 10). However, the other antennae can be Hex 2 HexNAc or Hex 3 HexNAc, likely to be identical to the ␣-galactosylated antenna found in the neutral glycans. It thus appears that the predominant bisected bi-and triantennary complex-type structures with core ␣6-fucosylation are highly heterogeneous due primarily to a combination of different capping moieties on the Gal␤1-4GlcNAc antennae, i.e. Gal␣1-4 -, Gal␣1-4Gal␤1-, and NeuAc␣2-6 -. In the case of Gal␣1-4Gal␤1-4Gal␤1-4GlcNAc termini, an additional ␤-Gal may be added to the LacNAc unit without the concomitant ␣-Gal capping. Although we have not analyzed all of the isomeric structures resolved by HPLC, the general conclusion may apply to the overall glycosylation heterogeneity of pigeon serum IgG. DISCUSSION The diversity of carbohydrate structures is manifest in nature at many different levels, but expression of different oligosaccharide structures among different species is an important subject of inquiry in glycobiology. Highly complex oligosaccharides are speculated to be generated by necessity for survival among hosts and microbes through the symbiotic-commensalpathogenic continuum (47,48). However, only a few case of species-specific oligosaccharide moiety had been investigated systematically. One of the better known examples is the Gal␣1-3Gal epitope, which is widely distributed in all mammals except human, apes, and Old World monkeys. This moiety has been reported to be absent from fish, amphibian, reptile, and bird fibroblasts but present on both cell surfaces and secreted glycoproteins of the noncatarrhine mammals (31,32,49). N-Glycans possessing Gal␣1-3Gal, found in mammalian glycoproteins, have some structural analogy to those possessing Gal␣1-4Gal, found in PEW glycoproteins (1), because both ␣3-Gal and ␣4-Gal are located on non-reducing termini of Lac-FIG. 8. Separation of PA-oligosaccharides from pigeon serum IgG by HPLC. A, total PA-oligosaccharides from pigeon IgG were separated into neutral, mono-, and disialyl oligosaccharides on the DEAE column. B, elution profiles of the neutral, mono-, and disialyl PA-oligosaccharides on the ODS column. Each peak was analyzed with mass spectrometry. Fraction numbers for N-glycans correspond to those in Tables II and III and Supplemental Material Tables S1 and S2. NAc. With the example of Gal␣1-3Gal in mind, we first examined the presence of Gal␣1-4Gal in pigeon serum, lymphocyte, and liver and then isolated one of the major serum glycoproteins possessing Gal␣1-4Gal, which was identified as IgG.
Moreover, analysis of oligosaccharide structures of pigeon IgG unveiled several unique glycosylation properties not found in other avian IgGs or in mammalian IgGs. First, little or no ␣-galactosylated oligosaccharides have been detected in normal serum IgG of non-primate mammals such as dog, cow, sheep, horse, rat, mouse, rabbit, and cat (42,50), even though these animals have functional ␣-1,3-galactosyltransferases (␣-1,3-GalT). A well conserved N-glycosylation site in mammalian IgG is located at Asn-297 (Eu numbering) on CH2 domains (51), and oligosaccharide chains can be seen occluding the cavity at the center of the Fc (52). The N-glycosylation at Asn-297 with biantennary complex-type oligosaccharides is indispensable, because it influences mammalian IgG in its thermal stability, interaction with Fc receptors, association with C1q, and induction of antigen-dependent cellular cytotoxicity (53)(54)(55). It is presumed that N-glycans in the cavity of Fc region cannot possess Gal␣1-3Gal termini, due to steric hindrance (56). If this is true, then the highly ␣-galactosylated features of pigeon serum IgG require a different explanation. The rationalization came from analysis of oligosaccharides in the pigeon IgG-CH3 domain that is structurally equivalent to the mammalian IgG-CH2 domain (36,39). The result indicated that no complex-type N-glycans were located in pigeon IgG-CH3 glycopeptides (Fig. 4C).
Second, about one-third of the total N-glycans from pigeon serum IgG was high mannose-type oligosaccharides, whereas normal mammalian serum IgG contain exclusively biantennary complex-type N-glycans (42). It should be pointed out that 61.7% of the high mannose-type glycans in pigeon IgG are mono-glucosylated as found in chicken (serum, egg yolk) (41) and quail (egg yolk) IgGs (43). Specifically, we demonstrated that pigeon IgG-CH3 glycopeptides contain only high mannosetype oligosaccharides. Because the cognate site specificity was also found in chicken serum IgG, 3 this site-specific N-glycosylation pattern might have resulted from the steric hindrance imposed by the unique conformational structures of avian IgG. The steric hindrance on avian IgG-CH3 may be more severe than that on mammalian IgG-CH2 and allow only limited process on the N-glycans.
Because N-glycans located on CH3 domain are exclusively high mannose-type, complex-type N-glycans were consequently located at other glycosylation sites on avian IgG. Interestingly, structural analysis of N-glycans revealed that the complex-type 3 N. Suzuki and Y. C. Lee, unpublished data.

TABLE II
Structures of high mannose-type PA-oligosaccharides from pigeon IgG based on HPLC and MS analysis a Relative quantity was calculated based on CLC-ODS elution profiles. b Mass/charge (m/z) of neutral oligosaccharides (non-permethylated) were detected as [M ϩ Na] ϩ . c PA-derivatized reference N-glycans whose elution positions on both CLC-ODS and Amide-80 columns and [M ϩ Na] ϩ coincide with the pigeon IgG PA-oligosaccharides were indicated on the table. Structures of the reference compounds were shown in Fig. 9. d Monosaccharides were denoted as follows: M, mannose; Glc, glucose; GN, N-acetylglucosamine.
Currently, no clear function has been assigned to Gal␣/␤1-4Gal in animal glycoproteins. The ␣-galactosyl residues are expressed in some species of birds (e.g. pigeon and swiftlet) but absent in chicken even though they belong to the same Class Aves, suggesting that such an epitope is apparently not uniformly required by all animals. Nevertheless, glycoproteins with Gal␣/␤1-4Gal are widely distributed or circulating in the body of the pigeon. The enigmatic presence of Gal␣1-4Gal, Gal␤1-4Gal, and Gal␣1-4Gal␤1-4Gal␤ may have some important function such as protection against specific carbohydrate-binding proteins expressed on infectious microbes. Alter-natively, because Gal␣1-4Gal, Gal␤1-4Gal, or Gal␣1-4Gal␤1-4Gal on non-reducing termini are strongly antigenic in animals that do not express the antigens by themselves, they may work for species-specific barrier against certain viral infections, as speculated for Gal␣1-3Gal (47,62,63). How these Gal␣/␤1-4Gal structures were acquired or lost during the course of diversification of birds can be investigated through structural analysis of the genes encoding Gal:␤-1,4-GalT and ␣-1,4-GalT.
Finally, it should be mentioned that biomolecules produced by birds inhabiting close to human habitat influence our lives. For example, pigeon fanciers' lung (PFL), one of extrinsic allergic alveolitis, is caused by the repeated inhalation of pigeon droppings and feather bloom (64). The pathogenesis of PFL is  Table II. b PA-derivatized reference N-glycans whose elution positions on both CLC-ODS and amide-80 columns and [M ϩ Na] ϩ coincide with the pigeon IgG PA-oligosaccharides were indicated on the table. Structures of the reference compounds are shown in Fig. 9.
c Key fragment ions detected in MALDI MS/MS analysis of the selected parents, the production mechanism of which is shown in Fig. S5A. In addition to the key fragment ions illustrated in Fig. S6, key y ions resulting from sequential loss of one or more of the antennae were also illustrated. As in Fig. 7, the facile loss of the bisecting GlcNAc is indicated by an arrow from the primary cleavage ions. The location of the different antennae on either ␣3or ␣6-arm of the trimannosyl core was not defined. Symbols used are as in Fig. 7 and S6. related to hypersensitivity reactions to pigeon antigens, one of which had been identified as pigeon intestinal mucin oligosaccharides (65). Although N-glycans on pigeon IgG has not been determined as key antigens for PFL, they deserve future scrutiny. Wild pigeons can be carriers of pathogenic bacteria. Indeed Shiga toxin-producing Escherichia coli strains had been detected in fecal samples from some pigeons (66). Because Gal␣1-4Gal is known to be a minimum structure recognized by several bacterial adhesins and enterotoxins including Shiga toxin (1), it may be worthwhile to assess the relation of Gal␣1-4Gal production in wild birds and their potentials as carriers of pathogenic bacteria.