Secretory IgA N- and O-glycans provide a link between the innate and adaptive immune systems.

Secretory IgA (SIgA) is a multi-polypeptide complex consisting of a secretory component (SC) covalently attached to dimeric IgA containing one joining (J) chain. We present the analysis of both the N- and O-glycans on the individual peptides from this complex. Based on these data, we have constructed a molecular model of SIgA1 with all its glycans, in which the Fab arms form a T shape and the SC is wrapped around the heavy chains. The O-glycan regions on the heavy (H) chains and the SC N-glycans have adhesin-binding glycan epitopes including galactose-linked beta1-4 and beta1-3 to GlcNAc, fucose-linked alpha1-3 and alpha1-4 to GlcNAc and alpha1-2 to galactose, and alpha2-3 and alpha2-6-linked sialic acids. These glycan epitopes provide SIgA with further bacteria-binding sites in addition to the four Fab-binding sites, thus enabling SIgA to participate in both innate and adaptive immunity. We also show that the N-glycans on the H chains of both SIgA1 and SIgA2 present terminal GlcNAc and mannose residues that are normally masked by SC, but that can be unmasked and recognized by mannose-binding lectin, by disrupting the SC-H chain noncovalent interactions.

Secretory IgA (SIgA) 1 is the major immunoglobulin responsible for protecting the mucosal surfaces against invasion by pathogens. In humans, mucosa covers a vast surface area (ϳ400 m 3 ), and the body produces more SIgA each day than all other antibodies combined (66 mg kg Ϫ1 day Ϫ1 ) (1). SIgA occurs mainly as a dimer in which the two IgA molecules are joined together via a small (16 kDa) J chain (joining chain) (2), which is linked to the terminal cysteine of one heavy (H) chain on each IgA. The dimeric IgA with attached J chain is produced in plasma cells close to the epithelium. The epithelial cells express the polymeric immunoglobulin receptor (pIgR) that binds to dimeric IgA; this complex is then translocated across the epithelial cell. During translocation disulfide bonding occurs between the pIgR and one H chain. On reaching the mucosal surface, the (50 -90-kDa) secretory component (SC) is cleaved from the pIgR transmembrane tail, and the whole IgA/J chain/SC (SIgA) complex is secreted (3). Thus, SIgA is a multipolypeptide complex originating from two cell types (3). This is in contrast to serum IgA, which is predominantly monomeric and lacks the J chain and SC.
There are two isotypic forms of IgA: IgA1 and IgA2. Both forms contain two conserved N-glycan sites per H chain, one at Asn 263 in the C␣2 domain and one on the terminal amino acid (Asn 459 ) of its 18-amino acid tail piece (compared with IgG, IgA has an extra 18 amino acids on the C-terminal of the H chain). There are two allotypes of IgA2, IgA2m(1) and IgA2m (2), that contain further conserved N-glycan sites: one on the C␣2 domain and one or two on the C␣1 domain respectively (1) (Fig.  1). SC is highly glycosylated, it has seven N-glycan sites, and sugars contribute up to 25% of its molecular mass, whereas the J chain has only one N-glycan site (3). In addition, IgA1 has a 23-amino acid, proline-rich hinge region with nine potential O-glycosylation sites (serine and threonine residues) of which three to five sites have been shown to be occupied in serum IgA1 (4 -7). IgA2 lacks this 13-amino acid hinge region and is not O-glycosylated.
A newborn child relies on passive immunity from the SIgA in its mother's milk until its own immune system has matured. The SIgA in colostrum and milk binds to microorganisms, their metabolic products and toxins, preventing their attachment to the gut epithelium and facilitating their expulsion in the feces, a process known as immune exclusion (8). The adhesion of many pathogenic organisms to mucosal membrane cells is mediated by adhesins on their surface. These are lectin-like receptors that can bind to complementary carbohydrate constituents expressed by the host tissues (9). For example, S-fimbriated Escherichia coli, which causes sepsis and meningitis in newborns, can be prevented from binding to epithelial cells, independently of the antigen-binding sites, by sialylated glycans on SIgA that bind to the pathogen (10). Fucose, linked ␣1-2 to galactose as in Lewis b and Lewis y epitopes, on SC N-glycans compete with Helicobacter pylori for binding to gastric receptors (11,12). Free SC also binds to E. coli (13,14) and toxin A from Clostridium difficile (15), and both free and SIgAbound SC interact specifically with a surface protein of Streptococcus pneumoniae (16,17). Type 1-fimbriated E. coli express a mannose-specific lectin that binds to SIgA (13). This has led to an increased interest in the use of orally administered recombinant SIgA for passive immunization against virulent pathogens such as C. difficile and Neisseria meningitis (18,19). However, accurately defining glycan structures is a prerequisite to understanding their binding properties, particularly because proteins made in different cell types are differently glycosylated, and this can have major implications when producing engineered antibodies (20).
In this paper we present the first total N-and O-glycan analysis of each of the different peptide chains (H, J, and SC) from the same sample of normal pooled human SIgA. Sensitive analytical procedures have enabled us to identify minor components of the glycan pool in addition to confirming the major glycan structures found by previous investigators (21)(22)(23)(24)(25)(26)(27)(28). The O-glycans on the H chain and the N-glycans on SC presented a wide range of epitopes for adhesin binding. Over 75% of the N-glycans on the J chain were sialylated, whereas over 66% of the N-glycans on the H chain were truncated complex structures with free terminal GlcNAcs.
On the basis of these data we have constructed a molecular model of SIgA1 with its glycans attached, in which the Fab arms form a T shape and the SC is wrapped around the H chains. This model shows that glycans cover most of the SIgA1 complex with the exception of the Fab regions. Each SIgA1 molecule has several sites for binding to pathogens. In addition to the four Fab antigen-binding sites (adaptive immunity), there are two O-glycosylated regions containing up to 10 glycans per region and SC with seven N-glycans, which present a wide range of sugar epitopes capable of binding to adhesins (innate immunity).
Our model of SIgA1 shows the SC wrapped around the H chains, masking the H chain N-glycans, which are truncated complex structures with free terminal GlcNAc residues. These GlcNAc residues are potential ligands for lectins such as mannose-binding lectin (MBL). MBL is a calcium-dependent serum lectin that is able to bind to D-mannose, L-fucose, GlcNAc, and N-acetylmannoseamine, recognizing two common equatorial hydroxyl groups (29). Binding of MBL to microorganisms, including viruses, bacteria, and yeast species, can induce activation of the complement system via the lectin pathway, thus leading to target opsonization. MBL represents a key component of the innate immune system, as illustrated by the increased susceptibility to infections occurring in MBL-deficient individuals (30). Recently, binding of the lectin domain of MBL to polymeric serum IgA has been reported (31). We show that MBL does not bind to SIgA at neutral pH, supporting the proposal that SC is masking the H chain N glycans, and that disruption the SC-IgA noncovalent interactions, by preincubation at pH 3, unmasks the H chain N-glycans, allowing MBL to bind.

EXPERIMENTAL PROCEDURES
Secretory IgA-Pooled human SIgA, purified from colostrum, was obtained from Sigma. The secretory IgA complex was reduced, alkylated, and then separated into SC, J chain, H chain, and light chain by reducing SDS-PAGE (80 ϫ 80 ϫ 1 mm, 10% BisTris NuPAGE gel, MES SDS running buffer (Invitrogen)). The protein bands were visualized by Coomassie staining. MultiMark molecular mass standards were used (Invitrogen). Relative amounts of IgA1 and IgA2 were determined by ELISA according to Ref. 32 using normal human serum with known concentrations of IgA1 (2.1 mg/ml) and IgA2 (0.2 mg/ml) as standards.
Identification of Gel Bands by Mass Spectrometry-Coomassiestained bands were excised and in-gel digested with trypsin (sequencing grade; Roche Applied Science) as described in Ref. 33. The mixtures of recovered tryptic peptides were desalted by loading onto a PepMap C18 0.3 ϫ 5 mm cartridge (LC packings; Presearch Ltd., Hitchin, UK) in water with 0.1% formic acid and then eluting with 80% acetonitrile, 0.1% formic acid at a flow of 0.2 l min Ϫ1 directly into a hybrid quadrupole time-of-flight mass spectrometer fitted with a nanospray source (Waters-Micromass Ltd., Manchester, UK). The peptides were sequenced from fragmentation data as described (34). A BLAST search of the NCBI data base was performed using MASCOT software.
Release and Fluorescent Labeling of Glycans-N-Glycans were released from excised gel bands by in-gel digestion of the protein with peptide: N-glycosidase F (Roche Applied Science) (35). Manual hydrazinolysis (36,37) was used to release O-glycans (60°C for 6 h) from the whole SIgA complex. Released glycans were fluorescently labeled with 2-aminobenzamide (2AB) by reductive amination according to the method of Bigge et al. (38) using an Oxford GlycoSciences Signal TM labeling kit (Oxford GlycoSciences, Abingdon, UK).
Analysis of Glycans by High Performance Liquid Chromatography (HPLC)-Normal phase (NP) HPLC was performed according to the low salt buffer system as previously described (39) using a 4.6 ϫ 250-mm GlycoSep-N column (Oxford GlycoSciences). The system was calibrated using an external standard of hydrolyzed and 2AB-labeled glucose oligomers to create a dextran ladder. Weak anion exchange HPLC (40) was performed using a Vydac 301VHP575 7.5 ϫ 50-mm column (Anachem Ltd., Luton, Bedfordshire, UK) according to the modified methodology (41). These HPLC methods are described in detail in Ref. 37.
Exoglycosidase Digestions-Arrays of exoglycosidases were used in combination with HPLC to determine the sequence, monosaccharide type, and linkage of sugar residues as described in Ref. 37   Mass Spectrometry of Glycans-Positive ion matrix-assisted laser desorption-ionization time-of-flight mass spectra were recorded with a Micromass TofSpec 2E reflectron time-of-flight mass spectrometer (Waters-Micromass) using a saturated solution of 2,5-dihydroxybenzoic acid in acetonitrile as the matrix, as described in Refs. 37 and 42. LC-ESI-MS and MS/MS spectra were recorded from a Waters CapLC interfaced with a hybrid quadrupole time-of-flight mass spectrometer fitted with a Z-spray electrospray ion source (Waters-Micromass) and operated in positive ion mode. A 1 ϫ 150-mm microbore NP HPLC column was packed with stationary phase material from a GlycoSep N column (Oxford GlycoSciences). The same solvents and gradient were used as for standard NP HPLC but with a flow rate of 40 l/min (37).
MBL Binding Experiments-MBL and polymeric serum IgA were purified from human donor plasma exactly as described previously (31). Binding of MBL to immobilized IgA was performed by ELISA, as described (31). In brief, IgA was coated on ELISA plates using a carbonate buffer (pH 9.6). As a negative control, the plates were coated with purified human serum albumin (Central Laboratory of Bloodtransfusion (Sanquin), Amsterdam, the Netherlands). After each step, the plates were washed with phosphate-buffered saline containing 0.05% Tween 20. MBL was diluted in BVBϩϩ (1.8 mM sodium 5,5-diethylbarbital, 0.2 mM 5,5-diethylbarbituric acid, 145 mM NaCl, 0.5 mM MgCl 2 , 1 mM CaCl 2 , 0.05% Tween 20, 1% bovine serum albumin, pH 7.5) and incubated in the plates for 1 h at 37°C. In some experiments, coated wells were pretreated with buffers of different pH (0.1 M glycine-HCl, pH 2.0 -5.0, containing 0.15 M NaCl) for 5-60 min at 37°C, followed by washing with phosphate-buffered saline/Tween and the addition of MBL as indicated above. Furthermore, inhibition experiments were performed using 10 mM EDTA or 50 mM D-mannose, which were preincubated with MBL for 40 min at room temperature before the addition of MBL to the plate. Binding of MBL was examined using monoclonal antibody 3E7 directed against MBL (mouse IgG1, kindly provided by Dr. T. Fujita, Fukushima, Japan), conjugated to digoxygenin (Roche Applied Science), followed by horseradish peroxidase-conjugated rabbit anti-digoxygenin antibodies (Fab fragments; Roche Applied Science). Enzyme activity of horseradish peroxidase was detected using 2,2Јazino-bis(3-ethylbenzthiazoline-6-sulfonic acid) (from Sigma), and the A values at 415 nm were measured.
Molecular Modeling-Molecular modeling was performed on a Silicon Graphics Fuel work station using InsightII and Discover software (Accelrys, San Diego, CA). The figures were produced using the program Molscript (43). Crystal structures used as the basis for modeling were obtained from the Brookhaven data base (44). N-and O-glycan structures were generated using the data base of glycosidic linkage conformations (45) and in vacuo energy minimization to relieve unfavorable steric interactions. The Asn-GlcNAc linkage conformations were based on the observed range of crystallographic values (46), the torsion angles around the Asn C␣-C␤ and C␤-C␥ bonds then being adjusted to eliminate unfavorable steric interactions between the glycans and the protein surface.

Determination of the Relative Amounts of IgA1 and IgA2 in
SIgA-The percentages of IgA1 (39%) and IgA2 (61%) in the total SIgA from pooled human colostrum were determined by ELISA (data not shown).

Separation and Identification of SIgA Component Proteins-
Pooled normal human SIgA from colostrum was resolved into SC, H, L, and J chains by SDS-PAGE (Fig. 2). Tryptic peptides were sequenced to confirm the identity of the proteins in the gel bands ( Table I). The J chain migrated with a higher apparent molecular mass than 16 kDa, in agreement with results by Chuang and Morrison (47).
N-Glycan Analysis- Fig. 2 shows the NP HPLC profiles of the pools of glycans released from the different gel bands. The glucose unit (GU) scale indicates the relative size of the glycans (the larger the GU, the larger the glycan). The N-glycans from the H chains ranged between GU 5 and 7; SC had the largest glycans (GU 6 -12), whereas the J chain glycans were GU 6 -9.
Because there is only one glycosylation site per J chain and one J chain per SIgA, the abundance of glycans is lower; the light chain had negligible levels of N-glycosylation.
The preliminary assignment of structures was made by comparing GU values with standards and confirmed by following the elution positions (measured in GU) of peaks through the different exoglycosidase arrays and by mass spectrometry (37,48). In Tables II-VI, the relative amounts of glycans in the pool released from each peptide can be found in the column headed "Undig." To detect the presence of a bisecting GlcNAc residue, both Jack bean ␤-N-acetylhexosaminidase and S. pneumonia ␤-N-acetylhexosaminidase digestions were performed. Jack bean ␤-N-acetylhexosaminidase does not digest glycans with bisecting GlcNAc, whereas S. pneumonia ␤-N-acetylhexosaminidase does digest bisected biantennary glycans but is inefficient at digesting tri-or tetra-antennary structures with GlcNAc ␤1-4 or 1-6 linked to mannose under the conditions used in this paper (data not shown). The presence of bisecting GlcNAc was also confirmed by tandem mass spectrometry (MS/MS).
H Chain N-Glycans-Over 75% of the H chain N-glycans contain a bisecting GlcNAc, 66% contain a free terminal Glc-NAc on one antenna, less than 20% of structures were fully galactosylated, less than 15% were sialylated (only ␣2-6 sialic acids detected), no glycans contained outer arm fucose residues, all galactose residues were ␤1-4-linked, about half the structures contained core fucose, and there were about 12% of oligomannose structures (Figs. 3 and 4 and Table II). The major structures were FcA2B (30%), A2B (21%), and FcA2BG1 (8%) (the notation is explained in footnote 1 of Table II). The glycans were of the biantennary complex type with a bisecting GlcNAc residue rather than triantennary, because they did not digest beyond A2B (GU 5.8) with Jack bean ␤-N-acetylhexosaminidase but were digested with S. pneumonia ␤-N-acetylhexosaminidase, and during LC-ESI-MS/MS, fragments were found that had lost the two GlcNAc-Man antennas but not the third bisecting GlcNAc (data not shown). Fig. 5 shows the profiles of N-glycans from the H chains of SIgA1 and SIgA2, purified from colostrum from a single human donor. The N-glycan profiles from SIgA1 and SIgA2 H chains were almost identical, even though SIgA2 has two or three more N-glycan sites, depending on the isotype, than SIgA1.
J Chain N-Glycans-Over 75% of the structures at the single N-glycan site on the J chain were sialylated (Figs. 3 and 6 and  Table III). Five major structures were detected in approximately equal proportions: biantennary complex structures FcA2G2S2, A2G2S2, FcA2G2S, A2G2S, and the hybrid structure FcMan4A1G1S. The di-sialylated biantennary structures contained one sialic acid ␣2-3-linked and one ␣2-6-linked (because all of the di-sialylated structures were digested to the mono-sialylated forms by Newcastle disease virus, which is specific for ␣2-3 sialic acids). The mono-sialylated structures contained either ␣2-3or ␣2-6-linked sialic acid. Bisecting GlcNAc was found in about half of the neutral structures but not in the sialylated structures. About half of all of the structures (neutral or charged) were core-fucosylated. No outer arm fucosylation was seen, and all galactose residues were ␤1-4-linked. Secretory Component N-Glycans-A much greater range of structures was found on SC than on the H or J chain (Figs. 3, 7, and 8 and Table IV). The majority of structures were fully galactosylated, nonbisected biantennary structures. Triantennary (11.7%) structures were also present, and a small (Ͻ1%) amount of tetraantennary (or polylactosamine) structures were detected by matrix-assisted laser desorption-ionization-MS. Over 70% of the glycans were sialylated, and the majority were mono-sialylated. The glycan pool was separated into neutral and mono-and di-sialylated fractions by weak anion exchange, and then these fractions were run again on NP HPLC (Fig. 8). The two di-sialylated peaks digested to A2G2 and FcA2G2 with A. ureafaciens sialidase, and, when subjected to Newcastle disease virus, about 20% lost one sialic acid and therefore originally contained one ␣2-3 and one ␣2-6 sialic acid; the remaining 80% had two ␣2-6 sialic acids. The mono-sialylated fraction consisted of a mixture of glycans carrying either ␣2-3 or ␣2-6 sialic acid. Over 65% of structures contained core fucose. Digestion with glycosidases of differing specificity (bovine testes ␤-galactosidase and S. pneumoniae ␤-galactosidase) gave different results, indicating that galactose was linked both ␤1-4 and ␤1-3 to GlcNAc. Digestion with almond meal ␣-fucosidase showed the presence of outer arm fucoses linked both ␤1-3 and ␤1-4 to GlcNAc (because fucose was removed from GlcNAc, which also had galactose either ␤1-3 and ␤1-4 linked, the fucose was assumed to be in the unoccupied linkage position). After digestion with all of the exoglycosidases listed in Table IV, except for bovine kidney fucosidase, 1.8% of the glycan pool was left at GU 7.2. This material digested with bovine kidney fucosidase, indicating the presence of fucose ␣1-2 linked to a galactose. LC-ESI-MS/MS analysis of the major mass peaks gave results consistent with the majority of structures with outer arm fucosylation and sialylation having fucose on one arm and sialic acid on the other. There was, however, some fucosylation and sialylation on the same arm (i.e. sialylated Lewis epitopes) because the fragment Hex-HexNAc-Fuc-Neu5Ac (m/z 803.2) was detected from fragmentation of FcA2G2FS (Hex 5 -HexNAc 4 -Fuc 2 -Neu5Ac-2AB) (data not shown). Notably there were a number of unusual structures identified with outer arm fucosylation but no core fucose. Thus, there is a wide range of different epitopes present on the SC N-glycans, including all of the different Lewis and sialylated Lewis epitopes.
O-Glycosylation- Fig. 9 shows the NP HPLC profile from the undigested glycan pool followed by a range of whole pool digestions. Over 50 peaks, from glycans ranging in size from two-(GU 1.8) to fifteen-monosaccharide residues (GU 12), were found, and many of the peaks contained more than one structure. We used a combination of different HPLC and MS strategies to identify these structures as described in Ref. 37. The HPLC profiles were very complex; therefore individual peaks were collected, digested separately, and then run on NP and reverse phase HPLC. Table V gives the results of a selection of digestions on the whole pool of O-glycans, with 33 structures fully characterized in the whole O-glycan pool. The major structures (present at over 3% and labeled in the top panel of Fig. 9)   Table II for abbreviations and footnotes. 2 Fucose linked ␣1-2 to Gal. 3 Composition Hex 7 HexNAc 6 ϩ 0,2,3, and 4 Fuc also detected as minor components by matrix-assisted laser desorption-ionization. These could be tetrantennary or polylactosamine structures.
Molecular Model of SIgA1-The molecular model (Fig. 10) was constructed from models of the IgA1 monomer, J chain, and secretory component. The IgA1 monomer model was based on the model of serum IgA1 previously published by us (6), with the orientation of the Fab domains adjusted to fit with the recent low angle x-ray and neutron scattering data obtained on serum IgA1 (49). It was also necessary to modify the conformations of the C-terminal tail pieces on building SIgA1 (see below). The glycan structures were also modified according to the glycan sequence data for SIgA presented here, with the most  Ta-bles V and VI for glycan structures)). No Fab glycosylation was found. The J chain was modeled based on the three-dimensional fold of superoxide dismutase (1cbl.pdb) (50), as suggested by secondary structure predictions and circular dichroism spectroscopy (51). The sequence homology between the J chain and superoxide dismutase is not high, so the model is intended only to indicate the general size and shape of the J chain. An FcA2G2S2 glycan was attached to residue 49 (SWISSPROT numbering, P01591). The SC consists of five  Table II for abbreviations). Table II for percentage areas). Aliquots of the total 2AB-labeled glycan pool were incubated with different exoglycosidases, as shown in each panel. Following digestion, the products were analyzed by NP HPLC. The major peaks have been annotated. The structures were allocated by their elution position measured in GU before and after digestion with exoglycosidases and reference to known GU values (37,39).

FIG. 5. NP HPLC profiles of N-glycans from H chains of SIgA1
and SIgA2. Purified total SIgA from colostrum from a single human donor was separated into SIgA1 and SIgA2 by Jacalin-agarose affinity chromatography (SIgA1 binds). ELISA tests were done to check the purity of the samples, the IgA 1 was 98% IgA1, and the IgA2 was 97% IgA2 (ELISA plates were coated with SIgA samples at a concentration of 5 g/ml (250 ng/well) in carbonate buffer pH 9.6; monoclonal antihuman IgA1-and IgA2-peroxidase conjugates (Nordic Immunological Labs, Tilburg, The Netherlands) were used at 1:1000 dilution; incubation with primary and secondary antibody was 1 h; 12 wells were run per sample). The samples were reduced, alkylated, and then run on a 10% BisTris gel. The N-glycans were released from the H chains by in-gel N-glycosidase F digestion, 2AB-labeled, and run on NP HPLC.
Ig-like domains connected by variable length peptides (52). Each individual domain was modeled separately, based on the crystal structure of the Ig-like domain with the highest sequence homology. The sequence homologies were generally in the region of 25-30% identity and 40% homology (using BLAST), so again the model provides only the overall size and shape of the domains. No attempt was made to model the linkers between the domains, except to put a limit on the possible distances between the domains. Glycans were added as follows: A2G2S2 glycans on residues 83, 186, and 499; an FcA2G2S2 glycan on residue 90; and FcA2G2S glycans on residues 135 and 469 (SWISSPROT numbering, P01833).
Dimeric SIgA1 was modeled from two IgA1 monomers. The relative positions of the IgA1 monomers were based on electron microscopy studies (53) showing that the dimer is linear with the four Fab domains in roughly the same plane and a separation between the two hinge regions of about 125 Å. This places the two Fc domains very close together, with insufficient room between them to accommodate the J chain. The J chain was, therefore, positioned asymmetrically to one side of the dimer with the glycan pointing away from the IgA1 monomers. The tail pieces were then modeled as random coil peptides, with the requirements that the C-terminal Cys on one tail of each monomer has to form a disulfide bond with the J chain and that the C-terminal Cys residues on the other tails have to bind to each other (54). The positioning of the SC is much less certain. Cys 467 on domain V of the secretory component forms a disul-fide bond with Cys 311 on the C␣2 domain of one of the IgA1 monomers (55), which fixes its position. In addition, residues on domain I of the SC have been shown to interact with the J chain (56) and one of the C␣3 domains (57,58). This has been modeled as the secretory component wrapping around the IgA1/J chain complex, so that domain V is bonded to one IgA1 monomer and domain I interacts with the other monomer.
MBL Binding Assays-Although the SIgA H chains contain terminal sugars known to be ligands for MBL, no binding of MBL to SIgA-coated microtiter plates was detected (Fig. 11A). In contrast a sample of polymeric serum IgA, with known MBL affinity, did bind MBL (31). This suggests that the SC in the SIgA complex may be preventing access of MBL to these H chain N-glycans. To determine whether pH changes could disrupt the association of SC with SIgA-H chains and thereby expose the GlcNAc and mannose bearing N-glycans on the H chains, the SIgA-coated plates were pretreated for 60 min with buffers of different pH. Preincubation at pH 2.0 or 3.0 resulted in MBL binding to SIgA but not to albumin (Fig. 11B). A time course of 5-60 min of preincubation with a pH 3.0 buffer showed increasing MBL binding to SIgA with time (Fig. 11C), indicating that the disruption of the noncovalent interactions between SC and IgA-H chains progressed over this period. Fig.  11D shows that this MBL binding to SIgA was a C-type lectinsugar interaction because preincubation of MBL with either EDTA or D-mannose prevented MBL-SIgA binding. Aliquots of the total 2AB-labeled glycan pool were incubated with different exoglycosidases, as shown in each panel. Following digestion, the products were analyzed by NP HPLC. The major peaks have been annotated. The structures were allocated by their elution position measured in GU before and after digestion with exoglycosidases and reference to known GU values (37,39). For example, the A2G2S peak (GU 7.90) moves to GU 7.20 (A2G2) after digestion with sialidase (A. ureafaciens sialidase (Abs)), whereas further digestion with galactosidase (bovine testes ␤-galactosidase (Btg)) moves the peak to GU 5.47 (A2), and finally digestion with S. pneumonia ␤-N-acetylhexoseaminidase moves the peak to GU 4.40 (Man3) (see Table III for percentage  areas and Table II for abbreviations). Table IV for percentage areas).

FIG. 7. NP HPLC profiles of N-glycans from SC following arrays of exoglycosidase digestions (see
whereas the majority of the N-glycans on the H chains of both SIgA1 and SIgA2 carry terminal GlcNAc residues. We show that these GlcNAc residues are masked from binding to MBL in native SIgA by SC but that they can be unmasked, allowing interaction with MBL by disrupting the SC-H chain noncovalent interactions by incubation at low pH. We hypothesize that unmasking of the H chain N-glycans may take place in vivo upon the interaction of SC with bacterial adhesins or cellular lectins. Subsequent lectin binding to the exposed GlcNAc residues would promote opsonization by complement and/or direct phagocytosis, leading to presentation of pathogen to the adaptive immune system.

N-Glycans of SC Present Many Epitopes for Bacterial Adhesin and Lectin
Binding-Our results show that SC presents a wide range of glycan structures, including all of the different Lewis and sialyl-Lewis epitopes that can potentially bind lectins and bacterial adhesins. We found galactose linked both ␤1-4 and ␤1-3 to GlcNAc; fucose linked ␣1-3 and ␣1-4 to GlcNAc and ␣1-2 to galactose, as well as both ␣2-3and ␣2-6-linked sialic acids. This is a larger range of epitopes than reported by previous workers (21, 24 -26). As well as having a general role of protecting both SC and the SIgA from proteases (59), these glycans on SC can specifically interact with adhesins and lectins. SC has been shown to bind to a range of bacteria via its glycans (e.g. H. pyroli (11,12), E. coli (10,13,14), C. difficile toxin A (15), and S. pneumoniae (16,17)), thereby inhibiting attachment and the subsequent infection of epithelial surfaces. The SC glycans are also involved in the localization of SIgA by anchoring the SIgA to mucus lining the epithelial surface through its carbohydrate residues. Phalipon et al. (60) showed that glycosylated SC was required to locate the SIgA to specific areas of the epithelium and that antibody specificity was required for binding to Shigella flexneri. To protect the mice from S. flexneri infection, the whole glycosylated SC-IgA complex was required. Burns et al. (61) also found that SC was needed with IgA antibodies to protect mice from rotavirus, because the IgA was only effective if given systemically, not when presented at the luminal side of the intestinal tract (without SC), indicating that IgA transcytosis (with the addition of SC) was required for viral inactivation in vivo. A further role for SC glycans is in binding to the lectin-binding domain of Mac-1 (CR3, CD11b/CD18) (62), which is involved in inducing SIgA signaling via the IgA receptor Fc␣R (CD89), leading to respiratory burst, phagocytosis, and cytokine secretion (reviewed in Refs. 63 and 64). Mac-1 has a broad specificity to sugars, including ␤-glucans (65), and has been shown to bind to recombinant SC, either as free SC or in the SIgA complex (62). The physiological roles of serum and SIgA are quite different, and these data show that the presence of the highly glycosylated SC has major effects on the biological functions of SIgA.
J Chain Conformation Requires an N-glycan-The J chain on dimeric IgA is essential for binding to pIgR and translocation across the epithelial cells (2,56). Deletion of the N-glycan site, by substitution of Asn with alanine, prevents IgA dimer formation (54), indicating that the N-glycan plays an important role in this process. We found that over 75% of the structures on this single N-glycan site were sialylated with a mixture of both ␣2-3and ␣2-6-linked sialic acids. Baenziger (28) also reported 85% sialylation (although all ␣2-6-linked sialic acids), and these data suggest that the charge on this glycan is important in maintaining the correct conformation of the J chain for presentation to pIgR.
The  Tables V and  VI for percentage areas). without (ϳ51%) or with one (ϳ15%) galactose and that less than 15% of structures are sialylated. These truncated glycans are very different from the mainly fully galactosylated, sialylated structures that have been found on normal serum IgA (6,66). Serum IgA is produced in plasma cells in the spleen or lymph and then distributed around the whole body. In contrast, SIgA is made in plasma cells local to the site of secretion. The difference in glycosylation between serum and secretory IgA is likely to be a reflection of the different glycosylation machinery operating in these different plasma cells. It has been shown that H chain N-glycans are needed to maintain proper conformation, because selective removal of these N-linked sites by mutagenesis results in degradation and reduced secretion of IgA (67) and markedly reduced dimer assembly (68). The SIgA H chain N-glycans on both SIgA1 and SIgA2 have 66% of glycans terminating in GlcNAc residues and 12% oligomannose structures. There are four N-glycans on SIgA1 and up to 20 on SIgA2; therefore these GlcNAc and mannose residues are multiply presented and may be expected to bind to lectins such as MBL, particularly if the SIgA is aggregated.
MBL Does Not Bind to Native SIgA-When we tested the ability of MBL to bind to SIgA coated onto microtiter plates, no binding was detected. We therefore concluded that the SC is masking the N-glycans on the H chains. SC is covalently bound at one end (domain V) to the C␣2 domain of one of the IgA1 H chains (55), and the other end (domain I) interacts noncovalently with both the J-chain (56) and one of the C␣3 domains on a H chain (57,58). The model in Fig. 10 shows these interactions with the SC wound around the dimeric IgA. The SC with its large N-glycans (shown in orange) thus obstructs the MBL from accessing the H chain N-glycans (shown in yellow). To find out whether disruption of the noncovalent interactions between SC and SIgA-H chains could unmask the underlying H chain N-glycans, the SIgA-coated plates were preincubated with low pH buffers. Indeed, preincubation at pH 3 allowed binding of MBL to SIgA, and longer preincubation times (0 -60 min) directly correlated with increasing binding, suggesting that the H chain N-glycans were being unmasked. We therefore hypothesize that such unmasking may take place upon the interaction of SC glycans with bacterial adhesins or cellular lectins. The binding of lectins to the SC glycans (up to seven glycans) may be stronger than the noncovalent interactions between SC and the IgA H chains and therefore pull one end of the SC away from the H chains. Subsequent MBL binding to This model illustrates that most of the molecule is covered with glycans with the exception of the Fab antibody-binding sites. This leaves the Fab regions clear for interaction with antigens (adaptive immunity), whereas the O-glycans on the hinge region of the H chains and the N-glycans on SC cloak the rest of the surface, free for interaction with bacterial adhesins (innate immunity). In this conformation, with the SC wrapped around the H chains, the N-glycans on the H chain (which bear terminal GlcNAc and mannose residues) are masked from binding with lectins such as MBL.
the exposed GlcNAcs and mannose residues would promote opsonization by complement and/or direct phagocytosis, leading to presentation of pathogen to the adaptive immune system. In addition to the potential interaction with soluble lectins such as MBL, sugars present on SIgA may also bind to lectin receptors present on phagocytic cells. In this respect it has been shown that SIgA, but not serum IgA, could be internalized by dendritic cells via mannose receptor (69), which is an important receptor for antigen uptake, and that this binding and uptake could be blocked by specific sugars (mannose, fucose, GlcNAc, but not galactose) or partially by antibody reactive with mannose receptor. Recognition and presentation of antigen by SIgA may involve modulation of the immune response via its binding to host lectins. In this respect, it is important that uptake of SIgA by dendritic cells, which was shown to be partially mediated by the mannose receptor, did not induce DC maturation (69). No data are currently available about the potential effect of antigen opsonization by MBL on phagocytosis and maturation of dendritic cells. Because a recent study showed that MBL can be produced on the apical side of intestinal epithelium in the mouse (70), an interaction between SIgA and MBL in the gut lumen is quite feasible. In view of the well known roles of both SIgA and MBL in host defense, we hypothesize that such an interaction will contribute to the protection against mucosal infection.
A Wide Range of O-Glycan Epitopes Are Available for Interaction with Bacterial Adhesins-We found over 50 different O-glycan structures, up to 15 sugars in size, of which we have fully sequenced 33 structures (13 of which were previously identified by Pierce-Cretel et al. (22,23)). As with the SC N-glycans, a wide range of structures were identified with many different sugar epitopes including galactose ␤1-4 and ␤1-3 linked; sialic acid ␣2-3 and ␣2-6 linked; and fucose ␣1-4, 3 and 2 linked, showing that these O-glycans (present on IgA1 but not IgA2) also present adhesin-binding sites. The T shape of IgA, revealed by x-ray resolution (49), indicates that these O-glycans would be exposed for interaction with bacterial adhesins. These SIgA O-glycans are much larger and more elaborate structures than the simple sialylated core 1 structures (6, 71) found on serum IgA and can be expected to have other functions in addition to stabilizing the hinge region. Evidence to support the proposal that these O-glycans can interact with bacterial adhesins in a similar way to those found on SC comes from Bos et al. (72). They established eight IgA-producing hybridomas from the mesenteric lymph nodes of mice and showed that these hybridomas recognize different but partially overlapping fecal bacterial populations. They suggested that this overlapping recognition was caused by interactions with conserved regions of the antibodies such as the glycans. The presence of the types of O-glycans we have shown could readily account for these overlapping interactions.
Differences in O-glycan structures between serum and secretory IgA are likely to be due to different glycosylation enzymes in the plasma cells located in the bone marrow/spleen/liver to those at mucosal surfaces. It is also plausible that plasma cells at different mucosal surfaces (e.g. lungs and upper or lower FIG. 11. Binding of MBL to immobilized IgA. SIgA or albumin was coated at 5 g/ml, and binding of purified MBL (diluted in BVBϩϩ) was assessed. A, different concentrations of MBL were added to wells coated with SIgA or human polymeric serum IgA with known MBL binding. Human albumin was used as a negative control. B, coated SIgA or albumin was pretreated for 60 min with buffers of different pH, as indicated. Binding of MBL (1.3 g/ml) was assessed as described under "Experimental Procedures." C, coated SIgA or albumin was pretreated with a pH 3.0 buffer for 5-60 min, as indicated, followed by assessment of MBL binding. D, coated SIgA or albumin was pretreated for 60 min with a pH 3.0 buffer followed by assessment of MBL binding in the presence or absence of EDTA or mannose. The data represent one of at least two similar experiments. The error bars represent S.D. of duplicate measurements. intestine) could produce SIgA glycosylation that is mucosal location-specific, because mucin glycosylation is known to differ with mucosal site (73). Differences in glycosylation of SIgA in different mucosal areas may provide a means of maintaining a balance with normal commensal flora without eliciting a deleterious immune response.
SIgA Participates in Both the Innate Adaptive Immune Systems-The molecular model presented in Fig. 10 summarizes the work presented in this paper. This shows that glycans cover most of the SIgA1 complex with the exception of the Fab regions. When viewed as a whole, each SIgA1 molecule has several sites for binding to pathogens. In addition to the four Fab antigen-binding sites (adaptive immunity), there are two Oglycosylated regions containing up to 10 glycans per region, and SC with seven N-glycans, which present a wide range of sugar epitopes capable of binding to adhesins (innate immunity). Thus, SIgA participates in both the adaptive and innate immune systems.
We have also shown that the N-glycans on the H chains of both SIgA1 and SIgA2 present GlcNAc and mannose residues that can be masked by SC. Exposure of these GlcNAc and mannose residues following disruption of the SC-H chain noncovalent interactions can promote recognition by soluble lectins, such as MBL, or by lectin receptors, such as the mannose receptor on macrophages and dendritic cells. We hypothesize that unmasking of the H chain N-glycans may take place upon the interaction of SC with bacterial adhesins or cellular lectins. Subsequent lectin binding to the exposed GlcNAc and mannose residues would promote opsonization by complement and/or direct phagocytosis, leading to presentation of pathogen to the adaptive immune system.