Identification of O-glycan Structures from Chicken Intestinal Mucins Provides Insight into Campylobactor jejuni Pathogenicity*

The Gram-negative bacteria Campylobactor jejuni is the primary bacteria responsible for food poisoning in industrialized countries, and acute diarrheal illness is a leading cause of mortality among children in developing countries. C. jejuni are commensal in chickens. They are particularly abundant in the caecal crypts, and poultry products are commonly infected as a result of cross-contamination during processing. The interactions between C. jejuni and chicken intestinal tissues as well as the pathogenic molecular mechanisms of colonization in humans are unknown, but identifying these factors could provide potential targets to reduce the incidence of campylobacteriosis. Recently, purified chicken intestinal mucin was shown to attenuate adherence and invasion of C. jejuni in the human colorectal adenocarcinoma cell line HCT-8 in vitro, and this effect was attributed to mucin O-glycosylation. Mucins from different regions of the chicken intestine inhibited C. jejuni binding and internalization differentially, with large intestine>small intestine>caecum. Here, we use LC-MS to perform a detailed structural analysis of O-glycans released from mucins purified from chicken large intestine, small intestine, and caecum. The O-glycans identified were abundantly sulfated compared with the human intestines, and sulfate moieties were present throughout the chicken intestinal tract. Interestingly, alpha 1–2 linked fucose residues, which have a high binding affinity to C. jejuni, were identified in the small and large intestines. Additionally, N-glycolylneuraminic/N-acetylneuraminic acid containing structures present as Sda-like epitopes were identified in large intestine samples but not small intestine or caecum. O-glycan structural characterization of chicken intestinal mucins provides insights into adherence and invasion properties of C. jejuni, and may offer prospective candidate molecules aimed at reducing the incidence of infection.

The Gram-negative bacteria Campylobactor jejuni is the primary bacteria responsible for food poisoning in industrialized countries, and acute diarrheal illness is a leading cause of mortality among children in developing countries. C. jejuni are commensal in chickens. They are particularly abundant in the caecal crypts, and poultry products are commonly infected as a result of crosscontamination during processing. The interactions between C. jejuni and chicken intestinal tissues as well as the pathogenic molecular mechanisms of colonization in humans are unknown, but identifying these factors could provide potential targets to reduce the incidence of campylobacteriosis. Recently, purified chicken intestinal mucin was shown to attenuate adherence and invasion of C. jejuni in the human colorectal adenocarcinoma cell line HCT-8 in vitro, and this effect was attributed to mucin O-glycosylation. Mucins from different regions of the chicken intestine inhibited C. jejuni binding and internalization differentially, with large intestine>small intestine>caecum. Here, we use LC-MS to perform a detailed structural analysis of O-glycans released from mucins purified from chicken large intestine, small intestine, and caecum. The O-glycans identified were abundantly sulfated compared with the human intestines, and sulfate moieties were present throughout the chicken intestinal tract. Interestingly, alpha 1-2 linked fucose residues, which have a high binding affinity to C. jejuni, were identified in the small and large intestines. Campylobactor jejuni infection is widespread in industrialized countries and is the greatest source of food poisoning worldwide. Infection and the resulting diarrhea are the most common cause of death among young children in developing countries, with an estimated 2.1-2.5 million cases per year in the USA. C. jejuni leads to severe gastroenteritis and is also linked to Guillain-Barré and irritable bowel disease (1,2). Contaminated poultry meat is the most common source of infection, with up to 88% of poultry products being infected (3). The bacteria are commensal in chickens and exist throughout their intestinal tract. However, they do not invade the epithelium of the gut. Purified mucin glycoproteins from the caecum and small and large intestine have been shown to exhibit inhibitory properties against the bacteria in vitro (4). The commensal relationship between C. jejuni occurrence in chickens and the inability to invade the intestinal tissues is poorly understood. Enhanced understanding of how the properties of the chicken mucosal barrier differ from humans; as well as changes in mucin composition along the chicken intestinal tract, have the potential to explain the commensal behavior of C. jejuni in this species. Such information could also provide leads for the development of agents to limit infection.
Intestinal mucus in vertebrates is a hydrated gel comprised mainly of very high molecular weight and polymeric secreted mucin glycoproteins that are heavily O-glycosylated. O-linked glycans constitute up to 80% of the mucin by weight and represent an abundant potential carbon source for the resident microflora. They also present targets for bacterial adhesion and chemotaxis that may be exploited by both commensal and pathogenic organisms. Intestinal mucus is a dynamic barrier layer that is mainly secreted by goblet cells. It forms a supramucosal layer to protect the gastrointestinal epithelial cells against infection. A density gradient of viscoelastic and highly hydrated mucins polymers exists from the inner to the outer mucus layer. This stratification is most clearly demonstrable in the colon, where a relatively tightly attached mucus layer proximal to the epithelium is normally devoid of bacteria, but a less cross-linked more superficial layer is heavily colonized (5). In humans, it has been shown that the abundance of bacteria is highest in the large intestine, where colonization of the mucosal epithelium is contested by a thick mucus layer (ϳ700 m) and more rapid mucus turnover (6). In the small intestine, however, where bacteria are much less abundant, the mucus layer is between 100 -400 m thick. O-glycosylation confers the mucus component of the mucosal barrier with much of the protective properties required to separate the abundant gut microflora from the immune system of its host. Defects in mucin glycosylation lead to severe inflammation and susceptibility to infection, and the glycans themselves have been shown to be ligands that can block the binding of microorganisms (7)(8)(9)(10)(11)(12)(13)(14). Thus, secreted gel-forming mucins are key components of intestinal defense, beyond which bacterial pathogens must normally penetrate to cause pathology through interaction with epithelial cells.
Several C. jejuni adhesion proteins required for colonization of chickens have been identified (15), but the mechanisms by which the bacteria initiate interactions with carbohydrates on mucosal surfaces remain unclear. However, many hypotheses can be made from previous studies where bacterial-carbohydrate interactions have been investigated (Table I). Work from Day et al. demonstrated that glycan array binding of two forms of the C. jejuni NCTC11168 strain; an original unpassaged isolate and one with multiple passages having adapted to laboratory conditions was influenced by temperature and oxygen availability. The different isolates bound structures con-taining fucose, mannose, terminal galactose, and N-acetylneuraminic acid (7). Additionally, bovine mucins and L-fucose are known chemoattractants for C. jejuni (8), and L-fucose is a substrate for C. jejuni growth (13). Glycans from human breast milk bearing the H(O) blood group antigen (which presents ␣1-2-linked fucose) also inhibits infection (12). D-glucose, Dmannose, and D-fucose, but not the L-sugar equivalents, inhibited the binding of C. jejuni to human colonic Caco-2 cells (16). Muc1, a membrane-bound mucin, is up-regulated during infection in mice, and knockouts are highly prone to C. jejuni infection with transepithelial translocation (11). These studies have provided considerable insight into C. jejuni behavior, but the specific relationship between the bacteria and mucin glycans in vivo remains unidentified.
Recently, Alemka et al. demonstrated that adherence and internalization of C. jejuni of the human intestinal cell line HCT-8 was inhibited by mucins purified from chicken caecum and small and large intestines in vitro (4). In this study, the strongest inhibition of binding and invasion was found with O-glycosylated mucins from the large intestine followed by the small intestine and caecum, respectively. Following incubation with small and large intestinal mucins, C. jejuni showed a 6.5-fold decrease in binding to HCT-8 cells and a Ͼ1,000fold and ϳ150-fold decrease in internalization, respectively. Mucin from the chicken caecum only exhibited a fivefold reduction in internalization. The presence of chicken mucin had no effect in bacterial viability. When mucin O-glycans from each chicken intestinal section were oxidized with sodium metaperiodate, binding and internalization by HCT-8 cells returned to levels observed in the untreated controls. These results illustrate that the O-glycans expressed on the secreted intestinal mucins of chickens can inhibit colonization regiospecifically. A detailed structural characterization of chicken intestinal mucin glycans from these different sites may provide insights into what makes this species resistant to the pathology caused by this organism in humans. Therefore, we now report the mucin O-glycan structures from  healthy chicken caecum, small intestine, and large intestine. O-glycan analysis by LC-MS identified structures that were not only unique to the different intestinal sections but also others that have never been identified in the human intestinal tract.

EXPERIMENTAL PROCEDURES
Chicken Mucin Isolation and Purification-Chicken intestinal sections (large intestine, small intestine, and caecum) from healthy animals were collected and epithelial mucins were purified from each. Chemicals used were of analytical grade quality and sourced from Sigma-Aldrich (St. Louis, MO) unless otherwise indicated. Mucins were isolated and purified in accordance with published methodology (4,17,18). Six hens were sacrificed, and the intestine was dissected from each. The tract was divided into sections of caecum, small intestine, and large intestine. Each section was then slit longitudinally, and the mucus at the epithelial surface was scraped into 7 M GuHCl in PBS using a scalpel. These samples were rolled at room temperature for up to 5 days to solubilize the mucus and allow the samples to become homogenous. Dithiothreitol was added at a final concentration of 0.01 M and incubated for 5 h at 37°C. Iodoacetamide was added to a final concentration of 0.025 M. The samples were centrifuged at 23,000g for 60 min at 4°C to remove insoluble components. The supernatants were retained and the pellets discarded.
The density of dispersed mucus was then adjusted to 1.42 g/ml using cesium chloride. Samples were then centrifuged at 65,000 rpm for 18 h at 10°C. 1 ml aliquots were removed along the gradient from top to bottom, and the density of each fraction was measured. The relative intensity of a 5 l aliquot was assessed from each fraction when slot blotted onto a polyvinylidene fluoride membrane and stained using periodic acid-Schiffs stain (PAS staining). Based on the density and PAS analysis, the fractions containing mucin were identified and pooled. High molecular weight mucin was then separated from these pooled samples by size exclusion chromatography. The column was packed with Sepharose Cl-4B to a volume of 240 ml, and the mobile phase used was 50 mM Tris, pH 7.5. The eluent was collected in 4 ml fractions. Mucin-rich fractions were determined by PAS analysis as described above. The fractions for each tissue type were pooled and desalted on 400m Sephadex G-25 columns before being freeze dried. Samples were then resuspended into sterile distilled water at 1 mg/ml of mucin and stored at Ϫ20°C until analysis.
O-glycan Release and Preparation-Oligosaccharides were released by reductive ␤-elimination with minor adaptation as previously described (19). In brief, chicken large and small intestinal and caecal O-glycans were chemically released from 20 g of mucin in 1 ml of 1.0 M NaBH 4 in 0.10 M NaOH for 16 h at 50°C. Release of O-glycans from poricine gastric mucin (Sigma-Aldrich) was equivalent. The ␤-elimination reaction was stopped with the addition of 1 l glacial acetic acid. Peptides were removed and samples were simultaneously desalted using a hand-packed bed of DOWEX AG-50 cation exchange resin (BioRad, Hecules, CA) on the top of a C18 Zip-tip (Millipore, Billerica, MA). Prior to the sample addition, the resin was washed with 1 ml methanol, primed with 1.0 M HCl, and equilibrated with 1 ml H 2 O. The sample solutions were added and spun for 5 min or until the entire sample was loaded. The resin bed was washed with three aliquots of 50 l H 2 O. Samples were dried and borates were removed by a series of five washes consisting of 60 l 1% acetic acid in MeOH followed by vacuum centrifugation between additions. Glycan samples were reconstituted in 10 l H 2 O prior to LC-MS analysis.
LC-MS-Mass spectrometry analysis was performed as described with minor adaptations (20). Briefly, analysis of glycan alditols was performed using a HTC-PAL autosampler (CTC Analytics AG, Zwingen, Switzerland), Agilent 1100 HPLC binary pump (Agilent Technol-ogies, Palo Alto, CA), and Thermo LTQ linear quadrupole ion trap mass spectrometer (Thermo Electron, San José , CA). HPLC analytical columns, consisting of 5 m porous graphitized carbon particles (Hypercarb, Thermo Hypersil-Keystone, Bellefonte, PA), were packed in a fused silica capillary (20 cm ϫ 180 m inner diameter; outer diameter 375 m; Polymicro Technologies, Phoenix, AZ). A linear gradient from 0 -40% acetonitrile was used consisting of 100% 8 mM NH 4 HCO 3 (solvent A) and 8 mM NH 4 HCO 3 in 80% acetonitrile/20% H 2 O (solvent B). The flow rate was between 8 -12 l/min generated by a preinjection flow splitter. MS data were acquired in negative mode, and MS 2 spectra were obtained automatically using the Xcalibur 2.1 MS software. Glycan ions were detected as [M-H] Ϫ and [M-2H] 2ions and elution time ranged between 6 -27 min. The collision-induced dissociation energy was set to 30%, and the spray voltage was 3.2 kV. FMS and MS 2 data were annotated manually with the assistance of the Unicarb-DB glycan database for spectral matching (http:// www.unicarb-db.com) (21).
Reverse Phase HPLC: Sialic Acid Speciation-The LudgerTag TM DMB sialic acid release and labeling kit was purchased from Ludger Ltd (Oxfordshire, UK). 100 g aliquot of purified large intestinal mucin was dried and 25 l 2 M acetic acid was added. It was then incubated at 80°C for 2 h. Samples were cooled, and 5 l was removed, transferred to a 0.5 ml centrifuge tube, and 20 l 1,2-diamino-4,5methylenedioxybenzene (DMB) labeling mix was added. The labeling mix was prepared by adding 440 l mercaptoethanol solution to the sodium dithionite vial and mixed until completely dissolved. This mixture was added to the DMB vial and mixed. The DMB labeling reaction was carried out for 3 h at 50°C. The reaction was quenched with the additional of 475 l water. The sialic acid quantitative standards were prepared as outlined in the kit protocol. Analysis was performed using a Waters XBridge TM C 18 3.5 m particle size 2.1 ϫ 150 mm column, Waters Alliance 2695 HPLC with a 2475 Multi-Wavelength Fluorescence Detector (Waters Corp., Milford, MA). Solvent A was 10 mM ammonium formate adjusted to pH 4.4, and solvent B was acetonitrile. The flow rate was 0.2 ml/min and separation was performed with a 40 min linear gradient running from 0% to 50% solvent B. The ex and em were 373 nm and 448 nm, respectively.

RESULTS
Here, we present a detailed analysis of O-glycans released from purified chicken caecum, small intestinal, and large intestinal mucins by LC-MS n analysis using porous graphitized carbon, which is a well-established technique (20) for assigning glycan structures. Annotation of fragmentation spectra was performed de novo where glycan fragments were assigned manually and checked with the aid of UniCarb-DB, an on-line database of LC-MS mucin O-glycan fragmentation data (21). The online Mucin Glycan Database from the Mucin Biology Group at the University of Gothenburg was also a resource for MS 2 spectral matching (22); however, not all glycan structures assigned in this study were represented in these databases. GlycoworkBench (23) was also used to verify glycan assignments as follows; MS 2 peak lists (fragment ions with a relative intensity Ͼ1%) were imported, and fragments were annotated from candidate structures. These data were uploaded to Unicarb-DB and are freely available. Branching and topology were established from MS 2 spectra, and the assignment of specific structural features (i.e. Lewis and blood group epitopes) was based on structural assumptions of the individual monosaccharide identification, namely Hex and HexNAc residues as galactose and N-acetylglucosamine, respectively.
Glycans were not modified prior to analysis and were observed as deprotonated species, primarily [M-H] Ϫ with minor [M-2H] 2ions seen for larger structures. A control sample of porcine gastric mucin (PGM) from Sigma Aldrich was run prior to analysis to ensure instrument settings were optimized to limit in-source fragmentation. Observed and theoretical massto-charge values with corresponding glycan structures are presented in Table II. In the absence of diagnostic fragment ions, as seen in the higher mass O-glycans where detailed assignments were not possible, only a compositional description was assigned and presented. Additionally, the relative abundance of each precursor ion from the combined spectra (spanning 6 -27 min) was determined. These values were calculated as a ratio of the measured intensity of each ion to the sum of all ion intensities. The distribution within each type (neutral, sialylated, sulfated, and sulfo-sialylated) was also determined under the same principle but with the ratio of each O-glycan structure as a ratio to the sum of peaks within each group. Glycan fragmentation vocabulary used was defined by Domon and Costello (24), and O-glycan cartoon representation was the CFG glycan nomenclature (Supplemental Fig. 1) (25). Structures were assigned (Table II) from collision-induced dissociation spectra and were found to be primarily extended core III (GlcNAc␤1-3GalNAc) and core IV (GlcNAc␤1-3[GlcNAc␤1-6]GalNAc) O-glycans. Core I and core II O-glycans were also detected in the caecum.
Mucin O-glycan Diversity Is Correlated to C. jejuni Colonization Density-The overall glycan compositions in each section of the intestinal tract examined, determined from the combined MS profiles (Fig. 1), showed the diversity and range in the number of O-glycans in the chicken intestinal tract. Mucin glycosylation heterogeneity increased from the caecum toward the large intestine, with 23 structures found in the caecum, 40 in the small intestine, and 60 in the large intestine. The increase in glycan microheterogeneity also corresponds with the greatest inhibition of C. jejuni binding to HCT-8 cells (i.e. large intestines glycans inhibited the most). There were 16 structures unique to the chicken large intestine compared with other sections. The intensity (signal strength) of the combined large and small intestine MS 1 spectra (6 -27 min) were comparable (NL: 1.12E1 and NL: 2.01E1), whereas the caecum O-glycan sample intensity was an order of magnitude less (NL: 4.31), indicating a lower degree of glycosylation of mucins in caecum.
The distribution of glycans by type in the large intestine was 26% neutral, 29% sialylated, 34% sulfated, and 11% sulfosialylated. O-glycan distribution in the small intestine was similar to the large intestines with 40% neutral, 23% sialylated, 33% sulfated, and 4% sulfated-sialylated glycans, while the caecum contained more sulfated (57%) and less sialylated (14%) with a similar amount of neutral (22%) and sulfatedsialylated (7%) glycans as the other two sections. Neutral glycans only represent a small number of structures identified in the chicken intestines, especially in the caecum. There were 15 neutral structures in the large intestines, 13 in the small intestines, and five in the caecum. Most noticeable between samples was the presence of highly anionic structures that contained sulfate and/or sialic acid residues.
O-glycans Were Predominantly Extended Core III and Core IV Extended with Both Type I and II LacNAc-O-glycan cores I, II, III, and IV were identified in all mucin sections (Fig. 2) but core III and core IV predominated based on ion intensities in the MS 1 spectra. Core III structural isomers were identified from the collision-induced dissociation fragmentation of the m/z 587 ion (HexHexNAcHexNAc), confirming the presence of an extended core III structure with both type I Gal␤1-3GlcNAc and type II Gal␤1-4GlcNAc residues. Core II (Gal␤1-3[GlcNAc␤1-6]GalNAc) structures were detected in the large and small intestines and comprised a small percentage (0.3% and 0.4%, respectively). The most intense fragment in the core II MS 2 spectrum was a 407 ion, resulting from the elimination of the galactose linked at the C-3 position of the reducing end GalNAcol. This characteristic C-3 fragmentation was used to identify the branching of neutral and sialylated core II and core IV structures from the MS 2 spectra throughout the analysis. The B type fragment ion of m/z 364 represents the loss (-223) of the reducing end GalNAcol residue, and this type of neutral loss was informative for assigning linear core types (i.e. core I and III) together with the 108 Da loss (C 3 H 8 O 4 ).
Evidence for both type I and II LacNAc extensions was based on diagnostic ions in MS 2 spectra and description of fragmentation mechanisms (20,26). Evidence for the Gal␤1-4GlcNAc linkage (type II) of a core III structure was supported by the presence of the m/z 263 ion ( 0,2 A -H 2 O) and m/z 221 ( 2,4 A) fragment (Fig. 2). However, the m/z 263 ion was also observed in the core III type I spectrum and is likely from coelution of the precursor ion during analysis. It is important to note that the 263 ion was detected in the MS 2 spectrum from the proposed core III type II structure in the small intestines but not in the type I spectrum (raw data are found in UniCarb-DB). When detected, the m/z 263 ion was used for differentiating type I or type II as was the case for the m/z 790 ion (core III, type II with an additional HexNAc). For the type I glycan, it was found that ions resulting from a neutral loss of 108 Da on the reducing end as described above could be further fragmented by a galactose neutral loss (Ϫ162) giving the type I diagnostic ion fragment of m/z 317. Furthermore, the m/z 202 ion (Y 2 /B 2 cleavage) was detected in the type I structure and not the type II. However, in the absence of diagnostic ions to specify either type, O-glycan structures were designated with a "Hex-3/4" linkage as seen in Table II.
Core IV structures were indicated by presence of a trisaccharide with the composition HexNAc 2 HexNAcol ([M -H] Ϫ 628 ion). Further evidence was provided from the MS 2 where a predominant Z type m/z 407 fragment showed that one GlcNAc was linked to the C-3 of GalNAcol, leaving the additional GlcNAc attached to the C-6 position. The characteristic 262 2,4 A cleavage from the reduced GalNAc indicative of a GlcNAc␤1-6GalNAc was also identified. Overall, the most common core O-glycans were type III and IV and subsequent structures throughout this study were primarily extensions of these structures.
The C. jejuni Chemo-Attractant ␣1-2 Fucose Is Present on Large and Small Intestine Mucins-Fucose-containing glycans were present throughout the intestinal tract and were identified with blood group H (Fuc␣1-2Gal␤1-) like and blood group B (Gal␤1-3(Fuc␤1-2)Gal-) like epitopes as well as ␣1-3/4 linkages present as Lewis a/x (Gal␤1-3/4(Fuc␤1-3/4)Gal-NAc-) epitopes. The sequence of blood group H-like structures in the large and small intestine is illustrated by the fragmentation of the m/z 895 ion where the fucose residue was linked to the penultimate hexose residue, presumably galactose (Fig. 3). Ions 731 and 749 (Z-and Y-type fragments) were from the neutral loss of a fucose, and ions 715 (Z-type) and 733 (Y-type) resulted from a hexose loss, but these ions alone were not diagnostic of a blood group epitope. The m/z 569 (Z-type) and m/z 731/749 (Z-and Y-type) fragment ions showed the fucose-hexose disaccharide is in a linear conformation. The 569 ion, which corresponds to the neutral loss of a hexose and fucose, contains the reducing end HexNAcol with a hexose and HexNAc residue. This ion could arise from a structure with a terminal hexose-fucose disaccharide, and together with Z-ion of m/z 407 and the m/z 389 ion (407 -H 2 O) support the presence of a blood group H-like epitope on the C-3 branch of the GalNAcol. Spectral and retention time matching to the equivalent m/z 895 ion isolated from the well-characterized blood group H expressing PGM standard provided further evidence for the presence of blood group H epitopes in chicken intestinal mucins (Fig. 3, inset). An additional blood group H containing structure was found in the large intestines (m/z 1041) as an extension of the m/z 895 structure with an additional fucose, and comparison to the PGM equivalent also supported this structural assignment. These two structures represent the only neutral blood group H-like glycans, but several blood group H-like sulfated and sialylated structures were also found ( Table II). A blood group B-like epitope was detected in the MS 2 spectra of m/z 1551 and 1567, and its structural assignment is discussed below.
Other neutral-fucosylated structures with ␣1-3/4 linked fucose were identified (m/z 733, 936, 1098, and 1244) in the small and large intestines only. The fragmentation spectra of the m/z 733 glycan ion from the large and small intestines were consistent with a Lewis a/x type epitope (Fig. 3). The predominant fragment ion at m/z 359 was produced from the loss of the hexose and fucose residues (a double Z-type fragment ion; m/z 389) together with a loss of the CH 2 O side group (-30 Da). Spectral matching of the 733 MS 2 to a previous report characterizing Lewis and blood group MS 2 fragments of five structures with identical compositions (20)    . Chickens lack the enzyme to produce NGNA residues but trace amounts have previously been reported (27), and the presence of NGNA identified by MS in this study was further confirmed by sialic acid speciation HPLC (Fig. 4A). Several O-glycans were found in the chicken large intestines that were not in the small intestines or caecum. These ions were m/z 1040, 1169, 1259, 1405, 1551, 1567 (Fig. 4B). These isomers had different retention times and MS 2 fragmentation spectra, confirming the structural assignments (Supplemental Fig. 3). The chromatogram of the m/z 1040 ion only showed a single peak, but the location of the residue was not confirmed.
The m/z 1551 ion and the corresponding N-glycolylneuraminic acid containing structure (m/z 1567) had matching fragmentation MS 2 spectra suggesting the structures were identical except for the sialic acid residue (Fig. 5). There is evidence for both blood group B-like epitopes and Sd a -like antigens that were determined from several important diagnostic ions. The presence of the 772 ion suggested substitution at the C-6 and was consistent with HexNAc-Hex-HexNAc linked to the reducing end HexNAcol (following the loss of sialic acid). Furthermore, the 466 ion results from the loss of sialic acid (Z-type) and a 0,2 A cross-ring cleavage of HexNAc residue linked to the reducing end HexNAcol. These ions in addition to the 569 fragment ion that comes from a HexNAc loss of the 772 fragment, verifies a HexNAc-Hex-HexNAc sequence. Additional evidence for Sd a -like-containing structures was also detected for the 1243/1259 and 1389/1405 ions. Sd a -like antigens appear to be a common feature for chicken large intestines, and Sd a antigens have been found in other species such as human, mice, and rat. For the blood group B-like (Gal␤1-3(Fuc␤1-2)Gal-) assignment, the determining fragments were from the 895 (hexose loss on the C-3 arm) and 730 (additional fucose loss) fragment ions. Together, these show the presence of a terminally linked hexose and fucose residues linked to the hexose at the C-3 position indicative of a blood group B epitope.
Chicken Intestinal Mucin Glycans Are Highly Sulfated-The most recurrent structural motif in all chicken intestinal sections was sulfation. There were 19 sulfated structures in the large intestines and 13 sulfated structures in the caecum and small intestine. Despite the relatively reduced diversity of nonsulfated structures in the caecum, surprisingly, the number of sulfated glycans was nearly equal in number to structures in the small and large intestines. MS 2 spectra of sulfated structures confirmed that sulfates were present on hexose and N-acetylhexosamine residues as shown from fragmentation of the most commonly observed sulfated ions (Fig. 6) and additional sulfated species (Supplemental Fig. 4). Sulfated residues were identified on hexose residues typically as Btype ions m/z 241 and on N-acetylhexosamine residues (Btype ions m/z 282). Sulfated O-glycans were found in most cases to be extensions of the neutral or sialylated glycans reported in Table II with the sulfate moiety located primarily on terminal residues presumably linked to the 3 or 6 position on galactose residues. We did not identify these linkages specifically, and thus, they were not reported.
Large Intestine Mucins Contain Keratan Sulfate-Like Lac-NAc Repeats-Four structures containing sulfated galactose ␤1-4N-acetylglucosamine repeats, characteristic of keratan sulfate, were identified. These structures (m/z 1178, 1324, 628 [M-2H] 2-, and 701 [M-2H] 2-) also contained one or two fucose residues linked to HexNAc residues, presumably forming Lewis a/x decoration of the sulfated galactose ␤1-4N-acetylglucosamine repeats. The assigned structures corresponded to extended core III glycans with sulfates on the hexose and/or HexNAc residues. The doubly charged ions 628 and 701 [M-2H] 2structures contained two sulfates linked to hexose/HexNAc residues and only differed by one fucose residue (Supplemental Fig. 5). Similar structures were identified with one sulfate (m/z 1324, 1178, and 1032) and two, one, and zero fucose residues. These structures were identified in all three intestinal sections.
Sialylated-Sulfated O-glycans-O-glycans with both sulfate and sialic acid residues were identified in the chicken mucin. There were seven structures in the large intestines, three in the small intestines, and two in the caecum. Five of the seven Interestingly, the location of sulfate was primarily on hexose residues in this class of structure.
Differences to Human Intestinal Mucin Structures-A previous report of human O-glycan structures isolated from different gut sections obtained from two male adults of A-Lewis b blood groups provides useful information for comparison of structures identified in this study. Evaluation of chicken mucin O-glycans to the equivalent sections in humans (caecum, ileum, and transverse colon) shows considerable differences in the type of glycans present in the gut (Fig. 7). The number of O-glycans was the highest in the ileum (n ϭ 56 and 40) and caecum (n ϭ 42 and 46), followed by the transverse colon (n ϭ 41 and 34). These values contrast to our findings specifically in the caecum section where we identified only 23 O-glycans, which corresponds to the lowest level of C. jejuni inhibition in vitro. Several structures were identified in chicken mucins that are not present in any human intestinal mucins. Overall, there were 33 structures unique to the chicken not reported in any section of the human gut. Of these, nine were unique to the large intestines where C. jejuni inhibition was the greatest.
When comparing the degree of sulfation to the human gut, it becomes evident that mucin sulfation is significantly more predominant in chicken (Fig. 7). We identified 19 sulfated structures in the large intestine (LI) and 13 in the small intestine (SI) and caecum. This contrasts to the amount of sulfation found in the human transverse colon (n ϭ 6 and 2), ileum (n ϭ 2 and 0), and caecum (n ϭ 2 and 6). Sulfated glycans in humans were most predominant in the sigmoidal colon (n ϭ 5 and 7) and rectum (n ϭ 7 and 8). The location of sulfate was found on both hexose (Gal) and HexNAc (GlcNAc) residues. Many sulfated structures were similar between species (i.e. extended core II and III) with the exception of extended core IV structures (m/z 870, 1016, and 1032) found in all three chicken sections. There were not any sulfated core IV glycans in the human gut.
Sialic-acid-containing glycans reported in the human intestinal tract were primarily extended core III, with the sialic acid residue linked at the 6 position on the reducing end GalNAc-ol. There were three structural isomers for the 1024 ion, two of which were present in all human sections. These two structures were a core III with the Neu5Ac linked to the reducing end GlcNAcol and a Lewis a/x epitope (Hex33/4(Fuc33/4)-HexNAc33(Neu5Ac36)HexNAcol). This structure was found in the chicken small and large intestines (Fig. 4), but an additional isomer with a terminal sialic acid was more predominant (Neu5Ac33Hex33/4(Fuc33/4)HexNAc33HexNAcol). This structure was reported in the sigmoidal colon of one patient. In general, we observed sialylation both at the reducing end GalNAcol but more commonly linked at terminal positions. DISCUSSION This analysis represents a follow-up study from previous work demonstrating that chicken mucin O-glycans inhibit binding and internalization of C. jejuni on the human colorectal HCT-8 cell line. Here, we identified the mucin O-glycan structures and identified unique characteristic structural features that offer insights that may enhance understanding of how chicken mucins inhibit C. jejuni infection. Most notable was the consistent expression of sulfates in all sections of the intestinal tract. Terminal epitopes that were identified as well as the degree of sulfation present on the entire O-glycan is highlighted in Table III. The most common epitope were Lewis a/x structures with and without sulfate followed by blood group H-like structures. Secondly, the overall number of glycans correlated inversely with the density of C. jejuni in vivo, with the caecum having the highest number of colony-forming units (10 10 cfu/gram), the lowest glycan diversity, and least inhibitory effects in in vitro studies. Above all, C. jejuni are unable to cross the mucosal barrier in any section of the chicken intestines, and therefore, bacterial resistance may be attributable to the presence of a common structural feature throughout the tract. The extent of chicken mucin sulfation is further emphasized when compared with results from intestinal sections isolated from two healthy human donors. As highlighted in Fig. 7, the number of sulfated structures in chicken caecum was 13 and represents roughly 57% of all O-glycans. By comparison, sulfation in the caecum from two healthy human samples was 13 and 5% (28). These numbers were comparable in the other sections: small intestines 33% (chicken) and 0 and 4% (human); large intestines 30% (chicken) and 6 and 15% (human). However, another report investigating the glycosylation of MUC2 isolated from 25 human colon biopsies classified 42 sulfated O-glycans (29), but it is important to note that these values represent all structures identified and not necessarily the number present in each individual.
Sulfation is a common modification of gut mucin O-glycans and can protect the host from mucin degradation by glycosidases (30,31). The link between sulfated mucins and C. jejuni remains uncharacterized, but inferences can be made based on previous studies where sulfation has been shown to confer protective properties. The NaS1 sulfate transporter knockout mouse that exhibits a low level of circulating serum SO 4 2showed a reduction in sulfomucins in goblet cells in the jejunum, ileum, and colon (32). The NaS1 Ϫ/Ϫ mouse had equal amounts of C. jejuni in the small and large intestines compared with NaS1 ϩ/ϩ mice, but the bacteria were detected in the liver of the null mice, suggesting that a decrease in available sulfate or potentially an overall reduction in mucin content promotes transepithelial transit of the organism. Additionally, gene expression studies in two healthy chicken lines that differ in C. jejuni resistance revealed a 3.346-fold increase in sulfotransferase activity in the more resistant animals as well as an increase in galactose 3-O-sulfotransferase-2 (0.104fold) and heparinsulfate 6-O-sulfotransferase-3 (0.412-fold)  (29), inhibits C. jejuni growth in vitro and significant up-regulation of mucin-degrading enzymes as well down-regulation of a sialic acid synthase (34). A thorough glycan array study using an attenuated C. jejuni strain adapted to laboratory conditions through multiple passages (11168-GS) and the original more invasive strain (11168-O) revealed the importance of sulfate in binding (7). Both strains bound fucosylated glycans, especially Gal␤1-3(Fuc␣1-4)GlcNAc (Lewis a ), but binding of the 11168-O strain to SO 3 -Gal␤1-3(Fuc␣1-4)GlcNAc (sulfo-Lewis a ) was significantly reduced. From these previous reports and from the prevalence of this epitope with and without sulfate (Table III), it appears that mucin sulfation may play a significant role in protecting the intestinal wall against C. jejuni adhesion and invasion. Continued studies should target the binding properties of C. jejuni against glycans with and without sulfate moieties to determine changes in affinity or identify epitopes with greater inhibitory effects. The degree of charged structures (sulfated, sialylated and sulfated/sialylated) increased from the caecum to the large intestines, with the number of sulfated and sialylated glycans most prevalent in the large intestines. One interesting result was the identification of N-glycolylneuraminic acid (Neu5Gc) residues in the large intestines despite the observation that chickens lack the gene required to synthesize Neu5Gc ((cytidine monophosphate (CMP)-N-acetylneuraminic acid hydroxylase (CMAH)), although trace amounts have been previously reported (27). Interestingly, chickens are the only other animals aside from humans that generate antibodies against injected Neu5Gc monosaccharides (35). Uptake and incorporation of dietary Neu5Gc into salivary mucin O-glycans has been observed in humans (36), and the adsorption of Neu5Gc from the diet may be the source of the O-glycans containing N-glycolylneuraminic acid in this study. Previous reports have shown that alterations in chicken diet affect mucin glycosylation in goblet cells and also C. jejuni colonization in broiler chicks (37). A diet supplemented with xylanase resulted in a reduction in C. jejuni in the intestines and an increase in sialic-acid-containing residues in goblet cells. Additionally, an increase in sulfation on the surface and upper crypt cells was observed with the xylanase supplemented diet. These results further support a possible correlation between an increase in anionic glycans and a decrease in C. jejuni colonization in broiler chicks, but the specific role of N-glycolylneuraminic acid residues in bacterial defense remains unclear. The biological significance of NGNA residues linked terminally as opposed to being linked to the GalNAc core (as seen with NANA residues) is not ostensibly obvious. One postulation is that substrate recognition of the enzyme responsible for terminal additions has a broad affinity for NGNA as well as NANA. We cannot infer the biological importance for the presence or absence of NANA/NGNA as it pertains to C. jejuni binding, but this result underscores the need for additional binding assays such as glycan arrays. Importantly, it may be possible to limit C. jejuni infection by shifting the mucin bilayer toward a more anionic O-glycan composition through altering broiler chick diets.
It is well known that a strong affinity between C. jejuni and fucose monosaccharides exists, where previous work has shown that C. jejuni bind milk oligosaccharides, particularly neutral ␣1,2-fucosylated structures (12). L-fucose, and hog gastric mucins have also been show to chemo-attract C. jejuni (8). Additionally, Chinese hamster ovary (CHO) cells overexpressing ␣1,2-fucosyltransferase (FUT1) inhibited bacterial binding; whereas wild-type CHO cells did not (12). FUT1 transfected CHO cells that express the H(O) Fuc␣1,2Gal␤1,4GlcNAc-R epitope showed greater C. jejuni binding compared with CHO cells transfected with FUT3 (␣1,3/4-fucosyltransferase) or FUT4 (␣1,3-fucosyltransferase). Our analysis identified fucose-containing structures with ␣1,2-fucose and/or ␣1,3-fucose and ␣1,4-fucose linkages. Interestingly, the site with the most neutral O-glycans with ␣1,2-fucose was in the large intestines where colonization population is the lowest. The expression of this linkage in mucins may therefore act as an inhibitory binding decoy in the chicken large intestine. However, this effect may be modulated where glycans carry charged residues. Interestingly, the greatest colonization by C. jejuni in the chicken gut occurs in the caecum where we did not detect neutral fucosylated structures. Ruiz-Palacio et al. state that ␣1,2-fucose-containing glycans may serve as a potential ligand in the initializing step toward infection. C. jejuni also express enzymes for the metabolic utilization of L-fucose (13). Currently, it remains unclear whether ␣1,2-fucose is the preferred substrate and first recognition step or acts as a decoy ligand to prevent infection.
The analysis of chicken intestinal mucin glycans provides a platform for greater understanding on how thematic structural differences in mucin glycosylation could explain how C. jejuni exists as a commensal within the mucus of the chicken gut but as a pathogen in the human gut. Based on this work, the following observations differentiate different sites with different levels of colonization in chickens: The site with the greatest inhibition in vitro and the least colonization in vivo (large intestines) had the greatest number of O-glycans as well as the most diversity of structures. Where colonization is highest and inhibition was weakest in vitro (caecum), there were far fewer O-glycan structures detected. There are two plausible postulations for the increased inhibition in the large intestines. The total increase in glycan diversity creates a barrier that the bacteria cannot bypass in order to invade the gut epithelium or that there are specific glycan structures/epitopes that inhibit the bacteria and eliminate their ability to travel through the outer mucosal barrier. However, considering that chicken mucins protect against C. jejuni in vivo, further observations can be made differentiating the characteristics of the mucus barrier in humans and chickens. First, the level of sulfation is much greater in chickens, and second, the overall charge density of anionic glycans is increased. The increase in anionic glycans may provide a shielding or charge-repulsion effect as well as maintaining the relatively low pH in the intestines (5.7-6.4) (38) that could also reduce the pathogenicity of the bacterium. Overall, the increase in anionic glycans through sialylation, sulfation, and sialylation/sulfation has already been shown to have protective effects. However, the mode in which the bacterium interacts with, or might be deterred by, these factors is not clear. Together, these observations provide rational avenues and a library of structures for further investigation of the chicken mucin glycome for glycan structures with therapeutic or prophylactic potential against C. jejuni.