Differential Carbohydrate Recognition by Campylobacter jejuni Strain 11168: Influences of Temperature and Growth Conditions

The pathogenic clinical strain NCTC11168 was the first Campylobacter jejuni strain to be sequenced and has been a widely used laboratory model for studying C. jejuni pathogenesis. However, continuous passaging of C. jejuni NCTC11168 has been shown to dramatically affect its colonisation potential. Glycan array analysis was performed on C. jejuni NCTC11168 using the frequently passaged, non-colonising, genome sequenced (11168-GS) and the infrequently passaged, original, virulent (11168-O) isolates grown or maintained under various conditions. Glycan structures recognised and bound by C. jejuni included terminal mannose, N-acetylneuraminic acid, galactose and fucose. Significantly, it was found that only when challenged with normal oxygen at room temperature did 11168-O consistently bind to sialic acid or terminal mannose structures, while 11168-GS bound these structures regardless of growth/maintenance conditions. Further, binding of un-capped galactose and fucosylated structures was significantly reduced when C. jejuni was maintained at 25°C under atmospheric oxygen conditions. These binding differences identified through glycan array analysis were confirmed by the ability of specific lectins to competitively inhibit the adherence of C. jejuni to a Caco-2 intestinal cell line. Our data suggests that the binding of mannose and/or N-acetylneuraminic acid may provide the initial interactions important for colonisation following environmental exposure.


Introduction
Carbohydrates (or glycans) that modify proteins and lipids play a key role in numerous cell recognition events, including those involved in the regulation of the immune system [1], and in the attachment of pathogenic organisms to host tissue [2]. Proteincarbohydrate interactions have been identified as adherence factors for numerous commensal and pathogenic bacteria including Pseudomonas aeruginosa [3,4,5], Helicobacter pylori [6,7,8,9,10] and Escherichia coli [11]. The identification and characterisation of carbohydrate binding proteins (or lectins) has been greatly enhanced through the development of glycan array technology [12]. Glycan arrays are fast becoming the technique of choice for identifying and elucidating the specificity of lectins [12,13,14], and have been successfully used to identify and characterise the interactions of viruses and bacteria with their glycan receptors [12,15].
Campylobacter jejuni, the most prevalent cause of gastroenteritis in developed countries, is a highly motile, Gram-negative spiral rod that requires microaerobic conditions for growth [16,17]. C. jejuni is a zoonotic pathogen, being a commensal organism in poultry and other wildlife. Poorly prepared poultry products represent the most common source of human infection; however, infection can also arise from contaminated water and unpasteurised milk [16]. C. jejuni is a thermophilic organism, requiring temperatures in the range of 32uC to 45uC for growth, but it can also survive at lower temperatures in the environment [18].
The pathogenic clinical isolate NCTC11168 was the first C. jejuni strain to be sequenced and has been a widely used laboratory model for studying C. jejuni pathogenesis [19]. However, more recent investigations have shown that continuous passaging of C. jejuni NCTC11168 dramatically affects its colonisation potential. That is, long-term passaging and oxygen-adaptation of the 11168genome sequenced strain (11168-GS) leads to poor colonisation potential in chicken models [20,21], while the same strain infrequently passaged (11168-original, 11168-O) remains a potent coloniser of chickens. Moreover, C. jejuni 11168-O was found to be more adherent and invasive in in vitro infection models and was observed to be more motile than 11168-GS [21]. Microscopic analysis of the two strains also demonstrated a shift in phenotypic appearance, with 11168-GS existing as straight rods rather than the normal spiral-shaped rod of 11168-O. Other differences were also noted in gene expression, with particular changes in genes involved in oxygen metabolism; however, serotyping, genotyping and subtractive hybridisation found the two strains to be clonal [21].
C. jejuni does not ferment or oxidise carbohydrates as a carbon source, instead relying on amino acids such as aspartate and serine [22]. Any observed carbohydrate binding can therefore be attributed to interactions important for adherence and/or colonisation rather than for energy acquisition. The differential adherence, invasion and colonisation observed for C. jejuni 11168-GS and 11168-O suggests that significant differences exist in how the two strains interact with host cells. Bacteria-host interactions depend not only on bacterial motility, but also on bacterial adherence factors such as lectins, and the surface composition of the host cell, including the degree and type of glycosylation [2,23]. The adherence phenotypes observed for C. jejuni 11168-GS and 11168-O may therefore reflect differential host cell carbohydrate recognition. The carbohydrate-mediated interaction of C. jejuni with host cell glycans has been reported. C. jejuni, for example, is chemotactic to fucose (Fuc) [24], and is known to interact with glycoproteins of the intestinal tract, such as mucins [25] and Lewis B-like structures present on human milk proteins [26,27]. The functional significance of these interactions is not entirely understood, and the complete range of sugars recognised, as well as the C. jejuni lectins involved, are yet to fully be identified.
This study aims to identify glycan structures bound by C. jejuni 11168-GS and 11168-O strains using glycan array technology and in vitro cell adherence assays, and to explore how these interactions are influenced by changes in temperature and oxygen condition. Conditions were chosen to mimic those encountered by C. jejuni through various growth and survival phases, including conditions required for survival in the environment (25uC, normal O 2 ), and in mammalian (37uC, microaerophilic) and avian (42uC, microaerophilic) hosts (Table 1). Here we report that distinct growth and survival conditions dramatically alter the degree and specificity of C. jejuni binding to various carbohydrate structures, including differential binding to mannose (Man) and N-acetylneuraminic acid (Neu5Ac)-containing glycans.

Glycan array analysis
Glycan arrays were performed using the non-colonising genome sequenced isolate of the C. jejuni strain NCTC11168 (11168-GS) and the colonising isolate 11168-O, grown or maintained at varying conditions as outlined in the Materials and Methods (summarised in Table 1). Our analyses found that the binding of C.
jejuni strains 11168-GS and 11168-O to galactose (Gal), Man, Fuc, and Neu5Ac-containing glycans was significantly affected by the conditions under which the bacteria were grown/maintained. In addition, significant differences in the carbohydrate-binding specificity were also observed between C. jejuni 11168-GS and 11168-O.
Binding to terminal Gal/GalNAc C. jejuni 11168-GS maintained at 25uC (11168-GS 25uC) exhibited little or no binding to terminal Gal-containing structures present on the array. Of particular note, no binding to the core Oglycan Tn antigen (Galb1-3GalNAca1-O-Ser; 1M) or glycans terminated with Gala1-3/4Gal (1K, 1N, 1O, 1P and 2A; see  table 2 for structures) was observed ( Figure 1A). The growth of 11168-GS at 37uC under a microaerobic atmosphere, which mimics mammalian host-like conditions (11168-GS 37uC), and at 42uC under a microaerobic atmosphere, which mimics avian hostlike conditions (11168-GS 42uC), resulted in significantly greater binding (P,0.05) to terminal Gal structures in comparison to 11168-GS 25uC. In fact, C. jejuni 11168-GS 37uC and 11168-GS 42uC bound all terminal Gal structures present on the array ( Figure 1A). However, structures such as Galb1-3GlcNAc, Galb1-3GalNAc, Galb1-3GalNAcb1-4Galb1-4Glc and Gala1-3Gal (1A, E, F and N respectively; Figure 1 and Table 2) were bound significantly less by 11168-GS 42uC in comparison to 11168-GS 37uC. Conversely, the structure Galb1-4GlcNAcb1-3Galb1-4Glc (1H, Figure 1 and Table 2) was bound significantly more by 11168-GS 42uC than 11168-GS 37uC. Even though some differences in the level of binding to various terminal Gal structures by C. jejuni 11168-GS grown at conditions that mimic mammalian and avian hosts were noted, there appeared to be no significant specificity requirement for a particular glycosidic linkage or underlying sugar.
With respect to Gal terminated glycans, binding of C. jejuni 11168-O ( Figure 1B) had few statistically significant differences to the binding observed for 11168-GS ( Figure 1A). For example the ability of C. jejuni 11168-O to bind terminal Gal structures generally increased as the growth temperatures increased (P,0.05). Specific differences however were observed between the two 11168 isolates grown at 37uC and 42uC. For example, 11168-O 37uC bound glycans 1H, 1I, 1J, 1K and 1M significantly more than 11168-O 42uC (P,0.05; Figure 1B). This is different than binding observed to the same glycans by 11168-GS grown at 37uC and 42uC, where no significant differences were observed ( Figure 1A). Analogous to that observed for 11168-GS binding to terminal Gal structures, even though some differences in the level of binding to various terminal Gal structures by C. jejuni 11168-O grown under mammalian and avian-like conditions were noted, no significant differences in the specificity for a particular glycosidic linkage or underlying sugar could be deduced. As with 11168-GS, 11168-O grown at 37uC and 42uC, but not at 25uC had significant binding to terminal Gala1-3/4Gal structures ( Figure 1B).

Binding to Mannose containing glycans
With respect to the ability of C. jejuni 11168 to bind Mancontaining structures, figure 2A shows that C. jejuni 11168-GS bound multiple Man structures at low but detectable levels irrespective of the growth conditions tested (Figure 2A). One exception was the binding of 11168-GS to a1-3-mannobiose (Mana1-3Man; 5D), which was noticeably reduced. Interestingly, there were no significant differences observed in the level and specificity of C. jejuni 11168-O 25uC and 11168-GS 25uC binding to Man. However, 11168-O grown at conditions mimicking mammalian and avian hosts had, in the majority of cases, significantly reduced Man-binding in comparison to 11168-GS grown under the same conditions. The only exception to this was 11168-O binding to glycans containing GlcNAcb (1-2)Man, particularly GlcNAcb1-2Mana1-6(GlcNAcb1-2Mana1-3))Man (5B), an important N-glycan biantennary core structure, which significantly increased as the growth temperature was increased, and to a lesser extent 5A ( Figure 2B). Considered together, the data suggests that a distinct difference in Man-binding potential exists between C. jejuni 11168-GS and 11168-O when grown at conditions that are associated with either the mammalian or avian hosts.
Even though in comparison to 11168-GS 25uC a larger subset of fucosylated glycans was bound, C. jejuni 11168-O 25uC like 11168-GS 25uC was unable to recognise Le a and Le x (7J and 7I, respectively; Figure 3B). As was seen for 11168-GS the binding of 11168-O grown at 37uC and 42uC was significantly higher (P,0.05) than that seen for 11168-O 25uC ( Figure 3B). However, unlike 11168-GS, there was no increase in 11168-O Fuc-binding as growth conditions were changed from 37uC microaerobic to 42uC microaerobic ( Figure 3B). One of the few fucosylated structures that was poorly bound by 11168-O was sulfo-Le a (8A, Figure 3B), which exhibited significantly reduced binding compared to 11168-GS grown under host-like conditions. These data highlight distinct differences in Fuc binding that is dependent both on the C. jejuni isolate (11168-GS or 11168-O), and the condition used to grow/maintain the bacteria.

Binding to sialylated glycans
Our glycan array analyses also identified significant differences in the ability of C. jejuni 11168 to bind to Neu5Ac-containing glycoconjugates, with Neu5Ac-binding being affected not only by the growth/maintenance conditions, but also the isolate (11168-GS or 11168-O). Overall, C. jejuni 11168-GS bound all Neu5Accontaining glycans present on the array ( Figure 4A); however, increased binding was observed for 11168-GS 37uC and 11168-GS 42uC in comparison to 11168-GS 25uC (P,0.05; Figure 4A). C. jejuni 11168-O maintained at 25uC was also able to recognise and bind all sialylated glycans. Significantly however, Neu5Acbinding was almost completely abolished when C. jejuni 11168-O was grown at 37uC and 42uC ( Figure 4B). The only exception to this was 11168-O binding to monosialyl, monofucosyllacto-Nneohexaose (Galb1-4(Fuca1-3)GlcNAcb1-6(Neu5Aca2-6Galb1-4GlcNAcb1-3)Galb1-4Glc; 10D; Figure 4B and Table 2), a fucosylated and sialylated biantennary glycan, which was significantly higher for 11168-O 37uC and 42uC in comparison to 11168-O 25uC ( Figure 4B). It is interesting to note that binding to sialylated structures was independent of the Neu5Ac linkage (a2-3, a2-6 or a2-8) and the nature of the sub-terminal sugar for both 11168-GS and 11168-O ( Figure 4). Taken together, these data indicate that significant differences exist between C. jejuni 11168-GS and 11168-O recognition of sialylated glycans when grown at conditions that are associated with either the mammalian or avian hosts.
Recently, a sialidase activity was observed associated with the C. jejuni CSTII sialyltransferase following its expression in E. coli [28]. Given the differences observed in Neu5Ac-binding between 11168-GS (significant Neu5Ac binding) and 11168-O (little or no Neu5Ac binding), and that such sialidase activity could release terminal Neu5Ac from sialylated glycans on the glycan array, additional array analyses were performed in the presence of the sialyltransferase inhibitor CMP, and the sialidase inhibitors Neu5Ac2en and 4-guanidino-Neu5Ac2en (Zanamivir). Representative experiments were performed using 11168-O 25uC and 11168-GS 42uC. These conditions were chosen to allow for testing of both strains at their respective maximum observed Neu5Ac binding capacity. The analyses showed that the addition of these known inhibitors did not significantly alter the ability of 11168-GS 42uC and 11168-O 25uC to bind Neu5Ac containing glycans (P.0.52, Figure 5A and B). In fact, the addition of sialyltransferase and sialidase inhibitors had no effect on any of the glycan binding profiles shown in figures 1-4 for both C. jejuni 11168-GS 42uC and 11168-O 25uC (P.0.18). In vitro Caco-2 cell adherence assays The specificity of glycan-mediated adherence of C. jejuni 11168 isolates were verified by in vitro Caco-2 binding assays following pre-treatment of cells with a competing lectin. The lectins selected were Erythrina cristagalli agglutinin (ECA; Galb1-4Glc/ GlcNAc.Gal), concanavalin A (ConA; Man), Ulex europaeus agglutinin 1 (UEA-1; terminal a1-2 Fuc), Limax flavus agglutinin (LFA; terminal sialic acid), Maackia amurensis agglutinin (MAA; Neu5Aca2-3.Gal) and Sambucus nigra agglutinin (SNA;-Neu5Aca2-6). The effect of pre-treating Caco-2 cells with selected lectins on C. jejuni 11168 adherence was determined relative to binding observed in the absence of lectin, which was set at 100%. As previously observed by Gaynor et al., (2004), we found that C. jejuni 11168-O (1.2% of inoculated bacteria) adhered to Caco-2 cells to a greater extent than 11168-GS (0.41% of inoculated bacteria) (P = 0.004) under standard assay conditions (see Materials and Methods).
We found that treatment with ECA reduced C. jejuni 11168 adherence to terminal lactose/Gal structures present on the surface of Caco-2 cells. C. jejuni 11168-GS 37uC adherence to Caco-2 cells pre-treated with ECA was reduced by greater than 85% (P = 0.02). No statistically significant difference was observed Table 2. Glycan structures present on array.
The affect of C. jejuni adherence to fucosylated structures on Caco-2 cells was determined using UEA-1. While no significant inhibition of adherence to UEA-1 pre-treated cells was observed for 11168-GS and 11168-O at 25uC (P.0.22, Figure 6A and B); 40-100% reduction in adherence of 11168-GS and 11168-O grown at both 37uC and 42uC was observed (P,0.05, Figure 6A and B).
Initial cell binding assays using LFA to neutralise all terminal Neu5Ac residues present on Caco-2 cells were performed to evaluate the involvement of Neu5Ac on in vitro adherence. Figure 6A shows that a significant decrease in the adherence to Caco-2 cells of 11168-GS 25uC (P = 0.02) and 11168-GS 37uC (P = 0.04) was observed, with a 75-100% reduction in comparison to non-treated cells. A reduction of 50% was also observed for 11168-GS 42uC adherence; however, this was not statistically significant (P = 0.09, Figure 6A). Based on these results the adherence of 11168-GS to Caco-2 cells was further assessed by pre-treating cells with other sialolectins, MAA and SNA (Neu5Aca2-3Gal.Gal and Neu5Aca2-6Gal-specific lectins respectively). Both MAA and SNA pre-treatment resulted in significantly reduced adherence of 11168-GS 25uC (P = 0.01 and 0.02, respectively) to Caco-2 cells ( Figure 6C). However, only MAA pre-treatment significantly reduced the adherence of 11168-GS 37uC to Caco-2 cells ( Figure 6A), suggesting that 11168-GS 37uC recognition of Neu5Ac on Caco-2 cells is a2-3 specific.
As expected, pre-treatment of Caco-2 cells with LFA significantly reduced the adherence of 11168-O 25uC (P = 0.02), while no significant reduction was observed for 11168-O 37uC (P = 1.0) and 42uC (P = 0.21) ( Figure 6B). Adherence assays using MAA and SNA as the neutralising lectin also revealed significant inhibition of 11168-O 25uC (P = 0.04 and 0.02, respectively), whereas no difference was observed with either MAA or SNA for 11168-O 37uC binding ( Figure 6B). Taken together, this suggests that 11168, both the original and genome sequenced isolates, maintained under environmental conditions bind Neu5Ac independent of linkage, whereas 11168-GS 37uC predominantly binds a2-3-linked Neu5Ac.
To further explore the significance of Neu5Ac in C. jejuni 11168 adherence to Caco-2 cells, the effect of Neu5Ac removal by sialidase treatment of Caco-2 cells was explored. Sialidase treatment of Caco-2 cells resulted in a 50% reduction in the adherence of both 11168-GS 25uC and 11168-O 25uC compared to non-sialidase treated controls (P = 0.03 and 0.02 respectively, Figure 7). Reduction in adherence to sialidase treated Caco-2 cells   was also observed for 11168-GS grown at 37u (40% reduction) and 42uC (12% reduction); however, only 11168-GS 37uC was significantly different to untreated controls (P = 0.02, Figure 7). The opposite was observed for the adherence of 11168-O 37uC and 42uC to sialidase treated Caco-2 cells with increases of 47% and 109%, respectively. However, only the increase observed for 11168-O 42uC was significantly different to normal adherence observed for untreated Caco-2 cells (P = 0.04).
Lectin competition combined with sialidase treatment (summarised in Table 3) provided independent verification of the glycan array data, with inhibition of C. jejuni adherence observed using lectins specific for Gal-, Fuc-, Man-and Neu5Ac-containing glycans and differential effects of sialidase treatment being observed between the C. jejuni 11168 isolates.

Discussion
Bacterial pathogens and commensal organisms both utilise adherence to host cell surface glycans for colonisation and/or initiation of disease [2]. In this study we demonstrate that glycan array technology represents a powerful tool for the identification and characterisation of novel bacteria-glycoconjugate interactions. Specifically, we were able to interrogate the carbohydrate binding potential of two C. jejuni isolates, the genome sequenced and noncolonising C. jejuni 11168-GS isolate, and the original virulent and colonising C. jejuni 11168-O isolate. Significantly, a number of glycans were shown to specifically bind both isolates in a manner dependent on growth/maintenance conditions, including Man, Fuc, Gal and Neu5Ac, with each verified using a cell-based adherence assays (summarised in Table 3).
C. jejuni 11168-GS and 11168-O bound terminal Gal structures independent of linkage and the nature of the sub-terminal carbohydrate, with the most significant binding observed at growth conditions that mimic those of the potential hosts. The specificity of Gal-binding by C. jejuni was confirmed using cell adherence assays, where a significant decrease was observed for 11168-GS grown at 37uC and 11168-O grown at 42uC following pre-treatment of Caco-2 cells with the lectin ECA. It should be noted that ECA binds lactose with greater affinity than other terminal Gal-glycans. Therefore, glycan structures similar to the array glycans 1B, 1G and 1H would be preferentially neutralised by ECA over other terminal Gal structures (e.g. 1B, 1C, 1D).
Gal-binding lectins are common throughout nature and are involved in cell-cell and cell-matrix adhesion, cell signaling, toxin binding and cytokine release [2,29,30,31,32]. Bacterial pathogens produce a range of Gal-specific lectins that are used in various ways to interact with the host tissues. Several bacterial toxins target Gal on cell surfaces [2,32], while other bacteria use Gal-binding lectins for adhesion, including enteroaggregative and enteropathogenic E. coli [33,34,35]. E. coli possesses an adhesin known as PapG, located on the flexible tip of P fimbriae, which binds Gala1-4Gal sequences present in the core of glycosphingolipids [36,37]. P. aeruginosa expresses a soluble Gal-specific lectin, PA-IL, which facilitates adherence to the intestinal epithelium and also leads to   increased permeability of the intestinal epithelial layer [4,38]. The nature of the lectin, and the function of Gal-binding by C. jejuni identified here awaits elucidation. However, our investigations suggest that C. jejuni NCTC11168 possesses either one Gal-specific lectin with broad substrate specificity able to bind both aand blinked Gal, or two highly specific lectins able to bind Gala1-3/ 4Gal and Galb-linked glycans separately that may be important for host (mammalian and avian) cell adherence. The finding that 11168-GS and 11168-O maintained at 25uC do not bind Gala1-3/4Gal sequences further suggests that this structure may be of particular importance in host adherence.
A previous study by Russell and Blake [39], showed that free Man was able to inhibit the adherence of C. jejuni to Caco-2 cells; however, the direct interaction of C. jejuni to Man has not previously been reported. Our present study has now demonstrated that C. jejuni 11168 is able to directly bind Man containing glycans in a specific manner. Interestingly, dramatic differences were observed between the ability of the GS and O 11168 isolates to bind Man. C. jejuni 11168-GS was able to bind a wide variety of Man structures independent of the growth/maintenance conditions. Conversely, C. jejuni 11168-O only bound Man after being maintained at 25uC under normal oxygen conditions. These results were confirmed by the ability of the Man-specific lectin ConA to inhibit the adherence of C. jejuni 11168 to Caco-2. Based on these findings it is therefore probable that C. jejuni 11168-O possesses a Man-binding lectin whose expression is dependent on the environmental temperature and/or atmospheric conditions encountered. Further, Man-recognition by C. jejuni 11168-O suggests that Man may be required for initial adherence to host tissues after environmental challenge, but may not be necessary for continued colonisation. Man is a ubiquitous component of Nglycan structures [40], used by many bacterial commensals and pathogens for adherence, such as uropathogenic E. coli that utilises the Man-specific lectin FimH located at the tip of type 1 fimbriae to bind host tissue [11,41]. The vast majority of Man binding lectins identified in bacteria are, like FimH, associated with fimbriae or pili, extracellular adhesins that C. jejuni is not known to possess [42,43]. While the Man-specific lectin of C. jejuni is unlikely to be of similar structure to previously identified fimbriae or pili lectins, the broad Man-binding specificity observed in C. jejuni is commonly observed in other bacterial species [41,44].
It is not surprising that C. jejuni 11168 grown under mammalian and avian host-like conditions recognized multiple fucosylated glycans, including Le antigens. Fucosylated glycoproteins are frequently found in the intestines of humans and other animals [45], with Le antigens representing a common fucosylated glycan.   Le antigens are classified as either type 1 (Le a and Le b ) or type 2 structures (Le x and Le y ) and are present, together with other fucosylated glycans, on mucins of the gastrointestinal tract [46]. Fucosylated glycans and free Fuc have been shown to interact with C. jejuni [24,26]. In fact, free Fuc and Fuc associated with the Oglycans of mucins are considered C. jejuni chemotaxis agents [24]. It is therefore likely that C. jejuni proteins involved in chemotaxis towards Fuc may be, at least in part, responsible for the binding of C. jejuni to fucosylated structures present on our glycan array. The ability of bacteria to bind to fucosylated host cell glycans has been linked to pathogenesis, with P. aeruginosa representing a well-characterised example. The P. aeruginosa Fuc-specific lectin (PA-IIL) has been structurally elucidated, and been shown to play a central role in Pseudomonas infection of immuno-compromised and cystic fibrosis patients [47]. PA-IIL facilitates the colonisation of the airways by adhering to the heavily fucosylated mucosal layer [47]. In C. jejuni, a mucosal pathogen, the importance of Fuccontaining glycoconjugates, particularly fucosylated glycoproteins found in human milk, was demonstrated by the ability of these structures to protect individuals against C. jejuni infection [48]. This protection against infection was attributed to the binding of C. jejuni to difucosylated terminal Le b structures and 2'fucosyllactose present on human milk glycoproteins [26]. The importance of this interaction is highlighted by our observation that 11168-GS and 11168-O were able to bind at least one of the terminal Le b or 2'fucosyllactose structures on our glycan array independent of the growth/maintenance conditions tested (Table 3, Figure 3A and 3B). Further, the complete inhibition of C. jejuni 11168-O 37uC adherence to Caco-2 cells by UEA-1 (terminal a1-2 Fuc binding lectin) confirmed the binding specificity of C. jejuni to terminal monofucosylated structures such as 2' fucosyllactose. It should be noted that C. jejuni 11168-O grown under mammalian host-like conditions was a poor binder of other non-Fuc capped glycans. However, the binding of lectins to a specific site on a cell surface glycoprotein may also result in the occlusion of other sites used by adhering bacteria, leading to a further decrease in binding to host cells.
The differential binding of fucosylated structures by C. jejuni grown or maintained under different conditions was also demonstrated by both the glycan array and UEA-1 pre-treatment of Caco-2 cells, with the most striking differences observed for both 11168-GS and 11168-O maintained at 25uC. Little or no binding was seen for both Le a and Le x under environmental conditions, with many other fucosylated structures showing reduced binding when compared to bacteria grown under hostlike conditions. UEA-1 pre-treatment of Caco-2 cells had no effect on the level of adherence of environmentally maintained 11168. These data suggest that environmentally maintained C. jejuni 11168 is a poor binder of terminally a1-2 fucosylated structures compared to the same bacteria grown under host-like conditions. These results not only highlight the importance of host glycan fucosylation for C. jejuni adherence to intestinal cells, but also that C. jejuni grown under host like conditions is likely to express a broad specificity Fuc-binding lectin that mediates this interaction.
Much like Man-recognition, binding to sialylated glycans was observed for C. jejuni 11168-GS regardless of growth/maintenance conditions and for 11168-O 25uC, with little or no binding observed for 11168-O when grown under conditions mimicking mammalian or avian host (Table 3, Figure 4A and 4B). Significantly, these differences between C. jejuni 11168-GS and 11168-O are not due to a modification of glycans by the action of a recently described C. jejuni sialyltransferase/sialidase activity [28], with no significant differences in Neu5Ac binding observed in the presence of sialyltransferase or sialidase inhibitors.
As proposed for Man-binding, Neu5Ac may be important for the initial contact between environmental C. jejuni and host tissues. Neu5Ac is another highly abundant carbohydrate on cell surface glycoconjugates, representing a common receptor for lectins expressed by many pathogenic bacteria species and viruses [49]. Cell surface mucins are typically highly sialylated, only becoming fucosylated in disease states such as cystic fibrosis where there is large turnover of cell surface components, including cell surface mucins [47,50]. The binding of C. jejuni to Neu5Ac-containing glycans present on our array was found not to be dependent on Neu5Ac-linkage or the nature of the sub-terminal sugar, regardless of the growth conditions tested and this was confirmed by LFA adhesion neutralisation assays. However, analyses using the Neu5Ac-linkage specific lectins SNA and MAA revealed selective specificity for a2-3-linked Neu5Ac, but only for C. jejuni 11168-GS grown at 37uC ( Figure 5A).
Sialidase treatment of Caco-2 cells further verified the role of Neu5Ac in the adherence of C. jejuni 11168 maintained under environmental conditions with significant reduction in adherence observed for 11168-GS 25uC and 11168-O 25uC. A significant reduction in the adherence of 11168-GS 37uC was also observed verifying the array and Neu5Ac-binding lectin competition data. Sialidase treatment of Caco-2 cells also affected adherence of 11168-O grown under host-like conditions; however, unlike 11168-GS an increase in adherence was observed. Caco-2 cells have been shown to express a2-3Neu5Ac, a2-6Neu5Ac, a1-2Fuc and Gal, with terminal Neu5Ac structures consistently being reported in all studies [51,52,53,54,55]. Treatment with sialidase, thus removing terminal Neu5Ac residues, would therefore unmask sub-terminal structures such as Le x , lactose and neo-lactose [51,52,53,55] enabling these glycans to be bound by C. jejuni. The observed increase in adherence of C. jejuni 11168-O 37uC and 42uC is therefore likely to be due to an increase in the number of these newly liberated glycans on the surface of Caco-2 cells following sialidase treatment. This is supported by the glycan array analysis that showed that Le x , lactose and neo-lactose were strongly bound by 11168-O grown under host like conditions (Table 3). These results indicate that 11168-O grown under hostlike conditions is more likely to adhere to niches of the gastrointestinal tract with lower levels of sialylated glycoconjugates in comparison to either 11168-GS or environmentally challenged 11168-O.
Throughout this study both isolates of C. jejuni 11168 bound type-1 (Galb1-3GlcNAc, Le a and Le b ) and type-2 (Galb1-4GlcNAc, Le x and Le y ) structures present on the array. This is of interest as type-2 glycans are ubiquitously distributed throughout tissues of the body, while type-1 glycans are expressed only within certain tissues including the digestive tract and pancreatic tissues [56]. The binding of type-1 glycans is to be expected as C. jejuni primarily infects/colonises the gastrointestinal tract of animals; however, in human disease C. jejuni is known to also cause systemic infections including meningitis and urinary tract infections. The ability of C. jejuni to bind type-2 glycans may be an important virulence factor in systemic C. jejuni infections.
With the exception of fucosylated structures such as Le b , very little was previously known about the glycans recognised by C. jejuni. Our study has now identified a number of additional carbohydrate structures commonly found on the mucins of the gastrointestinal tract that are bound by both the original C. jejuni NCTC11168 (11168-O) and genome sequenced (11168-GS) isolates. Glycan array and cell adhesion assays indicate that these newly identified glycan-bacteria interactions are important for adhesion and possibly colonisation of host tissues during both infection and commensal colonisation.
Differential binding between growth/maintenance conditions of 11168-O highlights an apparent adaptation of glycan binding potential from environmental to host-like conditions, with Gal/ Fuc binding being associated with host-like conditions and Man/ Neu5Ac being associated with environmental conditions (summarised in Table 3). This indicates that C. jejuni 11168-O has a preference for gastrointestinal niches within host tissues, with terminal Gal/Fuc being preferentially bound over Man/Neu5Ac. The most striking difference observed through our analyses was that the Man/Neu5Ac binding specificity of the C. jejuni 11168-O isolate varied significantly according to growth conditions, whereas this was not the case for the C. jejuni 11168-GS isolate. Much like 11168-O, C. jejuni 11168-GS grown under host conditions was observed to have increased binding to Gal/Fuc compared to environmentally maintained 11168-GS; however, 11168-GS grown under host like conditions still bound Man/Neu5Ac (Table 3). Therefore, given the data presented here, we postulate that the binding of Man/Neu5Ac may provide the initial interactions important for colonisation following environmental exposure. These interactions occur prior to bacteria adaptation through the expression of factors necessary for survival and growth in the avian and mammalian host. Presumably, changes in carbohydrate specificity from Man/Neu5Ac to Fuc/Gal through the expression of alternate lectins may then allow C. jejuni to adhere and colonise its preferred gastrointestinal niche.
Differential binding of glycans reported here, is likely to be one of several contributing factors to the differences in colonisation potential reported for the two C. jejuni 11168 isolates. [21]. Other factors contributing to the apparent lack of adaptation of C. jejuni 11168-GS to host-like conditions need to be considered, such as reduced motility of 11168-GS compared to 11168-O, and differences in the expression of genes involved in respiration (modulating C. jejuni's response to oxygen levels) and metabolism (modulating C. jejuni's response to temperature fluctuations) [21]. This report highlights the importance of further and detailed exploration of glycan-lectin interactions in the initiation and maintenance of C. jejuni colonisation and infection.

Preparation of SuperAmine Glass Slides
Propylamino-glass slides (SuperAmine2 glass slides, ArrayIt Microarray Technologies) were functionalised with a polyethylene glycol spacer moiety as previously described [57]. Glass slides were incubated for at least 18 h in a solution of Fmoc-8-amino-3,6dioxaoctanoic acid (10 mM), PyBOP (10 mM), diisopropylethylamine (20 mM) in dry DMF at room temperature in an anhydrous environment. After fuctionalization the slides were rinsed in DMF, the Fmoc protecting group removed by incubating the slides in piperidine (10% v/v) in DMF for 30 min, and the slides were subsequently rinsed in DMF. The slides were then activated with isocyanate functionality by incubation in a DMF solution of 1,6diisocyanatohexane (10% v/v) for 30 min at room temperature, rinsed once in DMF, once in THF, and centrifuged dry (1000 rpm for 2 min). The slides were stored under vacuum in desiccator for up to 1 week prior to use.

Preparation of Glycosylamines
Glycans were purchased from Dextra Laboratories (Reading, UK) and Glycoseparations (Moscow, Russia), and functionalised with an amine group using a modification of a previously described methodology by Vetter and Gallop [58]. One to five milligrams of each glycan and ammonium carbonate (1.3 equivalents) were dissolved in ,25% ammonium hydroxide (500 mL) in a sealed glass vial and incubated at 50uC for 3 days. The vials were unsealed, allowed to remain at 50uC for 2 h to dissipate excess ammonia and carbon dioxide, frozen and concentrated by lyophilisation. The glycosylamines were subsequently dissolved in anhydrous DMSO (100 mL), diluted further with an equal volume of anhydrous DMF and stored in sealed vials at 4uC until required.

Preparation of Glycan Arrays
Arrays were printed by the robotic non-contact dispensing (Piezorray Non-Contact Dispensing System, PerkinElmer) of 1500 pL of 10 mM glycosylamines in DMSO/DMF (1:1) onto isocyanate functionalised glass slides in replicates of four. Two identical sub-arrays were printed per slide with each sub-array comprising 352 spots (80 glycans and 8 controls in replicates of 4). Once the arrays had been printed they were placed in a vacuum desiccator fitted with a gas tap, which was evacuated under high vacuum. The vacuum was sealed off and a round-bottomed flask containing pyridine (,1 mL) attached via hosing and the desiccator was re-equilibrated with catalytic pyridine vapour for at least 8 h. The slides were then neutralised by incubating them in ethylene glycol (1 M) and pyridine (1% v/v) in DMF for 30 min at room temperature. The glycan array slides were rinsed in DMF, then ethanol and centrifuged dry (900 rpm for 2 min), at which stage they were ready for binding assays.
Bacterial Growth Conditions C. jejuni strain 11168 original (11168-O; Skirrow), obtained from Prof Diane Newell (Veterinary Laboratory Agency, London, UK), and genome sequenced strain (11168-GS; NCTC) were grown at either 37uC or 42uC for 36 h or 20 h, respectively (see Table 1), on Columbia agar supplemented with 5% defibrinated horse blood in microaerobic conditions (10% CO 2 , 5% O 2 in N 2 ). For the maintenance of 11168-O and 11168-GS under environmental conditions bacteria were first grown at 42uC for 20 h prior to being harvested in 10 mL of sterile fresh pond water and incubated at 25uC as described in Table 1. Gram-staining was performed as per standard procedures, and was used to assess the level of coccoid bacteria present following growth/maintenance at the conditions stated in Table 1. Only C. jejuni maintained under environmental conditions were used in further experiments when bacteria were found to be coccoid.

Labelling of Campylobacter jejuni
C. jejuni were harvested from agar plates using 1 mL of prewarmed Brucella broth and pelleted by centrifugation at 13,000 rpm for 1 min. The pellet was washed 5 times with prewarmed phosphate buffered saline pH 7.4 (PBS) to remove residual broth and agar, and diluted in PBS to 10 6 bacteria/ mL. To label 10 6 bacteria, 1 mL of diluted C. jejuni was pelleted by centrifugation at 13,000 rpm for 2 min and resuspended in 500 mL of 5 mM Carboxyfluorescein diacetate, succinimidyl ester (CFDA-SE, Molecular Probes) prepared in PBS and the labelling was performed using a modified version of that previously described [59]. Bacteria were then incubated protected from light either at room temperature for 90 min or at 37uC or 42uC for 30 min. Excess dye was removed by eight consecutive washes with 1 mL of PBS, with the final pellet being resuspended in 1 mL of PBS. Initial testing of CFDA-SE labelling was analysed with fluorescence activated cell sorting (FACS Calibur BD Scientific) with greater than 99% labelling observed. The extent of CFDAlabelling was routinely checked by fluorescence microscopy, with only bacterial samples observed to be greater than 90% labelling being used in subsequent experiments.

Binding Assays to Glycan Array
Each of the two sub-arrays printed per slide were isolated with a 1.762.8 cm (125 mL) Gene Frame (Abgene), into which 125 mL of CFDA-labelled bacteria in PBS was added. Studies of the effects of sialyltransferase and sialidase inhibitors were preformed in the presence of CMP, Neu5Ac2en and 4-guanidino-Neu5Ac2en (Zanamivir) at a concentration of 10 mM in PBS. Each sub-array was covered with a Gene Frame coverslip (Abgene) and incubated in a humidified chamber at 37uC for 30 min. Slides were washed with 40 mL of pre-warmed PBS flushed across the array with a transfer pipette. Duplicate arrays were performed using separate cultures of the bacterial strains under investigation.

Image Acquisition and Data Processing
Fluorescence intensities of array spots were measured using the ProScanArray Microarray 4-Laser Scanner (PerkinElmer) using the Blue Argon 488 excitation laser set to the FITC setting (494 excitation and 518 emission). Image analysis was carried our using the ProScanArray imaging software ScanArray Express (Perki-nElmer). Microsoft Excel was used to further process the data and to perform statistical analysis (Independent sample T test).

Caco-2 Cell Adherence and Lectin Neutralisation Assays
Adherence assays were performed using Caco-2 human intestinal cell lines. Cells were seeded at 10 5 cells/well in a 96well cell culture plates in minimal essential medium (MEM) for 48-72 h prior to bacterial challenge. Caco-2 cells were monitored prior to the assay to ensure a confluent monolayer of cells and that the cells had formed tight junctions. Sialidase pre-treatment of cells was carried out using 50 mU of Vibrio cholerae Sialidase (Sigma-Aldrich) for 2 h at 37uC in 5%CO 2 . Cells were pre-treated with the lectins UEA-1 (Sigma), ConA, ECA, MAA, SNA and LFA (EY Laboratories, Inc.) for 1 h, and subsequently washed 3 times with PBS. C. jejuni 11168, grown as described above, was added at a concentration of 5610 6 bacteria per well (viable counts were performed to verify the inoculums). Bacterial adherence was allowed to proceed for 1 h at 37uC under microaerobic conditions, with cells being subsequently washed three times with PBS. Caco-2 cells were then lysed with 0.1% Triton-X 100 in PBS and adherent bacteria enumerated by viable counts following plating onto Columbia agar with 5% horse blood and grown at 42uC for 36 h. All experiments were repeated a minimum of 3 times. Statistical analysis (independent sample Ttest) of viable bacterial counts was performed with Microsoft Excel.