Bacterial surface glycosylation is involved in many critical and diverse host-microbe interactions including adherence and immune modulation [1]. The analysis of bacterial surface glycosylation traditionally has been performed using nuclear magnetic resonance (NMR) spectroscopy, mass spectrometry (MS) and blotting based techniques [2-7]. Both NMR spectroscopy and MS are powerful techniques, but require relatively large quantities of highly purified glycan for structural analysis. Blotting methodologies, such as lectin blotting [7], have the advantage of only requiring small amounts of partially purified glycosylated protein or lipid for analysis, however only a single lectin interaction can be assessed per blot and the results can be ambiguous due to the difficulties in ensuring equimolar loading of test compounds into individual wells on the gel. In all cases these methodologies only provide low-throughput capabilities.

Lectin arrays consisting of diverse carbohydrate binding proteins covalently immobilised on glass microarray slides, have proven to be an important tool for evaluating cell surface glycosylation on whole bacteria and eukaryotic cells [1,8-10], eliminating the need to purify surface glycoproteins and lipids prior to analysis. However, assessing glycosylation on whole bacteria using lectin arrays may not be an accurate reflection of the bacterial glycosylation. Lectins are glycoproteins [11,12], therefore when immobilized on microarray slides they may themselves act as receptors for bacterial carbohydrate binding proteins. That is, bacterial glycan recognising adhesins [13- 17] may interact with carbohydrate structures present on immobilised lectins rather than lectins recognising bacterial glycoconjugates. In order to overcome this potentially significant limitation, we report here the analysis of semi-purified bacterial glycan using a discrete lectin array that requires only a small quantity of product for accurate, quick, and reproducible glycan structure determination. As the model systems for lectin array-based glycan structure determination we assessed both purified and crudely isolated C. jejuni lipooligosaccharide (LOS).

Campylobacter jejuni is a Gram-negative spiral rod bacteria that synthesizes both N-linked and O-linked glycans [18]. The strains of C. jejuni with that produce sialylated LOS (biosynthesis cluster types A, B, C, M and R) are capable of causing an autoimmune response against human gangliosides. This autoimmune response occurs in approximately 1:1000 people infected with C. jejuni leading to the development of the debilitating and life-threatening diseases Guillain- Barré syndrome (GBS) or Miller Fisher Syndrome (MFS) [19-21]. Our novel approaches to glycan structure identification permitted the elucidation of a clinically important ganglioside structures involved in molecular mimicry, the basis for the development of GBS.

Bacterial strains and growth conditions

The original isolate of C. jejuni NCTC 11168 (11168-O) previously characterized [22] was kindly supplied by D.J. Newell (Veterinary Laboratories Agency, Weybridge, UK). The human isolates C. jejuni 81-176 was donated by James G. Fox (Massachusetts Institute of Technology, Cambridge, Massachusetts, USA) and C. jejuni 224 was obtained from the Royal Melbourne Institute of Technology (Melbourne, Vic., Australia). C. jejuni were grown on blood agar, composed of Columbia agar containing 5 % (v/v) defibrinated horse blood and Skirrow's antibiotic supplement (Oxoid), under microaerobic conditions (5 % O₂, 10 % CO₂ and 85 % N₂) at 37°C for 48 h.

Lipooligosaccharide preparations

Crude LOS preparation: Blood agar-grown bacteria were harvested in 1 mL of sterile water, washed once in 1 mL of sterile water, and lysed by heating. Prior to lysis, samples were adjusted for bacterial number by OD₆₀₀ measurement of bacterial suspensions. Mini-preparations of LOS were prepared by treating the whole-cell extracts with proteinase K as described previously [23]. The LOS mini-preparations from single colonies were prepared by collecting and washing cells in 40 µL of sterile water followed by heat lysis. These preparations were diluted 10- fold prior to gel electrophoresis or lectin array analysis.

Purified LOS preparation: C. jejuni LOS was purified by subjecting the C. jejuni cell biomass to hot phenol-water treatment using 90 % (v/v) aqueous phenol at 65°C for 10 min [24], followed by enzymatic treatment as previously described [25]. The LOS preparations were adjusted to 15 µg/µL with distilled water prior to gel electrophoresis or lectin array analysis.

Electrophoretic analyses

Equal quantities of either LOS mini-preparations or purified LOS (~15 µg) were resolved on SDS-PAGE (5.5 % (w/v) and 10 % (v/v) stacking and separating acrylamide gels, respectively) containing 6 M urea and 0.3 mM tricine (Tricine-SDS-PAGE) as previously described [26]. Following electrophoresis at 20 V for 1 h to maximize stacking and then at 200 V for 30 min, gels were fixed and carbohydrate bands visualised by silver staining [27].

Lectin and western blotting

In addition to silver staining, fractionated C. jejuni LOS was electrotransferred from Tricine SDS-PAGE gels to Pall^® PVDF membranes using a Trans-Blot SD Semi-Dry Transfer Cell (Bio-Rad) at 30 V for 60 min. Membranes were subsequently probed using either horseradish peroxidase (HRP)-conjugated CTB (Cholera toxin subunit B) (3 µg/mL), HRP-conjugated PNA (Peanut Agglutinin) (5 µg/mL), or HRP-conjugated anti-G_M1 ganglioside IgG (diluted 1:3000) in PBS. Membranes were developed SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific) according to the manufacturer's instructions.

LOS biosynthesis cluster typing.

LOS biosynthesis clusters were typed using a PCR methodology based upon the published gene content of the LOS gene clusters [28,29]. Typing was performed using the primers and at annealing temperatures presented in (Table 1). Reactions were performed using 10 ρmoles of forward and reverse primer, 200 µM dNTPs, using the Phusion DNA Polymerase system (Finnzymes) to a final volume of 50 µL.

Table 1: Primers used in this study.

Lectin array preparation

Lectins were purchased from EY Laboratories and Sigma-Aldrich, and anti-ganglioside antibodies were purchased from Matreya (Table 2). Four identical sub-arrays per slide were printed essentially as previously described [30,31] onto epoxy functionalised glass slides (SuperEpoxy II; ArrayIt) at three different lectin concentrations (500 µg/ mL, 250 µg/mL and 125 µg/mL), and anti-ganglioside antibodies were printed at 1:100 to 1:10,000 dilutions across a 10-fold serial dilution. Printed slides were subsequently neutralised using ethanolamine as previously described [31]. Lectins and antibodies printed on our array and their published specificities are presented in Table 2.

Table 2: Results of the lectin array of phenol purified and crudely prepared LOS from C. jejuni strains NCTC11168, 81-176 and 224.

Lectin array analysis of C. jejuni LOS and whole C. jejuni

Purified and crudely isolated LOS was labelled using the lipophillic dye BODIPY^® TR methyl ester (BODIPY; Invitrogen). One mM BODIPY (6 µL) was added to 5 µg of LOS in 1x PBS containing 1mM CaCl₂ and 1mM MgCl₂ to a final volume of 30 µL and incubated at room temperature for 15 min prior to direct application to printed lectin arrays. Subarrays were separated using 25µL gene frames (Thermo Scientific). LOS was incubated on the arrays for 15 minutes, followed by 3 washes with 1x PBS containing 1mM CaCl₂ and 1mM MgCl₂. Controls for BODIPY and unlabelled LOS were also applied to the array and washed in the same way. Whole C. jejuni was labelled using BODIPY by the addition of 1 mM BODIPY to 10⁸ CFU of C. jejuni in 1xPBS and incubated at 37°C for 30 mins. Application of C. jejuni to the arrays and washing was performed as previously described [30].s Image acquisition and data processing was performed using the ProScanArray Microarray 4-Laser Scanner and the ProScanArray imaging software ScanArray Express from PerkinElmer as previously described [30]. AAnalysis was limited to presence or absence of binding to lectin spots across a 1:2 serial dilution rather then absolute binding levels. All positive binding spots were confirmed by visual inspection of the array and tested significantly above the background, which was confirmed by two-tailed unpaired T-test in Microsoft Excel. Lectin arrays were performed a minimum of twice per LOS sample.

The LOS of C. jejuni strain NCTC 11168 is known to mimic the human ganglioside structure, G_M1, and was therefore chosen to evaluate the viability of screening both phenol purified and crudely prepared LOS samples using our discrete lectin array. Initial analysis of the two alternatively isolated LOS samples from C. jejuni NCTC 11168 was performed using traditional SDS-PAGE silver staining and lectin blotting analysis (Figure 1). Silver staining revealed no differences in electrophoretic mobility between the phenol purified and crudely isolated LOS preparations (Figure 1A). Lectin blotting using HRP-conjugated PNA and CTB also revealed no differences in binding ability between sample preparations, with as expected [6] significant binding of both lectins (Figure 1B and 1C respectively). Taken together these data show that both LOS preparations have similar glycoconjugate components, specifically LOS that mimics G_M1 (as identified through PNA and CTB positive binding), and hence both LOS preparations provide an excellent tool for assessment of lectin array technology. Due to the fact that the C. jejuni LOS structure with molecular mimicry with human gangliosides been reported [6,32- 34], lectins were selected to generate a tailored array that would most effectively identify the individual carbohydrate motifs of structures containing monosialylganglioside mimicry (Table 2). The lectins selected included those known to bind the G_M1 mimic C. jejuni NCTC 11168 LOS (PNA and CTB) [6,34,35], and those assumed to bind based on the published structures of NCTC11168 and 81-176 [6,32], LFA (Limax flavus agglutinin), MAA (Maackia amurensis agglutinin), VAA (Viscum album agglutinin) and Jacalin (Jackfruit; Artocarpus integrifolia). Both DBA (Dolichos biflorus agglutinin) and ConA (Canavalia ensiformis agglutinin) were used as negative controls due to the absence of terminal GalNAc (recognised by DBA) and Man/ GlcNAc (recognised by ConA) in the published structures [6,32]. In addition, two specific anti-ganglioside antibodies were also included, one anti-G_M1 and one anti-G_M2 (Figure 2).

Figure 1: Silver-stained SDS-PAGE gel and Lectin blots of Crude and Phenol purified LOS preparations extracted from C. jejuni NCTC 11168. a). Silver stained LOS preparations. b). CTB blotting result for LOS preparations. c). PNA blotting result for LOS preparations. Lane 1 in each pane is Crude LOS preparation, while lane 2 is the phenol purified preparation.

Figure 2: Published structures of the LOS of C. jejuni NCTC 11168, 81- 176 and human GM1 and GM2. A). C. jejuni NCTC 11168 [6] and the expected binding for lectins/antibodies present on the array. B). C. jejuni 81-176 [32] and the expected binding for lectins/antibodies present on the array. N.B. Expected lectin binding is predicted from specificities listed in Table 2. Lower specificity interactions are also possible. Presented structures are the published majority structure for each strain [6,32]. C). Structure of GM1 (bolded sugar residues are identical to published C. jejuni 11168 structure shown in A). D). Structure of GM2 (bolded sugar residues are identical to published 81-176 structure shown in B).

Table 2 summarises the binding results obtained using our lectin array for both purified (11168 [P]) and crudely isolated (11168 [C]) LOS preparations. Even though binding of BODIPY labelled LOS to all lectins/antibodies with the exception of DBA, ConA and CTB was observed, some differences with respect to the level of binding was observed (Table 2; Figure 3). That is, three lectins were found to bind both LOS preparations down to 125 µg/mL (lowest concentration printed), the β-Gal recognising lectins PNA and VAA, and the a-Gal recognising Jacalin (structures recognised shown in Figure 2). The lack of LOS binding to DBA and ConA was anticipated, however the inability of either LOS preparation to binding immobilised CTB was surprising, particularly given that lectin blot analysis using HRPconjugated CTB showed significant binding (Figure 1B). Both antiganglioside antibodies were bound by both LOS preparations, however binding to anti-G_M2 was only observed at the highest concentration printed, while anti-G_M1 was bound down to the 1:1000 dilution (Table 2). This is not surprising given that the published C. jejuni LOS structure is known to be exclusively G_M1 under specified growth conditions [35]. The low binding therefore observed to anti-G_M2 may simply reflect some cross-specificity to the underlying G_M2 structure within the G_M1 mimic. In comparison, live C. jejuni 11168-O bound to all printed spots that contained protein including CTB (Figure 3C).

Figure 3: Analysis of live C. jejuni 11168 compared to purified LOS. A). Lectin array analysis of crude LOS extract prepared from C. jejuni 11168. B). Lectin array analysis of phenol purified LOS extract prepared from C. jejuni 11168. C). Lectin array analysis of live C. jejuni 11168. D). Graphical representation of the array analysis. Error bars equal one standard error of the mean. Any data not significantly different from background has been set to zero units. Lectins are listed in Table 2.

Further analysis was performed on LOS isolated using the crude method from other C. jejuni strains, one with a published LOS structure 81-176 (Structure shown in Figure 2) [32] and a strain without a published structure, C. jejuni 224. C. jejuni 224 was chosen for analysis as it has a LOS biosynthesis cluster (Class R; Table 3A) capable of producing molecular mimicry (Class A, B, C, R and M; Table 3B ) [28,29].

The use of lectin arrays to determine surface glycan structures on whole bacteria has been previously described [1-8]; however, to our knowledge, our approach of analysing fluorescently labelled isolated LOS on lectin array is completely novel, and offers a powerful analytical technique. This is particular the case, because the use of whole bacteria on lectin arrays may not accurately reveal the nature of the glycosylation present on the cell. As previously stated, lectins and antibodies are glycoproteins [11,12], therefore carbohydrate recognising adhesins present on bacteria [13-17] may bind these structures, rather than immobilised lectins binding to bacterial surface glycoconjugates. Testing of whole C. jejuni found binding to all lectins printed including CTB which had failed to bind in all other experiments (Figure 3; Table 2). Our novel approach of utilising isolated LOS overcomes this potentially significant shortcoming. Using the C. jejuni 11168 LOS as a model our discrete lectin array identified all carbohydrate components (Table 2) as predicted from the published structure (Figure 2).

The success of our approach depended largely on the dye used to label the isolated LOS. The lipophilic BODIPY TR methyl ester specifically interacts with the hydrocarbon tail of the LOS. The lack of binding observed to both DBA and ConA on the array indirectly confirms this selective labelling, given that any contaminating glycoconjugates (eg. glycoproteins) co-isolated with the C. jejuni LOS that could have been labelled with BODIPY would be expected to bind DBA immobilised on our array. That is, terminal α-D-GalNAc structures are common in N-linked C. jejuni glycoproteins [36], and if labelled, would have been detected on our lectin array.

Analysis was also performed on two other strains of C. jejuni for LOS molecular mimicry, 81-176 and 224. C. jejuni 81-176 was chosen because, like NCTC1168 [6], 81-176 has a published structure [32]. The results of the lectin array agreed with the known structure produced by 81-176, a G_M2 mimicking LOS [32]. C. jejuni 81-176 is known to produce several other structures including G_M3, GD1b and GD2, however, these structures are present in smaller amounts than the G_M2 mimicking structures [32]. Therefore it is unlikely these structures would be affecting the outcomes of the array analysis. A wider variety of anti-ganglioside antibodies may prove effective in identifying these less prevalent LOS structures.

C. jejuni 224 LOS analysed by lectin array indicated primarily G_M1 mimicry from the antibody binding but was not 100% confirmed by the binding observed for the other lectins present on the array. Levels of binding for PNA, VAA and Jacalin were lower than those seen for the known G_M1 mimic NCTC11168 (Table 2). However, the LOS structure produced by C. jejuni 224 was sialylated providing further evidence for gangloside mimicry. A previous study showed that the LOS from C. jejuni 224 was of the same size by electrophoresis as NCTC11168 and bound strongly by CTB [35]. This result together with the lectin array result implicates strongly that the LOS is a G_M1 mimic.

Although our lectin array data correlates well with the known C. jejuni 11168 and 81-176 LOS structures, a discrepancy between observed and expected binding results was noted. Specifically, no binding was observed to CTB immobilised on the lectin array, even though lectin blotting using CTB (Figure 1) showed a strong positive signal. This was a predictable outcome since CTB exists as a pentameric structure [37]. The complex pentameric structure of CTB may therefore be disrupted or constrained when covalently attached to the array through an epoxside-linkage. CTB is likely to be unsuitable for use on lectin arrays and its use restricted to lectin blotting analyses.

We also investigated the suitability of two different LOS preparations (phenol-purified and crudely isolated LOS) for analysis on our lectin array. Apart from a slight increase in non-specific background binding of the crude LOS preparation, compared to the phenol purified LOS (data not shown), no significant difference was observed using either LOS preparations. This suggests that a simple heat lysis and proteinase K digestion of C. jejuni or other bacteria is sufficient to allow rapid and sensitive screening using lectin arrays particularly of strains expressing ganglioside mimicry. The simple methodology reported here can, therefore, be used to rapidly evaluate whether clinical isolates have the potential to produce adverse autoimmune reactions as post infection sequelae, similar to those attributed to C. jejuni (GBS or MFS). Knowing whether clinically important bacterial strains are capable of inducing severe autoimmune responses may aid in prevention and/or early diagnosis of these debilitating post infection conditions.

The method reported here is also applicable, through minor modifications to the lectins/antibody specificities printed on the array, to the screen glycolipids from almost any other species of bacteria for almost any terminal glycoconjugate.