Brachyspira Species Avidity to Colonic Mucins from Pigs with and without Brachyspira hyodysenteriae Infection Is Species Specific and Varies between Strains

ABSTRACT

Brachyspira hyodysenteriae is commonly associated with swine dysentery (SD), a disease that has an economic impact on the swine industry. B. hyodysenteriae infection results in changes to the colonic mucus niche with massive mucus induction, which substantially increases the number of B. hyodysenteriae binding sites in the mucus. We previously determined that a B. hyodysenteriae strain binds to colon mucins in a manner that differs between pigs and mucin types. Here, we investigated if adhesion to mucins is a trait observed across a broad set of B. hyodysenteriae strains and isolates and furthermore at a genus level (B. innocens, B. pilosicoli, B. murdochii, B. hampsonii, and B. intermedia strains). Our results show that binding to mucins appears to be specific to B. hyodysenteriae, and within this species, the binding ability to mucins varies between strains/isolates, increases for mucins from pigs with SD, and is associated with sialic acid epitopes on mucins. Infection with B. hyodysenteriae strain 8dII results in mucin glycosylation changes in the colon, including a shift in sialic acid-containing structures. Thus, we demonstrate through hierarchical cluster analysis and orthogonal projections to latent structures discriminant analysis (OPLS-DA) models of the relative abundances of sialic acid-containing glycans that sialic acid-containing structures in the mucin O-glycome are good predictors of B. hyodysenteriae strain 8dII infection in pigs. The results emphasize the role of sialic acids in governing B. hyodysenteriae interactions with its host, which may open perspectives for therapeutic strategies.

INTRODUCTION

The gastrointestinal mucus layer is mainly composed of heavily glycosylated gel-forming mucins (1). Mucins can play a protective role against pathogens in a range of animal hosts (2). MUC2 is the predominant mucin expressed in the colon under normal or healthy conditions (3–5). However, mucin expression can change during certain parasitic and bacterial infections such as Salmonella enterica, Trichuris suis, and Brachyspira hyodysenteriae infections, resulting in the de novo production of MUC5AC, a mucin ubiquitous in the gastric mucosa (3, 6–8). In the pig colon, mucins predominantly carry N-acetylglucosamine (GlucNAc), N-acetylgalactosamine (GalNAc), fucose, and acidic structures, including sialylated and sulfated glycans (9, 10). The mucin O-glycans vary between pigs, with a predominance of mucin-type core 4 glycans (9).

Within the Brachyspira genus, several species can colonize pigs. The strongly hemolytic B. hyodysenteriae, B. suanatina, and B. hampsonii can cause severe mucohemorrhagic colitis (11–14). Some B. hyodysenteriae strains are weakly hemolytic, which has been associated with reduced virulence (15). B. innocens, B. pilosicoli, B. murdochii, and B. intermedia are considered weakly hemolytic, and infection can result in mild colitis and diarrhea (16–19). Additionally, B. pilosicoli can infect humans and is associated with intestinal spirochetosis (20).

The colonic pathogen B. hyodysenteriae is commonly associated with swine dysentery (SD), a mucohemorrhagic diarrheal disease that results in growth retardation and economic losses for the swine industry. The outcomes of infection with B. hyodysenteriae strain 8dII are a disorganized mucus layer structure with massive mucus induction characterized by increased expression of MUC2 and de novo synthesis of MUC5AC (3). Similar changes in mucin expression have also been reported in pigs challenged with B. hyodysenteriae strain B204 and B. hampsonii (21). We demonstrated that the increased expression of the local inflammatory mediators neutrophil elastase and interleukin 17 during in vivo infection with B. hyodysenteriae induces mucin production via mitogen-activated protein kinase 3 in vitro (22). Analysis of the mucin glycan profile during B. hyodysenteriae strain 8dII infection revealed changes in mucin glycosylation in the colon, resulting in a predominance of mucin-type core 2 and N-glycolylneuraminic acid (NeuGc) residues, in addition to shorter glycan chains likely attributed to faster mucin turnover (9). The changes in mucin glycosylation during infection are potentially favorable for B. hyodysenteriae, as mucins from pigs with SD enhance B. hyodysenteriae growth (23).

The massive mucus induction during B. hyodysenteriae infection increases bacterial adhesion epitopes on mucins (3). We have shown that B. hyodysenteriae strain 8dII adheres to colon mucins in a manner that varies between individuals and mucin types (3). The objectives of the present study were to determine if adhesion to mucins is a trait observed across B. hyodysenteriae strains and isolates and furthermore at a genus level. Our results show that adhesion to mucins appears to be specific to B. hyodysenteriae, and within this species, the adhesion ability to mucins varies between strains and increases to mucins from pigs with SD. Furthermore, bacterial adhesion was associated with sialic acid epitopes on mucins, pointing to their relevance in host-pathogen interactions.

RESULTS AND DISCUSSION

The strain-specific binding ability of B. hyodysenteriae is increased to colonic mucins from infected pigs.

We previously showed that B. hyodysenteriae strain 8dII binds to pig colon mucins in a manner that differed between individuals and mucin populations and increased with B. hyodysenteriae infection (3). However, it is unknown whether other B. hyodysenteriae strains bind to pig colon mucins and/or if strain specificity plays a role in adhesion to mucins. Here, we investigated the binding ability of a broad set of B. hyodysenteriae strains and isolates (n = 10) to mucins isolated from healthy control pigs (pigs A to F) and pigs with SD (pigs 1 to 5). The pigs were experimentally infected with B. hyodysenteriae strain 8dII, excreted B. hyodysenteriae in feces, had mucohemorrhagic diarrhea, and overexpressed MUC2 and MUC5AC (Table 1) (3). Pig colon mucins were isolated by isopycnic density gradient centrifugation, which separates the highly glycosylated (i.e., high-density) mucins from the less glycosylated nonmucin molecules. The resulting material was classified as soluble and insoluble in guanidinium hydrochloride (GuHCl), with the latter being further solubilized by reduction in dithiothreitol (DTT) but referred to as “insoluble” mucins.

TABLE 1.

Summary of colon mucins isolated from healthy and B. hyodysenteriae-infected pigs^d

Exptl group and pig ID	Presence of clinical signs of SD and B. hyodysenteriae excretion in feces	Overall mucin glycan content isolateda,b (Eu counts)	Fold change in MUC2 mucin quantity based on ELISAb,c		Fold change in MUC5AC mucin quantity based on ELISAb,c
			Soluble	Insoluble	Soluble	Insoluble
Control
A	No	5.23 × 108	1.3	1.2	1.5	0.9
B	No	1.45 × 109	6.7	4.1	3.2	1.2
C	No	5.92 × 108	1.6	0.8	1.2	0.6
D	No	1.16 × 109	0.6	0.7	0.8	1.1
E	No	2.39 × 108	0.7	0.2	0.6	0.2
F	No	1.14 × 109	0.4	1.9	0.6	1.7
B. hyodysenteriae infected
1	Yes	2.74 × 109	10.4	4.7	1.5	1.2
2	Yes	4.19 × 109	8.1	3.9	11.9	4.0
3	Yes	6.69 × 109	7.4	2.9	21.4	3.2
4	Yes	2.74 × 109	10.3	1.9	9.2	1.9
5	Yes	3.46 × 109	14.7	5.3	4.1	3.3

^aOverall pig mucin glycan content in the colon, including GuHCl-soluble and -insoluble mucins. Mucins were isolated by isopycnic density gradient centrifugation, and mucins insoluble in GuHCl were solubilized by reduction and alkylation, breaking the disulfide bonds that link the mucin monomers into large complexes (3). The glycan content in the mucin samples was determined using a microtiter-based assay detecting carbohydrates as periodate-oxidizable structures.

^bSignificantly different in infected pigs compared to controls (P < 0.05).

^cCompared to the median for noninfected controls.

^dSee reference 3. ELISA, enzyme-linked immunosorbent assay.

All B. hyodysenteriae strains and isolates were isolated from pigs with diarrhea, and the majority of them were strongly hemolytic, except for the moderately hemolytic strain M2 and the weakly hemolytic strain D28 (24). Half of the tested B. hyodysenteriae strains/isolates (n = 5), namely, 6bI, 10cI, D1, M1, and M2, bound to soluble and insoluble mucins isolated from B. hyodysenteriae-infected and control pigs (Fig. 1). B. hyodysenteriae B204, 21cI, D28, 4cI, and WA-1 displayed a binding signal not statistically different from the background signal and therefore were considered non-mucin-binding strains (Fig. 2). Of the mucin-binding B. hyodysenteriae strains/isolates, 6bI and M2 presented a binding signal per glycan unit of a magnitude similar to that of the previously tested strain 8dII (Fig. 1A and M), whereas strains 10cI, D1, and M1 presented a binding signal ≤2-fold lower than that of strain 8dII (Fig. 1D, G, and J) (3). Although the magnitude of the binding signal differed from that of strain 8dII, the binding pattern of the mucin-binding B. hyodysenteriae strains/isolates strongly correlated with the binding pattern of strain 8dII (Table 2). As mucin glycans provide a repertoire of epitopes for bacterial adhesion (25), and bacterial surface adhesins are primarily involved in glycan recognition, the mucin-binding B. hyodysenteriae strains/isolates potentially carry a common adhesin recognizing a specific glycan structure(s) present in the pig colon mucins.

FIG 1.

Binding of B. hyodysenteriae strains and isolates to pig colonic mucins. (A, D, G, J, and M) Binding pattern of B. hyodysenteriae strains/isolates 6bI, 10cI, D1, M1, and M2 to GuHCl-insoluble and -soluble mucins isolated from control pigs (pigs A and C to F) and B. hyodysenteriae-infected pigs with clinical signs of SD (pigs 1 to 5). (B, E, H, K, and N) Overall binding of B. hyodysenteriae strains/isolates (6bI, 10cI, D1, M1, and M2) to insoluble and soluble mucins isolated from control and B. hyodysenteriae-infected pigs. Bacterial binding to mucins was determined when there was a statistically significant difference between the signal for the assay and the background signal. After subtracting the background signal, the bacterial binding signal was normalized against the glycan value for that particular mucin, and data are expressed as median relative light units (RLU) per 1,000 glycan signals (Eu counts) with interquartile ranges (IQRs). *, P < 0.05; **, P < 0.01 (by a Mann-Whitney U test). (C, F, I, L, and O) Binding ability of B. hyodysenteriae strains/isolates 6bI, 10cI, D1, M1, and M2 relative to the total mucin content observed in pigs with SD (pigs 1 to 5) and control pigs (i.e., binding to mucin at a set concentration × the total amount of mucin recovered from that sample). Results are expressed as medians with interquartile ranges. *, P < 0.05; **, P < 0.01 (by a Mann-Whitney U test). The data shown are representative of results from at least two independent experiments.

FIG 2.

B. hyodysenteriae strains/isolates without binding ability to pig colonic mucins. Binding signals of B. hyodysenteriae strains and isolates 4cl, D28, 21cl, WA-1, and B204 to GuHCl-insoluble and -soluble mucins isolated from B. hyodysenteriae-infected pigs (pigs 1 to 5) and control pigs (pigs A to F) are shown. Results are expressed as mean relative light units (RLU) ± SEM. The dotted line shows the mean background signal. Data were analyzed by one-way ANOVA with Dunnett’s correction for multiple comparisons. The data shown represent results from at least two independent experiments.

TABLE 2.

Associations between mucin-binding B. hyodysenteriae strains and isolates^b

Comparison of binding B. hyodysenteriae strains/isolatesa	r	P value	CI
8dII vs 6bI	0.77	0.0002	0.48–0.91
8dII vs 10cI	0.69	0.0017	0.32–0.87
8dII vs D1	0.64	0.0045	0.24–0.85
8dII vs M1	0.56	0.0154	0.13–0.82
8dII vs M2	0.74	0.0004	0.42–0.89

^aComparisons to B. hyodysenteriae 8dII binding data from a previous study (3).

^bPearson’s product-moment correlation coefficient (r) was used to determine the relationship between the binding of B. hyodysenteriae strain 8dII and strains/isolates 6bI, 10cI, D1, M1, and M2 to pig mucins. CI, confidence interval.

Despite the differences in the magnitude of the binding signal, the adhesion of all five mucin-binding B. hyodysenteriae strains/isolates increased to the MUC5AC-rich soluble mucins isolated from pigs with SD compared to controls (P < 0.05) (Fig. 1B, E, H, K, and N). No binding was detected to the MUC5AC-rich mucins with the strains that were designated non-mucin-binding strains (Fig. 2). These results are in line with our observations of B. hyodysenteriae strain 8dII (3). On the contrary, statistically significant differences were not observed in bacterial binding to the MUC2-rich insoluble mucins isolated from pigs with SD versus controls (Fig. 1B, E, H, K, and N). When considering the total mucin content isolated from the experimental pigs (i.e., soluble and insoluble), the total binding ability of the B. hyodysenteriae strains/isolates to infected pig mucins increased between 6- and 8-fold (P < 0.05) (Fig. 1C, F, I, L, and O), compared to a 7-fold increase for strain 8dII (3). Thus, differences in B. hyodysenteriae binding reflect changes in mucin production and the mucin glycan fingerprint during SD (9), which in turn increase the availability of epitopes for bacterial binding and ultimately influence bacterial interactions with host mucins. However, it is unknown if B. hyodysenteriae adhesion to mucins benefits the host by preventing further colonization of the intestinal epithelium, as is the case for the gastric pathogen Helicobacter pylori (26, 27). Individuals with mucins with higher H. pylori avidity are associated with a lower pathogen burden in rhesus monkey stomach (27), and mucin binding has been shown to act as a releasable decoy limiting H. pylori colonization (26). Instead, the changes in the mucin glycan repertoire together with the aberrant mucin induction during SD point to a benefit for the survival of B. hyodysenteriae in the colonic mucus layer, as it can utilize mucins from infected pigs as a nutrient source for growth (23).

Species-specific binding of Brachyspira to pig colon mucins.

Since there is a lack of information regarding the interaction of weakly hemolytic Brachyspira species with intestinal mucins, we investigated the adhesion ability of B. innocens, B. pilosicoli, B. murdochii, B. hampsonii, and B. intermedia to colonic insoluble mucins isolated from pigs with SD and healthy controls. Our results show that the binding signals displayed by B. intermedia, B. innocens, B. murdochii, B. hampsonii, and B. pilosicoli were not statistically different from the background signal (Fig. 3A to K). Although we tested fewer strains/isolates of these species, it is plausible that bacterial binding to mucins is a trait exclusive to B. hyodysenteriae.

FIG 3.

Nonbinding Brachyspira spp. to pig colonic mucins and B. hyodysenteriae chemotaxis toward fucose. (A to K) The binding signals of Brachyspira intermedia isolates DGZ13, DGZ16, and DGZ19 (A to C); B. innocens isolates 13bI, 14aI, and 7bI (D to F); B. pilosicoli isolates DGZ6, DGZ7, and DGZ11 (G to I); B. murdochii isolate DGZ27 (J); and B. hampsonii isolate DGZ52 (K) to GuHCl-insoluble mucins isolated from B. hyodysenteriae-infected pigs (pigs 3 and 4) and control pigs (pigs A and C to E) were not different from the background signal. Analysis was performed with a subset of representative mucin samples due to the limited amount of material. Results are expressed as the mean relative light units (RLU) ± SEM for panels A to I and as the median RLU with interquartile ranges for panel J, due to a lack of normality for the latter data set. The dotted line shows the mean background signal. Results were analyzed by one-way ANOVA with Dunnett’s correction for multiple comparisons. The data shown are representative of results from two independent experiments. (L) B. hyodysenteriae strain/isolate chemotaxis toward 0.1 M fucose. The R_che value is the ratio of the median number of B. hyodysenteriae cells per milliliter in a capillary tube with fucose in anaerobic PBS/median number of B. hyodysenteriae cells per milliliter in a capillary tube with anaerobic PBS only. Results are expressed as medians with interquartile ranges. *, P < 0.05; **, P < 0.005 (by a Mann-Whitney U test, compared to the PBS control). Experiments were performed in triplicates, and the data shown are representative of results from at least two independent experiments per strain/isolate.

The binding of B. hyodysenteriae strains to pig colon mucins is associated with the presence of sialic acid on mucins.

B. hyodysenteriae infection with strain 8dII has been demonstrated to regulate mucin glycosylation, resulting in changes to the colonic mucin O-glycome (9). The mucin glycan structures characterized in pig colon mucins by mass spectrometry depicted greater glycan diversity among infected pigs than among noninfected pigs (9). In order to pinpoint glycan structures likely involved in B. hyodysenteriae adhesion to colonic mucins, we analyzed potential associations between the pig mucin glycan data set (obtained from B. hyodysenteriae-infected and healthy pig mucins) (9) and the mucin-binding data for B. hyodysenteriae strains/isolates 6bI, 10cI, D1, M1, and M2.

In line with B. hyodysenteriae strain 8dII (23), for all five strains/isolates, there was a statistically strong link between binding to mucins and the relative abundance of structures carrying α2,3 sialic acid (r_s = 0.8; P < 0.05), including NeuGcα2,3Galβ1,3(GlcNAcβ1,6)GalNAc (r_s = 0.8; P < 0.007). The higher abundance of N-glycolylneuraminic acid (NeuGc) found on mucins from pigs with SD and its association with bacterial binding could potentially explain the increased adhesion ability observed for B. hyodysenteriae strains to mucins from infected pigs. Our results suggest a common interaction across mucin-binding B. hyodysenteriae strains with sialic acid residues on the mucins. Moreover, we have validated sialic acid as an important adhesion epitope in B. hyodysenteriae interactions with mucins, as the removal of sialic acid inhibits B. hyodysenteriae strain 8dII adhesion to mucins (23). Orthogonal projections to latent structures (OPLS) models constructed to predict bacterial binding based on the relative abundance of mucin glycan structures indicate that structures containing sialic acid contribute substantially to binding (see Fig. S1 to S3 in the supplemental material). However, the binding models did not have a good predictive ability, which we defined as achieving a Q2Y value of >0.5. This may be due to the small sample size and because the mucins carry 5 to 20 sialic acid-containing structures, several of which may contribute to B. hyodysenteriae binding. Although blood group antigens are known to act as receptors for some pathogens, i.e., the H. pylori surface adhesin BabA binds to the Lewis b blood group antigen on gastric mucins (22), no associations between B. hyodysenteriae binding to mucins and the abundance of blood group antigens were established. Thus, it is unlikely that blood group antigens play a role in B. hyodysenteriae adhesion to intestinal mucins. For some pathogens, such as H. pylori, mucins with the ability to bind pathogens limit the number of bacteria in contact with the epithelium by acting as decoys hindering intimate interactions with the epithelial cells or aggregating the bacteria (26, 28). Whether adhesion to sialic acid structures plays a protective role against bacterial colonization of the intestinal epithelium remains unknown.

The sialic acid profiles differ between B. hyodysenteriae-infected and noninfected colons.

Although there were no differences in the overall sialylation levels of the mucin O-glycans from infected and noninfected groups, the NeuGc contents differed (9). Hierarchical cluster analysis of the relative abundances of sialic acid-containing glycans clustered the pigs based on infection status, which is visualized with a heat map (Fig. 4). Meanwhile, hierarchical cluster analysis using glycans without sialic acid did not cluster pigs as decisively according to infection status (Fig. S4). Furthermore, the OPLS discriminant analysis (OPLS-DA) model, which discriminates between samples from pigs with SD and control samples based on sialic acid-containing glycan structures, has a high predictive ability (Fig. 5A and Fig. S5). This model is better than the OPLS-DA model using all glycan structures (Fig. 5B and Fig. S6) or glycan structures without sialic acid (Fig. S7). Altogether, these results point to the relevance of sialic acid-containing glycans as good predictors of SD.

FIG 4.

Relative abundance of colonic mucin glycan structures containing sialic acid. Each cell in the heat map represents the mean relative abundance of soluble and insoluble mucin samples for one glycan in one pig. Each row represents one glycan drawn to the left according to the symbol nomenclature for glycans (SNFG). The dendrogram on top of the heat map shows the hierarchical agglomerative cluster analysis of pigs with SD and controls, with the complete linkage method according to Euclidean distance.

FIG 5.

OPLS-DA models for discrimination of B. hyodysenteriae infection status based on the relative abundance of mucin glycan structures. (A) Model built using only mucin glycan structures containing sialic acids. (B) Model built using all mucin glycans. (A1 and B1) Bar charts of model components. The OPLS-DA models consist of one predictive component (p1) and one or more orthogonal components (o1, o2, and o3). The orthogonal component(s) captures systematic variation unrelated to the response variable. R2Y shows how much of the variation in the response (infection status) that the component explains, and Q2Y shows how well the model is predicted to explain the response variation on a new data set based on cross-validation. For this study, a Q2Y value of >0.5 is considered to indicate a good predictive ability for the predictive component. (A2 and B2) Scatterplots of component scores. The component scores for each mucin sample in the predictive component, t1, are plotted against the component scores of the first orthogonal component, to1, in order to visualize how well the model predicts infection status based on the relative abundances of glycans. The percentage of the variation captured by the predictor component t1 is given in parentheses.

B. hyodysenteriae chemotaxis to fucose is independent of mucin-binding ability.

Bacterial motility and chemotaxis play an important role in pathogenicity and colonization of the host. B. hyodysenteriae carries a large number of genes involved in motility and chemotaxis, allowing this bacterium to penetrate the mucus layer and adapt to its intestinal habitat (29, 30). Previous studies have shown chemotaxis to mucins and their components, among which fucose is a strong chemoattractant for B. hyodysenteriae (31–33). As 78% of the neutral glycans present in the pig colon mucins were fucosylated, and their relative abundance was >50% in soluble and insoluble mucins (9), we analyzed the chemotactic ability of B. hyodysenteriae strains/isolates toward fucose. Chemotaxis toward fucose was observed for the mucin-binding strains/isolates 8dII, 10cI, M1, and M2 as well as for the non-mucin-binding strain/isolate D28 and positive-control strain B204 (Fig. 3L). Thus, no associations were established between mucin-binding and non-mucin-binding B. hyodysenteriae strains or isolates concerning chemotaxis toward fucose.

Conclusion.

After testing a range of Brachyspira hyodysenteriae strains and isolates, as well as several Brachyspira species (with exception of B. suanatina) associated with SD, our results suggest that binding to colonic mucins appears to be specific to B. hyodysenteriae, with varying binding abilities between strains/isolates. Intimate adhesion to the host is likely to be affected by the mucin-binding abilities of the strain, and therefore, it is possible that host responses to B. hyodysenteriae may differ between strains. We demonstrate that the repertoire of sialic acid-containing structures can be used to discriminate between samples from pigs with SD and control samples using both supervised and unsupervised machine learning techniques. We highlight the effect of massive mucin induction together with mucin glycosylation changes, including a shift in sialic acid-containing structures, during B. hyodysenteriae infection on bacterial binding, as adhesion was higher to mucins from pigs with SD and associated with sialic acid structures on the mucins.

MATERIALS AND METHODS

Experimental inoculation and sample collection.

Information regarding the experimental inoculation of pigs and sample collection procedures was reported previously by Quintana-Hayashi et al. (3). Briefly, five pigs (pigs 1 to 5) presented clinical signs of swine dysentery after oral inoculation with brain heart infusion (BHI) broth containing 10⁸ CFU of B. hyodysenteriae strain 8dII, while control pigs (pigs A to F) received sterile BHI broth. Infection was confirmed by clinical signs of mucohemorrhagic diarrhea and B. hyodysenteriae excretion in feces detected by quantitative PCR (qPCR). The pigs were sacrificed at day 40 postinoculation by anesthesia with a combination of xylazine at 4.4 mg/kg of body weight (Xyl-M 2%; VMD, Arendonk, Belgium) and zolazepam/tiletamine at 2.2 mg/kg (Zoletil 100; Virbac, Carros, France) and final euthanasia by intracardial injection of a formulation comprising embutramide, mebezonium iodide, and tetracaine (T61; Intervet, Brussels, Belgium) at 0.3 ml/kg. Midsection samples of the spiral colon of the five infected pigs and healthy controls were collected for mucin isolation. Fecal material was removed, and the tissues were rinsed with phosphate-buffered saline (PBS) containing a protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany), followed by snap-freezing and storage at −80°C.

Isolation and purification of pig colon mucins.

Mucin isolation from colon tissue samples was performed by isopycnic density gradient centrifugation as previously described, obtaining guanidinium hydrochloride (GuHCl)-soluble and -insoluble mucins (3). Briefly, frozen tissues were drenched with 10 mM sodium phosphate buffer (pH 6.5) containing 0.1 mM phenylmethylsulfonyl fluoride (AppliChem, Darmstadt, Germany). Once thawed, the mucosal surfaces were scraped with a microscope slide, dispersed with a Dounce homogenizer, and stirred slowly overnight at 4°C in ice-cold extraction buffer consisting of 6 M GuHCl (AppliChem, Darmstadt, Germany), 5 mM EDTA (Sigma-Aldrich, St. Louis, MO, USA), 5 mM N-ethylmaleimide (Alfa Aesar, Karlsruhe, Germany), and 10 mM sodium dihydrogen phosphate at pH 6.5. GuHCl-soluble mucins were obtained after centrifugation at 23,000 × g for 50 min at 4°C, and the remaining material was reextracted twice by stirring overnight at 4°C in extraction buffer. The remaining pellets contained the “insoluble” mucins, which were solubilized with 10 mM dithiothreitol (DTT) in reduction buffer (6 M GuHCl, 5 mM EDTA, 0.1 M Tris-HCl [pH 8]) for 5 h at 37°C. Finally, residues were alkylated overnight with 25 mM iodoacetamide (IAA; Alfa Aesar, Karlsruhe, Germany).

Both the GuHCl-soluble and -insoluble materials were dialyzed in 10 volumes of extraction buffer at 4°C, changing the dialysis solution three times in 24 h. Isopycnic density gradient centrifugation in cesium chloride (CsCl)–4 M GuHCl with a starting density of 1.39 g/ml was performed at 40,000 rpm for 90 h. The mucin-containing fractions were pooled and further purified from DNA by a second gradient in CsCl–0.5 M GuHCl. Approximately 25 mucin fractions were recovered per sample using a fraction collector equipped with a drop counter. Fractions were stored at 4°C until further analysis.

Analysis of mucin fractions.

The first and second CsCl gradient mucin fractions were analyzed as previously described (3). The glycan content in the GuHCl-soluble and -insoluble mucin samples was determined using a microtiter-based assay detecting carbohydrates as periodate-oxidizable structures. Nunc 96-well plates (Thermo Scientific, Waltham, MA, USA) were coated overnight at 4°C with mucin fractions diluted in 4 M and 0.5 M GuHCl. Plates were incubated with a 25 mM sodium metaperiodate solution diluted in sodium acetate (NaAc) for 20 min and blocked with a solution containing 50 mM Tris-HCl, 0.15 M NaCl, 90 μM CaCl₂, 4 μM EDTA, 0.01% NaN₃, and 2% bovine serum albumin at pH 8 for 1 h. The wells were then incubated for 1 h with a biotin hydrazide solution diluted 1/50 in NaAc, followed by europium-labeled streptavidin diluted 1/1,000 in Delfia assay buffer (PerkinElmer, Waltham, MA, USA). Finally, plates were incubated with Delfia enhancement solution for 5 min on a shaker. Between each step, the plates were washed three times with a solution containing 5 mM Tris-HCl, 0.15 M NaCl, 0.005% Tween 20, and 0.01% NaN₃ at pH 7.75, except for the final step, where plates were washed six times. The signal was measured in a Wallac 1420 Victor² microplate reader (PerkinElmer, Waltham, MA, USA) by time-resolved fluorometry.

Mucin sample preparation and concentration estimation.

Mucin sample preparation and concentration estimation were performed as previously described (3). Briefly, gradient fractions containing mucins were pooled to obtain one sample for each gradient. Mucin concentrations in pooled samples were determined by serial dilutions as well as a standard curve of a fusion protein of the mucin MUC1, 16TR, and IgG2a Fc (34), starting at a concentration of 20 mg/ml and using seven 1/2 serial dilutions in a carbohydrate detection assay as described above. The mucin concentrations were calculated from the standard curve. Setting the concentration based on the glycan content appears most appropriate as bacterium-mucin interactions largely occur via the mucin glycans (1). Although this is not an absolute measure of concentration, it can be used to ensure that the mucins are at the same concentration for comparative assays.

Brachyspira species and B. hyodysenteriae isolates and strains.

B. intermedia isolates DGZ13, DGZ16, and DGZ19; B. innocens isolates 13bI, 14aI, and 7bI; B. pilosicoli isolates DGZ6, DGZ7, and DGZ11; B. murdochii isolate DGZ27; B. hampsonii isolate DGZ52; and B. hyodysenteriae isolates 6bI and 21cI and strains 10cI, D1, M1, M2, D28, and 4cI were isolated from pig fecal samples. More information regarding these specific B. hyodysenteriae strains was reported previously by Mahu et al. (24). B. hyodysenteriae strains WA-1 and B204 are available from the ATCC. The bacterial strains and isolates were cultured on tryptone soy agar (Thermo Fisher Scientific, Waltham, MA, USA) plates supplemented with 5% sheep blood (Thermo Fisher Scientific, Waltham, MA, USA), 0.1% yeast extract (Merck, Darmstadt, Germany), 400 μg/ml spectinomycin, 25 μg/ml colistin, and 25 μg/ml vancomycin (AppliChem, Darmstadt, Germany) and incubated at 40°C under anaerobic conditions.

Bacterial adhesion to pig colon mucins.

White 96-well plates (Corning Life Sciences, NY, USA) were coated overnight at 4°C with 6 μg/ml mucins in 0.5 M GuHCl. Wells were washed three times with PBS containing Tween (PBS-T) and blocked with 200 μl of 5% fetal bovine serum (FBS) for 1 h. Bacteria were harvested from agar plates, washed in PBS, centrifuged at 2,500 × g for 5 min, and resuspended in PBS with 5% FBS. One hundred microliters of a bacterial suspension diluted to10⁸ bacterial cells/ml was added per well, and the plates were shaken during incubation for 2 h at 40°C in an anaerobic environment. Plates were washed three times with PBS-T and once with PBS. Subsequently, 100 μl of PBS was added to each well, followed by the addition of an equal volume of BacTiter-Glo reagent (Promega, Madison, WI, USA). Incubation proceeded for 5 min at room temperature. Relative light units (RLU) were measured in an Infinite M200 microplate reader (Tecan, Männedorf, Switzerland) with an integration time of 1,000 ms per well. Controls included wells without bacteria (PBS only) in mucin-coated wells and non-mucin-coated wells incubated with the bacterial suspension followed by the addition of PBS and the reagent. Bacterial binding to mucins was determined when there was a statistically significant difference between the binding signal for the assay and the background signal. In order to confirm that differences in bacterial binding to pig mucins were not due to variations in the mucin glycan contents between samples, a glycan detection assay as described above was simultaneously coated and performed with the binding experiments. Only samples with glycan values of 17,000 to 24,000 Eu counts were included to ensure that the analysis occurred within the linear range of the assay; for this reason, insoluble mucin sample B was excluded from the analyses. For strains that were identified to bind to mucins, the bacterial binding signal was normalized against the glycan value for that particular coating and mucin (represented on the y axis as RLU/1,000 Europium [Eu] counts). For the nonbinding strains, the data were not normalized against the glycan values, and the background data were not subtracted from the overall signal and instead are shown in the graphs as a dotted line to clearly show the basis for why these strains/isolates are considered nonbinders. Results were obtained from at least two independent experiments with five technical replicates for each mucin and plotted as relative luminescence per glycan unit.

Chemotaxis toward fucose.

For the chemotaxis assay, fresh cultures of B. hyodysenteriae strains/isolates 8dII, 10cI, M1, M2, 21cI, D28, 4cI, and B204 (positive control) were prepared by harvesting a 4-day-old culture plate with a sterile cotton swab and stirring the cotton swab in anaerobic BHI broth (Bio-Rad, Hercules, CA, USA) supplemented with 10% FBS (Thermo Fisher Scientific, Waltham, MA, USA). The bacterial cultures were incubated for 48 h under anaerobic conditions at 37°C on a rocking platform, and for each strain/isolate, three cultures were made. After incubation, cultures were microscopically examined for purity, and the optical density at 620 nm (OD₆₂₀) was measured. Cultures with an OD₆₂₀ of between 0.30 and 0.35 were included in the experiments. Subsequently, 1 ml of the B. hyodysenteriae culture was centrifuged at 3,000 rpm for 20 min, and the bacterial pellet was washed three times with anaerobic PBS and resuspended in anaerobic PBS overnight. Of this suspension, 400 μl was transferred into Eppendorf tubes with a puncture hole in the lid made with a heated 18-gauge needle. Capillary tubes containing either anaerobic PBS (negative control) or 0.1 M fucose in anaerobic PBS were placed in the puncture holes. The Eppendorf tubes with capillary tubes were incubated for 90 min in an anaerobic chamber. The experiments were performed in triplicates and repeated at least twice for each strain. After incubation, capillary tubes were removed from the Eppendorf tubes, and the content of the capillary tubes was removed for DNA isolation using a QIAamp DNA minikit (Qiagen, Hilden, Germany). The quantity of B. hyodysenteriae DNA in the capillary tubes was determined by qPCR using forward primer 5′-ACTAAAGATCCTGATGTATTTG-3′ and reverse primer 5′-CTAATAAACGTCTGCTGC-3′ based on the nox gene as described previously by La et al. (35). The following primer pair was designed to generate an amplicon to be used as a standard: forward primer 5′-GGGTTGGTGGCGTAGTTAAA-3′ and reverse primer 5′-TGCATTTCTATGCCAGCTTC-3′. Quantitative PCR was performed on a CFX96 real-time PCR (RT-PCR) system with a C1000 thermal cycler (Bio-Rad, Hercules CA, USA). Two microliters of DNA was suspended in a 10-μl reaction mixture consisting of SensiMix SYBR No-ROX (Bioline Reagents Ltd., UK), high-performance liquid chromatography (HPLC)-grade water, and primers at 0.5 μM. The PCR program consisted of denaturation for 10 min at 95°C followed by 40 cycles of 95°C for 30 s and 60°C for 30 s. Standards and samples were run in duplicate. The chemotaxis value (R_che) was calculated as the ratio of the median number of bacteria in the capillary tube with fucose to the median number of bacteria in the negative-control capillary (31).

Statistical analysis.

Statistical analysis was performed using GraphPad Prism version 8 software (GraphPad, La Jolla, CA, USA). Results are expressed as means ± standard errors of the means (SEM) for normally distributed data and medians with interquartile ranges (IQRs) for data that did not follow a normal distribution (determined using the D’Agostino-Pearson omnibus test). Spearman’s rank correlation coefficient (r_s) or Pearson’s correlation coefficient (r) was computed to assess the relationship between two variables. Data were analyzed using the Mann-Whitney test and one-way analysis of variance (ANOVA) wherever applicable, and P values of ≤0.05 were considered statistically significant.

R 4.0.3 (36) scripting in R studio 1.3.1093 (37) was used to perform hierarchical cluster analysis, draft heat maps of glycan relative abundances, as well as build and visualize OPLS(-DA) models (38, 39). The utilized R packages were ropls 1.20.0 (40), viridis 0.5.1 (41), pheatmap 1.0.12 (42), and the tidyverse toolkit 1.3.0 (43). The script and underlying data set can be found in the supplemental material. The heat map for the relative abundance of sialic acid-containing glycan structures was refined in Affinity Designer (Serif Europe Ltd., Nottingham, UK) and annotated with glycan structures drawn with DrawGlycan-SNFG (44). Hierarchical cluster analysis is an unsupervised machine learning technique; i.e., the algorithm tries to find patterns in unlabeled data. The hierarchical agglomerative cluster analysis of pigs, with the complete linkage method according to Euclidean distance, was visualized in the heat map with a dendrogram.

OPLS can be understood as a multivariate regression analysis extension of principal-component analysis (PCA) (38). This is similar to projections to latent structures (PLS) regression modeling except with only one predictive component for easier analysis of the results and with orthogonal components to capture the variation not related to the response variable. OPLS-DA is a discriminant analysis extension of OPLS (39), which, for example, can be used to create models for discriminating between infection states. In contrast to hierarchical cluster analysis, OPLS-DA is a supervised machine learning technique where the algorithm tries to figure out how the data are connected to the provided labels. There are various model diagnostics available for OPLS-DA, but for this study, we focused on R2Y and Q2Y. R2Y shows how much of the variation in the response that the component explains, and Q2Y shows how well the model is predicted to explain the response variation on a new data set based on cross-validation. For this study, we considered a Q2Y value of >0.5 to indicate a good predictive ability. The OPLS models presented in Fig. 5 were refined with Affinity Designer for easier interpretation. The underlying R script and figure output from it for OPLS(-DA) model construction are available in the supplemental material.