Mucin-Degrading Microbes Release Monosaccharides That Chemoattract Clostridioides difficile and Facilitate Colonization of the Human Intestinal Mucus Layer

Abstract

It is widely accepted that the pathogen Clostridioides difficile exploits an intestinal environment with an altered microbiota, but the details of these microbe–microbe interactions are unclear. Adherence and colonization of mucus has been demonstrated for several enteric pathogens and it is possible that mucin-associated microbes may be working in concert with C. difficile. we showed that C. difficile ribotype-027 adheres to MUC2 glycans and using fecal bioreactors, we identified that C. difficile associates with several mucin-degrading microbes. C. difficile was found to chemotax toward intestinal mucus and its glycan components, demonstrating that C. difficile senses the mucus layer. Although C. difficile lacks the glycosyl hydrolases required to degrade mucin glycans, coculturing C. difficile with the mucin-degrading Akkermansia muciniphila, Bacteroides thetaiotaomicron, and Ruminococcus torques allowed C. difficile to grow in media that lacked glucose but contained purified MUC2. Collectively, these studies expand our knowledge on how intestinal microbes support C. difficile.

Graphical Abstract

In the United States, Clostridioides difficile infection (CDI) affects millions of patients each year and represents an annual cost of over $1 billion.^1–5 Despite the immense human health burden of this antibiotic-associated disease and significant research effort dedicated to fighting CDI, the various mechanisms C. difficile employs to colonize the host and promote disease remain ill-defined. CDI typically results from a disruption in the gut microbiota following broad-spectrum antibiotic use,⁶ whereby C. difficile produces toxins that result in the clinical manifestations of CDI, which include diarrhea, inflammation, epithelial damage, and pseudomembranous colitis.^1–5 While much progress has been made on understanding the primary virulence factors produced by C. difficile, questions still remain in terms of colonization and C. difficile–microbe interactions.

The ability to interact with intestinal mucus constitutes an important colonization factor for several enteric pathogens.^7–13 In vitro, C. difficile has been shown to adhere to intestinal mucins,^14–18 and in vivo C. difficile has been observed within the mucus of mice^19–21 and humans.^22,23 C. difficile and other mucosa-associated microbes grow as microcolonies or intestinal biofilms in the mucus layer, indicating that mucus is likely the colonization site for C. difficile.^21,24–46 Intestinal mucus is primarily composed of a MUC2 core protein that is secreted from goblet cells.^47,48 MUC2 is highly O-glycosylated, which accounts for up to 80% of the mucin molecular weight. O-Glycans contain branched carbohydrates including N-acetylgalactosamine (GalNAc), N-acetyl-glucosamine (GlcNAc), galactose, fucose, and sialic acid (NANA).^49–53 Additionally, MUC2 is N-glycosylated, consisting primarily of linked mannose residues.⁵⁴ These O- and N-glycosylated residues on MUC2 could provide an environment rich in nutrients capable of supporting C. difficile growth.

Numerous bacterial species can use host-derived mucin glycans as growth substrates.^19,55–61 Mucin-degrading specialists such as Akkermansia muciniphila, Bacteroides thetaiotaomicron, B. fragilis, and Ruminococcus gnavus possess a large repertoire of glycosyl hydrolase enzymes that can degrade mucin glycans.^61–69 These released glycan oligosaccharides, such as NANA, GlcNAc, GalNAc, galactose, mannose, and fucose, can then be metabolized by the degrading microbe or by other resident microbes.^67,70–73 Enteric pathogens, in particular, are well-adapted to use mucin carbohydrates to colonize specific niches.^7,74–77 As a result, MUC2 glycans can serve as a nutrient source for bacteria that can utilize the mucus-derived sugars even if they lack mucin-degrading enzymes. Since glycans function as a nutrient source, multiple enteric pathogens including Vibrio cholera,^10,78,79 Campylobacter jejuni,^8,9,80 and Salmonella typhimurium^11,13,81 chemotax toward intestinal mucus and mucin oligosaccharides. Similar to other pathogens, C. difficile has also been shown to utilize carbohydrates,^82–86 but the precise interactions between C. difficile, mucins, and mucin-degrading microbes remain unclear. Here we examined C. difficile (ribotype 027)–human mucin interactions in vitro and identify the role of mucin-degrading microbes in promoting C. difficile growth.

RESULTS

C. difficile Adheres to Human MUC2 Glycans.

Highly O-glycosylated mucin proteins can serve as ligands for bacterial adhesins.⁸⁷ To examine if mucin glycans could serve as ligands for C. difficile, we examined the localization of fluorescently labeled C. difficile R20291 in the human mucin-producing cell line HT29-MTX (Figure 1A). C. difficile R20291 was fluorescently labeled with the cell-permeable compound carboxyfluorescein diacetate succinimidyl ester (CFDA-SE) and incubated with confluent layers of HT29-MTX for 1 h prior to immunostaining for MUC2. Fluorescence microscopy showed C. difficile (green) colocalized with MUC2 (purple) among HT29-MTX cells (nuclei blue) in Figure 1A–C. At high magnification (100×), a direct interaction between C. difficile and MUC2 was observed in vitro (Figure 1C), showing that C. difficile could interact closely with mucin ligands along the gut epithelium.

Figure 1.

C. difficile R20291 adheres to the glycans human and mouse MUC2. (A) A 3D projection of C. difficile R20291 fluorescently tagged with CFDA-SE (green) colocalized with human MUC2 (purple) in mucin-producing HT29-MTX cells by immunostaining. Nuclei are marked with Hoechst dye (blue) (n = 3 replicates). (B, C) Representative images of C. difficile R20291 at higher magnifications (40×, 100×). Scale bar = 10 μm (n = 3 replicates). (D) Representative images of CFDA-SE tagged C. difficile R20291 (green) adhering to human LS174T-derived MUC2 coated coverslips after 1 h incubation (scale bar = 100 μm) (n = 3 replicates). (E) Representative images of CFDA-SE tagged C. difficile R20291 (green) adhering to human LS174T-derived MUC2 coated coverslips with glycans removed by beta-elimination and acid hydrolysis (scale bar = 100 μm) (n = 4 replicates). (F) Mouse colon tissue sections stained with Hoechst (blue) and incubated with CFDA-SE tagged C. difficile (green) for 1 h with costaining of mucin glycans (UEA-1, purple) (scale bar = 100 μm)6666 (n = 3 mice). (G) Mouse colon tissue sections with glycans removed stained with Hoechst (blue) and incubated with CFDA-SE tagged C. difficile (green) for 1 h with costaining of mucin glycans (UEA-1, purple) (scale bar = 100 μm) (n = 3 mice). (H) Normal human colon incubated with CFDA-SE tagged C. difficile (green) for 1 h with costaining of mucin glycans (UEA-1, purple) (n = 72 patients; 3 patients shown). Yellow arrows highlight C. difficile within the mucus layer. Scale bar = 100 μm.

To identify the role of glycans in adherence, MUC2-coated coverslips were oxidized and underwent acid-hydrolysis and beta-elimination, effectively removing all N- and O-linked glycans.⁸⁸ Fluorescently labeled C. difficile was then incubated with control (untreated) or glycan-removed MUC2 coverslips for 1 h (Figure 1D, E). Robust adhesion of C. difficile was observed in untreated MUC2 coverslips (Figure 1D), while glycan removal greatly diminished the ability of C. difficile to adhere to MUC2 (Figure 1E). To confirm the adhesion of C. difficile to mucin glycans, we incubated slides containing 7 μm sections of fixed tissue from C57BL/6J mouse colons with fluorescently labeled C. difficile under anaerobic conditions at 37 °C for 1 h and examined glycan colocalization using the lectin Ulex Europaeus Agglutinin-1 (UEA-1), which recognizes fucose residues. We observed C. difficile within the UEA-1 positive mucus layer (Figure 1F). Interestingly, we rarely found C. difficile on the epithelium, despite the access of C. difficile to the tissue using this technique. To further demonstrate specificity for mucin glycans, tissue sections were oxidized and underwent acid-hydrolysis and beta-elimination as described above. Incubation of fluorescently labeled C. difficile with these glycan-removed slides resulted in little to no adhesion (Figure 1G). These data indicate that C. difficile adheres to the glycan component of purified MUC2 and murine intestinal MUC2. Finally, we employed a human normal colon array (US Biomax) to identify adhesion of C. difficile in human colon (Figure 1H). Although these biopsy specimens were fixed in 10% formalin, which compresses the mucus layer, we still observed C. difficile localized in mucus above goblet-cell-lined crypts. Consistent with our mouse colon staining, we observed limited number of C. difficile cells associated with the epithelium. These data suggest that C. difficile preferentially adheres to mucin glycans which may facilitate intestinal colonization.

C. difficile Colonizes Mucus in Antibiotic-Treated Human Fecal Bioreactors.

Once C. difficile adheres to the mucus layer, it likely interacts with other mucus-associated microbes to form a stable biofilm community.^{19,21,23,42,89} Fecal minibioreactor arrays (MBRAs) have been used to model C. difficile colonization of antibiotic-disrupted microbial communities using the antibiotic clindamycin and fecal samples of healthy adult donors.^85,90 We modified the MBRA model for mucus-coated coverslips and evaluated C. difficile colonization of mucus-associated microbial communities when treated with broad-spectrum antibiotics (Figure 2). We used 50 mL fecal bioreactors to test whether C. difficile-colonized coverslips coated with purified mucin (MUC2) from LS174T (n = 2 bioreactors) and HT29-MTX (n = 2) cells compared to coverslips with no MUC2 coating (mock, n = 2). Coverslips were suspended into established microbial communities after clindamycin treatment at the time of C. difficile inoculation. After 4 days, coverslips were removed and evaluated microscopically for biofilm formation. Imaging of the coverslips revealed biofilms established on MUC2-coated and mock-coated coverslips (Figure 2A), as determined by acridine orange staining. C. difficile within biofilm communities was identified by fluorescence in situ hybridization (FISH) staining. As seen in Figure 2B, C. difficile was observed in greater abundance in MUC2-coated coverslips compared to mock-coated coverslips.

Figure 2.

C. difficile interacts with other mucus-associated microbes in bioreactors. Bioreactors seeded with stool from healthy donors and treated with clindamycin (250 μg/mL), followed by C. difficile (10⁶ CFU). Bioreactors were fitted with inserts harboring mock or MUC2-APTS-coated coverslips. MUC2 was purified from human LS174T and HT29-MTX cell lines. (A) Biofilm status was confirmed by staining with acridine orange in mock and MUC2-coated coverslips (scale bar = 50 μm) (mock, n = 2; MUC2, n = 4). (B) The presence of C. difficile within biofilm communities was determined by FISH with a C. difficile specific probe (red) and counterstained with Hoechst (blue) (mock, n = 2; MUC2, n = 4). Community structure of the MUC2-coated coverslips were assessed by 16S rDNA sequencing at the level of (C) family and (D) genus (mock, n = 2; MUC2, n = 4). (E) CFUs of C. difficile in bioreactor supernatant overtime, confirming C. difficile invasion. (F) Abundance of Peptostreptococcaceae OTUs detected by 16S rRNA gene sequencing of mock and MUC2-coated coverslips (BLAST analysis indicates sequence of Peptostreptococcaceae OTU is 100% identical to C. difficile). (G) qPCR analysis of C. difficile housekeeping gene rpoA, back calculated to CFU by four-parameter logistic curve in mock and MUC2-coated coverslips. Student’s t test, *p < 0.05.

C. difficile Interacts with Mucin-Degrading Microbes in Human Fecal Bioreactors.

To evaluate the effects of mucus on microbial community composition and the gut microbes that aid in C. difficile mucus colonization, we sequenced the V4 variable region of 16S rRNA genes from DNA isolated from bioreactor coverslips. Family level (Figure 2C) and genus level (Figure 2D) relative abundances of microbial communities from mock coverslips were compared to that of mucus-coated coverslips. The dominant microbes of both mock and MUC2-coated coverslips were Bacteroidaceae (mock, 23.5%; MUC2, 30.6%), Fusobacteriaceae (mock, 21%; MUC2, 20.5%) and Clostridiaceae (mock, 16.7%; MUC2, 9.4%). In addition to Bacteroides, within MUC2 biofilms we identified several known mucin-degrading containing genera, including Parabacteroides, Porphyromonas, Ruminococcaceae, and Akkermansia. We speculated that the presence of these microbes within the mucus layer may promote the colonization of C. difficile.

Despite successful invasion of C. difficile in all bioreactors (Figure 2E), we observed varying levels of C. difficile in the inserted coverslips. Consistent with our FISH staining, we found that the overall abundance of C. difficile associated with mucus-coated coverslips was 7.4-fold greater than that for mock coverslips (Figure 2F). Importantly, the concentrations of C. difficile did not vary between mucus isolated from LS174T or MTX (LS174T, 70 ± 9.9 sequences per Peptostreptococcaceae OTUs vs HT29-MTX, 60.5 ± 6.4 sequences). These findings were confirmed using quantitative PCR (qPCR) with primers specificto C. difficile (Figure 2G), further confirming the preferential colonization of C. difficile to mucus-coated coverslips within a mixed microbial community.

Mucin Monosaccharides Chemoattract C. difficile.

Since we observed C. difficile in higher abundance in the MUC2-coated coverslips, we reasoned that C. difficile was able to sense mucins and chemotax toward the MUC2-coated coverslip. Chemotaxis is the directed movement of an organism in response to a chemical gradient that facilitates survival by enabling cells to move toward favorable growth conditions or away from toxic environments.^80,91 Chemotaxis toward intestinal mucus is thought to be crucial in the life cycle of motile enteric pathogens⁷ and could play a role in C. difficile infection. Mucin-degrading bacteria in biofilms release mucin glycan oligosaccharides, which can provide a key source of nutrients to other bacteria in the community, including C. difficile. Using a standard capillary chemotaxis assay,⁹² we investigated the ability of CFDA-SE-labeled C. difficile R20291 to migrate up the concentration gradient of the mucin oligosaccharides glucose, fucose, galactose, mannose, N-acetylglucosamine (GlcNAc), N-acetylgalactosamine (Gal-Nac), or N-acetylneuraminic acid (NANA) (Figure 3A). Chemotaxis of C. difficile up the monosaccharide concentration gradient would result in increased fluorescence intensity due to the accumulation of the CFDA-SE-labeled C. difficile. Capillary tubes containing chemotaxis buffer alone (negative control) or chemotaxis buffer with 100 mM glucose (positive control) or mucin monosaccharides were placed upright in the wells of a 96-well plate containing CFDA-SE-labeled C. difficile. After a 2 h anaerobic incubation at 37 °C, the capillary tubes were removed and fluorescence measured. Increased fluorescence intensity was observed in capillary tube contents for all monosaccharides compared to buffer control; however, the amount of fluorescence was monosaccharide-specific(Figure 3A). The mucin monosaccharide NANA provided the least fluorescence compared to buffer control (2.8-fold differences) and was 3.3-fold lower than glucose suggesting NANA is not a preferred chemoattractant for C. difficile. The greatest fluorescence resulted from C. difficile chemotaxis toward mannose (21-fold greater than glucose) followed closely by GlcNAc (18-fold greater than glucose). Galactose also elicited a greater chemotactic response from C. difficile than glucose (14-fold difference). In addition to monosaccharides, we also tested intact pig stomach mucus (MUC5AC/MUC6) and purified human MUC2 from HT29-MTX and LS174T cell lines. Fluorescent readings indicate that human-derived mucus served as a stronger chemoattractant for C. difficile than pig mucus (Figure 3A). Notably, human-derived mucus resulted in C. difficile chemotaxis comparable to glucose, while pig-derived mucus performed similarly to the mucin monosaccharides NANA and GalNAc. Addition of purified O-linked glycans from LS174T MUC2 also served a strong chemoattractant for C. difficile (13.7-fold increase). These data support that mucin glycan monosaccharides serve as chemoattractants for C. difficile.

Figure 3.

C. difficile R20291 chemotaxes toward and utilizes mucin components. (A) Capillary chemotaxis assay with C. difficile R20291 in chemotaxis buffer (10 M potassium phosphate, 0.1 mM EDTA) with or without 100 mM sugars, 1 mg/mL mucus (pig stomach MUC5AC/6, HT29-MTX derived MUC2, LS174T derived MUC2), or 1 mg/mL O-linked glycans isolated from LS174T MUC2. The amount of C. difficile in the capillaries following incubation was assessed by transferring the capillary contents to 96-well plates where fluorescence was read by a plate reader at excitation 428 nm/emission 528 nm. n = 8 wells per sugar, repeated four independent times. *p < 0.05, one-way ANOVA. (B, C) Live cell imaging analysis of ibidi chemotaxis slide using the ibidi “Motility” FIJI plugin. (C) Velocity and (D) distance traveled was measured from traces of individual C. difficile cells. n = 3 replicates, repeated five independent times. *p < 0.05, Student’s t test. (D) C. difficile R20291 was inoculated in the fully defined minimal medium CDMM at an OD_{600 nm} = 0.1. Samples were analyzed for OD_600nm at 1, 3, 6, 9, 12, and 24 h. CDMM was prepared without sugars or with 10 mM sugars (glucose, fucose, galactose, mannose, N-acetylglucosamine (GlcNAc), N-acetylgalactocosamine (GalNAc), or N-acetylneuraminic acid (NANA)). (E) Growth of C. difficile after 24 h in CDMM lacking sugars (0 mM) or containing 1, 10, or 100 mM of glucose or mannose. (F) Levels of AI-2 were determined by measuring V. harveyi luminescnence. V. harveyi was incubated with cell-free supernatant of C. difficile grown for 24 h in CDMM with 10 mM glucose or 10 mM mannose. Control represents cell-free supernatant AB media alone. (G) C. difficile inoculated at OD_600nm = 0.1 in CDMM containing glucose or mannose (1, 10, 100 mM) and incubated for 72 h in a MUC2 (LS174T)-coated 96-well plate. Crystal violet staining of mucin-based biofilms was assessed at A_600nm. (H, I) Vero cells were grown as 2D monolayers and incubated for 4 h with supernatant from 48 h cultures of C. difficile R20291 grown in CDMM with various carbohydrate sources (glucose, fucose, galactose, mannose, GlcNAc, GalNAc, NANA) or LS174T purified MUC2. Cell rounding was visualized by microscopy on a Nikon TiE with a 20× Plan Apo (NA 0.75) differential interference contrast objective, using a SPECTRA X LED light source and ORCA-Flash 4.0 sCMOS camera. Cell diameter (μm) was calculated from representative images of Vero cells following treatment (scale bar = 100 μm) (n = 3 per experiment, repeated two independent experiments). *p < 0.05, one-way ANOVA.

We further characterized the positive chemotactic response of C. difficile to mannose, the strongest chemoattractant tested, using live-cell imaging of ibidi μ-Slide chemotaxis slides. These slides are designed to generate a chemoattractant gradient for live imaging. CFDA-SE-labeled C. difficile R20291 were added in a chemotaxis agar to the channel and then chemotaxis buffer with or without an increasing mannose gradient was added to the flanking reservoirs. When C. difficile was exposed to chemotaxis buffer lacking mannose, cells displayed random unbiased movement within the chamber. However, when seeded with a mannose gradient increasing in concentration to the right-hand side of the chamber, imaging verified C. difficile cells exhibited directed motion toward mannose. Using the ibidi motility FIJI software, we tracked individual bacterial cells and showed that in the presence of a mannose gradient C. difficile cells moved with a higher velocity (PBS, 0.9 ± 0.4; mannose, 4.0 ± 2.1 μm/min) and traveled longer distances (PBS, 135.2 ± 88.6; mannose, 294.7 ± 128.7 μm) compared to chambers without mannose (Figure 3B, C). These data support the finding that mucin components, particularly mannose, behave as chemoattractants for C. difficile.

Mucin Monosaccharides Serve as Primary Carbon Sources for C. difficile.

Positive chemotactic behavior drives bacterial cells toward greater concentrations of beneficial chemicals and often these chemoattractants support bacterial growth. Having established that C. difficile can chemotax to mucin monosaccharides, we hypothesized that these sugars might serve as a nutrient source for C. difficile. To test this, we cultured C. difficile in CDMM supplemented with either no carbon source (negative control), 10 mM glucose (positive control), or 10 mM fucose, galactose, mannose, GlcNac, GalNAc, or NANA (Figure 3D). With 10 mM glucose as the primary carbon source in CDMM, C. difficile reached the stationary phase by 12 h with a peak OD₆₀₀ = 0.5 (Figure 3D, gray squares). C. difficile had only minor growth in the absence of glucose (Figure 3D, black circles), and the addition of mucin monosaccharides was sufficient to support C. difficile growth as individual primary carbon sources in CDMM. At 12 h, growth on all mucin monosaccharides was lower than that observed on glucose and cells had not reached stationary phase. However, cell density reached OD₆₀₀ values greater than 0.7 at 24 h for mucin monosaccharides mannose (green diamonds), GlcNAc (light green squares), and NANA (light blue squares), 1.4-fold greater than glucose (Figure 3D). Along with being the preferred chemoattractant, mannose also led to the greatest levels of total growth among mucin monosaccharides (Figure 3D). We further characterized C. difficile growth on mannose over a range of concentrations (1–100 mM) (Figure 3E). After 24 h, C. difficile growth increased with increasing concentrations of mannose and continued to produce greater cell density than glucose at the same concentration.

Our data indicated that C. difficile exhibits optimal growth in the presence of mannose as the primary carbon source; therefore, we examined if the presence of mannose had any effect on quorum-sensing compounds, because chemoattractant may increase bacterial numbers and thus AI-2 production. C. difficile secretes the universal quorum-sensing compound AI-2⁹³ and specific dietary sources are known to regulate AI-2 production.⁹⁴ Analysis of the Vibrio harveyi reporter strain BAA-1121, a mutant sensor strain that only senses autoinducer AI-2, revealed that C. difficile secretes AI-2 in the presence of glucose and mannose (Figure 3F). Interestingly, mannose did not elevate the levels of AI-2 above those observed for glucose. This data suggests that C. difficile does synthesize AI-2 in the presence of mannose as the sole carbon source. Finally, we questioned whether mannose influenced C. difficile colonization of mucus. To mimic the intestine, we coated 96-well plates with purified MUC2 from LS174T cells and examined C. difficile colonization by crystal violet staining after 72 h in CDMM with glucose or mannose (Figure 3G). The presence of glucose and mannose both enhanced C. difficile colonization of MUC2-coated wells compared to the CDMM no glucose/mannose controls. However, a significant increase in the density of colonized C. difficile was observed with 100 mM mannose compared to the equivalent glucose levels, indicating that mannose can bolster bacterial density and thus colonization. These data provide new evidence that mucin oligosaccharides can be a potent carbon source to promote C. difficile growth and colonization of the intestinal mucus layer.

Finally, we examined toxin production in CDMM with various carbohydrate sources after 48 h. To assess toxin activity, we incubated Vero cells with 1:1 C. difficile CDMM in DMEM for 4 h and collected images. Cell diameter calculations (Figure 3H) from representative Vero cell images (Figure 3I) revealed that all 48 h CDMM cultures harbored toxins. In this assay, no differences were observed between carbohydrate sources.

Mucin Degraders Promote C. difficile Growth When Mucin Is the Sole Carbon Source.

Bacteria must harbor the specific glycosyl hydrolase (GH) to enzymatically cleave mucin glycan structures and use mucin as a carbohydrate source. GH families cleave O-linked mucin-glycans including NANA (GH 33), GalNAc (GH 101, 129), GlcNAc (GH 84, 85, 89, 98), galactose (GH 2, 20, 42), or fucose (GH 29, 85). Additionally, the N-linked mucin glycans mannose can be removed with GH families 38 and 125. We queried the C. difficile R20291 genome for GH families using the Carbohydrate-Active enZYmes Database (CAZy). C. difficile R20291 harbored 13 GH families, only one of which was involved in mucin glycan degradation (Figure 4A). C. difficile R20291 was found to contain three gene copies of GH 38, which removes N-linked mannose residues. We next investigated whether these genes were conserved among C. difficile strains. When we examined 35 C. difficile genomes, we found that all genomes only harbored GH 38. Additionally, we found that all 35 genomes contained one to four gene copies of GH 38 (Figure 4B), indicating that the ability to remove mannose is a conserved function across all examined C. difficile members. Secreted mucins are covered with O-linked glycans, and the 3D structure of mucus often masks N-linked glycans. As a result, we speculated that GH38 from C. difficile would not be able to readily access N-linked glycans. Based on the lack of mucin-O-glycan degrading GHs, we predicted that C. difficile would be unable to grow in CDMM if mucus were the only carbon source. Consistent with this hypothesis, we found that C. difficile had minimal growth in CDMM without glucose, even when supplemented with mucus from pig stomach or human mucin-producing cell lines LS174T, HT29-MTX, or T84 (Figure 4C) (starting OD_600nm = 0.1; final OD_600nm = 0.18–0.22). These data indicate that C. difficile is able to use available mucin monosaccharides, but is unable to free these sugars from mucin on its own.

Figure 4.

C. difficile cannot enzymatically cleave intestinal MUC2, but can use oligosaccharides cleaved by A. muciniphila, B. thetaiotaomicron, and R. torques. (A) C. difficile R20291 genome analysis for glycosyl hydrolase (GH) families using CAZY. Y-axis represents the number of genes in each family. The green bar indicates GH families involved in mucin degradation. (B) Gene copies for GH Family 38 following genome analysis of 35 C. difficile genomes using CAZY. (C) C. difficile R20291 was inoculated in the fully defined minimal medium CDMM containing 1 mg/mL mucus (pig stomach, human LS174T, human HT29-MTX, human T84) or 10 mM glucose at OD_600nm = 0.1. OD_600nm was measured after 1, 3, 6, 9, 12, and 24 h of incubation. (D) A. muciniphila, B. thetaiotaomicron, R. torques, and R. gnavus genome analysis for mucin-degrading glycosyl hydrolase (GH) families using CAZy. (E) C. difficile R20291 was inoculated in CDMM at an OD_600nm = 0.1 with or without 1 mg/mL MUC2 (LS174T), with or without glucose. Separate cultures of A. muciniphila BAA-835, B. thetaiotaomicron ATCC 29148, and R. torques ATCC 2775 were coinoculated with C. difficile R20291 in CDMM containing 1 mg/mL MUC2 and no glucose at an OD_600nm = 0.1. Data represents OD_{600 nm} at 24 h growth. (F) qPCR analysis of C. difficile housekeeping gene rpoA, back calculated to CFU by four-parameter logistic curve from the same experiment. (G) C. difficile R20291 and A. muciniphila BAA-835 were inoculated in CDMM containing glucose at an OD_600nm = 0.1 with or without 1 mg/mL MUC2 (LS174T). After overnight incubation, cell-free supernatant was collected (“spent medium”) and used to supplement new cultures of CDMM with no glucose at 25%. Data represents OD_{600 nm} at 24 h growth. *p < 0.05, one-way ANOVA.

Mucin degradation has been shown to be a cooperative event by members of the gut microbiome.⁵⁵ Akkermansia, Bacteroides, and Ruminococcus have been well documented for their mucin-degrading ability.^{55,68,95–98} We queried seven A. muciniphila genomes, three Bacteroides thetaioatiomicron genomes, one Ruminococcus torques, and two Ruminococcus gnavus genomes using CAZy (Figure 4D). These genomes contained multiple genes for glycosyl hydrolases involved in mucin degradation (Figure 4D), with the highest gene copies of mucin-degrading GHs observed in Bacteroides. To identify whether mucin-degrading bacteria were capable of liberating mucin-glycans and cross-feeding C. difficile, we grew C. difficile in CDMM containing LS174T MUC2 as the sole carbon source in the presence or absence of A. muciniphila BAA-835, B. thetaiotaomicron ATCC 29148, and R. torques ATCC 2775 (Figure 4E). Consistent with our previous findings, we observed that C. difficile was unable to grow on MUC2 alone. However, addition of mucin-degrading microbes to media containing MUC2 resulted in increased optical density that was similar to or greater than that observed when C. difficile was grown in CDMM + glucose or CDMM + glucose + MUC2. Using qPCR, we verified increased C. difficile growth in mucin-containing CDMM with A. muciniphila (calculated CFU: 6.1 × 10⁶ ± 1.1 × 10⁶), B. thetaiotaomicron (CFU: 8.9 × 10⁶ ± 5.4 × 10⁵), and R. torques (CFU: 4.1 × 10⁶ ± 9.3 × 10⁴) compared to the CDMM with MUC2 alone (CFU 3.0 ± 5.2) (Figure 4F).

To confirm that degradation of mucin was responsible for the observed increase in C. difficile growth and not bacterial metabolites, we focused our experiments on characterizing interactions with A. muciniphila. We cultured C. difficile R20291 and A. muciniphila separately in CDMM containing 10 mM glucose with or without MUC2 overnight (Figure 4G) and used the filtered cell-free supernatants to assess whether hydrolyzed MUC2 byproducts produced by A. muciniphila could influence the growth of C. difficile. The following day, we added 25% of this cell-free supernatant from the overnight culture medium to new CDMM containing no sugars and inoculated with C. difficile at an OD_600nm of 0.1. As expected, C. difficile was unable to grow in CDMM without glucose, but it grew at moderate levels when cultured in the presence of 25% cell-free supernatant from C. difficile grown in CDMM with glucose. The addition of A. muciniphila cell-free supernatant previously cultured in CDMM with glucose supported C. difficile similarly to that of C. difficile cell-free supernatant containing glucose. However, cell-free supernatant from A. muciniphila cultured in CDMM with glucose and MUC2 increased the OD_600nm of C. difficile by 2.6-fold, indicating that mucin degradation and not simply A. muciniphila metabolites were responsible for increased C. difficile growth. These data point to the role of mucin-degrading microbes in promoting C. difficile growth and colonization of the intestinal mucus layer.

MUC2 and Mucin-Degrading Microbes Influence C. difficile Gene Expression.

Finally, we examined if the presence of mucin or mucin degradation by microbes could influence C. difficile gene expression. To address this question, C. difficile was cultured overnight in CDMM with 10 mM glucose, 10 mM mannose, or 10 mM glucose in the presence of 1 mg/mL LS174T purified MUC2. Alternatively, C. difficile was cocultured with A. muciniphila, B. thetaiotaomicron, or R. torques in CDMM containing MUC2 as the only carbon source (Figure 5). The presence of MUC2 or MUC2 and mucin-degrading microbes significantly increased expression of the cell wall protein 84 (cwp84) and surface layer protein A (slpA), both of which are involved in adhesion (Figure 5A, B). No differences in slpA or cwp84 were observed between C. difficile grown in 10 mM glucose and 10 mM mannose in the absence of MUC2. Additionally, MUC2 and to a greater degree (>20-fold increase) MUC2 with mucin-degrading microbes elevated gene expression of fliC and flgG that encode components of flagella (Figure 5C, D). Increased expression of genes important for chemotaxis (fliC and flgG) in C. difficile cells cultured in the presence of MUC2 and mucin-degrading microbes led us to hypothesize that cleavage of mucin glycans by A. muciniphila, B. thetaiotaomicron, or R. toques may increase concentrations of these chemoattractants and promote C. difficile chemotaxis. To test this, we exposed chemotaxis buffer to MUC2-treated A. muciniphila, B. thetaiotaomicron, or R. toques and allowed cells to secrete metabolites or mucin-degradation products into the buffer. After 2 h exposure, we used the filtered chemotaxis buffer to monitor C. difficile chemotaxis (Figure 5E). As a control, intact LS174T MUC2 in buffer was used. We observed that C. difficile chemotaxed toward intact MUC2, but that chemotaxis was enhanced by 1.5-fold and 1.9-fold in response to MUC2 preincubated with either A. muciniphila or B. thetaiotaomicron. We observed a trend toward elevated chemotaxis with R. torques and MUC2 compared to MUC2 alone, but it did not reach significance. With fewer GHs, it is possible that R. torques is less efficient at mucin monosaccharide cleavage than A. muciniphila or B. thetaiotaomicron, leaving the filtered chemotaxis buffer from R. torques less concentrated in mucin monosaccharides and therefore less efficient at chemoattracting C. difficile. These findings suggest that C. difficile is responsive to both intact MUC2 and monosaccharaides cleaved from MUC2 by mucin-degrading gut microbes. These data lead us to propose that mucin degradation by the gut microbiome promotes C. difficile chemotaxis, mucin adhesion, and colonization. Collectively, these data highlight the interplay between intestinal mucus, C. difficile, and other mucus-associated microbes.

Figure 5.

Intestinal MUC2 influences C. difficile gene expression C. difficile R20291 was inoculated in the fully defined minimal medium CDMM with glucose, mannose, 1 mg/mL human LS174T MUC2, and/or A. muciniphila BAA-835, B. thetaiotaomicron ATCC 29148, and R. torques ATCC 2775. After 16 h of incubation, samples were collected for RNA isolation and qPCR analysis. qPCR analysis of adhesion- related genes, (A) splA and (B) cwp84 or flagella-related genes (C) flgG and (D) fliC. (E) Chemotaxis analysis of CFDA-SE labeled C. difficile toward buffer controls, 1 mg/mL MUC2 (LS174T), or supernatant from A. muciniphila, B. thetaiotaomicron, or R. torques incubated with MUC2 in PBS for 2 h. n = 6 replicates, repeated three independent times. *p < 0.05, One-Way ANOVA.

DISCUSSION

Herein, we demonstrate the importance of intestinal mucins as a chemoattractant and energy source for C. difficile. Our work indicates that C. difficile is responsive to nutrient bioavailability and that C. difficile readily metabolizes mucin monosaccharides. However, we demonstrate that C. difficile does not harbor the GHs required for extensive degradation of mucin glycans and thus is not capable of generating these monosaccharides without the aid of other members of the gut microbiota. This work points to the role of other microbes in supporting C. difficile colonization. A major question in the field is which microbes play a supportive role in CDI and which ones simply co-occur with C. difficile.⁹⁹ In the antibiotic-treated gut, we propose that mucin-degrading microbes such as Akkermansia, Ruminoccocus, and Bacteroides create a gradient of mucin monosaccharides which chemoattracts C. difficile to its colonization site. Once at the intestinal mucus layer, C. difficile metabolizes mucin monosaccharides cleaved by members of this shifted microbiota.

Antibiotic treatment is known to perturb the microbiota and thus modulates mucin carbohydrate availability. An expansion of MUC2 into the lumen following antibiotic treatment has been documented.^100,101 Additionally, the concentrations of free fucose and sialic acid have been reported to reach high levels during antibiotic treatment.¹⁰² This may be due in part to an increased abundance of mucin-degrading Bacteroides and Akkermansia which have been shown to increase following antibiotic administration.^101,103 Additionally, postantibiotic luminal contents possess more mucin and host carbohydrate-degrading genetic modules than strains isolated before antibiotic treatment.¹⁰¹ The importance of carbohydrates in C. difficile infection has been shown in vivo, where a decreased abundance of murine cecal carbohydrates over time correlated with the expression of C. difficile genes for carbohydrate utilization.¹⁰⁴ Moreover, inoculation of mouse intestinal microbes into growth medium made from the cecal contents of gnotobiotic mice depleted carbohydrates and prevented C. difficile replication.¹⁰⁵ This same study demonstrated that supplementation with the monosaccharides glucose, GlcNAc, or NANA was sufficient to overcome this nutrient-depravation and enable C. difficile growth.¹⁰⁵ Based upon these previous results and the work presented here, we propose that C. difficile uses released mucin glycans (GlcNAc, NANA, and mannose) as a fuel source to drive its colonization in antibiotic-treated individuals.

Previous studies have found that mucin degradation is a community process.^{52,55,96,98,106,107} Bacterial species which do not harbor neuraminidases (enzymes that remove sialic acid/NANA from mucin) must rely on species that do contain these enzymes. Our in silico findings revealed that C. difficile does not harbor GH 33 neuraminidases or any other O-linked glycan removing enzymes. As a result, C. difficile must rely on other members of the gut microbiome to provide this function. Antibiotic exposure, a prerequisite for most C. difficile infection, shifts the microbiome structure.^86,104,108 Patients with C. difficile infection harbor mucin-degrading Escherichia, Ruminococcus, Streptococcus, and Akkermansia.^{27,55,109–113} Thus, it is possible that mucin degradation is involved in the pathogenesis of C. difficile. Conflicting data exists on the levels of Bacteroidetes in C. difficile infection. Some groups have observed increased Bacteroidetes levels in C. difficile infected patients, while other groups have observed decreased Bacteroidetes levels.^{22,23,109,111,114–116} In B. thetaiotaomicron monocolonized mice, B. thetaiotaomicron was shown to promote C. difficile infection likely through release of mucin-derived monosaccharides.¹⁰² These studies suggest that bacterial mucin degradation and mucin monosaccharides likely influence C. difficile infection. Another example of a microbe likely aided by mucin degradation is vancomycin-resistant Enterococcus. Similar to our studies with C. difficile, Enterococcus is unable to grow on purified mucin alone, but it has been reported to grow on mucin predigested by the mucin-degrader R. torques.¹¹⁷ Enterococcus dominated our bioreactor MUC2-biofilms (Figure 2). As a result, we speculate that like C. difficile, Enterococcus may take advantage of mucin-degradation to promote its expansion.

Removal of mucin glycans by microbes can destabilize the mucus gel, allowing microbes to come in closer proximity to the host.^118–120 However, in addition to the secreted mucins, like MUC2, epithelial cells are covered in adherent mucins.¹²¹ Similar to MUC2, adherent mucins like MUC1, MUC4, and MUC13 are also glycosylated and can serve as additional decoys for pathogens. Adherent mucins like MUC1 have been shown to be critical for C. jejuni infection by acting as a releasable decoy.⁸² Therefore, even if MUC2 levels are decreased, we would speculate that C. difficile could adhere to adherent mucins, which can be cleaved and released from the epithelium.

Our work highlights the potential importance of mucin-degrading microbes in the initial stages of C. difficile pathogenesis. We demonstrate that A. muciniphila, B. thetaiotaomicron, and R. torques can grow with and cross-feed C. difficile mucin glycans, indicating a level of synergy between these microbes. In general, A. muciniphila is considered a commensal microbe. Evidence of a dose-dependent benefitof A. muciniphila on human health is lacking, but several positive attributes have been attributed to A. muciniphila. Favorable effects of A. muciniphila on the host include improved glucose tolerance, prevention of obesity, amelioration of experimental alcoholic liver disease, and immune modulation.^122–127 However, overgrowth of A. muciniphila is speculated to cause mucin degradation and may promote tissue damage.¹²⁸ Additionally, A. muciniphila may inadvertently promote infection with enteric pathogens. A. muciniphila has been reported to stimulate epithelial access and initiate lethal colitis by the mucosal pathogen Citrobacter rodentium.¹²⁹ We similarly find that A. muciniphila may promote C. difficile colonization of the intestinal mucus layer.

Competition for nutrients among microbial communities has been suggested as a key barrier to enteric infection. When the microbiome is disrupted by antibiotics, as is commonly the case in C. difficile infection, nutrient competition is reduced. As a result, nutrients are more freely available, allowing C. difficile to establish a niche and cause infection. The ability of C. difficile to use carbohydrates as a nutrient source is well documented. Using Biolog Phenotype MicroArrays, Scaria et al. examined the ability of C. difficile strains 630, CD196, QCD-32g58, 54634, QCD-23m63, and R20291 to grow in rich BHI broth with various carbon sources.¹³⁰ All C. difficile strains were found to use simple sugars to support enhanced growth. The authors identified that all six strains had the highest levels of growth (OD_600nm > 2.2) in mannose, GlcNAc, and sialic acid, with moderate growth (OD_600nm = 1) with GalNAc and fucose. However, these studies were not performed with minimal media and glucose was present. We recently published work examining C. difficile 2015 carbohydrate utilization in our fully defined medium CDMM using Biolog Phenotype MicroArrays.¹³¹ Similar to the C. difficile R20291 data present herein, we found that C. difficile 2015 exhibited the highest growth with mannose and GlcNAC (OD_{600 nm} ~ 1), with lower growth observed with fucose and galactose. In these assays, we did not examine GalNAc or sialic acid. In another study using a fully defined medium and microarrays, C. difficile 2015 and 630 were found to have >1.5 fold growth advantage with mannose, GlcNAc, and sialic aicd.¹³² These studies confirm our present work and demonstrate that a variety of C. difficile strains and ribotypes can utilize mucin-related sugars, particularly mannose and GlcNAc. In the future, in vivo studies would be valuable to confirm the ability of C. difficile to use preferentially use sugars and examine whether mucin-degrading microbes can promote C. difficile in the setting of complex microbiota. Collectively, these findings provide an additional link between intestinal mucus, mucin degrading microbes, and C. difficile.

METHODS

Bacterial Strains and Culture Conditions.

Routine culturing of Clostridioides difficile strains R20291 (ribotype 027) and CD2015 (ribotype 027), B. thetaioatomicron ATCC 29148, and Ruminococcus torques ATCC 2775 was performed in BHIS (Brain Heart Infusion medium supplemented with 2% yeast extract and 0.2% cysteine). Akkermansia muciniphila ATCC BAA-835 was cultured in BHIS supplemented with 0.4% pig stomach mucin (Sigma # M2378). All cultures were grown in an Anaerobe Systems AS-580 anaerobic chamber supplied with a mixture of 10% CO2, 5% H2, and 85% N2 at 37 °C. Routine culturing of Vibrio harveyi BAA-1121 in Marine medium and Escherichia coli strains K12 and DH5α in LB medium were carried out aerobically at 30 and 37 °C, respectively.

To assess utilization of mucin-related carbohydrates, BHIS overnight cultures of C. difficile strains were subcultured to an optical density (OD_600nm) of 0.1 into the fully defined minimal media, CDMM¹³³ with varied carbon sources. CDMM composition contained 10 mM of either glucose, fucose, galactose, mannose, N-acetyl-glucosamine (GlcNAc), N-acetylgalactosamine (GalNAc), or N-acetyl-neuraminic acid (NANA) (all purchased from Carbsynth). Anaerobic growth of cultures at 37 °C was monitored over time (0, 3, 6, 9, 12, and 24 h) through measurement of OD_{600 nm} on a Smartspec Plus spectrophotometer (Biorad Laboratories Inc.).

Chemotaxis and Live Imaging.

Chemotaxis of C. difficile strains toward monosaccharides was evaluated. Overnight cultures were diluted 1:50 in fresh BHIS, incubated anaerobically at 37 °C for 3 h, and then washed in PBS. Cells were fluorescently tagged with 10 μM CFDA-SE (ThermoFisher) in PBS anaerobically at 37 °C for 1 h, washed in PBS, and resuspended to an OD_600nm = 2.0 in chemotaxis buffer (10 M potassium phosphate, 0.1 mM EDTA at pH 7.0) and incubated anaerobically for 1 h at 37 °C as previously described.⁹² Fluorescent C. difficile was then distributed into 96-well plates (200 μL/well) prior to the addition of capillary tubes (one tube per well, n = 8 tubes per group) containing chemotaxis buffer with or without 100 mM monosaccharides (glucose, fucose, galactose, mannose, GlcNAc, GalNAc, and NANA). This capillary tube setup was incubated anaerobically at 37 °C for 2 h,⁹² and then the capillary tube contents were collected and measured by a fluorescence plate reader (excitation 485 nm/emission 528 nm). Fluorescent microbes were also confirmed by microscopy.

For live imaging assays, CFDA-SE fluorescently tagged C. difficile was added to cooled chemotaxis buffer containing 0.05% agar and added to ibidi Chemotaxis slides (ibidi). Chemotaxis slides were placed in an Okolabs stage-top incubation chamber with CO₂ mixing and humidity control. The incubation chamber was placed on a Nikon TiE inverted wide field epifluorescence microscope (Nikon) with a motorized X, Y, and Z stage for software-controlled multi-position imaging. Cells were imaged with widefield epifluorescence using a 20× PlanFluor (NA 0.45) phase contrast objective or a 20× Plan Apo (NA 0.75) differential interference contrast objective, using a SPECTRA X LED light source (Lumencor). Images were recorded using an ORCA-Flash 4.0 sCMOS camera (Hamamatsu). Videos were analyzed with the “Chemotaxis and Migration Tool,” a plugin used with the NIH ImageJ processing system (FIJI).

Quorum Sensing AI-2 Assay.

The AI-2 bioluminescence assay was performed as previously described.^93,134 Briefly, V. harveyi BAA-1121 overnight cultures were diluted 1:5000 in fresh Autoinducer Bioassay (AB) liquid medium and aliquoted into a 96-well plate (100 μL/well). C. difficile cell-free conditioned CDMM (10% v/v) was added. Negative controls included uninoculated CDMM, uninoculated LB, and E. coli DH5α cell-free conditioned LB. E. coli K12 cell-free conditioned LB served as a positive control. Reactions were incubated at 30 °C shaking on a Synergy H1 microplate reader (BioTek) and luminescence measured every 15 min. Induction of luminescence was taken at the time when there was maximal difference between the positive and negative controls (3–5 h).

RNA Isolation, cDNA Preparation, and qPCR.

For gene expression studies, C. difficile was cultured anaerobically at 37 °C in CDMM (starting OD_600nm = 0.1) for 16 h. Cells (1 mL) were fixed in an equal volume of ice-cold methanol, pelleted at 16 000g for 30 s, and stored at −80°C until RNA isolation. RNA was extracted from bacterial pellets using the Zymo QuickRNA kit (Zymo) as previously described.¹³⁵ Briefly, pellets were resuspended in 100 μL of STE buffer (100 mM NaCl, 10 mM Tris-HCl, pH 8.0, 1 mM EDTA) and bead beaten with ~100 μL of 0.1 mm glass beads on a FastPrep bead homogenizer (MP Biologicals) for 20 s at 4.0 m/s. Next, Zymo Quick RNA lysis buffer was added and samples bead beaten as before. Beads were removed by centrifugation at 10 000g for 30 s. RNA was isolated from supernatants according to the manufacturer’s protocol. Complementary strands of DNA (cDNA) were prepared from 500 ng RNA using a SensiFast cDNA kit per manufacturer’s instructions (manufacturer). Quantitative real time PCR (qPCR) of cDNA (1 μg) was performed using FAST SYBR green (ThermoFisher) on a QuantStudio3 qPCR machine (Applied Biosystems). For quantification of C. difficile within fecal-bioreactor MUC2 biofilms, cycle of threshold values (C_T) for C. difficile-specific primers were matched to a standard curve of C. difficile CFUs.

Mammalian Cell Culture.

Human colonic mucusproducing adenocarcinoma cell line LS174T ATCC CL-188, T84 (ATCC CCL-248), HT29-MTX-E12 (Sigma 12040401), and T84 (ATCC CCL-248) were maintained in DMEM medium (ATCC) supplemented with heat-inactivated 10% fetal bovine serum (FBS) (Invitrogen) at 37 °C, 5% CO₂. Additionally, Vero (ATCC CCL-81) cells were grown in Dulbecco’s modified Eagle’s medium (ThermoFisher) supplemented with heat-inactivated 10% fetal bovine serum (FBS) at 37 °C, 5% CO₂. All cell lines were incubated in a humidified atmosphere at 37 °C, 5% CO₂ and routinely tested for Mycoplasma (Mycoplasma Detection Kit Lonza, cat# LT07–518).

Vero cells were used to assess C. difficile toxin activity as previously described.¹³¹ Briefly, Vero cells were grown to confluency on 96-well plates and treated with C. difficile CDMM supernatant (after 48 h growth) in DMEM at final concentration of 1:1 C. difficile supernatant to DMEM. Vero cells were imaged after 4 h incubation on a Nikon TiE inverted wide field epifluorescence microscope (Nikon) using a SPECTRAX LED light source (Lumencor), ORCA-Flash 4.0 sCMOS camera (Hamamatsu), and Nikon Elements Advanced Research v4.5 software. To determine cell rounding, FIJI (Formerly ImageJ; National Institutes of Health) was used to define cell diameter. Cell rounding was defined as a >50% reduction in Vero cell diameter. Data was collected in three regions/well, n = 3 wells; repeated two independent times.

MUC2 Purification.

MUC2 was collected from human mucin-producing LS174T and HT29-MTX cells as previously described.^136,137 Briefly, supernatant was collected from T300 flasks of cells and incubated with phenylmethylsulfonyl fluoride, EDTA, and iodoacetamide to prevent further degradation. Ice cold ethanol was added to precipitate crude mucin protein, and crude mucin was purified by guanidine chloride and cesium chloride isopycnic density gradient high-speed ultracentrifugation (70 Ti rotor, Beckman Coulter Inc.).^138,139 MUC2 fractions were identified by slot blot with a mouse anti-MUC2 antibody (Santa Cruz #sc-515032). Pooled MUC2 fractions were dialyzed, lyophilized, and resuspended in Hanks’ salt solution. MUC2 was adhered to glass coverslips (18 mm) using 3-aminopropyltriethoxysilane (APTS) as previously described.^140,141 APTS-treated coverslips were incubated with 1 mg/mL purified MUC2 overnight at 4 °C and air-dried. For MUC2 adhesion assays, C. difficile was fluorescently tagged with 10 μM CFDA-SE (ThermoFisher) in PBS anaerobically at 37 °C for 1 h, washed thoroughly and then incubated with MUC2-coated coverslips anaerobically at 37 °C for 1 h. Coverslips were washed and mounted, and C. difficile localization was examined by microscopy. O-Linked glycans were removed from purified MUC2 by beta-elimination (0.5 M NaBH₄ in 50 mM KOH, 50 °C overnight). MUC2 protein was isolated by the addition of ethanol, and the resulting supernatant containing O-linked glycans was lyophilized.

Human Fecal Bioreactors.

Fecal samples were collected from consenting adults who self-identified as healthy using protocols reviewed and approved by the Institutional Review Boards of Michigan State University and Baylor College of Medicine. The colonization of mucus coated coverslips by fecal microbial communities and C.difficile was tested in bioreactors. Replicate bioreactors and bioreactor medium (BRM2) were prepared as previously described (60) with the following modifications. Arabinogalactan, d-cellobiose, maltose, potassium phosphate monobasic, and potassium phosphate dibasic were added to BRM2. The reactors (independent 50 mL reactors in 100 mL media bottles) were inoculated with an anaerobic preparation of a 20% fecal slurries (final concentration, 3.5% [wt/vol]) and were operated in an anaerobic chamber with a 5% H₂,5% CO₂, 90% N₂ atmosphere. The fecal slurries were pooled from two different healthy donors. After 16 h of outgrowth in batch culture mode, continuous-flow cultivation was initiated at a flow rate of 6.25 mL/h (8 h retention time). After 24 h of flow, the reactor communities were treated twice daily with clindamycin (final concentration, 250 μg/mL) for 4 days. The reactors were transitioned to fresh BRM2 medium supplemented with a arabinogalactan, d-cellobiose, and maltose mixture on day 4 of clindamycin treatment and continued with this medium throughout the duration of the experiment. Simultaneously, coverslips coated with mucus isolated from either LS174T or HT29-MTX, or not coated were introduced into reactors (mock, n = 2; human MUC2 (LS174T and HT29-MTX), n = 4). All six reactors were challenged with vegetative cells of C. difficile CD2015 (clinical ribotype 027) that had been propagated in BRM2 prior to inoculation. After 3 days of coincubation, mucus-coated coverslips were removed for microbial community analysis as described below.

Bioreactor 16S rDNA Sequencing and Data Analysis.

DNA from MUC2-coated coverslips was extracted using the Zymo gDNA isolation kit (Zymo) according to the manufacturer’s instructors with the addition of two rounds of bead beating. For microbiome characterization, the V4 region of the 16S rRNA gene was amplified using the NEXTflex 16S V4 Amplicon-Seq Kit 2.0 (Bioo Scientific, Austin, TX) as previously described.¹⁴² Sequencing of the pooled 16S libraries was performed using a 500 cycle v2 chemistry kit on an Illumina MiSeq instrument (Illumina, San Diego, CA) at the Texas Children’s Microbiome Center following the standard Illumina sequencing protocol.¹⁴³ The barcode and primer sequences in the raw reads were trimmed using cutadapt (v1.10). Next, PhiX and human-derived sequences were removed using Bowtie2.¹⁴⁴ Sequences were then quality filtered using the LotuS pipeline (v1.462)¹⁴⁵ and processed as previously described.¹⁴⁶ OTUs that fail to classify as bacteria at the kingdom level and unclassified OTUs at the phylum level were excluded from the analysis, along with the mitochondrial and cyanobacterial sequences. The median sequencing depth was 22 866 reads (range, 1960–38 207). Resultant sequences were analyzed using standard bioinformatics approaches within the TCMC.

Next generation sequencing data were analyzed by several metagenomic methods including determination of bacterial diversity, evenness, richness, and relative abundance of the bacteria identified in each sample. Relative abundance of a bacterial taxon in a sample was calculated using the statistical analysis of metagenomic profiles (STAMP v12.3).¹⁴⁷ ANOVA was performed for multiple group comparisons. The p-values were corrected for multiple testing with the Benjamini–Hochberg method to control for false discovery rate. Diversity indices were computed using QIIME (Quantitative Insights Into Microbial Ecology, v1.9.1) software.¹⁴⁷ Sequences are available online on the NCBI BioSample database (accession numbers: SAMN15790567, SAMN15790568, SAMN15790569, SAMN15790570, SAMN15790571, and SAMN15790572).

Immunostaining.

For live cell adhesion assays, HT29-MTX cells were seeded at 5 × 10⁵ cells/well into Corning Costar 12-well cell culture plates (Corning, Sigma-Aldrich) containing polylysine coated coverslips and grown to confluence. HT29-MTX cells on coverslips were stained with Hoechst 33342 nuclear dye and incubated with live CFDA-SE-tagged C. difficile for 1 h at 37 °C, 5% CO₂. Following incubation, coverslips were gently washed 2× with PBS and fixed with Clark’s fixative for 30 min at RT. Mucin staining was performed with an anti-MUC2 antibody (dilution: 1:200, rabbit anti-MUC2, cat # sc-15334) incubated overnight at 4 °C. Donkey anti-rabbit-Alexa Fluor 488 diluted at 1:1000 (Life Technologies) was added and incubated for 1 h at RT. Alternatively, APTS-adhered MUC2 coverslips were incubated with CFDA-labeled C. difficile for 1 h at 37 °C and fixed in Clark’s fixative. HT29-MTX and MUC2-coated coverslips were mounted to slides with mounting media (Life Technologies). Slides were imaged on an upright wide field epifluorescence Nikon Eclipse 90i microscope (Nikon, Tokyo, Japan).

Fluorescent In Situ Hybridization (FISH).

C. difficile localization was examined in bioreactor coverslips with a C. difficile-specific FISH probe [6FAM] CATCCTGTACTGGCTCAC (Integrated DNA Technologies, IDT). Coverslips were hybridized for 45 min at 45 °C in a dark humidifying chamber. To identify other bacteria, Hoechst 33342 (Invitrogen) nuclear dye was added and incubated at RT for 10 min. Mounted coverslips were imaged on an upright wide field epifluorescence Nikon Eclipse 90i (Nikon, Tokyo, Japan).

Ex Vivo C. difficile Adherence Assays.

Carnoy’s fixed paraffin-embedded mouse colons were obtained from C57BL6/J mice (6–8 weeks of age) purchased from Jackson Laboratories and housed in the BCM animal facility (n = 3 mice). Animal care and experimental procedures were approved by the Institutional Animal Care and Use Committee at BCM (Protocol #AN-914). Mice were humanely euthanized by CO₂ with thoracotomy as a secondary method. Paraffin-embedded human colon samples were obtained from US Biomax (Colon normal tissue array #BN05014; n = 72 patients). Mouse and human colon sections were deparaffinized through a series of ethanol washes. CFDA-labeled C. difficile R20291 was incubated with deparaffinized tissue for 1 hat 37 °C. Epithelial cells were identified by staining nuclei with Hoechst 33342 for 10 min at room temperature. Mucus glycans were identified by staining with the lectin UEA-1 (Ulex Europaeus Agglutinin-1; Vector Laboratories, Rhodamine labeled #RL-1062; 1:200 dilution) for 30 min at room temperature.

To remove glycans from tissue and APTS-MUC2 coverslips, samples were processed as previously described.⁸⁸ Briefly, deparaffinized slides or coverslips were incubated with 0.1 M NaOH for 30 min at room temperature and oxidized by adding 100 mM NaIO₄ in 100 mM acetate buffer (pH 4.5) overnight at 4 °C. Reactive aldehydes were neutralized by incubating with 2% glycine solution for 30 min and beta-elimination was performed by adding 0.1 M NaOH for 30 min at room temperature. These slides and coverslips were then incubated with CFDA-labeled C. difficile and imaged on a Nikon Eclipse 90i instrument.

Identification of Microbial Glycosyl Hydrolases using the Carbohydrate Active Enzymes database (CAZy).

Bacterial glycosyl hydrolases (GH) were examined using the Carbohydrate-Active Enzyme (CAZy) database (http://www.cazy.org) as previously described.^148–151 GH copy numbers were collected from selected annotated genomes. Only GHs known to be involved in mucin degradation (GH 2, 20, 29, 33, 38 42, 84, 85, 89, 95, 101, 125, 129) were included for analysis.⁵⁵

Statistical Analysis.

GraphPad Prism software (ver. 8) (GraphPad Inc., La Jolla, CA) was used to generate all graphs. One-way or two-way ANOVA with a Bonferroni posthoc test was used for all assessment, except for ibidi chemotaxis analysis, which was performed by Student’s t test. The data are presented as mean ± SEM, with P < 0.05 (*).