Report

Proximal colon–derived O-glycosylated mucus encapsulates and modulates the microbiota

See allHide authors and affiliations

Science  23 Oct 2020:
Vol. 370, Issue 6515, pp. 467-472
DOI: 10.1126/science.aay7367

So much more to mucus

Mammals accommodate a dense community of metabolically active microorganisms in their gut. This is not a passive relationship, and host and microbe have antagonistic as well as mutualistic responses to each other. Using a whole-colon imaging method in mice, Bergstrom et al. looked at the role of colonic mucus in segregating the microbiota from host cells during elimination of feces (see the Perspective by Birchenough and Johansson). Host goblet cells synthesize two forms of mucin that differ in branched chain O-glycosylation and the site of production in the colon. A “thick” mucus in the proximal, ascending colon wraps the microbiota to form fecal pellets. Transit along the distal, descending colon is lubricated by “thin” mucus that transiently links with the thick mucus. Normal mucus encapsulation prevents inflammation and hyperplasia and thus is important for maintenance of a healthy gut.

Science, this issue p. 467; see also p. 402

Abstract

Colon mucus segregates the intestinal microbiota from host tissues, but how it organizes to function throughout the colon is unclear. In mice, we found that colon mucus consists of two distinct O-glycosylated entities of Muc2: a major form produced by the proximal colon, which encapsulates the fecal material including the microbiota, and a minor form derived from the distal colon, which adheres to the major form. The microbiota directs its own encapsulation by inducing Muc2 production from proximal colon goblet cells. In turn, O-glycans on proximal colon–derived Muc2 modulate the structure and function of the microbiota as well as transcription in the colon mucosa. Our work shows how proximal colon control of mucin production is an important element in the regulation of host-microbiota symbiosis.

In mammals, the colon houses a dense and diverse commensal microbiota, which has an impact on both health and disease (14). The colon mucus system modulates host-commensal symbiosis (510). Mucus is primarily a large polymeric network of Mucin-2 in mice and humans (Muc2 or MUC2, respectively), which is heavily O-glycosylated (11, 12). An established model is that colon mucus is primarily produced by goblet cells in the distal colon to initially form a single nonmobile inner mucus layer, tightly attached to the epithelium, which segregates microbiota from the colon mucosa (5, 13). The inner mucus layer subsequently becomes a loose mobile outer layer where microbiota reside (13, 14). However, how the mucus system forms and functions along the whole colon is unclear.

To address this question, we developed a whole-colon imaging method to analyze the origin and composition of mouse colonic mucus via sections of wild-type “colon coils” labeled with anti-Muc2 antibody, Maackia amurensis lectin II (MALII, recognizing α2,3-linked sialylated and sulfated glycans) (15), and the general bacterial probe EUB338 (Fig. 1A and fig. S1, A to C) (16). We discovered that MALII was negative in most proximal colon goblet cells but was strongly positive in distal colon goblet cells (Fig. 1, A and B). In contrast, MALII was negative in the mucus layers throughout the colon, except in a previously undescribed minor portion of the inner layer closest to the distal colon mucosal surface (fig. S2A). This suggests that (i) luminal mucus is mainly derived from the proximal colon, a finding supported by increased size and number of goblet cells and higher Muc2 expression in the proximal versus distal colon (Fig. 1C and fig. S2B); and (ii) the inner layer consists of two distinct entities of Muc2 from the proximal or distal colon goblet cells. Notably, we found that MALII primarily marked sulfated O-glycans on Muc2 (fig. S3). Transcriptome analysis of healthy proximal versus distal colon revealed a significant difference in their gene expression profiles (>1000 genes, including Muc2) even among similar cell types including goblet cells. Collectively, these data underscore the functional uniqueness of these two colon regions (fig. S1D).

Fig. 1 Proximal colon–derived mucus encapsulates microbiota-containing fecal pellets.

(A) Tiled immunofluorescent image of a longitudinal Carnoy’s fixed paraffin-embedded (CFPE) wild-type whole colon coil section (left) with goblet cells highlighted (blue-boxed images, right). Prox, Mid, and Dis denote proximal, middle, and distal colon, respectively. (B) Left: Enumeration of Muc2+MALII goblet cell numbers (GCs) in different colon regions. Right: Schematic GC Muc2 and MALII expression profile. (C) Left: Number of GCs per crypt. Right: Muc2 expression in colon epithelium [quantitative polymerase chain reaction (qPCR) normalized to Villin and Hprt]. (D) Magnified images of red boxed areas in (A). (E) Representative droplet digital PCR (ddPCR) scatterplot of the bacterial 16S rRNA gene of samples from wild-type distal colon with (P) or without pellets (inter-pellet, IP). H2O, negative control; +, positive control. (F) Mean ddPCR 16S rRNA copy number among replicates. Each point represents one animal; each color denotes the same animal. (G) Left: Alcian blue–stained sections of tissues and feces. Right: Violin plot of median mucus thickness (median, red; quartiles, black). (H) Top left: Z-stacked confocal images of cross sections of colon with pellet or an excreted pellet wrapped in peritoneum; the narrow image at right shows the corresponding y/z plane. Bottom left: Single-colored images of white-boxed regions. Right: Thickness of the mucus layers. Scale bars, 4 mm [(A), left]; 20 μm [(A), right, (D), (G), and (H)]. Results are representative of two (C) or three [(A), (B), (G), and (H)] independent experiments each with n = 4 or 5 mice per experiment, or n = 4 mice pooled from two independent extractions [(E) and (F)]. Data are means ± SD [(B), (C), (F), and (H)]. *P < 0.05, **P < 0.01 [two-tailed unpaired t test in (C), left, and (F); Mann-Whitney test in (C), right, (F), and versus b2 layer in (H)]; ns, not significant.

Whole-colon imaging showed that the luminal mucus layers were attached to bacteria-dense fecal pellets to create “bacteria-sparse” zones between pellets (Fig. 1, D to F, and fig. S2C) (16). This was confirmed by colonizing mice with a green fluorescent protein (GFP)–tagged gut symbiont, Bacteroides thetaiotaomicron (B. thetaGFP) (17), which was only found inside mucus-coated fecal pellets (fig. S4). These data indicate that proximal colon–derived mucus encapsulates bacteria in fecal pellets, which supports the idea that the mucus can associate with fecal pellets (18, 19). We found that the mucus structure and segregating functions on freshly excreted fecal pellets and on fecal pellets in situ within their corresponding colon tissues were identical (Fig. 1, G and H, and fig. S5D) (16). [A mucus barrier layer was also detected on fecal surfaces of healthy primates (baboons) and humans (fig. S6).] These findings in mice were validated using unfixed (fig. S5) or unprocessed (fig. S2D) pellet-containing colon tissues and excreted pellets, excluding fixation artifacts.

Further analysis of the formation of the mucus coating revealed a stepwise establishment of the fecal-associated mucus layers, with a Muc2+MALII layer associating with feces first in the proximal colon, followed by addition of a Muc2+MALII+ layer in the distal colon coinciding with appearance of Muc2+MALII+ goblet cells (fig. S7). The Muc2+MALII+ layer likely forms after compression as the fecal mass pushes it against the mucosal wall (fig. S8). Intact fecal mucus could be successfully extracted, and glycomics analysis further confirmed its similarity to proximal colon– versus distal colon–derived mucus (fig. S9 and table S1) (16). The mucus phenotype was reproduced in different inbred and outbred strains of mice of both sexes at different ages (fig. S10, A to E), as well as in rats (fig. S10, F and G). On the basis of these findings, we redefine the mucus system to reflect its origin, biochemical structures, and function (fig. S10H): The colon mucus system consists of a proximal colon–derived mucus niche” and “barrier” layer, the latter composed of a proximal colon–derived “b1 layer” and a distal colon–derived “b2 layer” (Fig. 1H and fig. S10H).

We then asked how the b1 and b2 mucus barrier layers contribute to host-microbiota interactions in the colon. More than 80% of the mouse colon mucin mass is derived from ubiquitously expressed core 1–derived and proximal colon–expressed core 3–derived complex mucin-type O-glycans (20, 21). To determine the contribution of these O-glycans from different colon regions to overall mucus function, we generated mice lacking both types of O-glycans preferentially in the proximal colon (TM-DKOProx), in the distal colon (TM-IEC C1galt1–/–) (20), and in both the proximal and distal colon (TM-DKO) (Fig. 2A and fig. S11) (16). Analysis of mucus barrier status in tissues and fecal pellets showed significantly reduced mucus thickness in distal colons of TM-DKOProx versus wild-type mice (Fig. 2A), characterized by loss of the b1 layer, whereas the b2 layer remained relatively intact (Fig. 2B). In contrast, wild-type mice had both b1 and b2 layers, TM-IEC C1galt1–/– mice had only a visible b1 layer, and TM-DKO lacked both (Fig. 2B). Loss of the b1 layer, but not the b2 layer, also significantly reduced the overall distance of the microbes to the mucosal tissue (Fig. 2B, lower panel). These results indicate that proximal colon–derived O-glycans form the b1 layer critical for segregating the microbiota from the host tissue.

Fig. 2 Proximal colon–derived mucin-type O-glycans govern the composition and function of the mucus barrier.

(A) Top left: Diagrams depicting simplified forms of core 1– or core 3–derived O-glycans or Tn antigen (GalNAc) in the different colon regions of mouse lines used for results below. Bottom left: Representative images of colon sections stained with Alcian blue. Arrows denote the mucus barrier layer. Top right: Proportion of Alcian blue+ goblet cells in different colon regions. Bottom right: Violin plot of median mucus thickness in distal colons. (B) Top: Z-stack confocal images of colon sections; b1, b2, and niche identify the different sublayers of the mucus. Bottom left: Mean mucus thickness in distal colons. Bottom right: Histogram showing frequency distribution of individual bacterial distances to the top of the distal mucosa surface shown at the top. (C) Left: Confocal tiling images of cross sections of fecal pellets stained with the lectin UEA1. Arrows mark the mucus layer. Right: Violin plot of mucus thickness of the entire mucus layer surrounding the fecal pellet represented at left. (D) Blended Z-stack confocal images of proximal colon cross sections. (E) Left: Representative images of hematoxylin and eosin–stained colon cross sections. Right: Graph of histologic colitis scores. TM, tamoxifen. Scale bars, 20 μm [(A), (B), and (D)], 500 μm (C), 50 μm (E). Results are representative of three independent experiments, each with n = 3 to 5 mice per group. Data are means ± SD [(B) and (E)] or median (red) and quartiles (black) [(A) and (C)]. *P < 0.05 [two-tailed Mann-Whitney test in (A), top right; Kruskal-Wallis test with Dunn’s multiple-comparisons posttest in (A), bottom right, and (C); one-way analysis of variance (ANOVA) with Tukey’s or Dunnett’s multiple-comparisons posttest in (B) and (E), respectively].

To confirm that fecal pellets were encapsulated by proximal O-glycosylated mucus, we performed complementary analyses, which demonstrated that (i) the mucus coating of excreted pellets was diminished in TM-DKOProx pellets relative to wild-type and TM-IEC C1galt1–/– mice (Fig. 2C) as a result of rapid degradation of proximally secreted mucus (Fig. 2D); (ii) colonized B. thetaGFP was not encapsulated in TM-DKOProx mice (fig. S12); (iii) the b1 and b2 layers were rescued on mucus-deficient fecal pellets transplanted into the proximal colon of wild-type but not TM-DKO mice (fig. S13); (iv) gavaged fluorescent beads were sequestered in the mucus-coated pellets within the colon of wild-type but not mutant mice (fig. S14, A to D); and (v) there was increased expression of 16S rRNA within the inter-pellet tissues of TM-DKOProx and TM-DKO mice relative to the wild type (fig. S14, E to G). The defective mucus coating was also observed in the native state (fig. S14, H and I). Physiopathologically, defective mucus encapsulation of fecal pellets in TM-DKOProx mice was associated with increased susceptibility to spontaneous colitis (Fig. 2E) as well as acute intestinal injury (fig. S15), although disease severity was worst in TM-DKO mice, indicating that both b1 and b2 layers contribute to mucosal homeostasis. As expected, TM-IEC C1galt1–/– mice developed disease as a result of their impaired mucus barrier function (Fig. 2, C and E, and fig. S16) (21). Notably, Il10–/– mice, an alternative model of microbiota-dependent colitis (22), had an impaired b1 layer, associated with altered proximal colon homeostasis and increased distal mucosal microbial intrusion, despite an intact b2 layer (fig. S17). Collectively, our data show that the proximal colon is essential in producing the O-glycosylated mucus that encapsulates the microbiota, limiting it from interacting with the mucosa and causing disease as it migrates distally (fig. S16).

To determine whether the microbiota influenced its own encapsulation by O-glycosylated mucus, we compared the mucus system in microbiota-replete or -deficient hosts. The fecal mucus coating in germ-free (GF) fecal pellets was marginal relative to specific pathogen-free (SPF) fecal pellets (fig. S18), which suggests that the microbiota promotes its encapsulation via mucus production. Comparison of mucus formation in GF versus conventionalized littermates (ExGF), and in SPF mice versus their littermates treated with broad-spectrum antibiotics (SPFAbx) (Fig. 3A) indicated that ExGF mice [day 21 after conventionalization (D21)] had a robust mucus layer similar to that of SPF mice (Fig. 3A). By contrast, SPFAbx mice, like GF mice, had a poorly formed mucus layer, which was restored when the microbiota was reconstituted (SPFRecon) (Fig. 3A). Longitudinal analysis of the fecal mucus coating in ExGF mice before and after conventionalization showed that the encapsulation was formed by D7 (Fig. 3, B and C). This result was reproduced in ExGF mice mono-associated with B. thetaGFP (fig. S19), indicating that a complex microbiota is not required for microbiota-induced formation of fecal-associated O-glycosylated mucus.

Fig. 3 The microbiota directs its own encapsulation by inducing Muc2 production in the proximal colon.

(A) Gross picture of colon (top left) and Alcian blue–stained mucus and goblet cells in colon sections (bottom left) of mice with and without a complex microbiota. Arrows mark mucus barrier layer. Right: Violin plot of median mucus barrier layer thickness in distal colon sections. (B) Left: Formation of the fecal mucus barrier over time in ExGF mice. Images show Alcian blue–stained excreted fecal sections. Insets show magnified images of mucus layer (arrows). Right: Violin plot of mucus thickness over time in the same mice. (C) Violin plot of median mucus barrier layer thickness of excreted feces from multiple GF and ExGF mice. (D) Confocal Z-stack images of distal colon (left) and excreted fecal pellet sections (right). Arrows denote the b2 layer; white line spans the thickness of the b1 layer; yellow dashed line denotes mucosal surface. Graphs of b1 and b2 layer thickness are at the right of each group of images. (E) Left: Western blot of tissue Muc2. Right: Densitometry of Muc2 signal. GAPDH, loading control. (F) Alcian blue–stained GF and ExGF colon sections. Arrowheads mark mucus secretion. (G) Confocal Z-stack images of GF and ExGF colon sections. Arrowheads, secreted mucus. (H) Tiled confocal images of excreted fecal pellet sections from mice before or after inoculation with a complex microbiota. Scale bars, 10 μm [insets in (B)], 20 μm [(A), (D), (F), and (G)], 500 μm (B), 1000 μm (H). Results are representative of three independent experiments, each with n = 3 to 5 mice per group. Data are median (red) and quartiles (black) [(A) to (C)] or means ± SD [(D) and (E)]. *P < 0.05 [Kruskal-Wallis test with Dunn’s multiple-comparisons posttest in (A), versus D0 in (B), and versus GF b1 layer in (D); two-tailed Mann-Whitney test in (C); one-way ANOVA with Tukey’s multiple-comparisons posttest (E)].

We further probed whether and how microbiota regulate proximal colon goblet cell function. The microbiota induced formation of the b1, but not the b2, mucus barrier layer in tissues and on fecal pellets in ExGF (Fig. 3D) and SPFRecon mice (fig. S20A). This correlated with higher Muc2 expression and secretion in proximal colons of colonized (ExGF, SPFRecon) versus microbiota-deficient littermates, as shown by Western blot (Fig. 3E and fig. S20C), Muc2 staining (Fig. 3F and fig. S20, B, D, and E), and glycan metabolic labeling (fig. S21). We further noted that the secretory response segregated the microbiota away from the tissue (Fig. 3G and fig. S20, D and E) and was independent of inflammasome signaling (fig. S21B), a known modulator of goblet cell secretory responses (23). Fecal mucus screening revealed that the microbe-induced encapsulation was observed only in mouse lines with complex O-glycosylation pathways retained in the proximal colon epithelium (i.e., ExGF, SPFRecon, and TM-IEC C1galt1–/–;Recon) (Fig. 3H). These studies indicate that the microbiota specifically induces the b1 mucus barrier layer by inducing Muc2 expression and secretion in proximal colon goblet cells, leading to microbiota encapsulation.

To determine the impact of the microbe-induced encapsulation of fecal pellets by proximal colon–derived O-glycosylated mucus on the microbial ecosystem and host responses, we performed system-level analyses of the microbiota composition, metabolic output, and host transcriptome in TM-DKOProx mice and their control littermates (three independent groups) with a normalized community (fig. S22, A and B) (16). Analysis of microbiota composition by taxonomic classification and quantitation of shotgun metagenomics from proximal and distal lumen contents (table S2) showed that, beyond group-specific features (PC1/2; Fig. 4A), there were differences between TM-DKOProx communities versus littermate controls common to all groups (PC3; Fig. 4A and fig. S22D), independent of colon region and α-diversity (fig. S22C). Among taxa reduced in TM-DKOProx versus control were Akkermansia muciniphila, a known mucin forager (24), Turicibacter sp. H121, and Bifidobacterium pseudolongum (Fig. 4B); among taxa increased were Ruminococcus champanellensis and members of genus Bacteroides (Fig. 4B and fig. S22D). Peak-to-trough analysis (25) on metagenomes of A. muciniphila and B. thetaiotaomicron indicated a contribution of proximally derived O-glycosylated mucins to microbial replication rates (fig. S23A).

Fig. 4 Loss of O-glycans in the proximal colon alters microbiota community structure and the host mucosal transcriptome throughout the colon.

(A) Principal components analysis (PCA) of species-level counts. The first three principal components (PC1, PC2, PC3) and percent of variance explained are shown; n = 5 mice per group. All mice received tamoxifen injections of the same dose, lot, and time point (16). (B) Relative abundance for selected taxa showing the highest PC3 loadings in (A); the y axis represents variance-stabilized data computed from species-level counts (Bracken classifier). (C) Partial least-squares discriminant analysis (PLS-DA) of microbial metabolites. (D) Metabolite correlation networks between control (n = 5) and TM-DKOProx (n = 5). (E) Venn diagram of pairwise-correlated metabolites. (F) Gene expression changes common to proximal and distal samples [left to right (LTR)–adjusted P < 0.01; only group 1 is shown for clarity]. Bars show relative (fold) changes in TM-DKOProx versus Control. Boxes mark genes expressed in colon epithelium; those specifically annotated in cell cycle and proliferation pathways (hypergeometric enrichment P < 10–12) are highlighted in green. The gray gradient highlights P values in decreasing significance. (G) TM-DKOProx versus control gene expression changes specific to distal samples. Bar plots show gene expression log2-transformed fold change (FC) in each group of mice (black and gray bars) for both proximal and distal colon; shown at right are selected top-ranked genes by global fold change, grouped by their molecular function.

Metabolic profiling (table S3) (16) of proximal fecal content from control and TM-DKOProx groups revealed that these community shifts were concomitant with an altered metabolic landscape (fig. S23, B and C) independent of cage effects (Fig. 4C). Correlation-based network analysis of intermetabolite relationships showed 3 times as many negative correlations and only 27 correlated metabolic pairs found in common between the pooled control and TM-DKOProx groups (Fig. 4, D and E, and table S4) (16). Among 50 unique primary metabolites detected in both groups, three were consistently observed only in TM-DKOProx mice: 3-hydroxybutyric acid, 2-oxoglutarate, and glucose-6-phosphate (fig. S23C).

Finally, despite intergroup variability (Fig. 4A), host transcriptomics in all groups of mice (table S5) revealed that proximal O-glycan loss led to systemic changes (in both proximal and distal regions) in a set of 53 genes, including many related to cell cycle and epithelial homeostasis (Fig. 4F and figs. S24 and S25) (16). Additionally, a set of 147 genes involved in the interaction between the microbiome and inflammation were up-regulated exclusively in distal colon of TM-DKOProx mice, whereas the same genes were only marginally or inconsistently regulated in the proximal colon (Fig. 4G) (16). Collectively, these results imply that the proximal colon O-glycans are sufficient to affect the mucosal ecosystem by altering (i) community structure by forging a growth-promoting or regulatory microenvironment for commensals, (ii) community function by altering metabolic output, and (iii) transcriptional homeostasis locally and in distal sites.

Our data indicate that the proximal colon–derived O-glycan–rich mucus forms both the niche and a primary barrier that encapsulates fecal materials and provides an enclosed ecosystem for microbiota (fig. S26). This model represents a major revision of the current mucus system model, which describes a mucus layer locally produced and attached to the distal colon tissue. The fecal-associated mucus provides new insights into microbiota metabolism and composition, and it may lead to noninvasive strategies such as fecal mucus screening for disease diagnosis.

Supplementary Materials

science.sciencemag.org/content/370/6515/467/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S26

Tables S1 to S6

References (2681)

MDAR Reproducibility Checklist

References and Notes

  1. See supplementary materials.
Acknowledgments: We thank G. Hansson and R. Newberry for reagents and mice; P. Kincade for critical comments; and R. Banks, M. Hart, M. Rheault, J. Gibon, and A. Norris for technical support. Funding: Supported by NIH grants R01DK085691 and GM103441, the Oklahoma Center for Adult Stem Cell Research, and the Stephenson Cancer Center of the University of Oklahoma. Author contributions: L.X. and K.B. conceived and designed the experiments, interpreted data, and wrote the manuscript; K.B., X.S., D.C., A.B., V.L., J.P.J., B.S., L.G., Y.K., C.H., W.Z., B.N., S.M., J.M.M., D.L.G., S.P., and N.K. performed experiments and analyzed data; and J.B., D.C., A.B., J.L.S., C.W.H., C.V.R., T.M.G., and R.P.M. commented on the project and contributed to the manuscript preparation. Competing interests: The authors have declared that no conflict of interest exists. Data and materials availability: All sequencing data have been deposited in Gene Expression Omnibus (GSE133257) and Short Read Archive (PRJNA657953). All other data are available in the manuscript or supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article