Neonatal microbiota colonization primes maturation of goblet cell-mediated protection in the pre-weaning colon

Abstract

Regulated host-microbe interactions are a critical aspect of lifelong health. Colonic goblet cells protect from microorganisms via the generation of a mucus barrier structure. Bacteria-sensing sentinel goblet cells provide secondary protection by orchestrating mucus secretion when microbes breach the mucus barrier. Mucus deficiencies in germ-free mice implicate a role for the microbiota in programming barrier generation, but its natural ontogeny remains undefined. We now investigate the mucus barrier and sentinel goblet cell development in relation to postnatal colonization. Combined in vivo and ex vivo analyses demonstrate rapid and sequential microbiota-dependent development of these primary and secondary goblet cell protective functions, with dynamic changes in mucus processing dependent on innate immune signaling via MyD88 and development of functional sentinel goblet cells dependent on the NADPH/dual oxidase family member Duox2. Our findings identify new mechanisms of microbiota-goblet cell regulatory interaction and highlight the critical importance of the pre-weaning period for the normal development of protective systems that are key legislators of host-microbiota interaction.

Graphical abstract

Figure 1.

Postnatal maturation of the colonic IML barrier. (A) Sampling time points in postnatal days (P) of neonatal, weaned, and adult Wistar rats. (B) IML thickness quantified by ex vivo needle measurement. (C) IML barrier function quantified by ex vivo microbead penetration. (D) Representative confocal z-stacks used to generate data shown in C. Images show x/z-axis cross-sections of colonic tissues overlaid with microbeads. Impenetrable mucus is indicated (yellow arrows). (E) Confocal micrographs of fixed colonic tissue sections stained for DNA (grey) and Muc2 (green). Dashed lines in high-magnification P2 and P9 images show tissue surface (orange) and microbiota border (blue). (F) IML growth rate quantified by ex vivo needle measurement. (G) Principal component analysis plot of complete mucus proteome data. Samples are color coded based on age-dependent IML ex vivo phenotype: P1–P2 (low barrier function, slow growth; pink), P3–P7 (high barrier function, slow growth; teal), and P9–Adult (high barrier function, fast growth; red). (H) Volcano plot comparing complete mucus proteome data from slow growth (P1–P7) and fast growth (P9–Adult) samples. Proteins significantly enriched in slow (teal) or fast (red) growth groups are indicated. (I) Abundance of Muc2 (left), Tgm3 (middle), and Clca1 (right) proteins at different ages from label-free quantification (LFQ) of complete mucus proteome data. LOWESS curves (purple) are indicated for each dataset. Spearman correlation R and P values derived from correlating age and protein LFQ are shown. (J) Correlation of ex vivo mucus growth rate and Clca1 mucus proteome abundance (LFQ) at different ages. Simple linear regression (purple) is indicated. Pearson correlation R and P value derived from correlating mucus growth rate and Clca1 LFQ are shown. (K) Ex vivo mucus growth and response to metalloprotease inhibition by EDTA (left) or Comp PIC (right) treatment at different ages. Graphs show changes in mucus growth rate (Δ mucus growth; right) in response to inhibitor treatment. (L) AB/PAS-stained distal colon tissue sections from animals at different ages. Yellow dashed line indicates the IML. (M) Quantification of IML thickness values at different ages based on images shown in L. Data represent n = 5–19 (A–J, L, and M) or n = 5 (K) animals per group, as indicated. All data are pooled from at least three independent litters. All error-bar graphs show median and interquartile range. Statistical comparisons between groups by Kruskal–Wallis and Dunn’s multiple comparison (B, C, F, K, and M) or Welch’s t test and Benjamini–Hochberg FDR correction (H); P < 0.05 (*), <0.01 (**). Image scale bars are 100 µm. AUC, area under the curve.

Figure S1.

R elated to Fig. 1 . (A) Ex vivo mucus growth and response to metalloprotease inhibition by EDTA treatment at different ages. Graph show mucus thickness values over time in response to EDTA treatment at t = 30 min (m). (B) Ex vivo mucus growth and response to serine/cysteine protease inhibition by Comp PIC treatment at different ages. Graph shows mucus thickness values over time in response to Comp PIC at t = 30 m. (C) Proteases belonging to different classes detected in mucus proteome data of at least one age group. (D) Volcano plot comparing protease abundance in slow growth (P1–P7) and fast growth (P9–Adult) samples. Proteases color coded by class; data points size proportional to label-free quantification (LFQ) abundance values. Data represent n = 5 (A and B) or n = 3–16 (C and D) animals per group. All error-bar graphs show median and interquartile range. All data are pooled from at least two independent litters or experiments.

Figure 2.

Postnatal IML maturation is driven by both microbiota-dependent and -independent factors. (A) Total and taxon-specific colonic bacterial load in fetal and postnatal Wistar rats (left) and C57BL/6 mice (right) quantified by 16S qPCR of stool DNA. IML formation period between P2–P3 in rats is indicated. (B) Confocal micrographs of fixed mouse colonic tissue stained for luminal bacteria by 16S FISH. (C) Confocal micrographs of fixed colonic tissue sections stained for DNA (grey) and Muc2 (green). Dashed lines in high-magnification images (purple boxes) show tissue surface (orange) and IML border (yellow). Stratified Muc2 layers present in ConvR but not GF P3 IML are indicated (arrowheads) in highest magnification images (red boxes). (D) Comparison of gene expression ratios between ConvR P14:P3 mice (age dependent) and P14 ConvR:GF mice (microbiota dependent) quantified by DESeq2 analysis of bulk colonic RNA sequencing data. Genes significantly regulated by age (orange) or by both age and colonization status (blue) are indicated. (E) Ex vivo mucus growth in GF, ConvD, or ConvR tissue before (pre) and after (post) serine/cysteine protease inhibition by Comp PIC or the cysteine protease-specific inhibitor E64. (F) Ex vivo analysis of MyD88^+/+ and MyD88^−/− IML thickness and barrier function by microbead penetration. Images are x/z-axis cross-sections of confocal z-stacks showing colonic tissues (grey) overlaid with microbeads (red). Impenetrable mucus is indicated (yellow arrows). (G) Ex vivo analysis of MyD88^+/+ and MyD88^−/− IML thickness based on images shown in F. (H) Ex vivo analysis of MyD88^+/+ and MyD88^−/− IML barrier function based on images shown in F. (I) Ex vivo mucus growth in MyD88^+/+ and MyD88^−/− tissue in response to metalloprotease (EDTA) or serine/cysteine protease (Comp PIC) inhibition. Data represent n = 4–5 (A–C, E, and I), n = 6–8 (F–H), or n = 3 (D) animals per group, as indicated. All data are pooled from at least two independent litters or experiments. All error-bar graphs show median and interquartile range. Statistical comparisons between groups by DESeq2 (D), Kruskal–Wallis and uncorrected Dunn’s test (E), or Mann–Whitney test (G–I); P < 0.05 (*), <0.01 (**). Image scale bars are 50 µm (B) or 100 µm (C). AUC, area under the curve.

Figure S2.

R elated to Fig. 2 . Confocal micrographs of fixed colonic tissue sections from ConvR P2 (left) and P3 (right) mice stained for DNA (grey) and Muc2 (green). Images are representative of n = 3–5 animals per group pooled from two independent litters. Image scale bars are 40 µm.

Figure 3.

Microbiota colonization and GC maturation. (A) Sampling time points in postnatal days (P) of neonatal, weaned, and adult ConvR and GF mice. (B) Illustration of postnatal gene expression patterns with positive or negative monotonic correlation to age. (C) Total number of GC-enriched genes with positive or negative monotonic correlation to age in both ConvR and GF mice (microbiota independent) or in only ConvR or GF mice (microbiota dependent) was determined by Spearman correlation analysis of DESeq2 normalized RNA sequencing data. (D) Proportion of microbiota-dependent gene expression patterns shown in C identified by standalone DESeq2 comparison of ConvR and GF samples at individual ages. Genes not identified in any pairwise comparison are classified as “not identified” (NID). (E) Comparison of Spearman correlation coefficients (R) and GC expression enrichment (log₂ GC:enterocyte [EC] expression ratios) of genes with significant positive (orange) or negative (blue) monotonic correlation to age. Plots show GC-enriched genes with microbiota-independent (left) or -dependent (middle, right) expression patterns. (F) Heatmap showing standardized expression (z-score) of selected GC-enriched genes with significant microbiota-dependent or -independent expression correlation to age. (G) Confocal micrographs of fixed colonic tissue from adult ConvR and GF mice stained for DNA (grey) and Apo-Muc2 (red). (H) Quantification of Apo-Muc2–positive cells as a fraction of total epithelial cells in adult ConvR and GF tissue based on images shown in G. (I) Quantification Apo-Muc2–positive cells as a fraction of total epithelial cells in P1–9 neonatal ConvR and GF tissues based on images shown in Fig. S1 C. (J) Whole-mount confocal imaging of neonatal and adult RedMUC2^98tr colon stained for DNA (blue), F-actin (green), and MUC2-mCherry (red). Crypt entrances (yellow line) and intercrypt transition zones between high and low GC density areas (white dashed line) are indicated. Images show x/y-axis maximum intensity projections. Data represent n = 2–4 (C–F), n = 4–7 (G–I), or n = 4 (J) animals per group, as indicated. All data are pooled from at least two independent litters or experiments. All error-bar graphs show median and interquartile range. Statistical comparisons between groups by Spearman correlation with Benjamini–Hochberg FDR correction (C–F), Mann–Whitney test (H), or two-way ANOVA and Fisher’s Least Significant Difference (LSD) (I); P < <0.001 (***), <0.0001 (****). Image scale bars are 50 µm.

Figure S3.

R elated to Fig. 3 . (A) Volcano plot of bulk mRNA sequencing data illustrating GC-enriched genes differentially expressed between ConvR and GF adult colon. (B) Reactome pathways significantly enriched in microbiota-regulated GC genes identified as positively correlated with postnatal age. (C) Confocal micrographs of fixed colonic tissue from adult ConvR and GF mice stained for DNA (grey) and Apo-Muc2 (red). (D) Quantification of mCherry⁺ GCs from colonic whole mounts based on images shown in Fig. 3 J. P3 samples are divided into late epithelium (LE) and early epithelium (EE). (E) Quantification of total epithelial cells/crypt in P1–9 neonatal and adult ConvR and GF tissues based on images shown in Fig. 3 G and Fig. S1 C. (F) Quantification of Apo-Muc2–positive cells as a fraction of total epithelial cells in adult GF and conventionalized (ConvD) mice at different time points after colonization. Data represent n = 4 (A–D) or n = 3 (D) animals per group. All error-bar graphs show median and interquartile range. All data are pooled from at least two independent litters or experiments. Statistical comparisons by one-way ANOVA and Fisher’s LSD; P < 0.05 (*), P < 0.01 (**). Image scale bars are 50 µm.

Figure 4.

senGC maturation is microbiota dependent. (A) Ex vivo mucus growth in adult GF and ConvR mouse colon after stimulation with bacterial MAMPs. (B) Ex vivo mucus growth dose response to P3CSK4 in adult GF and ConvR mouse colon. (C) Ex vivo mucus growth in adult GF and ConvR mouse colon stimulated with P3CSK4 in the presence or absence of senGC activation inhibitors. (D) AB/PAS-stained tissue sections from ex vivo experiments illustrated in A. Emptied upper crypt GCs (red arrowheads) and lower crypt cavitation (yellow arrowheads) indicated. (E) Whole-mount confocal imaging of adult GF mouse colon treated with fluorescent dextran tracer. Images show x/y-axis (upper panel) and x/z-axis (lower panel) cross-sections illustrating dextran uptake by an upper crypt GC (purple arrowhead). (F) Ex vivo mucus growth in neonatal (P3, 5, and 15) and postweaning (P33) rat colon stimulated with P3CSK4 in the presence or absence of Dynasore inhibitor. (G) Standardized expression of genes (columns) encoding known and predicted secreted proteins upregulated in mucus from P9-adult compared with P1–P7 rats (see Fig. 1 H) in GC subpopulations (rows) identified by scRNA-seq. “Secretion” row indicates evidence of secretion determined by prior annotation or in silico predication of classical or nonclassical secretion by SecretomeP. (H) Quantification of the frequency of Tgm3-expressing GCs as a proportion of the total GC population in neonatal (P3, P9, P14, and P19) and postweaning (P24) colonic tissue sections from ConvR mice. (I) Confocal micrographs of representative tissue sections from P3 and P14 ConvR mice stained for Tgm3 (green) or the epithelial border and GC-binding lectin WGA (grey) and the GC-specific lectin UEA1 (red). The epithelial surface (blue dashed line) and an individual GC from each image is indicated (yellow dashed line). Data represent n = 3–5 (A–F, H, and I) animals per group, as indicated. All data are pooled from at least two independent litters or experiments. All error-bar graphs show median and interquartile range. Statistical comparisons between groups by two-way ANOVA and Fisher’s LSD (A and C) or Kruskall–Wallis and uncorrected Dunn’s test (F and H); P < 0.05 (*), <0.01 (**), <0.001 (***), <0.0001 (****). Image scale bars are 50 µm (D and I) or 20 µm (E). # note: ConvR data displayed in A and B are reproduced from our previous publication (Birchenough et al., 2016) and are shown for illustrative purposes only.

Figure S4.

R elated to Fig. 4 . (A) Filled upper crypt GCs/crypt quantified from AB/PAS-stained tissue sections shown in Fig. 4 D. (B) Percentage of cavitated crypt bases quantified from AB/PAS-stained tissue sections shown in Fig. 4 D. (C) Processing pipeline for determining the expression of age-regulated mucus protein genes (see Fig. 1 H) in GC clusters identified by scRNA-seq. Databases and software used are shown in parentheses. Rn, R. norvegicus; Mm, M. musculus. (D) UMAP plot of scRNA-seq data from isolated colonic GCs highlighting previously annotated GC clusters (Nystrom et al., 2021). (E) Per GC cluster standardized gene expression (z-scores) of secreted proteins (n = 72) identified as significantly enriched in the P9-Adult compared with P1–P7 colonic mucus proteome (see Fig. 1, G and H; and Fig. 4 G). (F) UMAP plot illustrating log₂ expression of Tgm3 in colonic GC scRNA-seq data. Data from n = 5 replicates (A and B) or aggregated from n = 2 independent experiments (D–F). Graphs show median and interquartile range. Statistical comparisons between clusters by two-way (A and B) or one-way (E) ANOVA and Fisher’s LSD; P < 0.0001 (****).

Figure 5.

Microbiota-dependent induction of senGC function. (A) Total bacterial load in the colon of rats at different postnatal days quantified by 16S qPCR of stool DNA. (B) Principal coordinate analysis of postnatal rat microbiota beta diversity (unweighted UNIFRAC) based on metataxonomic 16S sequencing of DNA from stool samples. (C) Linear discriminant analysis (LDA) size effect analysis of bacterial taxa significantly enriched in stool from rats at different ages. Taxa enrichment in specific age groups is indicated. (D) Ex vivo mucus growth in adult conventionalized (ConvD) and B. fragilis monoassociated mouse colon stimulated with P3CSK4 in the presence or absence of senGC activation inhibitors targeting endocytosis (Dynasore) or inflammasome activation (Casp IP). (E) Total bacterial load in colon of conventionally raised (CR), ConvD, and monoassociated mice quantified by 16S qPCR of stool DNA. (F) Ex vivo mucus growth in adult MyD88^+/+ and MyD88^−/− ConvR, GF, and 4-wk (w) ConvD mouse colon stimulated with flagellin in the presence or absence of a senGC activation inhibitor targeting inflammasome activation (Casp IP). (G) Ex vivo mucus growth in adult Nlrp6^+/+ and Nlrp6^−/− ConvR, GF, and 4-wk ConvD mouse colon stimulated with P3CKS4 in the presence or absence of a senGC activation inhibitor targeting inflammasome activation (Casp IP). (H) Principal coordinate analysis of microbiota beta diversity (Bray–Curtis dissimilarity) based on metataxonomic 16S sequencing of DNA from ConvR and ConvD stool samples. (I) Linear discriminant size effect analysis of bacterial taxa significantly enriched in stool from mice with the senGC⁻ or senGC⁺ phenotype. (J) Principal coordinate analysis of microbiota beta diversity (Bray–Curtis dissimilarity) based on metataxonomic 16S sequencing of DNA from ConvR WT, ConvD WT, and ConvD MyD88^−/− and Nlrp6^−/− stool samples. (K) Relative abundance (RA) of the genus Mucispirillum in ConvD Nlrp6^+/+ and Nlrp6^−/− mice determined by metataxonomic 16S sequencing of stool DNA. (L) Standardized abundance (z-score) of bacterial taxa identified in F in 16S sequencing data from ConvD WT and ConvD MyD88^−/− and Nlrp6^−/− stool samples. Data represent n = 4–9 animals per group, as indicated. All data are pooled from at least two independent experiments or litters. Where relevant (A–C, E, and H–K) experimental groups are color coded by the absence (senGC⁻; blue) or presence (senGC⁺; brown) of the senGC-dependent secretory response. All error-bar graphs show median and interquartile range. Statistical comparisons between groups by two-way ANOVA and Fisher’s LSD (D, F, and G), Kruskal–Wallis and Dunn’s multiple comparison (A and E), PERMANOVA (B, H, and J), or Mann–Whitney test (K); P < 0.05 (*), <0.01 (**), <0.001 (***), <0.0001 (****).

Figure 6.

Induction of senGC maturation coincides with colitis protection. (A) Timing of colitogenic DSS challenge in relation to microbiota-dependent induction of senGC maturation in adult GF mice. (B) Tracking of day zero normalized mass changes in H₂O- and DSS-treated ConvD mice from senGC⁻ and senGC⁺ groups. (C) Photographs of colonic tissue dissected from different treatment groups at the termination of the DSS challenge experiment. (D) Quantification of total colon length in different treatment groups based on images shown in C. All data normalized to average colon length of the H₂O control animals for each group. (E) Quantification of CLN mass in different treatment groups. All data normalized to average CLN mass of the H₂O control animals for each group. (F) Quantification of spleen mass in different treatment groups. All data normalized to the average spleen mass of the H₂O control animals for each group. (G) Determination of mucosal bacterial load in different treatment groups by 16S qPCR of colonic tissue DNA. (H) Determination of bacterial translocation from the colon in different treatment groups by 16S qPCR of CLN DNA. (I) Combined histology scores from different treatment groups determined by histopathological scoring of fixed colonic Swiss roll tissue sections. (J) Individual scores for different histopathology metrics from different treatment groups. (K) Micrographs of AB-PAS–stained fixed colonic Swiss roll tissue sections used for histology scoring. Areas of ulceration (asterisk) and GC depletion (yellow arrowhead) in senGC⁻ DSS-treated tissue are indicated. (L) Ulcer coverage in DSS-treated mice as a fraction of total mucosal surface area. Data represent n = 4–7 animals per group, as indicated. All data are pooled from two independent experiments. All error-bar graphs show median and interquartile range. Statistical comparisons between groups by Kruskal–Wallis and uncorrected Dunn’s (B–J) or Mann–Whitney test (L); P < 0.05 (*), <0.01 (**), <0.001 (***), <0.0001 (****). Image scale bars are 2 cm (C) or 100 µm (K).

Figure S5.

R elated to Figs. 6 and 7. (A) Quantification of senGC-dependent secretion in mice prior (day [d]0), during (d1–2), or after (d3–4) exposure to DSS in drinking water by ex vivo mucus growth stimulated with P3CSK4 or carbachol (CCh). (B) Quantification of endocytotic GCs in fixed colon whole mounts of the same mice analyzed in A. (C) Volcano plot of bulk mRNA sequencing data illustrating differential gene expression between ConvD 3-wk (w) and ConvD 2w colonic tissue. (D) Representative micrographs of colonic tissue sections from Duox2^fl/fl and Duox2^ΔIEC mice after immunohistochemical staining for Muc2. Data represent n = 3–4 animals per group. All data are pooled from at two independent experiments. Statistical analysis by two-way (A) or one-way (B) ANOVA and Fisher’s LSD test; P < 0.01 (**), P < 0.001 (***), P < 0.0001 (****). Image scale bars are 100 µm.

Figure 7.

Microbiota induction of senGC maturation via regulation of Duox2. (A) Schematic of the senGC activation pathway highlighting known (black) and putative (red) pathway genes. (B) Expression of known and putative senGC genes in FACS-isolated colonic GCs and colonocytes determined by DESeq2 analysis of bulk RNA sequencing data. (C) Comparison of gene expression ratios between P22 ConvR:GF mice and adult 3-wk ConvD:GF mice quantified by DESeq2 analysis of bulk colonic RNA sequencing data. Genes significantly upregulated (red) or downregulated (blue) by microbiota exposure in both P22 and ConvD mice are indicated. (D) Proportion of unique and shared genes significantly regulated by microbiota exposure in P22 ConvR and adult 3-wk ConvD mice, based on data shown in C. (E) Comparison of microbiota-dependent expression of known and putative senGC activation pathway genes (A and B) in P22 ConvR and adult 3-wk ConvD mice. Subset of data shown in C. Genes not significantly regulated by microbiota in either group (grey) or genes regulated in either P22 ConvR (purple), adult ConvD (yellow), or both groups (teal) are indicated. (F) Relative expression (compared with GF) of Duox2 (left) and Nox1 (right) genes in ConvD (brown) and B. fragilis monoassociated (blue) mice from 1 to 4 wk (w) colonization. Expression determined by qRT-PCR of colonic RNA, normalized to Gapdh and Rplp0 expression. (G) Expression of Duox2 (left) and Nox1 (right) genes in postnatal ConvR (purple; P3–33) and GF (teal; P9–P33) determined by DESeq2 analysis of bulk colonic RNA sequencing data. (H) Confocal micrographs of fixed colonic tissue sections from ConvR WT mice stained for Duox2 (left) and Nox1 (right) mRNA by in situ RNA hybridization and counterstained by Epcam (grey). Duox2- or Nox1-expressing crypt regions are indicated (yellow arrowheads). (I) Confocal micrographs showing upper crypt GCs in fixed colonic tissue sections from ConvR, GF, and ConvD mice stained for Duox2 (red), mucus (UEA1; green), actin (grey), and DNA (blue). Intracellular Duox2 in GCs are indicated (yellow arrowheads). (J) Ex vivo mucus growth in Duox2^fl/fl and Duox2^ΔIEC colon tissue treated with carbachol (CCh), LPS, or P3CSK4. (K) Ex vivo mucus growth in WT colon tissue treated with P3CSK4 in the presence or absence of the Nox1 inhibitor ML171. Data represent n = 2–5 animals per group, as indicated. All data are pooled from at least two independent experiments or litters. All error-bar graphs show median and interquartile range. Statistical comparisons between groups by DESeq2 (B, C, and E), Kruskal–Wallis and Dunn’s multiple comparison (F), or two-way ANOVA and Fisher’s LSD (J and K); P < 0.05 (*), <0.01 (**), <0.001 (***), <0.0001 (****). Scale bars are 50 µm (H) or 5 µm (I). FC, fold change.