Using gnotobiotic mice to decipher effects of gut microbiome repair in undernourished children on tuft and goblet cell function

Abstract

Studies have implicated perturbations in the postnatal development of the gut microbiome as a contributing factor to childhood undernutrition. Compared to a standard ready-to-use supplementary food, a microbiome-directed complementary food (MDCF-2) designed to repair these perturbations produced superior improvements in ponderal and linear growth in clinical trials of Bangladeshi children with moderate acute malnutrition. Here, "reverse translation" experiments are performed where intact fecal microbiomes collected from trial participants before and at the end of treatment are introduced into female gnotobiotic mice just after delivery of their pups. Pups received diets designed to resemble those consumed by children in the trials to recreate "unrepaired" and "repaired" gut ecosystems. Analyses of the abundances of bacterial strains (metagenome-assembled genomes), their expressed genes, and metabolic products, combined with assessments of ponderal growth and intestinal epithelial lineage transcriptomes (single-nucleus RNA-Seq with follow-up immunocytochemistry) disclosed effects of MDCF-2 associated microbiome repair that cannot be determined, in part because "no treatment" control arms cannot be ethically incorporated into these trials. Specifically, microbiome repair in these mice produced significant increases in ponderal growth, changes in microbial gene expression consistent with a less virulent gut ecosystem and alterations in expression of i) components of cell junctions in the enterocytic and goblet cell lineages, ii) pathways for synthesis and secretion of eicosanoid immune effectors in chemosensory tuft cells, and iii) goblet cell pathways involved in glycosylation and secretion of mucin. Experiments of the type described can help formulate and test hypotheses about how microbiome repair affects host biology.

Keywords: barrier function; childhood undernutrition; goblet cells; gut microbiome-directed therapeutic foods; tuft cells.

Fig. 1.

MAG and microbial RNA-seq analyses. (A) Experimental design. (B) Line plots illustrating relative weight changes from postnatal day 15 (P15) to P32, using P15 as the reference point for comparison. Data are presented for both donor 1 and donor 2 experiments. Error bars represent SD. * indicates statistical significance with P < 0.05. (C–E) Heatmaps representing the relative abundance of colonized bacterial strains (MAGs) positively associated with WLZ in postpartum day 14 (C), postpartum day 32 dams (D), and postnatal day 32 pups (E) for the donor 1 experiment. Z-score was calculated by centering and scaling the log-transformed relative abundances of each MAG across all samples. Each column represents data from an individual animal. The inset in E provides detailed information about the WLZ-associated MAGs shown in the heatmap, including MAG ID, assigned species, β1 coefficient from the linear mixed effects model [WLZ ~ β1(MAG) + β2(study week) + (1|PID)] indicating the strength of WLZ association in the clinical trial, the Log2-fold difference in MAG abundance between “unrepaired” and “repaired” arms of the mouse experiment. The MAGs are ordered by descending β1 coefficient. Prevotella copri is highlighted in red.

Fig. 2.

Compass-based analysis of tuft cell metabolism in mice modeling unrepaired and repaired microbiomes. (A) Bar plot showing the number of metabolic reactions with significantly altered flux, as predicted by Compass, shared between donor 1 and donor 2 experiments. Data are presented for each segment of small intestine. (B) Heatmap representing the normalized number of Compass-predicted forward metabolic reactions with significantly altered flux that are shared between donor 1 and donor 2 experiments. Data are presented for select Recon2 metabolic subsystems in tuft cells and goblet cells from the duodenum, jejunum, and ileum. The blood group synthesis subsystem includes glycosylation reactions that are also involved in mucin glycosylation, as indicated by the parenthesis. Red indicates the normalized number of reactions with increased flux in the repaired arm while black indicates the normalized number of reactions with increased flux in the unrepaired arm. Color intensity represents the number of reactions with significantly increased flux in a given Recon2 subsystem within a specific cell type of an experimental arm, normalized to the total number of such reactions across all epithelial cell clusters in the same arm. The term “max” in the color code refers to a normalized value of 1. (C) Top panel is a schematic of tuft cell metabolic pathways predicted by Compass to exhibit significantly altered flux in response to unrepaired and repaired microbiomes. These pathways include arachidonic acid synthesis, prostaglandin synthesis (via the cyclooxygenase pathway), and leukotriene synthesis (via the lipoxygenase pathway). Bottom panel is a schematic representation of metabolic pathways involved in PIP2 synthesis and GPCR-phospholipid signaling in tuft cells, which facilitate the release of eicosanoid immune effectors. Enzyme group ‘a’ includes Pikfyve, Pip4k2a, Pip4K2b, Pip4K2c, Pip5k1a, Pip5k1b, Pip5k1c. Enzyme group ‘b’ includes Hcst, Pik3c2a, Pik3c2b, Pik3c2g, Pik3ca, Pik3cb, Pik3cd, Pik3cg, Pik3r1, Pik3r2, Pik3r3, Pik3r5. Enzyme group ‘c’ includes Plcb1, Plcb2, Plcb3, Plcb4, Plcd1, Plcd3, Plcd4, Plce1, Plcg1, Plcg2, Plch1, Plch2, Plcl1, Plcxd2, Plcz1. Enzyme group ‘d’ includes Inpp5a, Inpp5b, Inpp5d, Inpp5e, Inpp5j, Inppl1, Synj1. Green arrows indicate shared reactions with significantly increased flux in duodenal, jejunal, and ileal tuft cells in mice belonging to the “unrepaired” arm in both donor 1 and donor 2 experiments. Gray arrows represent reactions without significant predicted flux changes. Black arrows denote the signal transduction cascade associated with GPCR-phospholipid signaling. Blue text highlights the eicosanoid immune effectors synthesized in tuft cells. Green text represents enzymes involved in catalyzing the metabolic reactions.

Fig. 3.

Tuft cell hyperplasia and cecal butyrate level changes in response to an “unrepaired” gut ecosystem. (A) Quantification of tuft cell density across different intestinal regions based on whole-slide scanned images of DCLK1-immunostained sections. Data are presented as box and whisker plots for each donor and intestinal segment. *P < 0.05 (Mann–Whitney U test). (B) Lollipop plot illustrating the normalized enrichment score (NES) for mcSEED metabolic pathways with significantly altered expression from microbial RNA-seq data, as identified from GSEA. The length of each lollipop represents the NES, the color indicates the arm (unrepaired” or repaired) in which the pathway is enriched, and the size of the dots reflects the level of statistical significance of the enrichment (adjusted P-value). (C) Cecal butyrate levels (μmol/g cecal contents). (D) Fecal butyrate levels (μmol/g feces) in samples obtained from donor 1 and donor 2 prior to MDCF-2 treatment (unrepaired) and at the conclusion of the 3-mo intervention (repaired).

Fig. 4.

Goblet cell hyperplasia and mucin glycosylation in response to “unrepaired” gut ecosystem. (A) Schematic representation of four primary types of mucin glycosylation reactions predicted by Compass to have significantly increased flux in goblet cells exposed to the “unrepaired” microbiome. These include glycan elongation, branching, terminal modification, and fucosylation. Abbreviations: GlcNAc, N-acetylglucosamine; GalNAc, N-acetylgalactosamine; Gal, Galactose; Fuc, Fucose; Ser/Thr, Serine/Threonine. Green arrows indicate shared reactions with significantly increased flux in jejunal and ileal goblet cells present in the “unrepaired” arm across both donor 1 and donor 2 experiments. Gray arrows represent reactions with no significant changes in flux. Green text represents enzymes involved in catalyzing the metabolic reactions. (B) Goblet cell density across three intestinal regions based on whole slide scanned images of MUC2 immunostaining. Data are presented as box and whisker plots for each donor and intestinal segment. *P < 0.05 (Mann–Whitney U test).

Fig. 5.

Alterations in microbial virulence and epithelial cell tight junctions. (A) Lollipop plot of NES for virulence factor categories from VFDB with significant altered expression, as identified by GSEA from microbial RNA-seq data. (B) Representative immunostaining images showing the distribution of tight junction components ZO-1 and JAMA in the jejunums of P32 mice. Brown staining indicates positive immunoreactive signal, while blue staining is from nuclei counterstained with hematoxylin. (Scale bar, 50 μm in the Left panels and 20 μm the Right panels.) (C) Table summarizing the shared epithelial tight junctional interactions that ranked in the top 10% prioritization score across the duodenum, jejunum, and ileum in mice from both donor 1 and donor 2 experiments.