Precocious infant fecal microbiome promotes enterocyte barrier dysfuction, altered neuroendocrine signaling and associates with increased childhood obesity risk

Figures & data

Figure 1. Compositionally distinct gut microbiota classes (GMCs) in feces of 1-month-old infants exhibit differential microbiota maturity and relate to the relative risk (RR) of overweight/obesity (OW/OB) at age 2 years.

(a) Distinct GMCs identified by Dirichlet Multinomial Mixture modeling explain more than 10% of observed variation in infant fecal bacterial β-diversity (n = 349; PERMANOVA of Unweighted UniFrac distances). (b) BMI Z-score at age 2 years significantly differs between infants with distinct GMCs; In multivariate models adjusted for several confounding factors, GMC3 infants exhibit significantly higher BMI Z-scores in childhood. (c) Bacterial taxonomic comparison of GMC3 and GMC1 subjects; taxa exhibiting significant differences (zero-inflated negative binomial regression (ZINB); P_FDR < 0.05) in mean relative abundance are shown. Values are log₁₀-transformed for purposes of illustration. Bar height indicates the magnitude of between-group difference in relative abundance. Bacterial phyla are color coded as indicated. (d) Relative abundance of fungal genera differs across GMCs. (e) Microbiota-for-age Z-scores (MAZ) of higher-risk GMC3 infants are significantly greater than that of lower-risk infants (n = 349; Kruskal-Wallis; P < 2e-16), indicating accelerated microbiota maturation in high-risk infants. MAZ scores were calculated from a random forest model trained on the 50 most age-discriminatory bacterial OTUs of normal BMI infants. Pairwise comparisons were calculated using two-sided Wilcoxon rank sum tests in c. Boxplots indicated within violin plots represent the median (center), the 25^th and 75^th percentiles, and the smallest and largest values within 1.5 × the interquartile range (whiskers).

Table 1. Infant gut microbiota classes (GMCs) exhibit significantly different relative risk ratios (RRs) of developing OW/OB phenotypes at age 2 years.

	Microbial Community Types			RR^1 (95% Cl)
	GMC1(N = 141)	GMC2 (N = 130)	GMC3 (N = 78)	GMC3 vs. GMC1	GMC3 vs. GMC1	GMC3 vs. GMC2
Overweight or Obese at Age 2	16 (10%)	25 (19%)	21 (27%)	2.52 (1.33–4.79) P = 0.005	1.60 (0.85–3.03) P = 0.146	1.57 (0.91–2.73) P = 0.105

¹Adjusted for age (in days) at stool sample collection, maternal BMI at first measure during pregnancy, prenatal antibiotic and antifungal use, mode of delivery and breastfeeding status at 1-month. Between-group risk ratio significance values were calculated using log-binomial regression.

Figure 2. High-risk GMC3 and lower-risk GMC1 exhibit distinct metabolic productivity and functional capacities.

(a) Comparison of fecal samples from GMC3 (n = 32) and GMC1 (n = 28) by untargeted mass spectrometry identifies significant (P_FDR < 0.05) widespread metabolic differences. Many metabolic differences between GMCs corroborate with differential expression of targeted pathways identified in paired shotgun metagenomic data, including (b) glycerol and methyglyoxal degradation, pyruvate fermentation and (c) allantoin degradation. In b and c, enzymes (arrows) and metabolites (open circles and closed squares) with log₂ FC > |0.25| are color coded by GMC enrichments (red enriched in GMC3; blue enriched in GMC1). (d) Phenylacetate and (e) GABA levels differ between GMC3 and GMC1. Differences in normalized abundance between groups determined by zero-inflated compound Poisson (ZICP) regression (Supplementary Table S10a). In d and e, significance was calculated using Welch’s two-sided t-test, with P_FDR < 0.05 considered significant, and each dot represents an independent infant stool sample. Boxplots represent the median (center), the 25th and 75th percentiles, and the smallest and largest values within 1.5 × the interquartile range (whiskers).

Figure 3. Cell-free fecal products from GMC3 infant microbiomes who developed OW/OB phenotypes in childhood reprogram Caco-2 enterocyte transcription.

(a) GMC and OW/OB sample groupings explain a large proportion (33%) of observed variance in Caco-2 enterocyte response to cell-free fecal products of 1-month-old infants (Blue = GMC1 normal BMI, n = 4; Red = GMC3 OW/OB, n = 7; PERMANOVA of Euclidean distances). (b) Gene-set analysis of transcripts associated with intestinal epithelial cell states (Parikh et al. 2019) indicate enrichment of enteroendocrine precursors in response to cell-free fecal products of GMC3 OW/OB infants. (c) Heat map and (d) volcano plot of significantly (P_FDR < 0.05) differentially (log₂ FC > |1|) expressed genes observed following exposure of enterocytes to cell-free fecal products of GMC3 OW/OB (n = 7) or GMC1 Normal BMI (n = 4) infants. Genes associated with obesity phenotypes and distinct intestinal epithelial cell types (Parikh et al. 2019) are labeled in c. (e) Cell-free fecal products of GMC3 OW/OB (n = 8) infants reduce transepithelial electrical resistance (TEER) of Caco-2 cells following 48 hours of treatment compared with cell-free fecal products from GMC1 normal BMI (n = 7) infants (One-way ANOVA; P = 0.02). PBS-EDTA represents control exposure. For a-e, n indicates biologically independent infant samples. Each dot represents one independent infant fecal sample in a and e and one transcript in d. DESeq2 was used to calculate significant genes using a two-sided FDR and log₂ FC.

Data availability statement

All raw sequences are deposited in the European Nucleotide Archive (Study PRJEB52295 and PRJEB13896) and in the SRA Bioproject PRJNA648818. All additional datasets and materials are available from the corresponding author upon reasonable request.