Bifidobacterium longum and microbiome maturation modify a nutrient intervention for stunting in Zimbabwean infants

Abstract

Background: Small-quantity lipid-based nutrient supplements (SQ-LNS), which has been widely tested to reduce child stunting, has largely modest effects to date, but the mechanisms underlying these modest effects are unclear. Child stunting is a longstanding indicator of chronic undernutrition and it remains a prevalent public health problem. The infant gut microbiome may be a key contributor to stunting; and mother and infant fucosyltransferase (FUT) phenotypes are important determinants of infant microbiome composition.

Methods: We investigated whether mother-infant FUT status (n = 792) and infant gut microbiome composition (n = 354 fecal specimens from 172 infants) modified the impact of an infant and young child feeding (IYCF) intervention, that included SQ-LNS, on stunting at age 18 months in secondary analysis of a randomized trial in rural Zimbabwe.

Findings: We found that the impact of the IYCF intervention on stunting was modified by: (i) mother-infant FUT2+/FUT3- phenotype (difference-in-differences -32.6% [95% CI: -55.3%, -9.9%]); (ii) changes in species composition that reflected microbiome maturation (difference-in-differences -68.1% [95% CI: -99.0%, -28.5%); and (iii) greater relative abundance of B. longum (differences-in-differences 49.1% [95% CI: 26.6%, 73.6%]). The dominant strains of B. longum when the intervention started were most similar to the proficient milk oligosaccharide utilizer subspecies infantis, which decreased with infant age and differed by mother-infant FUT2+/FUT3- phenotypes.

Interpretation: These findings indicate that a persistently "younger" microbiome at initiation of the intervention reduced its benefits on stunting in areas with a high prevalence of growth restriction.

Funding: Bill and Melinda Gates Foundation, UK DFID/Aid, Wellcome Trust, Swiss Agency for Development and Cooperation, US National Institutes of Health, UNICEF, and Nutricia Research Foundation.

Keywords: Infant; Metagenome; Microbiome; Nutrition; Stunting.

Fig. 1

Infant growth trajectories and velocities by IYCF and mother-infant FUT2+/FUT3-phenotype. (A) LAZ by infant age from 1 to 18 mo (n = 792), stratified by IYCF (light grey, no IYCF; dark grey, IYCF) and mother-infant FUT2+/FUT3− phenotype. Lines illustrate average trajectories. Shaded areas are 95% confidence bands. (B) Violin plots of LAZ velocity from 6 to 18 mo of age, stratified by IYCF (light grey, no IYCF; dark grey, IYCF) and mother-infant FUT2+/FUT3− phenotype. Open circles with bars indicate mean LAZ velocity and 95% CIs.

Fig. 2

Infant growth trajectories and velocities by IYCF and infant gut Bifidobacterium longum relative abundance. (A) LAZ by infant age from 1 to 18 mo (n = 53), stratified by IYCF and Bifidobacterium longum relative abundance (light grey, ≤median relative abundance; dark grey, >median relative abundance). Thin lines illustrate individual trajectories. Thick lines show average trajectories. (B) Violin plots of LAZ velocity from 6 to 18 mo of age, stratified by IYCF (light grey, no IYCF; dark grey, IYCF) and mother-infant FUT2+/FUT3− phenotype. Open circles with bars indicate mean velocity and 95% CIs.

Fig. 3

Bifidobacterium longum strain clusters. Ordination plots of PCoA of Jaccard dissimilarities between PanPhlan3.0 pangenome profiles of the dominant Bifidobacterium longum strain in each fecal specimen (N = 284). Three clusters were identifiable. Individual strains are indicated by small circles. Strains in the same cluster are enclosed by a large ellipse and are differentiated by color (blue, B. infantis; dark green, B. longum longum; red, B. suis). Filled small circles indicate reference strains (light blue, B. infantis; pink, B. suis, light green, B. longum longum; grey, unclassified). Open small circles indicate SHINE strains.

Fig. 4

Heatmap of differentially abundant gene families or metabolic pathways between SHINE infant Bifidobacterium longum strains clusters. (A) Heatmap of UniProt gene family presence in SHINE Bifidobacterium longum strains (N = 284). Grey indicates UniProt gene family presence. The horizontal bar at the top indicates strain cluster (blue, B. infantis; dark green, B. longum longum; red, B. suis). Vertical bars from left to right indicate: UniProt gene families that differed between the B. infantis and B. longum longum cluster (red); UniProt gene families that differed between the B. infantis and B. suis cluster (red); Biological Process GO groups; CAZymes; and Transporter class. Only gene families that differed by two-sided Fisher's Exact test and with evidence of overrepresentation in a Biological Process, CAZyme or Transporter class by one-sided Fisher's Exact Test after FDR correction are presented. (B) Heatmap of MetaCyc pathway presence in SHINE Bifidobacterium longum strains (N = 284). Heatmap colors, horizontal and vertical bars are as for panel A. Only pathways that differed by two-sided Fisher's Exact test and with evidence of overrepresentation in a pathway type by one-sided Fisher's Exact Test after FDR correction are presented.