An approach for identifying complementary foods that promote healthy gut microbiota development in children with undernutrition. An approach for identifying complementary foods that promote healthy gut microbiota development in children with undernutrition

Culture-independent methods applied to fecal samples, collected monthly during the first two years of postnatal life from members of a Bangladeshi birth cohort living in an urban slum, combined with a machine learning method (Random Forests), have identified a set of bacterial strains whose relative abundances together define a normal program of assembly of the gut microbiota in healthy, biologically unrelated infants and children in this area (Subramanian et al., 2014). Applying this sparse Random Forests-derived model to Bangladeshi children with severe acute malnutrition (SAM), revealed that that they had perturbed microbiota development, with their gut communities appearing younger than those of chronologically age-matched children with healthy growth. Moreover, this immaturity was not repaired with existing therapeutic foods. (Subramanian et al., 2014). These findings raise several questions: To what extent is this identified program of gut community assembly generalizable to other children raised in different geographic locations and representing different cultural traditions? To what extent does the gut microbiota continue to develop beyond the first two years of life, and do the age-discriminatory strains present in the existing Random Forests-derived model of gut microbiota maturation persist beyond the first two years? What food or microbial interventions can be used to prevent or repair this immaturity and improve clinical outcomes? To address these questions, we applied our modeling approaches to (i) bacterial 16S rRNA datasets generated from 50 healthy children, each sampled monthly from birth through the end of postnatal year 2, who live in urban, peri-urban or rural areas of four other countries (India, South Africa, Peru and Brazil) where the burden of childhood undernutrition is great, and (ii) datasets obtained from 36 healthy children living in an urban slum in Bangladesh who were each sampled monthly from birth through 5 years of age. Single country Random Forests-derived models, as well as a model generated from 16S rRNA data aggregated from the four countries, were applied reciprocally to the different populations. The results revealed bacterial strains that define a program of gut microbiota development/maturation that is shared across children living in very distinct geographic and cultural settings. The extended time series analysis conducted in Bangladesh established that this developmental program is largely but not fully completed by the third year of postnatal life. We cultured and sequenced the genomes of nine age-discriminatory bacterial strains as well as seven SAM-associated taxa from Bangladeshi donors, and then fed gnotobiotic mice colonized with these cultured bacterial strains different combinations of commonly consumed Bangladeshi complementary foods. Analysis of 16S rRNA datasets generated from the recipient animals’ fecal microbiota disclosed a number of complementary food ingredient–bacterial strain associations, with levels of chickpea positively correlating with the relative abundances of the greatest number of age-discriminatory strains (and without untoward effects on the representation of any of SAM-associated strains). Comparing gnotobiotic mice colonized with the consortium of SAM-associated strains and treated with either Khichuri-Halwa, a locally produced therapeutic food used for SAM, or the chickpea-enriched microbiota-directed complementary food (MDCF) formulation, with or without subsequent administration of a ‘probiotic consortium’ of these nine age-discriminatory strains, disclosed that only the combination of the MDCF and ‘probiotic’ consortium produced a significant increase in microbial fermentation (notably levels of butyrate) and colonic Foxp3+ CD4+ regulatory T cells. Together, these results illustrate a shared, microbial, feature of human development and a way for identifying and characterizing microbiota-directed interventions designed to support, repair and/or prevent disruption of the normal program of community assembly.