Gut microbiome development and childhood undernutrition

Abstract

Forty-five percent of deaths among children under 5 years-of-age are associated with undernutrition. Globally, almost 200 million children exhibit the two major forms of undernutrition; wasting (low weight-for-height) or stunting (low height-for-age), with many affected by both. Undernutrition is not due to food insecurity alone. Growing evidence indicates that perturbed postnatal gut microbiome development contributes to its pathogenesis. This Perspective focuses on defining and repairing these defects in gut microbiome development. We describe an approach that involves analysis of well-phenotyped human cohorts, followed by preclinical studies using gnotobiotic animals colonized with microbiota from these cohorts. Additionally, these models can be used to identify therapeutic targets and candidates that can then be tested clinically. Furthermore, introducing pre-treatment microbiota from trial participants into gnotobiotic animals and re-enacting trial conditions allows mechanisms to be dissected. We highlight these recent advances as well as gaps in existing knowledge that present opportunities for future research.

eTOC

In their Perspective, Barratt et al. examine experimental and computational approaches for characterizing the perturbed gut microbiome development of undernourished children, and how preclinical models containing their microbiomes can yield insights about disease mechanisms, identify therapeutic targets, and be used to develop microbiome-directed therapeutic agents for clinical testing.

Introduction

Undernutrition in infancy and childhood increases the risk of mortality and is associated with a number of adverse outcomes, including persistent stunting, immune dysfunction, and neurocognitive deficits (Black et al., 2013; Guerrant et al., 2013). Nutritional status is typically defined by anthropometric measures of weight and height (length); the results are expressed in terms of Z-scores that indicate the degree of deviation of an individual’s value from the median value for age-matched members of the World Health Organization (WHO) Multicentre Growth Reference Study (MGRS; WHO 2009). Low weight-for-length (WLZ) is termed wasting, a form of acute malnutrition (undernutrition) that affects 45 million children. Acute malnutrition, in turn, is classified as ‘moderate’ (WLZ between −2 and −3) or ‘severe’ (WLZ less than −3). Stunting describes deficits in length-for-age (LAZ <−2), which affects 150 million children. While anthropometry is objective and relatively straightforward to perform, it is a ‘coarse’ characterization of nutritional status. This coarseness highlights the need for a more comprehensive definition of the underlying biological states of undernourished individuals in order to guide improved disease classification, decipher the mechanisms underlying pathogenesis, identify therapeutic targets, and better assess the effects of existing or new therapeutic modalities.

The first 1000 days of the life encompasses the time from conception to the end of the second postnatal year. This period constitutes a critical window of human growth and development. Various factors have been shown to influence the trajectory of development during this interval including maternal nutritional status and physiology, epigenetic modifications, exposure to environmental pathogens and feeding practices (Martorell, 2017). This Perspective focuses on another factor, the gut microbiome. We posit that childhood undernutrition provides a case study for delineating how perturbations in the development of the gut microbial community impact various aspects of infant and child physiology. A corollary is that “repairing” the microbiome offers an opportunity to link community components and their functions to various host systems that regulate growth. Achieving these insights requires knowledge obtained from both ‘forward’ and ‘reverse’ translational research (Shakhnovich, 2018). In the case of childhood undernutrition, this research includes: (i) defining normal microbial community development within and across different body habitats in a given populations of healthy infants and children, (ii) identifying and quantifying disruptions in community assembly and determining whether the degree of perturbation is significantly correlated with the degree of growth faltering, (iii) testing, initially in representative preclinical models, whether disrupted microbial community development is a cause or an effect of growth faltering, (iv) identifying growth-associated taxa and metabolic/signaling pathways expressed by the microbiome that could serve as targets for therapeutic interventions, (v) using these preclinical models that harbor microbiome components from the target human population to develop therapeutic candidates for repairing microbial community function and (vi) determining the efficacy and safety of therapeutic candidates by returning to the very human population(s) used to characterize the microbiome perturbations. If proof-of-concept and efficacy is established in step (vi), yet another step is to re-enact the clinical study in preclinical models using the pre-treatment microbiomes from participants to identify the molecular mediators that connect microbiome repair to alterations in the expressed properties of host systems. An additional interdisciplinary facet of this multi-step translational medicine ‘journey’ is the importance of early strategic focus on therapeutic approaches that are affordable, culturally acceptable and scalable. Moreover, the extent to which therapeutic targets are unique to a given population, or shared more broadly, needs to be investigated early given the efficiencies of scale that could be obtained with an effective, generalizable, affordable intervention that has broad cultural acceptability.

The maternal microbiome and nutritional status

Acquisition of an infant’s gut microbial community begins at birth. At the time of birth and during the subsequent postpartum period, there is exposure to microbes from a variety of environmental sources, although the principal origin of stably colonizing taxa is the maternal gut (Ferretti et al., 2018). Culture-independent techniques, including strain-level metagenomic profiling of maternal-infant dyads, have provided evidence for the similarity of bacterial strains present in maternal milk and the infant gut microbiota (Asnicar et al., 2017). A maternal reservoir of microbes that can be repeatedly inoculated into the infant gut during the period of breast feeding represents one of the explanations for why the contribution of breastfeeding to the composition of the infant gut microbiota is dose-dependent (Pannaraj et al., 2017; Sugino et al., 2021).

Current approaches for addressing nutritional deficits in pregnant and lactating women living in low-income settings typically focus on dietary diversification, biofortification of staple foods, and supplementation with multiple micronutrients (Dewey et al., 2016). Inadequate maternal education and empowerment, and lack of data on women’s health indicators represent some of the impediments to research in this area (Kinshella et al., 2021). While important, existing interventions have not taken into consideration the role of the pre-term maternal microbiome in determining nutritional status or of its functional changes during or immediately following pregnancy. Data on whether an ‘undernourished’ maternal microbiome contributes to impaired growth phenotypes in her offspring are also currently lacking. However, one recent study of mother-infant dyads conducted in Zimbabwe, in which shotgun sequencing was performed on DNA isolated from fecal samples collected during pregnancy and at 1-month postpartum, provided evidence that bacterial members of the maternal gut community and their metabolic capacities were predictors of birth weight and of ponderal growth at 1-month (Gough et al., 2021). Specifically, the analysis revealed that a greater relative abundance in the maternal microbiome of genes involved in (i) degrading plant polysaccharides represented in the maize-dominated maternal diet and (ii) biosynthesis of thiamine and folate were predictive of higher birth weight and improved neonatal growth. In contrast, genes contributing to biofilm formation in response to nutrient deprivation were associated with reduced birth weight and reductions in subsequent growth (Gough et al., 2021).

Reductions in the cost of shotgun sequencing of microbial community DNA, coupled with improvements in the algorithms used to generate metagenome-assembled genomes (MAGs) have set the stage for characterizing the genomic features of organisms represented in the microbiomes of mothers and their offspring. This capability permits more in-depth analyses of the effects of gestation, maternal parity, age and health status on microbiome composition. For example, a MAG-based analysis has demonstrated that a mother’s bacterial strains are more likely to persist in her infant at one year of age than microorganisms acquired from other sources; it also uncovered persistent differences between term and preterm infants in colonization by strains encoding virulence factors (Lou et al., 2021). MAGs also provide a platform for a major leap forward in the field of microbiome research – namely, advancing from its current focus on interpreting DNA datasets to interpreting microbial RNA-Seq datasets generated from suitably preserved microbiome samples. One hoped for outcome is that knowledge of the expression of MAG-associated metabolic pathways can be used to guide hypothesis-based analyses of microbial contributions to the metabolic features of mothers and their offspring (see below).

Diet and the developing microbiome

In addition to its microbial component, breastmilk contains a complex mixture of macronutrients and bioactive molecules; they include lactoferrin and secretory immunoglobulin IgA which serve to protect the newborn against pathogens (Ballard and Morrow, 2013), plus a plethora of cytokines, growth factors and microRNAs that help shape the immunologic and metabolic development of the infant (Moore and Townsend, 2019, Carrillo-Lozano et al., 2020). Human milk oligosaccharides (HMOs) constitute the third most abundant non-water component of milk (~10g/L). HMOs, which are largely unabsorbed during intestinal transit (Bode, 2012), have evolved to support the establishment of early colonizers of the infant gut such as Bifidobacteria. Among these early colonizers, Bifidobacterium longum subspecies infantis (B. infantis) is uniquely adapted to the gut of the breastfed infant; it possesses five distinct gene clusters (H1-H5) that encode a variety of glycoside hydrolases and oligosaccharide transporters; these loci enable this subspecies to metabolize all of the several hundred known HMOs (Underwood et al., 2015). The therapeutic importance of this ‘champion colonizer’ of the infant gut for those at risk for undernutrition is discussed below.

The interval between cessation of exclusive milk feeding and a fully weaned state is defined as the period of complementary feeding (Dewey and Brown, 2003). Undernutrition often unfolds during this period and is associated with ‘suboptimal’ complementary feeding practices, including a lack of dietary diversity and deficiencies in key micronutrients (Arimond and Ruel, 2004; Krebs, 2007). The World Health Organization recommends exclusive breastfeeding until 6 months of age, and continued provision of breastmilk along with complementary foods for up to 2 years. These foods have been organized into eight groups (including breastmilk), with a recommendation that at least five groups should be consumed on daily basis during the period of complementary feeding (WHO, 2019). In Western countries, where most of the research assessing the impact of complementary feeding on the gut microbiome has been conducted (Chehab et al, 2021), introduction of solid foods and the concomitant reduction in breastmilk intake are associated with an ‘acceleration’ of microbial community development towards an adult-like state (Bäckhed et al., 2015; Stewart et al., 2018).

The mechanisms that determine the trajectory of gut community assembly (succession) during the period of complementary feeding remain obscure, both from a microbial and host perspective. Moreover, current WHO guidelines for complementary feeding are not formulated based on knowledge of how members of the developing gut community sense, acquire and metabolize the molecular components of these foods. One approach for characterizing these mechanisms has been to (i) generate collections of cultured bacterial strains from fecal samples collected from a given donor prior to, during and at the conclusion of complementary feeding, (ii) sequence the genomes of members of the culture collection, (iii) perform in silico reconstructions of the representation, in each cultured strain, of pathways for carbohydrate utilization, amino acid and vitamin biosynthesis and other facets of nutrient/energy metabolism, and (iv) sequentially colonize germ-free animals (e.g., mice or piglets) with consortia representing different stages of community assembly – where the sequence of colonization with the different consortia is deliberately manipulated (Feng et al., 2020). Changes in the abundance, gene expression patterns and metabolic activities of consortium members, as a function of feeding gnotobiotic animals diets representing different phases of complementary feeding, are then monitored. Subsequent application of ‘feature selection’ methods to the resulting multi-omic datasets can provide insights about the niches (‘jobs’) of different community members (examples include machine learning techniques such as Random Forests, or statistical methods from the field of econophysics based on consistent patterns of covariation in the abundance of taxa). However, this type of approach is very labor intensive even when only focused on fecal samples, not to mention extension to samples from colonic and small intestinal habitats (Feng et al., 2020). Moreover, hypotheses based on observed correlations between the fitness of key taxa and their expression of metabolic pathways need to be explicitly tested using genetic tools. These genetic tools are often lacking, although emerging CRISPR-based approaches should be very helpful (Rubin et al., 2022).

One motivation for acquiring knowledge of the metabolic pathways represented within the genomes of members of the developing gut communities of healthy versus undernourished children, and determining how these pathways are expressed, is to relate the micro- and macronutrient content of different dietary components to their microbial metabolism, and/or to their to co-metabolism by microbe and host. As described below, one approach for repairing the perturbed gut microbial communities of children with undernutrition has involved deciphering how different complementary foods are utilized by different growth-promoting members of the gut community, and to use this knowledge to formulate ‘microbiota-directed’ complementary foods (MDCFs).

Repairing perturbed gut microbial development in undernourished children with microbial community-directed complementary foods

Combining serial fecal sampling of members of birth cohorts living in low- and middle- income countries with (i) culture-independent enumeration of bacterial composition and (ii) feature selection methods have identified a sparse group of ‘age-discriminatory’ bacterial taxa whose changing patterns of representation define a program of gut microbial community assembly that is shared between healthy children in different cohorts. This process appears to be largely completed by the end of the second postnatal year (Subramanian et al, 2014; Raman et al., 2019). Children with moderate or severe acute malnutrition have perturbed community development as judged by the representation of these taxa - resulting in community configurations that resemble, in several respects, those of chronologically younger children. This type of perturbation is incompletely repaired by standard nutritional interventions (Subramanian et al., 2014; Gehrig et al., 2019).

A shared, age-dependent signature of ‘metabolic maturation’ has also been identified using urine and plasma samples collected during the first two years of life from members of birth cohorts living in Peru, Bangladesh and Tanzania (Giallourou et al., 2020). Calculation of a ‘phenome-for-age-Z score’ (PAZ) using a subset of metabolites significantly associated with chronological age revealed that, as with microbiota immaturity, metabolic immaturity is linked to growth deficits (Giallourou et al., 2020).

To determine whether perturbed gut microbial community development is a cause rather than an effect of growth faltering, separate groups of just-weaned gnotobiotic mice were colonized with fecal microbiota samples collected from age-matched healthy children and those with acute malnutrition. All mice were given a diet representative of that consumed by the children whose microbiota were used for the colonization experiments. The results revealed that communities from undernourished children transmit impaired weight gain phenotypes, altered bone growth, as well as with immune and metabolic abnormalities (Smith et al., 2013; Kau et al., 2015; Blanton et al., 2016a). Application of machine learning methods to the transplanted communities revealed ‘growth-discriminatory’ bacterial taxa (defined as organisms whose abundances correlated with the growth phenotypes of recipient gnotobiotic animals). Co-administration of a consortium of these growth-discriminatory taxa with the intact microbiota from an undernourished donor produced changes in lipid and amino acid metabolism and augmented lean body mass in recipient gnotobiotic mice (Blanton et al., 2016a). Moreover, a number of these growth-discriminatory taxa are under-represented in the fecal microbiota of Bangladeshi children with moderate and severe acute malnutrition during the period of complementary feeding (Subramanian et al., 2014; Kau et al., 2015).

These findings provided a rationale for identifying approaches that target these growth-promoting bacterial species in undernourished children so that age-appropriate microbiome-host co-development can be restored. However, this envisioned approach for gut community repair raises intriguing questions. How ‘fixed’ is the functional configuration of a community that has co-existed with undernourished children for different periods of time? How fixed is the physiologic state or states of undernutrition/impaired growth? How does the degree of ‘plasticity’ of this state or states relate to the severity of microbial community disruption and the chronologic age of the child? Plasticity can be conceptualized as the capacity of growth-promoting taxa still represented in a perturbed microbial community to re-establish their health-promoting functional niche during and following completion of community repair. Given the complexity and dynamism of microbe-microbe and microbe-host interactions, it seems reasonable to ask how both the rate and degree of repair (‘catch-up development’) of a perturbed community relates to the degree of restoration of healthy growth. Moreover, the organisms and metabolic pathways that are therapeutic targets may be quite different depending upon host chronologic age (and their associated state of physiologic/metabolic/immune maturation). The biogeography of repair is also a key consideration; what regions of the intestine do growth-discriminatory organisms occupy in healthy infants/children versus those with ‘undernutrition’, and does their location reflect different ways that normal versus perturbed communities can affect growth? The latter consideration would impact strategies for formulation of therapeutic agents so that effective dosing along the length of the intestine can be achieved. The type of translational medicine pipeline, which is presented in Figure 1A,B and described in the next section, can be used to begin to address these questions.

Figure 1. A translational medicine pipeline for delineating the contributions of the human gut microbiome to undernutrition, identifying therapeutic targets, and developing therapeutic candidates.

(A) Schematic of the pipeline. (B) Hypotheses and considerations related to the pipeline.

Illustrations of how a translational medicine pipeline can be applied

Gehrig et al. (2019) described the development of culturally acceptable MDCFs for the treatment of moderate acute malnutrition (MAM) in 12 to 18-month-old Bangladeshi children during the weaning period. The approach involved preclinical screens of locally available complementary food ingredients in gnotobiotic mice colonized with a consortium of genome-sequenced, age- and growth-discriminatory bacterial strains cultured from fecal samples that were collected from Bangladeshi children in the very target population where MDCFs would be first tested. Ingredients that promoted the fitness of those strains that were deficient in the microbial communities of children with MAM were advanced into a series of secondary tests in gnotobiotic mice colonized with intact fecal communities from donors with MAM, followed by a tertiary test of efficacy in a gnotobiotic piglet model. Prototype formulations were subsequently tested in a 4-week pilot study in children with MAM where the outcome measures were microbiota repair and changes in levels of plasma protein biomarkers and mediators of different aspects of healthy growth (Gehrig et al., 2019; Raman et al., 2019). In a subsequent 3-month randomized controlled trial in 12 to 18-month-old children with MAM living in the same locale, those participants whose diets were supplemented with the lead formulation identified in the pilot study (called ‘MDCF-2’) exhibited more complete repair of their gut communities and superior weight gain than children who received a common ready-to-use supplementary food (RUSF) – even though the RUSF had higher energy density (Chen et al., 2021). Comparing the effects of the two treatments on levels of nearly 5000 plasma proteins revealed that proteins associated with ponderal growth (e.g., leptin), articular cartilage and bone formation [e.g., collagens, cartilage intermediate layer protein 2 (CILP2), adseverin, secreted frizzled-related protein 4 (SFRP4)] as well as skeletal muscle, vascular and nervous system development [e.g., thrombospondin-3 and thrombospondin-4, CDON (cell-adhesion associated, oncogene regulated)] were all elevated to a greater degree with MDCF-2 supplementation compared to RUSF treatment (Chen et al., 2021) (Figure 2A).

Figure 2. Effects of MDCF-2 intervention on ponderal growth-associated fecal bacterial taxa and the plasms proteins in Bangladeshi children with MAM.

(A) Coefficients of linear mixed effects model (± SEM) showing bacterial amplicon sequence variants (ASV) significantly correlated with ponderal growth (weight-for-length Z-score; WLZ). (B) Differential effects of MDCF-2 and RUSF on WLZ-associated proteins. Proteins are ordered by the log₂(fold-change) of the treatment effect of MDCF-2 over RUSF after three months of supplementation. Gene set enrichment analysis (GSEA) was used to calculate the enrichment of proteins whose abundances were increased more by MDCF-2 compared to RUSF for the set of proteins that were positively correlated with WLZ. The top 30 proteins are shown. Adapted from Chen et al., 2021.

This proof-of-concept study also used bacterial 16S rRNA gene amplicon sequencing to identify taxa whose abundances were significantly associated with improvement in WLZ. Twenty-one taxa were identified as being significantly positively associated with WLZ (Chen et al., 2021; Figure 2B). A number of these taxa, such Prevotella copri, are associated with the weaning phase microbiota of healthy Bangladeshi children but are depleted in age-matched children with acute malnutrition (Subramanian et al., 2014). Moreover, several other WLZ-associated taxa (e.g., Faecalibacteium prausnitzii, Dorea formicigenerans, Ruminococcus gnavus, Clostridium spp.) were ‘growth-discriminatory’ in the gnotobiotic mouse studies described above; they were among the bacterial targets used in the gnotobiotic animal-based screens of complementary food ingredients that led to development of MDCF-2 (Blanton et al 2016a; Gehrig et al., 2019).

These results provide evidence of the translatability of results from the preclinical models to the very population whose microbes were employed to create the model. They also set the stage for using fecal samples collected over the course of the clinical study to search for WLZ-associated MAGs, characterizing their genomic features and using microbial RNA-Seq to identify metabolic pathways in WLZ-associated MAGs that are differentially expressed in MDCF- compared to RUSF-treated children. This information would serve to direct biochemical analysis of the metabolism of MDCF components that are the targets of these pathways. As noted in Figure 1A,B, gnotobiotic mice can also be used for ‘reverse translation’ studies where pre-treatment ‘unrepaired’ and post-treatment ‘repaired’ communities from clinical trial participants can be introduced, enabling more detailed mechanistic analyses of the pathways that mediate the effects of the MDCF on various host cell lineages and organ systems. Another goal is to gain deeper understanding of structure-activity relationships for the MDCF formulations.

The translational medicine approach described in Figure 1 provides a roadmap for further work to establish the generalizability of the effects of MDCFs across different study populations and the impact of this type of intervention at different ages. It can also be used to search for additional food staples that are available and affordable in distinct geographic locales, and that possess similar or superior bioactivity to the tested MDCF. One anticipated return on investments in this type of ‘multi-omic’ analysis of microbial community and host features is development of cost-effective point-of-care diagnostics. In addition, a more refined definition of the biological state of a given individual or cohort of individuals would enable a more targeted approach to treatment, or for prevention in cases where there is known risk for disease in a population.

Next-generation probiotics for undernutrition and the impact of strain-level variation

The efficacy of MDCFs will be limited if the extent of community perturbation is such that microbial therapeutic targets are absent or severely depleted. Under these circumstances, community restoration may require supplementation with a probiotic, or a probiotic followed by an MDCF, to induce and sustain a growth-promoting gut microbiome. The selection of microbial strains for this purpose underscores another challenge encountered when considering community repair. In an illustrative example, 3–6-month-old Bangladeshi infants with severe acute malnutrition (SAM) were found to have pronounced deficiencies in Bifidobacterium longum, subspecies infantis (B. infantis) (Barratt et al., 2022). A randomized controlled clinical trial in these infants involved daily treatment for 4 weeks with a commercially available strain of B. infantis that had been cultured from a child living in the US, or a placebo (lactose). Infants that received this strain of B. infantis exhibited a significantly faster rate of weight gain and an associated reduction in fecal levels of biomarkers of intestinal inflammation compared to infants who had received the placebo (Barratt et al., 2022). However, even after treatment, the level of colonization of B. infantis was ~10–100-fold lower than in age-matched, healthy infants who lived in the same locale and whose breastmilk intake was significantly greater than in the SAM infants. Follow-up studies involved culturing a B. infantis strain from a healthy Bangladeshi child living in the same community where the clinical study was performed. This strain had genomic features that distinguished it from most other reported B. infantis strains – namely, a number of genes involved in utilization of non-HMO carbohydrates present in the breast-milk poor diet of these infants with SAM. A reverse translation experiment was subsequently conducted where a fecal microbiota sample, obtained prior to treatment from a child with SAM enrolled in this trial, was introduced into germ-free female mice during the peripartum period. A subset of the dams was also orally gavaged with a mixture of the Bangladeshi- plus the USA-derived B. infantis strains. Analysis of dam-to-pup transmission of these microbial consortia disclosed that the Bangladeshi B. infantis strain had greater fitness in the SAM microbiota than did the USA strain; this greater fitness occurred in the context of a diet representative of that consumed by 6-month-old Bangladeshi children with SAM that included powdered cow’s milk and plant-based ingredients (Barratt et al., 2022).

These findings highlight the need for considering the impact of strain-level variation in the selection and development of future probiotics (Duar et al., 2020). Strains that are adapted to host habitats with environmental exposures resembling those experienced by the target population of undernourished children may provide higher engraftment efficiencies, and thus more impactful health benefits. This consideration raises important questions for next-generation probiotic development. Are ‘geo-adapted’ strains that have evolved in one population optimal for treating other populations? Should multi-strain consortia be used to optimize the chance of engraftment in the gut microbial communities of recipients representing different populations, thus reducing the need for multiple, distinct, ‘geo-adapted’ individual strain formulations?

Role of the small intestinal microbiota in undernutrition

Much of the work to date in the field of gut microbiome research has focused on analyses of fecal samples due to their relative ease of collection. The small intestinal microbiota has remained an underexplored ‘wilderness’ due to the difficulty in obtaining samples. Environmental enteric dysfunction (EED) is a small intestinal enteropathy of unknown etiology (Prendergast and Humphrey, 2014; Budge et al., 2019). EED was first described in adult Peace Corps volunteers who had returned to the US with diarrhea and intestinal malabsorption, having resided in areas with high fecal-oral contamination. Affected individuals underwent esophagogastroduodenoscopy (EGD) so that proximal small intestinal mucosal biopsies could be obtained. The biopsies revealed histopathologic changes characterized by diminished villus height and number, disruption of the epithelial barrier and a chronic inflammatory infiltrate in the underlying lamina propria (Lindenbaum et al., 1966; Lunn et al., 1991; Campbell et al., 2003). Given the challenges of performing EGD in children with undernutrition, studies of EED have relied on fecal and plasma biomarkers with varying degrees of reproducibility across populations (Mutasa et al., 2021); this has made it difficult to define the contribution of EED to their growth faltering.

Recent studies have focused on children with stunting who have failed commonly employed nutritional interventions; they underwent EGD in order to recover fluid from the lumens of their proximal intestine (duodenums) and duodenal mucosal biopsies (Mahfuz et al., 2017). Comparing plasma and duodenal mucosal proteomes and correlating these data with anthropometry and the results of culture-independent quantification of the absolute abundances of bacterial taxa present in their duodenal aspirates, revealed biomarkers and candidate mediators of enteropathy (Chen et al., 2020). Notably, the absolute abundance of a core group of 14 bacterial taxa, including members of the genera Veillonella, Streptococcus, Gemella, Haemophilis and Neisseria, were negatively correlated with linear growth (LAZ) in the children from whom they were recovered (Chen et al., 2020) (Figure 3). Moreover, the absolute abundances of these taxa were positively correlated with various innate immunity and inflammation-associated proteins in duodenal biopsies, including lipocalin-2 (LCN-2), resistin (RETN) and matrix metalloproteinase (MMP8), along with antimicrobial peptides (cathelicidin antimicrobial peptide [CAMP] and chitinase-3-like protein 1 [CHI3L1]) (Figure 3). In addition, there was a negative correlation between the levels of these bacterial taxa and levels of numerous enterocyte proteins (e.g., cadherin-related family member 5 [CDHR5], which is required for brush border formation (Chen et al., 2020). These findings are similar to those obtained from transcriptional analysis of duodenal biopsies from Zambian children with SAM; the RNA-Seq data disclosed reduced expression of genes associated with the villus brush border and intestinal transport, suggesting a impaired capacity for nutrient absorption (Kelly et al., 2021). Finally, a consortium of organisms cultured from the duodenal aspirates of the Bangladeshi children with EED produced small intestinal enteropathy and elevated biomarkers of intestinal inflammation in recipient gnotobiotic mice consuming a diet representative of that consumed by the children from whom the small intestinal microbes had been obtained (Chen et al., 2020). Of note, many of the genera identified in these Bangladeshi children with EED were also reported in the duodenums of stunted children living in sub-Saharan Africa (Vonaesch et al., 2018).

Figure 3. The top 10 positive correlations between the abundances of environmental enteric dysfunction (EED)-associated small intestinal taxa and levels of duodenal proteins in biopsies from a cohort of stunted Bangladeshi children (BEED study) with histopathological evidence of EED.

A larger size and darker color circle represent a stronger correlation. Adapted from Chen et al., 2020.

Further assessments of the generalizability of these findings across other populations is hampered by the challenges of sampling the small intestinal microbiota, especially in vulnerable populations of undernourished children, and in comparator groups of healthy children, living in the same locale, for whom EGD has no clinical justification. This limitation emphasizes the pressing need for developing less invasive sampling techniques, including those where the proximity of sample collection within the small intestine to the site of disease can be verified, and where sampling can be repeatedly and safely performed over time in children, as well as their mothers (Thompson et al., 2017; Otuya et al., 2018; Tang et al., 2020). Preclinically, the recent development of in vitro models, such as microfluidic ‘Gut-on-a-Chip’ systems that can incorporate (i) epithelial cell populations derived organoids generated from intestinal biopsies of a given patient with confirmed pathologic and defined host genotypic features, and (ii) microbial community members from the same individuals, offer much promise for modeling microbiome-host signaling and identifying new therapeutic targets (Kasendra et al., 2018; Jalilli-Firoozinezhad et al., 2019).

Concluding remarks

Childhood undernutrition is a manifestation of the multi-generational impact of poverty and social inequality. There are many contributing factors, but evidence is emerging for the role microbial communities transmitted from a mother to her offspring, and the perturbed development of this microbial ‘organ’ during the critical first two years of life. While the mechanisms by which members of the microbial community in different regions of the gut contribute to host growth/metabolism remain to be fully elucidated, a microbial view of the pathogenesis of undernutrition could help explain why, despite considerable effort, current nutritional interventions and initiatives to improve water, sanitation and hygiene practices have had limited effectiveness (Dewey, 2016; Goudet et al., 2019). Characterization of the microbial communities of undernourished women and their transmission to offspring requires prolonged, complex, adequately powered longitudinal studies; it also requires linkage of these observational studies to preclinical intergenerational transmission models representative of the human cohorts being characterized so that disease mechanisms and therapeutic targets can be identified. Critically, in populations where the burden of maternal and child undernutrition is great, focusing on those who have avoided undernutrition (i.e., have exhibited ‘positive deviance’) is a first and necessary step in defining normal (Das et al., 2022).

Characterizing the contributions of perturbations in microbial communities distributed along the length of the intestine to growth faltering in infants and children will be challenging. However, it holds the promise of helping to evolve therapeutic strategies for more precise repair of defects in the co-development of these communities and their hosts – strategies that will, hopefully, be predicated on a multi-faceted consideration of host biological state and adequate biogeographical coverage of therapeutic targets by therapeutic agents.

Achieving a balance between more personalized/precision microbial community-targeted interventions versus generalized nutritional approaches that have traditionally been implemented on a broad scale, across a broad range of ages (e.g., 6–24 months) represents a significant challenge, especially with respect to internationally funded programs. Moreover. the long-term consequences of interventions that have the potential for intergenerational impact need to be determined; this includes consideration of myriad ethical, safety, regulatory, and societal issues (Blanton et al., 2016b). Not least, viewing the microbiome as a target for ‘repair’ presents unique challenges to the current approach for classification/regulation of foods and probiotics (Green et al., 2017, De Simone, 2019). Nevertheless, surmounting the challenges and conducting type of studies envisaged provide an opportunity to improve growth and developmental outcomes in ways that traditional interventions have thus far failed to achieve.