Effects of microbiota-directed foods in gnotobiotic animals and undernourished children – Science Magazine

Metabolic phenotypes

Targeted MS of plasma samples obtained at enrollment revealed high levels of ketones, nonesterified fatty acids (NEFA), and medium to long even-chain acylcarnitines (Fig. 2, A and B, and table S2), which is consistent with the known acute malnutrition-induced lipolytic response that raises circulating fatty acids and activates fatty acid oxidation (11). By discharge, this metabolic feature had normalized, whereas levels of a number of amino acids had increased significantly, including the gluconeogenic amino acid alanine; the branched-chain amino acids leucine, isoleucine, and valine; plus products of branched-chain amino acid metabolism [C3 (propionyl)-carnitine and their ketoacids] (Fig. 2, A to C). These findings suggest that the increased protein provided by the therapeutic foods prompted a switch from fatty acid to amino acid oxidation, leading to repletion of fat depots, increases in plasma leptin (Fig. 2A), and weight gain (table S1B). However, 6 months after treatment, multiple plasma amino acids and their metabolites had declined to levels comparable with those at admission, whereas fatty acids and fatty acid–derived metabolites remained at similar concentrations to those observed at discharge (Fig. 2, A to C). Insulin-like growth factor 1 (IGF-1) levels did not change significantly during this period (Fig. 2A), potentially explaining the absence of a signature of pronounced lipolysis that had been observed at enrollment. Although the suppression of lipolysis at 6 months after discharge suggests a sustained effect of nutritional resuscitation, the fall in essential amino acids and the lower level of IGF-1 compared with that found in similarly aged healthy children from the same community (44.5 versus 69.4 ng/mL; P = 0.02, t test) may contribute to the observed failure to achieve catch-up growth.

The plasma proteome

Significant correlations between levels of plasma proteins, anthropometric indices, plasma metabolites, and host signaling pathways regulating key facets of growth are described in the supplementary text, results (table S3, A and B)—for example, components of the growth hormone (GH)–IGF axis, including soluble growth hormone receptor (also known as growth hormone binding protein), multiple IGF binding proteins (IGFBPs), and regulators of IGFBP turnover (the metalloprotease pappalysin-1 and its inhibitor stanniocalcin-1).

The gut microbiome

A sparse RF-derived model of normal gut microbiota development comprising 30 bacterial taxa [operational taxonomic units (OTUs)] and obtained from 25 healthy members of a birth cohort living in Mirpur (table S4, A to C) was applied to bacterial V4-16S rDNA datasets generated from fecal samples serially collected from the children in the SAM study (n = 539 samples). This model allowed us to define microbiota-for-age z-scores (MAZ) as a function of treatment arm and time [9.3 ± 3.7 samples/child (mean ± SD)]. The MAZ score measures the deviation in development of a child’s microbiota from that of chronologically age-matched reference healthy children on the basis of the representation of the ensemble of age-discriminatory strains contained in the RF-derived model (2). Significant microbiota immaturity was apparent in the SAM and post-SAM MAM groups (Fig. 1C and table S5A). Moreover, MAZ-scores in this SAM cohort were significantly correlated with WHZ, HAZ, and WAZ [Pearson correlation coefficient (r) = 0.16, P = 0.0004; r = 0.13, P = 0.003; and r = 0.10, P = 0.02, respectively]. The MAZ score was not different at discharge but improved 1 month later (P = 0.0051 versus admission, Mann-Whitney test). This improvement could reflect increased dietary diversity, reduced antibiotic usage (table S1B), and/or other factors associated with returning to the home environment. MAZ-scores did not change significantly thereafter (Fig. 1B).

A number of the age-discriminatory strains were significantly correlated with anthropometric indices as well as with plasma proteins involved in biological processes that mediate growth. We also identified significant negative correlations between these taxa and mediators of systemic inflammation and anorexia/cachexia. Bifidobacterium longum (OTU 559527) had the greatest number of significant correlations (114) [table S3C; further discussion is available in supplementary text, results].

The effects of the therapeutic food interventions on the representation of metabolic pathways in the gut microbiome were defined by shotgun sequencing of 331 fecal DNA samples obtained from 30 members of the Mirpur birth cohort with consistently healthy anthropometry and 15 of the 54 children enrolled in the SAM study (table S5B); these 15 children were selected according to their age (12 to 18 months) and that we had corresponding plasma metabolomic and proteomic datasets for at least two of the three time points sampled. The abundances of microbial genes that mapped to pathways in the microbial communities SEED (mcSEED) database (12)—related to metabolism of amino acids, carbohydrates, fermentation products, and B vitamins and related cofactors—were first defined in healthy children sampled monthly from birth to 2 years of age. A set of age-discriminatory metabolic pathways (mcSEED “subsystems” or pathway modules) was identified. The resulting sparse RF-derived model (fig. S1, A and B, and materials and methods) allowed us to assign a state of development (functional age or “maturity”) to the fecal microbiomes of the 15 children treated for SAM. Relative functional maturity was significantly correlated with MAZ, WHZ, and WAZ scores during the course of the trial (Pearson r and P values are MAZ, r = 0.55, P < 0.0001; WHZ, r = 0.30, P = 0.0011; WAZ, r = 0.23, P = 0.013). At enrollment, and just before administration of therapeutic foods, children with SAM had more immature microbiomes [one-way analysis of variance (ANOVA) P = 0.0002; Dunnett’s multiple comparisons test for healthy versus SAM adjusted P values at the two time points, 0.027 and 0.0001, respectively]. There was a statistically significant improvement in functional maturity from initiation of therapeutic food treatment to discharge, and at 1 and 6 months after discharge (Tukey’s multiple comparisons test; adjusted P values = 0.039, 0.0028, and 0.025, respectively). However, this improvement was not sustained at later time points (fig. S1D). Comparing the relative abundances of the 30 most age-discriminatory pathways at six time points revealed that the SAM microbiome had significantly reduced representation of (i) amino acid metabolic pathways, including those involved in isoleucine, leucine, valine biosynthesis, and uptake; (ii) several carbohydrate utilization pathways (arabinose and arabinosides, rhamnose and rhamnogalacturonan, and sialic acid); and (iii) multiple pathways involved in B-vitamin metabolism, including “niacin/NADP (nicotinamide adenine dinucleotide phosphate) biosynthesis” (fig. S1E and table S5C). The observed underrepresentation of age-discriminatory OTUs and metabolic pathways in the gut communities of children with post-SAM MAM provided the rationale for developing a pipeline to test complementary food ingredients for their ability to repair this immaturity.

Microbial community responses

Community profiling by means of short read shotgun sequencing (COPRO-seq) of cecal DNA revealed that compared with MS/KH, consumption of the MDCF prototype resulted in significantly higher relative abundances of a number of weaning-phase age-discriminatory taxa, including Faecalibacterium prausnitzii, Dorea longicatena, and B. luti (P < 0.01; Mann-Whitney test) (Fig. 3A and table S9B). This prototype did not promote the fitness of the SAM donor-derived strains, with the exception of Escherichia fergusonii.

We used targeted MS to quantify cecal levels of carbohydrates, short-chain fatty acids, plus amino acids and their catabolites (table S10, A to D). Germ-free animals served as reference controls to define levels of cecal nutrients that, by inference, would be available for bacterial utilization in the different diet contexts. There were several noteworthy findings: (i) Levels of butyrate and succinate were significantly higher in colonized animals consuming MDCF compared with MS/KH (Fig. 3B and table S10B). (ii) There were no statistically significant diet-associated differences in levels of any of the amino acids measured in germ-free animals, but when compared with their colonized MS/KH–fed counterparts, colonized MDCF-consuming animals had significantly elevated cecal levels of six amino acids classified as essential in humans (the three branched-chain amino acids leucine, isoleucine, and valine plus phenylalanine and tryptophan) (Fig. 3C and table S10C). And (iii) two tryptophan-derived microbial metabolites that play important roles in suppressing inflammation and are neuroprotective, 3-hydroxyanthranillic acid (3-HAA) and indole-3-lactic acid (16–21), were significantly elevated in colonized animals fed MDCF compared with their MS/KH–treated counterparts (table S10D).

Findings from RNA-sequencing (RNA-seq) analysis of the transcriptional responses of community members to the two diets based on Kyoto Encyclopedia of Genes and Genomes (KEGG)– and mcSEED-derived annotations of the 40,735 predicted protein-coding genes present in consortium members are described in tables S9C and S11, A to C, and supplementary text, results, and in silico predictions of their ability to produce, use, and/or share nutrients are provided in table S6, D and E. For example, community-level analysis revealed specific members manifested MDCF-associated increases in expression of genes involved in biosynthesis of the essential amino acids, including branched-chain amino acids (Ruminococcus obeum and Ruminococcus torques), and generation of aromatic amino acid metabolites (R. obeum, R. torques, and F. prausnitzii) (table S11C, ii).

Host effects

Serum levels of IGF-1 were significantly higher in colonized mice that consumed the initial MDCF prototype compared with those that consumed MS/KH. This effect was diet- and colonization-dependent, with germ-free animals exhibiting significantly lower levels of IGF-1 in both diet contexts (Fig. 3D). Serum insulin levels were also higher in colonized animals that consumed MDCF compared with MS/KH [800.7 ± 302.9 ng/mL (mean ± SD) versus 518.7 ± 135.1 ng/mL, respectively; P = 0.06; unpaired t test].

IGF-1 binding to its receptor tyrosine kinase, IGF-1R, affects a variety of signal transduction pathways, including one involving the serine/threonine kinase Akt/PKB, phosphatidylinositol-3 kinase (PI3K), and the mammalian target of rapamycin (mTOR). Absorption of several amino acids from the gut—notably, branched-chain amino acids and tryptophan—leads to activation of mTOR (22). Colonized animals fed MDCF had significantly higher levels of hepatic phosphoSer473-Akt, which is consistent with activation of Akt by IGF-1 signaling through the PI-3K pathway (Fig. 3E). Levels of phospho–AMPK (5′ adenosine monophosphate-activated protein kinase) were not significantly affected by diet (Fig. 3E), suggesting that Akt phosphorylation is not caused indirectly by altered hepatic energy status. Phosphorylation of hepatic Jak 2 (Tyr1007/1008) and mTOR (Ser2448), which are involved in IGF-1 production, was significantly increased in colonized mice consuming MDCF (Fig. 3E), whereas phosphorylation of STAT5, also implicated in IGF-1 production, was not significantly altered.

Previous studies of adult germ-free mice reported increases in serum IGF-1 after their colonization with gut microbiota from conventionally raised mice; increased IGF-1 levels were also associated with increased bone formation (23, 24). Micro-computed tomography (μCT) of mouse femurs revealed a significant increase in femoral cortical bone area in MDCF-fed animals; the effect was both diet- and microbiota-dependent (Fig. 3F).

We used targeted MS to quantify levels of amino acids, acyl–coenzyme As (acylCoAs), acylcarnitines, and organic acids in serum, liver, and gastrocnemius muscle (table S12). Products of nonoxidative metabolism of glucose and pyruvate (lactate from glycolysis and alanine from transamination of pyruvate, respectively) were significantly lower in mice fed MDCF compared with mice fed MS/KH; this was true for alanine in serum, skeletal muscle, and liver and for lactate in liver (table S12, A to C and H). Oxidative metabolism of glucose is associated with nutritionally replete, anabolic conditions. These findings are consistent with the observed elevations of the anabolic hormone IGF-1 in MDCF-fed compared with MS/KH–fed mice. MDCF-fed mice had significantly higher circulating levels of branched-chain amino acids than those of their MS/KH–fed counterparts (Fig. 3G and table S12, A to C). Skeletal muscle C5 carnitine and the closely related metabolite C5-OH/C3 carnitine were significantly higher in animals consuming MDCF (Fig. 3G and table S12F). In liver, C3 and C5 acylcarnitines were significantly lower in MDCF-treated mice (Fig. 3G and table S12E), suggesting that the more nutritionally replete state associated with MDCF may act to limit branched-chain amino acid oxidation in this tissue.

Gut mucosal barrier function

Epithelium and overlying mucus from the proximal, middle, and distal thirds of the small intestine were recovered with laser capture microdissection (LCM) (Fig. 4D). Listed in table S15C are the 30 most abundant OTUs identified by means of V4-16S rDNA analysis of LCM mucosal DNA obtained from the different small intestinal segments within a given diet group and between similarly positioned segments across the different diet treatments. For example, Mirpur(PCSB) produced a statistically significant increase in the relative abundance of F. prausnitzii in the proximal two-thirds of the small intestine, without significantly affecting the proportional representation of a milk-associated age-discriminatory Bifidobacteria OTU (Fig. 4, C and D).

Gene expression was characterized in the jejunal mucosa (Fig. 4D, SI-2 segment) recovered by LCM from mice belonging to all six treatment groups. Significant differences in expression were categorized based on enriched Gene Ontology (GO) terms for “Molecular Function.” In colonized mice, Mirpur(P) and Mirpur(PCSB) significantly up-regulated genes assigned to “cadherin binding” (GO: 0045296) and “cell adhesion molecule binding” (GO: 0050839) compared with Mirpur-18 (table S17A). The diet effect was colonization-dependent; there were no significant differences in expression of these genes or these GO categories in germ-free mice consuming supplemented versus unsupplemented diets (table S17). (Further discussion is available in the supplementary text, results, and histochemical and immunohistochemical analyses of tissue sections prepared along the length of the small intestines of these mice are provided in fig. S6). On the basis of its effects on microbial community organismal composition, gene expression, and gut barrier function, we deemed Mipur-18 supplemented with the four lead complementary foods [Mirpur(PCSB)] superior to that supplemented with just peanut flour [Mirpur(P)].

Effects on biological state

To contextualize the biological effects of the dietary interventions, we performed quantitative proteomics (SOMAscan) on plasma collected from 21 12- to 24-month-old Mirpur children with healthy growth phenotypes (mean age, 19.2 ± 5.1 months; WHZ, 0.08 ± 0.58; HAZ, –0.41 ± 0.56, WAZ, –0.12 ± 0.60) and 30 children with SAM before treatment (Fig. 1A, B1 sample; WHZ < –3; mean age 15.2 ± 5.1 months) [metadata associated with the healthy, SAM, and MAM (MDCF trial) cohorts are provided in table S22]. We rank-ordered all detected proteins according to fold differences in their abundances in plasma collected from healthy children compared with children with untreated SAM. The top 50 most differentially abundant proteins (P < 10⁻⁷; R package “limma”) that were significantly higher in healthy children were designated “healthy growth-discriminatory,” whereas the top 50 differentially abundant proteins that were significantly higher in children with SAM were designated “SAM-discriminatory” (table S23A). We next compared the mean difference for each protein in the pre- versus post-intervention plasma samples for all children in each MDCF/RUSF treatment group. Proteins were then ranked according to the fold differences of the pre- versus post-treatment levels in each of the four groups (table S23B), and these treatment effects were mapped onto the 50 most healthy growth-discriminatory and 50 most SAM-discriminatory proteins. Strikingly, MDCF-2 elicited a biological response characterized by a marked shift in the plasma proteome toward that of healthy children and away from that of children with SAM; MDCF-2 increased the abundance of proteins that are higher in plasma from healthy children and reduced the levels of proteins elevated in SAM plasma samples (Fig. 6B).

Aggregating proteomic datasets from the combined cohort of 113 children with SAM, MAM, and healthy growth phenotypes for whom plasma samples were available, we identified a total of 27 plasma proteins that were significantly positively correlated with HAZ and 57 plasma proteins that were significantly negatively correlated with HAZ [absolute value of Pearson correlation > 0.25, false discovery rate (FDR)–corrected P value < 0.05]. Among the treatments, MDCF-2 was distinctive in its ability to increase the abundances of a broad range of proteins positively correlated with HAZ, including the major IGF-1 binding protein IGFBP-3, growth hormone receptor (GHR), and leptin (LEP) (Fig. 6C). Growth differentiation factor 15 (GDF15) was reduced after 4 weeks of dietary supplementation with MDCF-2 (Fig. 6C). This transforming growth factor–β superfamily member, which was negatively correlated with HAZ, has been implicated in the anorexia and muscle wasting associated with cancer and with chronic heart failure in children; it was elevated in children with SAM and positively correlated with their lipolytic biomarkers NEFA and ketones (supplementary text, results). Peptide YY, an enteroendocrine cell product elevated in SAM plasma that reduces appetite and negatively correlated with HAZ, was also decreased by MDCF-2.

We identified GO terms that were enriched among the group of treatment-responsive proteins and ranked them according to the P value of their enrichment (table S23C). Proteins belonging to GO terms significantly higher in healthy compared with SAM plasma samples were deemed “healthy growth-discriminatory,” whereas those that were significantly higher in SAM were deemed “SAM-discriminatory” (fold-difference >30%; FDR-adjusted P value <0.05). This analysis revealed multiple healthy growth-discriminatory proteins associated with GO processes “osteoblast differentiation” and “ossification” that were increased by supplementation with MDCF-2 (Fig. 7A and table S23C). Examples include key markers or mediators of osteoblast differentiation [osteopontin (SPP1), bone sialoprotein 2 (IBSP), and bone morphogenetic protein 7 (BMP7)] as well as matrix metalloproteases (MMP-2 and MMP-13) involved in terminal differentiation of osteoblasts into osteocytes and bone mineralization.

A number of plasma proteins categorized under the GO process “CNS development,” including those involved in axon guidance and neuronal differentiation, were also affected by MDCF-2 supplementation. Levels of the SAM-discriminatory semaphorin SEMA3A, a potent inhibitor of axonal growth, decreased with this treatment, whereas healthy growth-discriminatory semaphorins (SEMA5A, SEMA6A, and SEMA6B) increased (Fig. 7B). Other healthy growth-discriminatory proteins whose abundances increased with MDCF-2 included receptors for neurotrophin (NTRK2 and NTRK3) plus various axonal guidance proteins [netrin (UNC5D), ephrin A5 (EFNA5), roundabout homolog 2 (ROBO2), and SLIT and NTRK-like protein 5 (SLITRK5)] (Fig. 7B). Expression of a number of neurotrophic proteins belonging to these families has been reported to be influenced by nutrient availability in primates (29).

Compared with healthy children, the plasma proteome of children with SAM was characterized by elevated levels of acute phase proteins [such as C-reactive protein (CRP) or interleukin-6 (IL-6)] and inflammatory mediators, including several agonists and components of the nuclear factor–κB (NF-κB) signaling pathway (Fig. 7C). These components include the pro-inflammatory cytokines IL-1β, tumor necrosis factor–α (TNF-α), and CD40L, plus ubiquitin-conjugating enzyme E2 N (UBE2N), which is involved in induction of NF-κB– and mitogen-activated protein kinase (MAPK)–responsive inflammatory genes (30). MDCF-2 supplementation was associated with reductions in the levels of all of these SAM-associated proteins (Fig. 7C).

Effects on the microbiota

Our analysis of fecal microbiota samples revealed no significant change in the representation of enteropathogens within and across the four treatment groups (fig. S7A and table S21F). MDCF-2–induced changes in biological state were accompanied by increases in the relative abundances of several weaning-phase taxa, including OTUs assigned to F. prausnitzii (OTU 851865) and a Clostridiales sp. (OTU 338992) that are closely related to taxa ranked first and second in feature importance in the sparse Bangladeshi RF-derived model of gut microbiota maturation (fig. S7, B and C). MDCF-2 supplementation was associated with a significant decrease in B. longum (OTU 559527) (fig. S7B), which is ranked third in feature importance in the RF-derived model and discriminatory for a young, milk-oriented microbiota. None of the other members of the 30 OTU model showed significant changes. By contrast, MDCF-1 did not produce significant increases in any of the taxa in the model. The other two formulations were each associated with a significant change in just one member [an increase in the relative abundance of an early age-discriminatory OTU (Streptococcus; ranked 30th) with MDCF-3 supplementation, and a decrease in another OTU (Enterococcus faecalis; ranked 29th) with RUSF supplementation] (table S4B).

MAZ scores were not significantly different between groups at enrollment, nor were they significantly improved by any of the formulations. Interpretation of this finding was confounded by unexpectedly high baseline microbiota maturity scores in this group of children with MAM [MAZ, –0.01 ± 1.12 (mean ± SD)] (table S22) compared with a small, previously characterized Mirpur cohort with untreated MAM (2). Hence, we developed an additional measure of microbiota repair (31). This involved a statistical analysis of covariance among bacterial taxa in the fecal microbiota of anthropometrically healthy members of a Mirpur birth cohort who had been sampled monthly over a 5-year period. Using approaches developed in the fields of econophysics and protein evolution to characterize the underlying organization of interacting systems with seemingly intractable complexity, such as financial markets, we found that the gut community in healthy children could be decomposed into a sparse unit of 15 covarying bacterial taxa termed an “ecogroup” (31). These ecogroup taxa include a number of age-discriminatory strains in the Bangladeshi RF-derived model (such as B. longum, F. prausnitzii, and Prevotella copri). We used the ecogroup to show that in addition to its effects on host biological state, MDCF-2 was also the most effective of the four treatments in reconfiguring the gut bacterial community to a mature state similar to that characteristic of healthy Bangladeshi children.