Infant Gut Microbes and Diet-Derived Metabolites Linked to Childhood Asthma Risk

Scientists have long suspected that the trillions of bacteria colonizing a newborn’s intestines play a role in shaping whether that child will one day develop asthma, but the molecular messengers connecting the gut to the airways have remained frustratingly elusive. Now, an exploratory study drawing on a large Tennessee birth cohort has taken a step toward closing that gap by pairing infant gut microbiome profiles with detailed chemical inventories of the same children’s urine and stool. The analysis, published in the journal Metabolomics, suggests that many of the metabolites associated with later asthma risk are not produced by the body alone but originate in food and are processed, or co-metabolised, by gut bacteria. The findings do not prove cause and effect, yet they sketch a plausible biochemical bridge between what infants eat, which microbes colonize their guts, and how their immune systems are primed in the first months of life.

The research team, led by Kedir N. Turi of Indiana University with collaborators at Vanderbilt University Medical Center, UT Southwestern, and the University of Wisconsin, mined data from the Infant Susceptibility to Pulmonary Infections and Asthma following Respiratory Syncytial Virus Exposure cohort, known as INSPIRE. Between 2012 and 2014, the study enrolled 1,946 infants in the southeastern United States, of whom 1,166 could be followed to age six. Within that group, 216 children, or 18.5 percent, were diagnosed with asthma by age six using validated questionnaires that captured physician diagnoses, medication use, symptoms, steroid treatment, and acute health care visits. Rather than analyzing the entire cohort, the investigators focused on subsets with layered molecular data: 402 infants had stool microbiome profiles from 16S ribosomal RNA sequencing, 199 had urine metabolomics, and 58 had stool metabolomics. After requiring both microbiome and metabolome data plus the six-year asthma outcome, the overlapping analytical samples shrank to 89 infants for urine and 58 for stool.

Methodologically, the study is a case study in taming high-dimensional biology. The researchers first filtered rare bacterial taxa, retaining 55 amplicon sequence variants after discarding those absent from more than 90 percent of samples, and normalized microbial abundances to correct for sequencing depth. Metabolite measurements, generated with Metabolon’s untargeted ultra-high-performance liquid chromatography-tandem mass spectrometry platforms in positive, negative, and polar ionization modes, were quantile normalized, log transformed, and Pareto scaled. To identify which features actually discriminated between children who later developed asthma and those who did not, the team deployed sparse partial least squares discriminant analysis inside a nested cross-validation framework, with a stratified five-fold outer loop, ten repeats of a five-fold inner loop, and all feature selection confined strictly to training folds to prevent information leakage. Model quality was assessed by the area under the receiver operating characteristic curve, with confidence intervals estimated from 2,000 bootstrap resamples.

The discriminating power of the individual datasets was modest, and the authors are candid about that. The gut microbiome alone yielded a mean AUC of 0.55, essentially barely better than a coin flip, while urine metabolites reached 0.61 and stool metabolites 0.60. But modest prediction was not the point. The goal was to shortlist candidate features for mechanistic hypothesis generation. Among the fifteen most important microbial taxa were Eisenbergiella massiliensis, members of the Clostridium innocuum group, Thomasclavelia ramosa, unclassified Veillonella, Enterobacteriaceae, Staphylococcus, Streptococcus, and Bacteroides species, along with Bilophila wadsworthia and Collinsella aerofaciens. Enrichment analysis showed these taxa overlapped with microbial communities previously implicated in asthma, cystic fibrosis, diabetes, necrotizing colitis, and inflammatory bowel disease, although most disease-pathway terms did not survive correction for multiple testing.

The metabolite side of the ledger told an equally suggestive story. Key urinary discriminators included argininate, kynurenine, 2-methylhippurate, benzoate, and butenoylglycine, hinting at perturbations in retinol metabolism, branched-chain amino acid biosynthesis, and pantothenate and coenzyme A pathways. In stool, the standout compounds were overwhelmingly lipids: cholesterol, fructose, palmitate, myristate, the fatty acid esters of hydroxy fatty acids known as PAHSA and LAHSA, stearoyl sphingomyelin, myristoyl dihydrosphingomyelin, desmosterol, and ceramide. These point toward steroid biosynthesis, fatty acid metabolism, sphingolipid signaling, and caffeine metabolism, pathways tied to epithelial barrier function, membrane organization, and inflammatory regulation. None of the pathway enrichments remained statistically significant after false discovery rate adjustment, a limitation the researchers acknowledge, but the directional signals aligned with known asthma biology.

The heart of the paper lies in the correlations between these shortlisted taxa and metabolites, assessed with Spearman rank correlation and Bonferroni correction. Between stool bacteria and stool metabolites, the relationships were numerous and concentrated in lipid chemistry. Staphylococcus species were negatively correlated with the anti-inflammatory FAHFA lipids PAHSA and LAHSA, yet positively correlated with cholesterol and desmosterol. Enterobacteriaceae tracked with myristoyl dihydrosphingomyelin, Veillonella with stearoyl sphingomyelin, and Thomasclavelia ramosa was inversely related to ceramide. Dorea species correlated negatively with stearate, Bacteroides with 3,7-dimethylurate, and the Clostridium innocuum group with 5,6-dihydrothymine. By contrast, the cross-compartment analysis linking stool bacteria to urine metabolites was strikingly weak, with only two significant negative correlations: Collinsella aerofaciens with 2-methylhippurate and Bilophila wadsworthia with argininate.

Interpreting these patterns, the authors emphasize the ecology of the newborn gut rather than any single causal arrow. Staphylococcus is a signature genus of the early facultative-anaerobe stage of colonization and a dominant member of the breast milk microbiome, while cholesterol and desmosterol are abundant in breast milk and crucial for infant growth and immune cell function. Prior laboratory work has shown that cholesterol stimulates the proliferation of Staphylococcus species, which can scavenge host lipids to fuel growth. The observed positive correlation between the bacterium and these sterols may therefore reflect a milk-fed gut environment in which diet, microbial succession, and host lipid metabolism co-vary, rather than a direct mechanistic relationship. Similarly, the negative association between Staphylococcus and FAHFAs, lipid mediators with anti-inflammatory and immune-modulating properties that also occur in human milk, invites speculation that early colonization patterns influence the availability of these protective compounds.

The sphingolipid and purine findings reinforce the dietary thread. Sphingomyelins are among the most abundant lipids in human breast milk and infant formula, and Enterobacteriaceae and Veillonella are classic pioneer colonizers that thrive during the milk-dominated stage of infancy, so their positive correlations with sphingomyelin species may simply mark a shared developmental timeline. The inverse relationship between Bacteroides and 3,7-dimethylurate, a byproduct of xanthine and caffeine-like compounds, is particularly intriguing: xanthines reach infants through maternal caffeine passing into breast milk or through formula ingredients, while Bacteroides species, which flourish on human milk oligosaccharides and are acquired in greater abundance during vaginal birth, may accelerate the clearance of these compounds. To check whether these interpretations were plausible, the team ran the eleven significantly correlated metabolites through the MetOrigin database, which assigns putative sources. Seven were flagged as likely co-metabolised by host and microbiome, two as solely microbial, and source enrichment pointed overwhelmingly to food-derived compounds processed through microbial metabolism.

The authors are careful to frame the work as exploratory. The metabolomics subsample was deliberately enriched for early wheeze and allergic sensitization, so the discrimination metrics do not represent predictive performance in the general population. The 16S rRNA sequencing resolves bacteria only to genus level in many cases, stool sampling captures luminal rather than mucosa-associated communities, the stool metabolome subset of 58 infants is small enough that even nested cross-validation may yield unstable estimates, and no direct dietary intake data were collected, making all diet-related discussion speculative. Correlations, by definition, establish neither direction nor causation. Still, by following a longitudinally characterized birth cohort from the first months of life to a hard clinical outcome at age six, and by integrating microbiome and metabolome data from both gut and urine, the study offers a concrete, testable hypothesis: the infant gut environment shaped by feeding mode and microbial colonization produces a metabolite milieu, rich in diet-derived lipids and microbial co-metabolites, that helps calibrate the developing immune system and, in doing so, may tilt the odds toward or away from childhood asthma. Confirmation will require replication and experiments, but the biochemical suspects are now named.

Subject of Research: Integrative analysis of the infant gut microbiome and metabolome in relation to childhood asthma risk

Article Title: Exploratory integrative analysis of infant gut microbiome and metabolome suggests diet- and microbiome-derived metabolites associated with childhood asthma

Article References: Turi, K. N., Wu, X., Li, Y., Rosas-Salazar, C., Gebretsadik, T., Shilts, M. H., Das, S. R., Gern, J., Xu, Y., & Hartert, T. V. (2026). Exploratory integrative analysis of infant gut microbiome and metabolome suggests diet- and microbiome-derived metabolites associated with childhood asthma. Metabolomics, 22(5), Article 162. https://doi.org/10.1007/s11306-026-02540-3

Image Credits: AI Generated

DOI: 10.1007/s11306-026-02540-3

Keywords: asthma, gut microbiome, metabolomics, infants, birth cohort, breast milk, FAHFA lipids, sphingolipids, Staphylococcus, Bacteroides, gut-lung axis, 16S rRNA sequencing