How Mothers Seed the Infant Gut Microbiome
A Dutch birth cohort maps maternal-to-infant bacterial transmission through the first year and finds an eczema prediction signal, but the observational study cannot prove prevention or justify changing delivery or feeding plans.
The 60-second version
In a Dutch cohort, the maternal gut was the main measured source of infant gut strains, while the infant microbiome changed rapidly across the first year.
Key points
- The study sequenced 4,526 stool samples from 714 mother-infant pairs and integrated 474 clinical and exposure variables.
- Median strain sharing was 65% early and 25% by twelve months as environmental strains accumulated.
- Delivery and feeding mode correlated with infant microbiome patterns, but neither was randomized.
- Maternal microbiome features moderately predicted eczema inside the cohort; external validation and causal evidence are absent.
- Funding and conflicts included Nestlé support, two Nestlé-affiliated authors and a disclosed speaker fee.
Verdict. A valuable longitudinal atlas and hypothesis generator, not a basis for changing delivery, feeding or probiotic care.
Bottom lineA detailed map of the first year
Researchers followed 714 mother-infant pairs from pregnancy through one year after birth and sequenced 4,526 stool samples. The maternal gut was the main measured reservoir for infant gut strains, while infant communities changed quickly as new environmental strains appeared.
Delivery mode and feeding mode were the strongest biological correlates of infant microbiome composition. The maternal microbiome also predicted infant eczema within this cohort, but that signal has not yet been externally replicated or shown to be causal.
Study designWhat was measured
| Maternal stool | 1,587 samples collected during pregnancy, at birth and three months postpartum. |
|---|---|
| Infant stool | 2,939 samples collected at seven timepoints from two weeks to twelve months. |
| Other maternal sites | Ultra-deep sequencing of 82 vaginal samples and 90 breast-milk samples. |
| Context | Clinical records and questionnaires covering 474 health, diet and exposure variables. |
The cohort design is a major strength because repeated samples reveal trajectories that a single swab cannot. It is still observational: exposures were not randomly assigned, and associations can reflect health, medical care, behavior and environment together.
TransmissionThe maternal gut was the main reservoir
Matching bacterial strains, not just species names, gave the team a more specific view of sharing. Across 81 species with enough comparisons, a median 33% of species were shared per mother-infant pair, compared with a median of zero in unrelated pairs.
Breast milk and vaginal samples sometimes contained matching strains, but these routes were less frequent in the measured subsets. The study did not measure every possible source, so 'main reservoir' means the largest among the maternal sites tested, not the sole origin of infant microbes.
The earliest infant microbiome was a starting point, not a fixed endpoint: two-week clusters did not reliably predict communities later in the first year.
AssociationsDelivery, feeding and eczema
After time and technical variation, delivery and feeding mode were prominent correlates of infant microbiome composition and function. These variables are not isolated treatments: medical indications, maternal health and household conditions can influence both the decision and the microbiome.
| Eczema model using species | AUC 0.74 in leave-one-out internal validation. |
|---|---|
| Eczema model using pathways | AUC 0.73 in leave-one-out internal validation. |
| Eczema model using diversity | AUC 0.68 in leave-one-out internal validation. |
| Clinical status | No external validation, intervention or approved test. |
The authors describe the eczema association as a first report and call for replication in diverse populations and work on causal pathways. Moderate AUC values do not show that changing the microbiome would prevent eczema.
LimitsWhat should not be inferred
- No causal treatment claim: the study observed natural variation; it did not assign probiotics, diets, delivery routes or feeding methods.
- One population: the relatively healthy Dutch cohort may not represent higher-risk pregnancies or other healthcare settings.
- Different denominators: vaginal and milk analyses used subsets of 82 and 90, much smaller than 714 pairs.
- Short horizon: follow-up ended at one year, so later health effects were not established.
- No personal prescription: medically indicated caesarean delivery and individual feeding decisions should not be changed on the basis of this paper.
Next stepWhat the finding is useful for
The study supplies a strong reference map for testing how maternal strains establish themselves and persist. It also identifies an eczema prediction signal that can now be challenged in independent cohorts.
For families and clinicians, the action today is not to change care. Delivery and feeding decisions should remain individualized; microbiome-based prevention requires replicated prediction and randomized intervention evidence.