Scientists unveil global patterns in early-life gut microbial evolution

Using data from over 3,000 infant samples worldwide, this groundbreaking study unlocks the secrets of gut microbiome development, setting the stage for improved child health diagnostics globally.

Study: Early life microbial succession in the gut follows common patterns in humans across the globe. Image Credit: Corona Borealis Studio / Shutterstock

In a recent study published in the journal Nature Communications, researchers characterized generalizable early-life gut microbial colonization patterns.

The human gut microbiome is an ecosystem of diverse microbes that interact with one another and form partnerships with the host. An infant’s gastrointestinal tract is colonized for the first time at birth, and microbial succession ensues in the first years of life. Environmental exposure and natural history shape age-dependent aspects of this succession.

Initial colonizers include microbes that metabolize oligosaccharides in human milk and scavenge simple molecules. Later, the introduction of a solid, diverse diet provided an opportunity for a more diverse microbial community. Recurrent colonization and succession patterns across life stages have revealed consistent links between microbiome development and chronology.

Substantial health associations could ensue when the gut microbial community does not match the expected stage for a child’s age. To measure this temporal mismatch, a reference developmental trajectory of the gut microbiome in early life and a method to measure deviation from such a trajectory is necessary. One solution is to develop age estimation models using gut microbial sequences. Such models could also help elucidate normative, healthy gut development.

The study and findings

In the present study, researchers estimated child age using gut microbial relative abundances from metagenomes. First, they examined infant gut microbiome developmental trajectories using a pooled dataset aggregating over 3,000 stool samples sequenced with shotgun metagenomic sequencing from 1,827 healthy infants across 12 studies. The samples included a broad representation of both high- and low-income countries to enhance the model’s robustness. Metagenomes spanned 12 countries across four continents. Samples collected during 2–18 months of age were included.

The researchers trained a five-fold cross-validated random forest model on features derived from the microbial community composition; inputs were α-diversity (Shannon index) and the relative abundances of species present in ≥5% of samples. This analysis included 3,153 samples and 149 species. The model generated predictions reaching a Pearson correlation of 0.8 and a root mean square error (RMSE) of 2.56 months with ground truth values. The model’s robustness was further validated using leave-one-dataset-out cross-validation and an external dataset, yielding consistent RMSE scores.

Next, the team analyzed the importance of variables to identify the top predictive features. The 35 highest-ranking predictors, i.e., 23% of inputs, accounted for 70% of the cumulative variable importance. Of these, 25 correlated positively with age, while 10 correlated negatively. The Shannon index, a measure of microbial diversity, emerged as the third most important predictor, reflecting the increasing complexity of the gut microbiome with age. Further, all of the top predictive taxa except one were present in all major cohorts.

In addition, Anaerostipes hadrus and Faecalibacterium prausnitzii were the taxa with the highest prediction scores across all cohorts. Individually, these species correlated positively with age. Conversely, another important predictor group, including Bifidobacterium breve and Bifidobacterium longum, showed a negative correlation with age. Likewise, specific Lachnospiraceae species also formed another important predictor group, correlating negatively with age.

Next, a leave-one-out cross-validation was performed to assess the model’s generalizability across data sources. As such, new versions of the original model were trained, excluding each data source. This yielded a mean RMSE of 3.03 months. Additionally, validation on an independent dataset comprising 66 infant metagenomes produced an RMSE of 1.55 months, demonstrating the model’s ability to generalize beyond its training data.

The team grouped a subset of samples by location and binned them by age to estimate monthly prevalences of 34 top taxonomic predictors of gut chronology. This revealed remarkably similar succession patterns between all locations, suggesting that many succession patterns were likely universal.

Finally, the team measured the consistency of the direction of abundance transitions of enzyme commission numbers (ECs) between earlier and later specimens from the same individual. As such, a transition score was generated by combining the predominance and direction of this abundance change. Transaldolase was the least-scoring EC, followed by succinate coenzyme A ligase and pyridoxal kinase.

Stratification of ECs decreasing in abundance with age showed a preponderance of functions associated with B. breve, B. longum, Escherichia coli, and Ruminococcus gnavus. By contrast, ribosomal protein S12 methylthiotransferase was the highest-scoring EC, followed by coproporphyrinogen dehydrogenase.

Stratification of ECs increasing in abundance with age showed a more diverse array of fastidious anaerobes, such as F. prausnitzii, Blautia wexlerae, A. hadrus, Blautia obeum, Prevotella copri, and Dorea longicatena. Compared to prior studies, the list of the top 1.5% of ECs also included novel transitions, such as the increase in malate dehydrogenase and the decrease in pyridoxal kinase, underscoring dietary and functional evolution.

Conclusions

The findings indicate that microbial succession in the early-life gut microbiome for key taxa follows common patterns across socioeconomic and geographical settings. These patterns were consistent and robust enough to be learned by a microbiome age model, allowing for generalization beyond cohort boundaries. Functional analyses corroborated trends in microbial genes involved in dietary exposures and feeding transitions.

Despite these strengths, the study noted limitations, including the exclusion of metadata on feeding practices and socioeconomic factors, which could have further enhanced model precision. Overall, the model provided a normative “microbiome age” benchmark to assess early gut maturation.