The team enrolled 337 healthy infants shortly after birth and followed them through their two-dose Rotarix series, administered at 2 and 4 months of age according to routine immunization schedules at each site. Of these, 328 infants, or 97.3 percent, completed visits through 6 months. To measure vaccine response, the researchers used two complementary indicators. The first was seroconversion, defined as at least a threefold increase in serum rotavirus-specific IgA between the pre-vaccination visit at a median of 54 days of life and the post-vaccination visit at a median of 188 days. The second was vaccine-strain shedding, detected by polymerase chain reaction in stool collected within 14 days after the second dose. Overall, 153 of 328 infants, or 46.6 percent, seroconverted. Shedding was less common: of 216 infants with stool analyzed after the second dose, only 79, or 36.6 percent, had detectable vaccine virus. Neither measure differed significantly between countries, a striking result given that absolute IgA levels did vary by site.

To understand what distinguished responders from non-responders, the team sequenced shotgun metagenomes from 816 stool samples collected at 1, 2, and 4 months of life, ultimately analyzing 765 high-quality samples from 324 infants. As expected, microbial richness increased with age, and community composition differed by delivery mode, diet, and country. But the critical signal emerged in the first month. Marginal PERMANOVA testing revealed that seroconversion status correlated with gut community composition at month 1, explaining roughly 1.1 percent of variation in Aitchison distance, while no relationship existed at months 2 or 4. Functional diversity based on MetaCyc pathway profiles showed the same temporal restriction. When the researchers applied distance-based redundancy analysis conditioned on country, delivery mode, diet, and age, they recovered a gradient inverse to seroconversion, and Klebsiella was the top contributor, explaining 8.6 percent of total variation along the first axis.

Differential abundance analysis sharpened the picture. At month 1, the species Klebsiella michiganensis and Phytobacter ursingii were overrepresented in the guts of vaccine non-responders, and these findings were reproducible across multiple statistical methodologies. No species at months 2 or 4 were associated with seroconversion status. The functional profile of non-responders was equally distinctive: 61 MetaCyc pathways were enriched in non-responders, dominated by aromatic compound degradation, particularly of toluene and catechol, alongside ureide biosynthesis, trimethylamine degradation, and catabolism of arginine, ornithine, asparagine, glutamate, and threonine. Crucially, pathways involved in lipopolysaccharide, or LPS, biosynthesis were overrepresented in non-responders, with no LPS pathways enriched in responders. Gene-level analysis of KEGG orthologs stratified by species confirmed that Klebsiella species carrying LPS biosynthetic genes were enriched in non-responders, directly linking the taxonomic and functional signatures.

To capture the overall ecological configuration of the infant gut, the researchers applied Dirichlet Multinomial Mixture modeling to cluster communities into discrete gut community types at each time point. At month 1, one type, designated cs-GCT3, was enriched in Klebsiella, while another, cs-GCT6, was depleted in both Klebsiella and Escherichia. After accounting for country, diet, and breastfeeding, seroconversion status was significantly associated with month 1 gut type. Seroconverters had lower odds of belonging to the Klebsiella-dominant cs-GCT3 and the highest odds of belonging to cs-GCT6, which contained the greatest absolute percentage of seroconverters at 74.0 percent. Proportionally fewer cs-GCT3 infants, only 16 percent, were seropositive after vaccination. The Klebsiella-dominant type was also less likely to show vaccine virus shedding than any other type except cs-GCT2, which was enriched in both Klebsiella and Escherichia. These patterns were restricted to month 1, indicating that the gut configuration in the first weeks of life, not the composition at the time of vaccination, predicted later response.

The researchers then turned to fecal calprotectin, a clinical biomarker of intestinal inflammation, to test whether these microbial patterns corresponded to an inflamed gut. Calprotectin did not differ between countries at month 1, but by month 2 it was elevated in Peruvian infants compared to those in Panama and the United States. More tellingly, infants with vaccine virus shedding detected after the second dose had significantly lower calprotectin at month 1, linking reduced early inflammation with enhanced mucosal replication of the vaccine virus. Differential abundance analysis showed that calprotectin at month 1 was negatively associated with Bifidobacterium longum and, more weakly, Bifidobacterium saguini, though not with Klebsiella or Escherichia. Calprotectin was also elevated at month 1 among infants in the Klebsiella- and Escherichia-enriched cs-GCT2 compared to the demographically similar but pathobiont-depleted cs-GCT1.

The mechanistic hypothesis that emerges from these data centers on LPS, the endotoxic outer membrane component of Gram-negative bacteria. Klebsiella overabundance in the first month of life was accompanied by enrichment of LPS biosynthetic potential, and the authors propose that this drives a pro-inflammatory mucosal environment that impedes vaccine take. The discussion in the paper situates the finding within a broader literature: neonatal antibiotic exposure in an Australian cohort was associated with decreased antibody responses to multiple vaccines through reductions in Bifidobacterium, while oral vaccine failure in Indian infants has been linked to Enterobacteriaceae enrichment and Bifidobacterium depletion. Non-polio enterovirus colonization and increased gut inflammation have similarly been tied to reduced poliovirus and rotavirus vaccine responses in South Asia and Central America. Animal models reinforce the framework, with neonatal microbiome disruption in mice impairing both mucosal and systemic vaccine responses in a manner that could be rescued through microbial restoration.

Notably, the high-response gut type cs-GCT6 was not characterized by enrichment of genera typically considered probiotic. The authors suggest that protection against vaccine hypo-responsiveness may result from exclusion of pro-inflammatory taxa rather than enrichment of beneficial ones, a subtle but important reframing. Consistent with this, probiotic supplementation did not improve Rotarix immunogenicity among infants in India in a prior randomized study, and reduced Proteobacteria abundance at birth correlated with improved vaccine shedding following neonatal rotavirus vaccination in Indonesia and Malawi. The relationship between shedding and seroconversion also appears paradoxical here: the authors propose that an inflamed gut may reflect neutrophil-driven clearance of the vaccine virus without concurrent generation of effective adaptive immune memory, since enhanced neutrophil activity can suppress immune memory through competition with dendritic cells for antigen or inhibition of IgA production.

The study has clear limitations. Seroconversion measured in blood represents systemic rather than localized gut response, and the degree of seroconversion may be confounded by pre-existing immunity, including transplacentally acquired maternal IgG, which the team did not assess. Human milk contains immunologic and compositional factors that may influence response, and the study was observational, so a causal relationship between gut microbiome and vaccine response cannot be established. Machine learning models built with LightGBM failed to predict seroconversion from microbiome data alone, with holdout area under the receiver operating characteristic curve not exceeding 50 percent at any time point, underscoring that context-specific ecological configurations rather than globally generalizable signatures drive outcomes. The Klebsiella-dominant gut type was also overrepresented by cesarean-delivered Panamanian infants, lacking a direct comparator among Peruvian infants.

Despite these caveats, the findings carry substantial implications for precision vaccine approaches. The authors argue that microbiome-targeted strategies, including synbiotic combinations, selective prebiotics to support Bifidobacterium, or phage-based depletion of Klebsiella, may offer greater specificity than broad probiotic supplementation, and that the timing of such interventions should align with the neonatal window during which the microbiome and immune system are most dynamically interacting. By identifying a microbial-metabolic-inflammatory axis centered on Klebsiella colonization and early-life gut inflammation, the study reframes oral vaccine failure not as a mystery of geography but as a consequence of which bacteria claim the infant gut in its first weeks, opening a concrete path toward interventions that could improve vaccine performance where it is needed most.

Subject of Research: Association between Klebsiella-dominant infant gut microbiome communities and impaired oral rotavirus vaccine response

Article Title: Klebsiella-dominant gut communities are associated with impeded oral rotavirus vaccine response in infants

Article References: Seymour, C. O., Ecklu-Mensah, G., Lee, Y., Nicols, A. R., Cheng, W. A., Jumarang, J. M., Marentes Ruiz, C. J., Navarro, J., Navarro, R., Rodriguez-Angeles, Y., Durand, D., Allard, S. M., Taylor, Z. W., De León, R. G., Ochoa, T. J., Gilbert, J. A., & Pannaraj, P. S. (2026). Klebsiella-dominant gut communities are associated with impeded oral rotavirus vaccine response in infants. Nature Communications, 17(1), Article 10783. https://doi.org/10.1038/s41467-026-78148-x

Image Credits: AI Generated

DOI: 10.1038/s41467-026-78148-x

Keywords: rotavirus vaccine, gut microbiome, Klebsiella, infants, seroconversion, lipopolysaccharide, fecal calprotectin, gut inflammation, Bifidobacterium, oral vaccines, Nature Communications, vaccine efficacy