Integration of metagenome-assembled genomes with clinical isolates reveals genomic signatures ofKlebsiella pneumoniaein carriage and disease

Klebsiella pneumoniaeis an opportunistic pathogen causing diseases ranging from gastrointestinal disorders to severe liver abscesses. While clinical isolates ofK. pneumoniaehave been extensively studied, less is known about asymptomatic variants colonizing the human gut across diverse populations. Genome-resolved metagenomics has offered unprecedented access to metagenome-assembled genomes (MAGs) from diverse host states and geographical locations, opening opportunities to explore health-associated microbial features. Here we analysed 662 human gut-derivedK. pneumoniaegenomes (319 MAGs, 343 isolates) from 29 countries to investigate the population structure and genomic diversity ofK. pneumoniaein carriage and disease. Only 9% of sequence types were found to be shared between healthy and disease states, highlighting distinct diversity across health conditions. Integrating MAGs nearly doubled gut-associatedK. pneumoniaephylogenetic diversity, and uncovered 86 lineages without representation among >20,000Klebsiellaisolate genomes from various sources. Genomic signatures linked to pathogenicity and carriage included those involved in antibiotic resistance, iron regulation, restriction modification systems and polysaccharide biosynthesis. Notably, machine learning models integrating MAGs and isolates more accurately classified disease and carriage states compared to isolates alone. These findings showcase the value of metagenomics to understand pathogen evolution with implications for public health surveillance strategies.