Coupling time-aware SNP thresholds with genetic markers to define bacterial transmission chains in hospital surveillance

Hospital settings can act as reservoirs for pathogens, with many nosocomial bacteria surviving for extended periods of time and spreading via contaminated environments, healthcare workers, or medical equipment. Infection prevention and control in hospital settings relies on accurately identifying transmission events to prevent further transmission. Whole-genome sequencing (WGS) offers the possibility to accurately identify bacterial strains and define transmission routes, yet traditional analyses often rely on fixed single nucleotide polymorphisms (SNP) thresholds that fail to account for the accumulation of variations over extended periods of time. We analyzed WGS surveillance data from bacterial isolates collected over five years in two hospitals in Germany and Denmark. By leveraging longitudinal sampling and repeated isolates from persistent infections, we estimated in vivo evolutionary rates from different strains belonging to three species: Escherichia coli, Klebsiella pneumoniae and Pseudomonas aeruginosa. Using species-level mean molecular clock rates, we developed a time-aware framework that defines transmission threshold based on expected SNP accumulation over time. Using this approach, transmission clusters were detected in 0.3% of E. coli, 9.3% of K. pneumoniae and 3.5% of P. aeruginosa isolates. To understand the genetic factors underlying epidemic strain potential, we compared epidemic lineages part of transmission chains with sporadic lineages. We found that epidemic lineages of K. pneumoniae and E. coli had higher virulence scores than sporadic strains, with enrichment of siderophores and adhesion genes, while resistance scores were similar. Genome-wide association analyses revealed hundreds of variants associated with epidemic status, particularly in replication, recombination and repair mechanisms, as well as metabolism related functions. With the predicted virulence and resistance scores we could easily observe predicted phenotypic changes within transmission clusters, including the loss of virulence factors in an outbreak, and the emergence of resistance after antibiotic treatment.