Horizontal Gene Transfer Shapes Resistome Concordance Across Urinary Tract and Bloodstream Infection Isolates

Background: Timely identification of effective empirical therapy is essential for the management of bloodstream infections yet culture-based diagnostics remain slow and often inaccessible, especially in resource-limited settings. Because urinary tract infections frequently cause urosepsis, they are most frequently caused by Escherichia coli (E. coli), Klebsiella pneumoniae (K. pneumoniae), and Enterococcus faecium (E. faecium). A key clinical question is whether urinary antimicrobial resistance (AMR) profiles reliably be used to predict bloodstream resistance patterns as this would be clinically relevant in resource poor countries.<br><br>Methods: We analysed large global genomic datasets of E. coli, K. pneumoniae, and E. faecium from 3,370 urine and 4,675 blood isolates across 6 World Health Organisation (WHO)-designated regions. Four (4) molecular dimensions were evaluated: (1) AMR gene families, (2) phenotypic AMR profiles (antibiotic classes and individual drugs), (3) plasmid replicon families, and (4) multi-locus sequence types. Concordance between urine and blood isolates was quantified using Spearman correlation (ρ), concordance correlation coefficients (CCC), Jaccard indices, and Bray-Curtis dissimilarity. A pangenome analysis of all three organism groups was carried out to assess accessory genome overlaps.<br><br>Results: Urinary and bloodstream isolates from all three organism groups demonstrated very high concordance in resistomes and predicted phenotypic AMR. Preliminary analyses demonstrated broadly overlapping predicted resistance patterns between urine and blood within each species. Pangenome-based gene accumulation and Heaps’ law modelling showed open pangenomes across all three species indicating sustained acquisition of novel gene clusters at late sampling depths with the highest openness in K. pneumoniae (γ=0.277; 95% interval 0.211-0.342) followed by E. coli (γ=0.221; 0.202-0.235) and E. faecium (γ=0.203; 0.108-0.249). Distribution of accessory gene content revealed substantial overlap between infection sources, with minimal source-associated separation. Quantitative concordance metrics confirmed very high agreement between urine and blood isolates for AMR gene families, AMR classes, predicted antibiotic-level resistance, and plasmid replicon profiles (Spearman’s ρ ≥0.86; CCC ≥0.98 for most comparisons). In contrast, multi-locus sequence type concordance was low across species, which translates to limited clonal overlap between urine and blood groups.<br><br>Interpretation: The high resistome concordance between urinary and bloodstream resistomes across three key uropathogens suggests that resistance circulates within a shared, mobile gene pool spanning infection sites. This decoupling of resistance from the clonal background highlights the central role of horizontal gene transfer in shaping AMR epidemiology. Urine-derived genomic data may therefore offer a scalable and timely proxy for population-level resistance surveillance, especially in regions where invasive sampling is constrained, strengthening early warning systems for emerging resistance threats.