A multi-hospital, clinician-initiated bacterial genomics programme to investigate treatment failure in severe Staphylococcus aureus infections

Abstract

Bacterial genomics is increasingly used for infectious diseases surveillance, outbreak detection and prediction of antibiotic resistance. With expanding availability of rapid whole-genome sequencing, bacterial genomics data could become a valuable tool for clinicians managing bacterial infections, driving precision medicine strategies. Here, we present a clinician-driven bacterial genomics framework that applies within-patient evolutionary analysis to identify in real-time microbial genetic changes that have an impact on treatment outcomes of severe Staphylococcus aureus infections, a strategy that is increasingly used in cancer genomics. Our approach uses a combination of bacterial genomics and antibiotic susceptibility testing to identify and track bacterial adaptive mutations that underlie microbiologically documented treatment failure (i.e. ongoing positive cultures [persistent infection] or new positive cultures after initial response [recurrent infection]). We show the potential added value of our approach to clinicians and propose a roadmap for the use of bacterial genomics to advance the management of severe bacterial infections.

Fig. 1. Overview of study approach and results.

A Proposed approach to identify and track bacterial adaptive mutations underlying antibiotic treatment failure. WGS: whole-genome sequencing. B Maximum-likelihood phylogenetic tree of 798 S. aureus bacteraemia strains showing the genetic background of isolates from antibiotic treatment failure. Of note, all same-patient isolates belong to monophyletic clades. MRSA: methicillin-resistant S. aureus. MSSA: methicillin-susceptible S. aureus. C Timeline, observed phenotypic changes, and number of within-host mutations (relative to a baseline isolate). Oxa oxacillin, MIC minimum inhibitory concentration, CFZ cefazolin, S susceptible, R resistant. D distribution of isolate collection days (after baseline isolate) between persistent and recurrent case. E Distribution of within-host mutation pairwise counts. Source data are provided as a Source Data file.

Fig. 2. Within-host evolution analysis of adaptive episodes.

A Detailed within-host evolution analysis of four cases with evidence of adaptive mutations. The first panel shows the location of the identified mutation. The second panel shows the statistical significance of the enrichment (adaptive signal: Poisson regression, one-sided likelihood ratio test, no adjustment for multiple comparisons) in a large-scale within-host evolution analysis of 400 S. aureus infection episodes. Dot and bars are coloured according to the functional association of the mutated gene. MIC minimum inhibitory concentration. B, C Reconstruction of within-host evolutionary trajectories of two cases with >2 clinical isolates. The maximum-likelihood trees are inferred from a core-genome alignment on the internal reference. Trees were rooted using the closest available isolate from the bacteraemia collection used for the global phylogeny. Tips are annotated with the day of collection, with the baseline isolate displayed as the day 0 isolate. Inferred emergence of non-synonymous substitutions and truncations is annotated on the internal nodes or branches. B Within-host phylogenetic tree of Case C/Nov 2019 persistent and relapsing bacteraemia with progressive oxacillin MIC increase and emergence of small colony variants. C Within-host phylogenetic tree of Case A/May 2019: no phenotypic changes. Source data are provided as a Source Data file.

Fig. 3. Overview of the RedCap survey of 25 infectious diseases specialists.

A Clinicians assessment of probability of adaptive mutations, report readability, and added value, using a score between 0 and 100. The box plots represent median, interquartile range (hinges), and 1.5 times the interquartile range (whiskers), with outliers represented as individual dots. B Changes in recommended antibiotic duration (based on an ordinal score) after providing additional phenotypic and genomic information. C–E antibiotic switch (C, E) and source control (D, F) recommendations before and after reading additional phenotypic and genomic information, for cases with (E, F) and without adaptive mutations (C, D). Source data are provided as a Source Data file.