Yes / Background. Bacterial bloodstream infection (bBSI) is one of the leading causes of death in critically ill patients and accurate diagnosis is therefore crucial. We here report a 16S metagenomics approach for diagnosing and understanding bBSI. Methodology/Principal Findings. The proof-of-concept was delivered in 75 children (median age 15 months) with severe febrile illness in Burkina Faso. Standard blood culture and malaria testing were conducted at the time of hospital admission. 16S metagenomics testing was done retrospectively and in duplicate on the blood of all patients. Total DNA was extracted from the blood and the V3–V4 regions of the bacterial 16S rRNA genes were amplified by PCR and deep sequenced on an Illumina MiSeq sequencer. Paired reads were curated, taxonomically labeled, and filtered. Blood culture diagnosed bBSI in 12 patients, but this number increased to 22 patients when combining blood culture and 16S metagenomics results. In addition to superior sensitivity compared to standard blood culture, 16S metagenomics revealed important novel insights into the nature of bBSI. Patients with acute malaria or recovering from malaria had a 7-fold higher risk of presenting polymicrobial bloodstream infections compared to patients with no recent malaria diagnosis (p-value = 0.046). Malaria is known to affect epithelial gut function and may thus facilitate bacterial translocation from the intestinal lumen to the blood. Importantly, patients with such polymicrobial blood infections showed a 9-fold higher risk factor for not surviving their febrile illness (p-value = 0.030). Conclusions/Significance. Our data demonstrate that 16S metagenomics is a powerful approach for the diagnosis and understanding of bBSI. This proof-of-concept study also showed that appropriate control samples are crucial to detect background signals due to environmental contamination. / This work was supported by the Flemish Ministry of Sciences (EWI, SOFI project IDIS). / This paper has been subject to a correction. Please see Correction file above.

Bacterial bloodstream infections (BSIs) are a major healthcare problem causing high mortality and economic cost. BSIs require an immediate initiation of antibiotic therapy as any delay is associated with a mortality increase. With the emergence of antimicrobial resistance, the choice of the appropriate antibiotic becomes increasingly difficult, thus creating an urgent need for new diagnostics, ideally to be done at the point of care. The current gold standard is blood culture with subsequent susceptibility testing although several molecular methods have recently entered the market. However, in many instances there is a discrepancy between the in-vitro data provided by the test and the outcome of antimicrobial therapy in-vivo because current diagnostics fail to take into account the impact of the environment in the patient such as the immune system, pharmacokinetics and pharmacodynamics or bacterial fitness. In this thesis, it was hypothesised that the measurement of the bacterial gene expression after the beginning of antibiotic therapy might be a more accurate indicator of the therapy outcome because it reflects the bacterial response under in-vivo conditions. In the first part of the thesis the expression of a set of pre-defined mRNA markers was investigated under various conditions. Experiments conducted with clinical E. coli isolates incubated in human whole blood revealed an excellent correlation between the gene expression, the treatment outcome, the antibiotic susceptibility and the genetic background for three different classes of antimicrobial drugs. The second part of the thesis describes the extraction of bacterial RNA from human whole blood specimen. The effect of different agents for the lysis of human blood cells and the impact of co-purified human RNA were analysed and a method for high yield extraction of undegraded bacterial RNA was established. The third part of the thesis investigates two methods for the sensitive measurement of the bacterial gene expression. This is relevant because the bacterial loads in BSI patients are extremely low. For genes with high gene expression levels both methods yielded reliable results but were unable to quantify the expression of the previously investigated mRNA markers due to their low copy numbers. Other approaches, especially those based on single cell measurements, might be able to overcome the problem in the future and should be explored in greater detail. Overall, the foundations for a future diagnostic test based on the measurement of the bacterial gene expression have been laid in this work. Future work should address the mRNA quantification and further evaluate the connection between gene expression and therapy outcome, e.g. in animal models. A future diagnostic test should also fulfil point-of-care requirements. This will include integrated sample preparation and quantification as well as a time-to-result in the range of a few minutes.