Plasma metagenomic sequencing to detect and quantify bacterial DNA in ICU patients suspected of sepsis: a proof-of-principle study.

INTRODUCTION Timely recognition of sepsis and identification of pathogens can improve outcomes in critical care patients but microbial cultures have low accuracy and long turnaround times. In this proof-of-principle study, we describe metagenomic sequencing and analysis of non-human DNA in plasma. We hypothesized that quantitative analysis of bacterial DNA (bDNA) levels in plasma can enable detection and monitoring of pathogens. METHODS We enrolled 30 patients suspected of sepsis in the surgical trauma ICU and collected plasma samples at time of diagnostic workup for sepsis (baseline), and 7 and 14 days later. We performed metagenomic sequencing of plasma DNA, and used computational classification of sequencing reads to detect and quantify total and pathogen-specific bDNA fraction. To improve assay sensitivity, we developed an enrichment method for bacterial DNA based on size selection for shorter fragment lengths. Differences in bDNA fractions between samples were evaluated using t test and linear mixed-effects model, following log transformation. RESULTS We analyzed 72 plasma samples from 30 patients. 27 samples (37.5%) were collected at the time of infection. Median total bDNA fraction was 1.6 times higher in these samples compared to samples with no infection (0.011% and 0.0068%, respectively, p < 0.001). In 17 patients who had active infection at enrollment and at least one follow-up sample collected, total bDNA fractions were higher at baseline compared to the next sample (p < 0.001). Following enrichment, bDNA fractions increased in paired samples by a mean of 16.9 fold. Of 17 samples collected at the time when bacterial pathogens were identified, we detected pathogen-specific DNA in 13 plasma samples (76.5%). CONCLUSION Bacterial DNA levels in plasma are elevated in critically ill patients with active infection. Pathogen-specific DNA is detectable in plasma, particularly after enrichment using selection for shorter fragments. Serial changes in bacterial DNA levels may be informative of treatment response.