Enrichment of Microbial DNA in Plasma to Improve Pathogen Detection in Sepsis: A Pilot Study

Abstract

Background: Diagnosis of sepsis and timely identification of pathogens in critically ill patients remains challenging. Plasma metagenomic sequencing to detect microbial cell-free DNA (mDNA) has shown promise, but low abundance of mDNA in plasma limits sensitivity and necessitates high sequencing depth. mDNA is shorter and more fragmented than human cell-free DNA. Here, we evaluated whether combining single-stranded DNA (ssDNA) sequencing library preparation and size selection can enrich mDNA and improve pathogen detection.

Methods: We prospectively enrolled 48 trauma patients and collected daily blood samples during the first 10 days of intensive care unit (ICU) admission. For patients with culture-proven infections, we extracted plasma DNA, prepared double-stranded DNA (dsDNA) and ssDNA sequencing libraries, and applied size selection to exclude fragments >110 bp. Following sequencing, we performed taxonomic classification, and evaluated differences in mDNA fractions and in sensitivity for pathogen detection (compared to background noise).

Results: We analyzed 46 plasma samples from 5 patients who developed culture-proven infections, including 17 samples coincident with positive microbial cultures. Size-selected ssDNA libraries showed the total mDNA fraction 204-fold higher on average than conventional dsDNA libraries (P < 0.0001). However, for pathogen-specific DNA (at the genus level), the highest sensitivity was observed in size-selected dsDNA (82%), compared to dsDNA (41%), ssDNA (71%), and size-selected ssDNA (35%) library preparations.

Conclusions: Our results demonstrate that combining ssDNA library preparation together with fragment size selection improves mDNA yield, potentially reducing sequencing requirements. However, at the genus level, this combination also increases background noise, which limits sensitivity for pathogen detection.

Figure 1:

Workflow for improving the recovery of microbial cell-free DNA using different library preparation protocols and size-selection. (A), Plasma samples were processed using semi-automated magnetic bead-based extraction methods, followed by dsDNA and ssDNA library protocols, size-selection to exclude DNA fragments greater than 110 bp in length, and sequencing. (B), Quantification of pooled libraries before and after application of size-selection for dsDNA and ssDNA library protocols. Created with BioRender.com.

Figure 2:

Comparison of microbial DNA fraction between paired samples (n=46) using double-stranded DNA (ds-DNA) and single-stranded DNA (ss-DNA) library preparation before and after size-selection (23-fold, p < 0.0001 for dsDNA size-selected compared to dsDNA and 5-fold, p < 0.0001 for ssDNA size-selected compared to ssDNA).

Figure 3:

Comparison of microbial fraction and cell-free DNA extraction yield in 1 ml of plasma across the four library preparation approaches.

Figure 4:

Detection rate in paired samples across 4 library preparation methods in samples coincident with positive cultures. Detectable: MAD scores were significant (> 2.5 MADs above median) for genus-specific mDNA fraction. Undetectable: Includes both not detectable and not significant above background noise. Table below barplot in the figure indicates detectable versus undetectable for 17 coincident plasma samples.

Figure 5:

Longitudinal patient-specific pathogen DNA quantification using the dsDNA size-selected library preparation. Filled circles indicate detection of pathogen DNA significantly above background (see methods). Grey indicates non-significant pathogen DNA detection, white indicates no detection. Boxes below each plot indicate microbiology cultures. Color ranges indicate antimicrobial treatment regimens.