Metagenomic sequencing of cerebral spinal fluid (CSF)

Pathogen detection in central nervous system infections Nicola Cumley, Josh Quick, Thomas Brier, Sam Wilkinson, Chris Kent, Zaki Hassan-Smith, Nicholas Loman, Ghaniah Hassan-Smith Background Pathogen detection rates remain low in central nervous system infection (CNSI), metagenomic sequencing has the potential to increase diagnostic sensitivity, particularly in unusual or partially treated infections. We analysed the potential for cell free nucleic acid to enrich pathogen signal in CSF from patients with and without CNSI. Organisms detected were compared to the organisms identified by the clinical microbiology laboratory. We developed a simple. bioinformatic filtering strategy to remove background contamination. Methods Samples of CSF were obtained from patients requiring CSF analysis. Total nucleic acid was extracted from the supernatant (cell free) and the cell pellet of every sample. Half of the eluate underwent cDNA synthesis and amplification using SMART-9N. DNA and cDNA were library prepped for sequencing using the Illumina platform. Organisms were identified using KrakenUniq. Mapped reads were normalised to reads per million (RPM) and an E-value was calculated to enable threshold values for filtering the data. The presence of organisms was confirmed using alignments to reference genomes. Results 32 patients were recruited. RPM and horizontal genome coverage were compared between organisms that were detected in both cell free and cell pellet fractions. Coverage was increased in the cell free fraction in 72% of organisms. Neisseria meningitidis had a striking increase in coverage breadth in the cell free fraction of 88% compared to 4% in the cell pellet fraction. The organisms detected in both fractions matched the results obtained by the clinical microbiology laboratory in 75% patients. These included fungal, bacterial and viral pathogens. Discussion This study has demonstrated the cell free fraction of CSF provides pathogen nucleic acid enrichment as measured by increased genome recovery. Using a simple and robust filtering strategy reads mapped to organisms could be confidently identified and compared to clinical microbiology results.