A metagenomics-based diagnostic approach for central nervous system infections in hospital acute care setting

Abstract

The etiology of central nervous system (CNS) infections such as meningitis and encephalitis remains unknown in a large proportion of cases partly because the diversity of pathogens that may cause CNS infections greatly outnumber available test methods. We developed a metagenomic next generation sequencing (mNGS)-based approach for broad-range detection of pathogens associated with CNS infections suitable for application in the acute care hospital setting. The analytical sensitivity of mNGS performed on an Illumina MiSeq was assessed using simulated cerebrospinal fluid (CSF) specimens (n = 9). mNGS data were then used as a training dataset to optimize a bioinformatics workflow based on the IDseq pipeline. For clinical validation, residual CSF specimens (n = 74) from patients with suspected CNS infections previously tested by culture and/or PCR, were analyzed by mNGS. In simulated specimens, the NGS reads aligned to pathogen genomes in IDseq were correlated to qPCR C_T values for the respective pathogens (R = 0.96; p < 0.0001), and the results were highly specific for the spiked pathogens. In clinical samples, the diagnostic accuracy, sensitivity and specificity of the mNGS with reference to conventional methods were 100%, 95% and 96%, respectively. The clinical application of mNGS holds promise to benefit patients with CNS infections of unknown etiology.

Figure 1

Analytical sensitivity of mNGS to detect pathogens in CSF specimens is comparable to qPCR. CSF specimens (n = 9) negative by standard microbiological methods were spiked with a range of viral and bacterial pathogens at varying concentrations as described in the Materials and Methods or left unspiked and simultaneously tested along with a nuclease free water (NFW) sample by pathogen specific qPCR and by mNGS as described in the Materials and Methods. The approximate titer of pathogens (A) or qPCR C_T (B) values were plotted against mNGS read counts of pathogens obtained after analysis in IDseq.

Figure 2

Schematic of mNGS laboratory and bioinformatics workflow. Total nucleic acids (NA) from CSF specimens were extracted after spiking pUC19 plasmid as an internal control (IC) on an automated extraction platform, Qiasymphony (Qiagen). However, any other extraction platform or manual, spin column based methods can also be used for DNA extraction. DNA concertation was determined using a Qubit fluorometer (Thermofisher). NGS library preparation and clean up were performed according to manufacturer’s instructions with the exception that 96-well plates were replaced by single tubes or PCR tube strips because single sample was processed. NGS library quantification, normalization, sequencing and data analysis were performed as described in the materials and methods. TT total time, HT hands-on-time, RT reaction time, TAT turnaround time. NT r nucleotide reads, NT L nucleotide alignment length in bp, NT Z score nucleotide Z score, NR r non-redundant reads.

Figure 3

Heatmap of taxa identified in the validation set of specimens with IDseq after applying CSF background and customized threshold filters. mNGS was performed and data were analyzed as described in Fig. 2. Heatmaps were generated based on log₁₀[NT total reads]. Different filter sets were applied for bacteria and eukaryotes versus viruses and E. coli reads were manually derived from IDseq after manual review of data. All positive and negative results by mNGS are shown by ‘+’ and ‘−’ signs, respectively.