Clinical Metagenomic Sequencing for Diagnosis of Meningitis and Encephalitis

Abstract

Background: Metagenomic next-generation sequencing (NGS) of cerebrospinal fluid (CSF) has the potential to identify a broad range of pathogens in a single test.

Methods: In a 1-year, multicenter, prospective study, we investigated the usefulness of metagenomic NGS of CSF for the diagnosis of infectious meningitis and encephalitis in hospitalized patients. All positive tests for pathogens on metagenomic NGS were confirmed by orthogonal laboratory testing. Physician feedback was elicited by teleconferences with a clinical microbial sequencing board and by surveys. Clinical effect was evaluated by retrospective chart review.

Results: We enrolled 204 pediatric and adult patients at eight hospitals. Patients were severely ill: 48.5% had been admitted to the intensive care unit, and the 30-day mortality among all study patients was 11.3%. A total of 58 infections of the nervous system were diagnosed in 57 patients (27.9%). Among these 58 infections, metagenomic NGS identified 13 (22%) that were not identified by clinical testing at the source hospital. Among the remaining 45 infections (78%), metagenomic NGS made concurrent diagnoses in 19. Of the 26 infections not identified by metagenomic NGS, 11 were diagnosed by serologic testing only, 7 were diagnosed from tissue samples other than CSF, and 8 were negative on metagenomic NGS owing to low titers of pathogens in CSF. A total of 8 of 13 diagnoses made solely by metagenomic NGS had a likely clinical effect, with 7 of 13 guiding treatment.

Conclusions: Routine microbiologic testing is often insufficient to detect all neuroinvasive pathogens. In this study, metagenomic NGS of CSF obtained from patients with meningitis or encephalitis improved diagnosis of neurologic infections and provided actionable information in some cases. (Funded by the National Institutes of Health and others; PDAID ClinicalTrials.gov number, NCT02910037.).

Conflict of interest statement

Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.

Copyright © 2019 Massachusetts Medical Society.

Figures

Figure 1.. Overview of the Study.

Panel A shows the flow of patients through the study. Panel B shows 8 participating sites. The size of the circle is proportional to the number of patients enrolled at a given site. Panel C shows the protocol for the metagenomic next-generation sequencing (NGS) assay. After samples of cerebrospinal fluid (CSF) are received in the clinical laboratory, nucleic acid (DNA and RNA) is isolated, followed by construction of a metagenomic NGS library and sequencing. The metagenomic NGS data are analyzed with the use of an automated computational pipeline (Sequence-based Ultra-Rapid Pathogen Identification [SURPI+]), with results reported in the electronic medical record (EMR) after review by the laboratory director. CHCO denotes Children’s Hospital Colorado; CHLA Children’s Hospital Los Angeles; CNMC Children’s National Medical Center; SJCRH St. Jude Children’s Research Hospital; UCD University of California, Davis; UCLA University of California, Los Angeles; UCSF University of California, San Francisco; and ZSFGH Zuckerberg San Francisco General Hospital.

Figure 2 (facing page).. Results of Metagenomic NGS Testing and Clinical Effect.

Panel A shows the proportion and categories of established diagnoses in the study patients. A diagnosis was made in 103 of 204 patients (50.5%) after routine clinical workup and metagenomic NGS testing of CSF. A total of 58 infections (pink circles) were identified in 57 patients (27.9%). Conventional testing included culture, polymerase-chain-reaction (PCR), serologic (antibody), and antigen testing of CSF and other body fluids or tissues. Diagnoses in the “Other” category included resolving treated infection, idiopathic intracranial hypertension, posterior reverse encephalopathy syndrome, postneurosurgical (chemical) meningitis, and hemophagocytic lymphohistiocytosis. In Panel B, a plot shows the number and percentage of patients with high DNA or RNA background at designated intervals of CSF cell counts. The proportion of samples with high background (defined as samples in which the normalized read counts corresponding to the internal spiked DNA or RNA control did not meet preestablished thresholds) increases with increasing cell count. Panel C shows supplementary metagenomic NGS analyses discussed during meetings of the clinical microbial sequencing board (CMSB). Panel D shows clinician feedback for cases diagnosed solely by metagenomic NGS. Panel E shows the clinical effect of cases diagnosed solely by metagenomic NGS. The specific effect of metagenomic NGS results on the initiation, discontinuation, or length of antibiotic or antiviral treatment is described. EBV denotes Epstein–Barr virus, HEV hepatitis E virus, HIV-1 human immunodeficiency virus type 1, IV intravenous, and SLEV St. Louis encephalitis virus.

Figure 3 (facing page).. Supplementary Metagenomic NGS Analyses.

Supplementary analyses of the metagenomic NGS data were performed and results discussed during weekly teleconferences with the clinical microbial sequencing board. The asterisk denotes the column on the interactive SURPI+ heat map corresponding to the patient’s CSF sample, and pop-up windows highlight the cell corresponding to the given species hit (see Supplementary Appendix for additional details). For Panels A and B, the green tracing corresponds to the coverage at a given nucleotide position (y axis, left), and the purple tracing corresponds to the pairwise identity (y axis, right) after automated mapping by SURPI+ of metagenomic NGS reads to the most closely matched viral reference genome in the National Center for Biotechnology Information (NCBI) nucleotide (nt) database. Panel A shows prediction of resistance to antiviral drugs. Mapping HIV-1 reads from a patient CSF sample to the most closely matched genome in the reference database shows that the complete viral genome can be assembled (middle), thus enabling prediction of antiviral drug resistance (right). Predicted z scores were obtained with the use of Web-based geno2pheno software. The z scores corresponding to a subset of commonly prescribed antiretroviral drugs (black) are shown relative to reference z-score ranges for susceptible (green), intermediate (yellow), or resistant (orange) phenotypes. 3TC denotes lamivudine, ABC abacavir, ATZ/r ritonavir-boosted atazanavir, DRV/r ritonavir-boosted darunavir, EFV efavirenz, LPV/r ritonavir-boosted lopinavir, NNRTI nonnucleoside reverse-transcriptase inhibitor, NRTI nucleoside reverse-transcriptase inhibitor, NVP nevirapine, PI protease inhibitor, TDF tenofovir, and ZDV zidovudine. Panel B shows viral genotyping. The viral genome in an enterovirus B–positive case was assembled from metagenomic NGS reads, and the specific viral strain was identified as coxsackievirus B5 by SURPI+ (right). Panel C shows longitudinal tracking of viral infection. MW polyomavirus, originally identified in stool from children with diarrhea, was detected in an immunocompromised child presenting with acute meningoencephalitis. The finding was thought to be of unclear clinical significance, although no other infectious cause was identified. Zero and 12 reads to MW polyomavirus were detected in two CSF samples obtained 3 months later, during a second hospitalization for documented varicella–zoster virus (VZV) uveitis. Panel D shows accurate species identification. Assembly of the full-length 16S rRNA gene from metagenomic NGS reads enabled phylogenetic analysis and assignment of the species as Streptococcus mitis. A phylogenetic tree was obtained by aligning 25 representative S. mitis and 25 representative Streptococcus pneumoniae strains (with Streptococcus pyogenes as an outgroup) with the patient’s 16S rRNA sequence with the use of MAFFT at default settings, followed by tree construction with the use of PhyML. Panel E shows analysis of antibiotic-resistance genes. Such genes were identified by alignment of Enterobacter aerogenes (now renamed Klebsiella aerogenes) metagenomic NGS reads to the comprehensive antibiotic-resistance database. Panel F shows the detection of pathogen reads below the reporting threshold, with heat maps corresponding to two pathogens (Mycobacterium bovis and astrovirus MLB2) that were not reported as positive by metagenomic NGS because the number and distribution of reads did not meet preestablished thresholds. In Panels E and F, AmpC denotes class C β-lactamase, and wgs the NCBI whole-genome shotgun database.