ARTICLE IN BRIEF
In a small series of patients with undiagnosed chronic meningitis, “metagenomic” next-generation sequencing identified the causative organism in every case, and in one case identified a pathogen that had not been previously linked to central nervous system vasculitis.
Establishing a definitive diagnosis in a patient with chronic meningitis can be a challenge; in fact it is estimated that only a third of patients reaching a tertiary care center will ever know exactly what they are infected with, or whether they are infected at all.
But a new study in the April 16 online edition of JAMA Neurology suggests that may all change. In a small series of patients with undiagnosed chronic meningitis, “metagenomic” next-generation sequencing identified the causative organism in every case, and in one case identified a pathogen that had not been previously linked to central nervous system vasculitis.
“In acute meningitis, the standard diagnostic process— running individual tests for a small handful of likely suspects — works pretty well to identify the infectious organism,” said the lead study author Michael R. Wilson, MD, assistant professor of neurology at the University of California, San Francisco (UCSF). “But in chronic meningitis, the organisms tend to be more unusual and harder to diagnose;” indeed, that difficulty precludes the effective treatment that would prevent long-term infection.
“If the real infectious organism is not on your list, the candidate approach won't work,” Dr. Wilson said.
That challenge led Dr. Wilson, first-co-authors Jeffrey Gelfand, MD, and bioinformatics graduate student Brian O'Donovan, and senior author Joseph DeRisi, Phd, all of UCSF, to test whether an unbiased approach might be better for the toughest cases. They used a new technique called metagenomics, which analyzes the genetic material of every organism in a sample, no matter its taxonomic classification, abundance, or likely origin. Metagenomics has been used to study a wide variety of environmental samples, revealing unexpected microorganismic diversity in everything from seawater to the human gut (For more on metagenomics, see the box, “The Science Explained: Metagenomics”).
High-throughput metagenomics depends on so-called next-generation sequencing technologies, which rapidly generate the genomic sequences of any piece of DNA or RNA in a sample, and bioinformatics, which compares those sequences to giant databases of genomic sequences from thousands of organisms, looking for the nearest match.
“With this approach, you can look for any infection with a single test, and you can find unexpected things,” Dr. Wilson said. Since there is no culturing involved, there is no bias for the relatively small number of microbes that can be grown in the lab.
For the study, Dr. Wilson and O'Donovan generated a large database, with rows for organisms and columns for samples. “Each cell in the spreadsheet has the abundance of that organism in that sample,” he said, which is derived from the sequencing results — the stronger the sequencing signal is, the more abundant the organism is in the sample.
Dr. Wilson said: “For each organism, you can create a distribution of its abundance across all samples, so that when you get a new sample, you can ask how unusual it is to find it at this level of abundance, based on its distribution in previous samples. An unusually high abundance suggests it may be the culprit.”
However, Dr. Wilson pointed out, “the blessing, and the curse, of an unbiased approach is that you see everything,” including many fragments of nucleic acids that are not from the cerebrospinal fluid (CSF), but are instead inevitable contaminants introduced along the way, often shed by medical personnel during collection or from the skin of the patient. Thus, the database also contained columns for sterile water, reagents, and confirmed noninfectious CSF, to serve as controls. Organisms found in controls can then be discounted when they turn up in patient samples.
For the study, Dr. Wilson examined CSF from seven patients with undiagnosed, chronic meningitis, with or without encephalitis. In the clinical laboratory, sequencing library prepation and data generation typically takes an average of 72 hours, but the metagenomic analysis — removing human and redundant sequences, applying sequence quality filters, aligning sequences to microbial databases, identifying specific organisms, comparing them to controls, and ranking the remaining candidates — was typically completed in under 20 minutes.
From the shortlist of possible infectious agents, the actual cause was reported as number one for six patients, and number two for the seventh. One patient, who had a positive tuberculin skin test and a positive serum cysticercosis antibody test but without cysts or calcifications on brain magnetic resonance imaging (MRI), had been treated unsuccessfully for tuberculosis. Sequencing and metagenomic analysis revealed Taenia soleum (pork tapeworm) as the top candidate; he was treated successfully with antihelminthic therapy.
A second patient contributed by colleagues from the Cleveland Clinic with symptoms for seven months but with negative results for bacterial and fungal cultures was treated unsuccessfully for an autoimmune disorder. He subsequently tested positive for Aspergillus, and metagenomic analysis indicated the likely species was A. oryzae, the only reported case of CNS vasculitis from this species. In this patient, the top hit from the analysis was GB virus C.
“The patient was infected with the virus, but the neurologic problems didn't fit with that infection,” Dr. Wilson said. “This test can detect three to five simultaneous pathogens, and then you need to do some thinking based on the clinical context about which is most relevant for the symptoms you see. Sometimes it's easy, sometimes it's not.”
UCSF now has a lab certified to offer metagenomic testing, and other groups around the country are developing similar protocols, Dr. Wilson said. Metagenomics is not likely to completely replace standard meningitis testing in the near future, he added, for reasons of both cost and speed. But that day may be coming, as newer sequencing technologies evolve. “I think it is very realistic to think that in the next five years you might get a same-day turnaround.”
The metagenomic approach offers a major advantage over the standard diagnostic approach to central nervous system infections, said Jennifer Lyons, MD, chief of the division of neurological infections and inflammatory diseases at Brigham and Women's Hospital and assistant professor of neurology at Harvard Medical School in Boston. With standard CSF assays requiring 500 microliters or more per pathogen, standard lumbar puncture retrieves 500 microliters or less, barely enough for a single test, “there is no way to comprehensively evaluate for CNS pathogens,” she commented. “The big strength of this study is that the method is not only novel, but is very useful. Once this is accessible on a wider scale, it promises to be a much better way to diagnose nervous system infections than the methods we currently have.”
Avindra Nath, MD, MBBS, FAAN, chief of the Section of Infections of the Nervous System at the National Institute of Neurologic Disorders and Stroke, added: “This approach could be very important” in arriving at a diagnosis must faster, potentially reducing empiric testing of treatments in the absence of a diagnosis “with the possibility of doing the patient harm.” However, he noted, a current bottleneck for expansion of metagenomics beyond a few specialized centers is the lack of access to this technology and the lack of personnel trained in the bioinformatics that can turn a mountain of sequence data into the identification of a true pathogen. “But if you have a patient with meningitis or encephalitis, and you don't find an etiology by the regular investigations, this should be considered,” he said.
Kenneth L. Tyler, MD, FAAN, professor and chair of neurology at the University of Colorado, who wrote an accompanying editorial in JAMA Neurology, told Neurology Today: “One of the hardest problems in chronic meningitis is that much of our testing depends on thinking of the candidate pathogen,” leading to both errors of omission — leaving out the actual pathogen — and errors of commission, by ordering too many tests, with high costs and the possibility of false positives. The promise of metagenomics is that “it doesn't depend on your thinking of the candidate pathogen a priori.”
Furthermore, the ability to identify novel organisms, or ones that have mutated in some way to make them newly pathogenic, offers additional benefits. “This is not just an interesting research approach,” Dr. Tyler said. “It is a potentially very valuable clinical tool.”
THE SCIENCE EXPLAINED: METAGENOMICS
WHAT IT IS: Metagenomics is the analysis of all the genetic material — RNA or DNA or both — in any sample, to identify all the organisms present and to characterize their abundance.
THE TECHNIQUE: A sample is analyzed with “next-generation” sequencing, in which the genetic material is fragmented to allow for rapid, massively parallel sequencing. The resulting sequences are then built back up, by aligning overlapping regions, to reveal the sequences of the original genetic material. The organisms contributing these sequences are then identified by comparing them to databases of the genomes of all known organisms. The relative abundance of each organism present is determined by comparing the total number of sequence fragments from each organism in the dataset.
HOW IT IS USED: By identifying the full breadth of species present, metagenomics has revolutionized the understanding of microbial communities wherever they occur. Metagenomics has been used to analyze microbial life in deep sea vents, toxic waste dumps, seawater, the human gut, any many other environments. In each case, the diversity and complexity of the microbial communities was greater than previously thought.