SAN FRANCISCO (GenomeWeb) – For some 40 to 60 percent of patients suffering from meningitis, a cause is never pinpointed. That's a statistic that researchers at the University of California, San Francisco think they can change by performing metagenomic sequencing on patients' cerebral spinal fluid.
This week, a team led by Joe DeRisi's laboratory at UCSF and the Chan Zuckerberg Biohub described in a study published in JAMA Neurology how a metagenomic next-generation sequencing approach could find the cause of chronic meningitis in seven cases that illustrate the various complexities associated with diagnosis. DeRisi's team focuses on the research side of infectious disease diagnostics, developing and testing new methods, but works closely with another UCSF group led by Charles Chiu, which launched a clinical metagenomic NGS assay for meningitis last year.
The cases tested are "highly illustrative of the kinds of conundrums that clinicians face when dealing with neurological disease," DeRisi said. Not only is it difficult to identify a pathogen, but often patients' symptoms are consistent not only with a potential infectious agent but also with autoimmune disorder.
"It's not straightforward how to rule one out, and they have very different treatment implications," DeRisi said. For instance, patients with autoimmune disorders often need to be treated with steroids. The longer they go without steroids, the worse the outcome. But on the other hand, steroids can be very damaging if given to a patient who actually has a bacterial infection, he said.
Many of these indications "are very costly and have poor outcomes," DeRisi added, which was one reason why his lab decided to focus on developing methods for faster diagnosis. Neurological diseases in general have a high economic and health burden, "and of those, meningitis and encephalitis are at the top," he said.
DeRisi added that the seven cases in this study illustrate the complexity of cases that physicians deal with and the potential for metagenomic sequencing to sufficiently address them.
For instance, the researchers analyzed patients ranging from 10 years to 55 years of age who had diagnoses that included a range of etiologies across fungal, bacterial, and viral species.
DeRisi described one case of a 26-year-old female who had been suffering from lower back pain that eventually led to foot drop, an abnormality in which the sufferer has trouble lifting her foot, often due to a neurological or sciatic issue.
An MRI showed a cyst at the top of her spine and brainstem. "Every attempt was made to culture the organism from her spinal fluid," DeRisi said. The sample was also sent out for testing via both 18S and 16S rRNA sequencing and nothing was found. Metagenomic sequencing identified reads that matched the fungus Candida dubliniensis. The other tests failed, DeRisi said, because the species is "such an unusual form of fungus." So even though so-called "universal PCR" primers were used, he added, they failed.
In the study, the researchers were able to identify pathogens in all seven cases, but DeRisi stressed that it was not designed to analyze the sensitivity or specificity of the test, but merely to illustrate the technology's potential on a diverse range of cases.
The assay described in the JAMA Neurology study is the research version of the test, while Chiu is leading the team implementing the clinical version and plans to publish results on an approximately 300-patient study that will describe the sensitivity and specificity of that test.
The prospective clinical trial for the test began in June 2016 under a project called Precision Diagnosis of Acute Infectious Diseases with funding from the California Initiative to Advance Precision Medicine. The trial included seven participating hospitals, and last year, Chiu described preliminary results as "very promising."
Michael Wilson, lead author of the JAMA Neurology study, said that the research and clinical teams work closely together, and although the assays that the two groups use are conceptually very similar in that they both use metagenomic sequencing, there are some differences.
"A clinical test has to be locked down," Wilson said, "so we on the research side are constantly playing with new computational changes and molecular tweaks to try and make the test more sensitive, faster, and cheaper," he said. "So one year from now, if we've accumulated enough substantial improvements to the test in the research lab, it might make more sense to revalidate a clinical test that incorporates some of those changes."
One technique that the group used in the recent study but that is not used in the clinical assay is a method for depleting unwanted sequences. The method, depletion of abundant sequences by hybridization, or DASH, makes use of the Cas9 enzyme to target and cut out unwanted sequences. The UCSF team described the method in Genome Biology in 2016.
Wilson said that although there are commercial kits for removing unwanted sequences, like ribosomal RNA or mitochondrial RNA, those kits all require multiple nanograms of RNA. "With spinal fluid, we don't have that luxury," he said, with clinical samples typically containing picograms of RNA.
Instead, the DASH protocol is implemented further downstream, after initial amplification has taken place. In the method, the researchers use RNA as starting material, but convert that to DNA for a sequencing library. After the RNA has been converted to DNA and amplified, the researchers use DASH. Wilson said that in their case, the team was obtaining a lot of unwanted mitochondrial sequences from the spinal fluid, so they designed the guide RNAs to target those sequences, and then used the Cas9 enzyme to cut them.
Another important facet of the recent study is that the researchers included sufficient controls to rule out background contamination. Including control samples is critical, DeRisi said, because although CSF is thought to be contaminant free, the researchers have found that skin microbiota often end up in the sample. To control for possible contamination, the researchers sequenced 24 water control samples and 94 CSF samples from uninfected patients. The background microbial signature consisted predominantly of bacteria commonly found in soil, skin, and environmental flora.
Wilson said that the water controls enabled cataloging the contaminants in the lab and the reagents, while the samples from uninfected patients "put together a picture of the most common contaminants from spinal fluid." The spinal fluid itself is not contaminated, Wilson added. When physicians use a needle to draw spinal fluid, they first disinfect the skin, which kills the bacteria, but dead bacterial DNA can still stick to the needle and show up in the sample that is tested.
Sequencing the controls tells the researchers both the types of bacteria that are common as well as their abundance, Wilson said. That way, when a patient sample is sequenced, the results can be compared to that of the controls, which helps to rule a lot of things out as a potential cause of the person's meningitis.
"That helps triage the data," Wilson said, "and flags the organisms that are really unexpected."
In an accompanying editorial, also published in JAMA Neurology, Kenneth Tyler of the University of Colorado's department of neurology wrote that the inclusion of those control samples was important for avoiding false-positives, and that in fact, when applied to previously published cases of proposed novel causes of meningoencephalitis, or putative identification of normal brain microbiota, the identified microbial agents would likely have been considered contaminants."
Nonetheless, he added, the study does not answer the question of whether NGS for all CNS infections "will prove to be faster, cheaper, more sensitive and specific, and less invasive, and whether it will ultimately provide better outcomes for these frequently challenging conditions.