Skip to main content
Premium Trial:

Request an Annual Quote

Johns Hopkins Team Demonstrates Feasibility of Metagenomic Sequencing to Diagnose Brain Infections


NEW YORK (GenomeWeb) – In a pilot study, researchers from Johns Hopkins University have prospectively demonstrated that next-generation sequencing of brain or spinal cord biopsies can identify pathogenic microbes in individuals suspected of having an infection-induced neurologic disorder.

The group reported the results of the metagenomic sequencing assay on 10 individuals in the journal Neurology: Neuroimmunology & Neuroinflammation. For three individuals, the assay was able to give a definitive diagnosis that was confirmed via diagnostic testing. For an additional five, the NGS approach helped the clinician understand the neuropathology of the patients' disorders, and in two patients the technique did not contribute to a diagnosis.

Steven Salzberg, a senior author of the study and professor of biomedical engineering at Johns Hopkins, told GenomeWeb that his group is now working on demonstrating that metagenomic sequencing can be used to diagnose the causative agent from other types of infections, like skin and respiratory infections. In addition, he said, the team will continue to validate the approach for neurologic conditions, which will require additional studies demonstrating the reproducibility, reliability, and precision of the assay.

Carlos Pardo-Villamizar, associate professor of neurology and pathology at Johns Hopkins and a senior author of the study, told GenomeWeb that an eventual test could be run in about 72 hours for a cost per sample in the range of $1,000.

Improving diagnostics for neurologic conditions would be particularly helpful because more than half of all inflammatory disorders of the central nervous system go undiagnosed. Having an appropriate diagnosis helps ensure that the clinician treats the patient appropriately. For instance, infection-caused neurologic disorders often present with similar symptoms as ones caused by autoimmune disorders. But the two have drastically different treatments.

Metagenomic sequencing may be a good tool to improve the diagnostic rate because it is such a broad approach and does not require "guessing" at the potential pathogen, Salzberg said. Many other researchers in the field are looking to use metagenomic sequencing for diagnostics for that purpose as well. For instance, Cincinnati Children's Hospital published a protocol for analyzing fecal samples to detect drug-resistant bacteria in order to prevent the spread of infection. And Charles Chiu's lab at the University of California, San Francisco has been working on a metagenomic sequencing assay to diagnose infections.

The Johns Hopkins sequencing assay was applied alongside standard diagnostic techniques in 10 patients who suffered from an inflammatory neurologic disorder and whose clinicians thought it could be caused by an infectious agent. The patients had a range of symptoms, including encephalitis, meningitis, seizures, and brain lesions.

Aside from standard metagenomic sequencing, the team used an analysis pipeline called Kraken, which was developed in Salzberg's lab, in order to quickly align reads from metagenomic data to specific taxonomic classes using a database consisting of 2,817 bacterial genomes, 4,383 viral genomes, and 26 single-cell pathogen genomes.

The Kraken pipeline takes just 30 minutes, Salzberg said. In addition, the researchers have since added steps to filter out bacterial species that are common contaminants, like species that are often found on human skin, he said.

In three cases, the team was confident that the assay found the causative agent, and the results were confirmed by standard diagnostic tests.

One case was a 67-year-old woman who had lung disease as well as brain and spinal lesions. The researchers obtained brain biopsies, and sequencing identified 15 reads that matched the Mycobacterium tuberculosis genome. Pathology studies identified tissue with a condition called necrotizing granuloma — essentially an area of inflammation in which tissue has died — which is known to be caused by tuberculosis, although the specific microbiologic tests did not identify the microorganism. Nevertheless, the patient responded to anti-tuberculosis treatment.

This case really demonstrated "the value of NGS as the standard microbiological approaches such as cultures and standard histopathological stains were unable to demonstrate any evidence of TB." Pardo-Villamizar said.

In another case, a 52-year-old man was experiencing partial seizures and weakness in his lower extremities. An MRI identified a brain lesion. In this case, the researchers sequenced his RNA because a viral infection was suspected. Although the assay identified reads from many different bacterial species, no one species dominated. In contrast, they obtained 8,944 reads that mapped to the JC polyomavirus, out of 8,954 total viral reads. The whole genome of the virus was covered at an average depth of over 200. The team confirmed the diagnosis via immunostaining.

The third diagnosed case was a 44-year-old woman who had previously undergone organ transplantation and immunosuppression. She had facial paralysis and brain lesions, one of which appeared to be CNS toxoplasmosis. Clinicians assessed her cerebrospinal fluid for pathogens, but that did not turn up anything. Sequencing of a CSF sample found very few reads that matched to bacteria, and the ones that it did match to were all known skin bacteria or other contaminants. Only 20 reads matched to viral genomes, but of those 20, 18 matched to Epstein Barr virus. The team confirmed that indeed the woman had that virus.

In these two cases, Pardo-Villamizar said that although standard diagnostic methods did find the answer, it was necessary to use multiple methods, which was time consuming and would have required the physician "to be suspicious about such diagnosis and the pathologist evaluating the cases to screen all spectrum of possibilities of opportunistic infections by special histopathological  and imunocytochemical techniques combined with PCR or other molecular diagnostic strategies."

In five cases, sequencing provided clinically useful results, without providing an exact diagnosis.

Salzberg said that aside from bacterial or viral infections, neurologic disorders can be caused by autoimmune problems. In that case, clinicians would treat the patient with steroids. But, because steroid treatment is "the worst thing for an infection," Salzberg said, in cases where the NGS assay is negative, it can "give clinicians who suspect an autoimmune disease additional confidence."

In this study, the NGS assay was not used to make treatment decisions, Salzberg added, but was simply run alongside standard diagnostic methods to demonstrate its feasibility. In the cases where the NGS assay was negative, the additional microbiology and pathology tests were also negative, he said, which was what the clinician used to determine how to treat the patient.

In two cases, neither the NGS assay nor any additional diagnostic tests were able to find a conclusive result. For both cases, clinicians still suspected that the patients suffered from a bacterial infection, Salzberg said, and treated them with broad-spectrum antibiotics.

In one case, although the NGS assay was indeterminate, it did contain a high proportion of reads that aligned to the bacterial genome Delftia acidovorans. Typically, Delftia acidovorans is not pathogenic, although it has been shown to be so in rare cases, usually in patients who are immunocompromised or who have become infected in hospital settings. The researchers noted that the finding could be a contaminant, as the other tests were all negative. However, the patient responded to antibiotics, leading them to suspect that it could have indeed been the cause.

"A negative result isn't definitive, but it does give you one more piece of information," Salzberg said.

In addition, Pardo-Villamizar noted that these two patients were eventually diagnosed with glial tumors.

Although the study demonstrated that the metagenomic sequencing approach was feasible and had the potential to improve on current diagnostic methods for infection-induced neurologic disorders,, which have a success rate of less than 50 percent, Salzberg said that there is still a long way to go before the test could be used clinically.

One major issue, Salzberg said, is with contamination. Using a metagenomic sequencing approach, the vast majority of the sequencing reads will be human, leaving just a small proportion containing the bacterial and viral reads that could be the source of the infection. It will be important to demonstrate that when the test identifies a microbe or virus as the causative agent, it is not simply identifying something that was picked up in the process of sample collection and running the test, he said.

Already, he said, the researchers have identified a suite of species that are known to be common on the skin and are nonpathogenic, as well as known contaminants in the sequencing reagents themselves, and filter those reads out in the analysis.

Eliminating contamination will be even more important when the group moves into other disease areas. Brain tissue, for example, tends to be relatively free of microbes compared to skin, so there is much less background noise in those samples, Salzberg said.