NEW YORK – A team of researchers at Stanford University is fine-tuning a technique to diagnose respiratory infections by measuring metabolic signatures via mass spectrometry.
In a presentation at the Next Generation Dx Summit on Wednesday, Benjamin Pinsky, an associate professor in the department of pathology at Stanford's School of Medicine, laid out the workflow for the method, which was also described in a paper published last week in EBioMedicine.
According to Pinsky, viruses can induce metabolite alterations in human cells that can be detected using liquid chromatography mass spectrometry, or LC-MS. Those alterations, resulting in metabolic signatures from the host, may allow for the diagnosis of infections, in a way analogous to transcriptomic signatures.
The compound signatures "can represent what is happening with the cell and whether or not the cell may be infected," he added.
According to the study, metabolomics "represents a change in paradigm from the routine clinical virology diagnostics as it detects host metabolic response rather than directly detecting the pathogen."
The study looked at two cohorts — one biomarker discovery cohort and one validation cohort — to assess the test performance for the diagnosis of influenza and "to identify key metabolites for classification" of influenza-infected and uninfected people, the researchers wrote.
The overall workflow of the method for the discovery cohort involved filtering the primary specimen — in this case the viral transport media from a nasopharyngeal swab for influenza — and adding a solvent.
The compounds in the solution were then separated using liquid chromatography/quadrupole time-of-flight, or LC/Q-TOF, mass spectrometry to determine features of the metabolites, such as retention time and mass-to-charge ratio. Peak intensity, which correlates with the concentration of the analyte, is also measured, Pinsky said.
After liquid chromatography separates the metabolites from the solution in the samples, electrospray ionization charges the metabolites. A mass spec analyzer is then able to detect that charged metabolite, Pinsky said.
Metabolites for evaluation can be charged or uncharged, and laboratorians generally need to use different types of columns and separate specimen preparation procedures to separate out the charged and uncharged molecules.
An in-line two-column method, developed by fellow Stanford researchers and described in a paper published in the Journal of Chromatography last year, allows those metabolites to be separated without having to perform those procedures separately.
Though the in-line method "seems simple," Pinsky said, "it is quite an innovation." The in-line method "would reveal distinct signatures for the diagnosis of infectious disease," the researchers wrote.
With the discovery cohort, the team performed untargeted analysis via LC/Q-TOF to determine the compounds that would be used for detection. The metabolites were identified by comparing what's found in the cell to reference metabolite databases, Pinsky said in his presentation.
The data generated from the LC/Q-TOF testing was also used to develop and validate machine-learning algorithms that could classify influenza infection status and create an interpretation method for biomarker discovery. The method was developed to "differentiate between infected and not infected" people, Pinsky said.
The metabolic profile of a sample included the mass-to-charge ratio, retention time, and relative abundance, and that profile was matched to the influenza status of the samples to teach the machine which features correlated to infection, the researchers noted.
The top 20 features correlating to infection were determined based on which ones recorded the highest performance with the machine-learning model. The top two ion features were pyroglutamic acid and an in-source fragment ion of pyroglutamic acid, which are lower in specimens from people with influenza, the researchers found.
Many of these features overlap and multiple components are needed to generate the signature, Pinsky said. That signature, which consisted of those same top 20 ion features, was validated via tandem mass spectrometry testing on upper respiratory tract specimens.
In their research, the Agilent 6460 Triple Quadrupole mass spec was used in the validation cohort to program the limited signature and "see if that limited signature actually work[ed] as a discriminator" between positive and negative people, Pinsky said.
Both sensitivity and specificity for the LC/Q-TOF method determining the difference between influenza-positive and influenza-negative patients was above 90 percent, the researchers found. In the validation cohort, tandem mass spec testing "demonstrated sustained high performance," the researchers found.
Although for the paper the researchers studied influenza, they couldn't distinguish between influenza A or B, Pinsky said. In the retrospective study, they found a high accuracy for the diagnosis of acute infection with an amino acid signature of 20 targets and were able to differentiate acute infections from other viral infections and non-infected people with elevated inflammatory markers, he added.
LC/Q-TOF testing is complex and inaccessible to many labs, which is why the researchers chose to test the validation cohort using tandem mass spectrometry — a more affordable and common instrument. Because the performance was still high on the tandem mass spec instrument, labs may be more willing to apply the method, "enabling cost-effective testing," the researchers wrote.
The potential advantages of the method are that it is performed directly from specimens, requires a limited sample volume with reagents that are "almost free," and has a rapid turnaround time of two to three minutes, Pinsky said.
It could also be adapted to point-of-care settings, he said. While mass spectrometers are unwieldy and generally for use in clinical laboratories, "mini" mass spec instruments, such as those used by the Transportation Security Administration to detect trace amounts of explosive residue at airports, can be deployed to perform this testing.
"Although they're looking for quite different compounds than one that might be found in a human cell, [there are] mass spectrometers that are just as small … and that can be used to detect the types of compounds we're looking at here in these experiments," Pinsky said.
However, "further work will be required to determine the optimal number of biomarkers" for point-of-care testing, the researchers added.
As for limitations, the team noted that the study was only performed at one institution, so results might not be generalizable to other patient populations. In addition, only influenza samples were compared in the discovery cohort, so the team couldn't extrapolate the method to other respiratory viruses or bacterial or viral coinfections.
The study also didn't look at the respiratory metabolic profiles of healthy people as negative controls, which could help isolate the metabolites that change in response to acute infection.
Next steps include using the method in a prospective study and independently validating the signature, Pinsky said.
While the team wants to develop a biomarker signature for SARS-CoV-2 and is working on it, Pinsky said he doesn't have final data for this use yet. However, he added that the team has tested 275 nasopharyngeal samples with LC/Q-TOF and 100 samples for confirmation via LC-MS.
Pinsky said applying a similar metabolic approach to SARS-CoV-2 will be difficult. "It's certainly not entirely clear whether this is going to work," he said, and the technique may get tripped up distinguishing between that virus and seasonal coronaviruses or distinguishing between influenza and SARS-CoV-2.
The methods used in the study have patents either granted or pending for Stanford, Pinsky noted.