Mass Spectrometry Reveals Respiratory Viral Infection Biomarkers.

Acute respiratory infections are the commonest disease in humans; they are most often caused by viruses and, to a smaller extent, by bacteria (Denny, 1995). Distinguishing these causes is important because viral infections are usually self-limiting and generally do not require any specific treatment while bacterial infections can often lead to complications and are treated with anti-bacterial medications (Dowell et al., 1998, Gozales et al., 2001). However, ruling out a viral etiology is not trivial because of complexities of diagnostic tests and because of a poor correlation between detection of virus and symptoms even with a positive test. This is because the virus could colonize the body but not lead to an infection. One approach to identifying the infection-causing pathogen is to detect the host-response to that pathogen.

Gene expression changes in the peripheral blood can be used to detect responses elicited in the host by the pathogen. Indeed, host gene expression profiles based on microarray profiling studies that can distinguish between bacterial and viral infections in patients with acute respiratory symptoms have already been developed (Tsalik et al., 2016). However, such RNA-based detection methods have a number of limitations including cost, time for processing and complexity. To circumvent some of these limitations, Burke et al. came up with an innovative approach using mass spectrometry-based proteomics technology to detect proteins that are involved in host-response to viral infection from nasopharyngeal lavage (NPL) (Burke et al., 2017). To discover such biomarker proteins in NPL, they designed an experiment where they recruited healthy volunteers and infected them with human rhinovirus, H3N2 subtype of influenza A virus or sham. Using global mass spectrometry-based proteomics, they first identified candidate proteins that were differentially expressed in the nasal cavity upon viral infection. Next, they designed multiple reaction monitoring assays to validate a set of candidate proteins in a high-throughput and targeted fashion on a larger number of samples. Finally, they derived a classifier using a machine learning algorithm to distinguish infected from uninfected individuals to obtain a 75% true positive rate and 97.46% true negative rate.

This study is important for several reasons: First, NPL can be used as a specimen for diagnosing acute respiratory viral infections instead of more invasive samples such as venipuncture. Second, detection of proteins as biomarkers reported in this study is more attractive than RNA signatures. Third, the success of mass spectrometry-based approach used for the discovery and validation steps shows the feasibility of this approach for other disorders as well. Last, a large cohort was used and a classifier was able to select the best biomarkers for prediction.

This report illustrates the power of applying technologies such as mass spectrometry-based proteomics to address clinical needs such as accurate diagnosis of acute respiratory viral infections. Although there are numerous studies that catalog candidate biomarkers discovered using mass spectrometry (Raghothama and Pandey, 2008), very few describe their rigorous testing in patients for clinical use. This paper shows how collaborative efforts between statisticians, basic scientists and clinicians can result in biomarker discovery, which could someday be applicable to clinical practice. Importantly, this approach can easily be extended to other diseases. It should be noted that this study was done only for two types of viruses and it was not investigated if the biomarkers discovered in this study can be used to test NPL samples from subjects infected by other viruses or bacteria. Thus, there is a possibility that the false positive rate will turn out to be higher when it comes to real clinical samples, which necessitates further testing.

Regardless, as such investigations become commonplace in pursuit of more precise and individualized medicine, we are apt to discover biomarkers that will enable us to define patient subpopulations more accurately. Although biomarker discovery has traditionally been focused on blood samples, the availability of newer and more robust detection platforms will allow more and more non-invasive methods of testing for diagnosing medical conditions in the future.

The authors declare no conflicts of interest.