Over the past decade, omics has revolutionized diagnostic microbiology, with genomic applications in diagnostics and epidemiology, and proteomic applications of Matrix-Assisted Laser Desorption/Ionization Time of Flight Mass Spectrometry (MALDI-TOF) to identify bacteria and fungi. The development of diagnostic applications of a third “omic,” metabolomics, has lagged behind the other 2. However, recent studies suggest that there may be an important diagnostic role for metabolomics as well.

Principle of Metabolomics

When living organisms metabolize organic compounds, hundreds to thousands of chemically complex products/by-products called 'volatile organic compounds' (VOCs) are produced. VOCs have high vapor pressure at room temperature and can be detected as " odors," "smells" or "aromas," which can be subsequently used for identification purposes when a sufficiently sensitive method of analysis is applied. Using metabolomic techniques, VOCs can be detected from a wide variety of sample types, including isolates of pathogens, expired breath, sputum, feces, saliva, nasopharyngeal and throat swabs and vaginal discharges.

Metabolomic analysis relies on sensors for detecting VOCs and a processor to identify them. Following identification, data analysis searches for unique VOC patterns associated with specific disease states or organism identification. A goal of data analysis is to detect specific VOCs or biomarkers that would be associated with specific disease states or pathogens. In the absence of specific biomarkers, VOC patterns from clinical specimens or isolates are compared to data libraries consisting of known VOC patterns found in specific disease states or clinical isolates. This principle is similar to that used to identify bacteria and fungi using MALDI-TOF MS.

Using Odors to Detect Pathogens and Infectious Diseases

Although the term “metabolomics” may be novel to many diagnostic microbiologists, using odors to detect infectious diseases or identify microorganisms is as old as medicine and diagnostic microbiology themselves. The human nose naturally functions as the “sensor,” the olfactory nerves, the “processor,” and the brain performs pattern recognition as the “identifier” of specific infectious diseases or organisms. Since antiquity, putrefaction and abscesses have been associated with foul odor, which is now known to be caused by specific odiferous metabolites, such as cadaverine and putrescine. Few medical students will forget the fetid breath of the first patient they encounter with an anaerobic lung abscess.

After overcoming the initial queasiness upon encountering the ”special smell” of a clinical microbiology laboratory, medical laboratory science students soon learn to recognize the smells of the frequent visitors: the grape-like odor of Pseudomonas aeruginosa, the caramel sweet odor of Streptococcus anginosus, the earthy odor of higher order bacterial genera (such as Nocardia and Mycobacterium) and the sweet odor of Candida. Individuals processing specimens from patients with Clostridioides difficile infections may notice a “horse barn” odor, and VOC analysis suggests high levels of the metabolite 4-methylphenol may contribute to this rather unpleasant smell. A positive “whiff test” is 1 of 4 Amsel criteria for the diagnosis of bacterial vaginosis. The whiff test is positive if a “fishy” odor is present due to the release of amines when 10% KOH is added on a slide of the patient’s vaginal discharge.

Animals’ Roles in Pathogen Detection

Animals, such as dogs and rats, have much more sensitive olfactory systems than humans. Dogs are used to track people and detect drugs, explosives and weapons. A recent study even used dogs to detect C. difficile in a hospital environment and demonstrated a sensitivity of detection of 92%. Dogs have also been used to detect bacteriuria in urine specimens with a sensitivity of screening close to 100% and specificity ranging between 90 to 95% for the 5 study dogs. Using dogs to detect bacteruria may not be practical, but employing them as comfort animals that are also capable of detecting C. difficile in hospital environments is an interesting idea.

African giant pouch rats have been trained to detect Mycobacterium tuberculosis on sputum slides from individuals suspected to have tuberculosis. In a large study with specimens obtained from over 10,000 persons and a smear positivity rate of 13%, rats detected 95% of the positive samples. In addition, the rats detected M. tuberculosis in 1,335 additional patients who were smear negative on initial screen. When smears from these 1,335 patients were re-examined microscopically, 620 were found to be positive. The rat screening increased detection by 44%, with an overall specificity of 90%. It would be interesting to compare rat screening with the more sensitive PCR method for detection of M. tuberculosis. Because rats are more cost-effective and do not require complex testing infrastructure or equipment, this diagnostic alternative is particularly attractive for low-income settings.

Metabolomic Studies Using Specific Biomarkers

Four studies reflect the promise and problems with using metabolomics as a diagnostic tool. All 4 studies used a combination of either gas (GC) or liquid chromatography (LC)-mass spectrometry (MS). These types of instrumentation have large footprints and require operators with advanced technical skills. Furthermore, specimen collection, transport and quality challenges limit current use to research settings only.

Metabolomics to Diagnose Pseudomonas aeruginosa

The initial study of interest provides proof of principle. This study used 2-aminoacetophenone as a biomarker for detection of Pseudomonas aeruginosa in the breath of individuals with cystic fibrosis. Using LC-MS, scientists found that 15/16 CF individuals with P. aeruginosa were positive for 2-aminoacetophenone, while only 5/17 healthy controls and 4/13 P. aeruginosa negative CF individuals were positive. Furthermore, 2-aminoacetophenone breath concentrations were found to be much higher in individuals with P. aeruginosa, indicating that quantitative 2-aminoacetophenone levels could be used to differentiate P. aeruginosa-positive from negative samples. Unfortunately, quantitation of analytes adds significant complexity to such assays, and costs versus benefits are important considerations in the implementation of any assay.

Subsequent studies found consumption of corn chips resulted in high breath levels of 2-aminoacetophenone soon after consumption, which disappeared after 2 hours. This is not surprising, since some individuals describe the grape-like odor of P. aeruginosa as smelling like corn chips. Ensuring that patients have not eaten or drank in the 2 hours preceding testing has become standard practice for breath test analyses. Because P. aeruginosa is reasonably easy to recover and identify from the respiratory tract, the use of metabolomics for P. aeruginosa detection has not been further developed for diagnostic use.

Metabolomics to Diagnose Invasive Pulmonary Aspergillosis

Invasive pulmonary aspergillosis (IPA) is a particularly difficult diagnostic problem. Because Aspergillus spp. are ubiquitous in the environment, the upper airways can be colonized and give false positive culture results. More importantly, false negative sputum and bronchial alveolar lavage cultures may occur because the organism tends to invade endothelial cells. The optimal diagnostic approach requires invasive examination of lung tissue, either histologically or by culture. Unfortunately, the patient population most likely to develop IPA are patients with malignancy or bone marrow transplants: two populations that are more likely to be thrombocytopenic, which precludes them from lung tissue biopsies.

A recent study in patients with IPA (n=34) and an appropriate negative control group (n=30) used breath analysis by GC-MS to demonstrate that the presence of any metabolite from a 4-metabolite signature consisting of α- or β-trans-bergamotene, trans-geranylacetone or β-vatirenene had a 94% sensitivity and 93% specificity for IPA detection. The advantage of this technique is the ease of specimen acquisition. However, standardization of specimen collection and reproducibility of the assay are just a few of the unanswered questions about its utility.

Metabolomics to Diagnose Bacterial Vaginosis

A healthy microbiome with a lactobacilli pre-dominance has long been associated with vaginal health. Bacterial vaginosis (BV) is an important clinical example of dysbiosis. BV dysbiosis occurs when lactobacilli predominance is disrupted and replaced by a variety of anaerobic bacteria. BV has been associated with increased likelihood of transmission and acquisition of both HIV and other sexually transmitted infections, as well as increased risk of premature labor. Diagnostic tools for BV diagnosis currently include either a gram stain scoring system (Nugent score), which is operator dependent and associated with significant error, plus a “whiff” test (Amsel criteria as described above).

Metabolomic analysis using GC-MS analysis of vaginal swabs showed that specimens with high ratios of 2-hydroxyisovalerate: tyrosine and γ-hydroxybutyrate: tyrosine ratios were able to predict BV with 91% accuracy, suggesting that these 2 metabolites could be used as biomarkers for BV. However, because the Nugent scoring system, which is highly operator dependent, was used as a reference method in the study, these data should be viewed cautiously, and further validation is needed.

Metabolomics to Diagnose Influenza

Perhaps one of the most exciting metabolomic applications is the diagnosis of influenza from nasopharyngeal swabs. Decreased levels of the biomarker pyroglutamate in the nasopharyngeal specimens of influenza virus-infected patients were reliably detected using LC-MS and machine learning. The authors of the study speculate that changes in glutathione metabolism in influenza virus-infected cells was responsible for the decreased levels of pyroglutamate, which is a by-product of glutathione degradation. It is important to recognize that there are many unanswered questions about this application, and rigorous clinical trials still are required; however, the potential of this application appears promising.

Metabolomic Studies Using the “Electronic Nose”

The electronic nose (E-nose) is a hand-held device using specific biosensor arrays that react with VOCs present in odors to produce different electronic signal patterns. These patterns are compared to preexisting pattern libraries in order to identify specific substances. The patterns range from simple to highly complex. The technique, which was first used in 1982, has a wide array of applications in the food industry, as well as detection of different chemical exposures and environmental hazards, but diagnostic microbiology applicationshave been more limited. E-nose infectious disease detection studies tend to have small numbers of subjects and are not designed to detect specific organisms. Furthermore, although the studies cited here all used expired breath as their sample type, significant variation in sample collection and patient populations exists amongst electronic nose applications, making comparison between studies difficult to impossible.

Since electric noses have been studied for close to 40 years with little data of relevance to diagnostic microbiology, why the interest in this device? With the increasing desire for “near patient” testing, a hand-held device that could detect biomarkers for different infectious agents that bring patients to emergency departments and urgent care facilities would be of great interest. The advantage of a handheld device is portability, allowing it to be used at the bedside. The ability to test easy-to-collect specimens, such as expired breath, nasopharyngeal or throat swabs or urine reduces invasiveness and risk to the patient. Finally, rapid test turn-around time decreases the time between patient arrival and administration of a targeted and effective treatment.

The Future of Metabolomics

Many challenges and questions remain in the field of metabolomics. Can devices be developed with biosensor arrays for detection of specific infectious diseases and/or pathogens? If so, when will pattern libraries and/or biomarkers for specific pathogens be identified and made available? Is metabolomics just too far behind the infectious disease diagnostic applications of genomics and proteomics to be largely applicable for the diagnosis of infectious diseases? Time will tell, as long as curious scientists are willing to take a chance to investigate the answers.