Proceedings of the Summer Institute on Symptoms and Omics

Metabolomics, a relatively recent field of research, offers nurse scientists a powerful new way to investigate physiologic pathways underlying health and disease and identify a biological basis for symptoms or health risks, as is consistent with nursing’s focus. Metabolomics methods are particularly useful because the metabolome is the final downstream product of gene transcription and the closest to the phenotype of the entire biological system (Harrigan & Goodacre, 2003). Another advantage of metabolomics methods, compared to other omics methods, is that changes in the metabolome are frequently amplified relative to changes at the level of the genome, transcriptome, and proteome (Harrigan & Goodacre, 2003). Despite its potential, metabolomics is currently underutilized by nurse researchers. A recent scoping literature review presented by Laura Kimble, PhD, RN, FNP-C, FAAN, Clinical Professor at Emory University School of Nursing and Nicole Carlson, PhD, CNM, Assistant Professor at Emory University School of Nursing (Kimble et al., 2020) suggests that while uptake is emerging, the current state of nursing science still reflects a scarcity of nurse scientists with an established program of metabolomics research.

The Emory School of Nursing P30 Center for the Study of Symptom Science, Metabolomics, and Multiple Chronic Conditions (P30NR018090) is dedicated to supporting the conduct of research to identify metabolites and metabolic pathways associated with symptoms of fatigue, depression, and anxiety in African Americans with multiple chronic conditions. The center is based on a novel metabolomics framework developed by one of the center’s founders, Elizabeth Corwin, PhD, FNP-BC, now Vice Dean of Strategic and Innovative Research at Columbia University School of Nursing, to study symptoms in persons with multiple chronic conditions. This theoretical framework posits that tissue injury in individuals with multiple chronic conditions may activate metabolic pathways initiating symptoms and/or clusters of symptoms, and that a variety of covariates (i.e., clinical characteristics, sex, stress) may influence this effect (Corwin et al., 2021). The model focuses on specific symptoms such as fatigue, depression, and anxiety, all of which are known to accompany chronic conditions and impact health-related quality of life (Corwin et al., 2021).

To increase exposure of nurse researchers to metabolomics methods, the Center launched its first annual Summer Institute on Symptoms and Omics (SISO), May 13–14, 2021, to describe how metabolomics methods can improve symptom science research and introduce researchers to basic metabolomics methods (www.sisoconference.com). This article presents highlights of the virtual event.

Types of Metabolomics Analyses

Dr Carlson described three metabolomics approaches: targeted, untargeted, and semi-targeted analyses. A targeted analysis is used when the researcher seeks to identify and quantify specific known metabolites. An untargeted analysis is agnostic and allows researchers to measure every metabolite, both known and unknown, within a sample. The semi-targeted approach falls somewhere between these approaches allowing for the quantification of known metabolites alongside the added discovery of similar unknown metabolites.

Dean Jones, PhD, Director of the Clinical Biomarkers Laboratory at Emory University, elaborated on the advantages and disadvantages of both targeted and untargeted analyses. Targeted methods can be used to measure individual metabolites for hypothesis testing. They can measure large numbers of metabolites with relative quantification or with absolute quantification as needed. Untargeted methods uncover features of up to a thousand confirmed metabolites, about half of which are currently unidentified. Although untargeted methods offer the opportunity, with a high degree of accuracy, to identify novel associations of metabolites or metabolic pathways with certain symptoms or conditions, targeted methods are better suited when absolute accuracy is needed and when there is a hypothetical anticipation of the involvement of certain identifiable molecules and pathways. When choosing a metabolomics approach, researchers should keep in mind the trade-off between the coverage of untargeted methods and the absolute accurate quantification of targeted methods.

Targeted Metabolomics Use in Research

Margaret Heitkemper, PhD, RN is the Chair of the Department of Biobehavioral Nursing and Health Informatics at the University of Washington School of Nursing and was the keynote speaker of the SISO conference. Dr Heitkemper described her research program focused on irritable bowel syndrome (IBS) and associated symptoms. Poor sleep quality, including frequent awakenings, fatigue, and not feeling rested in the morning are symptoms commonly reported by women with IBS. Given the evidence that serotonin, the most well-known tryptophan metabolite, is involved in pain sensitivity, gut motility, gut function, and secretion (Costedio et al., 2007) and the common thinking that tryptophan is a precursor to melatonin, Dr Heitkemper conducted a targeted metabolomics study investigating the role of tryptophan metabolites in IBS symptoms and explored the association of these metabolites to sleep measures (Burr et al., 2019).

Using nocturnal blood samples collected from both IBS and healthy control subjects exposed to a social stressor, Heitkemper’s findings suggest that with the exception of nicotinamide, tryptophan metabolites decreased across the night in both controls and women with IBS (Burr et al., 2019). She also identified that compared to the control group, nicotinamide levels were higher and indole-3-lactic acid levels lower in the IBS group (Burr et al., 2019). Because blood levels of indole-3-lactic acid can indicate bacterial metabolism, these findings prompted an examination of the contribution of the gut microbiome to peripheral levels of tryptophan metabolites in a targeted metabolomics study of fecal samples. Results of this study are pending. Heitkemper’s work illustrates the utility of targeted metabolomics to explore underlying molecular pathways in symptom science research.

Untargeted Metabolomics Use in Research

While the capability of untargeted metabolomics studies to generate hypotheses is well understood, this type of metabolomics research also has the potential for future clinical applications. Jones described colleague and Assistant Professor of Emory University School of Medicine, Miriam Vos’ study of pediatric non-alcoholic fatty liver disease (NAFLD) (Khusial et al., 2019). Traditional diagnostic procedures for NAFLD are expensive or invasive, so Vos’s group explored an alternative approach using untargeted metabolomics. Vos and colleagues identified 11 different metabolites and metabolic features, most of which were confirmed as known metabolites. This panel of metabolites shows promise for detecting NAFLD, particularly when combined with clinical phenotyping data, suggesting that developing a low-cost minimally invasive evaluation of NAFLD in youth may be possible.

One of the exciting characteristics of untargeted metabolomic studies is that it is possible to discover features of a chemical with potential importance before knowing what the actual chemical is. As an example, Jones described how Senior Bioinformatics Analyst, Ken Liu, PhD, studied the metabolic differences between conventional and germ-free mice. The top two metabolic features found in the liver of the conventional mice reflected a compound (C₁₈H₁₇NO₂) with Carbon-13. This metabolite was previously unknown. It has since been discovered to be valerobetaine. Liu confirmed that valerobetaine is produced by the microbiota of the gut in conventional mice. The administration of valerobetaine to conventional mice inhibited carnitine in the liver and serum and decreased fatty-acid dependent mitochondrial respiration. Liu further found that valerobetaine is a diet-dependent obesogen. Mice on a standard diet, administered valerobetaine, do not gain weight, but those on a Western diet do gain substantial amounts of weight. Liu went on to look at valerobetaine levels in humans and found that valerobetaine levels were significantly higher in individuals with BMI greater than 30. Liu’s work is currently under review. These brief study descriptions illustrate the potential of untargeted metabolomics for discovery and, with additional testing and validation, clinical use.

How to Conduct a Metabolomics Study

Since the field of metabolomics is just starting to emerge in nursing research applications, many nurse researchers are unfamiliar with the methodology of metabolomics studies. What follows is a methodology primer.

Study Design

According to Emory University School of Nursing Professor, Vicki Hertzberg, PhD, the first thing researchers need to do before planning a study is to establish a specific, clear hypothesis or research question that the study will attempt to answer. The hypothesis should drive the study design, the standard operating procedures, and the metabolomic approach to be used.

One of the first study design questions that should be considered is: how many participants should be included in a metabolomics study? Carolyn Accardi, PhD, Program Director, Clinical Metabolomics in Emory Clinical Biomarkers Laboratory, stated that inherent biologic variability in the metabolome will often drive up the number of study subjects needed. Often studies utilizing untargeted metabolomics platforms plan to include 40 or more subjects in a study design to adequately compare groups. For targeted approaches, power analyses for quantified metabolites of interest will optimally determine sample size. Hertzberg added further considerations that may influence sample size such as whether randomization will be incorporated, or whether stratification will be required for factors that are unable to be controlled. Early and continuous input from biostatistical/data science collaborators is critical for study conceptualization, logistics, and data analysis.

Standard Operating Procedures to Minimize Variability

Hertzberg also stressed that along with these decisions, researchers should establish standard operating procedures (SOPs). This is especially important if researchers are conducting multi-center investigations. SOPs are also useful for when modifications need to be made. For instance, Laren Narapareddy, PhD, RN and Brittany Butts, PhD, RN, both Assistant Professors at Emory University School of Nursing, conducted studies during the COVID-19 pandemic in 2020. They developed an SOP that included recruiting subjects remotely and asking them to collect their own samples under carefully crafted protocols. A description of their remote protocol is currently under review (Butts et al.). The goal of standardizing procedures is to make sure that they are followed consistently from the first participant to the last participant.

As Accardi described, carefully and thoughtfully developed SOPs ensure that biological and analytical variability will be minimized and that study results are as robust as possible. Biologic variability is the variation from one individual study subject to another, for example, differences among individuals. The human metabolome is highly responsive to environmental conditions. In addition to metabolome changes due to diurnal variation, the metabolome is influenced by a range of factors such as time since the subject’s last meal, physical activity, and the season of the year. Study designs can aim to minimize effects of such variability by collecting samples at consistent times of the day to address diurnal variations or by collecting samples based on the known fasting or postprandial state of the subject, as well as controlling sample collection in relation to the time since the subject’s last meal. Other factors influencing the metabolome include exercise, stress, diet, medications, lifestyle factors, and geography, and these should be considered when designing study protocols and inclusion and exclusion criteria to help minimize variation.

Analytical variability concerns collection and chemical analytical methods. There are three stages of sample collection: extracting and obtaining the sample, storing the sample (samples should be stored at −80°C as soon as possible), and preparing the sample for analysis. Timing and specimen collection protocols and SOPs should be kept as consistent as possible from one sample collection to the next. Mishandlings or inconsistencies at any stage of sample collection can introduce significant error and variability. While there are often strict protocols in place to prevent error and minimize variability, the success of these protocols depends on how rigorously they are followed. Any deviations in collection procedures should be documented to aid in the interpretation of results. Analytic methods using liquid-chromatography mass spectrometry (LC-MS) instrumentation should follow established laboratory SOPs for the specific sample type.

Sample Types

According to Accardi, metabolomics analysis can be applied to a variety of different specimen types. Many metabolomics studies focus on the analysis of biofluids, particularly blood and plasma as these are thought to represent systemic measures of the human metabolome. Urine, saliva, and dried blood spots are also suitable for metabolomics assays and may be preferable in some settings because they are obtained by less invasive collection methods. Research questions may target the collection of other specific biospecimens such as cerebral spinal fluid, breast milk, synovial fluid, interstitial fluid, or exhaled breath condensate. Protocol development must include early communication with the lab for sample volume or weight requirements needed for the chosen matrix.

Solid tissue samples may also be collected and prepared from a range of human specimens, such as from colon, small intestine, brain, liver, pancreas, adrenal glands, parathyroid gland, and placenta obtained by human biopsies, as well as a larger range from experimental animal models. Cultured cell specimens for experimental metabolomics research can explore whole cell metabolomics or cell media and cellular extract metabolome changes. These can also enable more detailed cellular metabolome studies focused on mitochondrial metabolome or nuclear or lysosomal extracts.

Each category of sample can break down into smaller sub-categories or require choices to be made regarding collection method. For instance, if researchers collect blood, should they store whole blood, plasma, or serum? If they collect urine, would spot urine, first morning urine, or 24-hour urine best serve their research purpose? Should saliva be collected by passive drool or swab? What considerations go into whether to collect stool or rectal swab? Sample considerations should be informed by the original hypothesis and the study design that was planned around it as well as associated logistical advantages and challenges.

Analysis of Samples

Assistant Professor of Environmental Health at Emory’s Rollins School of Public Health, Donghai Liang, MPH, PhD detailed an untargeted metabolomics analysis workflow (Liang et al., 2018, 2019). Samples are first analyzed using liquid chromatography-mass spectrometry (LC-MS), gas chromatography-mass spectrometry (GC-MS), or nuclear magnetic resonance (NMR) spectroscopy. The raw data generated are then analyzed by an algorithm that extracts metabolic signals to create a feature table containing the key information of tens of thousands of metabolic features. The feature table is composed of three components: mass to charge ratio, retention time, and the intensity of the feature signal, which may serve as a surrogate for the relative concentration of the metabolite in the biospecimen.

Next, a metabolome wide association study (MWAS) should be conducted to identify metabolic signals associated with the research outcome of interest. There are numerous statistical approaches and bioinformatic tools available for this step. An experienced biostatistician is critical at this point to conduct rigorous, robust, and reproducible analyses that control for confounding factors, covariates, random effects (in the case of repeated measures) and adjust for the false positive discovery rate.

At this point, the analysis has likely yielded many more unidentified metabolites than identified metabolites. Pathway analysis and network module analysis predict the biological function of the unknown-but-significant features, identifying those that are enriched on pathways relevant to the outcome of interest. Features of interest can then be matched to authentic reference standards to confirm their chemical identity. Finally, the confirmed metabolites are examined for any dose–response relationship with the outcome of interest, and underlying molecular mechanisms are further explored by looking in literature to see how these metabolites might contribute to the outcome of interest (Li et al., 2021).

Conclusion

Nurse researchers are uniquely poised to incorporate omic methods like metabolomics into their investigations of symptoms and intervention development. Our keynote and plenary conference speakers, Drs Heitkemper and Jones, illustrated the power and flexibility of metabolomics to provide deeper insight into normal physiological processes and altered metabolism related to disease, symptoms, environment, and behavior, ultimately leading to the discovery of biomarkers to improve diagnostics and disease management.

As Drs Hertzberg, Accardi, and Liang’s presentation highlighted, metabolomics methods require rigorous protocols and strong data analytic support. Nurse scientists interested in metabolomics research should be aware of the cost of analysis, as well as infrastructure and interdisciplinary collaboration required to conduct rigorous metabolomics research.

Kimble and Carlson had additional suggestions to facilitate the uptake of metabolomics among nurse researchers. Schools of nursing should consider incorporating metabolomics into their curricula by training interested faculty and students, facilitating team science, and building infrastructure. Training should include courses, workshops, and bootcamps for doctoral students, postdocs and faculty on omics science fundamentals, methods, and advanced statistical programing with R or Python. Promoting the practice of interdisciplinary and collaborative team science gives nurse researchers the opportunity to develop productive working relationships with professionals ranging from molecular biologists to biochemists and data scientists. Finally, for schools of nursing to successfully integrate metabolomics into their programs of study, they will need to build basic infrastructure like lab space to process specimens, and freezers to store specimens, as well as develop relationships with institutional cores and/or commercial facilities that can conduct metabolomics analyses.

ORCID iD

Irene Yang, PhD, RNhttps://orcid.org/0000-0001-7873-0212