Applications of Metabolomics to Precision Nutrition Open Access

Optimal nutrition promotes prolonged good health, while poor nutrition contributes to chronic diseases such as cardiovascular disease, diabetes, obesity, or diseases of undernutrition such as maramus and kwashiorkor. Many developed countries have implemented nationwide mandatory mineral or vitamin supplementation along with population-wide recommended dietary allowances (RDAs) of certain foods (fruits and vegetables, dairy, and meat). These RDAs are set at a level that meets the needs for essential vitamins, minerals, amino acids, and fatty acids for maintaining good health in 97.5% of a specific population. Mandated micronutrient supplementation and nationwide RDAs have helped to significantly improve population-wide health in many countries. While RDAs are set as a goal for usual intake at the individual level, not everyone benefits from these RDAs in the same way. Indeed, there can be significantly different interindividual (cross-sectional differences between individuals) or intraindividual (longitudinal differences within a single individual) variations or responses to the same foods or the same nutrients [1]. These dietary responses are affected by age, genetics, gut microbiota, metabolism, physiology, physical activity, and lifestyle [2]. It is this realization of nutritional “individualism” that has led to the push toward precision nutrition.

Precision nutrition is a branch of nutrition science that aims to use genomics, proteomics, and/or metabolomics to assess an individual’s response to specific foods or dietary patterns. By assessing these responses at a comprehensive, molecular level, it is possible to determine the most effective diet or lifestyle interventions to improve health or prevent and even treat specific diseases [3]. A key focus of both precision nutrition and precision medicine is on proactive intervention and prevention rather than reactive salvaging or saving. Catching a problem early and working to fix it usually prevents long-term or irreparable damage. Given its potential impact, the National Institutes of Health has promoted precision nutrition as the best strategy to develop clinically relevant and meaningful diet solutions for individuals and populations that share common physiological, behavioral, or sociocultural characteristics [4].

The emergence of precision nutrition has largely been tied to the emergence of sufficiently fast, cheap, and comprehensive genomics, proteomics, and metabolomics technologies. It is through these technologies that the necessary molecular measurements to precisely assess food composition, identify individual genetic propensities, or evaluate individual responses to diet and lifestyle interventions have become available. While most reviews on precision nutrition have focused on the use of genomics or transcriptomics technologies to advance the field [5, 6], in this review we shall focus on the role of metabolomics in precision nutrition. As we will show, metabolomics is vital to nearly every aspect of precision nutrition. Metabolomics is a field of omics science that uses cutting-edge analytical chemistry techniques and advanced computational methods to characterize complex biochemical mixtures [7]. It can be used to comprehensively characterize the thousands of chemicals in foods [8], to identify food byproducts in human biofluids or tissues [9], to characterize nutrient deficiencies or excesses [10, 11], to monitor biochemical responses to dietary interventions [12], to track long-term or short-term dietary habits [13], and to guide the development of nutritional therapies [14].

In this review, we will briefly introduce metabolomics and describe how it has become a key technology for nutritional science. We will then show how metabolomics has been used to advance the field of precision nutrition by providing some notable examples or case studies. First, we will describe how metabolomics played a key role in launching the field of precision nutrition through the diagnosis and dietary therapy of individuals with inborn errors of metabolism (IEMs). Next, we will describe how metabolomics can be used to comprehensively characterize the full chemical complexity of many key foods and how this is adding new meaning to the concept of molecular nutrition. Third, we will describe how metabolomics is being used to identify biomarkers of food intake (BFIs), and how this could open the door to a more objective and quantitative assessments of an individual’s diet and their response to dietary changes. Finally, we will describe how metabolomics is being coupled with other omics technologies to develop custom diets and lifestyle interventions that are leading to positive health benefits for several chronic diseases. This review will conclude with a brief discussion regarding the future of metabolomics in precision nutrition.

Metabolomics is the science of studying the metabolome. It is a branch of analytical chemistry that focuses on characterizing small molecules (molecular weights of <1,500 Da) found in tissues, cells, or biofluids. Unlike genomics, transcriptomics, or proteomics, in which a single instrument is often sufficient to perform the necessary measurements, metabolomics requires a broad array of instrumentation. Over the past 15 years, 3 main technologies have emerged as the primary workhorses in metabolomics: nuclear magnetic resonance (NMR) spectroscopy [15], gas chromatography-mass spectrometry (GC-MS) [16], and liquid chromatography MS (LC-MS) [17, 18]. Each technique provides broad coverage of many classes of organic compounds, including lipids, amino acids, sugars, biogenic amines, and organic acids. As a general rule, NMR is best at identifying and quantifying high abundance metabolites, while GC-MS and LC-MS are best at detecting lower abundance metabolites (shown in Fig. 1). Although these technologies each have their own advantages and disadvantages, numerous studies have shown how they may be used to complement each other [17, 19]. Indeed, the use of multiple technologies greatly broadens the level of metabolite coverage that can be achieved and the types of samples that can be studied.

Metabolomic approaches can either be targeted or untargeted. With targeted metabolomics, selected metabolites are identified and quantified by comparison to known chemical standards to help develop biomarkers or test hypotheses. With untargeted metabolomics, there is less focus on compound identification and quantification but more emphasis on novel compound discovery and discovery-based research [19]. Both untargeted and targeted metabolomics have their advantages and disadvantages, but given the importance attached to bioactive compound identification and quantification, there is a growing preference for quantitative metabolomics in many areas of food science and nutrition research. Indeed, targeted, quantitative metabolomics is now widely used in food composition analysis [8, 20, 21], the identification of BFIs [13, 22], the detection and monitoring of nutritional deficiencies or metabolic/nutrient disorders [10, 11, 23], the estimation of dietary intake [24-26], and the development of dietary recommendations for wellness and chronic disease prevention [14].

Perhaps, the most successful example of precision nutrition, to date, has been the application of targeted metabolomic methods toward newborn screening. Nearly every newborn child in the developed world is subject to a dried blood spot test that uses MS. These MS-based metabolomic tests can measure up to 40 different metabolites, including amino acids, organic acids, and acylcarnitines [27]. These metabolomic assays permit the diagnosis of up to 30 different metabolic disorders or IEMs [28]. Newborn screening is far and away the most successful example of any omics technology making its way to the clinic. Based on national testing data reported by many countries, we estimate that over the past 25 years >400 million children have been subject to newborn screening and up to 1 million have had their lives improved through early detection and intervention. In many cases, IEMs detected by MS-based newborn screening can be treated through customized dietary interventions that either supplement or eliminate a certain nutrient that causes the health effects arising from the IEM. Not only does MS-based newborn screening help in diagnosing or even predicting disease but the same metabolomic techniques can also be used to determine the optimal nutritional therapy, to monitor each individual’s response, and to customize the composition of the medical food to suit the patient.

There are a number of examples with metabolomics and metabolite-guided dietary therapy succeeding in improving the lives of IEM patients. For instance, precision nutrition has long been used to treat patients with maple syrup urine disease (MSUD). MSUD is an autosomal recessive IEM that is usually first detected via MS-based newborn screening for high blood levels of branched-chain amino acids (BCAAs). MSUD affects an individual’s ability to break down BCAAs such as leucine, isoleucine, and valine [29]. As a result, these BCAAs and their toxic byproducts (α-keto acids) buildup in the body. If untreated, MSUD can cause convulsions and permanent brain damage [30]. To prevent these BCAAs and keto acids from accumulating, MSUD patients are given specially developed medical foods that have been formulated to be free of leucine, isoleucine, and valine. These custom foods or beverage mixes contain all the amino acids (except the BCAAs) and appropriate protein content, vitamins, minerals, omega 3-polyunsaturated fatty acids, carbohydrates, and other fats needed for proper growth and development of affected individuals. Metabolomics is not only used to diagnose MSUD and guide the nutrient intervention but it is also used throughout an MSUD patient’s life as each patient must test their urine daily for the buildup of keto acids using an at-home metabolomic colorimetric test, such as the dinitrophenyhydrazine assay, a nonquantitative screening test for keto acids [29].

Precision nutrition, guided by metabolomics, is also used to treat individuals with phenylketonuria (PKU). PKU is an autosomal recessive IEM that is usually first detected via MS-based newborn screening for high blood levels of phenylalanine. PKU affects a patient’s ability to break down the amino acid phenylalanine [31]. The excess phenylalanine is converted to toxic metabolites (phenylketones) which accumulate in the blood and brain. If untreated, PKU can cause irreversible brain damage, seizures, and behavioral, emotional, and social problems [32]. To prevent the buildup of phenylketones, PKU patients must adhere to medical diets that are low in phenylalanine. Furthermore, they must avoid foods containing the artificial sweetener aspartame (which is converted to phenylalanine upon digestion). As those with PKU will not receive all the essential nutrients from their restricted diet, they must drink age-specific medical beverages free of phenylalanine that includes all other amino acids, vitamins, minerals, carbohydrates, and fats to meet all their energy requirements and to support proper growth and development of affected individuals. PKU patients are regularly monitored via metabolomic methods for the buildup of phenylalanine and phenylketones through monthly blood tests.

Yet another example of an IEM that can be treated via metabolomics-guided precision nutrition is biotinidase deficiency (BTD). BTD is an autosomal recessive IEM that can be detected via MS-based newborn screening for low biotin levels. BTD prevents the vitamin biotin from being recycled by the body [33]. Biotin helps with the breakdown of protein, fats, and carbohydrates. If left untreated, those with BTD can develop breathing and balance problems, vision and hearing loss, and alopecia and skin rashes. BTD patients can avoid these neurological and dermatological disorders through oral biotin supplements [34]. Blood tests are not usually performed to monitor BTD patients; however, periodic monitoring for dietary noncompliance may be performed with younger individuals. Thus, for IEMs, metabolomics offers precise, personalized diagnoses that help catch the condition early so that precise diets, customized lifestyle changes, or individualized medical foods can be developed or prescribed. Indeed, metabolomics has been enabling precision nutrition for the proactive treatment of IEMs for >2 decades.

Understanding what nutrients and micronutrients are in our food is key to understanding their health-promoting or health-harming properties. This kind of molecular understanding is also key to the main goals of precision nutrition. Traditionally most food analyses have focused on measuring about 30–40 macronutrient or essential nutrients. These measurements are normally performed by national food agencies such as the USDA or Health Canada. Most developed countries have national food composition databases that provide detailed measurements of these nutrients for thousands of raw or prepared foods.

However, most foods actually contain thousands of compounds, not dozens [35]. Indeed, the average fruit or vegetable is estimated to contain >15,000 different compounds from >100 chemical classes with concentrations ranging from femtomolar (e.g., vitamins) to millimolar (e.g., sugars) [35]. Many of the unmeasured or less measured compounds in foods include micronutrients such as polyphenols, which have effects that could lead to important health benefits [36], terpenes which give many foods their flavor and aroma [37], and various pigments which give foods their color. The only way to detect, identify, and quantify this rich chemical diversity in foods is through metabolomics. A diagram illustrating standard metabolomic workflow for analyzing food samples is shown in Figure 2. Comprehensive MS and NMR metabolomics studies on milk [17], beef [38], bananas [39], wine [40], beer [41], rice [42], and tomatoes [43] have led to the identification of hundreds to thousands of previously undocumented chemicals in these foods and beverages. This work has led to the creation of a number of online food constituent or food metabolome databases such as Phenol-Explorer [44], PhytoHub [45], and FooDB [46]. Phenol-Explorer contains detailed data on 501 polyphenols from 459 different foods along with data on the effects of food processing on polyphenol contents on 161 polyphenols from 155 foods. PhytoHub includes data on > 1,800 food phytochemicals (including many terpenes) and their metabolites from 356 different plant foods. FooDB, which is the largest of these databases, has information on >71,000 chemicals in nearly 800 different raw or lightly processed foods. FooDB is particularly notable for including information on the health benefits, flavor, color, and aroma characteristics for many of these food compounds. These food metabolome databases, which primarily capture metabolomics-derived food composition data, are becoming key resources for guiding precision nutrition and informing nutrition scientists. This is because they provide rich, actionable information about the natural, unnatural, and even the unexpected compounds that can be found in common foods. This information can be used to formulate foods for medical purposes and advance our understanding of foods that have the desired ingredients at the desired concentrations to maximize health benefits.

For precision nutrition to deliver on its promises, it is vital to have a detailed understanding of an individual’s diet and their overall nutritional status. Nutritional assessment has been traditionally done through surveys such as 24-h dietary recalls, food frequency questionnaires, or dietary diaries. However, these methods have many inherent errors and limitations. Problems with recall bias, subjectivity, deliberate deception, memory lapses, inability to estimate portion sizes, lack of preparation details, and food processing information can often lead to incorrect or incongruous data. Due to the limitations of these methods, there has been a strong push to use more objective analytical methods that are more precise and more reliable to measure an individual’s food consumption. This led to a major initiative launched in 2013 called the Food Biomarker Alliance (FoodBAll). The goal of FoodBAll was to use metabolomics to identify biomarkers of food intake (BFIs) [47]. FoodBAll has created an inventory of metabolite markers in blood, urine, or other tissues/biofluids that can be used to reliably identify what a person has eaten and how (rapidly or slowly) that food has been metabolized. Understanding exactly what a person has eaten, how much they have eaten, and how it has been metabolized is critical to doing precision nutrition. Indeed, without this kind of reliable information, the development of personalized dietary interventions or the measurement of personalized dietary responses is almost impossible.

To catalog these BFIs, members of the FoodBAll consortium systematically reviewed the literature and conducted independent metabolomic studies to identify dozens of BFIs for many classes of foods [17, 48-52]. The FoodBAll consortium also developed protocols and definitions for BFI identification and validation [53, 54]. Many of these BFIs are now listed in the HMDB [55], MarkerDB [56], and in Exposome-Explorer [24]. Some of the more useful BFIs identified include urinary 1-methylhistidine and 3-methylhistidine as BFIs of meat intake [57], urinary proline-betaine as a BFI of citrus intake [58], urinary TMAO as a BFI of fish consumption [59], serum daidzein as a marker of soy intake [60, 61] and urinary 2-furoylglycine as a BFI for coffee consumption [62]. In many cases, the performance of these biomarkers in terms of their sensitivity, specificity, and area under the receiver operating characteristic curve is quite impressive. In every case, these BFIs must be measured using targeted metabolomic methods.

BFIs offer nutrition scientists an unbiased, scientifically rigorous approach to molecularly assessing food consumption and avoiding the failings of food questionnaires. They also open the door to understanding individual differences in terms of dietary habits, food metabolism, and gut microbial activity. This kind of information is critical to implementing precision nutrition strategies. Indeed, without a solid understanding of what a person consumes and how their body processes, it is almost impossible to develop effective dietary or lifestyle interventions that could improve a person’s health or well-being [63].

Metabolomics is now being used to understand the diet-related mechanisms associated with many chronic conditions such as cardiovascular disease, diabetes, and obesity [64-66]. Among the most impressive examples of how metabolomics and precision nutrition have complemented each other has been in diabetes control or prevention. Diabetes is rapidly becoming one of the most common life-threatening conditions in the developed world. Approximately 10% of adults in the United States are diabetic and 35% are prediabetic [67]. To explore how metabolomics, metagenomics, and precision nutrition could work to control diabetes, Zeevi et al. [68] integrated these omics techniques with machine learning to develop customized dietary recommendations for controlling elevated postprandial blood glucose. In performing this study, the authors collected metabolomic-based blood parameters, anthropometric data, physical activity data, self-reported dietary intake, and gut microbial composition and function from 800 healthy and prediabetic participants. Participants were given standardized meals (glucose, bread, bread and butter, bread and chocolate, and fructose), and their blood glucose levels were continuously monitored for 1 week. Even though all participants ate the same food, they found a high interpersonal variability in postprandial glycemic responses. Using the data collected from these 800 individuals, they developed a machine-learning algorithm to accurately predict glycemic responses for each participant (based on their omic profile) after consuming different types of food. This algorithm was later successfully validated in another cohort of 100 individuals. This work led to the establishment of one of the first precision nutrition companies, DayTwo Inc., which specializes in using both individual metabolomics and microbial genomics data to create custom diets that can control or prevent prediabetes [69].

In another example of metabolomics being used to guide precision nutrition, Anwar et al. [70] conducted a study on 40 healthy adults with normal blood glucose levels to predict the risk of developing type 2 diabetes. At the beginning of the study, blood samples were collected from participants to quantify metabolite levels using LC-MS-based metabolomics. The blood metabolite levels were then used to calculate health risk scores for the development of type 2 diabetes, insulin resistance, and associated comorbidities. Based on the analysis of their health risks scores, participants were provided with health reports with several lifestyle recommendations including personalized diets (guided by their dietary preferences and food allergies), exercise (based on physical limitations), and nutritional supplements to normalize their blood metabolite levels. After following the recommendations for 100 days, a follow-up metabolomic assessment showed significant reductions in the health risks associated with type 2 diabetes and associated comorbidities. Larger studies of up to 1,000 North American adults showed similar positive results when participants were given personalized nutrition and lifestyle recommendations based on baseline dietary behavior, preferences, and health markers measured via metabolomics [71]. While most metabolomics-driven precision nutrition studies have been limited to controlling diabetes, prediabetes, and/or obesity, it is clear that these approaches are having positive effects. Given that diabetes, prediabetes, and/or obesity affect >30% of American adults and given that they also lead to heightened risks for heart disease, stroke, and cancer, these success stories for precision nutrition and metabolomics are highly significant.

The application of metabolomics to diverse areas of nutritional and food science research has significantly extended our knowledge about the many chemicals in food. This work has also revealed the diversity of human metabolic responses to these chemicals. As highlighted in this review, the development of specialized food composition databases and the identification of robust BFIs using metabolomic platforms have provided new opportunities for researchers to design evidence-based dietary recommendations and to assess their effectiveness on human health. While the use of metabolomics to guide precision nutrition has largely been limited to treating IEMs and diabetes, the success of these metabolomics-informed interventions suggests that similar kinds of lifestyle interventions could be used to treat or prevent many other chronic conditions.

Certainly, there are a number of limitations and challenges to using metabolomics in nutritional research (extensively discussed by Maruvada et al. [72]). However, increased adoption is happening and important advances are being made. For instance, recent improvements in NMR and MS-based technologies for targeted metabolomics combined with much more automated metabolomic analyses have made these techniques much cheaper and more widely accessible. Indeed, the projected cost to quantitatively measure up to 600 metabolites in a drop of blood or urine is predicted to be <USD 30. These dramatic cost reductions should open the door to widespread adoption and use in precision nutrition applications. As a result, a number of new precision nutrition companies are emerging that are taking advantage of these pricing dynamics. However, a key limitation of metabolomics being applied to diet and nutrition studies is the relatively short-term dietary window (a few hours to a few days) that blood, urine, or stool samples provide. New developments in the area of hair [73-75], fingernail, and toenail metabolomics [76] offer the possibility of analyzing and interpreting much longer term dietary patterns. We expect that, as more metabolomics and nutrition data are collected and more BFIs continue to be identified, it is likely that many more robust connections to health outcomes will become apparent. While this may take some time, these unique datasets and these important health connections will go a long way to making precision nutrition more ubiquitous and more useful to a larger segment of the population.

The authors have no conflicts of interest to declare.

Funding for this work has been provided by Genome Canada, the Canada Foundation for Innovation (CFI), Western Economic Diversification, and the Canadian Institutes of Health Research (CIHR). A.H.K. was funded by postdoctoral fellowships from Alberta Innovates and CIHR.

D.S.W. conceived the idea for the manuscript. M.L., A.H.K., and P.Z. performed the literature review and prepared the first draft of the manuscript. D.S.W. edited and revised the final manuscript. All the authors approved the final version of the manuscript.