Paving the Way to Precision Nutrition Through Metabolomics

Nutrition is an interdisciplinary science that studies the interactions of nutrients with the body in relation to maintenance of health and well-being. Nutrition is highly complex due to the underlying various internal and external factors that could model it. Thus, hacking this complexity requires more holistic and network-based strategies that could unveil these dynamic system interactions at both time and space scales. The ongoing omics era with its high-throughput molecular data generation is paving the way to embrace this complexity and is deeply reshaping the whole field of nutrition. Understanding the future paths of nutrition science is of importance from both translational and clinical perspectives. Basic nutrients which might include metabolites are important in nutrition science. Moreover, metabolites are key biological communication channels and represent an appealing functional readout at the interface of different major influential factors that define health and disease. Metabolomics is the technology that enables holistic and systematic analyses of metabolites in a biological system. Hence, given its intrinsic functionality, its tight connection to metabolism and its high clinical actionability potential, metabolomics is a very appealing technology for nutrition science. The ultimate goal is to deliver a tailored and clinically relevant nutritional recommendations and interventions to achieve precision nutrition. This work intends to present an update on the applications of metabolomics to personalize nutrition in translational and clinical settings. It also discusses the current conceptual shifts that are remodeling clinical nutrition practices in this Precision Medicine era. Finally, perspectives of clinical nutrition in the ever-growing, data-driven healthcare landscape are presented.

. These endeavors led to the emergence of the concept of metabotyping or metabolic phenotyping, which stratify individuals based on their metabolic and phenotypic patterns into metabotype subgroups with high metabolic similarity. Thus, this metabolic similarity could be used for population stratification to define tailored dietary or medical intervention (8)(9)(10). Two main technologies are used in metabolomics, Nuclear magnetic spectroscopy (NMR) and mass spectrometry (MS) combined or not with a gas phase or liquid phase separation method (11). NMR spectroscopy has the advantage to be nondestructive, rapid, highly reproducible, and robust but it lacks sensitivity. MS-based methods are more sensitive, but they are destructive and might have lower throughput especially if combined with a separation dimension. These techniques recover global, unbiased, and comprehensive chemical information from the assessed biological samples. Metabolomics workflows are based on the previous knowledge of the metabolites to analyze. A targeted approach is defined as a quantitative analysis when absolute concentrations are determined or a semiquantitative analysis when relative intensities are determined for a set of metabolites. The untargeted approach is primarily based on the qualitative or semiquantitative analysis of the most possibly detectable metabolites in the biological sample. Untargeted metabolomics is often used for biomarker discovery studies. Depending on the study, different parameters should be taken into account; defining the metabolomics approach (targeted vs. untargeted), biological samples, sample size, pooling, experimental conditions (i.e., observational studies, exploratory studies, time series), sampling conditions and storage, analytical platforms, and standardized sample preparation procedures. To translate the metabolomics data into actionable information, data analysis. The choice of the most appropriate data analysis strategy depends mostly on the underlying investigation aim. In mechanistic studies, the structure of the data descriptions of the built model are more important than its predictive power to classify new samples. In diagnosis or prognostic applications, the classification performances of the model are crucial. It is also to be noted that sample sizes and model validation using external validation sets are mandatory to avoid overfitting of the models to avoid spurious conclusions (12). In this review we will focus on the use of metabolomics in human studies to achieve personalized nutrition or precision nutrition (PN) which refers to the design of customized nutritional recommendations to prevent or to treat nutrition-related disorders (13).

PATHWAYS TOWARD PRECISION NUTRITION Nutritional Research
Nutritional research focuses on unveiling the interrelationships between diet and health to prevent or treat diet-related disorders (14). It is established that the individual nutritional needs vary largely according to anthropometrics such as age, gender and also on physiological states and other biological and environmental attributes (15). Furthermore, the response to dietary interventions exhibits substantial inter-individual variation. In this era of precision medicine, it is time to tailor dietary intervention and advice at the individual level such as biomarkers related to red meat (16) or bread consumption (17). To effectively understand diet and lifestyle interaction in health and disease, robust, and informative biomarkers for nutritional status are mandatory (18,19). However, different challenges hamper the application of nutritional biomarkers. First, there is no consensus on nutritional biomarker definition, its assessment and use. Furthermore, even the well-validated biomarkers lack consistency to support strong recommendations (20,21). To overcome these constraints, several initiatives have been launched to promote the identification and development of new nutritional biomarkers by defining methods, conditions, and settings of nutritional biomarker quantification to set standardized and actionable protocols (22,23). In addition to this organizational and strategic endeavors, deep and extensive research is still mandatory to provide a deep molecular understanding of nutrients effects on health and disease. Recently, a comprehensive definition of personalized nutrition has emerged that describes three main levels in precision nutrition: level 1 is dietary advice based on dietary intake; level 2 adds to level 1 phenotypic metrics and laboratory tests; level 3 goes deep by adding, to the previous levels, the genetic information layer (22). Large-scale initiatives such as the Food4me performed randomized control trial to assess improvements in personalized dietary advice compared to conventional approaches (7). The results showed changes in dietary behavior related to customized nutritional advice. However, from a public health perspective, these results are hard to scale up. This work intends to present an update on the applications of metabolomics to personalize nutrition in translational and clinical settings. It also discusses the current conceptual shifts that are remodeling clinical nutrition practices in this Precision Medicine era.

Traditional Dietary Assessment Techniques and Their Limitations
For decades, nutritional research has been a crucial pillar to unveil diet-health relationships at both individual and population scale. However, consistency, validation and reproducibility of the dietary assessments have been the great challenges (19). For a long time, food diaries, 24 h recalls and Food Frequency Questionnaire (FFQ) are used as self-report dietary assessment methods in human nutrition studies (24). Multiple days weighted food records are often used in controlled dietary studies (25). However, these dietary records are time consuming and tedious for both participants and investigators, and lack accuracy and consistency (25,26) which limits their integration in large-scale studies. Alternatives such as 24 h recalls may be used. This method may include a structured collection of detailed information about food intake during the previous 24 h. A major drawback is that these recalls are subjected to inter-day nutritional variation (25). For largescale studies, self-administered FFQ are preferred to evaluate food intake. All these methods (food records, recalls, and FFQ) require an intensive preparation before implementation. However, the managing and processing of a well-validated FFQ is rather smooth for large studies. Still, inaccurate portion size estimation, socially distorted answers, lack of objectivity and accuracy in self-reported data, and errors in food intake composition hamper FFQ and make it prone to record errors (27,28). The implementation of high-throughput omics tools enables the holistic and integrative interrogation of nutrition by identifying nutrition-specific biomarkers to objectively assess dietary intake. These biomarkers may provide actionable information to fill in the gaps of self-reported food intake given the tight connection between metabolite production, food intake, the microbiome, and health status (29). Indeed, stratification of individuals into groups can be achieved through the analysis of their respective metabolic profiles. This approach of grouping subjects based on their metabolic phenotype was coined "metabotyping" and has been used in several of studies (30,31). This review will examine this concept in the field of personalized nutrition.

Pathways Toward Precision Nutrition
The goal of precision or personalized nutrition (PN) is the design of customized nutritional recommendations to prevent or to treat nutrition-related disorders (13). PN aims to set dynamic nutritional recommendations based on individual attributes and external environmental factors. Hence, PN strategies should include genomics information, other factors such as dietary and physical activity patterns, metabolome, and microbiota. The post-genomic era generated various genome-wide association studies (GWAS) that allow the identification of genomic factors that might unveil the inter-individual metabolic response variability to specific diets. Various genes and polymorphisms have been defined as relevant factors to explain diet-specific metabolic responses (32)(33)(34)(35)(36). However, there is weak clinical evidence to establish comprehensive and integrated framework for personalized nutritional interventions in clinical settings (37). Despite this, different successful implementations already made it to the clinic. For example, in inborn errors of metabolism (IEM) which are a group of about 500 rare genetic diseases due to defects in a given biochemical pathway due to the deficiency or abnormality of an enzyme, its cofactor, or a transporter, leading to an accumulation of the substrate or lack of the product. In IEM, nutritional strategies are crucial in treatment of some disease such as phenylketonuria screening (38). Interventional nutrition has allowed the implementation of specific diet and saved patients from devastating outcomes for them and their families. In the wellness sphere, genetic tests are used to define slow or fast metabolizers (33). While genomicbased customized nutrition is already being implemented, PN might lack sufficient evidence for full integration into clinical setting (39).

METABOLOMICS AS A KEY ENABLER OF PRECISION NUTRITION
Metabolomics is a key tool to investigate the effect of food on the individual's health. By identifying food-derived biomarkers, inter-individual variability in metabolizing same foods in health and disease states. Metabolomics could enhance nutrition assessment through three main pathways: dietary intake biomarker discovery, diet-related diseases exploration through cohort studies, and dietary intervention assessment through metabolic patterns. A summary of discussed studies are presented in Table 1.

Dietary Intake Biomarker Discovery Using Intervention Studies
Dietary intervention studies are based on consuming specific food followed by biofluid collection either postprandial or following a short-term intervention (54). Urine is the most used biological sample due to its convenience compared to plasma and serum. This strategy unveiled potential food intake biomarkers such as citrus fruit (40), red meat (41), coffee (42), tea (43), sugar-sweetened beverages (44), and wine (45). Even though, several food specific biomarkers have been reported they still lack independent large-scale validation studies. It also worth noting that most of these biomarkers are rapidly secreted in urine. Thus, they are mostly acute food intake biomarkers. Hence, given this temporal limit, longer-term biomarkers of food intake are crucially needed. Currently, an increasing number of studies are assessing biomarkers in serum and plasma.

Dietary Biomarker Discovery Using Cohort Studies
Large-scale cohort studies are also valuable tools to discover dietary biomarkers. Dietary data may be collected using traditional methods to identify consumer profiles related to specific food (e.g., low and high consumers, consumers, and nonconsumers). In this case, metabolomic profiles are compared between these subgroups to unveil potential dietary biomarkers. Given their practicability, Self-reported dietary assessment methods are often used in large-scale studies tend to rely on which are known to be notoriously prone to error. It is worth noting these studies are mainly association studies and don't allow causal inference (55). Intervention studies are, therefore, needed to validate the potential metabolite as a specific food intake biomarker. Several cohort studies identified intake biomarkers such as fish (46), red meat (16,56), whole-grain bread (17), and walnuts (47). Furthermore, metabolomics has also been used in populational stratification (57)(58)(59). Recently, Wittenbecher et al. (16) applied serum metabolomics to unveil significant relationships between different red meat intake biomarkers with risk of type-2 diabetes. The authors state that high levels of ferritin, low glycine, and altered hepatic derived lipids in the circulation were associated with both total red meat consumption and diabetes risk. This first study assessing a large set of metabolites revealed the link between red meat intake and diabetes risk and thus achieved an important step in biomarker discovery in precision nutrition. Unveiling these links between will allow deeper understanding potential metabolic pathways as disease drivers. These pathways could be validated through more interventional nutrition studies using targeted approaches for biomarker validation.  suggest that medium-to long-chain acylcarnitines show opposite patterns related to fruits and desert intake. Whereas, shortchain acylcarnitines and amino acids, were positively correlated with saturated fat (53). These studies show the potential of metabolomics is not only a tool dietary pattern compliance, but also to identify and assess diet and disease relationships (78)(79)(80).

INTERROGATING MICROBIOTA-DIET CROSS TALKS THROUGH METABOLOMICS
Gut microbiota profiling is gaining great interest in nutritional interventions to assess dietary effects on gut microbiome ecosystem diversity (81). The core idea is to tailor nutritional interventions by optimizing the richness and diversity of the gut microbiota (82). Diet may effect on the gut microbiota and its impact on the microbial metabolome is very appealing.
Diet may serve as a substrate that can be processed by the gut microbiota for the production of small molecules to sustain host-microbiome interactions (83)(84)(85). For example, it has been reported that the majority of microbiota-derived short-chain fatty acids is absorbed by the host (86) and contributes to host nutrition (87). The FRUVEDomics Study is a behavioral interventional trial. It aims to identify metabolomics-and microbiome-based nutritional risk factors. The study included 36 participants that were randomized into three intervention groups, fruits and vegetables diet, associated either with low-fat or low carbohydrates (88). The results suggested that metabolic syndrome is correlated to higher Firmicutes to Bacteroidetes ratio (88). Other studies highlight the relevance of interrogating the gut microbiota to achieve precision nutrition (89,90). These studies suggest that increased fasting plasma levels of trimethylamine (TMA), produced by gut microbiota, is associated with increased risk of atherosclerosis. These studies led to targeted recommendations of nutrition advice such as reducing red meat intake (91) for subjects with gut microbial ecosystem that could convert red meat related nutrients into proatherogenic molecules. Other general recommendations may also be questioned, such as using artificial sweeteners, as reported by Suez et al. (92) who reported that high sweetener intake may lead to glucose intolerance in subjects with sensitive gut microbiota (92). However, their results seem to be controversial given the used high dose of sweetener (93,94). Another way to take advantage of this microbiota-host interaction through diet and thus, through metabolites, is to use prebiotics which are substrates that are selectively and used by host microorganisms and led to better health for the host. They are another appealing mean to take advantage of microbiome modulation (95). Some studies suggest a potential therapeutic use of prebiotics (96)(97)(98)(99). For example, fiber-rich diet is associated with an increased Prevotella/Bacteroides ratios and improved glucose metabolism (100). However, diet is a highly diverse biochemical space, often tuned by the microbiota (101). Thus, dietary interventions may be personalized based on this fine-tuning of microbiome of diet (102). Zeevi et al. used an elegant strategy integrating microbiome, clinical, and dietary data to build predictive models to shape customized dietary recommendations for healthier glycaemia (18). Using genome-scale metabolic modeling, Shoaie et al. elegantly modeled the interactions of diet, gut microbiota, and host metabolism. The authors successfully predicted microbiome metabolic responses related to a dietary intervention in obese subjects and formally validated their predictions using fecal and blood metabolomics data (103). Given this microbiota plasticity, microbiome-targeted interventions are very appealing approach to personalize diet intervention. However, this microbiota plasticity may be dependent on previous dietary habits (104) and baseline microbial populations, which could hamper microbiota-directed intervention responses (105). Thus, high resolution definition of key diet-responsive microbiome is very important to enhance the prediction performance of dietary interventions. Despite current challenges that hamper effective clinical translation of microbiota-related to nutrition interventions, new analytical, and computational developments may help overcome these limits. Indeed, its integration with other omics such as proteomics and metabolomics for better and more accurate functional profiling (18,106). However, still largescale studies are needed to draw robust conclusions. Moreover, controlled studies are mandatory to characterize environmental factors independent of diet which may be crucial in modeling the gut microbiota ecosystem. These evidence-based requirements will help to open promising opportunities in designing personalized nutritional strategies to shape the precision nutrition era.

PRECISION NUTRITION THROUGH INTEGRATIVE OMICS
As above mentioned, food patterns and eating behaviors are extremely complex that include multilayer informational flows (107). Gene-diet interactions might play a role in the development of and protection against chronic diseases (108). For example, the antioxidant content of the Mediterranean diet could a protective mechanism since antioxidants modulate gene expression (109). It is well known that oxidation and inflammation processes are interconnected and may contribute to different diseases the physiopathology (110

CHALLENGES AND FUTURE PERSPECTIVES TOWARD PRECISION NUTRITION
One of the raised limits of PN is that most studies are observational rather than from randomized controlled trials (RCTs) with assessed clinical end points. Limited information is so far available to confirm that PN produce more accurate and sustained changes in behavior than currently used approaches and do these PN strategies led to better health (7,19,120). So far, no large scale and long term personalized nutrition study has been carried out (19,121). The logistical burden and costs of large-scale nutrition intervention studies with disease risk as outcomes hamper their implementation. Thus, the use of diet changes, or established biomarkers of disease (glycated hemoglobin A1c, blood pressure, or cognitive assessment) as outcomes may be potential surrogates (121). Advancement of the young discipline of PN, including metabotyping, will require new perspectives. Setting strong theoretical foundations by identifying key individual attributes that drive the personalization process. Evidence for cost effectiveness of well-designed interventional studies is also fundamental. Moreover, the introduction of a regulatory framework is mandatory to gain trust of health professionals and policy makers and enhance public protection. This will need a substantial enhancement in the scientific evidence using robust random control trials. These efforts should include multidisciplinary teams, comprising clinicians, behavioral psychologists, nutritionists, computer scientists, and biomedical scientists (19). Furthermore, the integration of other "omics" to provide deeper mechanistic insights will be crucial to improve the evidence of precision nutrition (103,122).

CONCLUSION
Metabolomics proved to be a valuable tool for the measurement of biochemical changes associated health changes related to diet. It is also, highly, promising in identification of nutritional biomarkers to monitor nutritional intervention studies. The greatest challenge for metabolomics research is its integration with other omics and phenotypic data (123). However, better standardization of metabolomics data is mandatory to effectively implement its functional and operational integration. This will enhance our knowledge of diet-health relationships. Nutrition research is intrinsically multidisciplinary, and so is metabolomics. It requires collaboration among research scientists with overlapping expertise areas including nutritionists, clinicians, nurses, bioinformaticians, analytical scientists, statisticians and chemists, and many other stakeholders. This expertise integration is vital to develop the knowledge to establish the evidence-based precision nutrition. The greatest challenge to cracking the relationships between food and health is to decipher the high inter-individual variability responses to food intake. The new frontier of the nutritional sciences lies in our ability to predictably engineer our physiologic networks for diet, health, and disease. This will ultimately allow fine tuning of diet intervention and health monitoring (18). Efforts must be done to support evidence-based nutritional research and achieve effective diet-based disease prevention.

AUTHOR CONTRIBUTIONS
AT performed literature search and drafted the manuscript. SB critically revised and edited the manuscript. All authors approved the final manuscript.