Hostname: page-component-78c5997874-s2hrs Total loading time: 0 Render date: 2024-11-08T20:24:35.220Z Has data issue: false hasContentIssue false

Metabolomics as a tool in the identification of dietary biomarkers

Published online by Cambridge University Press:  25 May 2016

Helena Gibbons
Affiliation:
Institute of Food and Health, UCD School of Agriculture and Food Science, University College Dublin (UCD), Belfield, Dublin, Ireland UCD Conway Institute of Biomolecular Research, UCD, Belfield, Dublin, Ireland
Lorraine Brennan*
Affiliation:
Institute of Food and Health, UCD School of Agriculture and Food Science, University College Dublin (UCD), Belfield, Dublin, Ireland UCD Conway Institute of Biomolecular Research, UCD, Belfield, Dublin, Ireland
*
*Corresponding author:Professor L. Brennan, email [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Current dietary assessment methods including FFQ, 24-h recalls and weighed food diaries are associated with many measurement errors. In an attempt to overcome some of these errors, dietary biomarkers have emerged as a complementary approach to these traditional methods. Metabolomics has developed as a key technology for the identification of new dietary biomarkers and to date, metabolomic-based approaches have led to the identification of a number of putative biomarkers. The three approaches generally employed when using metabolomics in dietary biomarker discovery are: (i) acute interventions where participants consume specific amounts of a test food, (ii) cohort studies where metabolic profiles are compared between consumers and non-consumers of a specific food and (iii) the analysis of dietary patterns and metabolic profiles to identify nutritypes and biomarkers. The present review critiques the current literature in terms of the approaches used for dietary biomarker discovery and gives a detailed overview of the currently proposed biomarkers, highlighting steps needed for their full validation. Furthermore, the present review also evaluates areas such as current databases and software tools, which are needed to advance the interpretation of results and therefore enhance the utility of dietary biomarkers in nutrition research.

Type
Irish Section Postgraduate Meeting
Copyright
Copyright © The Authors 2016 

Dietary biomarkers and the concept of metabolomics

The contribution of diet to the increasing burdens of CVD, diabetes, obesity and cancers has been recognised since the 1970s( Reference Mozaffarian and Ludwig 1 ). Selected foods and nutrients as well as dietary patterns are now known to interact with various metabolic processes contributing to a reduction or an increase in the risk of disease( Reference Bingham 2 ). For example, it is well established that high salt consumption raises blood pressure( Reference Kotchen, Cowley and Frohlich 3 ) and high consumption of red meat has been associated with increased incidence of type 2 diabetes( Reference Feskens, Sluik and Fau - van Woudenbergh 4 , Reference Pan, Sun and Bernstein 5 ), CVD( Reference Hu, Rimm and Stampfer 6 ) and cancers( Reference Cross, Leitzmann and Gail 7 ). In contrast, dietary patterns such as the dietary approaches to stop hypertension diet, which emphasises consumption of fruit and vegetables, low-fat dairy foods and whole grains and reduced intake of red meats and sugars has been shown to decrease blood pressure and CVD risk( Reference Appel, Moore and Obarzanek 8 , Reference Sacks, Svetkey and Vollmer 9 ). Similarly, the Mediterranean diet which emphasises high fruit, vegetable and olive oil consumption has been shown to reduce CVD and type-2 diabetes risk( Reference Estruch, Ros and Salas-Salvado 10 , Reference Salas-Salvadó, Bulló and Babio 11 ). As diet is a key environmental risk factor, the identification and targeting of dietary factors with the greatest prospective for reducing or increasing disease risk is of major scientific and public health importance( Reference Mozaffarian, Appel and Van Horn 12 ). It is therefore essential that dietary assessment methods are reliable and accurate for the advancement of our understanding of the links between diet and health.

Diet is traditionally measured via self-reporting methods such as FFQ, 24-h recalls and weighed food diaries. There is however a number of methodological issues associated with each of these assessment methods, including energy underreporting, recall errors and difficulty in assessment of portion sizes( Reference Bingham 2 , Reference Kipnis, Midthune and Freedman 13 , Reference Dhurandhar, Schoeller and Brown 14 ). Such errors can lead to reduced power, underestimated associations and false findings which may contribute to inconsistencies in the field of nutritional epidemiology( Reference Dhurandhar, Schoeller and Brown 14 , Reference Marshall and Chen 15 ). In an effort to address some of these measurement issues, the use of dietary biomarkers, which are found in biological samples and are related to ingestion of a specific food or food group, have emerged( Reference Brennan, Gibbons and O'Gorman 16 ). Currently dietary biomarkers exist for salt, protein, sucrose/fructose intake (sodium/nitrogen/sucrose and fructose measured in 24 h urine samples) and energy expenditure (the doubly labelled water technique)( Reference Bingham 2 , Reference Tasevska, Runswick and McTaggart 17 ). These dietary biomarkers can be used in conjunction with traditional dietary assessment methods to improve the accuracy of dietary intake measurement and can also be used to more accurately associate dietary intake with disease risk and nutritional status( Reference Potischman 18 ).

In recent years, metabolomics has developed as a key technology for the identification of new dietary biomarkers. Metabolomics provides a powerful approach for the comprehensive description of all low molecular weight molecules present in biological samples( Reference Brennan, Gibbons and O'Gorman 16 ). In metabolomics research, the analytical platforms predominantly used are NMR spectroscopy and MS coupled with a chromatographic step, for example, GC or liquid chromatography. Each of these techniques is associated with a number of advantages and disadvantages, for example MS-based techniques have high sensitivity and therefore may detect metabolites below the detection limit of NMR spectroscopy; however, sample treatment is necessary before MS-based analysis, while little or no pre-treatment is required for NMR( Reference Wishart 19 ).While in the past many articles detailed the advantages and disadvantages of different approaches there has now been a realisation that using one platform alone will not give complete coverage of the metabolite profile; therefore, a combination of technologies and approaches is usually recommended for optimal coverage. Analysis of metabolomic data is commonly performed using multivariate statistics and there are an increasing selection of databases and tools available to assist in the interpretation of these multivariate results( Reference Scalbert, Brennan and Manach 20 ).

Examination of the literature reveals that there are three approaches generally employed for dietary biomarker discovery. These can be summarised as: (i) acute or medium interventions where participants consume specific amounts of a test food and biological samples are collected post consumption, (ii) cohort studies where metabolic profiles are compared between consumers and non-consumers of a specific food and (iii) the analysis of dietary patterns and metabolic profiles to identify nutritypes and biomarkers. Although these study designs have led to the identification of a number of biomarkers in the literature in recent years, each of these approaches has a number of limitations associated with it. Awareness of these is important in the interpretation and potential use of such biomarkers. Therefore the objective of the present review is to give an overview of currently proposed biomarkers and secondly the present review aims to critique the current literature in terms of approaches for dietary biomarker discovery, highlighting steps needed for their full validation.

Dietary biomarker discovery using intervention studies

Dietary intervention studies involve participants consuming specific amounts of a test food in a single meal (acute intervention) or for a short-to-medium term intervention the test food is consumed in repeated meals. In this approach, baseline and postprandial biofluids are collected and following analysis, potential biomarkers are identified. This approach has led to the identification of a number of putative biomarkers of specific foods and beverages as summarised in Table 1. An excellent example of a biomarker successfully identified using this approach is proline betaine, a robust biomarker of citrus fruit intake. Proline betaine was originally identified by Atkinson et al.( Reference Atkinson, Downer and Lever 21 ) and following this Heinzmann et al. performed an acute intervention study with a mixed-fruit meal, which consisted of apples, grapes, oranges and grapefruit( Reference Heinzmann, Brown and Chan 22 ). Eight participants consumed standardised meals over 3 d and on the second day the mixed-fruit meal was consumed( Reference Heinzmann, Brown and Chan 22 ). Urine samples were collected and analysed using NMR spectroscopy. Following multivariate analysis proline betaine was identified as a potential biomarker. To assign the origin of urinary proline betaine excretion after the mixed-fruit meal, concentrations of proline betaine in fruits and fruit juices were measured. Concentrations of proline betaine were higher in citrus fruit compared with other commonly available fruit and fruit juices tested. The urinary excretion profile of proline betaine was then measured in six individuals after consumption of orange juice. This biomarker was confirmed using data from participants in the INTERMAP UK cohort and demonstrated a high sensitivity and specificity for citrus fruit consumption (90·6 and 86·3 %, respectively)( Reference Heinzmann, Brown and Chan 22 ). Lloyd et al. also identified proline betaine and a number of biotransformed products in postprandial urine samples after consumption of 200 ml orange juice as part of a standardised test breakfast( Reference Lloyd, Beckmann and Favé 23 ). Subsequent biomarker validation demonstrated sensitivities and specificities of 80·8–92·2 and 74·2–94·1 %, respectively, for elevated proline betaine in high reporters of citrus fruit consumption( Reference Lloyd, Beckmann and Favé 23 ). Following on from these acute studies, a medium-term intervention study used MS to profile the urinary metabolomes of twelve volunteers that consumed orange juice regularly for 1 month as part of their habitual diet. Proline betaine was again identified as a potential marker of citrus fruit( Reference Pujos-Guillot, Hubert and Martin 24 ). Considering the range of studies that consistently report proline betaine as a marker of citrus fruit intake the evidence base is strong to support its use.

Table 1. Summary of putative biomarkers identified using a metabolomics approach in intervention studies

LC, liquid chromatography; ESI, electrospray ionisation; qTOF, quadrupole time-of-flight; LTQ, linear trap quadrupole; FIE, flow infusion electrospray ionisation; FTICR, Fourier transform-ion cyclotron resonance; qTOF, quadrupole time-of-flight; UPLC, ultra-performance liquid chromatography; HILIC, hydrophilic liquid interaction chromatography; TMAO, trimethylamine-N-oxide; LTQ-FT. linear ion trap-Fourier transform mass spectrometer; HHPAA, 2-hydroxy-N-(2-hydroxyphenyl)acetamide; HPAA, N-(2-hydroxyphenyl)acetamide; HBOA, 2-hydroxy-1,4-benzoxazin-3-one; HHPAA, 2-hydroxy-N-(2-hydroxyphenyl)acetamide.

A number of research groups have also used dietary interventions to investigate biomarkers of cruciferous vegetables( Reference Edmands, Beckonert and Stella 25 Reference Andersen, Kristensen and Manach 27 ). Andersen et al. performed a controlled cross-over meal study with nine brassica-containing New Nordic Diet meals in seventeen subjects( Reference Andersen, Reinbach and Rinnan 26 ). The 24 h urine samples were collected and analysed by ultra-performance liquid chromatography–quadruple time-of-flight–MS. To investigate the food sources of the biomarkers found in the meal study, a range of small single food studies were performed with three to four participants in each. Using a sensitivity and specificity analyses to select the most promising biomarkers, a range of conjugated isothiocyanates were identified as biomarkers of brassica intake( Reference Andersen, Reinbach and Rinnan 26 ). Further biomarkers of other foods, including fish were also identified( Reference Andersen, Reinbach and Rinnan 26 ). To validate the biomarkers from this study, Andersen et al.( Reference Andersen, Kristensen and Manach 27 ) carried out a 6-month parallel dietary intervention study where 107 participants were randomised into two distinct dietary patterns. Combining liquid chromatography–MS data from 24 h urine samples and data from 3-d weighed dietary data the present study again identified conjugates of isothiocyanates as brassica biomarkers. However, using this approach it was only possible to verify 23 % of potential biomarkers observed in the previous-meal studies( Reference Andersen, Kristensen and Manach 27 ). As this was a less controlled intervention that included a wider selection of foods with varied amounts of intake and different preparation methods, it highlights the need for the validation of biomarkers in different subjects and study settings( Reference Andersen, Kristensen and Manach 27 ).

A number of red meat and fish biomarkers have been identified using this intervention approach( Reference Cross, Leitzmann and Gail 7 , Reference Lloyd, Fave and Beckmann 28 , Reference Stella, Beckwith-Hall and Cloarec 29 ). Most recently, metabolomics has been applied to compare the different effects of meat and fish on the plasma metabolome( Reference Ross, Svelander and Undeland 30 ). Ross et al.( Reference Ross, Svelander and Undeland 30 ) carried out an intervention study analysing the differences in the postprandial plasma metabolic response to meals containing baked beef, baked herring and pickled herring. Seventeen males consumed three test meals in a crossover design with 1 week washout between the meals. Postprandial blood plasma samples were taken over 7 h and analysed by GC–MS. Concentrations of 2-aminoadipic acid, β-alanine and 4-hydroxyproline were significantly higher following the beef meal compared with the baked herring meal. Herring intake led to a greater plasma postprandial response from DHA and cetoleic acid compared with beef( Reference Ross, Svelander and Undeland 30 ). However, further studies are needed to confirm these dietary biomarkers and decipher their specificity.

Dietary biomarker discovery using cohort studies

Searching for new dietary biomarkers in cohort studies requires the use of self-reported dietary data to identify low and high consumers of a specific food. Following this, the metabolomic profiles are compared between low and high consumers and potential biomarkers are identified. Putative biomarkers of foods, identified using this approach, are presented in Table 2. Work in our laboratory combined this approach with an acute intervention to identify and confirm a panel of biomarkers indicative of sugar-sweetened beverage (SSB) intake( Reference Gibbons, McNulty and Nugent 31 ). Heat map analysis was performed to identify correlations between NMR spectral regions and SSB intakes in the cohort study. A panel of four biomarkers: formate, citrulline, taurine and isocitrate were identified as markers of SSB intake. Following the acute consumption of the SSB all four metabolites were shown to increase in the urine and the panel of biomarkers were successfully identified in the SSB( Reference Gibbons, McNulty and Nugent 31 ). Another study using this cohort study approach, analysed the correlations between serum profiles and dietary data collected using FFQ in participants from the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial( Reference Guertin, Moore and Sampson 32 ). The application of untargeted metabolomics to this epidemiologic data set detected thirty-nine metabolites of known identity that were correlated with a total of thirteen dietary groups, for example citrus intake was associated with stachydrine, chiro-inositol, scyllo-inositol and N-methyl proline, fish with 3-carboxy-4-methyl-5-propyl-2-furanpropanoic acid, DHA and EPA, peanut intake with tryptophan betaine and 4-vinylphenol sulphate and coffee intake was associated with trigonelline-N-methylnicotinate and quinate( Reference Guertin, Moore and Sampson 32 ). To complicate interpretation further, the intake of foods is highly correlated making identification of specific biomarkers difficult and this highlights the need for the validation of biomarkers. The majority of biomarkers identified using cohort studies have been predominantly identified in urine, this study demonstrates the potential use of serum samples in dietary biomarker discovery. However, the proposed biomarkers identified are only based on associations and some biomarkers were not food specific, for example DHA was correlated with fish and rice intake. Further validation in intervention studies is therefore necessary to demonstrate responsiveness to intake.

Table 2. Summary of putative biomarkers identified using a metabolomics approach in cohort studies

FIE, flow infusion electrospray ionisation; FTICR, Fourier transform-ion cyclotron resonance; LC, liquid chromatography; ESI, electrospray ionisation; qTOF, quadrupole time-of-flight; LTQ, linear trap quadrupole; UHPLC, ultra-HPLC; CMPF, 3-carboxy-4-methyl-5-propyl-2-furanpropanoic acid; FIA, flow injection analysis; PC aa, diacyl phosphatidylcholines; PC ae, acylalkyl phosphatidylcholines; Lyso-PC, lysophosphatidylcholines; SM, sphingomyelin; HPAA, N-(2-hydroxyphenyl) acetamide; HHPAA, 2-hydroxy-N-(2-hydroxyphenyl) acetamide; HMBOA, 2-hydroxy-7-methoxy-2H-1, 4-benzoxazin-3-one; HBOA, 2-hydroxy-1,4-benzoxazin-3-one; HPPA, 2-hydroxy-N-(2-hydroxyphenyl) acetamide; DHPPA, 3-(3,5-dihydroxyphenyl) propanoic acid; DHPPTA, 5-(3,5-dihydroxyphenyl) pentanoic acid.

Wittenbecher et al. also demonstrated the use of serum samples when identifying biomarkers of red meat intake in a subset of participants from the European Prospective Investigation into Cancer and Nutrition-Potsdam cohort (n 2047)( Reference Wittenbecher, Mühlenbruch and Kröger 33 ). Total red meat consumption was assessed using FFQ and serum samples were analysed using a targeted metabolomics approach. Ferritin, glycine, four diacyl phosphatidylcholines, eleven acylalkyl phosphatidylcholines, two lysophosphatidylcholines and two sphingomyelins were associated with total red meat consumption and six of these biomarkers were also found to be associated with type-2 diabetes risk( Reference Wittenbecher, Mühlenbruch and Kröger 33 ). This is the first study evaluating a large set of metabolites as potential mediators of the association between red meat intake and diabetes risk, however, dietary information relied on estimates of habitual consumption over the past year by FFQ and metabolites were measured at a single time point. Furthermore, total red meat was defined as processed and unprocessed meat and therefore did not identify biomarkers of specific types of meat. Additional study is essential to validate the biomarkers identified and to further dissect such relationships with disease risk.

Biomarkers of bread intake have also been investigated in 155 subjects from the PERIMED study( Reference Garcia-Aloy, Llorach and Urpi-Sarda 34 ). A 137-item FFQ was used to stratify subjects into three groups: non-consumers of bread (n 56), white-bread consumers (n 48) and whole-grain bread consumers (n 51). Fasting urine samples, analysed by untargeted HPLC–quadruple time-of-flight–MS, identified higher concentrations of compounds, including benzoxazinoids and alkylresorcinol metabolites and compounds produced by gut microbiota (enterolactones, hydroxybenzoic and dihydroferulic acid metabolites) in bread consumers. 2, 8-dihydroxyquinoline glucuronide was also found to be more abundant in whole-grain bread consumers( Reference Garcia-Aloy, Llorach and Urpi-Sarda 34 ). The biomarkers identified are based on a FFQ; therefore further validation is essential to demonstrate a direct relationship with bread consumption.

Dietary biomarker discovery using dietary patterns

The third approach, analysing dietary patterns and metabolomic profiles to identify nutritypes (i.e. metabolic profiles that reflect dietary intake) and biomarkers have been demonstrated by a number of research groups (see Table 3). One of the first examples emerged from our laboratory when a k-means cluster analysis was performed on self-reporting dietary data and three distinct dietary patterns, which were associated with unique food intakes were identified( Reference O'Sullivan, Gibney and Brennan 35 ). Dietary clusters were reflected in the urinary metabolomic profiles of the 125 participants and a number of metabolites were identified and linked to the intake of specific food groups( Reference O'Sullivan, Gibney and Brennan 35 ). These nutritypes have the potential to aid dietary assessment by unobjectively classifying people into certain dietary patterns. Further work within our research group, applying the concept of using biomarkers to reflect dietary patterns, has focused on lipidomics, a subfield of metabolomics that concentrates on the global study of lipids( Reference O'Gorman, Morris and Ryan 36 ). Dietary data, measured by FFQ and lipid profiles measured from serum samples, in thirty-four Metabolic Challenge Study participants were used for this analysis. Principal component analysis reduced lipid profiles into lipid patterns and these were regressed against dietary data to identify biomarkers related to the intake of certain foods and nutrients. Six lipid patterns were identified including lipid pattern 1 which was found to be highly predictive of dietary fat intake (AUC = 0·82), lipid pattern 4 which was highly predictive of alcohol intake (AUC = 0·81) and lipid pattern 6 which had a reasonably good ability to predict dietary fish intake (AUC = 0·76). Lysophosphatidylcholine alkyl C18 : 0 (LPCeC18 : 0) was identified as a potential biomarker of alcohol consumption and lysophosphatidylethanolamine acyl C18 : 2 (LPEaC18 : 2) and phoshatidylethanolamine diacyl C38 : 4 (PEaaC38 : 4) were identified as potential biomarkers of fish intake( Reference O'Gorman, Morris and Ryan 36 ). This approach demonstrates the utility of serum in the identification of key dietary factors that influence the lipidomic profile. However, again validation of the biomarkers through use of intervention studies is needed.

Table 3. Summary of putative biomarkers identified using dietary patterns and metabolomic profiles

ADD, Average Danish Diet; PCA, principal component analysis; ESI, electrospray ionisation; LPCeC18 : 0, lysophosphatidylcholine alkyl C18 : 0. LPEaC18 : 2, lysophosphatidylethanolamine acyl C18 : 2. NND, New Nordic Diet; PEaaC38 : 4, phoshatidylethanolamine diaclyl C38 : 4. TMAO, trimethylamine-N-oxide; UPLC, ultra-performance liquid chromatography; qTOF, quadrupole time-of-flight; RRR, reduced rank regression; FIA, flow injection analysis.

Most recently, Andersen et al. used an untargeted metabolomics approach to distinguish between two dietary patterns with the purpose of developing a compliance measure( Reference Andersen, Rinnan and Manach 37 ). In a parallel intervention study, 181 participants were randomly assigned to follow a New Nordic Diet or an Average Danish Diet. The 24 h  urine samples were collected, analysed by ultra-performance liquid chromatography–quadruple time-of-flight–MS and partial least-squares discriminant analysis was applied to develop a compliance model for Average Danish Diet and New Nordic Diet based on the most discriminative features detected in urine. This resulted in a robust model with a misclassification rate of 19 %( Reference Andersen, Rinnan and Manach 37 ). Metabolites characterising the Average Danish Diet and the New Nordic Diet are listed in Table 3. The present study demonstrates the potential of metabolomics in discovering biomarkers indicative of dietary patterns, but furthermore it highlights a promising approach that may be used to develop compliance measures that cover the most important discriminant metabolites of complex diets.

Limitations of current approaches/study designs

In general, metabolomic-based approaches have produced reasonably robust models for dietary biomarker identification. However, following the discovery of any biomarker, validation in an independent study is critical to enable the generalisability of the results. This validation step is essential because factors which may not be present in traditional dietary assessment methods, including genetic factors, lifestyle and physiological factors, dietary factors, the biological sample or the analytic methodology could skew biomarker measures of dietary intake( Reference Jenab, Slimani and Bictash 38 ). For many of the study designs discussed, validation of the biomarker is often absent, making it difficult for the translation of these biomarkers into practice.

It has been proposed that the confirmation of dietary biomarkers should occur in two stages, firstly the dose–response effect should be included in intervention studies and secondly the suitability of the candidate biomarker in a free-living population should be investigated using a (controlled) habitual diet( Reference Kuhnle 39 ). Evaluation of the dose–response relationship is critical as it allows for the assessment of the suitability of the biomarker over a range of intakes( Reference Scalbert, Brennan and Manach 20 ). Unfortunately, in many studies, this important step is often absent. Biomarkers identified using samples from cohort studies do not assess the direct relationships of food amounts consumed and levels of biomarkers and do not demonstrate responsiveness to intakes, therefore the relationship is only an association( Reference Brennan, Gibbons and O'Gorman 16 ). Such studies should ideally be combined with intervention studies to demonstrate direct relationships and dose–response relationships. Conversely, dietary biomarkers identified within acute intervention studies advantageously allow for the examination of dose–response relationships; however, to date few studies have incorporated such designs.

When using self-reporting dietary data from cohort studies in the biomarker discovery process, one should be aware of reporting errors and the potential for missing important correlations and attenuation of results. May et al. investigated the metabolomic profiles of participants consuming a high-phytochemical diet compared with a diet without fruit and vegetables in a randomised controlled trial and also investigated the metabolomic profiles of participants in a cross-sectional study, where high and low fruit and vegetable diets were identified based on 3-d food records and FFQ. The intervention study found forty-six putatively annotated ions, with MS/MS fragment ion support that were differentially abundant between the two intervention diets; however, within the cross-sectional study only one compound annotated with MS/MS support was identified using the 3-d food records and there were no metabolites that significantly separated groups based on FFQ data( Reference May, Navarro and Ruczinski 40 ). This therefore demonstrates the drawbacks of using self-reported data in dietary biomarker discovery. Furthermore, when using cohort studies to identify or confirm biomarkers it is imperative that it is acknowledged that many of the foods consumed are highly correlated and therefore biomarkers identified may not be specific to the particular food of interest( Reference Scalbert, Brennan and Manach 20 ). Following identification of putative biomarkers from cohort studies we recommend that the relationship is confirmed using an intervention study in a dose–response manner where the sensitivity and specificity of the biomarkers can also be assessed. The importance of such a step is key to the validation of the biomarkers and important to support their use.

Use of acute and medium-term interventions is not without limitations in terms of dietary biomarker identification: many of the biomarkers identified using this approach are markers of acute intake. For example, proline betaine is excreted rapidly in urine and excretion is almost complete ≤24 h( Reference Heinzmann, Brown and Chan 22 ). These acute biomarkers may therefore only be valid for people that regularly and frequently consume the particular foods. The identification of dietary biomarkers that reflect habitual intake requires longer-term studies. Furthermore, it must also be noted that the majority of the acute and medium-term intervention study designs involve only a small number of participants( Reference Heinzmann, Brown and Chan 22 , Reference Pujos-Guillot, Hubert and Martin 24 , Reference Heinzmann, Holmes and Kochhar 41 ). The proposed dietary biomarkers identified using these approaches therefore cannot always be extrapolated to population studies in free-living individuals. However, this can be in part be dealt with by confirmation in cohort studies with a diverse range of characteristics.

While considering the earlier described limitations in study designs, there is also the need for development of databases and software tools to advance the interpretation of metabolomics results and therefore enhance the utility of dietary biomarkers in nutrition research. Current databases such as the Human Metabolome Database provides access to an online database containing detailed information about small molecule metabolites (>40 000) found in the human body( Reference Wishart, Jewison and Guo 42 ). Since it was first described in 2007, it is constantly being expanded and updated and has become a valuable resource that contains spectroscopic, quantitative, analytic and physiological information about human metabolites( Reference Wishart, Jewison and Guo 42 ). The Food Metabolome Database is another database of >28 000 food constituents that contains information about food sources and food concentrations( Reference Wishart and Suhre 43 ). This resource provides an aid for the identification of new metabolites that are reflective of food intake. While this resource is valuable, the identification of metabolites originating from food remains difficult and there is a need for sharing of databases to aid identification. Most recently, a comprehensive and electronically accessible human urine metabolome database, which includes quantitative concentrations of metabolites in urine samples was established( Reference Bouatra, Aziat and Mandal 44 ). This database also represents a significant development and resource for biomarker identification and quantification. Other new software tools include BAYESIL; this system provides fully automated and fully quantitative NMR-based metabolomics of complex mixtures( Reference Ravanbakhsh, Liu and Bjordahl 45 ). This will have a significant impact on NMR spectroscopy and NMR-based metabolomics.

Conclusion

The use of dietary biomarkers in nutrition research holds great promise. However, prior to having a suite of reliable dietary biomarkers that could be used in nutrition research a number of validation steps need to considered. Furthermore, the challenges identified in this review need to be acknowledged and addressed. Appropriate validation steps are essential, otherwise the robustness of biomarkers will remain uncertain and the translation of these biomarkers into practice will be challenging. Longer-term studies are also needed for the identification of dietary biomarkers reflective of habitual dietary intake. Until well-validated biomarkers are identified it is unlikely we will see uptake by the research community of the emerging biomarkers. The challenge for the researchers working in this field, in the coming years, will be to develop a suite of well-validated biomarkers. To this end the Joint Programming Initiatives funded programme FoodBall will address some of these issues and pave the way forward (http://foodmetabolome.org/). They may also have the potential for the assessment of compliance to dietary interventions in both a clinical and a research setting. Ultimately these dietary biomarkers will be used to further elucidate the proposed links between certain foods and disease.

Financial Support

The authors would like to acknowledge the following funding: FP7 Project NutriTech (289511), SFI (14/JP-HDHL/B3075) and ERC (647783).

Conflict of Interest

None.

Authorship

H. G. drafted the outline of the manuscript, conducted the literature search and drafted the manuscript. L. B. was responsible for critically reviewing the manuscript. Both authors read and approved the final manuscript before submission.

References

1. Mozaffarian, D & Ludwig, DS (2010) Dietary guidelines in the 21st century – a time for food. JAMA 304, 681682.Google Scholar
2. Bingham, SA (2002) Biomarkers in nutritional epidemiology. Public Health Nutr 5, 821827.Google Scholar
3. Kotchen, TA, Cowley, AW & Frohlich, ED (2013) Salt in health and disease – a delicate balance. N Engl J Med 368, 12291237.Google Scholar
4. Feskens, EJ, Sluik, D, Fau - van Woudenbergh, GJ et al. (2013) Meat consumption, diabetes, and its complications. Curr Diab Rep 13, 298306.Google Scholar
5. Pan, A, Sun, Q, Bernstein, AM et al. (2011) Red meat consumption and risk of type 2 diabetes: 3 cohorts of US adults and an updated meta-analysis. Am J Clin Nutr 94, 10881096.Google Scholar
6. Hu, FB, Rimm, EB, Stampfer, MJ et al. (2000) Prospective study of major dietary patterns and risk of coronary heart disease in men. Am J Clin Nutr 72, 912921.Google Scholar
7. Cross, AJ, Leitzmann, MF, Gail, MH et al. (2007) A prospective study of red and processed meat intake in relation to cancer risk. PLoS Med 4, e325.Google Scholar
8. Appel, LJ, Moore, TJ, Obarzanek, E et al. (1997) A clinical trial of the effects of dietary patterns on blood pressure. N Engl J Med 336, 11171124.Google Scholar
9. Sacks, FM, Svetkey, LP, Vollmer, WM et al. (2001) Effects on blood pressure of reduced dietary sodium and the dietary approaches to stop hypertension (DASH) diet. N Engl J Med 344, 310.Google Scholar
10. Estruch, R, Ros, E, Salas-Salvado, J et al. (2013) Primary prevention of cardiovascular disease with a Mediterranean diet. N Engl J Med 368, 12791290.Google Scholar
11. Salas-Salvadó, J, Bulló, M, Babio, N et al. (2011) Reduction in the incidence of Type 2 diabetes with the mediterranean diet: results of the PREDIMED-Reus nutrition intervention randomized trial. Diab Care 34, 1419.Google Scholar
12. Mozaffarian, D, Appel, LJ & Van Horn, L (2011) Components of a cardioprotective diet: new insights. Circulation 123, 28702891.Google Scholar
13. Kipnis, V, Midthune, D, Freedman, L et al. (2002) Bias in dietary-report instruments and its implications for nutritional epidemiology. Public Health Nutr 5, 915923.Google Scholar
14. Dhurandhar, NV, Schoeller, D, Brown, AW et al. (2015) Energy balance measurement: when something is not better than nothing. Int J Obes 39, 11091113.Google Scholar
15. Marshall, JR & Chen, Z (1999) Diet and health risk: risk patterns and disease-specific associations. Am J Clin Nutr 69, 1351s1356s.Google Scholar
16. Brennan, L, Gibbons, H & O'Gorman, A (2015) An overview of the role of metabolomics in the identification of dietary biomarkers. Curr Nutr Rep 4, 304312.Google Scholar
17. Tasevska, N, Runswick, SA, McTaggart, A et al. (2005) urinary sucrose and fructose as biomarkers for sugar consumption. Cancer Epidemiol Biomarkers Prev 14, 12871294.Google Scholar
18. Potischman, N (2003) Biologic and methodologic issues for nutritional biomarkers. J Nutr 133, 875s880s.Google Scholar
19. Wishart, DS (2008) Metabolomics: applications to food science and nutrition research. Trends Food Sci Technol 19, 482493.Google Scholar
20. Scalbert, A, Brennan, L, Manach, C et al. (2014) The food metabolome: a window over dietary exposure. Am J Clin Nutr 99, 12861308.Google Scholar
21. Atkinson, W, Downer, P, Lever, M et al. (2007) Effects of orange juice and proline betaine on glycine betaine and homocysteine in healthy male subjects. Eur J Nutr 46, 446452.Google Scholar
22. Heinzmann, SS, Brown, IJ, Chan, Q et al. (2010) Metabolic profiling strategy for discovery of nutritional biomarkers: proline betaine as a marker of citrus consumption. Am J Clin Nutr 92, 436443.Google Scholar
23. Lloyd, AJ, Beckmann, M, Favé, G et al. (2011) Proline betaine and its biotransformation products in fasting urine samples are potential biomarkers of habitual citrus fruit consumption. Br J Nutr 106, 812824.Google Scholar
24. Pujos-Guillot, E, Hubert, J, Martin, JF et al. (2013) Mass spectrometry-based metabolomics for the discovery of biomarkers of fruit and vegetable intake: citrus fruit as a case study. J Proteome Res 12, 16451659.Google Scholar
25. Edmands, WMB, Beckonert, OP, Stella, C et al. (2011) Identification of human urinary biomarkers of cruciferous vegetable consumption by metabonomic profiling. J Proteome Res 10, 45134521.Google Scholar
26. Andersen, M-BS, Reinbach, HC, Rinnan, Å et al. (2013) Discovery of exposure markers in urine for Brassica-containing meals served with different protein sources by UPLC–qTOF-MS untargeted metabolomics. Metabolomics 9, 984997.Google Scholar
27. Andersen, MB, Kristensen, M, Manach, C et al. (2014) Discovery and validation of urinary exposure markers for different plant foods by untargeted metabolomics. Anal Bioanal Chem 406, 18291844.Google Scholar
28. Lloyd, AJ, Fave, G, Beckmann, M et al. (2011) Use of mass spectrometry fingerprinting to identify urinary metabolites after consumption of specific foods. Am J Clin Nutr 94, 981991.Google Scholar
29. Stella, C, Beckwith-Hall, B, Cloarec, O et al. (2006) Susceptibility of human metabolic phenotypes to dietary modulation. J Proteome Res 5, 27802788.Google Scholar
30. Ross, AB, Svelander, C, Undeland, I et al. (2015) Herring and beef meals lead to differences in plasma 2-aminoadipic acid, β-alanine, 4-hydroxyproline, cetoleic acid, and docosahexaenoic acid concentrations in overweight men. J Nutr 11, 24562463.Google Scholar
31. Gibbons, H, McNulty, BA, Nugent, AP et al. (2015) A metabolomics approach to the identification of biomarkers of sugar-sweetened beverage intake. Am J Clin Nutr 101, 471477.Google Scholar
32. Guertin, KA, Moore, SC, Sampson, JN et al. (2014) Metabolomics in nutritional epidemiology: identifying metabolites associated with diet and quantifying their potential to uncover diet-disease relations in populations. Am J Clin Nutr 101, 10001011.Google Scholar
33. Wittenbecher, C, Mühlenbruch, K, Kröger, J et al. (2015) Amino acids, lipid metabolites, and ferritin as potential mediators linking red meat consumption to type 2 diabetes. Am J Clin Nutr 101, 12411250.Google Scholar
34. Garcia-Aloy, M, Llorach, R, Urpi-Sarda, M et al. (2014) Nutrimetabolomics fingerprinting to identify biomarkers of bread exposure in a free-living population from the PREDIMED study cohort. Metabolomics 11, 155165.Google Scholar
35. O'Sullivan, A, Gibney, MJ & Brennan, L (2011) Dietary intake patterns are reflected in metabolomic profiles: potential role in dietary assessment studies. Am J Clin Nutr 93, 314321.Google Scholar
36. O'Gorman, A, Morris, C, Ryan, M et al. (2014) Habitual dietary intake impacts on the lipidomic profile. J Chromatogr B, Anal Technol Biomed Life Sci 966, 140146.Google Scholar
37. Andersen, M-BS, Rinnan, Å, Manach, C et al. (2014) Untargeted metabolomics as a screening tool for estimating compliance to a dietary pattern. J Proteome Res 13, 14051418.Google Scholar
38. Jenab, M, Slimani, N, Bictash, M et al. (2009) Biomarkers in nutritional epidemiology: applications, needs and new horizons. Hum Genet 125, 507525.Google Scholar
39. Kuhnle, GG (2012) Nutritional biomarkers for objective dietary assessment. J Sci Food Agric 92, 11451149.Google Scholar
40. May, DH, Navarro, SL, Ruczinski, I et al. (2013) Metabolomic profiling of urine: response to a randomised, controlled feeding study of select fruits and vegetables, and application to an observational study. Br J Nutr 110, 17601770.Google Scholar
41. Heinzmann, SS, Holmes, E, Kochhar, S et al. (2015) 2-Furoylglycine as a candidate biomarker of coffee consumption. J Agric Food Chem 63, 86158621.Google Scholar
42. Wishart, DS, Jewison, T, Guo, AC et al. (2013) HMDB 3.0 – the human metabolome database in 2013. Nucleic Acids Res 41, D801D807.Google Scholar
43. Wishart, D (2012) Systems biology resources arising from the human metabolome project. In Genetics Meets Metabolomics: from Experiment to Systems Biology, pp. 157175 [Suhre, K, editor]. New York, NY: Springer New York.Google Scholar
44. Bouatra, S, Aziat, F, Mandal, R et al. (2013) The human urine metabolome. PLoS ONE 8, e73076.Google Scholar
45. Ravanbakhsh, S, Liu, P, Bjordahl, TC et al. (2015) Accurate, fully-automated nmr spectral profiling for metabolomics. PLoS ONE 10, e0124219.Google Scholar
46. Lang, R, Wahl, A, Stark, T et al. (2011) Urinary N-methylpyridinium and trigonelline as candidate dietary biomarkers of coffee consumption. Mol Nutr Food Res 55, 16131623.Google Scholar
47. van Velzen, EJJ, Westerhuis, JA, van Duynhoven, JPM et al. (2009) Phenotyping tea consumers by nutrikinetic analysis of polyphenolic end-metabolites. J Proteome Res 8, 33173330.Google Scholar
48. Daykin, CA, Duynhoven, JPMV, Groenewegen, A et al. (2005) Nuclear magnetic resonance spectroscopic based studies of the metabolism of black tea polyphenols in humans. J Agric Food Chem 53, 14281434.Google Scholar
49. Van Dorsten, FA, Daykin, CA, Mulder, TPJ et al. (2006) Metabonomics approach to determine metabolic differences between green tea and black tea consumption. J Agric Food Chem 54, 69296938.Google Scholar
50. Wang, Y, Tang, H, Nicholson, JK et al. (2005) A Metabonomic strategy for the detection of the metabolic effects of chamomile (Matricaria recutita L.) ingestion. J Agric Food Chem 53, 191196.Google Scholar
51. Tulipani, S, Llorach, R, Jáuregui, O et al. (2011) Metabolomics unveils urinary changes in subjects with metabolic syndrome following 12-week nut consumption. J Proteome Res 10, 50475058.Google Scholar
52. Cross, AJ, Major, JM & Sinha, R (2011) Urinary biomarkers of meat consumption. Cancer Epidemiol Biomarkers Prev 20, 11071111.Google Scholar
53. Lehtonen, H-M, Lindstedt, A, Järvinen, R et al. (2013) 1H NMR-based metabolic fingerprinting of urine metabolites after consumption of lingonberries (Vaccinium vitis-idaea) with a high-fat meal. Food Chem 138, 982990.Google Scholar
54. Vázquez-Fresno, R, Llorach, R, Alcaro, F et al. (2012) 1H-NMR-based metabolomic analysis of the effect of moderate wine consumption on subjects with cardiovascular risk factors. Electrophoresis 33, 23452354.Google Scholar
55. van Dorsten, FA, Grün, CH, van Velzen, EJJ et al. (2010) The metabolic fate of red wine and grape juice polyphenols in humans assessed by metabolomics. Mol Nutr Food Res 54, 897908.Google Scholar
56. Jacobs, DM, Fuhrmann, JC, van Dorsten, FA et al. (2012) Impact of short-term intake of red wine and grape polyphenol extract on the human metabolome. J Agric Food Chem 60, 30783085.Google Scholar
57. Johansson-Persson, A, Barri, T, Ulmius, M et al. (2013) LC–QTOF/MS metabolomic profiles in human plasma after a 5-week high dietary fiber intake. Anal Bioanal Chem 405, 47994809.Google Scholar
58. Rasmussen, LG, Winning, H, Savorani, F et al. (2012) Assessment of dietary exposure related to dietary GI and fibre intake in a nutritional metabolomic study of human urine. Genes Nutr 7, 281293.Google Scholar
59. Bondia-Pons, I, Barri, T, Hanhineva, K et al. (2013) UPLC–QTOF/MS metabolic profiling unveils urinary changes in humans after a whole grain rye versus refined wheat bread intervention. Mol Nutr Food Res 57, 412422.Google Scholar
60. Beckmann, M, Lloyd, AJ, Haldar, S et al. (2013) Hydroxylated phenylacetamides derived from bioactive benzoxazinoids are bioavailable in humans after habitual consumption of whole grain sourdough rye bread. Mol Nutr Food Res 57, 18591873.Google Scholar
61. Hjerpsted, JB, Ritz, C, Schou, SS et al. (2014) Effect of cheese and butter intake on metabolites in urine using an untargeted metabolomics approach. Metabolomics 10, 11761185.Google Scholar
62. Zheng, H, Yde, CC, Clausen, MR et al. (2015) Metabolomics investigation to shed light on cheese as a possible piece in the french paradox puzzle. J Agric Food Chem 63, 28302839.Google Scholar
63. Lloyd, AJ, Beckmann, M, Haldar, S et al. (2013) Data-driven strategy for the discovery of potential urinary biomarkers of habitual dietary exposure. Am J Clin Nutr 97, 377389.Google Scholar
64. Rothwell, JA, Fillâtre, Y, Martin, JF et al. (2014) New biomarkers of coffee consumption identified by the non-targeted metabolomic profiling of cohort study subjects. PLoS ONE 9, e93474.Google Scholar
65. Myint, T, Fraser, GE, Lindsted, KD et al. (2000) Urinary 1-methylhistidine is a marker of meat consumption in black and in white California seventh-day adventists. Am J Epidemiol 152, 752755.Google Scholar
66. Edmands, WMB, Ferrari, P, Rothwell, JA et al. (2015) Polyphenol metabolome in human urine and its association with intake of polyphenol-rich foods across European countries. Am J Clin Nutr 102, 905913.Google Scholar
67. Garcia-Aloy, M, Llorach, R, Urpi-Sarda, M et al. (2014) Novel multimetabolite prediction of walnut consumption by a urinary biomarker model in a free-living population: the PREDIMED study. J Proteome Res 13, 34763483.Google Scholar
68. Bouchard-Mercier, A, Paradis, A-M, Rudkowska, I et al. (2013) Associations between dietary patterns and gene expression profiles of healthy men and women: a cross-sectional study. Nutr J 12, 113.Google Scholar
69. Altmaier, E, Kastenmüller, G, Römisch-Margl, W et al. (2010) Questionnaire-based self-reported nutrition habits associate with serum metabolism as revealed by quantitative targeted metabolomics. Eur J Epidemiol 26, 145156.Google Scholar
70. Peré-Trepat, E, Ross, AB, Martin, F-P et al. (2010) Chemometric strategies to assess metabonomic imprinting of food habits in epidemiological studies. Chemometrics Intell Lab Syst 104, 95100.Google Scholar
71. Floegel, A, von Ruesten, A, Drogan, D et al. (2013) Variation of serum metabolites related to habitual diet: a targeted metabolomic approach in EPIC-Potsdam. Eur J Clin Nutr 67, 11001108.Google Scholar
Figure 0

Table 1. Summary of putative biomarkers identified using a metabolomics approach in intervention studies

Figure 1

Table 2. Summary of putative biomarkers identified using a metabolomics approach in cohort studies

Figure 2

Table 3. Summary of putative biomarkers identified using dietary patterns and metabolomic profiles