Metabolomics: applications to food science and nutrition research

Trends in Food Science & Technology 19 (2008) 482e493

Review

Metabolomics: applications to food science and nutrition research David S. Wisharta,b,c,* a

Department of Computing Science, University of Alberta, Edmonton, AB, Canada T6G 2E8 b Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada T6G 2E8 c National Institute for Nanotechnology, 11421 Saskatchewan Drive, Edmonton, AB, Canada T6G 2M9 (Department of Computing Science, University of Alberta, 2-21 Athabasca Hall, Edmonton, AB T6G 2E8, Canada. Tel.: D1-780-492-0383; fax: D1-780-4921071; e-mail: [email protected]) Metabolomics is an emerging field of ‘‘omics’’ research that focuses on high-throughput characterization of small molecule metabolites in biological matrices. As such, metabolomics is ideally positioned to be used in many areas of food science and nutrition research. This review focuses on the recent trends and potential applications of metabolomics in four areas of food science and technology: (1) food component analysis; (2) food quality/authenticity assessment; (3) food consumption monitoring; and (4) physiological monitoring in food intervention or diet challenge studies.

Introduction Metabolomics is a relatively new field of ‘‘omics’’ research concerned with the high-throughput identification and quantification of small molecule (<1500 Da) metabolites in the metabolome (German, Hammock, & Watkins, 2005). The metabolome is formally defined as the collection of all small molecule metabolites or chemicals that can be found in a cell, organ or organism. These small

* Corresponding author. 0924-2244/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.tifs.2008.03.003

molecules can include a range of endogenous and exogenous chemical entities such as peptides, amino acids, nucleic acids, carbohydrates, organic acids, vitamins, polyphenols, alkaloids, minerals and just about any other chemical that can be used, ingested or synthesized by a given cell or organism. Metabolomics, which is also known as metabonomics (Nicholson, Lindon, & Holmes, 1999) or metabolic profiling (Niwa, 1986), only became possible as a result of recent technological breakthroughs in small molecule separation and identification. These include robust, high-resolution mass spectrometry (MS) instruments for precise mass determination, high-resolution, high-throughput NMR (nuclear magnetic resonance) spectrometers, capillary electrophoresis (CE) and ultra-high pressure liquid chromatography (UPLC and HPLC) systems for rapid compound separation, as well as new software programs to rapidly process spectral or chromatographic patterns (Dunn, Bailey, & Johnson, 2005; Wishart et al., 2001). Equally important to the development of metabolomics has been the compilation of electronic databases containing both descriptive and spectral information on the constituent chemicals found in different metabolomes (Cotter, Maer, Guda, Saunders, & Subramaniam, 2006; Kopka et al., 2005; Moco et al., 2006; Smith et al., 2005; Wishart et al., 2007). These hardware and software innovations have made it possible to detect and characterize not just one or two small molecules at a time but dozens of small molecule metabolites in just minutes (Dunn et al., 2005; Moco et al., 2006; Wishart et al., 2001). As with many new ‘‘omics’’ fields, there is often a tendency towards rapid technological proliferation followed by a longer period of consolidation and methodological refinement. Metabolomics is no exception. Many different metabolomics technologies exist (NMR, GC-MS, LC-MS, CE-MS) and many different analytical approaches can be useddeach with their own advantages and disadvantages (Table 1). In fact, over the past few years, two distinct schools of thought have emerged for processing and interpreting metabolomic data (Fig. 1). In one version (called the chemometric approach) chemical compounds are not generally identifieddonly their spectral patterns and intensities are recorded, statistically compared and used to identify the relevant spectral features that distinguish sample classes (Nicholson et al., 1999; Trygg, Holmes, & Lundstedt, 2007). These statistical comparisons and feature

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

483

Table 1. Comparison of different metabolomics technologies Technology

Advantages

Disadvantages

NMR Spectroscopy

                         

     

Not very sensitive Expensive instrumentation Large instrument footprint Cannot detect or ID salts and inorganic ions Cannot detect non-protonated compounds Requires larger (0.5 mL) samples

     

Sample not recoverable Requires sample derivitization Requires separation Slow (20-30 min/sample) Cannot be used in imaging Novel compound ID is difficult

       

Sample not recoverable Not very quantitative Expensive instrumentation Slow (20-30 min/sample) Poor separation resolution and reproducibility (vs. GC) Less robust instrumentation than NMR or GC-MS Limited body of software and databases for metabolite ID Novel compound ID is difficult

GC-Mass Spectrometry

LC-Mass Spectrometry

Quantitative Non-destructive Fast (2-3 min/sample) Requires no derivitization Requires no separation Detects all organic classes Allows ID of novel chemicals Robust, mature technology Can be used for metabolite imaging (fMRI) Large body of software and databases for metabolite ID Compatible with liquids and solids Robust, mature technology Relatively inexpensive Quantitative (with calibration) Modest sample size need Good sensitivity Large body of software and databases for metabolite ID Detects most organic and some inorganic molecules Excellent separation reproducibility Superb sensitivity Very flexible technology Detects most organic and some inorganic molecules Minimal sample size requirement Can be used in metabolite imaging (MALDI) Can be done without separation (direct injection) Has potential for detecting largest portion of metabolome

identification techniques usually involve unsupervised clustering (Principal Component Analysis or PCA) or supervised classification (Partial Least Squares Discriminant Analysis or PLS-DA). Formally, PCA is a dimensional reduction technique that allows one to easily plot, visualize and cluster multiple metabolomic data sets based on linear combinations (known as principal axes) of their shared features. As a clustering technique, PCA is most commonly used to identify how one sample is different from another, which variables contribute most to this difference, and whether those variables contribute in the same way (i.e. are correlated) or independently (i.e. uncorrelated) from each other. In contrast to PCA, PLS-DA is a supervised classification technique that can be used to enhance the separation between groups of observations by rotating PCA components such that a maximum separation among classes is obtained. The basic principles behind PLS-DA are similar to that of PCA, but in PLS-DA a second piece of information is used, namely, the labeled set of class identities. While chemometric approaches like PCA and PLS-DA, on their own, do not permit the direct identification or quantification of compounds they still allow an unbiased (or untargeted), chemically comprehensive comparison to be made among different samples. In the other approach to metabolomics (variously called quantitative metabolomics or targeted profiling) the focus is on attempting to identify and/or quantify as many compounds in the sample as possible. This usually done by

comparing the sample’s NMR or MS spectrum to a spectral reference library obtained from pure compounds (Serkova et al., 2007; Weljie, Newton, Mercier, Carlson, & Slupsky, 2006; Wishart et al., 2001). Once the constituent compounds are identified and quantified, the data are then statistically processed (using PCA or PLS-DA) to identify the most important biomarkers or informative metabolic pathways (Weljie et al., 2006). Depending on the objectives and instrumental capacity, quantitative metabolomics may be either targeted (selective to certain classes of compounds such as lipids or polyphenols) or comprehensive (covering all or almost all detectable metabolites). Both chemometric and quantitative metabolomics have their advantages and disadvantages, but given the importance attached to bioactive compound identification, there is a growing preference for quantitative metabolomics in many areas of food science and nutrition research (Gibney et al., 2005; Moco et al., 2006). Until recently, most of the work in metabolomics or metabonomics has focused primarily on clinical or pharmaceutical applications such as drug discovery (Kell, 2006; Watkins & German, 2002), drug assessment (Lindon, Holmes, Bollard, Stanley, & Nicholson, 2004), clinical toxicology (Griffin & Bollard, 2004; Schnackenberg & Beger, 2006) and clinical chemistry (Moolenaar, Engelke, & Wevers, 2003; Wishart et al., 2001). However, over the past few years, metabolomics has also emerged as a field of increasing interest to food and nutrition scientists (Gibney et al., 2005). It’s not hard to understand why.

484

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

Fig. 1. A) Schematic illustration of the chemometric approach to metabolomics. In this example multiple blood samples are analyzed via MS or NMR and then the spectra are processed using principal component analysis. After identifying significant differences, the most informative peaks in the spectra are identified using a variety of methods. B) Schematic illustration of the quantitative approach to metabolomics. In this example the biofluid spectrum is annotated and most (all) of the compounds in the sample are identified and quantified. This information is then used to perform multivariate statistical analysis allowing the most important biomarkers and pathways to be identified.

Because metabolomics allows the simultaneous characterization of large numbers of chemicals in biological matrices, it actually offers food and nutrition scientists a superb opportunity to acquire a far more detailed and comprehensive molecular picture of food composition, of food consumption and of the molecular biological consequences of different diets. In other words, metabolomics essentially opens the door to studying many aspects of ‘‘molecular nutrition’’, including: (1) food component analysis; (2) food quality/authenticity detection; (3) food

consumption monitoring; and (4) physiological monitoring in food intervention or diet challenge studies. Metabolomics in Food Component Analysis Traditionally food component analysis involves identifying and classifying food constituents into very broad categories such as proteins, fats, carbohydrates, fiber, vitamins, trace elements, solids and/or ash. However, with the advent of metabolomics, foods and beverages are now being analyzed with considerably more chemical detail,

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

with hundreds or even thousands of distinct chemical identities being detected and/or identified in certain foods (Moco et al., 2006; Ninonuevo et al., 2006; Wishart et al., 2007). The potential to chemically ‘‘deconstruct’’ foods and beverages into their chemical constituents offers food chemists a unique opportunity to understand the molecular details of what gives certain foods and drinks their unique taste, texture, aroma or color. An equally compelling opportunity exists for nutritional scientists to precisely identify bioactive ingredients in foods and better understand their potentially beneficial (or harmful) consequences. From the perspective of a metabolomics researcher, most foods can essentially be viewed as complex chemical mixtures consisting of various metabolites and chemical additives in a solid, semi-solid or liquid matrix. Some ‘‘manufactured’’ foods consist of just 10e20 different chemicals (artificial juices, soft drinks, purified vegetable oils) while other foods consist of hundreds of compounds (milk, cheese) and still others may have thousands of compounds (fruits, meats and most prepared foods). The number and diversity of chemical compounds in foods can be better appreciated if we consider the following numbers: (1) the human metabolome (which is expected to be similar to what could be found in most red meats) contains more than 3000 detectable (>1 pM) endogenous metabolites (Wishart et al., 2007); (2) the milk metabolome contains more than 200 different oligosaccharides (Ninonuevo et al., 2006); (3) the edible plant metabolome likely consists of more than 10,000 different detectable compounds with more than 2000 nutrient metabolites and more than 8000 non-nutrient phytochemicals having already been identified (Ehrman et al., 2007); and (4) the U.S. FDA maintains a list of more than 3000 approved food additives (of which 2000 are well-defined chemicals) that can be found in processed foods. This compilation is often referred to as ‘‘Everything’’ Added to Food in the United States (EAFUS) (http://www.cfsan.fda.gov/wdms/eafus.html). Taken together, this collection of 15,000þ food components (which consists of both artificial and natural nutrients as well as artificial and natural non-nutrients) is often referred to as the nutrition metabolome (Gibney et al., 2005) or simply the food metabolome. The food metabolome is characterized not only by considerable chemical diversity (>100 major chemical classes), but also by considerable variations in chemical abundance. Indeed, for any given food item these constituents may range in concentration from femtomolar (vitamins, food additives and certain phytochemicals) to millimolar (sugars, salts) values. Detecting, identifying and quantifying so many different compounds over such enormous (1 trillion-fold) concentration ranges represents a significant technical and computational challenge. Nevertheless, with the advent of new technologies and expanded spectral libraries, metabolomics could add significant new knowledge to the study of health and disease processes.

485

In particular, an impressive number of natural foods, spices and beverages have already been the subject of detailed metabolomic-based component analysis, including milk (Agabrielet al., 2007; Hu, Furihata, Kato, & Tanokura, 2007; Ninonuevo et al., 2006), grapes (Mattivi, Guzzon, Vrhovsek, Stefanini, & Velasco, 2006; Pereira et al., 2006), tomatoes and tomato juice (Moco et al., 2006; Tiziani, Schwartz, & Vodovotz, 2006), rhubarb (Ye, Han, Chen, Zheng, & Guo, 2007), beer (Almeida et al., 2006; Nord, Vaag, & Duus, 2004), celery seeds (Kitajima, Ishikawa, & Satoh, 2003), coriander (Ishikawa, Kondo, & Kitajima, 2003) as well as many other herbs and spices (Ishikawa, Takayanagi, & Kitajima, 2002; Matsumura, Ishikawa, & Kitajima, 2002). These analyses used a combination of NMR, GC-MS, LC-MS and CE techniques to identify up to 100 different phytochemicals (Ye et al., 2007) or 200 different carbohydrates (Ninonuevo et al., 2006) in selected fruit, vegetable or beverage samples. The use of multiple separation and detection technologies generally allows for a far broader range of metabolites to be detected, with NMR being particularly useful for detecting high abundance polar metabolites and GC-MS or LCMS being best for low abundance non-polar or semi-polar metabolites. It is also worth noting that the majority of food-based metabolomics studies published to date have focused on liquids or liquid extracts (i.e. beverages) rather than an analysis of food solids. This is primarily because most nutrients can be easily extracted into liquids and that most chemical detection technologies are optimized for working with liquids. Recent metabolomic investigations of milk (both human and bovine) have been particularly revealing. These studies have shown the potential of high resolution Fourier Transform (FT) mass spectrometry to identify more than 200 different oligosaccharides, many of which had not been previously characterized before (Ninonuevo et al., 2006). LC-MS methods has also been used to characterize some 15 different vitamins, terpenoids and other phytochemicals in bovine milk as a way of tracking herd characteristics, feeding practices and monitoring milk quality (Agabriel et al., 2007). In addition to these MS-based studies, NMR methods have also been shown to permit rapid, nondestructive identification and quantification of nearly a dozen different milk fatty acids, free sugars, acetylcarbohydrates and other small molecule metabolites (Hu et al., 2007). Given the speed and accuracy with which many of the measurements were made, it seems likely that NMR spectroscopy could soon replace the more traditional and time-consuming techniques used to characterize milk. Studies of tomatoes and tomato juices using LC-MS methods (Moco et al., 2006) and NMR (Tiziani et al., 2006) have shown how these two methods can reveal complementary information about different constituents in tomatoes. LC-MS methods, in combination with appropriate reference databases, allow the ready identification of many (>60) non-polar, low abundance secondary

486

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

metabolites such as polyphenols, flavo-glucosides and alkaloids, (Moco et al., 2006). On the other hand, NMR was found to be particularly useful in the identification and quantification of the lycopene and carotenoid isomers that could not be distinguished by LC or LC-MS methods (Tiziani et al., 2006). Other NMR studies have identified and quantified many of the more abundant (mostly polar) compounds in tomatoes including sugars, flavanoid glycosides, amino acids and organic acids (Le Gall, Colquhoun, Davis, Collins, & Verhoeyen, 2003). Variations in some of these tomato-specific phytochemicals are largely responsible for the subtle variations in flavors, aromas, ripening stages and colors in tomatoes. They are also responsible for their potential prostate cancer-fighting properties (Tiziani et al., 2006). Variations in the more abundant, polar metabolites (sugars, acids, etc.) are responsible for the more obvious tomato taste variations. Over the coming years it is likely that food composition studies will become much more common with far more comprehensive metabolomic analyses being done of many of the economically or pharmaceutically more important fruits (apples, bananas, oranges, strawberries, pineapples, mangos, dates, pomegranates, blueberries), vegetables (potatoes, cabbage, carrots, avacados, spinach, broccoli, cauliflower, onions, corn), grains (wheat, barley, oats, flaxseed, rye), nuts (almonds, walnuts, peanuts, cashews), processed foods (cheeses, yogurt, wines, vegetable oils, fish oils), meats (beef, chicken, fishes and crustaceans) and, of course, many nutraceutical foods or beverages (coffee, chocolate, gingko, green tea, ginger, ginseng, garlic). It is also likely metabolomic-based food component studies will allow food scientists to more precisely follow the consequences of different preparation (frying vs. baking; steaming vs. boiling) and preservation (freezing, drying, smoking, refrigerating) processes on key food components. These studies will also help guide food scientists in the breeding, selection or modification of better crops, the breeding and feeding of livestock or fishstock as well as the preparation of many processed foods. Metabolomics in Food Quality/Authenticity Assessment The capacity to identify food constituents using metabolic profiling can also be used to assess both food adulteration and food quality. In particular, the detection of adulterated on contaminated food products exploits the fact that certain chemicals or certain concentrations of chemicals are quite characteristic of certain types of juices, extracts and oils (Ogrinc, Kosir, Spangenberg, & Kidric, 2003). Similar chemical composition characteristics can also be exploited to distinguish between food products with desirable characteristics that cannot otherwise be detected by flavor, aroma or color (such as unsaturated fat content, a favorable amino acid profile, increased vitamin content, enriched phytochemical content). Food quality assessment also impacts food quality control. In fact,

metabolomic techniques may find their greatest use in the food industry in monitoring quality control or batch-tobatch product reproducibility. To date most reported metabolomic studies on food adulteration and food quality have focused on characterizing oils or beverages, such as vegetable oils (Hutton, Garbow, & Hayes, 1999; Ogrinc et al., 2003; Prestes et al., 2007), fish oils (Aursand, Standal, & Axelson, 2007), fruit juices (Cuny et al., 2008; Cuny, Le Gall, Colquhoun, Lees, & Rutledge, 2007; Le Gall, Puaud, & Colquhoun, 2001), wines (Aursand et al., 2007) and beers (Almeida et al., 2006). Fruit juice adulteration is relatively common as it is very difficult to detect via taste or color. For instance, orange juice can be blended with lower-cost grapefruit juice without any obvious changes to flavor or color. However, the presence of grapefruit juice in a presumptively pure orange juice product can have serious consequences to individuals on certain medications. Grapefruit juice has a number of coumarin-like flavanoids and other powerful CYP450 inhibitors (Griennavar, Poulose, Jayaprakasha, Bhat, & Patil, 2006). Therefore, the detection and prevention of this kind of adulteration is particularly important. Traditional approaches to detect this kind of juice blending use fairly lengthy and labor intensive HPLC methods. Recently an NMR-based chemometric approach has been developed that uses Independent Component Analysis, a variant of PCA (Cuny et al., 2007). This approach is fundamentally based on analyzing selected regions of the 1H NMR spectrum that are known to contain distinguishing flavanoid glycoside signals The method was tested on 92 different juices, including 59 pure orange juices, 23 grapefruit juices and 10 blends. It was found to be quite accurate, with 98% of the samples being correctly classified. Furthermore, this NMR method took less than half the time of the standard chromatographic approaches. Another common issue with juice preparation is the distinction between freshly squeezed juices and those derived from pulp washes. As with grapefruit juice, pulp wash can be added to freshly squeezed orange juice as a way of reducing production costs, without being easily detected by taste, texture or color changes. To address this issue, Le Gall et al. (2001) used 1H NMR in combination with principal component analysis to distinguish between more than 300 samples of freshly squeezed and pulp-wash orange juice. The method was found to be nearly 95% accurate. Furthermore, a specific biomarker compound, dimethylproline, was found to exist at significantly higher levels in pulp-wash juice relative to freshly squeezed juice. The development of a test to detect this modified amino acid could make pulp-wash detection very simple and inexpensive. NMR has also been successfully used in monitoring batch-to-batch quality and production site differences in beer (Almeida et al., 2006). Large multi-national or pannational breweries must prepare their beers at many

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

different geographic locations. Given the desire to produce similar tasting and similar quality beers at these different sites, it is important to develop ways that can perform quality control not just at the level of color, alcohol content or specific gravity, but at a more detailed molecular level. The study by Almeida et al. (2006) used NMR in combination with principal component analysis to reveal that different production sites, presumably producing the same beer, could be easily distinguished based on their content of lactic acid, pyruvic acid, dextrans, adenosine, inosine, uridine, tyrosine and 2-phenylethanol. Quantifying these compounds also allowed the authors to identify one production site that exhibited much greater variability in these compounds (and therefore had less quality control) than the other production sites. These methods could also be adapted by other multinational corporations to improve quality control in soft drink production, juice production and vegetable oil manufacturing. Similar methods might also be used to monitor the quality control of functional foods and nutraceuticals (i.e. food extracts that are claimed to have positive medicinal effects) that are harvested from different geographic locations. To date, most published food/beverage adulteration studies have employed NMR in combination with PCA, artificial neural networks (ANNs) or other multivariate statistical techniques to identify and classify samples. However, it is also possible to use mass spectrometry to detect food adulteration. In particular, the use of isotope ratio mass spectrometry (IRMS) has long been used to detect the adulteration of wines and olive oils, as well as the detection of added beet sugar and exogenous citric acid to fruit juices (Ogrinc et al., 2003). IRMS is a specialized form of mass spectrometry, in which the MS instrument is used to measure the relative abundance of isotopes or isotope fractionation in a given sample. Isotope abundances can vary with the geographic location, physical source, metabolic reaction or age of the material being analyzed. Consequently isotope abundances, when measured against a known internal isotope reference (IIR) can be used to identify the biochemical or geographic origins of many compounds. Given the multinational nature of food production and processing along with the growing concerns over food safety and food quality, it is likely that more stringent and more precise food monitoring systems will have to be put in place over the coming years. The demonstrated potential of metabolomic techniques to rapidly and inexpensively identify adulterated beverages and oils, as well as their potential to accurately monitor previously undetectable quality control issues certainly suggests that metabolomics could soon play a major role in many aspects of food quality assessment and quality control. Metabolomics in Food Consumption Monitoring Monitoring or measuring food consumption is one of the most important and challenging tasks for nutritional

487

scientists. In an ideal world, nutrition scientists would like to control precisely what a person eats or monitor everything that a person has consumed, much the same way that we can control or monitor the diets of caged lab animals. This approach would certainly allow for important dieteresponse findings to be made. Unfortunately humans are not lab rats and closed-facility diet interventions with human subjects are incredibly expensive to perform. Most of the published findings on the positive or negative consequences of certain foods or diets are derived from epidemiological studies of large population cohorts who fill out food frequency questionnaires (FFQs). Unfortunately FFQ compliance for prospective diet studies and FFQ recall for retrospective diet studies has not been particularly good (Brantsaeter et al., 2007). Furthermore, these food frequency questionnaires do not allow for direct measures of plasma levels of circulating nutrients or micronutrients, which may vary from individual to individual by factors of 10 or more (Gibney et al., 2005; Hong & Mitchell, 2004; Zeisel, 2007). Given the shortcomings of FFQs, an emerging application of metabolomics is in food consumption monitoring. This particular approach takes advantage of the knowledge being acquired in food component analysis and food quality measurement (see above) and applies it to the analysis of blood, urine or saliva from volunteers in diet intervention studies. This is very similar, both in design and concept, to using metabolomics in drug safety or drug efficacy assessment (Lindon et al., 2004; Gibney et al., 2005). However, in the case of nutritional monitoring, the ‘‘drugs’’ are relatively ill-defined chemical mixtures of nutrients and non-nutrients (i.e. food). The utility of metabolomics-based food consumption monitoring comes not only through an improved read-out of food consumption patterns, but also through better interpretation of the true physiological consequences of a given diet (or drug). As will be discussed in the next section, metabolomic studies aimed at identifying physiological effects of food intervention trials must be able to distinguish spectral changes in the urine or blood arising from diet-specific non-nutrient ‘‘spill over’’ (such as peaks arising from flavanoids, alkaloids or food additives) from those spectral changes arising from diet-induced physiological changes, such as peaks arising from changed endogenous metabolites or endogenous signaling molecules. Failure to distinguish these differences will lead to serious misinterpretation of any diet intervention study. Indeed, if the diet-derived signals are not properly accounted for, all that one may learn from an extensive metabolomic study is the fact that two groups ate differentlydan obvious fact that needs no secondary confirmation. As a general rule, the best biofluid to study nutrient intake or to identify food-specific biomarkers is urine (Gibney et al., 2005; Walsh, Brennan, Malthouse, Roche, & Gibney, 2006). Urine is essentially the body’s liquid waste repository and any nutrient or non-nutrient that is not

488

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

needed or present in excess will find its way into the urine. Indeed, a major difference between urine and blood is the ratio of metabolites to non-metabolites (i.e. phytochemicals, cooking by-products, xenobiotics and other compounds that cannot be productively metabolized by the body). In almost every case reported to date, urine has a significantly higher proportion of non-metabolites than blood. For instance, a recent study on biomarkers of onion consumption found that 11 quercitin metabolites were detectable in urine at relatively high abundance, whereas only 5 lower abundance quercitin metabolites could be found in blood (Mullen, Boitier, Stewart, & Crozier, 2004). Table 2 provides a list of a number of the food-specific biomarkers that can be found in blood or urine. In some cases the mere presence of the compound indicates consumption of a certain food, in other cases, the concentration of the compound is significantly elevated for a period of

time. Most food-specific biomarkers are present in the blood or urine for 5e10 hours, with some persisting as long as 48 hours (Seeram et al., 2006). Many of these compounds are phytochemicals or non-nutrients that are not metabolized by the body. Others are actually secondary metabolites having been tagged for elimination as glycosylated, glucuronidated or sulfated conjugates. In many respects, the body treats these compounds much the same way as it treats drugs or other xenobiotics. Hence the same concepts behind drug testing and drug monitoring can be used in food consumption monitoring. The difference is that drugs are usually relatively pure compounds with relatively few metabolites. Foods are not pure compounds and consequently have many more constituents and many more metabolites. Some, however, are quite unique to certain foods or food processing events and these are the ones that are listed in Table 2.

Table 2. Summary of food consumption and diet biomarkers Food/diet

Biomarker

Biofluid (Level)

Reference(s)

Black tea

Gallic acid 4-O-methylgallic acid Epicatechin Kaempferol Gallic acid 4-O-methylgallic acid Caffeic acid Chlorogenic acid Naringenin Ethyl glucuronide Ethyl phosphate Ethyl sulfate Hippuric acid Phloretin Naringenin Hesperetin Ellagic acid Dimethylellagic acid glucuronide Naringenin Hesperetin Allylmercapturic acid

Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (decrease) Urine (increased) Urine (increased) Urine (increased) Blood (increase) Blood (increase)

Mennen et al., 2006 Mennen et al., 2006 Ito et al., 2005 Brantsaeter et al., 2007 Mennen et al., 2006 Mennen et al., 2006 Mennen et al., 2006 Mennen et al., 2006 Ito et al., 2005 Bicker et al., 2006 Bicker et al., 2006 Bicker et al., 2006 Bertram et al., 2007 Mennen et al., 2006 Mennen et al., 2006 Mennen et al., 2006 Seeram et al., 2006 Seeram et al., 2006

Urine (increased)

Lysinoalanine N(e)-fructolysine N(e)-carboxymethllysine Alpha-carotene Quercetin glucuronides Quercetin Methyl quercitin Trimethylamine Trimethylamine-oxide 1-methylhistidine Creatine Carnitine Trimethylamine-oxide Salicyluric acid Salicylic acid Lactate Alanine Citrate

Urine (increased)

Mennen et al., 2006 Brantsaeter et al., 2007 De Rooij, Boogaard, Rijksen, Commandeur, & Vermeulen, 1996 Somoza et al., 2006 Somoza et al., 2006 Somoza et al., 2006 Brantsaeter et al., 2007 Hong & Mitchell 2004 Hong & Mitchell 2004 Hong & Mitchell 2004 Lee et al., 2006 Lee et al., 2006 Myint et al., 2000 Stella et al., 2006 Stella et al., 2006 Stella et al., 2006 Lawrence et al., 2003 Lawrence et al., 2003 Zuppi et al., 1998 Zuppi et al., 1998 Zuppi et al., 1998

Wine Coffee Cocoa Alcohol

Milk Apple Grapefruit Orange Pomegranate

Citrus fruit Garlic Cooked protein

Cooked vegetables Cooked onions

Fish Meat-rich diet

Vegetarian diet Carbohydrate rich diet

Urine (increased)

Blood (increase) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased) Urine (increased)

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

In the not-too-distant future, this list will no doubt expand quite significantly as more compounds are identified and more foods are studied through different metabolomic approaches. It is also likely that food consumption biomarkers will evolve from being just one or two relatively non-specific compounds to a far more specific ‘‘profile’’ consisting of multiple compounds. Likewise the use of concentration data in conjunction with information on the time since, or the amount of, consumption should allow much more precise measures of metabolite flow and flux as well as the identification of individual differences in metabolite processing or absorption. In time, metabolomic techniques might be able to replace some of the short-term food consumption questionnaires used in prospective diet challenge or food intervention studies. Metabolomics in Physiological Monitoring of Diet and Nutrition Studies Over the past century great strides have been made by food scientists and biochemists in identifying the essential nutrients needed for human growth and viability. Improved dietary guidelines and mandated food supplementation with essential minerals and vitamins has been remarkably successful in treating most nutritional ‘‘deficiencies’’ (Zeisel, 2007). Today nutritional scientists are challenged with finding new ways of treating or preventing diseases brought on by nutritional ‘‘oversufficiencies’’ such as obesity, diabetes, chronic inflammation and cardiovascular disease (German et al., 2005). They are also being asked to identify bioactive food components that potentially increase life expectancy, reduce weight, enhance physical or mental performance and prevent diseases such as atherosclerosis, heart disease, cancer and arthritis. In addressing these challenges, nutrition scientists may choose one of two routes to explore the effect of diet or

489

bioactive food components on human health. One is through retrospective epidemiological studies that attempt to correlate long-term food consumption with health status or disease susceptibility. The other is through shorter-term prospective food intervention or diet challenge experiments. While many encouraging correlations between diet and improved health status have been identified through epidemiological studies (German et al., 2005; Gibney et al., 2005; Zeisel, 2007), these correlations do not necessarily imply causation. Consequently diet-challenge experiments, with a focus on quantitatively measuring molecular biomarkers and physiological consequences, have become the preferred route to exploring the causes behind the beneficial or detrimental effects of key nutrients. Genomic, proteomic and, most recently, metabolomic techniques are now commonly used in many diet-challenge or food intervention experiments (Gibney et al., 2005; Zeisel, 2007). Depending on the physiological response being measured, metabolomic assays can be used to measure small molecule biomarkers of oxidative stress, redox potential, anti-oxidant activity, inflammation and cardiovascular disease risk (Table 3). While proteomic and transcriptomic markers also exist for tracking these responses, the advantage of using small molecule biomarkers is that the same small molecule monitoring techniques can also be used to look at the short-term concentration changes of diet related chemicals or metabolites. In other words, only one technology platform needs to be employed to measure food inputs and physiological outputs. In contrast to the urinary polar compounds that can be used as biomarkers of food consumption (Table 2) the compounds that are used as biomarkers of physiological response to foods (Table 3) are typically nonpolar and most are found in blood. This is because blood is generally a more stable and reliable indicator of physiological

Table 3. Physiological response biomarkers in blood and urine Compound

Biofluid (level)

Indication

Reference(s)

8-isoprostane F2a

Urine (increased)

Thompson et al., 2006

8-hydroxy-2-deoxyguanosine Malondialdehyde (MDA) Glutathione (reduced) Hydrogen peroxide Low density lipoprotein (LDL)

Lymphocytes/Urine (increased) Urine (increased) Plasma (decreased) Urine (increased) Plasma (increased)

Lipid peroxidation Oxidative stress DNA oxidative damage Oxidative stress Oxidative stress Oxidative stress Increased risk for CVD

High density lipoprotein (HDL) Triacylglycerol (TAG) Homocysteine Total cholesterol alpha-tocopherol (Vitamin E) Ascorbic acid (Vitamin C) Thromboxane B2 Leukotriene B4 Prostaglandin E2 Uric Acid

Plasma Plasma Plasma Plasma Plasma Plasma Plasma Plasma Plasma Plasma

Increased risk for CVD Increased risk for CVD Increased risk for CVD Increased risk for CVD Decreased anti-oxidant potential Decreased anti-oxidant potential Inflammation Inflammation Inflammation Inflammation; Oxidative stress

Fardet et al., 2007 Fardet et al., 2007 Bogani et al., 2007 Castro, Barroso, & Sinnecker, 2005 Castro et al., 2005 Castro et al, 2005 Chiuve et al., 2007 Castro et al., 2005 Castilla et al., 2006 Castilla et al., 2006 Bogani et al., 2007 Garg et al., 2007 Sa´nchez-Moreno et al., 2004 Sa´nchez-Moreno et al., 2004

(decreased) (increased) (increased) (increased) (decreased) (decreased) (increased) (increased) (increased) (increased)

Thompson et al., 2006

490

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

processes. Furthermore, blood carries many more non-polar molecules than urine. Not all of the compounds listed in Table 3 have been traditionally detected by NMR or MS methods. In fact, some are frequently detected using immunoassays (thromboxanes, leukotrienes) or measured via centrifugation protocols (LDL, HDL). However, advances in MS-based lipidomic technologies (Wiest & Watkins, 2007) and lipoprotein characterization (Otvos, 2002) are now allowing these molecules to be detected and/or quantified using relatively standard, high-throughput MS or NMR techniques. One of the first studies to use metabolomic techniques in a diet-intervention trial involved studying the effects of soy-derived proteins and soy isoflavones (Solansky et al., 2003). 1H NMR and chemometric techniques were used to analyze the blood plasma changes in pre-menopausal women on a multi-week soy diet. The results showed that significant differences could be detected in lipoproteins, amino acid and carbohydrate profiles that related to metabolic pathways responsible for osmolyte fluctuation and energy metabolism. A similar study by Fardet et al. (2007) explored the influence of whole-grain and wheat flour diets on rats using both NMR-based chemometric methods. These studies showed that the whole-grain fed animals had higher urinary levels of Kreb’s cycle intermediates, aromatic acids and hippuric acid. These metabolites may indicate a shift in both the animal’s basal metabolic rate as well as an alteration to the gut microflora metabolism. NMR studies on plasma and liver showed increases in glutathione (reduced) and betaine, which are general indicators of good redox status and lower oxidative stress. Because many of the effects of diet intervention are likely to be more subtle and the biomarkers of interest will be at lower concentrations, many of the more recent studies have focused on using more sensitive MS methods or immunoassays. These include studies on the effects of fruit and vegetable diversity on the levels of oxidative biomarkers (Thompson et al., 2006), the influence of concentrated red grape juice on oxidative stress, lipidemic and inflammatory markers in blood (Castilla et al., 2006), the effect of extra virgin olive oil on plasma inflammatory and oxidative stress markers (Bogani, Galli, Villa, & Visioli, 2007), the consequences of macadamia nut consumption on markers for cardiovascular disease (Garg, Blake, Wills, & Clayton, 2007) and the influence of gazpacho soup consumption on plasma biomarkers of oxidative and inflammatory stress (Sa´nchez-Moreno et al., 2004). All of these studies showed some statistically significant, positive effectdlargely confirming the results of previous retrospective epidemiological studies. Many of these studies used both small molecule and large molecule protein biomarkers to measure the physiological effects of interest. With an improved understanding of the bioactive ingredients in many foods it is likely that future studies on the physiological effect of foods, or diet-intervention trials, will tend to use less heterogeneous sources of food or

selected bioactive components as diet supplements. The use of pure or partially purified bioactive components often makes the physiological interpretations easier and the results more reproducible. These kinds of studies are already being performed. For instance, Chiuve et al. (2007) studied the influence of two specific diet-derived molecules, betaine and choline, on reducing plasma homocysteine levels. Homocysteine is a well-established risk factor for cardiovascular disease and many other chronic diseases. Similar studies have been with other pure bioactives such as folate, omega-3 polyunsaturated fatty acids and a variety of purified fruit polyphenols on humans, animals and a variety of cell systems (Gibney et al., 2005; Vivancos & Moreno, 2007; Zeisel, 2007). While human health and disease are ultimately influenced by long-term exposure to diet, the use of short-term diet challenges to measure their projected effects on long-term health outcomes is perhaps the most effective (and cost-efficient) route to using metabolomics in understanding dietehealth relationships. The use of glucose, fructose (especially high fructose corn syrup), fat (cholesterol and/or trans-fatty acid) and high-calorie diet-challenge experiments along with their monitoring via sensitive metabolomic techniques will likely reveal some key insights into the temporal development of chronic diseases such as obesity, diabetes, cardiovascular disease and metabolic syndrome. Likewise, the use of fish/olive oil, flaxseed, cruciferous vegetable, polyphenolrich fruit and low-calorie diet-challenge experiments may reveal equally important insights into their efficacy for treating metabolic syndrome, combating cardiovascular disease and extending life expectancy. Diet-challenge experiments, if properly and comprehensively monitored by the more sensitive (i.e. GC-MS and LC-MS) metabolomic methods, will also help to reveal the breadth of individual responses to these different dietary perturbations. No doubt some ‘‘high-risk’’ individuals will show much greater or longer lasting perturbations for certain disease-risk biomarkers or metabolites whereas other ‘‘low-risk’’ individuals will show almost none. While metabolomics promises to provide a far more quantitative and chemically detailed understanding of phenotypic changes, the integrated use of transcriptomic, SNP-profiling, copy-number variation (CNV), gut metagenomic and proteomic/enyzmatic studies in these diet-challenge experiments will also allow for a much more complete understanding of the inter-individual differences or proclivities due to genetic and epigenetic effects. As noted earlier, there can be considerable diversity in individual responses to diet along with the considerable diversity of diets within the human population. This diversity certainly makes metabolic profiling and metabolite ‘‘standardization’’ difficult. Standardization and inter-subject comparison is also made more complicated by the fact that age, gender and diurnal effects also influence the concentrations and even the presence of certain metabolites. The creation of dedicated metabolite or metabolomic

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

databases that include these gender-specific ranges, agespecific variations or temporal effects could greatly help in this regard. Likewise the creation and sharing of comprehensive, longitudinal metabolomic (and other omic) data sets collected from well-designed and well-documented diet-challenge experiments would definitely make this sort of database development (and maintenance) much easier. It could also make group comparisons and metabolite calibrations much more robust and their conclusions much more reproducible. Conclusions Metabolomics is ideally positioned to be used in many areas of food science and nutrition research including food component analysis, food quality/authenticity assessment, food consumption monitoring and physiological monitoring in food intervention studies. However, the potential impact of metabolomics is still limited by two factors: (1) technology and (2) databases. In terms of instrumentation, it is clear that significant improvements need to be made to make metabolite detection and quantification technology more robust, automated and comprehensive. While promising advances have been made, current techniques are only capable of detectig perhaps 1/10th of the relevant metabolome. This expanded breadth and depth of coverage is particularly important in food and nutrition studies. Many more reference spectral or chromatographic databases on metabolites, food components and phytochemicals need to be developed and made public. It is only through these databases that nutritionally relevant compounds can be routinely identified or quantified. Indeed a comprehensive effort, similar to that undertaken to annotate the human metabolome (Wishart et al., 2007), needs to be made to complete and annotate the ‘‘food metabolome’’. Similar efforts also need to be directed towards creating publicly accessible, comprehensive nutritional phenotype databases that include quantitative metabolomic (and other omic) data collected from diet-challenge or food intervention experiments. While these kinds of endeavors may take years to complete and cost millions of dollars, hopefully the food science community (and its funding agencies) will find a way of coordinating its activities to complete these efforts. Indeed, having public resource like a food metabolome database or a nutritional phenotype database could be as valuable to food scientists as GenBank has been to molecular biologists. Acknowledgements I would like to thank the Canadian Institutes for Health Research (CIHR), as well as Genome Alberta and Genome Canada for their ongoing financial support. References Agabriel, D., Cornu, A., Journal, C., Sibra, C., Grolier, P., & Martin, B. (2007). Tanker milk variability according to farm feeding practices:

491

vitamins A and E, carotenoids, color and terpenoids. Journal of Dairy Science, 90, 4884e4896. Almeida, C., Duarte, I. F., Barros, A., Rodrigues, J., Spraul, M., & Gil, A. M. (2006). Composition of beer by 1H NMR spectroscopy: effects of brewing site and date of production. Journal of Agriculture and Food Chemistry, 54, 700e706. Aursand, M., Standal, I. B., & Axelson, D. E. (2007). High-resolution 13C nuclear magnetic resonance spectroscopy pattern recognition of fish oil capsules. Journal of Agriculture and Food Chemistry, 55, 38e47. Bertram, H. C., Hopee, C., Petersen, B. O., Duus, J. O., Molgaard, C., & Michaelsen, K. F. (2007). An NMR-based metabonomic investigation on effects of milk and meat protein diets given to 8-year old boys. British Journal of Nutrition, 97, 758e763. Bicker, W., Lammerhofer, M., Keller, T., Schuhmacher, R., Krska, R., & Lindner, W. (2006). Validated method for the determination of the ethanol consumption markers ethyl glucuronide, ethyl phosphate and ethyl sulfate in human urine by reversed-phase/weak anion exchange liquid chromatography-tandem mass spectrometry. Analytical Chemistry, 78, 5884e5892. Bogani, P., Galli, C., Villa, M., & Visioli, F. (2007). Postprandial antiinflammatory and antioxidant effects of extra virgin olive oil. Atherosclerosis, 190, 181e186. Brantsaeter, A. L., Haugen, M., Rasmussen, S. E., Alexander, J., Samuelsen, S. O., & Meltzer, H. M. (2007). Urine flavonoids and plasma carotenoids in the validation of fruit, vegetable and tea intake during pregnancy in the Norwegian Mother and Child Cohort Study (MoBa). Public Health and Nutrition, 10, 838e847. Castilla, P., Echarri, P., Davalos, A., Cerrato, F., Orgeta, H., Teruel, J. L., et al. (2006). Concentrated red grape juice exerts antioxidant, hypolipidemic, and anti-inflammatory effects in both hemodialysis patients and healthy subjects. American Journal of Clinical Nutrition, 84, 252e262. Castro, I. A., Barroso, L. P., & Sinnecker, P. (2005). Functional foods for coronary heart disease risk reduction: a meta-analysis using a multivariate approach. American Journal of Clinical Nutrition, 82, 32e40. Chiuve, S. E., Giovannucci, E. L., Hankinson, S. E., Zeisel, S. H., Dougherty, L. W., Willett, W. C., et al. (2007). The association between betaine and choline intakes and the plasma concentrations of homocysteine in women. American Journal of Clinical Nutrition, 86, 1073e1081. Cotter, D., Maer, A., Guda, C., Saunders, B., & Subramaniam, S. (2006). LMPD: LIPID MAPS proteome database. Nucleic Acids Research, 34(Database issue). D507eD510. Cuny, M., Le Gall, G., Colquhoun, I. J., Lees, M., & Rutledge, D. N. (2007). Evolving window zone selection method followed by independent component analysis as useful chemometric tools to discriminate between grapefruit juice, orange juice and blends. Analytica Chimica Acta, 597, 203e213. Cuny, M., Vigneau, E., Le Gall, G., Colquhoun, I., Lees, M., & Rutledge, D. N. (2008). Fruit juice authentication by 1H NMR spectroscopy in combination with different chemometric tools. Analytical and Bioanalytical Chemistry, 390, 419e427. De Rooij, B. M., Boogaard, P. J., Rijksen, D. A., Commandeur, J. N., & Vermeulen, N. P. (1996). Urinary excretion of N-acetyl-S-allyl-Lcysteine upon garlic consumption by human volunteers. Archives of Toxicology, 70, 635e639. Dunn, W. B., Bailey, N. J., & Johnson, H. E. (2005). Measuring the metabolome: current analytical technologies. Analyst, 130, 606e625. Ehrman, T. M., Barlow, D. J., & Hylands, P. J. (2007). Phytochemical databases of Chinese herbal constituents and bioactive plant compounds with known target specificities. Journal of Chemical Informatics and Modeling, 47, 254e263. Fardet, A., Canlet, C., Gottardi, G., Lyan, B., Llorach, R., Remesy, C., et al. (2007). Whole-grain and refined wheat flours show distinct metabolic profiles in rats as assessed by a 1H NMR-based metabonomic approach. The Journal of Nutrition, 137, 923e929.

492

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493

Garg, M. L., Blake, R. J., Wills, R. B., & Clayton, E. H. (2007). Macadamia nut consumption modulates favourably risk factors for coronary artery disease in hypercholeterolemic subjects. Lipids, 42, 583e587. German, J. B., Hammock, B. D., & Watkins, S. M. (2005). Metabolomics: building on a century of biochemistry to guide human health. Metabolomics, 1, 3e9. Gibney, M. J., Walsh, M., Brennan, L., Roche, H. M., German, B., & van Ommen, B. (2005). Metabolomics in human nutrition: opportunities and challenges. American Journal of Clinical Nutrition, 82, 497e503. Griennavar, B., Poulose, S. M., Jayaprakasha, G. K., Bhat, N. G., & Patil, B. S. (2006). Furocoumarins from grapefruit juice and their effect on human CYP3A4 and CYP1B1 isoenzymes. Bioorganic and Medicinal Chemistry, 15, 2606e2612. Griffin, J. L., & Bollard, M. E. (2004). Metabonomics: its potential as a tool in toxicology for safety assessment and data integration. Current Drug Metabolism, 5, 389e398. Hong, Y. J., & Mitchell, A. E. (2004). Metabolic profiling of flavonol metabolites in human urine by liquid chromatography and tandem mass spectrometry. Journal of Agriculture and Food Chemistry, 52, 6794e6801. Hu, F., Furihata, K., Kato, Y., & Tanokura, M. (2007). Nondestructive quantification of organic compounds in whole milk without pretreatment by two-dimensional NMR spectroscopy. Journal of Agriculture and Food Chemistry, 55, 4307e4311. Hutton, W. C., Garbow, J. R., & Hayes, T. R. (1999). Nondestructive NMR determination of oil composition in transformed canola seeds. Lipids, 34, 1339e1346. Ishikawa, T., Kondo, K., & Kitajima, J. (2003). Water-soluble constituents of coriander. Chemical and Pharmaceutical Bulletin, 51, 32e39. Ishikawa, Takayanagi, T., & Kitajima, J. (2002). Water-soluble constituents of cumin: monoterpenoid glucosides. Chemical and Pharmaceutical Bulletin, 50, 1471e1478. Ito, H., Gonthier, M. P., Manach, C., Morand, C., Mennen, L., Remesy, C., & Scalbert, A. (2005). Polyphenol levels in human urine after intake of six different polyphenol-rich beverages. British Journal of Nutrition, 94, 500e509. Kell, D. B. (2006). Systems biology, metabolic modelling and metabolomics in drug discovery and development. Drug Discovery Today, 11, 1085e1092. Kitajima, J., Ishikawa, T., & Satoh, M. (2003). Polar constituents of celery seed. Phytochemistry, 64, 1003e1011. Kopka, J., Schauer, N., Krueger, S., Birkemeyer, C., Usadel, B., Bergmuller, E., et al. (2005). Bioinformatics, 21, 1635e1638. [email protected]:, the Golm metabolome database. Lawrence, J. R., Peter, R., Baxter, G. J., Robson, J., Graham, A. B., & Paterson, J. R. (2003). Urinary excretion of salicyluric and salicylic acids by non-vegetarians, vegetarians and patients taking low dose aspirin. Journal of Clinical Pathology, 56, 651e653. Le Gall, G., Colquhoun, I. J., Davis, A. L., Collins, G. J., & Verhoeyen, M. E. (2003). Metabolite profiling of tomato (Lycopersicon esculentum) using 1H NMR spectroscopy as a tool to detect potential unintended effects following a genetic modification. Journal of Agriculture and Food Chemistry, 51, 2447e2456. Le Gall, G., Puaud, M., & Colquhoun, I. J. (2001). Discrimination between orange juice and pulp wash by 1H Nuclear Magnetic Resonance spectroscopy: identification of marker compounds. Journal of Agriculture and Food Chemistry, 49, 580e588. Lee, M. B., Storer, M. K., Blunt, J. W., & Lever, M. (2006). Validation of 1H NMR spectroscopy as an analytical tool for methylamine metabolites in urine. Clinica Chimica Acta, 365, 264e269. Lindon, J. C., Holmes, E., Bollard, M. E., Stanley, E. G., & Nicholson, J. K. (2004). Metabonomics technologies and their applications in physiological monitoring, drug safety assessment and disease diagnosis. Biomarkers, 9, 1e31.

Matsumura, T., Ishikawa, T., & Kitajima, J. (2002). Water-soluble constituents of caraway: aromatic compounds, aromatic compound glucosides and glucides. Phytochemistry, 61, 455e459. Mattivi, F., Guzzon, R., Vrhovsek, U., Stefanini, M., & Velasco, R. (2006). Metabolite profiling of grape: flavanoids and anthocyanins. Journal of Agriculture and Food Chemistry, 54, 7692e7702. Mennen, L. I., Sapinho, D., Ito, H., Bertrais, S., Galan, P., Hercberg, S., et al. (2006). Urinary flavonoids and phenolic acids as biomarkers of intake for polyphenol-rich foods. British Journal of Nutrition, 96, 191e198. Moco, S., Bino, R. J., Vorst, O., Verhoeven, H. A., de Groot, J., van Beek, T. A., et al. (2006). A liquid chromatography-mass spectrometry-based metabolome database for tomato. Plant Physiology, 141, 1205e1218. Moolenaar, S. H., Engelke, U. F., & Wevers, R. A. (2003). Proton nuclear magnetic resonance spectroscopy of body fluids in the field of inborn errors of metabolism. Annals of Clinical Biochemistry, 40, 16e24. Mullen, W., Boitier, A., Stewart, A. J., & Crozier, A. (2004). Flavonoid metabolites in human plasma and urine after the consumption of red onions: analysis by liquid chromatography with photodiode array and full scan tandem mass spectrometric detection. Journal of Chromatography A, 1058, 163e168. Myint, T., Fraser, G. E., Lindsted, K. D., Knutsen, S. F., Hubbard, R. W., & Bennett, H. W. (2000). Urinary 1-methylhistidine is a marker of meat consumption in Black and White California Seventh-day Adventists. American Journal of Epidemiology, 152, 752e755. Nicholson, J. K., Lindon, J. C., & Holmes, E. (1999). ‘Metabonomics’: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica, 29, 1181e1189. Ninonuevo, M. R., Park, Y., Yin, H., Zhang, J., Ward, R. E., Clowers, B. H., et al. (2006). A strategy for annotating the human milk glycome. Journal of Agriculture and Food Chemistry, 54, 7471e7480. Niwa, T. (1986). Metabolic profiling with gas chromatography-mass spectrometry and its application to clinical medicine. Journal of Chromatography, 379, 313e345. Nord, L. I., Vaag, P., & Duus, J. O. (2004). Quantification of organic and amino acids in beer by 1H NMR spectroscopy. Analytical Chemistry, 76, 4790e4798. Ogrinc, N., Kosir, I. J., Spangenberg, J. E., & Kidric, J. (2003). The application of NMR and MS methods for detection of adulteration of wine, fruit juices and olive oil. A review. Analytical and Bioanalytical Chemistry, 376, 424e430. Otvos, J. D. (2002). Measurement of lipoprotein subclass profiles by nuclear magnetic resonance spectroscopy. Clinical Laboratory, 48, 171e180. Pereira, G. E., Gaudillere, J. P., Pieri, P., Hilbert, G., Maucourt, M., Deborde, C., et al. (2006). Microclimate influience on mineral and metabolic profiles of grape berries. Journal of Agriculture and Food Chemistry, 54, 6765e6775. Prestes, R. A., Colnago, L. A., Forato, L. A., Vizzotto, L., Novotny, E. H., & Carrilho, E. (2007). A rapid and automated low resolution NMR method to analyze oil quality in intact oilseeds. Analytica Chimica Acta, 596, 325e329. Sa´nchez-Moreno, C., Cano, M. P., de Ancos, B., Plaza, L., Olmedilla, B., Granado, F., et al. (2004). Consumption of highpressurized vegetable soup increases plasma vitamin C and decreases oxidative stress and inflammatory biomarkers in healthy humans. The Journal of Nutrition, 134, 3021e3025. Schnackenberg, L. K., & Beger, R. D. (2006). Monitoring the health to disease continuum with global metabolic profiling and systems biology. Pharmacogenomics, 7, 1077e1086. Seeram, N. P., Henning, S. M., Yang, Y., Suchard, M., Li, Z., & Heber, D. (2006). Pomegranate juice ellagitannin metabolites are

D.S. Wishart / Trends in Food Science & Technology 19 (2008) 482e493 present in human plasma and some persist in urine for up to 48 hours. The Journal of Nutrition, 136, 2481e2485. Serkova, N. J., Zhang, Y., Coatney, J. L., Hunter, L., Wachs, M. E., Niemann, C. U., et al. (2007). Early detection of graft failure using the blood metabolic profile of a liver recipient. Transplantation, 83, 517e521. Smith, C. A., O’Maille, G., Want, E. J., Qin, C., Trauger, S. A., Brandon, T. R., et al. (2005). METLIN: a metabolite mass spectral database. Therapeutic Drug Monitoring, 27, 747e751. Solansky, K. S., Bailey, N. J., Beckwith-Hall, B. M., Davis, A., Bingham, S., Holmes, E., et al. (2003). Application of biofluid 1H nuclear magnetic resonance-based metabonomic techniques for the analysis of the biochemical effects of dietary isoflavones on human plasma profile. Analytical Biochemistry, 323, 197e204. Somoza, V., Wenzel, E., Weiss, C., Clawin-Radecker, I., Grubel, N., & Erbersdobler, H. F. (2006). Dose-dependent utilization of caseinlinked lysinoalanine, N(epsilon)-fructolysine and N(epsilon)carboxymethyllysine in rats. Molecular Nutrition and Food Research, 50, 833e841. Stella, C., Beckwith-Hall, B., Cloarec, O., Holmes, E., Lindon, J. C., Powell, J., et al. (2006). Susceptibility of human metabolic phenotypes to dietary modulation. Journal of Proteome Research, 5, 2780e2788. Thompson, H. J., Heimendinger, J., Diker, A., O’Neill, C., Haegele, A., Meinecke, B., et al. (2006). Dietary botanical diversity affects the reduction of oxidative biomarkers in women due to high vegetable and fruit intake. The Journal of Nutrition, 136, 2207e2212. Tiziani, S., Schwartz, S. J., & Vodovotz, Y. (2006). Profiling of carotenoids in tomato juice by one- and two-dimensional NMR. Journal of Agriculture and Food Chemistry, 54, 6094e6100. Trygg, J., Holmes, E., & Lundstedt, T. (2007). Chemometrics in metabonomics. Journal of Proteome Research, 6, 469e479. Vivancos, M., & Moreno, J. J. (2007). Effect of resveratrol, tyrosol and beta-sitosterol on oxidised low-density lipoprotein-stimulated oxidative stress, arachidonic acid release and prostaglandin E2

493

synthesis by RAW 264.7 macrophages. British Journal of Nutriton, 98, 1e9. Walsh, M. C., Brennan, L., Malthouse, J. P. G., Roche, H. M., & Gibney, M. J. (2006). Effect of acute dietary standarization on the urinary, plasma and salivary metabolomic profiles of healthy humans. American Journal of Clinical Nutrition, 84, 531e539. Watkins, S. M., & German, J. B. (2002). Metabolomics and biochemical profiling in drug discovery and development. Current Opinion in Molecular Therapeutics, 4, 224e228. Weljie, A. M., Newton, J., Mercier, P., Carlson, E., & Slupsky, C. M. (2006). Targeted profiling: quantitative analysis of 1H NMR metabolomics data. Analytical Chemistry, 78, 4430e4442. Wiest, M. M., & Watkins, S. M. (2007). Biomarker discovery using high-dimensional lipid analysis. Current Opinion in Lipidology, 18, 181e186. Wishart, D. S., Querengesser, L. M. M., Lefebvre, B. A., Epstein, N. A., Greiner, R., & Newton, J. B. (2001). Magnetic resonance diagnostics: a new technology for high-throughput clinical diagnostics. Clinical Chemistry, 47. 1918e1921. Wishart, D. S., Tzur, D., Knox, C., Eisner, R., Guo, A. C., Young, N., et al. (2007). HMDB: the human metabolome database. Nucleic Acids Research, 35(Database issue), D521eD526. Ye, M., Han, J., Chen, H., Zheng, J., & Guo, D. (2007). Analysis of phenolic compounds in rhubarbs using liquid chromatography coupled with electrospray ionization mass spectrometry. Journal of the American Society of Mass Spectrometry, 18, 82e91. Zeisel, S. H. (2007). Nutrigenomics and metabolomics will change clinical nutriction and public health practice: insights from studies on dietary requirements for choline. American Journal of Clinical Nutrition, 86, 542e548. Zuppi, C., Messana, I., Forni, F., Ferrari, F., Rossi, C., & Giardina, B. (1998). Influence of feeding on metabolite excretion evidenced by urine 1H NMR spectral profiles: a comparison between subjects living in Rome and subjects living at arctic latitudes (Svaldbard). Clinica Chemica Acta, 278, 75e79.