Challenges in Nutrition

From a “Metabolomics fashion” to a sound application of metabolomics in research on human nutrition

The philosopher Karl Popper described one particular aspect of science as “…its need to grow…” and it is “the way of its growth which makes science rational and empirical.” [1].

Biomedical science is based on empirical evidence and progress is perceived as high research activities and the accumulation of data. Overall, this progress is driven by new technologies and methods, their broad applications, growth of theoretical knowledge, ambition of scientists, power of institutions, high research funding as well as fast and effective content distribution by publications.

Today, many scientists see the metabolome as the closest representation of the phenotype [2], as it aims to provide a global perspective on metabolism. Metabolomics addresses metabolites, their metabolism and interactions between different classes of biomolecules; it is part of an integrative “omics” approach. Metabolomics may add to precision medicine and, thus, personalized health [3, 4]. As far as human nutrition is concerned, metabolomics is viewed as a “key analytical tool in human nutritional studies” that addresses the effects of specific diets, foods, and bioactive compounds, and so, promises to improve nutritional care and dietary treatment [5].

According to Karl Popper again, “science is valued, admittedly, for its practical achievements; but it is even more highly valued for its informative content” [1]. Today, there are multiple human data bases for the identification of individual metabolites. The data volume is impressive which also reflects that metabolites are easily measured. Does the mere accumulation of metabolomic data provide explanations? Today, we may give a more negative answer to that question. Compared with our present knowledge, it is likely that data-driven and network-based approaches may not necessarily support a better understanding of human metabolism. In addition, when compared with traditional risk scores the use of metabolite profiles as risk estimates is below expectations.

The issue is partly because interpreting results from metabolomics in a biological context is a difficult task. More and more data and the application of artificial intelligence (AI) applied to that data feeds speculations and the generation of newer hypotheses rather than increase our understanding of metabolism. To take a more general view, the big data approach with its dependence on algorithms and machine learning carries the risk to produce data which cannot be retranslated into normal “language” to explain to ourselves what they really mean. Then, the accumulation of data may not explain anything but may give us a transient illusion of understanding only.

It might be argued that metabolomics is a new technology which adds to an explosion of publications where some are good and others are not (i.e., there is a lot of noise). However, it remains questionable whether research needs that noise to generate sound applications of a new technology at the end. This article questions the informative content of some recent applications of metabolomics to human nutrition and provides ideas for its future applications. Our critical view does not exclude that metabolomics may hold the potential to add to our knowledge, e.g., to identify diagnostically informative biomarkers. Our present disenchantment about metabolomics is explained by its so far poor applications to nutrition research. Although we admire and understand the spirit of discovery, we consider some present research activities in this area as a “Metabolomic fashion” rather than an achievement.

What is metabolomics about?

In biofluids, tissues or cells, metabolomics identifies and quantifies low molecular weight molecules including sugars, amino acids, lipids, organic acids, and nucleotides. As has been already said, “The metabolome is more than the metabolites that are part of known metabolic pathways. Instead, it is all the chemicals with molecular weights less than 1500 Da that can be detected by metabolite measurement technologies, i.e., mass spectrometry (MS) or nuclear magnetic resonance analysis of biofluids or tissues” [6] where arbitrary signal intensity measured by ion counts depends on the concentration of the metabolite. Changes in ion counts reflect relative amounts which can be transferred into absolute concentrations of metabolites by the concomitant use of standards. In a standard procedure on plasma samples up to 600 metabolites are identified leaving about 10,000 peaks unidentified [7]; thus, the full set of metabolites analyzed remain ill-defined.

When compared with proteins and genes, metabolites (or intermediates or substrates) are small and readily available molecules. Since they have simple structures these make them good candidates for screening their activity. The diagnostic purposes of measuring a wide array of metabolites include

  • the diagnostics of inborn errors of diseases,

  • the discovery of early alterations in intermediary metabolism,

  • the identification of previously unknown metabolites,

  • finding biomarkers of either food and nutrient intake or the functional responses of the body (e.g., in terms of alterations in the metabolic flux and its control) or the susceptibility or resilience to a variety of chronic diseases,

  • to predict health risks and mortality and

  • to define surrogate endpoints for intervention studies [3].

Metabolites are produced and metabolized by enzymatic reactions within the cells of the body. Metabolite profiles in the blood, urine, and tissues are interpreted on the basis of biochemical knowledge about intracellular metabolism and transfer of substrates between organs and tissues. When metabolite profiles are integrated within metabolic pathways, certain metabolites related to enzyme activities and gene function characterizing certain phenotypes may be identified. However, metabolism within cells and between organs and tissues is complex and multifactorial. Therefore, data analysis benefits from a real world-application of AI in biomedical science [8, 9].

Untargeted metabolomics i.e., untargeted semi-quantified analyses of all measurable analytes [8]; provide an unbiased approach to characterize variance in metabolite levels in certain conditions, e.g., fasting vs. refeeding. Crucial points to be considered include

  • the implementation of standard procedures and workflows,

  • the probe itself, e.g., blood, urine or tissue probes, the sampling procedure and the analyses by liquid chromatography high-resolution MS,

  • the identification and quantitative assessment of individual metabolites,

  • the nature of unidentified peaks,

  • statistical analysis methods and

  • assumptions about the activity and direction of metabolic pathways of interest.

Untargeted metabolomics have been largely used as a screening tool for association with diets, metabolic risks and diseases. Metabolite profiles in blood, other biofluids or tissue samples have been assessed in cohorts of healthy humans and patients. The data can be used as reference as well as prospectively, when incidences of metabolic disturbances and/or diseases can be compared with alterations of metabolite levels observed during the observation period. The identification of key metabolites or metabolite profiles which are considered as characteristics and/or modulators of metabolism and the characterization of certain phenotypes is validated by targeted metabolomics i.e., targeted quantification of groups of chemically characterized and biochemically annotated metabolites [8]; in an experimental setting or in large scale studies. So far, in human nutrition, metabolites which are in the core of energy, glucose, lipid and amino acid metabolism have received the most attention in research.

In most studies, metabolite intercorrelations and their responses to interventions were assessed while the performance of biomarkers was tested using receiver operator characteristic curves to address the sensitivity and specificity of the prediction of health risks, mortality or dietary intake as suitable outcomes. Statistical methods to analyze metabolome data include univariate (e.g., linear) regression, multivariable regression analysis with adjustments for covariates, metabolome-wide association study, clustering, principal component analysis (PCA, identifying linear combinations of metabolites which may differ between samples) and network analyses. Using “clustered heat maps” (i.e., visually representing pairwise correlations), metabolites assessed in a sample showing similar patterns group together and their metabolic responses can be followed (e.g., during fasting vs. refeeding).

An extended protocol can be used to translate metabolite profiles into a metabolic pathway activity, i.e., the set of metabolic fluxes per time or the so-called “fluxome,” its structure and its control, which had originally created that metabolome [7]. This is addressed by extending the analytical platform by the use of isotopic tracers (e.g., 13C-glucose) and kinetic computational models to quantitate the flux through a metabolic pathway. Then, the accumulation of the label is identified within the fraction of the metabolite (e.g., 13C glucose in the glucose fraction) and its degradation products (e.g., 13C-lactate in the lactate fraction). During a metabolic steady state, the ratio of the labeled to the unlabeled fraction may be taken as a first and qualitative estimate of the metabolic flux where this calculation needs to take into account the half-time of the label and the metabolic pool size [10].

Alternatively, analyses performed after an isotopic steady state had been reached allow quantification of the metabolism of individual substrates. Then, the tracer infusion rate, its enrichment in the plasma, the ratio between labeled and unlabeled substrate (13C glucose on glucose) and its changes by (the endogenous or exogenous) production and/or utilization of that metabolite are measured. This approach can be extended by using a selective labeling of molecules (e.g., glucose can be labeled at its different C- or H-atoms) where the cellular catabolism of the different carbon or hydrogen units differs and, thus, allows the characterization individual pathways of that metabolite (e.g., glycolysis vs. pentose phosphate pathway). Labeling can be followed throughout the whole metabolic pathway (e.g., the 13C-glucose label can be identified within the intermediates of glycolysis and the Krebs cycle). This adds to the issue of recycling of labels (13C-lactate generated from the degradation of 13C-glucose recycles to 13C-glucose again by hepatic gluconeogenesis).

Faced with the complexity of intermediary metabolism, the direct and indirect associations between the circulating metabolites and the tissue metabolite profile need sophisticated matrix calculations and carefully thinking when one plans such kind of investigations.

A critical look at some recent applications of metabolomics in human nutrition

Many concerns about metabolomics have to do with random variation, both analytical and within-individual, which can lead to spurious findings. To discuss some issues related to the applications of metabolomics we refer to some selected experiences in the area of human nutrition.

  • Metabolomics in blood samples as biomarkers and risk estimates: The metabolome has been addressed in obese subjects including 1969 largely European ancestry twins enrolled in the TwinsUK registry [11]. Non-targeted metabolomics and whole-genome sequencing were performed to identify metabolic and genetic signatures of obesity. Sample selections followed a routine clinical protocol. Obesity resulted in profound perturbation of the metabolome. Nearly a third of the assayed metabolites was associated with changes in BMI. The single metabolite with the most significant association with BMI was urate. Predicting BMI again from the 49 BMI-associated metabolites had a specificity of 89.1%, and a sensitivity of 80.2%. 650 metabolites explained about 50% while a subset of 49 of the best markers explained 43% of variance in BMI. Using a PCA in the full set of about 1000 metabolites, different metabolite signatures associated with BMI were characterized, which were related to the metabolism of nucleotides, amino acids, peptides, lipids, carbohydrates, vitamins, and cofactors and energy. E.g., the plasma concentration of succinyl-carnitine was taken as a measure of the Krebs cycle activity within cells and, thus, whole body energy metabolism. In addition, no significant associations were seen between any single metabolites and polygenic risk scores or even MC4R carrier status in either the entire population or in obese individuals only. This result implied that metabolites are unlikely to be intermediate phenotypes that explain the underlying genetics of obesity.

    A meta-analysis of 27 cross-sectional and 19 prospective studies on the association between metabolites and either the metabolic syndrome or type 2 diabetes mellitus [12] showed that the risk was mainly associated with changes in the plasma levels of lipids (phospholipids, sphingomyelin, triglycerides) and amino acids (branched chain and aromatic amino acids, glycine and glutamine). As far as lipids were concerned, the results showed that high triglycerides were associated with increased risk, while long chain fatty acids with double bonds were associated with a lower risk. Risk estimates of branched chain and aromatic amino acids were found to range between 1.26 and 1.36. Surprisingly, intermediates of carbohydrate metabolism were not found as markers. By contrast, in one study [13], glucose, fructose, and inositol were associated with a higher risk of diabetes mellitus whereas the levels of sugar alcohols reflected a lower risk. All together these data suggested that the metabolites identified so far not only are altered but also exhibit prospective associations with clinically relevant disturbances in glucose metabolism. However, when compared to the fasting glucose level, a family history, of diabetes mellitus or a calculated diabetes risk score the variance in blood concentrations of all the diabetes-associated metabolites did not add to the diagnostic and/or risk estimates of type 2 diabetes [5]. However, the “new” biomarkers may be (at least in part) unrelated to the traditional risk estimates and, thus, may extend risk profiles in future. Up to now, the data do not provide any deeper insights into the metabolic disturbances observed in diabetic patients.

    Fatty liver disease (FLD) is a predictor of insulin resistance and type 2 diabetes mellitus. Metabolites associated with FLD may thus add to a suitable risk score. Positive associations with FLD were found for the levels of certain amino acids (branched chain and aromatic amino acids, glutamate) and urate while cysteine-glutathione showed a negative association [14]. The association between the plasma concentrations of these metabolites and liver fat content was stronger in overweight and obese subjects. Altogether, a metabolomic score calculated from 15 metabolites explained about 24% of the variance in liver fat content suggesting that a considerable proportion of the variance in liver fat is not “explained” by the circulating metabolites investigated.

    Another study addressed metabolic markers associated with 5–10 years all cause mortality. This meta-analysis included 12 well characterized cohorts of a total of 44,168 subjects and identified 14 biomarkers related to lipid metabolism, glycolysis, body water and its tonicity, and inflammation [15], i.e., lipids (lipids in chylomicrons, large VLDLs, small HDL; polyunsaturated fatty acids, acetoacetate), amino acids (branched chain and aromatic amino acids and histidine), lactate, albumin and glycoprotein acetyls. An increase of one unit of a calculated biomarker score was associated with a 2.73-fold increased mortality risk. When compared to conventional risk factors (e.g, elevated blood pressure and cholesterol levels) the score based on metabolomics only marginally improved the prediction of mortality (i.e., the AUC of the receiver operating curves increased from 0.772 to 0.837).

  • Metabolomics in blood samples as measures of intracellular metabolism: The basal and 1-year changes in the plasma concentrations of intermediates of glycolysis- and the Krebs cycle-related intermediates have been compared with the incidence of insulin resistance and type 2 diabetes mellitus in a nested case-control study on overweight and obese subjects with a high risk for cardio-vascular disease [16]. As for the baseline concentrations, several plasma metabolites were associated with incident type 2 diabetes including hexose monophosphate, phosphoenolpyruvate, pyruvate, lactate, alanine, glycerophosphate, and isocitrate. The weight sum of all metabolites resulted in an adjusted Hazard ratio of 1.30 per 1-SD increase. Baseline lactate and alanine concentrations were also associated with the 1-year change in insulin resistance (as measured by the HOMA index). Comparing the 1-year changes in the plasma metabolite concentrations in a control group of 210 subjects with 58 cases of incident Type 2 Diabetes mellitus the global score of metabolites resulted in an adjusted Hazard ratio of 3.57 per 1-SD increase (with hexose monophosphate, 3-phosphoglycerate, lactate, citrate, aconitate, isocitrate, fumarate, and malate as metabolites included in the score). This was in line with another large community-based cohort study, where the plasma concentrations of some Krebs cycle intermediates (i.e., aconitate, isocitrate, malate) were shown to be related to with both worse cardiovascular health and decreased longevity as defined as living to age >80 years [17].

    These data suggest that plasma metabolomics can provide an understanding of metabolic pathways within cells. However, it remains unclear how intermediates of glycolysis and the Krebs cycle, which are compartmentalized and protein-bound within the cell, can escape intact cells to become measurable in the blood. Since in those studies metabolic fluxes have not been determined concomitantly, the meaning of the plasma concentrations and their changes as a reflection of intra-cellular and inter-organ/tissue metabolism remains obscure.

  • Metabolomics in tissue samples: To address compartments that are the place of intermediary metabolism or the origin of its intracellular perturbation, metabolomics had been assessed in tissue samples. Up to now, most studies on metabolite profiles in tissue samples had been performed in experimental animals. E.g., metabolomics was used to assess hepatic acyl-CoA compounds in response to the high fat diet in mice [18]. While the tissue concentrations of many acyl-CoAs remained constant, two short-chain Acyl-CoAs, Malonyl-CoA and propionyl-CoA, and four medium to long chain acyl-CoAs (C10:3, C16, C18:1, C18:2-CoA) increased. These changes were associated with widespread changes in the hepatic concentrations of metabolites related to different metabolic pathways, while metabolites related to central metabolic pathways (e.g., the Krebs cycle) did not change in response to fat diet. Using statistical analyses, changes in the tissue concentrations of acyl-CoAs were correlated with a number of other metabolites. Among others, malonyl-CoA was correlated with the tissue levels of NAD+ and glutamine. The authors took their data as evidence that acyl-CoA-profiling adds to the discovery of metabolic changes in response to diets. As to methodological caveats, mice were sacrificed by cervical dislocation, the tissue was snap frozen in liquid N2 and stored at −80 °C before the analyses of metabolites. Since different acyl-CoAs show different degradation rates in the liver extract, the authors addressed this issue by using a few of labeled labeled acyl CoAs as a standard. The results suggested that the methods used did not prevent acyl-CoA degradation during tissue preparation.

    Metabolomics had also been used to address changes in brown adipose tissue (BAT) energy metabolism in response to graded caloric restriction compared with ad libitum fed animals [19]. With weight loss, the mass of BAT decreased but it was better conserved than white adipose tissue. Although metabolic fluxes were not assessed, the data suggested that lipid and glucose metabolism as well as uncoupling of the electron transport chain increased in response to fasting (and the fasting-associated decrease in body temperature). Concomitantly, a total of 883 differently expressed metabolites were identified with most of them related to Krebs cycle activity, fatty acid degradation, nucleic acids, antioxidants and catecholamine metabolism altogether indicating that BAT activity is increased by hunger. In that study, the analyses of tissue metabolite levels had been done in already frozen and re-thawed tissue samples.

    High-fructose corn syrup stimulates intestinal tumor growth which may be independent of the occurrence of obesity and metabolic syndrome [20]. This is explained by an increased formation of fructose-1-phosphate and glycolysis within tumor cells and the liver which was associated with (i) a decrease in ATP levels, (ii) an increased purine degradation, and (iii) increased fatty acid synthesis supporting tumor growth. Labeling both, fructose and glucose, with 13C, the effects of both sugars could be differentiated from each other suggesting that glucose inhibits the degradation of fructose-1-phosphate by competing for aconitase activity, thus, resulting in an acute drop in cellular ATP concentrations. The results provide evidence for the deleterious metabolic effects of the combination of dietary glucose and fructose. As to the methods, after euthanization, metabolites were extracted from the frozen liver, small intestinal epithelium, and tumor tissue, thus, taking into account the methodological caveats of tissue sampling (see below) the concentrations of metabolites in the tissues do not represent those under normal metabolic condition (see below).

    High selenium intake induced hyperglycemia and hyperinsulinemia and, thus, may induce type 2 diabetes mellitus [21]. Feeding pigs with supra-physiological amounts of seleno-methionine increased liver selenium concentrations 4.2-fold. Metabolite profiles were analyzed in serum, liver, muscle, heart, and kidney. With increased tissue concentrations of selenium, the activities of gluconeogenic enzymes increased whereas glycolytic enzyme activities decreased. This was associated with variable increases in the relative contents of intermediates of glycolysis, pentose phosphate pathway, and fatty acid synthesis suggesting a decrease in glucose utilization with a concomitant increase in lipid synthesis which resembled a diabetes-like phenotype. Concomitantly, the concentration of lipids in the serum were not affected by selenium treatment. For analyses, pigs were killed through electric shock and exsanguinated. The organs were quickly collected, frozen with N2, and stored at −80 °C. Again, the concentrations of metabolites in the tissues do not represent those under normal metabolic condition (see below).

  • Metabolomics as biomarkers of dietary exposure: Self-reports of dietary intake tend to be biased [22]. Validating dietary intake needs objective measurements which allow a prediction of the intake of energy or of individual nutrients or foods [23]. E.g., assuming a stable body weight and, thus, during energy balance, energy intake is validated by 24 h energy expenditure as measured by doubly-labeled water. Accordingly, at nitrogen balance, protein intake data can be compared with 24 h urinary nitrogen excretion. Other examples include the measurements of the urinary excretion rate of sodium, potassium, iodine, selenium, fluoride and chloride for validation of their respective intakes. In addition, erythrocyte and plasma concentrations of vitamins have been taken as qualitative biomarkers for the intake of folate and vitamin C with correlation coefficients between intake data and their biomarkers ranged from 0.4 to 0.7 [22].

    More recently, metabolomics in blood samples have been used as biomarkers of dietary exposure e.g., [24,25,26]. These include biomarkers originating from specific foods (reflecting the bioavailability of nutrients or food additives) and endogenous metabolites (reflecting the effect of nutrients or foods upon microbial action in the gut and/or metabolism and exchange within and between organs and tissues). Evaluations were based on cross-sectional correlations [24] or controlled dietary interventions [25, 26].

    Metabolomics responses to diets varying in fat and carbohydrate content in a controlled 4 week intervention study identified changes in the plasma concentrations of 152 metabolites related to lipid (e.g., triacylglycerol, ß-hydroxybutyrate) and amino acid metabolism (e.g., branched-chain amino acids) which could be used again to identify the respective test diet [25]. The metabolome predicted the correct diet with 95% accuracy. To test generalizability, the authors also analyzed the correlations between self-reported diet and plasma metabolite profiles in 1840 individuals of the Framingham Heart Study where 103 diet-associated metabolites were identified. These data suggested that the specific results of that controlled intervention study were not in accord with other studies.

    In a further study, 1-year changes in the lipidome in response to an intervention with the Mediterranean Diet had been followed in a case-control study of patients with cardiovascular diseases [26]. From a total of 202 lipid species 20 or 17 lipids responded to the intervention with the Mediterranean Diet supplemented with either extra virgin olive oil or nuts. Although the lipid composition of the intervention diets differed, only cholesterol ester with poly-unsaturated fatty acid (i.e., 20:3), remained significant after multiple testing. The authors took this single finding as evidence that their dietary intervention is reflected by the lipidome.

    Using two metabolomic platforms, a total of 522 metabolites were identified with 102 correlations between >1 metabolite and 18 foods, beverages, and supplements [24]. The correlations ranged from 0.44 for fish and 0.55 for citrus. The authors concluded that these biomarkers may complement self-reported food intake data.

    It has been proposed that clusters of metabolomic profiles (so-called “metabotypes”) may correspond to clusters of dietary patterns [27]. This would significantly add to nutritional epidemiology. However, presently this area remains a high research priority.

    To summarize, the comparison between dietary intake data and the plasma metabolome is flawed on both sides i.e., using protocols or questionnaires dietary exposure cannot be measured with accuracy and the association of plasma or urine metabolites with intake of nutrients or foods is at best moderate [22].

A critical look behind the “Metabolomic fashion”

For future studies the following points should be considered:

  • Blood and tissue samples: To avoid heterogeneities the use of fasting vs. nonfasting blood and tissue samples, their appropriate collection, handling, and storage has to be ensured [5, 8]. As a general rule, blood and urine samples have in common that they address compartments that are neither the place of intermediary metabolism nor the origin of its intracellular perturbations. This idea does not exclude the possibility that a plasma or urine “fingerprint” can be detected which might be useful for clinical prediction without advancing fundamental knowledge. However, the informative value of metabolomics in blood and urine is questionable since intermediary metabolism differs between the different organs and tissues within the body which also have a certain hierarchy with respect to specific substrates as well as the metabolic situation [28, 29].

    Comparing concentrations of different metabolites of a group of macronutrients (e.g., lipids) in a single blood sample is complicated by their different half-lives (e.g., 3 min in the case of free fatty acids, 5–15 min for triglycerides within chylomicrons, 2 to 4 h for VLDL, 4 to 6 days for LDL-cholesterol, 6–8 days for HDL-cholesterol while individual sphingosine lipids may have longer half-lives). This is textbook knowledge [30, 31]. Therefore, at a certain time point the concentrations of individual lipids and their changes in response to a diet cannot be directly compared with each other, thus, limiting the informative value of the lipidome with respect to lipid metabolism.

    Since it is unclear how metabolites fixed to intracellular membranes (e.g., intermediates of the Krebs cycle) can leave a healthy cell and become measurable in the blood, it is likely that these observations reflect cellular leakage or secretion of these intermediates. This may be a measure to control cellular growth or considered to reflect the symbiosis between cells within organs and tissues. This hypothetical and yet unproven idea would add to other dimensions of the possible meaning of metabolic profiles which have to be seen in the context of metabolic fluxes, cellular growth as well as symbiosis between cells.

    As far as methods of extracting metabolites from tissue are concerned, it is a requirement that samples should be fresh frozen in liquid N2 at the time of collection to prevent further metabolism and stored at −80 °C [32,33,34]. To do so, most authors refer to published standards to minimize perturbations of metabolism due to anoxia, stress, the freezing technique and narcosis e.g., [6]. When collecting frozen tissues of any type, the aim is to avoid causing any stress and to freeze the specimen in as short a period as possible following excision. In experimental animals, CO2 euthanization or neck fracture are still frequently used although these procedures most probably provoke a metabolic tsunami in all cells of the animal. Concomitantly, severe stress to the animal results in immediate (within seconds) changes in the level of hormones as epinephrine, glucagon, cortisone, and growth hormone elevating blood glucose concentrations derived from immediate break down of liver glycogen, circulating in the blood within seconds and, consequently, acting in all tissues of the body.

    Metabolic alterations due to tissue sampling are further increased when the sample is simply transferred into liquid N2 since it takes several seconds until the whole sample is frozen, e.g., for liquid nitrogen, 16.3 s are necessary to decrease the temperature of an immersed, 800 mg weighing tissue sample from 38° to 0 oC [34]. This is due to the physical effect of N2 bubbles isolating the specimen. Since the “flash frozen clamp technique” significantly reduces the time of tissue freezing [32], it is the best method for preparing tissues for metabolomics analysis. This is because of the shifting pH and the redox state in cytosol and mitochondria, plus anoxia, which is followed by an immediate drop of cellular ATP levels associated with an increase in tissue concentrations of ADP and AMP [32, 33]. Thus, within seconds the whole metabolic picture is changed into “survival glycolysis” via metabolite-regulated allosteric enzymes. Finally, the organic solvent (i.e., its mixture with an acid to denature enzymes within the probe to prevent degradation and production of metabolites) used for tissue preparation may not immediately stop catalytic activity which impacts the concentrations of high-energy compounds like ATP and NADH [7].

    It has been assumed by researchers, that when comparing differences between groups (e.g. with vs. without a diet intervention) to make conclusions, all inevitable perturbations to metabolism upon tissue harvesting will occur in the same time frame in the control and the treated groups. However, comparing incorrectly sampled data in two groups does not make any sense at all.

  • Study protocols: Not controlling for composition and effect of diets or the time of day of sample collection leads to excessive inter-individual variation in metabolite profiles and/or between group differences [6]. In addition, even a defined diet (e.g., a low carbohydrate diet) may have a considerable variability in food choices and, thus, the composition of the diet which adds to its effect on plasma metabolite levels. In most metabolomics studies samples were obtained in the so-called “basal state,” i.e., awake; no voluntary muscle activity, no physical activity before sampling; thermoneutrality; “normal” body temperature; during the follicular phase of the menstrual cycle; after an overnight fast, to minimize the effects of variations from the absorption of dietary components. However, a “basal state” cannot be considered as a strictly controlled metabolic situation, because of inter- and intra-individual variances in lifestyle determinants of metabolism. A period of fasting for 8–12 h affects hepatic rather than extrahepatic metabolism resulting in a catabolic and, thus, a nonsteady state with hepatic glycogen becoming the major substrate of energy metabolism. The kinetics of glycogen mobilization depend on the amount of the initial glycogen stores, the exact duration of fasting and additional determinants of hepatic glycogen metabolism (e.g., sleep duration, stress, and physical activity) altogether explaining a considerable intra- and inter-individual variability of the metabolome. It is evident that a sound application of metabolomics requires a strict definition of all lifestyle variables.

    Addressing the dynamics of metabolic control sophisticated protocols and methods are needed for a detailed discussion see [29, 35]. This is because within days, weeks, or even months after an intervention, changes in body composition, metabolism and its endocrine and inflammatory correlates reflect control and adaptations of metabolism to a varying extent. To get insights into metabolic control, there is a need of investigations within a tight time frame.

  • Meaning of metabolite concentrations: Today, different metabolite signatures, e.g., associated with the BMI or energy balance, leave scientists with some confusion. While an association between serum urate levels, insulin resistance and BMI are known, the meaning of some metabolites measured in plasma like succinyl-carnitine as a measure of energy metabolism, see [11] or cortisone as a measure of lipids [11], remains obscure. It is worthwhile to remember that, presently, interpretations of metabolite concentrations are frequently due to mere mathematical associations and clustering rather than to a comprehensive assessment of metabolic fluxes.

    In patients, the plasma concentration of numerous metabolites may change dramatically in the disease course. However, a low plasma or even a low tissue concentration of a metabolite may reflect metabolic adaptations rather than its “true” deficiency questioning the need of its substitution. In fact, early provision of glutamine into hypermetabolic and critically ill patients with low tissue and plasma concentrations of this amino acid did not improve but worsened clinical outcome [36]. This finding points out to issues of associations and causation which impact the interpretation of metabolomics [27]. Using a cross-sectional study design it remains unclear whether metabolite profiles reflect (i) epiphenomena associated with alterations of metabolic fluxes or (ii) outcomes of the nutritional and metabolic issues addressed.

    Modern biomedical research is driven by new methods and reductionism, i.e. the view “as scientific only such concepts as…could be reduced to elementary experiences” [1]. It is assumed that the whole is equal to the sum of it’s parts. Today’s fashion in biomedical science is to reduce its concepts and phenotypes to the data obtained by “omics”-methodologies which are taken as numerous parts. However, a more and more detailed view at the molecular and cellular level carries the risk of adding to more and more speculations rather than to end up in a complete picture of whole body metabolism and its biological control. In fact, the accumulation of “omics”-data are no substitute for integrative biology.

    To do better, we suggest that integrative concepts like “functional body composition” intend t to (i) integrate metabolomics into the anatomical and physicochemical structures of the body, (ii) demonstrate their interrelations with metabolic and endocrine functions and, finally, (ii) refer to their impact on systemic outcomes like body temperature, respiration, urine production and excretion, heart rate and blood pressure [37]. Taking an integrative view it is likely that “omics”-strategies (including metabolomics) cannot be successful on its own.

    Our textbook knowledge about the biochemistry of intermediary metabolism is still a safe way to go see [30, 31]. Metabolomics are about measuring metabolite concentrations. What are metabolites about? It is already well known from basic biochemistry that even (i) under controlled conditions (e.g., during fasting and refeeding high carbohydrate diets) and (ii) when all individual metabolites of a given metabolic pathway e.g., of glycolysis, Krebs cycle and ATP-production have been measured within hepatocytes, these metabolites do not reflect metabolic fluxes nor do they give any insights into metabolic control [38,39,40]. This is because, metabolite levels cannot be interpreted without knowledge of metabolic fluxes and, thus, the “direction” of metabolism.

    It is known that the intra-cellular flux through individual metabolic pathways is not a “one-way flow” of metabolites [30, 31]. E.g., in the liver gluconeogenesis and glycolysis occur at the same time with differences in the net balance between the two, i.e., during fasting there is increased hepatic glucose output whereas during refeeding there is an increased hepatic glucose uptake and production of pyruvate and lactate. Faced with the corresponding key enzyme activities (e.g., hepatic phosphofructokinase activity) the intra-cellular changes in the levels of its substrate in response to fasting or refeeding are within the range of their km-value [41]. Physiologically, there are transient changes in the intracellular metabolite levels only which are due to a transient limitations in metabolic capacity [40].

    Metabolite levels have to be seen in the context of metabolic fluxes. Net metabolic balance (of glucose, lipids and protein) of the body can be assessed in a controlled setting of indirect calorimetry combined with N-excretion while the specific flux of individual substrates is measured by isotope tracing and dilution techniques (e.g., using labeled substrate molecules). In addition, arterio-venous differences in metabolite concentrations can be measured across body regions (e.g., the splanchnic region), organs and tissues to assess their net metabolic balances [42,43,44]. The kinetics of uptake of specific metabolites is assessed by a positron emission tomography scan using radioactive tracers (e.g. 18F-2-Fluor-2-deoxy-D-glucose as a measure of glucose uptake and phosphorylation e.g., in skeletal muscle and BAT) [45]. As far as intracellular energy metabolism is concerned, phosphorus-31 magnetic resonance spectroscopy (31P-MRS) in organs and tissues is an in vivo measure of mitochondrial function e.g., phosphor creatine breakdown and ATP flux [46]; while with the concomitant assessment of oxygen uptake, the ratio of phosphorylation (i.e., the P/O-ratio) can be calculated.

    Detailed characterization of metabolism provides a sound basis for integrating metabolomics into metabolism rather than statistical calculations combined with knowledge of metabolic maps only. At any given metabolic flux, the concentrations of individual metabolites of a pathway can be low, medium or even high, reflecting the limited impact of metabolite levels as a measure to understand the physiology of intermediary metabolism. As a consequence, metabolomics on its own can provide only limited information about fluxes and control of intermediary metabolism.

    With the start of modern biomedical research, classical concepts and findings of the physiology and biochemistry of intermediary metabolism may look like yesterday’s knowledge. Method-driven big data approaches have frequently replaced “targeted” analyses of individual metabolic pathways. We may remember that during the last two generation scientists have established (i) suitable methods and applications for tracking individual metabolites by the use of isotope labeling and (ii) direct and indirect balance techniques including a detailed assessment of body composition integrating endocrine and metabolic functions [10, 47,48,49]. From that point of view, some of today’s hasty applications of metabolomics may look like a step backwards.

  • Data interpretation: Present research is characterized by two phenomena, “simplification” (in the context of this article it is meant as to make metabolism and it is control easier to understand) and “complexification” (i.e., to end up again in a complex picture of metabolism and it’s control) [1, 50]. As far as “simplification” is concerned, there are many assumptions e.g., that obesity is measured by BMI [51]; which are more or less taken as granted. However, these assumptions are not self-evident and depend on methods, metrics and concepts which are mutually agreed between scientists. Going that way of “simplifications” is very productive and may end up in a more and more publications and a so-called complex “understanding” of the issues addressed. However, the dynamics of “simplification” and, thus, the generation of complexity (i.e., “complexification”) are due to the scientists themselves (e.g., by defining standards, references, methods and measurements).

    It is surprising that a systems biology approach using large scale “omic”-analyses at the levels of the genome, the epigenome, the transcriptome, the proteome, the metabolome und the metagenome still uses that kind of “simplifications.” These concerns have been addressed recently for genome-wide association studies on obesity [52] pointing out weaknesses of the big data approach. However, particular simplifications may also be productive, e.g., adding clinical measures to simple anthropometric estimates (e.g., height) may increase productivity in the control of undernutrition at the population level [49]. A productive simplification refers to different purposes like epidemiology, population health and policy making rather than to the physiology of metabolism and its implications for clinical nutrition.

  • Blue eyed optimism about AI: Today, AI and machine learning algorithms are considered important for modern biomedical research, e.g., it has been proposed that AI has the potential “to revolutionize clinical sciences” [9]. However, the application of AI to metabolomics depends on the identified metabolites extracted from a specific sample and on the quality of that data. In addition, AI is about algorithms based on mathematics and statistics only. Modeling techniques to find associations as well as to predict and classify risks and outcomes are mostly about linear and logistic regressions, neural network analyses to approximate body functions. Amplifying and enhancing predictions and finally decision trees using categorical and numerical data are applied to define groups and classes according to health risks and diagnosis. Since AI is about algorithms only, it may not be meaningful at all. It describes what patterns of data more or less may fit to each other. Thus, strictly spoken, AI by itself cannot find the “true” answers regarding pathways, control and disturbances of metabolism. Taking a negative view, AI may add to “complexity” (and thus the “unknown”) rather than to an understanding for the benefits of science.

What about the future of metabolomics?

It is still a major challenge making sense of the metabolomic data. With application, “fingerprinting” should not be confused with mechanistic understanding. Improving the meaning of metabolite profiles related to (i) health risks or (ii) the characterization of a certain metabolic phenotype or (iii) the understanding of metabolism future investigations should include

  • a strictly controlled “basal” condition, intervention, blood and tissue sampling and probe analyses,

  • tracing metabolism by labeled isotopes to address both, metabolite fluxes, and metabolite concentrations,

  • the use of direct and indirect balance techniques including body composition analysis and indirect calorimetry to quantify substrate selection and utilization and, thus, the “net” metabolic flux at the whole body level to characterize catabolism or anabolism,

  • assessing cellular concentrations of metabolites rather than merely trusting on statistical associations between concentrations of intermediates measured in blood or urine samples only,

  • integrating metabolite profiles and fluxes into a network of data obtained from systemic physiology and other omics technologies, i.e., genomics, proteomics, metagenomics.

Today, some scientists may be more or less lost in “omics”-technologies, big data approaches, computational modeling, and integrative platforms. Much research seems to be driven by reductionism, new methods, and technologies, high research funding, high productivity of research centers, interests of publishers, and striving for profits. It presently looks like that the high dynamics of research and the “pressures” of research funding may not leave much time for some self-reflection about the reductionist approach, the whole business concept of today’s research and what an alternative way of thinking may have to offer. Faced with all the caveats which concern the sensibility and validity of metabolomics in many of its present applications we may ask a simple question: Why do scientists insist on continuing with their research when conceptual and methodological issues are apparent? The answer to that question may relate to autonomy and sovereignty of scientists. Scientists feel their independent sovereignty within their small world of research. This is about self-efficacy and a sense of agency. In addition, it is within the basic understanding of science that a chosen path has to be followed up to its end. A method-driven approach is also supported by our journals which are highly competitive and profit-oriented. From that point of view, metabolomics is a good business model and has to be seen as a story of success without competitors. However, questionable applications of metabolomic methodologies to nutrition research invites us to keep a self-critical eye on what we are doing.

In the end, such a “Metabolomic fashion” may remind us of Hannah Arendt’s critical view on the “truths” of modern scientific world published more than 60 years ago [53]: “The trouble concerns the fact that the “truths” of the modern scientific world view, though they can be demonstrated in mathematical formulas and proved technologically, will no longer lend themselves to normal expression in speech and thought…In this case, it would be as though our brain, which constitutes the physical, material condition of thoughts, were unable to follow what we do, so that from now on we would indeed need artificial machines to do our thinking and speaking. If it should turn out to be true that knowledge (in the modern sense of know-how) and thought have parted company for good, then we would indeed become helpless slaves, not so much of our machines as of our knowledge….”


  1. 1.

    Karl Popper. Conjectures and refutations. Growth Sci Knowl 2002;291:136. 374. Routledge Classics, Oxford UK

    Google Scholar 

  2. 2.

    Guijas C, Montenegro-Burke JR, Warth B, Spilker ME, Siuzdak G. Metabolomics activity screening for identifying metabolites that modulate phenotype. Nat Biotechnol. 2018;36:316–20.

    CAS  PubMed  PubMed Central  Google Scholar 

  3. 3.

    Schüssler Fionza-Rose SM, Contrpois K, Moneghetti KJ, Zhou W, Mishra T, Mataraso S, et al. A longitudinal big data approach for precision health. Nat Med. 2019;792:792–804.

    Google Scholar 

  4. 4.

    Hou Y-CC, Yu H-C, Martin R, Cirulli ET, Schenker-Ahmed M, Hicks M, et al. Precision medicine integrating whole-genome sequencing, comprehensive metabolomics and advanced imaging. PNAS. 2020;117:3053–62.

    CAS  PubMed  Google Scholar 

  5. 5.

    Ulaszewska MM, Weinert CN, Trimigno A, Portmann R, Lacueva CA, Badertscher R, et al. Nutrimetabolomics: an integrative action for metabolomic analyses in human nutritional studies. Mol Nutr Food Res. 2019;63:1800384.

    CAS  Google Scholar 

  6. 6.

    Barnes S, Benton HP, Casazza K, Cooper SJ, Cui X, Engler J, et al. Training in metabolomics. I Designment the experiment, collecting and extracting samples, and generating metabolomics data. J Mass Spectrom. 2016;51:461–75.

    CAS  PubMed  PubMed Central  Google Scholar 

  7. 7.

    Jang C, Chen L, Rabinowitz JD. Metabolomics and isotope tracing. Cell. 2018;173:822–37.

    CAS  PubMed  PubMed Central  Google Scholar 

  8. 8.

    Playdon MC, Joshi AD, Tabung FK, Cheng S, Henglin M, Kim A, et al. Metabolomics analytics workflow for epidemiological research: perspectives from the consortium of metabolomics studies (COMETS). Metabolites. 2019;9:145.

    CAS  Article  PubMed Central  Google Scholar 

  9. 9.

    NN. Why the metabolism field risks missing out on AI revolution. Nat Metab. 2019;1:929–30.

    Google Scholar 

  10. 10.

    Wolfe R, Chinkes DL. Isotope tracers in metabolic research: principles and practice of kinetic analysis. 2nd edition. Hoboken, New Jersey, US: Wiley&Sons; 2004.

    Google Scholar 

  11. 11.

    Cirulli ET, Guo L, Swisher CL, Shah N, Huang L, Napier LA, et al. Profound perturbation of the metabolome in obesity is associated with health risk. Cell Metab. 2019;29:488–500.

    CAS  PubMed  PubMed Central  Google Scholar 

  12. 12.

    Guasch-Ferre M, Hruby A, Toledo E, Clish CB, Martinez-Gonzales M, Salas-Salvado J, et al. Metabolomics in prediabetes and diabetes: a systematic review and meta-analysis. Diabetes Care. 2016;39:833–46.

    CAS  PubMed  PubMed Central  Google Scholar 

  13. 13.

    Drogan D, Dunn WB, Lin W, Buijsse B, Schulze MB, Langenberg C, et al. Untargeted metabolic profiling identifies altered serum metabolites of type 2 diabetes mellitus in a prospective, nested case control study. Clin Chem. 2015;61:487–97.

    CAS  PubMed  Google Scholar 

  14. 14.

    Koch M, Freitag-Wolf S, Schlesinger S, Borggrefe J, Hov JR, Jensen MK, et al. Serum metabolomic profiling highlights pathways associated with liver fat content in a general population sample. Eur J Clin Nutr. 2017;71:995–1001.

    CAS  PubMed  Google Scholar 

  15. 15.

    Deelen J, Kettunen J, Fischer K, van der Spek A, Trompet S, Kastenmüller K, et al. A metabolic profile of all-cause mortality risk identified in an observational study of 44,168 individuals. Nat Commun. 2019;10:3346.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  16. 16.

    Guasch-Ferre M, Santos JL, Martinez-Gonzales MA, Clish CB, Razquin C, Wang D, et al. Glycolysis/gluconeogenesis- and tricarboxylic acid cycle-related metabolites, Mediterranean diet, and type 2 diabetes. Am J Clin Nutr. 2020. pii: nqaa016.

  17. 17.

    Cheng S, Larson MG, McCabe EL, Murabito JM, Rhee EP, Ho JE, et al. Distinct metabolomic signatures are associated with longevity in humans. Nat Commun. 2015;6:6791.

    CAS  PubMed  PubMed Central  Google Scholar 

  18. 18.

    Liu X, Sadhukkan S, Sun S, Wagner GR, Hirschey MD, Qi L, et al. High-resolution metabolomics with acyl-CoA profiling reveals wide-spread remodeling in response to diet. Mol Cell Proteom. 2015;14:1489–500.

    CAS  Google Scholar 

  19. 19.

    Green CL, Mitchell SC, Derous D, Wang Y, Chen L, Han J-DJ, et al. the effects of graded levels of calorie restriction: XIV. Global metabolomics screen reveals brown adipose tissue changes in amino acids, catecholamines, and antioxidants after short-term restriction in C57BL/6 mice. J Gerontol A Biol Sci Med Sci. 2019;XX:1–12.

    Google Scholar 

  20. 20.

    Goncalves MD, Lu C, Tutnauer J, Hartman TE, Hwang S-K, Murphy CJ, et al. High-fructose corn syrup enhances intestinal tumor growth in mice. Science. 2019;363:1343–9.

    Google Scholar 

  21. 21.

    Zhang K, Han Y, Zhao Q, Zhan T, Li Y, Sun W, et al. Targeted metabolomics analysis reveals that dietary supranutritional selenium regulates sugar and acylcarnitine metabolism homeostasis in pig liver. J Nutr. 2020;150:704–13.

    PubMed  Google Scholar 

  22. 22.

    Dhurandhar NV, Schoeller D, Brown AW, Heymsfield SB, Thomas D, Sörensen TIA, et al. Energy balance measurement: when something is not better than nothing. Int J Obes. 2015;39:1109–13.

    CAS  Google Scholar 

  23. 23.

    Satija A, Yu E, Willett WC, Hu FB. Understanding nutritional epidemiology and its role in policy. Adv Nutr. 2015;6:5–18.

    PubMed  PubMed Central  Google Scholar 

  24. 24.

    Mazzili KM, McClain KM, Lipworth L, Playdon MC, Sampson JN, Clish CB, et al. Identification of 102 correlations between serum metabolites and habitual diet in a metabolomics study of the prostate, lung, colorectal, and ovarian cancer trial. J Nutr. 2020;150:694–703.

    Google Scholar 

  25. 25.

    Esko T, Hirschhorn JN, Hsu Y-HH, Feldman HA, Ebbeling CB, Ludwig DS, et al. Metabolomic profiles as reliable biomarkers of dietary composition. Am J Clin Nutr. 2017;105:547–54.

    CAS  PubMed  PubMed Central  Google Scholar 

  26. 26.

    Toledo E, Wang DD, Ruiz-Canela M, Clish CB, Razquin C, Zheng Y, et al. Plasma lipidomic profiles and cardiovascular events in a randomized intervention trial with the Mediterranean diet. Am J Clin Nutr. 2017;106:973–83.

    CAS  PubMed  PubMed Central  Google Scholar 

  27. 27.

    Gibney M, Allison D, Bier D, Dwyer J Uncertainty in human nutrition research. Nat Food. 2020.

  28. 28.

    Müller MJ, Geisler C, Hübers M, Pourhassan M, Bosy-Westphal A. Body composition-related functions: a problem-oriented approach. Eur J Clin Nutr. 2019;73:166–71.

    PubMed  Google Scholar 

  29. 29.

    Müller MJ, Enderle J, Bosy-Westphal A. Changes in energy expenditure with weight gain and weight loss in humans. Curr Obes Rep. 2016;5:413–23.

    PubMed  PubMed Central  Google Scholar 

  30. 30.

    Frayn K, Evans R. Human metabolism. 4th Ed., Hoboken-NJ, US: Blackwell-Wiley; 2019.

    Google Scholar 

  31. 31.

    Newsholme EA, Leech AR. Biochemistry for the medical science. 2nd. Ed. Hoboken-New Jersey, US: John Wiley&Sons; 1988.

    Google Scholar 

  32. 32.

    Seitz HJ, Faupel RP, Kampf SC, Tarnowski W. Influence of different types of narcosis and neck fracture on the concentration of glycolytic intermediates and related compounds in rat liver. Arch Biochem Biophys. 1973;158:12–18.

    CAS  PubMed  Google Scholar 

  33. 33.

    Faupel RP, Seitz HJ, Tarnowski W, Thiemann V, Weiss CH. The problem of tissue sampling from experimental animals with respect to freezing techniques, anoxia, stress and narcosis. Arch Biochem Piophys. 1972;148:509–22.

    CAS  Google Scholar 

  34. 34.

    Wollenberger A, Ristau O, Schoffa CT. Eine einfache Technik der extrem schnellen Abkühlung größerer Gewebestücke. Pflüger’s Arch Ges Physiol. 1960;270:399–412.

    CAS  Google Scholar 

  35. 35.

    Soares M, Müller MJ. Resting energy expenditure and body composition. Critical aspects for clinical nutrition. Eur J Clin Nutr. 2018;72:1208–14.

    CAS  PubMed  Google Scholar 

  36. 36.

    Heyland D, Muscedere J, Wischmeyer PE, Cook D, Jones G, Albert M, et al. A randomized trial of glutamine and antioxidants in critically ill patients. N Engl J Med. 2013;368:1489–97.

    CAS  PubMed  Google Scholar 

  37. 37.

    Müller MJ, Braun W, Pourhassan M, Geisler C, Bosy-Westphal A. Application of standards and models in body composition analysis. Proc Nutr Soc. 2016;75:181–7.

    PubMed  Google Scholar 

  38. 38.

    Exton JH, Lewis SB, Ho RJ, Robison GA, Park CR. The role of cyclic AMP in the interaction of glucagon and insulin in the control of liver metabolism. Ann NY Acad Sci. 1971;185:85–100.

    CAS  PubMed  Google Scholar 

  39. 39.

    Gumaa GA, McLean P, Greenbaum AL. Compartmentation in relation to metabolic control in liver. Essays Biochem. 1971;7:39–86.

    CAS  PubMed  Google Scholar 

  40. 40.

    Seitz HJ, Müller MJ, Krone W, Tarnowski W. Coordinate control of intermediary metabolism in rat liver by the insulin/glucagon ratio during starvation and after glucose refeeding. Regulatory significance of long-chain acyl-CoA and cyclic AMP. Arch Biochem Biophys. 1977;183:647–63.

    CAS  PubMed  Google Scholar 

  41. 41.

    Seitz HJ, Kaiser M, Krone W, Tarnowski W. Physiologic significance of glucocorticoids and insulin in the regulation of hepatic gluconeogenesis during starvation in rats. Metabolism. 1976;25:1545–155.

    CAS  PubMed  Google Scholar 

  42. 42.

    Wahren J, Ekberg K. Splanchnic regulation of glucose production. Ann Rev Nutr. 2007;27:329–45.

    CAS  Google Scholar 

  43. 43.

    Coppack SW, Fisher RM, Humphreys SM, Clark ML, Pointon JJ, Frayn K. Carbohydrate metabolism in insulin resistance: glucose uptake and lactate production by adipose and forearm tissues in vivo before and after a mixed meal. Clin Sci. 1996;90:409–15.

    CAS  Google Scholar 

  44. 44.

    Müller MJ. Hepatic energy and substrate metabolism: a possible metabolic basis for early nutritional support in cirrhotic patients. Nutrition. 1998;14:30–8.

    PubMed  Google Scholar 

  45. 45.

    Selberg O, Burchert W, vd Hoff J, Meyer GJ, Hundeshagen H, Radoch E, et al. Insulin resistance in liver cirrhosis. Positron-emission tomography scan analysis of skeletal muscle glucose metabolism. J Clin Invest. 1993;91:1897–902.

    CAS  PubMed  PubMed Central  Google Scholar 

  46. 46.

    Johannsen DL, Marlatt KR, Conley KE, Smith SR, Ravussin E. Metabolic adaptation is not observed after 8 weeks of overfeeding but energy expenditure variability is associated with weight recovery. Am J Clin Nutr. 2019;110:805–13.

    PubMed  PubMed Central  Google Scholar 

  47. 47.

    Umplepy AM. Tracing lipid metabolism. The value of stable isotopes. J Endocrinol. 2015;216:G1–10.

    Google Scholar 

  48. 48.

    Davis PSW. Stable isotopes: their use and safety in human nutrition studies. Eur J Clin Nutr. 2020;74:362–5.

    Google Scholar 

  49. 49.

    Charidemou E, Ashmore T, Griffin JL. The use of stable isotopes in the study of human pathophysiology. Int J Biochem Cell Biol. 2017;93:102–9.

    CAS  PubMed  Google Scholar 

  50. 50.

    Ulijaszek S. Productive simplification in the use of anthropometric nutritional status. Eur J Clin Nutr. 2020;74:359–61.

    PubMed  Google Scholar 

  51. 51.

    Müller MJ, Braun W, Enderle J, Bosy-Westphal A. Beyond BMI: conceptual issues related to overweight and obese patients. Obes facts. 2016;9:193–205.

    PubMed  PubMed Central  Google Scholar 

  52. 52.

    Müller MJ, Geissler C, Blundell J, Dulloo A, Schutz Y, Krawczak M, et al. The case of GWAS of obesity: does body weight control play by the rules? Int J Obes. 2018;42:1395–405.

    Google Scholar 

  53. 53.

    Hannah Arendt. The human condition. 1st Ed. Chicago, IL, US: University Press; 1958.

    Google Scholar 

Download references


We wish to thank Prof. Hans-Joachim Seitz, University Hospital Hamburg Eppendorf, Hamburg, Germany, for referring to his previous and fundamental work on the effects of anoxia, narcotics, euthanization, tissue sampling, and its preparation on the tissue concentrations of intermediates. Many thanks to Prof. John Blundell, Institute of Psychological Sciences, University of Leeds, Leeds, UK, for in depth discussions, ongoing motivation and his critical views on the benefits of the big data research. In addition, thanks to Prof. Mario Soares, School of Public Health, Faculty of Health Sciences, Curtin University, Perth, WA, Australia, for his valuable discussion and proof reading of the ms.

Author information




MJM had the idea and wrote the manuscript; ABW contributed to a critical discussion and impacted the final version of the manuscript.

Corresponding author

Correspondence to Manfred J. Müller.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Müller, M.J., Bosy-Westphal, A. From a “Metabolomics fashion” to a sound application of metabolomics in research on human nutrition. Eur J Clin Nutr 74, 1619–1629 (2020).

Download citation