See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/8940020

Nutrigenomics: Exploiting systems biology in the nutrition and health

arenas

Article in Nutrition · February 2004

DOI: 10.1016/S0899-9007(03)00207-7 · Source: PubMed

CITATIONS READS

122 1,408

1 author:

Ben van Ommen

TNO
320 PUBLICATIONS 13,422 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ben van Ommen on 16 December 2017.

The user has requested enhancement of the downloaded file.

INTRODUCTION

Nutrigenomics: Exploiting Systems Biology in the

Nutrition and Health Arenas
Ben van Ommen, PhD
From TNO Nutrition and Food Research, Zeist, The Netherlands

NUTRITION DISCOVERS GENOMICS www.genmapp.org). This allows for an integrated biological in-
terpretation of the observed changes, thereby strengthening the
Many diseases and disorders are related to suboptimal nutrition in pure statistical analysis or the results (Figure 2).
terms of deficits of essential nutrients, imbalance of macronutri-
ents, or even toxic concentrations of certain food compounds. In
their classic approach, nutrition scientists have dealt with this SYSTEMS BIOLOGY MAKES FULL USE OF
relation by studying the interaction of food and nutrition in human NUTRIGENOMICS DATA
intervention studies and using biomarker approaches to determine
the effect. Biochemical and molecular knowledge and technologies We begin to realize that, by using these differential display meth-
have gradually been integrated in explaining the observations ods, the vast majority of the collected data are not exploited.
made in these human studies and in underpinning postulated Multiple minor changes remain unobserved, because only the
mechanisms by in vitro and animal research. On the other end, the eye-catching differences are elaborated. This straightforward trend
biomedical research arena has unraveled a good number of mo- is stimulated by the lack of adequate statistical tools able to cope
lecular “disease mechanisms.” Currently, the two disciplines are with these new types of data sets, allowing the researcher to judge
well on their way to closely interact.1 Thus, we realize more and which of the gene expression changes are really significant. Can it
more that the nutrition and health relationship is solidly anchored be that a treasure of information is still hidden in the outcomes of
in interactions on the levels of DNA, RNA, protein, and metabo- transcriptomics, proteomics and metabolomics studies, waiting to
lites (Figure 1). be further investigated? A new way of dealing with these data is
Now that the complete human genome sequence has been currently taking shape, with the aim of making optimal use of all
unraveled, knowledge of the function of all individual human available information and thus describing “complete” biological
genes and their interaction is rapidly increasing. Technologies are processes. This new approach is called “systems biology.”8 –10
being developed that allow the simultaneous determination of the Many commercial and academic initiatives have been launched to
expression of many thousands of genes at the mRNA (transcrip- exploit this area. Major progress is envisioned through a system-
tomics) and protein (proteomics) levels. Current DNA microarray atic inventory of all relevant parameters by using genomic tech-
technology allows the simultaneous expression analyses of almost nologies and application of new bioinformatics tools together with
the complete human genome. Proteome analysis is following sev- extensive data warehousing to unravel mechanisms and define
eral tracks in its attempts to characterize the complete set of biomarker sets (Figure 3).
proteins of a tissue, such as the classic two-dimensional gel elec-
trophoresis, various LC-MS applications, and antibody arrays.2
Also, the analytical power of separating and identifying low- SYSTEMS BIOLOGY WAS MADE FOR NUTRITION
molecular-weight compounds is rapidly increasing and applied in
The point is, nutrition is not like pharmacology or toxicology,
nutrition studies as “metabolomics.”3–5 Although the methods to
where major effects can be observed, because the xenobiotic was
deal with this overwhelming amount of data and information are
designed to act on a single receptor with high affinity and strong
still in their infancy, initial examples of application of these
effects, or where dose-related pathologic effects are induced with
technologies in nutritional sciences have been published.6,7
related strong effects on transcriptomic changes. Our diet consists
Usually, these technologies are applied in a “differential dis-
of complex mixtures of many possibly bioactive chemical com-
play” mode, i.e., by comparing two situations (e.g., diseased versus
pounds, chronically administered in different compositions, and
healthy, treated versus untreated). In this way, the complexity in
with a multitude of biological effects. The vast majority of these
data is drastically reduced by examining only differences. This
biological responses are mediated through effector genes, effects
results in the identification of new receptors, possible biomarkers,
on enzyme concentration or activity, and changes in metabolite
etc., and great expectations exist regarding this approach.6 Of
concentration (Figure 1). Transcriptomics, proteomics, and
course, the major advantage of this approach is the (relatively)
metabolomics will gain in sensitivity not only because classical
open detection system paving the way for new mechanistic dis-
detection limits are lowered, but much more because multiple
coveries. This issue of Nutrition contains quite a number of ex-
minor changes taken together in new bioinformatics approaches
amples of these applications.
create a new sense of sensitivity. Multivariate statistical methods
The abundance of data allows not only the identification of
will become of major importance. Next to cluster analysis, in
individual genes, proteins, and metabolites that are differently
which the effects of single compounds or mixtures on individual
present in the samples but also the grouping of the observed
gene classes can be studied, tools such as principal component
changes into functionally or mechanistically related blocks. In-
analysis are ever more being applied to study effects on the
deed, software is being developed that visualizes the gene expres-
complete transcriptome or metabolome.
sion changes according to biochemical pathways (see, e.g.,

BIOMARKERS OF EARLY EFFECT

Correspondence to: Ben van Ommen, PhD, TNO Nutrition and Food
Research, PO Box 360, 3700 AJ Zeist, The Netherlands. E-mail: Many chronic “old-age” diseases and disorders are related to
[email protected] nutrition in the sense of prevention or of promotion. In the nutri-

Nutrition 20:4 – 8, 2004 0899-9007/04/$30.00

©Elsevier Inc., 2004. Printed in the United States. All rights reserved. doi:10.1016/j.nut.2003.09.003
Nutrition Volume 20, Number 1, 2004 Nutrigenomics in Nutrition and Health 5

FIG. 1. Health effects of food compounds are related mostly to specific interactions on a molecular level. SNP, single nucleotide polymorphism.

tion research that focuses on this relation, human intervention populations, where in fact the effect of nutrition on the disease
studies are performed by applying biomarkers to determine the state is being determined (therapy instead of prevention). Alterna-
effect of the nutritional intervention. A major dilemma arises in tively, very costly longitudinal studies need to be performed,
this type of study. Nutrition is intended to be involved in the very where large cohorts of healthy volunteers are being followed with
early stages of the (prevention of the) onset of the disease, but we nutritional intervention into the disease state.
hardly have biomarkers that are accurate, specific, and sensitive Here, the need for a new concept of biomarkers becomes
enough to determine effects before the early onset of the pathol- obvious. We would like to study the effect of nutrition in the
ogy. Thus, these studies are compromised by selecting patient healthy state and measure very early effects that predict the

FIG. 2. Schematic presentation of the pathways involved in apoptosis, with gene expression ratios alongside each gene. The ratios are derived from a 24-h
exposure comparison of 20 ␮M of eicosapentaenoic acid with 20 ␮M of linolenic acid in a cell culture system (colonic Caco-2 cells). Gene expression was
measured with an Agilent oligonucleotide array containing 17 000 human genes. The pathway presentation was made by Genmapp
(www.genmapp.org).TNF, tumor necrosis factor; TNFR1, tumor necrosis factor receptor-1.
6 van Ommen Nutrition Volume 20, Number 1, 2004

FIG. 3. Nutrigenomics and nutritional systems biology apply the same set of technologies. The nutrigenomics approach then extracts relevant differences,
which become leads for further mechanistic research. The nutritional systems biology approach aims at a complete description of the physiologic response
by exploiting the complete data sets, thus targeting a new concept of biomarker.

chronic effect in terms of prevention or promotion of a disease. quently, multiple minor differences, which (taken one by one)
Dietary therapy of atherosclerosis with antioxidant vitamins may would not have any significance, together allow for discrimination
be less effective than prevention of chronic metabolic stress between samples. The availability of such bioinformatics tools has
through optimal nutrition. a major impact on nutrigenomics, because nutrition in general does
Hence, the concept of the nutrition and health link is fully not provoke major changes in gene expression, but rather induces
appreciated only if uncoupled from a biomedical “therapy-like” multiple minor changes. Our daily food may well contain hundreds
approach and linked to the awareness that multiple minor changes of different bioactive compounds, each with its own profile of
in metabolism and its biochemical regulation contribute to the
onset of chronic nutrition-related disorders such as obesity, type 2
diabetes, cardiovascular disorders, osteoporosis, and chronic in-
flammatory syndromes. In other words, in this area, maintaining
optimal metabolism is of key importance for improving health and
preventing diseases.

DESCRIBING HOMEOSTASIS IS THE KEY

The application of “genomics” technologies in nutrition studies
may provide the key for this dilemma. In combining many subtle
changes into new biomarkers, the biomarker becomes much more
sensitive and as a consequence allows for very early detection of
changes. This has enormous potential for nutritional research.
Biomarkers will change from describing a disease state (e.g.,
plaque formation in atherosclerosis) or negative effect (e.g., oxi-
dative DNA damage) into describing subtle changes in health in
positive and negative ways. In other words, by exploiting “holis-
tic” data sets, healthy homeostasis and related early changes can be
carefully described. This means that, in exploiting nutrition in
disease prevention and health stimulation, nutritional science no
longer depends on the (often irreversible) disease end point, but
can use normal physiologic conditions as dynamic and reversible FIG. 4. Principal component analysis of the gene expression of 17 000
situations to work with. genes of the Caco-2 cell line after a 24-h incubation with LA, AA, and
Mathematical tools to deal with the complexity of the large data EPA. The complete set of gene expressions of each array is projected in
sets from transcriptomics or metabolomics experiments start to two principal components and represented by a plus sign, showing similar
become applicable. Many multivariate statistical applications patterns of behavior of the three fatty acids. The lines and numbers indicate
based on principal component analysis become routine in this type the relative contribution of the gene expressions to the relative contribution
of some specific gene expressions to the differences between the incuba-
of work (Figure 4 shows an application of principal component tions, indicated by circles. Lines pointing toward a specific sample indicate
analysis in transcriptomic analysis). Here, the relative contribution genes responsible primarily for the differences between this sample and
of all parameters (e.g., gene expression values) in the difference others at the opposite side of the panel (Yvonne Dommels, unpublished
between samples is calculated and presented. Differences between work). AA, arachidonic acid; EPA, eicosapentaenoic acid; LA, linoleic
samples thus are the result of all individual differences. Conse- acid; PC, principal component.
Nutrition Volume 20, Number 1, 2004 Nutrigenomics in Nutrition and Health 7

bioactivity, retraceable in gene expression, protein concentrations,

and metabolite profiles.
These described profiles or fingerprints are a start and have
already shown to be very sensitive. But this exercise has to be
continued. A major objective of nutritional systems biology will be
to describe physiologic homeostasis at cellular and organ levels. If
homeostasis is defined as the dynamic equilibrium between all
relevant metabolites, with underpinning gene expressions and pro-
tein activities, the “omics” technologies with their related multi-
variate statistical bioinformatics will give the nutritional scientists
an extremely powerful tool to describe this homeostasis. Ho-
meostasis will describe the healthy system, and perturbations,
based on a multitude of subtly changing parameters (which, if
taken apart, have no statistical relevance), can be traced and used
as fingerprint biomarkers of prevention.
Of course, this is easier said than done. The concept of using
description of homeostasis for biomarker purposes is valid, but
many hurdles need to be surmounted. In human plasma, intra-day
variation of many compounds, even independent of nutrition sta- FIG. 5. Vitamin C has been reported to have a beneficial effect on the
tus, is very large. Effects of age, sex, environment, genetic differ- progression of osteoarthritis. Through the vitamin C– dependent enzyme
ences, etc., need to be taken into account, in addition to nutritional lysyl-hydroxylase, vitamin C is involved in the hydroxylation of proline to
variation and development of (pre-)disease stages. Statistical meth- form hydroxyproline in the synthesis of collagen. Vitamin C also seems to
ods for describing these changes are largely lacking, and insight be required for glycosaminoglycan synthesis, by acting as a carrier of
sulfate groups. Depletion of sulfated proteoglycans is one of the earliest
into longitudinal behavior on a metabolome-wide scale has never manifestations of osteoarthritis. Relative deficiency of vitamin C may
been investigated. Thus, before these concepts can maturate, a impair the production and biochemical quality of articular cartilage.13 To
huge pile of work will need to be done. The good thing is that the investigate this possibility, a longitudinal intervention study was carried
discipline of systems biology, although especially fit for multidi- out. Hartley guinea pigs, which develop osteoarthritis during aging, re-
mensional problems and approaches in nutritional science, is also ceived a low dose (2.5–3 mg/d) that exceeded the minimum amount
discovered within other disciplines, and a number of generic tools necessary to prevent scurvy, a medium dose (30 mg/d), or a high dose (150
and methodologies will be developed in common. mg/d) of vitamin C from the age of 4 mo. Urine samples were collected at
age 12 mo and subjected to NMR analysis in which needed endogenous
and exogenous metabolites of vitamin C were eliminated from the NMR
spectra, leading to more universal osteoarthritis-related changes. Multivar-
METABOLOMICS IN NUTRITION RESEARCH iate data analysis was carried out on the NMR spectra, and the resulting
score plot is shown. The urinary NMR spectra of the guinea pigs treated
Metabolomics entails the complete and quantitative assessment of with variable vitamin C doses differed, as is clear from the differences in
the metabolome, the set of metabolites that makes up the low- their group positions on the score plot. This difference implies that the
molecular-weight fraction of cells, tissues, or body fluids. Tech- urinary metabolic composition and thus osteoarthritis metabolism are af-
nological advances in NMR and various mass spectrometry appli- fected by low versus medium versus high doses of vitamin C. Hence, from
cations indeed allow for a quite accurate assessment of large an NMR perspective, vitamin C had a noticeable effect on the development
portions of the metabolome. This emerging technology of metabo- of osteoarthritis, assuming that the guinea pigs did not have any
lomics seems to be well placed to become of major importance in comorbidity.5
nutrition research, for a number of reasons. First, many metabolites
are part of our nutrition or metabolites derived from food com-
pounds. Second, the metabolome can be regarded as the functional deviations in its code from the “standard gene.” Of course, not all
readout of transcriptomic and proteomic changes. Third, different of these polymorphisms have a functional impact. A relatively
body fluids are readily available from human studies in nutrition small number of these polymorphisms has serious health implica-
research (in contrast to human tissues that need to serve as a source tions and may even be lethal. This is the domain of clinical
for mRNA). Fourth, many “disease targets” for nutrition research genetics. Many polymorphisms, however, have only a mild effect
are directly related to the metabolome (e.g., chronic metabolic on the functionality of the resulting protein. It is here that, within
stress, syndrome X, diabetes, obesity, cardiovascular diseases, certain limits of “health,” a large variety in response to nutrition is
inflammation-related diseases, and osteoporosis). The biomarker observed. For example, the plasma cholesterol concentration is
concept as described above can be exploited with metabolomics, only partly determined by the cholesterol intake through nutrition.
where fingerprints of patterns of many metabolites together are Extremely high concentrations are observed related to specific
indicative for early changes in physiology or onset of a pathology gene mutations. These persons run a high risk of cardiovascular
(Figure 5 shows an example). Progress is being made in this area, problems and undergo drug therapy to lower plasma cholesterol.
with initial published examples being available.4 Major break- Apart from these extremes, a large variation in plasma cholesterol
through of nutritional metabolomics and its incorporation into concentration exists, with an underpinning of known and unknown
nutritional systems biology are expected with the advancements in genetic polymorphisms. Here, the possible interplay between a
data handling, especially in the area of data preprocessing of the number of polymorphisms may be, at least in part, accountable for
complex spectra derived from LC-MS and GC-MS applications. the variations between the boundaries of accepted plasma choles-
terol concentrations.11,12 In studying the effect of cholesterol-
lowering nutritional intervention, it may be wise to take these
THE ADDED COMPLEXITY OF NUTRIGENETICS subpopulations into account. Numerous other examples of inter-
individual genetic differences related to nutrition and health are
Of course, not all individuals react identically to nutrition. If known. Many single nucleotide polymorphisms resulting from
nutrigenomics describes changes in gene expression related to a inheritance or as spontaneous mutations are involved in the type 2
specific nutritional intervention, deviations in genes will have an diabetes phenotype. Inflammation-related single nucleotide poly-
impact on these transcriptome changes and ultimately on the morphisms, like the mutations in the interleukin genes, can also be
physiologic function. On average, each of our genes contains 10 influenced by nutrition (see other papers in this issue).
8 van Ommen Nutrition Volume 20, Number 1, 2004

efficacy in the relevant physiologic human context, including

interindividual variation. The multiparameter biomarkers will be
sufficiently sensitive to be used at physiologic conditions and will
be able to identify subtle changes from the homeostasis. Thus, in
developing the area of nutrigenomics, new methods of safety
evaluation for food compounds may be implemented.

CONCLUSION
Although relatively new technologies, the various genomics appli-
cations searching for new receptors and pathways already have
found their way to many nutritional applications. Moreover, the
new science of nutritional systems biology is emerging, taking up
FIG. 6. Nutritional efficacy is subject to external and internal variabilities. the challenge of exploiting all available data generated by genom-
The human genome affects the relation between nutrition and health in two ics technology in a complete description of a biological system. As
ways: 1) genetic variation, resulting in interindividual differences in re- a consequence, this new paradigm is ideally fit for the evaluation
sponse, with implications toward susceptible subgroups in the population of many subtle changes in biological activity as triggered by
(nutrigenetics), and 2) the effect of the many bioactive compounds in our nutrition. In this case, a multitude of bioactive compounds acts
nutrition on gene expression and the resulting changes in physiology simultaneously and chronically in constantly changing
(nutrigenomics). combinations.
Having said this, we realize that tools on the level of data
handling and evaluation are mostly absent. New bioinformatics
Paradoxally, food itself may contribute to this diversity, be- will be necessary to make this dream come true. Fortunately, the
cause there are many examples in which nutritional compounds road to nutritional systems biology is full of applications that can
directly cause DNA damage or modulate susceptibility (in the be used now during this passage. Indeed, it is through emerging
positive and negative sense) against DNA damage through regu- examples in the field of nutrigenomics that we begin to clearly see
lation of specific pathways involved in the many processes in- the road we need to take.
volved in these events.

REFERENCES
PERSONALIZED NUTRITION?
1. Vidal H. Gene expression in visceral and subcutaneous adipose tissue. Ann Med
Based on these interindividual differences, it is tempting to spec- 2001;33:547
ulate on “personalized nutrition” based on genotyping differences. 2. Yi EC, Marelli M, Lee H, et al. Approaching complete peroxisome character-
Of course, without stressing the genetic background of variation of ization by gas-phase fractionation. Electrophoresis 2002;23:3205
nutritional response, specific subgroups are already targeted with 3. Shi H, Vigneau-Callaha KE, Shestopalov AI, Milburt PE, Matson WR, Kristal
BS. Characterization of diet-dependent metabolic serotype: primary validation of
“subgroup nutrition” (e.g., cholesterol-lowering margarines). A
male and female serotypes in independent cohorts of rats. J Nutr 2002;132:1046
debate is arising as to whether nutrition should enter into the area 4. Watkins SM, German JB. Towards the implementation of metabolic assessment
of linking genetic differences with tailor-made nutrition. Apart of human health and nutrition. Curr Opin Biotech 2002;13:512
from the social, ethical, and communication issues involved, from 5. Lamers RAN, DeGroot J, Spies-Faber EJ, et al. Identification of disease and
a scientific point of view, a big challenge is ahead of us in nutrient related metabolic fingerprints in osteoarthritic guinea pigs. J Nutr 2003;
validating the combined action of these minor-impact polymor- 133:1776
phisms and their practical effect on the relation between nutrition 6. Müller M, Kersten S. Nutrigenomics: goals and strategies. Nat Rev Genet
and health (Figure 6). 2003;4:315
7. Van Ommen B, Stierum R. Nutrigenomics: exploiting systems biology in the
nutrition and health arena. Curr Opin Biotech 2002;13:517
8. Ideker T, Galitsky T, Hood L. A new approach to decoding life: systems biology.
COMBINED SAFETY AND EFFICACY EVALUATION Annu Rev Genomics Hum Genet 2110;2:343
9. Wolkenauer O. Systems biology: the reincarnation of systems theory applied in
Risk-benefit analysis aims at defining optimal intakes of bioactive biology? Brief Bioinform 2001;2:258
food components. To establish maximum benefit and minimal risk, 10. Kitano H. Systems biology: a brief overview. Science 2002;295:1662
effects of bioactive food components need to be assessed in the 11. Froguel P, Boutin P. Genetics of pathways regulating body weight in the
context of the diet, at relevant doses of intake, and as they occur development of obesity in humans. Exp Biol Med 2001;226:991
over long periods in the entire human body. Risk-benefit evalua- 12. Stein O, Thiery J, Stein Y. Is there a genetic basis for resistance to atheroscle-
tion needs to aim at optimization of nutrition and analysis at rosis? Atherosclerosis 2002;160:1
relevant physiologic doses of intake (rather than extrapolation 13. McAlindon TE, Jacques P, Zhang Y, et al. Do antioxidant micronutrients protect
from non-relevant high-dose animal toxicity data). Thus far, how- against the development and progression of knee osteoarthritis? Arthritis Rheum
1996;39:648
ever, adequate tools are lacking, and a classic toxicologic approach
14. Dahlquist KD, Salomonis N, Vranizan K, Lawlor SC, Conklin BR. GenMAPP,
usually has been followed in assessing potentially harmful effects a new tool for viewing and analyzing microarray data on biological pathways.
of food compounds. The nutritional systems biology concept de- Nat Genet 2002;31:19
scribed above (functional genomics applied to integrated analysis
of biological systems) can be used for assessment of safety and doi:10.1016/j.nut.2003.09.003

View publication stats