Functional genomic approaches (e.g. transcriptomics, proteomics and metabolomics) have, for a number of years, been heralded as presenting enormous potential for advancement of nutritional research by providing new insights into diet-gene interactions and for the development of new robust biomarkers of nutritional status(Reference Elliott and Ong1–Reference Kussmann, Raymond and Affolter5). The technologies available for functional genomic research continue to expand and be refined. This progress constantly improves the scope for nutrigenomic research. But to what extent have the ‘omics’ so far lived up to their hype in advancing our understanding of diet-gene and nutrition-health interactions? This article will focus on technical progress with transcriptomic methods, their application to date for nutrition research, with particular emphasis upon micronutrients, and future needs and opportunities.
Current status of transcriptomic technologies
The first microarrays were developed over 10 years ago(Reference Schena, Shalon, Davis and Brown6) but the realisation of the full potential of transcriptomics has been held up to some extent by the inevitable need for adequate technology development. In the interim, there have been problems associated with data quality, data analysis methods, genome coverage and genome annotation. Now these issues appear to have been largely resolved. For example, concerns about the ability to obtain reproducible results across different array platform technologies no longer appear to be legitimate since the MicroArray Quality Control project demonstrated that is it possible to obtain high concordance in inter-site and cross-platform comparisons using well established commercial array platforms(Reference Guo, Lobenhofer and Wang7). The range and scope of methods for analysing array data have also improved dramatically over the last few years so that it is now feasible to reliably identify even quite subtle alterations in the mRNA levels of discrete genes and to start to interpret alterations in gene expression profiles in the context of pathways, biological processes and cell signalling mechanisms. Examples of such approaches are discussed below.
Transcriptome coverage is no longer a genuine concern, since (i) the number of genes in mammalian genomes has proven to be far lower than initially expected and (ii) platform technologies have improved to a point at which the number of features that can be represented on a single array far exceeds the number of genes in a genome. Indeed, this has made it possible to extend the scope of microarrays to examine more than just mRNA levels for all genes. For example, array technologies exist that cover every individual gene exon, thus enabling expression analysis at the level of gene splice variants(Reference Xing, Kapur and Wong8, Reference Gardina, Clark and Shimada9). In fact, DNA arrays are also being applied to perform a wider range of analyses including genome-wide DNA methylation profiling, high-throughput genotyping and haplotying, comparative genomic hybridization for copy number determination and microRNA expression profiling (Fig. 1).
Fig. 1 Schematic representation of the current range of applications for DNA microarray-based technologies in the broader context of structural and functional ‘omic’ research areas.
Another equally important issue has been the development and adoption of (i) standards for capturing microarray and associated study metadata (MIAME)(Reference Brazma, Hingamp and Quackenbush10), standards and systems for exchanging microarray data (MAGE-OM and MAGE-ML)(Reference Spellman, Miller and Stewart11), (iii) open access resources, such as GEO and ArrayExpress(Reference Parkinson, Sarkans and Shojatalab12, Reference Barrett, Suzek and Troup13), that make published microarray data freely available to all, and (iv) initiatives to provide research communities with common sets of descriptors (controlled vocabularies) arranged in ontologies (http://obi.sourceforge.net). Together, these efforts reflect the viewpoint that, given the significant level of resource that has to be put into microarray studies and the huge volume of data generated, this process should be considered a long-term investment for the benefit of the whole research community. These standards, systems and structures provide key elements of a framework to make the most of the stored data: facilitating the development of new analytical methods, helping in the training of new researchers and providing the scope for detailed structured searches of stored data and mining of data combined from multiple studies(Reference Rocca-Serra and Elliott14).
Thus, overall, it seems reasonable to assert that transcriptomic technology has finally achieved a level of technical maturity that means it is feasible for committed researchers to generate high quality experimental data and, with appropriate statistical and bioinformatics expertise, to extract robust biological information from that data.
Current application of transcriptomics in nutrition research
In the context of nutrition research, transcriptomic methods have already been applied fairly extensively, most commonly using the general study design format depicted in Fig. 2.
Fig. 2 Schematic representation of the most common experimental format employed for nutritional studies employing transcriptomics analysis.
Tables 1 and 2 illustrate the types and range of studies published in the context of micronutrient research, listing examples of microarray studies that have examined patterns of gene expression associated with varying iron and zinc exposure or status, respectively. Similar tables could equally be drawn up for many other nutrients, micronutrients and non-nutrient components of the diet.
Table 1 Examples of published transcriptomic studies examining the effects of iron on gene expression profiles
Table 2 Examples of published transcriptomic studies examining the effects of zinc on gene expression profiles
Two key observations can be made from the information in these tables. The first is that all the transcriptomic studies listed have been performed in animal models, mammalian cell lines or primary human cells ex vivo, but not in samples derived from human nutritional intervention studies. The second is that, whist these studies have served to identify a host of genes that are regulated at the level of mRNA by different nutritional components, this information has not yet been combined to provide a single coherent overview and has not yet led to the development of new biomarkers. These two observations hold true in general across the spectrum of nutrition research and are discussed in more detail in the following sections.
The scope for transcriptomics-based nutritional studies in human subjects
Proponents of transcriptomic approaches should particularly note the general absence of microarray data obtained from human nutritional intervention studies. Presumably this reflects a general reticence to undertake such time-consuming and resource-intensive work without confidence that it will yield useable and valuable data: limitations in access to target tissues, technical challenges of isolating sufficient high quality RNA and the analytical problems of dealing with inter-individual variation in human subjects are all valid concerns. Nevertheless, the limited evidence that is available suggests that transcriptomic analysis in nutritional intervention studies is feasible and can be highly informative.
The most obvious source of cellular RNA for such studies is blood since this is readily obtained and, as such, ideally suited to gene-expression based biomarker development. However, just as with most solid tissues, blood contains a number of different cell types, each with their own unique gene transcript profile. This must be taken into consideration when performing gene transcript analyses with blood samples. Although red blood cells (erythrocytes) are by far the most abundant cell type in blood, they pose a number of technical problems for gene expression analysis. Firstly, mature erythrocytes are anucleate and, therefore, are unable to modify gene transcription in response to environmental changes. Instead, gene transcript profiles of erythroytes are effectively set in the immature reticulocyte. Secondly, the quantity and quality of RNA in an erythrocyte will depend on its age since no new mRNA can be synthesized after the cell loses its nucleus and RNA present at this stage will inevitably undergo some degradation over time. Finally, haemoglobin mRNA predominates in reticulocytes, reflecting the highly specialised function of red cells. This predominance of a single message can hamper microarray analysis.
For these reasons, white blood cells (leukocytes) are commonly used for gene transcript analysis. Leukocytes have been described as scouts, maintaining continuous surveillance for signs of infection or other threats(Reference Whitney, Diehn, Popper, Alizadeh, Boldrick, Relman and Brown15). Equally, they are exposed to acute and chronic fluctuations in plasma composition arising from the sporadic and varied nature of the diet and the metabolic consequences of exposure to different levels and types of dietary components. Thus they represent a potentially valuable and widely used tool for nutritional intervention studies with human subjects both as direct targets for studying effects on immune and inflammatory processes and as surrogates for direct sampling of less accessible tissues.
There are a number of different types of leukocytes (generally categorised into neutrophils, eosinophils, basophils, lymphocytes and monocytes although some of these categories can be further sub-divided). The relative proportions of these different cell types can vary somewhat from sample to sample, with knock-on effects on the ‘average’ gene expression profiles(Reference Radich, Mao and Stepaniants16). This noise can be reduced by isolation and analysis of specific leukocyte sub-sets although care should be taken to avoid, or at least minimise, the risk of introducing artefacts in gene expression that could arise from environmental stresses applied to the cells during the additional isolation steps(Reference Eady, Wortley, Wormstone, Hughes, Astley, Foxall, Doleman and Elliott17).
Analyses of normal variation in leukocyte gene expression suggest that it is best to design studies in which subjects act as their own controls, since the extent of intra-individual variation, in the absence of any intervention, is markedly lower than that of inter-individual variation(Reference Whitney, Diehn, Popper, Alizadeh, Boldrick, Relman and Brown15–Reference Eady, Wortley, Wormstone, Hughes, Astley, Foxall, Doleman and Elliott17). Using such an approach, van Erk and co-workers were able to characterise the specific acute effects of a high-protein and a high-carbohydrate breakfast on the expression of a range of genes in blood leukocytes obtained from human volunteers(Reference van Erk, Blom, van Ommen and Hendriks18). The response to the high carbohydrate meal was notably associated with changes in expression of genes involved in glycogen metabolism, whereas the response to the high protein breakfast was noted to include altered expression of genes involved in protein biosynthesis.
By employing microarray analysis of RNA obtained from blood lymphocytes of post-menopausal women before and after dietary supplementation with high-dose purified soy isoflavones, Niculescu and co-workers were able to detect altered expression of a large number of genes, including induction of genes associated with cyclic cAMP signalling and cell differentiation and decreased expression of genes associated with cyclin-dependent kinase activity and cell division. They also found that the response to isoflavone supplementation was dependent on the volunteers' capacity to produce equol. This observation is interesting since equol is a bacterial metabolite of the isoflavone daidzein to which some of the beneficial effects of isoflavones have been tentatively attributed. The capacity to metabolise daidzein to equol is dependent the microflora of the gut, such that approximately 30–40 % of humans produce it. Thus the data from this study could provide new insights into functional effects of equol in relation to the apparent beneficial effects of isoflavones(Reference Niculescu, Pop, Fischer and Zeisel19).
The recent study of Pagmantidis and co-workers, demonstrated that a 6 week dietary supplementation of human volunteers with moderate levels of selenium elicited small but detectable changes in the expression of blood lymphocyte genes, in particular of genes encoding proteins that function in protein biosynthesis(Reference Pagmantidis, Meplan, van Schothorst, Keijer and Hesketh20). Overall, these data clearly support the potential value of transcriptomic analysis of human tissue samples in human nutritional intervention studies.
Analytical methods for transcriptomic studies and the development of new nutritional biomarkers
As noted above, microarray analysis of samples from nutritionally relevant studies in animal and cell models have led to the identification of many genes that are regulated at the mRNA level by exposure to different dietary components. This information has helped to provide new insights into the range of biological processes that can be affected by different nutritional parameters but a great deal more can and should be done in this regard. Closer inspection of the literature suggests that, in the past, there has been a widespread tendency to allocate insufficient time and resources for the data analysis phase of transcriptomic studies (phase 2 in Fig. 1), with many studies achieving little more than generating lists of up and down-regulated genes. This is entirely understandable, since the complexity of the analytical process is often not fully appreciated until it is attempted firsthand. Such vast datasets require time, unique expertise and advanced computing tools to interpret properly. However, the number of people with relevant expertise and the range of analytical tools have increased and become more widely accessible and this, in turn, is reflected in the more detailed analyses performed in the more recent studies. Functional and pathway analysis, literature mining tools, are now employed routinely, providing a clearer biological context for the results obtained and, just as importantly, facilitating cross-study comparisons(Reference Coppin, Darnaud, Kautz, Meynard, Aubry, Mosser, Martinez and Roth21–Reference Mootha, Lindgren and Eriksson26). Improvements in data quality and data normalisation strategies together with the development of powerful statistical approaches, such as gene set enrichment analysis, have combined to enhance the sensitivity of transcriptomic methods to such an extent that they genuinely compete with real-time RT-PCR, widely considered as the ‘gold standard’, in detecting subtle changes in gene expression(Reference Mootha, Lindgren and Eriksson26, Reference Subramanian, Tamayo and Mootha27). This is particularly important for nutrition research where only small changes in gene expression are generally expected to be brought about by dietary modification but, nevertheless, such changes may have profound effects in the long term on health.
In terms of gene expression-based biomarker development, progress to date has been more disappointing. In a few cases, results from nutritionally-relevant transcriptomic studies performed in vitro or in animal models have been taken forward into human studies to the test the effects of nutritional interventions on the mRNA levels of individual genes in circulating blood leukocytes using techniques such as real time RT-PCR(Reference Aydemir, Blanchard and Cousins28). Such studies provide further evidence that nutrient regulated changes in the expression of specific genes can be detected in vivo in samples from human subjects. But, although a number of potential gene expression-based nutrient sensitive biomarkers have been identified, these suffer from potential confounding effects that undermine their value. This reflects a fundamental problem with the specificity of single genes as biomarkers since expression of most (perhaps all) individual genes can be regulated by more than one environmental factor.
An interesting alternative that may help to overcome this problem is the development of markers based on expression profile ‘signatures’ rather than single genes. Such ‘signatures’ (characteristic patterns of differential gene expression) are effectively measures of cell phenotype, and can be used to look for novel biomarkers in cells that have been exposed to different levels of micronutrients. Approaches designed to identify characteristics gene expression ‘signatures’ are being developed and tested widely in the context of medical research, in particular cancer research(Reference Naderi, Teschendorff and Barbosa-Morais29–Reference Xu, Tan, Winslow and Geman32). Although still in a comparatively early developmental stage, the applicability and power of such approaches should also be evaluated in the context of nutrition.
Conclusions
Even during their initial developmental phase, transcriptomic approaches have already provided many valuable new insights into diet-gene interactions. Now the technology has reached a sufficient level of maturity that it can be used as a routine tool by all researchers who wish to exploit it to produce high quality data. The emphasis needs to shift towards more extensive, disciplined and detailed examination of the data generated by using advanced statistical, data- and literature-mining tools. The development of universal standards and infrastructures for microarray data capture, storage and sharing will be vital in ensuring that data produced can used to it full potential. Although not yet comprehensively tested, the scope for human nutritional intervention studies, which employ transcriptomic analysis of accessible tissues, appears very promising. This kind of study represents an exciting avenue of investigation for the development of new nutritional biomarkers.
Acknowledgements
The Institute of Food Research is funded by the Biotechnology & Biological Sciences Research Council (BBSRC), UK. Dr Elliott is a member of NuGO The European Nutrigenomics Organisation: linking genomics, nutrition and health research (NuGO, CT-2004-505944) a Network of Excellence funded by the European Commission's Research Directorate General under Priority Thematic Area 5 Food Quality and Safety Priority of the Sixth Framework Programme for Research and Technological Development. The content of this article presents no conflicts of interests.
