On 13 December 1902, Archibald E. Garrod published a landmark paper in The Lancet in which he described his observations of individuals with alkaptonuria, also known as black urine or black bone disease( Reference Garrod 1 ). By careful investigation of the histories of thirty-two known examples, he noted that nineteen occurred in seven families and that the condition was more common in offspring from marriages between first cousins( Reference Garrod 1 ). This helped to establish the genetic basis of the condition, but more than 90 years elapsed before a Spanish team established that mutations in the HGD gene, which encodes the enzyme homogentisate 1,2 dioxygenase are causal for the disease( Reference Fernández-Cañón, Granadino and Beltrán-Valero de Bernabé 2 ). Garrod's research led to the realisation that the laws of Mendelian inheritance apply to Homo sapiens (alkaptonuria is a classical example of an autosomal recessive disease) and initiated the study of inborn errors of metabolism, which we now know are individually rare but collectively common. Garrod's paper was subtitled ‘A study in chemical individuality’ and reflected his discovery that human biological individuality can be manifested as differences in the chemistry of body fluids( Reference Garrod 1 ), what we would now describe as the metabolome. He wrote ‘…just as no two individuals of a species are absolutely identical in bodily structure neither are their chemical processes carried out on exactly the same lines…’( Reference Garrod 1 ).
Almost 60 years later, Dr JA Roper from the University of Glasgow presented the first paper at a Nutrition Society meeting on the links between nutrition and genetics entitled ‘Genetic determination of nutritional requirements’( Reference Roper 3 ). Focussing largely on what had been learned from studies in bacteria, Roper concluded that studies in microorganisms had brought ‘…a clearer understanding of the relationship of genotype and nutritional requirements’ and that this had ‘…opened a new approach and a new way of thought to a large aspect of genotype-environment interaction’( Reference Roper 3 ).
Despite this insight and foundation in basic science, there was relatively little progress in understanding genotype–nutrition interactions in human subjects until initiation of the Human Genome Project, which was launched in 1990. On 26 June 2000, with a great deal of razzmatazz at a joint trans-Atlantic press conference, Mr Bill Clinton and Mr Tony Blair announced that the first draft of the human genome had been completed. This ushered in the ‘big science’ era in biology and led to extravagant claims about the ways in which outcomes from genomics research would revolutionise diagnosis, treatment and prevention of disease. Within a year, Peregrin predicted that the future of nutrition would be ‘nutrigenomics’ i.e. study of the interactions between genes and nutrition( Reference Peregrin 4 ). Throughout the first decade of this century, the science of nutrigenomics developed rapidly, stimulated by major infrastructure projects including the European Nutrigenomics Organisation( Reference Astley 5 ) and by increasing access to concepts, tools and resources from other areas of biology.
For the purposes of this review, I will consider nutritional genomics (or nutrigenomics) as that branch of science concerned with all types of interactions between nutrition and the genome and which is characterised by the application of high-throughput genomic (or genome-related) tools. Nutritional genetics (or nutrigenetics) is a sub-set of nutrigenomics, which focuses on understanding how genomic variants interact with dietary factors and the implications of such interactions.
Nutrigenetics: impact of genetic variation
Early tests of Garrod's prediction that different genetic variants could lead to different phenotypes involved studies of candidate gene variants (usually SNP) in small numbers of people who might not be representative of the target population. Unsurprisingly, the findings from such studies could be difficult to replicate and were likely to be subject to publication bias because researchers and journals were more likely to publish ‘positive’ findings. This area of science was transformed by the availability of tools, which enabled very large numbers of genetic variants to be interrogated simultaneously in large population groups. What is now known as genome-wide association studies is a very powerful approach for investigating genotype–phenotype relationships and for example, was used to discover the role of FTO gene variants in adiposity( Reference Frayling, Timpson and Weedon 6 ). The mechanism through which altered FTO genotype influences adiposity is unclear. An early study of ninety seven Scottish children suggested that the FTO risk variant predisposes to obesity through hyperphagia or a preference for energy-dense foods( Reference Cecil, Tavendale and Watt 7 ). However, a recent systematic review and meta-analysis of fifty-six studies reporting on 213 173 adults found that the FTO risk allele is associated with lower (not higher) reported total energy intake and with altered patterns of macronutrient intake( Reference Livingstone, Celis-Morales and Lara 8 ). Although statistically significant, these differences were small and it is uncertain whether the associations are independent of dietary misreporting( Reference Livingstone, Celis-Morales and Lara 8 ).
Unlike the ‘simple’ Mendelian genetics, which is responsible for inborn errors of metabolism such as alkaptonuria or phenylketonuria, obesity and other diet-related diseases are seldom due to single genetic variants. Indeed, recent analyses suggest that at least ninety-seven variants are involved in body fatness (Fig. 1a)( Reference Fu, Hofker and Wijmenga 10 ) and together these explain <3 % of the variance in BMI( Reference Locke, Kahali and Berndt 9 ). The affected genes are involved in multiple pathways within the central nervous system (regulating e.g. food intake and/or satiety) and in aspects of metabolism including lipid metabolism and adipogenesis (Fig. 1b)( Reference Fu, Hofker and Wijmenga 10 ). In addition, variants in genes involved in cell biology and cell signalling and in RNA binding/processing are also associated with adiposity risk (Fig. 1b)( Reference Fu, Hofker and Wijmenga 10 ). Variants in forty-nine loci are associated with differences in the distribution of body fat( Reference Shungin, Winkler and Croteau-Chonka 11 ) with high BMI/low waist : hip ratio and high BMI/high waist : hip ratio giving the classical pear and apple shapes, respectively (Fig. 1a)( Reference Fu, Hofker and Wijmenga 10 ). These fat distribution-related genetic variants are found in genes involved in adipogenesis, angiogenesis and regulation of transcription (Fig. 1b)( Reference Fu, Hofker and Wijmenga 10 ) but how they determine differences in patterns of body fat accumulation (body sites where adipocytes expand preferentially) remains to be elucidated. Information on such genetic variants can be used to develop genetic risk scores( Reference Belsky, Moffitt and Sugden 12 ), which may provide a valuable overall measure of the genetic component of obesity risk for use in diet–gene interaction studies.
Fig. 1. (Colour online) Genetic contribution to human adiposity (figure from Fu et al.( Reference Fu, Hofker and Wijmenga 10 )). (a) Human body shapes caused by different amounts and different distribution of body fat. (b) Examples of genetic variants, pathways and processes associated with fat accumulation and fat distribution. WHR, waist:hip ratio.
The early phase of nutrigenetics research was characterised by studies of candidate gene variants in small numbers of participants, often with relatively crude characterisation of dietary intake or nutritional exposure. Given that lifestyle factors such as nutrition play a major role in the aetiology of most common complex diseases( Reference Hunter 13 ), it took a surprisingly long time for those using genome-wide association studies approaches to begin to consider how dietary (and other environmental factors) might interact with genotype to modulate human phenotypes( Reference Thomas 14 ). There are now several studies which confirm that the adiposity risk associated with the FTO gene is attenuated by higher physical activity( Reference Vimaleswaran, Li and Zhao 15 , Reference Celis-Morales, Marsaux and Livingstone 16 ). In addition, investigations using the outcomes from genome-wide association studies demonstrate that dietary factors interact with genotype to determine adiposity. For example, Qi et al. showed that intake of sugar-sweetened beverages interacts with genetic risk of obesity to amply inter-individual differences in BMI( Reference Qi, Chu and Kang 17 ). Similar interactions were reported for fried foods( Reference Qi, Chu and Kang 18 ). Dietary patterns may also interact with genetic risk score to determine adiposity related outcomes( Reference Nettleton, Follis and Ngwa 19 ).
Mapping of the first human genome, announced by Clinton and Blair in 2000, had taken 10 years work by international teams of researchers and cost >$100 million. Very rapid advances in the technologies for direct sequencing of DNA have meant that the cost and time required to obtain the complete sequence of a human genome have fallen dramatically. By 2013, it was possible to sequence a human genome for about $5000( Reference Collins and Hamburg 20 ) and in 2016 this can be achieved by a robot within 24 h for <$1000. Such advances have facilitated the analysis and cataloguing of sequence variation in very large numbers of individuals and, for example, the Exome Aggregation Consortium announced recently details of sequence variation of all protein-coding regions (exomes) of 60 706 individuals of diverse ancestries( Reference Lek, Karczewski and Minikel 21 ).
Copy number variants
Investigation of human genomes has revealed several different types of genomic structural variation( Reference Mullally and Ritz 22 ). In addition to the widely-studied SNP, such variation includes gene deletion, inverted gene sequences, copy number variation (CNV; multiple copies of the same gene) and segmental duplication (duplication of a larger segment of the genome containing two or more genes)( Reference Mullally and Ritz 22 ). CNV are very common, are found on all chromosomes and depending on the level of stringency applied in their definition, account for 4·8–9·5 % of the human genome( Reference Zarrei, MacDonald and Merico 23 ).
CVN affect multiple cellular processes and the earliest discovery of a nutrition-related CNV was in AMY1, the gene encoding salivary amylase( Reference Perry, Dominy and Claw 24 ). AMY1 CNV are functionally important because AMY1 copy number is correlated positively with salivary amylase protein concentration( Reference Perry, Dominy and Claw 24 ). Furthermore, Perry et al. argued that diet may have been a factor influencing natural selection of CNV in AMY1 because individuals from populations with high-starch diets have, on average, more AMY1 copies than those with traditionally low-starch diets( Reference Perry, Dominy and Claw 24 ). In independent studies, increased AMY1 copy number was positively associated with amylase gene expression and serum enzyme concentration and perhaps surprisingly, reduced AMY1 copy number was associated with increased BMI and increased risk of obesity( Reference Falchi, El-Sayed Moustafa and Takousis 25 ). This association between higher copy number for AMY1 and lower obesity risk was confirmed in a study of Mexican children( Reference Mejía-Benítez, Bonnefond and Yengo 26 ). The human genes encoding salivary amylase (AMY1) and pancreatic amylase (AMY2) have very high-sequence identity and are found close together on chromosome 1( Reference Carpenter, Dhar and Mitchell 27 ). Although Falchi et al. reported no, or only weak, associations between AMY2 copy number and BMI( Reference Falchi, El-Sayed Moustafa and Takousis 25 ), Carpenter et al. argued that technical limitations in the approach used by Falchi et al. and the correlation in copy number between AMY1 and AMY2 mean that the specific link between amylase gene CNV and adiposity remains uncertain( Reference Carpenter, Dhar and Mitchell 27 ).
Although CNV in at least eighty-four genomic loci have been associated with human adiposity, of these, only four have been associated with BMI or obesity in multiple studies( Reference Peterson, Maes and Lin 28 ). In an analysis of 1850 European-Americans and 498 African-Americans, including CNV in genetic models of BMI category risk did not add significantly to the variation explained by SNP( Reference Peterson, Maes and Lin 28 ). In contrast, a study of Chinese children showed that higher genetic risk score, based on CNV at three loci (10q11·22, 4q25 and 11q11), was associated with higher risk of obesity( Reference Zhang, Li and Wang 29 ). The latter study also showed that dietary preferences, e.g. a meat-dominant diet, may interact with the CNV at 10q11·22 to increase obesity risk( Reference Zhang, Li and Wang 29 ).
Nutrigenomics: understanding mechanisms
In a review of the emerging science of nutrigenomics, Michael Muller and Sander Kersten predicted that ‘… applied wisely, it will promote an increased understanding of how nutrition influences metabolic pathways and homeostatic control, how this regulation is disturbed in the early phase of a diet-related disease and to what extent individual sensitizing genotypes contribute to such diseases’( Reference Muller and Kersten 30 ). This aim of increasing understanding of the molecular actions of nutrients and other dietary components and of their roles in the maintenance of normal and disturbed cellular homeostasis remains a central ambition of nutrigenomics. Many nutrigenomics investigations involve largely hypothesis-free study designs in which differences between two or more conditions (e.g. dietary treatments) are explored using genome-wide analyses at the transcriptome, proteome, metabolome and/or epigenome levels. Then, using sophisticated bioinformatics tools, attempts are made to identify genes, pathways and processes, which differ between the conditions and to use this information to illuminate the mechanisms responsible for these different conditions. Such mechanistic studies are often carried out in experimental animals, e.g. the nematode Caenorhabditis elegans or mice, because of the potential for greater control over confounding variables. For example, Lange et al. used genome wide transcriptomic approaches to investigate the effects of five types of dietary fibre on gene expression in the murine colonic mucosa( Reference Lange, Hugenholtz and Jonathan 31 ). They observed that resistant starch (which was not completely fermented in the colon) evoked a distinctive transcriptional response whilst expression profiles responses to the other four fibre sources were similar( Reference Lange, Hugenholtz and Jonathan 31 ). Importantly this study identified the nuclear receptor PPAR-γ (a member of a superfamily of ligand-inducible transcription factors) as an important regulator of the gene expression responses in the colonic mucosa( Reference Lange, Hugenholtz and Jonathan 31 ).
Classical meal feeding paradigms have been used extensively in nutrigenomic investigations. For example, in the fasted state, twenty-one healthy young men were given shakes containing predominately SFA or PUFA and post-prandial genome-wide gene expression changes in peripheral blood mononuclear cells were quantified using a microarray approach( Reference Bouwens, Grootte Bromhaar and Jansen 32 ). Ingestion of PUFA decreased, whilst SFA intake increased, expression of genes involved in liver X receptor signalling and PUFA also increased expression of genes related to cellular stress responses( Reference Bouwens, Grootte Bromhaar and Jansen 32 ). Such findings may help to explain the differential effects of these dietary lipid classes on metabolic processes and, ultimately, on health outcomes.
Nutrigenomics studies may also attempt to integrate data from different omics approaches to provide a more holistic understanding of an issue. Recently, we embedded such a nutrigenomics investigation within the Biomarkers of Risk in Colon Cancer study( Reference Greaves, Nooteboom and Elson 33 ). Since low selenium (Se) status is associated with higher risk of large bowel cancer, we tested the hypothesis that adverse effects of low Se would be evident in the apparently-normal colorectal mucosa( Reference Méplan, Johnson and Polley 34 ). We used both proteomic and transcriptomics approaches to investigate gene expression in colorectal mucosal biopsies from twenty-two healthy adults who were discordant for plasma Se concentration but matched for BMI, age and sex( Reference Méplan, Johnson and Polley 34 ). Integrating proteomics and transcriptomics datasets revealed that the low Se group had reduced inflammatory and immune responses and cytoskeleton remodelling, which suggests that suboptimal Se status may reduce cellular capacity to respond to inflammatory and oxidative stresses and, therefore, increase bowel cancer risk( Reference Méplan, Johnson and Polley 34 ).
Nutrition and the gut microbiome
Over the past decade, there has been an explosion of interest in the impact on health of microbes in, and on, the human body. It is becoming apparent that there are complex interactions between the microbiome, the immune system and whole body metabolism and that dietary factors may be central to, and may modulate, many of these interactions (Fig. 2)( Reference Verma, Hontecillas and Abedi 35 ). Our largest, most dense (>1011 cells per g contents) and metabolically active microbial community is in the large bowel and this flora is dominated by anaerobic bacteria belonging to two phyla, the Firmicutes and Bacteroidetes( Reference Louis, Hold and Flint 36 ). Unsurprisingly, the composition and metabolic activity of the large bowel microflora is strongly influenced by food intake as reported in both observational( Reference Claesson, Jeffery and Conde 37 ) and intervention( Reference Cotillard, Kennedy and Kong 38 ) studies. However, establishment of causal relationships between the human gut microflora and health outcomes such as obesity( Reference Turnbaugh and Gordon 39 ) or cancer( Reference Louis, Hold and Flint 36 ) is more challenging.
Fig. 2. (Colour online) Schematic representation of interactions between nutrition, the micriobiome, the immune system and metabolism (figure from Verma et al.( Reference Verma, Hontecillas and Abedi 35 )).
Recent research has shown that maternal malnutrition is associated with major changes in the infant gut microbiome and that these effects may be mediated by sialylated breast-milk oligosaccharides( Reference Charbonneau, O'Donnell and Blanton 40 ). These sialylated breast-milk oligosaccharides (and, perhaps other milk-derived components, which escape small bowel digestion) induce transcriptional responses in particular bacterial species, e.g. Bacteroides fragilis, which initiate a ‘cross-feeding’ cascade affecting other members of the microbiota (Fig. 3)( Reference Bashiardes, Thaiss and Elinav 41 ). Encouragingly, in animal models (mice and piglets), the addition of sialylated breast-milk oligosaccharides improved growth and development( Reference Charbonneau, O'Donnell and Blanton 40 ) but translational benefits in human infants remain to be demonstrated.
Fig. 3. (Colour online) Sialylated oligosaccharides in the breastmilk from well-nourished mothers are substrate for gut bacterial species such as Bacteroides fragilis and support a cross-feeding cascade with other species. This may promote growth and development of the infant through provision of bacterially-derived metabolites and other factors (figure from Bashiardes et al.( Reference Bashiardes, Thaiss and Elinav 41 )).
Nutrigenomics: applications to improve public health
Novel biomarkers of dietary intake
A major impediment in applying the outcomes from nutrition research to improve public health is the limitation in methodologies available for assessing dietary intake( Reference Penn, Boeing and Boushey 42 ). Here nutrigenomics approaches may advance the field through the application of metabolomics to discover novel biomarkers of food intake( Reference Favé, Beckmann and Draper 43 ). This approach is predicated on the concept that individual foods contain large numbers of (often) unique metabolites (food metabolome), which, after digestion, absorption and further metabolic processing give rise to characteristic metabolites in body fluids including blood, saliva and urine (Fig. 4)( Reference Favé, Beckmann and Draper 43 , Reference Scalbert, Brennan and Manach 44 ). Such metabolites can be detected using both targeted and untargeted metabolomics approaches. However, most biomarker discovery work to date has employed untargeted approaches and NMR and/or chromatography prefaced MS-based technologies( Reference Favé, Beckmann and Draper 43 , Reference Scalbert, Brennan and Manach 44 ). Given the chemical complexity of the metabolites in foods and the multiplicity of changes induced during digestion and metabolism, identification of the specific food-derived metabolites present in body fluids remains challenging( Reference Scalbert, Brennan and Manach 44 ). We have developed a standardised test meal-based protocol( Reference Favé, Beckmann and Lloyd 45 ) which has proved suitable for detection of a number of novel urinary biomarkers of food intake( Reference Lloyd, Beckmann and Favé 46 – Reference Beckmann, Lloyd and Haldar 48 ). Examples of putative biomarkers of specific foods identified through the use of untargeted metabolomics approaches are shown in Table 1. Foods such as sucrose, which are essentially single molecules and that, after digestion, give rise to metabolites that are difficult to distinguish from endogenous metabolites, pose particular challenges in using metabolomics approaches( Reference Beckmann, Joosen and Clarke 49 ).
Fig. 4. (Colour online) The metabolite composition of body fluids including urine, blood and saliva reflects the metabolite composition of consumed foods after digestion and metabolism. This food metabolome can be used to identify biomarkers of intake of individual foods and of eating patterns (figure from Scalbert et al. ( Reference Scalbert, Brennan and Manach 44 )).
Table 1. Examples of putative novel biomarkers of intakes of specific foods identified using untargeted metabolomics approaches*
* Data from Scalbert et al. ( Reference Scalbert, Brennan and Manach 44 ).
Given the speed of progress in using metabolomics to discover and validate novel biomarkers of food intake, which can be detected and quantified in body fluids such as urine, blood and saliva, it seems probable that we will be able to use these new biomarkers to obtain comprehensive and objective information on dietary exposure. This has potential to revolutionise dietary assessment by reducing the cost of making the measurements whilst simultaneously increasing the quality and reliability of the collected data. This will then shift the research focus to the development of easy-to-use technologies for self-collection of urine (and possibly other body fluids) at home to facilitate repeat assessments of dietary exposure and so provide a better measure of habitual dietary intake.
Personalised nutrition to enhance behaviour change
At the level of the individual, improving nutrition focuses on changing behaviours: what, when, where, with whom and how much to eat. Conventional approaches to achieving dietary change often use a one-size-fits-all approach e.g. ‘eat at least 5 portions of fruits and vegetables daily’. Such approaches can be effective but often result in only modest improvements in food intake( Reference Lara, Hobbs and Moynihan 50 ). A personalised nutrition (PN) approach is based on the hypothesis that knowledge of key characteristics of those to whom the intervention is being delivered will help in making the intervention more relevant and may increase motivation to make, and to sustain, the desired dietary change( Reference Celis-Morales, Lara and Mathers 51 ). Given the importance of interactions between diet and genetics in determining health, the use of genotypic information in designing and delivering personalised dietary advice has been a goal since the beginning of the modern nutrigenomics era( Reference Peregrin 4 , Reference Joost, Gibney and Cashman 52 ).
In a study involving 138 healthy young men and women, Nielsen & El-Sohemy tested the idea that provision of genotype-based dietary advice would result in bigger changes in dietary behaviour than advice to follow the generic Canadian dietary guidelines( Reference Nielsen and El-Sohemy 53 ). At 12 months follow up, those participants who had been informed that they carried the risk allele of the ACE gene and, therefore, should limit their sodium intake reported greater reductions in salt intake than the control group (generic dietary advice)( Reference Nielsen and El-Sohemy 53 ). However, there were no differences in intake for any of the other three targeted dietary components i.e. caffeine, vitamin C and added sugars( Reference Nielsen and El-Sohemy 53 ).
The Food4Me PN intervention study was a Europe-wide randomised controlled trial, which tested the hypothesis that providing personalised dietary advice will improve dietary intakes and markers of health, including body weight and waist circumference( Reference Celis-Morales, Livingstone and Marsaux 54 ). Adults (1607) from seven European countries were randomised to: (i) conventional dietary advice (Control) or to PN advice based on: (ii) individual baseline diet; (iii) individual baseline diet plus phenotype (anthropometry and blood biomarkers); or (iv) individual baseline diet plus phenotype plus genotype (five diet-responsive genetic variants)( Reference Celis-Morales, Livingstone and Marsaux 54 ). Participants in the study were recruited via the internet, collected and uploaded data via the web and collected biological samples (blood and buccal cells) at home using kits sent by post from the research team. In addition, they received dietary advice and other information from the research team by email and via the web( Reference Celis-Morales, Livingstone and Marsaux 54 ). At 6 months follow-up, participants randomised to the PN treatments showed significantly greater improvements in dietary intake, consumed less red meat, salt and saturated fat, and increased folate intake and had higher Healthy Eating Index scores than those randomised to the Control arm( Reference Celis-Morales, Livingstone and Marsaux 55 ). There was no evidence that including phenotypic or phenotypic plus genotypic information enhanced the effectiveness of the PN advice( Reference Celis-Morales, Livingstone and Marsaux 55 ). The Food4Me PN randomised controlled trial demonstrates that a personalised approach can produce greater improvements in eating behaviour than conventional one-size-fits-all approaches. However, at least in the context of this study, the nature of the information used to personalise that advice did not affect the outcome.
The Food4Me PN randomised controlled trial showed that the internet can be used to collect information on relevant participant characteristics and to deliver PN to large numbers of people so that this approach is potentially scalable. This could be a significant advance on conventional face-to-face interventions where the resource implications of collecting and processing the relevant information could mean that such PN interventions would be limited to more affluent sections of society. It is important that interventions should aim to narrow, rather than to exacerbate, health inequalities. In that context, digital-based technologies for intervention delivery may offer several potential advantages including convenience, scalability and reduced costs and so have the potential to narrow health inequalities( Reference Celis-Morales, Lara and Mathers 51 ). In addition, in the Food4Me PN randomised controlled trial, we observed that the quality of anthropometric data collected and recorded by participants themselves was high( Reference Celis-Morales, Livingstone and Woolhead 56 ) and that participants could use dried blood spot cards at home to collect blood for nutrient status and metabolite measurements( Reference Hoeller, Baur and Roos 57 ). These technical innovations offer considerable promise not only for future translational research aimed at developing and implementing PN interventions but also in conducting a wide range of nutrition research under real world conditions.
New opportunities for nutrigenomics
Mini-guts and other organoids
Better models facilitate better hypothesis testing and promote better science( Reference McKay and Mathers 58 ). Recent developments in stem cell biology offer exciting opportunities for developing new models for use in nutrigenomics research. Stem cells are undifferentiated cells within multi-cellular organisms, which are capable of unlimited replication and whose progeny can differentiate into several different cell types( Reference Lajtha 59 ). For example, hematopoietic stem cells within bone marrow give rise to all types of blood cells including myeloid (e.g. macrophages, erythrocytes and platelets) and lymphoid (e.g. T cells, B cells and natural killer cells) cell lineages. Research over the past three decades has resulted in the development of appropriate culture conditions and protocols for the generation of many cell lineages from both embryonic( Reference Keller 60 ) and adult stem cells( Reference Snykers, De Kock and Rogiers 61 ). In addition, the Nobel Prize winning discovery( 62 ) that differentiated cells such as fibroblasts can be reprogrammed to an embryonic-like state to produce what are known as induced pluripotent stem cells( Reference Takahashi and Yamanaka 63 ) produced a step change in understanding of the regulation of cell differentiation and laid the foundation for the new translational science of regenerative medicine.
Nutrigenomic research in individual adult human cell types is no longer restricted to using tumour-derived or transformed cell lines and can employ specific cell types derived by differentiating embryonic stems cells and induced pluripotent stem cells. Potentially even more exciting is the use of stem cells to derive three-dimensional cultures of organoids( Reference Huch and Koo 64 ). When grown under appropriate culture conditions, cell progeny from embryonic stems cells or induced pluripotent stem cells self-organise into three-dimensional organoids, which are small versions of their in vivo counter-parts( Reference Huch and Koo 64 ). Clevers and coauthors were the first to demonstrate that appropriate culture of single stem cells from the small intestine (identified by the gut epithelial stem cell marker Lgr5) would self-organise to produce crypt-villus organoids containing multiple differentiated cell types( Reference Sato, Vries and Snippert 65 ). Similarly, single Lgr5+ve cells from the stomach could be cultured in vitro to generate efficiently long-lived organoids resembling the mature pyloric epithelium( Reference Barker, Huch and Kujala 66 ). This approach has now been used to derive organoids from multiple organs and tissues including mammary gland, liver, pancreas and lung, which recapitulate key features of the corresponding human organs (Fig. 5)( Reference Huch and Koo 64 ). Human cerebral organoids, derived from pluripotent stem cells, which develop various discrete, although inter-dependent, brain regions have also been described( Reference Lancaster, Renner and Martin 67 ). Such organoids are likely to be better models of the complexity of the in vivo situation and their use provides nutrigenomics researchers with huge opportunities to undertake well-controlled experiments using very tractable models to investigate the mechanisms through which food components modulate development and function in both health and disease.
Fig. 5. (Colour online) Organoids can be grown starting from stem cells derived from multiple adult human tissues. These provide excellent, tractable models for study of the effects of nutrition on growth and development in health and in disease [figure from Huch & Koo( Reference Huch and Koo 64 )].
Genome editing
Reductionist, mechanistic research on diet–gene interactions has been accelerated by the ability to target specific genes using gene knock out and knock in approaches. Until recently, such approaches have been limited by side effects (off-target effects), which may not yield the desired phenotype. However, this field is being revolutionised by the development of much more specific genome editing techniques including use of zinc finger nucleases, transcription activator-like effector nucleases and clustered, regularly interspaced, short palindromic repeats (CRISPR)( Reference Maeder and Gersbach 68 ). In particular CRISPR, which uses short RNA sequences to target specific genomic locations, together with CRISPR-associated (Cas) enzymes can be highly specific in directing DNA cleavage( Reference Maeder and Gersbach 68 ). Another advantage of the CRISPR-Cas approach is that it can be multiplexed to facilitate the deletion and/ or insertion of multiple genomic domains. Proof of principle of the utility of CRISPR-Cas in primates was demonstrated first by a Chinese team at Nanjing Medical University who reported the birth of two macaques (Ningning and Mingming) in which the PPAR-γ and RAG-1 genes had been knocked out using the CRISPR/Cas9 system at the one-cell embryo stage( Reference Niu, Shen and Cui 69 ). Importantly, this simultaneous disruption of two target genes was achieved in one step with no evidence of off-target mutagenesis( Reference Niu, Shen and Cui 69 ).
The use of genome editing approaches such as CRISPR-Cas to target specific genes or processes in stem cell-derived organoids provides biologists, including nutrition scientists, with unparalleled opportunities for innovative, mechanistic research. Because organoids can be derived from adult stem cells( Reference Huch and Koo 64 ), this approach can begin to elucidate the mechanisms responsible for inter-individual differences in metabolism or other functions and the potential impact of nutritional factors. In principle, genome editing techniques could be used to correct nutrition-related inborn errors of metabolism such as alkaptonuria, phenylketonuria or cystic fibrosis( Reference Huch and Koo 64 ). Using such approaches in somatic tissues would have parallels with applications in regenerative medicine and are likely to be relatively non-controversial. However, the potential to use genome editing to ‘correct’ germ-line defects is much more revolutionary and raises significant ethical questions( Reference Baltimore, Berg and Botchan 70 ). On 1 February 2016, researchers at the Francis Crick Institute, London received permission from the UK Human Fertilisation and Embryology Authority to use CRISPR–Cas9 technology in human embryos for early-development research( Reference Callaway 71 ). This was the first endorsement of such research by a national regulatory authority and is aimed at new treatments for infertility. In this example, the modified embryos will be destroyed after 7 d but there is speculation that it may not be long before the first CRISPR-modified human baby is born elsewhere in the world( Reference Callaway 71 ).
Conclusions
Nutrigenomics is a young science, which is developing well. Much of the earlier hype around possible applications of the science was unhelpful and raised expectations which have not been realised as quickly as some would have hoped. However, major advances have been made in quantifying the contribution of genetic variation to a wide range of phenotypes and it is now clear that for nutrition-related phenotypes, such as obesity and common complex diseases, the genetic contribution provided by SNP alone is often modest. There is still much to be done to understand the roles of less well explored types of genomic structural variation, e.g. CNV, and of interactions between genotype and dietary factors in phenotype determination. Metagenomics-based studies suggest that the gut (and other) microbiota may have unexpected, and substantial, effects on multiple aspects of human health and that nutrition may be the most important mediator of the composition and function of our commensal flora. New tools, including stem cell-based approaches and genome editing, have huge potential to transform mechanistic nutrition research. Whilst recognising that some of these technologies bring with them significant ethical questions, nutrition researchers should grasp the opportunity to address questions of how nutrition modulates function with implications for health at all stages of the life-course. Finally, the application of nutrigenomics research offers substantial potential to improve public health e.g. through the use of metabolomics approaches to identify novel biomarkers of food intake and of dietary patterns, which will lead to more objective and robust measures of dietary exposure. In addition, nutrigenomics may have applications in the development of personalised nutrition interventions, which may facilitate larger, more appropriate and sustained changes in eating (and other lifestyle) behaviours. When combined with dried blood spot and spot urine-based measurements and with digital technologies to provide scalability, implementation of such personalised nutrition interventions could make a major contribution to better health and to reducing health inequalities worldwide.
Acknowledgements
I am grateful for the contributions of my many collaborators within Newcastle University and elsewhere which have stimulated my interest in nutrigenomics and facilitated the development of this research area both practically and conceptually.
Financial Support
My research in nutrigenomics has been supported by the BBSRC (BB/H011471/1), EU Seventh Framework Programme (265494) and Medical Research Council (MR/J010308/1).
Conflicts of Interest
None.
Authorship
J. C. M. is the sole author of this paper which is based on his Plenary Lecture at the Nutrition Society Summer Meeting 2016.
