Dr J. A. Lovegrove, Department of Food Biosciences, School of Chemistry, Food Biosciences and Pharmacy, University of Reading, Whiteknights PO Box 226, Reading RG6 6AP, UK. Tel.: +44 (0)118 378 6418 Fax: +44 (0)118 931 0080 E-mail: firstname.lastname@example.org
Cardiovascular disease (CVD) is responsible for significant morbidity and mortality in the Western and developing world. This multi-factorial disease is influenced by many environmental and genetic factors. At present, public health advice involves prescribed population-based recommendations, which have been largely unsuccessful in reducing CVD risk. This is, in part, due to individual variability in response to dietary manipulations, that arises from nutrient–gene interactions (defined by the term ‘nutrigenetics’). The shift towards personalized nutritional advice is a very attractive proposition, where, in principle, an individual can be given dietary advice specifically tailored to their genotype. However, the evidence-base for the impact of interactions between nutrients and fixed genetic variants on biomarkers of CVD risk is still very limited. This paper reviews the evidence for interactions between dietary fat and two common polymorphisms in the apolipoprotein E and peroxisome proliferator-activated receptor-γ genes. Although an increased understanding of how these and other genes influence response to nutrients should facilitate the progression of personalized nutrition, the ethical issues surrounding its routine use need careful consideration.
Conflict of interests, source of funding and authorship
The authors declare that they have no conflict of interest.
No funding declared.
JAL and RG prepared the manuscript. All authors critically reviewed the manuscript and approved the final version submitted for publication.
Cardiovascular disease (CVD) is the leading cause of death worldwide, with mortality rates within the UK being amongst the highest in the world (BHF, 2007; Allender et al., 2007). Our existing knowledge of risk factors for CVD is extensive, and underlies well-established and accepted guidelines for the primary and secondary prevention of this disease. However, further understanding of the aetiology and efficacy of treatment of this complex multi-factorial condition will require exploration of how genetic factors interact with the environment. The possibility of offering personalized nutrition advice to the individual is an attractive option for dietitians and nutrition scientists and is becoming practicable with the emergence of nutritional genomics. This developing field promises to revolutionize dietetic practice, with dietary advice prescribed according to an individual’s genetic makeup to prevent, mitigate or cure chronic disease (Kaput & Rodriguez, 2004). Nutritional genomics is a general term encompassing two fields of research: ‘nutrigenomics’ and ‘nutrigenetics’. Although the concepts of nutrigenomics and nutrigenetics are intimately linked, their meanings and purpose are fundamentally different for understanding the relationship between diet and genes. Nutrigenomics focuses on how specific nutrients or dietary constituents affect gene expression, protein and metabolite concentration. Nutrigenetics is concerned with the effects of fixed genetic variation [e.g. the effect of inheriting a particular variant of a gene (single nucleotide polymorphisms; SNP)] on an individual’s responsiveness to a particular diet or nutrient (Ordovas & Mooser, 2004; Corthesy-Theulaz et al., 2005). However, despite these discrete definitions, nutrigenomics and nutrigenetics are terms that are often used synonomously to describe the study of nutrient–gene interactions.
Nutritional genomic research is progressing rapidly, with investigations into the prevention and treatment of major chronic diseases such as CVD, cancer, type 2 diabetes, asthma, obesity, Crohn’s disease and osteoporosis to name but a few. This present review concentrates on the effect of genetic variation in relation to risk factors for CVD (nutrigenetics). To illustrate the complexity and confounding issues associated with nutrigenetics and the interpretation of study outcomes, polymorphisms of two genes [apolipoprotein E (apoE) and peroxisome proliferator-activated receptor (PPAR)-γ)] will be considered within the context of human ethics.
Present and future strategies for disease management
Diagnosis of disease at onset, usually by the identification of clinical symptoms or biochemical biomarkers such as raised plasma glucose, is the first step in the management of disease. Although general nutritional therapy or lifestyle advice, where effectively ‘one size fits all’, is normal practice, this approach has had limited success in benefiting the patient, either because of failing public health recommendations or a lack of patient compliance and motivation. However, with the advancement of nutrigenomic and nutrigenetic research, a shift to a ‘personalized nutrition’ strategy appears to be attainable. Some researchers predict that, in the future, a genetic profile will be as routine and commonplace as having a blood cholesterol test, leading to a change from generalized treatment to early detection and prevention based on an individual’s genetic predisposition. In this way, nutrigenetics and nutrigenomics would herald an unprecedented advance in the management and prevention of CVD but also present some ethical dilemmas at the same time.
Although humans share the same genome, there are many common variations in codon sequences amongst nutritionally relevant genes. It is estimated that there are over 10 million common SNPs in the human genome (McVean et al., 2005). Some common SNPs, which are therefore of Public Health relevance, occur in 5–50% of the population. Most individuals are heterozygous for greater than 50 000 SNPs across their genes (Hinds et al., 2005). A number of these SNPs result in the alteration of the end product of gene expression, proteins, with altered structure or function. Studying the possible ways in which combinations of numerous SNPs may influence metabolic responses to specific nutrients and nutrient requirements has been a practical impossibility until very recently. Within the space of a few years, major technical advances have made such work a reality. Two common strategies in nutrigenetic research include an analysis of candidate genes or a genome-wide linkage screen.
Candidate gene approach
The candidate gene approach involves the selection and study of biologically relevant genes. Genetic polymorphisms in these genes, known as SNPs, can alter susceptibility to a disease. Likewise, dietary constituents may preferentially interact with a particular gene variant to influence disease risk. Candidate or ‘susceptibility’ genes should meet one or more of the following conditions: genes that are chronically activated during a disease state and have been previously demonstrated to be sensitive to dietary intervention; genes with functionally important variations; genes that have an important hierarchical role in biological cascades; polymorphisms that are highly prevalent in the population (usually >10% for public health relevance); and/or genes with associated biomarkers, rendering clinical trials useful (Mutch et al., 2005).
To reduce the number of SNPs in an analysis, it is of great benefit to identify haplotypes (haploid genotype). Haplotypes are a set of closely linked genetic markers present on one chromosome, which tend to be inherited together (as they are not easily separated by recombination) with high linkage disequilibrium. Haplotypes can be identified by patterns of SNPs with the use of HapMaps. Over the past few years, an International Scientific Consortium has characterized patterns of SNP linkage in haplotype blocks (Hinds et al., 2005). The identification of a few alleles in a haplotype block can unambiguously identify all other SNPs that reside on that stretch of DNA. Utilization of these ‘tag SNPs’ excludes the need to measure all SNPs in the haplotype and facilitates practical SNP analysis for the nutrition scientist.
Genome-wide linkage screen
A genome-wide, linkage screen determines polymorphisms in the complete genome and relates these to a dependent variable such as body weight. This identifies genes that have a statistically significant relationship with the variable of interest. Although this method is often criticised for being nonhypothesis driven and a ‘fishing exercise’, it has advanced scientific knowledge by uncovering unexpected links between genes and CVD risk factors. An example of this is the recent link reported between the nonfunctional FTO gene, the SNP RS993609, and the incidence of obesity (Frayling et al., 2007). It was reported that the A allele of this SNP was associated with an increased risk of being overweight or obese compared to the T allele (30% and 70% in carriers of one and two alleles, respectively) in a white European population of adults and children (38 759 participants). The association was mediated through changes in fat mass and was observed from the age of 7 years upwards, with an interstitial deletion in the same region causing obesity (Stratakis et al., 2000).
There are an increasing number of published studies that have investigated the nutrigenetic links between CVD risk factors and dietary components. The following are examples of nutrient–gene interactions for which there is a strong evidence base.
Polymorphisms in apoE
The population response to changes in dietary fat intake has been extensively studied. A considerable degree of heterogeneity has been observed between individuals in response to the same dietary intervention. However, this variability has rarely been addressed in any detail and has often been ascribed to variation in dietary compliance between participants. A classic example of this is the large variation in the concentration of serum low-density lipoprotein-cholesterol (LDL-C) in response to fish oil supplementation. The cardioprotective effects of the fatty acids in fish oil, eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) are well recognized (Burr et al., 1989; Bucher et al., 2002; Kris-Etherton et al., 2003). However, a potentially deleterious increase in LDL-C (5–10%) has been consistently reported after moderate to high doses of fish oil (>2 g day−1 EPA + DHA) (Minihane et al., 2000; Brady et al., 2004; Lovegrove et al., 2004). Despite this small, but significant increase in LDL-C, closer examination of the responses revealed a marked inter-individual variation. Figure 1 illustrates the range of LDL-C responses for 74 subjects following a 2.5 g day−1 EPA + DHA supplement for a 6-week period. There was a mean increase in LDL-C of 4.1%, yet the spread of individual responses was substantial, with 33 of the 74 subjects demonstrating a lower serum LDL-C and the remaining 41 demonstrating a higher LDL-C (range −40 to +113%) following fish oil intervention [data collated from Brady et al. (2004) and Lovegrove et al. (2004)]. This heterogeneous response to a change in dietary fat, may be attributed to a number of factors, including age, gender, baseline LDL-C levels, disease status and drug use. However, recent evidence strongly suggests that variation in a number of key genes may also be important, including common variants of the apoE gene.
The most convincing evidence to date for genotypic effects on dietary response comes from the extensively studied apoE gene variant. The apoE protein has a central role in lipoprotein metabolism, being involved in chylomicron metabolism, very low-density lipoprotein synthesis and secretion, and in the cellular removal of lipoprotein remnants from the circulation (Mahley, 1988; Minihane et al., 2007). This gene locus is polymorphic, with 84 gene variants being characterized to date (http://www.ncbi.nlm.nih.gov). The best known polymorphism is the common and widely characterized apoE epsilon missense mutations, which results in three allelic isoforms, namely ε2, ε3 and ε4. The proteins produced from the different isoforms differ in the amino acid present at residue 112 (rs429358) and 158 (rs7412), with apoE2 containing Cys at both sites, apoE3 112 Cys/158 Arg, and apoE4 Arg at both positions (Rall et al., 1982; Weisgraber, 1990). The prevalence of this SNP varies in different populations. It is reported that approximately 61% of Caucasians are homozygous E3/E3, 25% are E4 carriers (E3/E4 or E4/E4), 12% are E2 carriers (E2/E2 or E2/E3), with the remaining 2% having an E2/E4, genotype (Eichner et al., 2002; Song et al., 2004; Singh et al., 2006) (Fig. 2).
The impact of apoE genotype on CVD risk has been extensively investigated over the past 30 years. A meta-analysis has been published recently that summarizes the overall findings from studies using a variety of endpoint measures (Song et al., 2004). A mean increase in CVD risk of 40–50% was observed in E4 carriers [overall odds ratio (OR) = 1.42] relative to the wild-type E3/E3 genotype, with no apparent differences for either the E2 and E3 subgroups (OR = 0.98) (Song et al., 2004) (Fig. 2). Although a causal mechanism to link E4 with increased CVD risk has not been fully elucidated, the association has been ascribed to a higher concentration of LDL-C. This is believed to arise from apoE4 isoform having a relatively higher affinity for its membrane (LDL/chylomicron remnant) receptor and feedback inhibition of receptor activity in E4 carriers (Kesaniemi et al., 1987; Demant et al., 1991). Other mechanisms relating to reduced antioxidant status may also be operative (Minihane et al., 2007).
The majority of studies support an interaction between apoE genotype and CVD risk, yet there is inconsistency in the evidence for this and other nutrient–gene interactions. This may be due to the retrospective genotyping of study cohorts. In this situation, the genotype-diet-LDL-C interactions were not the primary outcome, resulting in the less frequent genotypes being under-represented. Consequently, although clear trends may have been evident, many of the studies were under-powered to detect a significant genotype–treatment effect. It is therefore important for nutrigenetic studies to consider the use of prospective genotyping to increase statistical and discriminatory power.
Although a number of previous studies have observed effects of apoE genotype in response to dietary total fat and saturated fatty acid (SFA) manipulation (Huang et al., 1999; Masson et al., 2003; Minihane et al., 2007), only one study to date has examined the apoE genotype–dietary fat-LDL-C association using prospective recruitment by genotype. Sarkkinen et al. (1998) reported a significant effect of apoE genotype on the plasma lipid response to a low fat diet, with a 5%, 13% and 16% reduction in LDL-C in E3/E3, E3/E4 and E4/E4 males, respectively. Other studies have examined the association between apoE genotype and fish oil (EPA/DHA) on LDL-C responses. In a study by Minihane et al. (2000), it was observed that a mean increase of 7.1% in LDL-C for the group as a whole was solely attributable to a 16% rise in LDL-C in the apoE4 participants, and it was speculated that apoE genotype may, in part, predict the blood lipid response to fish oil intervention (Fig. 3). Variable effects of EPA and DHA on LDL-C have been reported previously. Two prospectively genotyped studies recently completed by Minihane and colleagues, demonstrated that: (i) apoE-fish oil-LDL-C interactions are only evident at intakes greater than 2 g EPA + DHA per day (A. M. Minihane, personal communication) and (ii) it is the DHA rather than EPA in fish oils that is responsible for the LDL-C raising effects in E4 individuals (A. M. Minihane, personal communication).
PPAR-γ is a nuclear transcription factor involved in the regulation of a number of key genes in relation to fat metabolism and inflammation (Auwerx, 1999; Rosen & Spiegelman, 2001). It is expressed predominantly, although not exclusively, in adipose tissue where it regulates adipocyte differentiation and fat metabolism through a complex programme of gene expression (Cecil et al., 2006). PPAR-γ is also present in other tissue, such as muscle, monocytes and the endothelium, where it also plays a role in controlling insulin sensitivity, blood pressure and inflammation (Memisoglu et al., 2003a). As a consequence of these regulatory functions, PPAR-γ is involved in the development of obesity and is a prime candidate gene for CVD nutrigenetic research (Cecil et al., 2006). A common polymorphism of the PPAR-γ gene (pro12ala) has been widely studied and may be associated with increased adiposity and insulin resistance (Gonzalez Sanchez et al., 2002) but, interestingly, with a decreased risk of metabolic syndrome and type 2 diabetes (Altshuler et al., 2000; Frederiksen et al., 2002; Soriguer et al., 2006). A more efficient suppression of lipolysis and lower circulating nonesterified fatty acids (NEFA) concentrations have been observed in pro12ala carriers during a hyperinsulinaemic-euglycaemic clamp (Stumvoll et al., 2001). However, no relationship between fasting NEFA levels and the polymorphism was found in the general population in another study (Vaccaro et al., 2002).
Despite extensive research into the interactions between pro12ala and dietary fat intake, the reported outcomes are somewhat conflicting. In a study by Memisoglu et al. (2003b), total dietary fat intake had no effect on the body mass index (BMI) of individuals with the pro12ala polymorphism compared to wild-type (pro12pro) individuals. Carriers of the pro12 ala variant had higher BMI than noncarriers within the lowest quintile of total fat intake. Total fat intake was directly correlated with plasma HDL-C among pro12ala variant carriers; by contrast, among pro12pro homozygotes, total fat intake was inversely correlated with HDL-C and total cholesterol. Intake of SFA was associated with increased BMI in both genotypes, whereas there was no association of MUFA intake and BMI in pro12pro individuals, but an inverse association in the 12ala polymorphism carriers. Robitaille et al. (2003) explored the effect of diet and the PPAR-γ gene on components of the metabolic syndrome and also reported that total dietary fat intake correlated positively with BMI only among homozygous wild-type pro12pro individuals. Similar results were observed when saturated fat intake was considered. Total fat and saturated fat intakes were also positively correlated with fasting glucose, waist circumference and the TC : HDL-C ratio, and negatively with HDL-C only in pro12pro carriers. These two studies were unable to confirm the interaction between the ratio of dietary polyunsaturated fatty acid (PUFA) to SFA (P : S ratio) and this polymorphism as observed by Luan et al. (2001) who reported that BMI and fasting insulin were inversely related to P : S ratio in ala12 carriers, but not in pro12 homozygotes (Fig. 4). Nevertheless, when the analysis was repeated for total fat intake, they found no interaction of genotype with BMI or fasting insulin (Luan et al., 2001). Soriguer et al. (2006) provided evidence to support the existence of an interaction between this polymorphism and obesity according to diet. Obese subjects with the ala12 allele who consumed fewer MUFA were more insulin resistant. This study was undertaken in a Mediterranean population with high intake of MUFA, whereas the previous studies comprised populations with diets rich in PUFA. In a 3-year longitudinal study in which subjects with impaired glucose tolerance were randomized to an exercise and low fat diet group versus a control group (Lindi et al., 2002), ala/ala homozygotes lost more weight than pro/pro and pro/ala subjects on the intervention arm. Pro12ala individuals lost similar amounts of weight to pro/pro subjects following hypoenergetic diets in a study by Nicklas et al. (2001), although they had more carbohydrate oxidation, less fat oxidation and a greater response in insulin sensitivity. Poirier et al. (2000) found no significant interaction between pro12ala and BMI, glucose or fat tolerance, or fasting levels of glucose, lipids or insulin in young healthy subjects. This finding suggested that the differential effect that this polymorphism has on weight and related metabolic disorders may only become apparent later in life. In a fish oil supplementation study, carriers of the ala12 allele presented with a greater decreases in plasma TAG after n-3 PUFA supplementation, when total fat intake was <37E% or the intake of SFA was below 10E% (Lindi et al., 2003). There was no evidence of a differential effect of n-3 PUFA supplementation by genotype on fasting insulin or glucose.
These studies suggest that the PPAR-γ pro12 ala polymorphism can modulate the association between dietary fat and cardiovascular risk factors. Thus, this and related SNPs are under intense investigation as a potential therapeutic target for obesity and insulin resistance. However, the association is not simple, with differential nutrient–gene–environment–gender interactions being multifaceted and extremely difficult to interpret. Therefore, there is a need for rigorous data analysis and the pooling of all available data from other sources to gain a full insight into to the functional role of a particular polymorphism. (Olefsky & Saltiel, 2000).
Future challenges in nutrigenetic research
The study of nutrigenetics is in its infancy. Many studies published in this area have only considered one SNP in a single gene, with little consideration being given to multiple nutrient–gene–environmental interactions (Sarkkinen et al., 1998; Minihane et al., 2000; Olefsky & Saltiel, 2000; Robitaille et al., 2003). Although this is scientifically valid, and invaluable for the elucidation of causative mechanisms in disease, multiple gene–nutrient–environment–gender interactions will be required for developing specific personalized nutritional advice. The collation of data in haplotype databases and biobanks is expensive and difficult to establish, but is a necessity if nutrigenetic research is to progress.
Standardized protocols in nutrigenetics are not yet established, the comparison of studies is challenging and conclusions are often difficult to draw. As discussed previously, studies are often retrospective in design and thus of insufficient power to detect nutrient–gene associations. Prospective genotyping increases the power to resolve these associations and should be used whenever possible. With any research, publication bias results in positive associations being reported more often than negative associations. This has applied to nutrigentic studies and created a false impression of the level of significance of many nutrient–gene interactions.
There are numerous ethical issues and unavoidable assumptions that need to be considered before personalized nutritional can become routine practice. First, it is important to consider whether the genetic tests and personalized food products would be affordable, cost-effective and socially acceptable. It is also of concern that only the well educated and affluent would benefit. The open accessibility of genetic information to third parties has major implications for the availability of health insurance and increased premiums.
Moreover, it is still unknown whether people will want to undertake genetic tests or even understand the concept of such technology. A survey was conducted by Cogent Research in 2003 on 1000 Americans (CogentResearch, 2003) in which 62% of respondents reported they had never heard of ‘nutrigenomics’. However, if specific products did arise from nutrigenomic research, those interviewed did express interest in an in-depth well-being assessment and also a strong interest in vitamins, fortified foods and natural foods. More research is required to determine whether individuals would want to undergo such tests, and for understanding the value to the individual of an increased awareness of personalized nutrition regimens. There is already a large gap between the existing dietary guidelines and what people actually eat (Henderson et al., 2003). Knowledge of being at higher than average risk of CVD may motivate people to actually make positive changes to their diets. However, genetic testing could undermine current healthy eating messages, by implying that only those with the ‘risky gene’ need to eat a healthy diet. These are important unanswered questions that must be addressed if personalized nutritional advice is ever to become part of mainstream disease prevention and treatment. It may be that the interactions between genotype–phenotype and the environment are just too complex to be properly understood from human dietary intervention studies.
There is resistance to the use and perceived effectiveness of personalized nutrition that is based on genomics, and whether this can offer a solution to diseases caused by a diet that is inappropriate for health (Cannon & Leitzmann, 2005). It has been suggested that it may be more beneficial to use current risk factors as a basis for population screening and the management of CVD (McCluskey et al., 2007). There has also been dialogue on the social, economic and environmental causes of CVD, shifting the emphasis away from dietary intake to food manufacturing as being more effective in disease management (Cannon & Leitzmann, 2005).
Progression of knowledge in the fields of nutrient–gene interactions promises a future revolution in preventative health care. However, although there is increasing evidence for interactions between diets/nutrients, genes and environmental factors, there are inconsistencies in the evidence that will limit the application of nutrigenetics in diet-related disease in the immediate future. In addition to the need for adequately powered intervention studies, greater attention should be given to ethical issues, such as the public’s acceptance of genetic testing and the economics of this relatively new science.
We thank Dr Bruce Griffin for his constructive comments on the manuscript.