Gene–Diet Interactions on Lipid Levels: Current Knowledge in the Era of Genome-Wide Association Studies
- First Online:
- Cite this article as:
- Walker, C.G. & Jebb, S.A. Curr Nutr Rep (2012) 1: 123. doi:10.1007/s13668-012-0017-z
Plasma lipids are important risk factors for cardiovascular disease. Both genetic factors and diet are known to regulate lipid levels, and there has been a longstanding interest in how genes may interact with diet to modulate changes in lipid levels. Genome-wide association studies have recently identified the genes most strongly associated with variation in lipids within a population. In this paper, the current knowledge on gene–diet interactions to regulate lipid levels is discussed in light of these studies. Future genome-wide studies are required that specifically identify genes that are important modulators of lipid levels in response to dietary change. Some methodologic challenges inherent in these studies are discussed.
KeywordsGenome-wide association studiesGWASGene–diet interactionsLipidsLipid metabolismVariationCholesterolTriglycerideDietary intervention
An adverse lipid profile of high triglycerides, total and low-density lipoprotein (LDL) cholesterol, and low high density (HDL) cholesterol has been associated with atherosclerosis and cardiovascular disease [1–3]. At a population level, diet is an important determinant of the lipid profile, but a great degree of interindividual variation has been consistently noted, particularly in the lipid responses to dietary interventions. This variation has been attributed to differences in genetic background, in addition to age, gender, and ethnicity. In the past, a number of likely candidate genes have been explored, but no genes have been identified that can consistently account for a large number of these interindividual differences. In recent years, genome-wide association studies (GWAS) have shed light on genetic mechanisms underlying the regulation of lipid levels; however, it is becoming apparent that further research is required, with studies more specifically designed to assess diet–gene interactions in a more comprehensive way.
Dietary Modulation of Lipids
Population-level effects of diet composition on lipid profile are well-established. Reductions in total fat, saturated fatty acids, and trans-fatty acids are all effective in reducing total cholesterol, LDL cholesterol, and triglycerides [4–6]. However, a reduction in total dietary fat intake achieved by replacement with carbohydrates has been associated with an unfavorable reduction in HDL cholesterol and increased triglycerides [5, 7]. Conversely, replacement of trans- or saturated fatty acids with monounsaturated fatty acids has been found to maintain HDL cholesterol . Increased dietary intake of n-3 polyunsaturated fatty acids has been associated with lower triglycerides . In addition to changes in dietary composition, a reduction in total energy leading to weight loss is associated with reductions in cholesterol and triglycerides, and an increase in HDL cholesterol associated with weight loss [10, 11].
However, population-level effects of diets on lipid levels conceal significant heterogeneity at an individual level . This is often attributed (in part) to genetic variation [23, 24]. Previous genetic studies, in particular studies of monogenic lipid disorders, have helped characterize specific lipid metabolic pathways that may be modulated by diet [2, 24, 25]. Further advances in genetic analysis and in particular gene–diet interactions will help further elucidate these pathways and identify novel pathways.
Genetic Modulation of Lipids
Heritability estimates for variation in lipid levels within a population range from 30 % to 60 % . In addition to the effect on the variation in lipids within a population, it is also assumed that genetic factors contribute to interindividual differences in response to dietary changes that cannot be explained by sex, age, and ethnicity [3, 23]. Over the past few decades, studies have been conducted to investigate candidate genes within specific lipid metabolic pathways (Fig. 1) in order to identify whether mutation or variation in these genes can explain the differences in lipid levels. Reviews of these candidate gene studies can be found elsewhere [2, 24]. In summary, gene variants have been identified in apolipoproteins, and genes involved in cholesterol synthesis and efflux, lipolysis, lipogenesis, and triglyceride and cholesterol transfer (Fig. 1). The genes that have most consistently been found to account for some of the variation in population lipid levels are APOE, CETP, and APOA5 .
However, there are fewer data relating to the genetic influence on lipid changes in response to diet, and no genes have consistently been shown to account for a large proportion of the interindividual variation in lipid responses to dietary change. The APOE polymorphism, which has often been described as a genetic modifier of dietary lipid responses , has shown inconsistent results between studies. In a systematic review of studies that involved responses to a modification of dietary saturated fat, only 11 of the 46 intervention studies showed evidence of modification in response by APOE genotype . In the same systematic review, variations in other genes previously found to be moderators of lipid responses to dietary change, including APOB, APO A-I, C-III, and A-IV gene cluster, LDL receptor and lipoprotein lipase, were also found to have inconsistent results. The majority of studies that found significant associations were small (n < 50) studies and therefore possibly a result of false-positive chance .
GWAS Identified the Genes That Are Most Important in Variation of Lipids Within a Population
Top 25 genes most strongly associated with lipid traits in GWAS
Lipid metabolic pathways
P for lead traitc
HDLC, TC, LDLC, TG
Cholesteryl ester/triglyceride exchange
P < 10−380
TG, TC, HDLC, LDLC
Major lipoprotein component of HDL, promotes cholesterol efflux
P < 10−240
Hepatic VLDL efflux
P < 10−170
LDLC, TC, HDLC
Major lipoprotein in chylomicrons and important in clearance of TG-rich lipoproteins
P < 10−147
Improved glycemic regulation at the expense of high hepatic TG production 
P < 10−133
Cell surface receptor for apo B, which regulates endocytosis of cholesterol-rich LDL
P < 10−117
Endothelial protein that promotes uptake of cholesterol-rich lipoproteins into peripheral tissues
P < 10−115
Major lipoprotein of LDL, docks to LDLR for cholesterol delivery to peripheral tissues
P < 10−114
HDLC, TC, TG
Hepatic protein that hydrolyzes triglyceride and catalyzes receptor-mediated lipoprotein uptake
P < 10−96
Transcription factor that binds carbohydrate response elements to increase transcription of genes involved in glycolysis, lipogenesis, and VLDL secretion
P < 10−58
TC, LDLC, HDLC
Scaffold protein with high hepatic expression. Affects lipogenic gene expression, hepatic lipogenesis, VLDL formation, and plasma lipids 
P < 10−55
Endothelial lipase that hydrolyzes HDL
P < 10−49
Selective sterol absorption in GIT and hepatic sterol excretion
P < 10−47
Rate-limiting enzyme in cholesterol synthesis pathway
P < 10−47
TG, TC, LDLC
Hepatic secretory factor that inhibits LPL and endothelial lipase
P < 10−43
TC, TG, LDLC
Scaffold protein (role in lipid metabolism unknown)
P < 10−38
Catalyzes cholesterol efflux from peripheral tissues and HDL formation
P < 10−33
Conversion from cholesterol to cholesteryl ester for formation of HDL
P < 10−33
Regulates LDLR activity
P < 10−28
TC, LDLC, TG
Receptor protein involved in engulfment of apoptotic cells (role in lipid metabolism unknown)
P < 10−28
HDLC, TC, LDLC
Protein of high hepatic expression that can reduce HDL levels [33•]
P < 10−25
TG, HDLC, TC, LDLC
Fatty acid biosynthesis
P < 10−24
Plasma protein (role in lipid metabolism unknown)
P < 10−24
Phospholipid transfer protein; plasma protein that transfers phospholipids from triglyceride-rich lipoproteins to HDL
P < 10−22
GalNAc-transferase involved in o-linked oligosaccharide biosynthesis, which can regulate HDL levels [33•]
P < 10−21
Loci identified in GWAS now account for about 25 % to 30 % of genetic variance in each trait (or 10 %–13 % of total variance in the trait) [33•]. There are likely to be more SNPs with small effect sizes that have not yet been identified with current GWAS methods [33•]. Rare SNPs of larger effect size will also not have been captured by GWAS and will contribute to the estimate of genetic variance . Variations in loci do not occur in isolation, and combinations of polymorphisms are likely to act together to modulate lipid levels. It has been demonstrated that SNPs identified in GWAS have a cumulative effect on lipid traits, such that a linear increase in the number of risk alleles for a given trait corresponded with an increasingly more adverse lipid profile [39•, 40, 41]. However, this simple additive approach does not explore the potentially multiplicative effects of two or more SNPs. A simple example is seen for the APOE polymorphism, which has been identified in GWAS to be associated with all lipid traits [29–31, 33•], but with an effect size of 0.18 mM per risk allele for LDL cholesterol [33•]. However, two SNPs define the common ApoE, and when combined, the difference between individuals of the lowest risk isoform compared with those of the highest risk isoform was greater than 1 mM for LDL cholesterol . Variance in a number of SNPs, in a number of genes, and of a number of different lipid metabolic pathways is likely to have complex and interacting effects. To fully explore these interactions in a comprehensive and systematic way will require novel data analytical methods and possibly increased computing power.
GWAS do not take into account environmental modifiers (eg, diet) in the identification of genes associated with a given trait. A number of different methods for analyzing the data are beginning to emerge to address this. Manning and colleagues  have proposed a joint meta-analytical method for simultaneously testing gene and gene–environment interactions in cross-sectional data in order to identify genetic loci that are impacted upon by a given environmental exposure on a trait of interest. These analyses could be conducted in studies in which data on dietary exposures are available in very large cohorts, along with GWAS data. However, this method is limited to testing a single dietary exposure at a time, and it remains to be seen how effective this approach will be in identifying gene–diet interactions that modify lipid levels. In a separate approach, Igl and colleagues [44•] performed a GWAS with a lifestyle-adjusted model that included certain dietary covariates (eg, game meat, non-game meat, fish and milk products) in their models to identify SNPs associated with lipid levels. This approach identified SNPs that were sensitive to the inclusion of environmental factors into the model (gene + diet) but did not assess how diet could interact with genes (gene × diet) to modify blood lipids.
Diet–Gene Interactions to Modify Lipids for Individual Genes Identified in GWAS
The effect of a number of genes that were identified in GWAS to modify lipid responses to diets has been studied previously from a candidate gene approach. These include CETP, APOA1, APOE, LDLR, LPL, APOB, LIPC, LIPG, ABCG5/8, and LCAT (Table 1, Fig. 1), and these findings have been comprehensively reviewed elsewhere [23, 24]. Although there is some evidence of gene–diet interactions, there is considerable heterogeneity in response, and it is noted that these gene–diet interaction studies were generally of short duration, with a small sample size, and from studies not designed for this purpose. Therefore, these findings need to be replicated in further studies, including well-designed, adequately powered randomized controlled trials. However, for some of the top genes identified in GWAS, in which the roles in lipid metabolism have been clearly established (Table 1, Fig. 1), there have been surprisingly few reports of gene–diet interactions. For example, despite the clearly defined role that HMGCR plays in the regulation of cholesterol synthesis, there has been a scarcity of reports on variation in this gene interacting with dietary factors to modulate lipid levels. In one cross-sectional analysis of interactions between dietary fat and a variant of the HMGCR gene, carriers of the variant appeared to have higher serum triglyceride levels to greater saturated fat intake but no effect on other lipids . There has also been a scarcity of reports on effects of dietary interactions with ANGPTL3, PLTP, PCSK9, and ABCA1 (Table 1).
Of the genes identified in GWAS that had not been implicated previously in the regulation of triglyceride and cholesterol levels (Table 1), some preliminary studies into possible gene–diet interactions have been conducted. In observational, cross-sectional studies, variants in the FADS genes (which encode desaturase enzymes in long-chain fatty acid biosynthesis) have been reported to interact with different levels of dietary n-3 and n-6 polyunsaturated fatty acids to alter serum cholesterol levels. In people who consumed a high proportion of n-3 polyunsaturated fatty acids (compared with the population median), the variant allele was associated with lower serum total and non-HDL cholesterol, translating to a protective effect [46, 47]. However, as these findings were cross-sectional, the impact of these gene variants on lipid concentrations over time or in response to changes in diet remains to be determined. Also, in a cross-sectional analysis in a small cohort (n = 379), there was no interaction observed between variants of the GCKR gene and the level of polyunsaturated, saturated, or monounsaturated fat in the diet on plasma triglycerides . In an intervention trial, variation in the GCKR gene was shown to modify triglyceride responses to an intensive lifestyle intervention of increased physical activity, weight loss, and a low-fat diet such that there was a greater reduction in carriers of the variant . However, it cannot be determined from this study whether diet, physical activity, or the effects of weight loss (or all factors) are important determinants. In acute postprandial studies, individuals who carry the GCKR risk variant were found to have an elevated postprandial triglyceride response to a high-fat meal (in combination with another variant in the apolipoprotein A-V [APOA5] gene) [50, 51], indicating a probable cumulative effect of high-fat diets to increase triglyceride levels in individuals with variants in these genes.
For many of the novel loci recently identified to be associated with lipids, the gene within the loci underlying the mechanistic effect may not have been identified, and the mechanism underlying the regulation of lipid levels is unknown (Table 1). These loci have not yet been studied in terms of gene–diet interactions to determine the modulation of effect on lipid levels.
There are some notable examples of genes that have been identified through candidate gene studies as important moderators of diet to modulate lipid levels, but that were not identified in GWAS. These include the transcription factors peroxisome proliferator-activated receptor (PPAR) alpha and gamma, which regulate the expression of several genes involved in lipid metabolism . Variants in the PPAR alpha gene have previously been shown to interact with polyunsaturated fatty acid intake to enhance improvements in total and LDL cholesterol  and triglycerides  among people with a higher proportional intake of n-3 and n-6 polyunsaturated fatty acids. A PPAR gamma SNP was also found to enhance the reduction in serum triglyceride in response to a dietary intervention to increase n-3 polyunsaturated fat intake . It is likely that these genes become important modifiers of lipid levels in response to particular dietary exposures, and were therefore not identified in GWAS that did not take into account diet. More examples of such genes are likely to be discovered when dietary factors are included.
The proportion of variance in lipid responses to dietary change that can be explained by variation in a single gene is not likely to be greater than 10 % . We have examined the potential cumulative effect of SNPs identified in GWAS to modify lipid responses to a dietary intervention of reduced dietary saturated fat. We showed that a cumulative genetic predisposition score composed from simple addition of these SNPs did not modify the response to changes in dietary saturated fat [39•]. We had insufficient power to explore the effects in individual SNPs, but a preliminary analysis revealed no obvious modifying effect of any of the individual SNPs [39•]. A similar approach was used to examine the combined effects of nine common genes on the change in lipid responses to statin therapy [55•]. In this analysis, the combined genetic score appeared to modify the response to statin therapy such that the response was greater—the higher the genetic predisposition to low HDL cholesterol and high LDL cholesterol—in women only [55•]. Although this study was not diet related, it illustrates that the modifying effect of the interaction between genes and environmental modifiers will depend on the mechanism by which the environmental factor acts to alter plasma lipids. Because different dietary factors act to regulate lipid levels via different metabolic pathways, the effect of SNPs on the change in lipids is likely to be complex, with different combinations of SNPs responding differently to various dietary exposures. New analysis and modeling techniques are required to comprehensively investigate interactions of multiple SNPs and multiple dietary components.
Studying the effects of genes that modify lipid responses to diets in a controlled intervention setting, by its nature, means the cohort is limited to a small size, which can limit power to detect gene–diet interactions. The use of large, longitudinal, population-based studies may offer an alternative. Following GWAS that identified the SNPs most strongly associated with body mass index (BMI), a study was conducted that used the cumulative score of these SNPs to test gene–lifestyle interactions . This study effectively demonstrated that physical activity (assessed by a self-administered questionnaire) attenuated the genetic predisposition to BMI in a cross-sectional, population-based sample of 20,430 individuals, and furthermore limited the change in BMI over time . A similar approach could be used with lipid-associated genes identified in GWAS in which measures of dietary intake are also available. Dietary exposures such as total fat, saturated fat, and high carbohydrate, which are known to impact upon lipid levels, could be used to test the effect on the cross-sectional association between genetic predisposition and an unfavorable lipid profile, and the change in lipids over time.
GWAS on Interventions
The majority of GWAS to date have been conducted in cross-sectional data; it is therefore unsurprising that the loci identified as most important for variation of population lipid levels may have less of a moderating effect on changes in plasma lipids in response to diet [39•]. Some small GWAS have been performed on dynamic lipid responses to lipid-lowering drugs [55•, 57, 58]. A combined analysis of such studies has shown that there was no overlap between the SNPs associated with variation in the lipid levels of the cohort at baseline or after treatment compared with those that associated with the change in lipid levels . This illustrates that the genes that are most important in modifying changes in lipids in response to lipid-lowering drugs are likely to be different from those that underlie differences in population lipid levels. Likewise, dietary-induced changes are likely to act via different mechanisms to pharmaceutical agents. Due to the nature of highly controlled dietary intervention trials, the size of the cohort is small in terms of genetic analyses. To combat this, it is likely that data from a number of similar studies would need to be combined to achieve the required power to assess diet–SNP interactions. Challenges to be encountered by this approach will be the heterogeneity in trial design with regard to changes in diet, participant characteristics, duration, and rigor.
GWAS have been central to recent advances in the understanding of genetic mechanisms underlying variation in blood lipid levels. However, as of yet, these studies have done little to enhance our understanding of gene–diet interactions and their modifying effect on blood lipids. Data analytical tools are being developed to incorporate dietary factors into GWAS studies in order to identify genes that interact with dietary factors to modify lipids. It is emerging that SNPs identified in cross-sectional GWAS may not be the most important modifiers of lipid changes in response to changes in diet or other environmental factors, and GWAS are now beginning to be conducted to identify the SNPs related to change in measured traits. These methods applied to large, controlled dietary intervention trials will help identify SNPs that interact with particular dietary exposures, although it is likely that data from many trials will need to be combined to obtain adequate power. The SNPs identified may differ, depending on the nature of the dietary exposure (eg, energy reduction, saturated fat reduction, increase in carbohydrate), and this in turn will increase our understanding of how these different diets alter blood lipids and the pathways involved. It is clear that advances will be required in the way data are analyzed in order to handle the vast amount of data that will take into account the complex interactions between different dietary exposures and different genes.
The authors wish to thank Dr. L. Hodson for discussions on lipid metabolism. Both authors are supported by the UK Medical Research Council (grant code U105960389).
Dr. Jebb has served on advisory boards for Tanita Ltd., PepsiCo, Nestle, Coca-Cola, Californian Almond, and Heinz; has been employed by the Medical Research Council; has received grant support from the NPRI, World Cancer Research Fund International, MRC Public Health Sciences Research Network, Tanita Ltd., Weight Watchers International, the Food Standards Agency, and the European Union; and has received payment for lectures given (including service on speakers’ bureaus) and preparation of nutrition-related articles for Rosemary Conley Enterprises.
Dr. Walker reported no potential conflicts of interest relevant to this article.