Epigenetics and obesity: the devil is in the details

Obesity is highly prevalent in most industrialized nations where labor-saving devices and calorically dense diets are common [1]. The impressive results from randomized clinical trials of intensive lifestyle modification on short- to medium-term weight loss [2, 3] give us good reason to believe that traditional hunter-gatherer or subsistence farming lifestyles might be a panacea for the obesity epidemic. However, programs of lifestyle modification are notoriously difficult to implement and maintain on a large scale for prolonged durations, and the public health message of "eat less and exercise more" appears to have fallen on deaf ears. Thus, despite the apparently simple explanation and remedy for obesity, this knowledge is not enough. So we are saddled with a challenge, which is to unravel the mechanisms by which obesity emerges and to understand how its presence causes disease and death, with the hope that somewhere within the details hides the solution to the problem.

Although obesity is widespread, certain ethnic groups appear much more susceptible than others to the condition [4]. Striking differences in the rates of obesity are often seen between genetically distinct subpopulations, such as American Indians and people of northern European ancestry living within comparable environments, indicating that obesity may be the consequence of gene × environment interactions. More concrete evidence of such interactions has emerged from epidemiological cohort studies, most notably for the interaction between a polymorphism at the FTO (fat mass and obesity gene) locus and physical activity levels [5]. Nevertheless, even where rare examples of statistically reliable gene × environment interactions have emerged from epidemiological studies, causal inference is often difficult and little is revealed about the underlying mechanisms driving the interaction.

So what mechanisms might underlie the gene × environment interactions observed in epidemiological studies or clinical trials? Clearly, the manner in which an interaction is defined and quantified will affect our ability to understand the cellular mechanisms which underlie it. However, assuming the interaction is a consequence of environmental effects (for example, smoking, physical activity, or dietary whole grains intake) on a disease trait that are dependent on the genotype of an individual, there are probably at least two closely related mechanisms that might give rise to the interaction (Figure 1).

The first involves differential transcription rates across genotypes in response to environmental stimuli. During aerobic exercise, as an example, subsets of genes, particularly those involved in oxidative energy metabolism, are upregulated [6]. If the rates of transcription or translation of these genes were to differ by genotype, one might, at a population level, observe a gene × lifestyle interaction on a clinical phenotype such as blood glucose concentrations. The second mechanism by which genes and environmental factors might interact involves epigenetic factors, such as DNA methylation and histone modifications.

The word epigenetics describes the phenomena of inherited changes in gene function that occur independently of changes in the nucleotide sequence [7]. Initially, it was believed that epigenetic modifications were unidirectional, but recent studies have demonstrated that the epigenome is in fact highly dynamic, changing in response to nutrient availability, physical exercise and aging, among other exposures [8‐22]. While nearly all cells in the body have the same nuclear genome, different cell types have their own epigenomes, a characteristic essential for the development of cell-specific phenotypes. In differentiated mammalian cells, DNA methylation occurs primarily to cytosines in CpG-dinucleotides [23]. DNA methylation of gene promoters has also been implicated in transcriptional silencing, mainly through repressed transcription factor binding to gene promoters or by recruiting methyl-CG-binding proteins, which in turn recruit histone deacetyltransferases (HDACs) and corepressors. By virtue of the cell's capacity for histone modification, it can control its chromatin structure and either activate or suppress the transcription of its genes. Early-life nutrition represents an intriguing example of how environmentally augmented epigenetic events might affect an individual's response to metabolic load and disease susceptibility in adulthood, with numerous studies lending weight to this hypothesis [14‐16, 24, 25]. Nevertheless, our understanding of how epigenetic events early in life influence the development of obesity and its comorbidities remains fairly rudimentary.

This month in BMC Medicine, Wang et al. [26] report a study in which they examined DNA methylation in isolated blood leukocytes and its relationship with obesity-induced immune dysfunction in seven lean and seven obese African American male adolescents. Specifically, the authors examined DNA methylation in peripheral blood leukocytes at approximately 27,000 CpG sites spread across more than 14,000 genes and compared methylation patterns between the two groups. Epigenetic studies are often constrained in their size by the high costs of the methylation chips (although this obstacle is receding as the technology becomes less expensive). This, in combination with the multitude of hypothesis tests that are performed in multiplex experiments and the corresponding procedures to correct for type 1 errors, render almost all existing epigenetic studies underpowered to detect statistically robust effects. Such is the case with the study conducted by Wang et al.; indeed, none of the findings from the first phase of their experiment remained statistically significant after correction for multiple testing. Despite this, the authors carried forward their most promising findings (defined as either genes yielding association P values ≤2 × 10^-4 or those with differences in DNA methylation ≥27.1%) for replication in a cohort composed of 45 obese and 46 lean individuals. In these replication analyses, two of six genes were differentially methylated in obese compared with lean individuals; the levels of DNA methylation for UBASH3A (ubiquitin-associated and SH3 domain-containing A) and TRIM3 (tripartite motif-containing 3) were higher and lower, respectively, in obese compared with lean individuals, findings that were directionally consistent with those from the first phase of the study.

Interestingly, recent genome-wide scans have implicated DNA variants proximal to UBASH3A, which encodes a T-cell signaling- and activation-regulating protein [27], in the development of type 1 diabetes [28]. Thus, the findings of Wang et al., when placed in context with existing genetic data, suggest that UBASH3A may play a role in obesity-induced immune dysfunction. Intriguingly, TRIM3, which belongs to the superfamily of TRIM proteins, is also involved in immune response [29], which may explain why differential methylation patterns of these genes were visible in blood leukocytes.

The study by Wang et al. is small and probably underpowered to detect all but the largest differences in DNA methylation. It may be, therefore, that with a larger sample size, statistically significant differences in DNA methylation patterns for many other genes that are smaller in magnitude than for UBASH3A and TRIM1 would be observable. The detection of other CpG sites and genes might also be facilitated by the application of methods that afford greater genomic coverage, such as deep sequencing combined with either bisulfite-treated DNA or immunoprecipitation of methylated DNA.

While peripheral blood leukocytes are attractive for epigenetic studies, not least because they can be easily obtained, there are many other cell types and tissues involved in the pathogenesis of obesity and its comorbidities, such as adipose tissue, skeletal muscle and hypothalamus, the examination of which could be well worthwhile in the context of epigenetics. Along these lines, Bouchard et al. [17] recently provided some of the first evidence that DNA methylation in adipose tissue differs in people who respond well and those who respond poorly to caloric restriction for weight loss. Elsewhere, a high-fat diet lasting 5 days was shown to affect DNA methylation of a major transcriptional coactivator (peroxisome proliferator-activated receptor-γ coactivator 1-α) of genes involved in oxidative energy metabolism [13]. Furthermore, in studies examining early life overnutrition and obesity, DNA methylation of the hyperthalamically expressed pro-opiomelanocortin promoter, which plays an important role in hunger and satiation [30], was observed. While these studies imply that epigenetic modifications cause obesity, metabolic disease and its complications, it is virtually impossible to establish whether changes in methylation precede the development of obesity or vice versa. Indeed, these relationships may not be causal at all, but consequences of confounding by factors correlated with obesity and DNA methylation, such as physical inactivity, nutrition or smoking.

In summary, Wang et al.'s study provides tentative evidence that DNA methylation at two loci, UBASH3A and TRIM3, may be implicated in the pathogenesis of obesity. Replication of these findings in independent settings will be necessary to ensure that these findings are true positives, and to fairly conclude that the relationships are causal will require appropriately designed experimental studies. Because of these and other hurdles facing the field of epigenetics, identifying a meaningful clinical application for epigenetics in the prevention or treatment of obesity is likely to remain more vision than reality for some time to come.