DNA methylation markers in obesity, metabolic syndrome, and weight loss

ABSTRACT

The fact that not all individuals exposed to the same environmental risk factors develop obesity supports the hypothesis of the existence of underlying genetic and epigenetic elements. There is suggestive evidence that environmental stimuli, such as dietary pattern, particularly during pregnancy and early life, but also in adult life, can induce changes in DNA methylation predisposing to obesity and related comorbidities. In this context, the DNA methylation marks of each individual have emerged not only as a promising tool for the prediction, screening, diagnosis, and prognosis of obesity and metabolic syndrome features, but also for the improvement of weight loss therapies in the context of precision nutrition. The main objectives in this field are to understand the mechanisms involved in transgenerational epigenetic inheritance, and featuring the nutritional and lifestyle factors implicated in the epigenetic modifications. Likewise, DNA methylation modulation caused by diet and environment may be a target for newer therapeutic strategies concerning the prevention and treatment of metabolic diseases.

Introduction

Obesity has been defined as an abnormal or excessive fat accumulation that contributes to the development of many associated comorbidities such as type 2 diabetes (T2D), cardiovascular diseases (CVDs), dyslipidaemia, hypertension, liver steatosis, or metabolic syndrome (MetS), among others [1]. Obesity is attributed to the maintenance of a positive energy imbalance between caloric input and output [2]. Several factors could affect the energy equation, such as lifestyle behaviours (dietary habits, exercise, and sleep patterns) [3,4], social factors (educational level and economic status) [5], endocrine disorders (hypothyroidism) [6], or prescription of certain medication (corticosteroids) [7]. The current notion on the pathogenesis of obesity is that not all individuals under the same environmental factors develop the disease, so it has been suggested that genetics play a role [8]. However, genetic risk loci identified by genome-wide association studies only explain part of the genetic variance in disease risk, suggesting that more factors must be taken into account to understand the multifactorial pathology of obesity. Epigenetics has therefore, been proposed as a factor involved in obesity development because it is influenced by both genetic and environmental factors [9]. Most recently, epigenetics has offered new explanations about the mechanisms, where environmental factors modify the expression of genes related to obesity and its comorbidities [10].

As obesity has been attributed to a positive energy balance [2], reduction of energy intake, changes in macronutrient distribution, increase of physical activity, behavioural approaches, and pharmacological or surgical treatments are commonly used to produce a negative energy balance [3]. Although many strategies have been investigated for inducing weight lowering, individual response varies widely. A better understanding of the contribution of diet and nutrients to epigenetics is necessary, as weight loss is crucial for reducing the negative medical consequences of obesity. This review describes the current advances in the use of DNA methylation in the assessment of metabolic risk and the personalization of the clinical management of obesity.

Epigenetics

Epigenetics is defined as the processes that alter gene activity without nucleotide sequence modification, but including chromatin structure changes as a direct consequence [11].

The most relevant epigenetic mechanisms involved in gene activity regulation are DNA methylation, histone modifications, and non-coding RNAs. The most extensively studied epigenetic mark in the mammalian genome in relation to gene expression regulation has been found to be DNA methylation [12], which is an epigenetic mechanism involving the covalent addition of a methyl group (-CH3) onto the five position of cytosine, resulting in 5-methylcytosine. Although in mammals methylation is generally restricted to cytosine linked by a phosphate to guanine (CpG site), a small percentage of methylation may occur in non-CpG sites (CHG and CHH, where H = A, C, or T) [13,14]. Similar to methylation, hydroxymethylation is an epigenetic mechanism that replaces, at the C5-position in cytosine, the hydrogen atom by a hydroxymethyl group.

Mammalian genomes present approximately 28 million CpG sites, among which 60–80% are generally methylated. Genomic regions with a high CpG density are known as CpG islands (CGIs). The majority of the CGIs are located in gene promoter regions, but they also appear in gene bodies, often acting as alternative promoters [15]. Contrary to CpG sites, most of the CGIs in promoter regions are usually unmethylated to maintain transcription of the active gene. A differentially methylated region (DMR) is an area of the genome where multiple adjacent CpGs show different methylation status between phenotypes. DMRs are regarded as possible functional regions involved in gene transcriptional regulation [16]. In this context, differential regional DNA methylation is more biologically interpretable and statistically powerful than CpGs measured individually [17]. These characteristics allow DMR to be powerful biomarkers of biological and pathogenic processes, or pharmacological responses to a therapeutic intervention.

Obesity, related comorbidities, and DNA methylation

Obesity is considered the pandemic of the 21st century (WHO, 2017). The prevalence of obesity closely tripled between 1975 and 2016. In 2016, more than 1.9 billion adults were overweight and 650 million were obese (WHO, 2017). Also among children and adolescents, the obesity prevalence has hugely increased in the last decades. A recent meta-analysis concerning 2,416 population-based studies in 128.9 million children and adolescents found an increase in the prevalence of obesity worldwide from 0.7 % in 1975 to 5.6% in 2016 in girls, and from 0.9% in 1975 to 7.8% in 2016 in boys, suggesting that in 2022 child and adolescent obesity would surpass moderate and severe underweight [18].

Despite extensive efforts, obesity prevalence continues to increase in every country. At the same time, a number of studies have tried to deepen into the molecular origin of obesity and related disorders, focusing on epigenetic signature, during the different stages of life [19–26]. In this context, research on novel epigenetic biomarkers is required to understand the pathophysiology of obesity and its clinical manifestations. These investigations described more than a hundred of genes distributed over all chromosomes (except sex chromosomes) (Figure 1). The lack of genes in sexual chromosomes, which is common with studies focused on other pathological conditions, is due to the methodological processing of the microarray data (which traditionally excludes sex chromosomes from the analysis), although the few studies including sex chromosomes did not find relevant genes with DNA methylation changes in obesity-related conditions.

Figure 1.

Karyogram representing genes loci whose DNA methylation levels have been associated with obesity (blue), metabolic syndrome features (green), and weight loss (red). The genes represented in Figure 1 have been described in Tables 1, 2, 3, and 4, where additional information about each gene is included.

Transgenerational epigenetic inheritance in obesity

DNA methylation modifications may be dynamic and altered as a consequence of environmental stimuli or may be stable and passed on to next generations [27]. Several studies, especially in animal models, reported that exposure to environmental factors in one generation caused phenotypical effects in unexposed future generations and considered DNA methylation mechanism in origin. This phenomenon is called transgenerational epigenetics inheritance [28]. This transfer is explained by an incomplete erasure of the epigenetic modifications during gametogenesis and early embryogenesis, some of which are transferred to the offspring [28]. A few cases of this phenomenon have been identified in mammals. For example, an intracisternal A-particle retrotransposon inserted in the agouti gene results in ectopic expression of agouti protein, causing yellow fur, obesity, diabetes, and a raised tendency to tumours in the next generations of mice [29]. Due to the public health implications of obesity and metabolic diseases, a number of investigations have studied the epigenetic inheritance of nutritional metabolic risk in mammals, especially in rodents. The adverse perinatal environment can affect the phenotype of the offspring and subsequent generations. For example, rats suffering a continuous 50% caloric restriction during 50 generations, exhibited higher obesity-related metabolic disturbances associated with modified epigenetic marks in insulin-2 (Ins2) gene promoter region [30]. Interestingly, these alterations in metabolism and epigenetic signature were not reversed after two generations of normal chow (restoration of normal nutrition) [30]. On the other hand, the exposure to chemical agents is also capable of promoting epigenetic transgenerational inheritance in rodents. For instance, exposure to plastic-derived endocrine disruptors and hydrocarbons in gestating F0 generation females derived in an increase of total abnormalities, such as testis disease, obesity, and ovarian diseases, in F3 generation male and female rodents [31,32]. These abnormalities in subsequent new-borns were associated with altered global DNA methylome in sperm of male rats [31,32].

In humans, although no studies have clearly demonstrated the transgenerational epigenetic inheritance of metabolic abnormalities, several investigations have observed the relationship between nutritional imbalances and adverse health outcomes. One of the most studied famines happened in the populated western provinces of the Netherlands during the winter of 1944–1945 (The Dutch Hongerwinter). The distributed food rations dropped to below 1000 Kcal and were reduced to extremely 500–600 Kcal [33]. Infants who were conceived during this period presented higher rates of obesity [34], hypercholesterolemia [35], and/or cardiovascular disease [36] in adulthood. Interestingly, these health problems developed in the first generation were also found in the second generation [37]. The DNA isolated from these subjects showed below-average methylation of insulin growth factor 2 (IGF2) and higher expression of the hormone, even six decades later [38]. Further studies described an increase in the methylation of 6 other genes in the Dutch Hongerwinter subjects, including IL-10, LEP, ABCA1, GNASAS1, MEG3, and INSIGF2, some of which are associated with cholesterol transport, ageing, or schizophrenia [39]. Other famine period occurred in a rural area in the Chittagong division of Bangladesh in 1974–1975 as a consequence of severe monsoon. Young adults exposed to famine during gestation presented low weight and hyperglycaemia, and significant differences in DNA methylation at seven metastable epialleles (genomic regions that exhibit inter-individual epigenetic variation), such as Vault RNA 2–1 (VTRNA2-1), paired box 8 (PAX8), PR/SET domain 9 (PRDM9), near ZFP57 zinc finger protein (ZFP57), near BolA family member (BOLA), and exonuclease 3ʹ-5ʹ domain containing 3 (EXD3) [40]. In rural Gambia the nutrient availability fluctuate with wet and dry seasons creating annual hungry periods during wet season (July–October). The studies in children gestated during famine periods demonstrated persistent DNA methylation changes in metastable epialleles [41] that were associated with different metabolic alterations in the offspring. Although nutrition during pregnancy affects the phenotype in the offspring and can affect the third generation, transgenerational epigenetic inheritance during more than three generations is not clear in humans. In any case, these studies could be included in the context of the DOHaD (developmental origins of health and disease) paradigm, which suggests that environmental conditions during embryonic development and early life modify disease susceptibility during adult life [42].

In conclusion, further epidemiological studies controlling the influence of prenatal and postnatal nutrition and environment, and well-designed experiments in humans, are needed to fully understand transgenerational epigenetic inheritance in humans in relation to obesity and other metabolic disturbances.

Factors affecting DNA methylation in the context of obesity

A number of studies have found evidence that epigenetic mechanisms are involved in the onset and development of obesity and its metabolic complications. Many of these epigenetic mechanisms are modulated by the environment, and the responsible factors can be divided into two major categories: nutritional and physiological.

Nutritional factors

In the last years, different dietary patterns, nutrients and food components have been related to epigenetic processes that may contribute to an increase in the susceptibility of obesity development and other metabolic disorders. The sensitivity of the epigenetic system to nutritional factors occurs mainly during periods of developmental transition because it is the time when epigenetic marks are critically modified [43]. Thus, inappropriate feeding practices and nutrient deficiencies during pregnancy and lactation might have long-term consequences in offspring’s health [44]. The available studies explained above (i.e., Dutch Hunger Winter and Gambian children) provide a bright example of how early and mid-gestation can lead to metabolic dysregulation later on life, having been associated with cardiovascular diseases, obesity or hyperglycaemia [37]. Furthermore, as one-carbon metabolism depends on the dietary methyl donors, DNA methylation can be influenced by folate, methionine, betaine, or choline availability during this stage. Thus, methyl donor levels in the maternal diet have been shown to modify the establishment of the DNA methylation profile in the foetus and have adverse metabolic effects during infancy and childhood [39,40]. For example, maternal deficiency of folate has been associated with higher inflammatory mediators (interleukins, TNF-α, and monocyte chemoattractant protein-1), but also with an increased risk of cardiovascular diseases, obesity and insulin resistance in the offspring, in both humans and rodents [45–47]. Maternal choline and betaine deficiencies also alter the methylation status of genes that participate in growth (IGRF2, Crh, and Nr3c1) or vascular function (VEFGC and Angpt2), resulting in developmental abnormalities, blood vessel malformation, or increased predisposition for developing steatohepatitis in the offspring [48–51]. These investigations are focused on epigenetic changes related to the development of metabolic abnormalities in offspring whose mothers have suffered nutritional stress. However, the paternal diet, lifestyle, and obesity and diabetes may also induce alterations in DNA methylation signatures in the offspring that can lead to different metabolic consequences [52]. In humans, the Överkalix study showed that grandparental access to food can modify diabetes risk and all-cause mortality in grandchildren [53]. The Uppsala Multigeneration Study found an elevated mortality (mostly from cancer) in grandsons whose paternal grandfathers had good access to food just before puberty [54]. These studies propose that maternal and paternal malnutrition can result in an obesity phenotype in the offspring and suggest that nutritional factors affect DNA methylation in humans. The malnutrition caused by overnutrition (when the intake of nutrients is oversupplied) is also able to induce epigenetic alterations in the promoters of key metabolic genes that could affect gene expression and predispose to metabolic alterations [52]. Studies about the effects of high-fat diets have associated diets rich in saturated and trans-fatty acids with damages in intestinal permeability, obesity, and inflammation as well as cardiovascular abnormalities [55,56]. For instance, palmitic acid induces global hypermethylation in human pancreatic islets, impairing insulin secretion and increasing the risk of type 2 diabetes [57]. Supplementation with saturated fatty acids and elaidic acid (trans fatty acid) induces TNF hypomethylation and PPARG1 hypermethylation, increasing and decreasing the expression of these genes and leading to an inflammatory environment in adipose tissue [58].

On the other hand, different in vitro and animal studies have reported that specific phytochemicals (i.e., catechins, curcumin, quercetin, genistein, resveratrol, and sulforaphane) are able to reverse adverse epigenetic markings by acting on DNA methylation, histone modifications, and microRNA expression (revised in [23]). It is possible that some of the beneficial metabolic effects of these compounds could be mediated by epigenetic mechanisms, due to their ability to inhibit DNMT activity by increasing S-adenosylhomocysteine levels or inhibiting the catalytic site of DNMTs [59]. For example, in humans, inverse associations have been found between flavonoid intake and body weight [60], suggesting that polyphenol-targeted epigenetics might be an attractive approach for obesity prevention. Nevertheless, few intervention studies in humans have found a positive effect of polyphenol supplementation on body weight. In any case, studies in humans are still ongoing and more information is needed about amounts, combinations and influence of gut microbiota on the epigenetic effects of phytochemicals in the context of obesity.

Obesity-related physiological factors

Several obesity-related physiological factors have also been associated with the dysregulation of epigenetic processes. Most of them have been studied in vitro but it is extremely difficult to study them isolatedly in humans. In obese patients, it is usual to find more than one of these factors, and their effects on the epigenome are also masked by the commented nutritional factors.

Low-grade chronic inflammation is considered an essential link between obesity and its comorbidities. Moreover, in some studies inflammation has been reported as a predictor of weight gain [61], whereas a lower expression of genes related to inflammation and immune response has been associated with better response to weight loss treatments [62]. Many studies have demonstrated that inflammatory mediators are able to induce changes in the epigenetic landscape. In addition, DNA methylation also has a functional role in the monocyte-to-macrophage differentiation and macrophage polarization [63]. Similarly, dynamic DNA methylation and demethylation activity is necessary to develop and differentiate T linage [64]. However, the effects and amounts of the different molecules, the interactions among them and the differential effects on the different cell types are still unknown.

Hypertrophy and hyperplasia of adipocytes in the obese adipose tissue decrease oxygen availability and induce hypoxic stress, which is considered to be an important mechanism involved in obesity-related chronic inflammation, IR and mitochondrial dysfunction [65]. Hypoxia-inducible factors (i.e., HIF3A) are both, regulated by epigenetic mechanisms, and related to adipose tissue dysfunction [65], which makes hypoxia a good epigenetic target to treat obesity-related complications. Hypoxic conditions favour the increase of reactive oxygen species and oxidative stress, both described as important inflammatory mediators [66]. Oxidative stress is the result from an imbalance between tissue oxidants and antioxidants, and is usually exacerbated in obese individuals. Not only oxidative stress, but also weight loss and dietary antioxidants, which may help to decrease this phenomenon, have been reported to induce changes in the epigenetic landscape [67].

Obesity is almost always associated with IR, where poor glycaemic control has been linked to epigenetic alterations that could be involved in the development of diabetes-associated comorbidities [68]. According to this, obese subjects would be more susceptible to episodes of transient hyperglycaemia that could impair cell structure and function by means of DNA and histone epigenetic alterations. In this context, a recent study has analysed the epigenetic signatures associated with IR, measured by the Homeostatic Model Assessment of IR (HOMA-IR), and has revealed that 478 CpGs were differentially methylated between individuals with HOMA-IR ≤ 3 and > 3, especially in genes that are mainly involved in glucose- and insulin-related pathways [69].

Psychological stress has become a typical trait of the modern way of life. It involves chronic activation of the neuroendocrine systems and has been associated with increased prevalence of obesity [70]. In this context, cortisol-based and perceived psychological measures of stress may be important predictors of obesity-related risks [71]. In relation to this, stress, particularly at early stages of life, has been associated with hormonal alterations and epigenetic modifications that can persist into adulthood and be a causative factor of obesity. For example, it has been demonstrated that DNA methylation mediates the impact of exposure to prenatal maternal stress on body mass index (BMI) and central adiposity in children [72].

Among the obesity-related physiological factors, disruption of the circadian rhythm may play a key role in obesity development. In mammals, the synchronization between circadian rhythms and environmental stimuli, including daily rhythms of natural light, external temperature and food intake, is driven by the circadian system, a hierarchical multilevel organization [73]. Sleep deprivation, shift work, bright light exposure at night, and unusual eating time have been related to obesity and diabetes [74]. Many metabolic functions, such as glucose homeostasis, lipid metabolism and fasting/feeding cycles, are regulated by the internal clock system [73]. In this context, several studies have described that different nutritional and environmental factors, including obesity, are able to affect the DNA methylation pattern of the clock genes that regulate circadian rhythm in the hypothalamus and peripheral tissues [75]. On the other hand, it has been also described that insufficient sleep (short sleep duration or insomnia) is associated with loss of DNA methylation, which could be related to impairments in processes related to neuroplasticity, neurodegeneration, and cardio-metabolic condition [76].

Moreover, there is growing evidence that a number of endocrine-disrupting chemicals (such as phthalates, p,p’-dichlorodiphenyldichloroethylene and bisphenol A), heavy metals (such as lead or cadmium), toxins, and drugs, particularly in early-life, may contribute to obesity development and metabolic impairment, at last in part, by altering epigenetic profiles and gene expression regulation [77,78]. However, more research in cells and humans is needed to understand the intimate mechanisms.

Human gastrointestinal tract is densely populated by bacteria, fungi, and viruses among others, collectively called gut microbiota [79]. These organisms play important physiological roles in digestion, vitamin synthesis, and metabolism [80]. The microbiome is a well-known epigenetic regulator, in part because of the production of bioactive molecules, such as short-chain fatty acids (propionate, butyrate, and acetate), and because of its capacity to modulate the inflammatory response [81]. Gut microbiota composition has been associated with a number of diseases including obesity [82] and is the perfect link between host genetics, diet, and epigenetics. In recent years, many studies are being conducted to analyse these interactions, with special emphasis on topics such as the role of maternal and neonatal nutrition on microbiota composition, the early establishment of the foetal and neonatal microbiome, the ability to reshape the epigenome of the inflammatory cells, the identification of the microbial metabolites that affect the mammalian epigenetic machinery, and the potential epigenetic effects on genes regulating important metabolic functions such as appetite, beta cell function and the browning of adipose tissue [83,84].

In addition, the decline in physical activity is one of the alleged perpetrators for the modern epidemic of obesity. Physical exercise has been demonstrated to induce genome-wide changes in DNA methylation in humans, both in adipose tissue [85] and skeletal muscle [86], potentially affecting cell metabolism and inducing health-enhancing phenotypic adaptations. However, there is not enough information about the amounts and types of physical activity that can be more beneficial for obesity prevention and treatment in relation to their effects on the epigenome.

Finally, other factors that could potentially affect the epigenome in relation to obesity risk, because they can exert an influence on energy intake and energy expenditure, are hormonal and neuropeptide imbalances, anxiety and depression, infections, ambient temperature, artificial sweeteners, air and noise pollution. However, all of them need to be demonstrated in humans.

DNA methylation markers in obesity

Numerous cross-sectional studies have reported a significant association between obesity or adiposity status, and DNA methylation (Table 1). For example, the offspring of obese mothers presented several CpG sites differentially methylated in cord blood in comparison with offspring from normal weight mothers [87]. Other regions studied, were mainly hypomethylated in obese children and located in the gene body region, and revealed a unique cluster of obese individuals that was differentiated from the normal weight children [88]. In addition, some of these genes are implicated in lipid and glucose metabolism, differential body size and body composition in children [89]. Another analysis in adult population found methylation changes in carnitine palmitoyltransferase 1A (CPT1A) gene, associated with obesity and T2D [90]. CPT1A is also implicated in the control of fasting triglycerides (TG) and very low-density lipoprotein (VLDL) levels [91]. These studies suggest an important role for this gene in obesity and metabolic complications.

Table 1.

Cross-sectional studies analysing the association between DNA methylation and obesity.

Study population	Sample type	Target gene	Major finding	Age/Sex adjusted	Reference
Children (n = 2,861)	Cord blood	Genome-wide	28 CpG differentially methylated between offspring of obese and normal weigh mothers. 0.01–1.84% change in DNA methylation per 1 kg/m2 of BMI Maternal weight modifies offspring epigenome	x/✓	[87]
Obese (n = 6) and non-obese (n = 6) children	PBC	21orf56, ZNF154, SDK1, KIAA0146, SKIV2L, GPR125, SORBS2,C14orf70, POLR3E, CTBP2,DLGAP2,CAPS2,GIMAP1, RNF213,MND1,SRM,TGM6,WDR27	Different DNA methylation profile between obese/non-obese children Markers for early diagnosis of obesity	-/-	[88]
Children (n = 374)	PBC	SNED1(IRE-BP1),KLHL6, WDR51A(POC1A),CYTH4-ELFN2, CFLAR, PRDM14, SOS1, ZNF643(ZFP69B),ST6GAL1, C3orf70, CILP2, MLLT4	Association between DNA methylation and BMI, fat mass, lipid, and glucose metabolism 1% change in DNA methylatin per 0.1–0.4% of variables	✓/✓	[89]
Obese (n = 24) and non-obese (n = 23)	PBC	CERCAM, DPYD, IL12A, ZNF35, ZNF362, TSC22D2, CBX6, FOXF1, PSMD7, H1FX, PRRC2C, MSI1, COL4A1, NBPF3, USP5, PLOD2, TLE3, RPS24, DVL3, POLD3	20 genes differentially methylated between obese and non obese adults 2.8–30.8% differences in DNA methylation between groups Biomarkers for the understanding of obesity	✓/✓	[147]
Adults (n = 547)	Blood	15 DMP and CPT1A	Association between DNA methylation and BMI, obesity, and abdominal adiposity	✓/✓	[148]
GOLDN (n = 991), ARIC (n = 2,106), and FHS (n = 442) cohorts	CD4+ T-cells	CPT1A, CD38, PHGDH	Association between DNA methylation and BMI and WC	✓/✓	[90]
Popgen biobank (n = 49)	PBC	MCHR1	DNA methylation associated with BMI DNA methylation contributes to the age-related specific effects	x/-	[149]
Non-obese (n = 90) and obese (n = 171) children	PBC	POMC	Hypermethylaton associated with obesity 37.5% differences in DNA methylation between groups Appetite regulation	✓/✓	[150]
Obese (n = 7) and non-obese (n = 7) adults	PBL	UBASH3A, TRIM3	Hypermethylation of UBASH3A and hypomethylation of TRIM3 genes in obese subjects compared with lean controls 2.5–10.1% differences in DNA methylation between groups	✓/✓	[26]
Adults (n = 40)	PBMC	SERPINE1, LINE-1	Association between hypermethylation of LINE-1 and SERPINE1, and WC and BMI	✓/✓	[151]
RESMENA (n = 48) and OBEPALIP (n = 25) cohorts	WBC	GPR13, ITGB5	Association between DNA methylation and BMI > 10% differences in DNA methylation between groups	✓/✓	[152]
Qatari descent (n = 30) and TwinsUK (n = 32) cohorts	Blood	CPT1A, TXNIP, ABCG1	Association between hypomethylation of CPT1A and TXNIP and obesity Association between hypermethylation of ABCG1 and BMI 2.7–3.77 rate of change of T2D characteristics as methylation changes	✓/✓	[112]
Men with hypercholesterolemia (n = 61) and morbid obese men(n = 30)	VAT and blood	ADRB3	Association between hypermethylation of ADRB3 gene and visceral obesity and fat distribution in VAT, and LDL-C and waist-to-hip ration in blood	✓/-	[93]
Obese (n = 50) and non-obese girls (n = 50)	Saliva	NAV3, MC2R	Association between hypermethylation and BMI 11.6% differences in DNA methylation between groups	-/-	[96]
Adults (n = 431)	Saliva	LEP	Association between DNA methylation and BMI	✓/-	[94]
Children (n = 92)	Saliva	Genome-wide	Association between 17 CpGs methylation and BMI	✓/✓	[95]
Adults (n = 43)	Blood and oral mucosa	GAP43, ATP2A3, ADARB2	Association between genes methylation and overweight	x/x	[97]
NEST cohort (n = 92)	Blood	PLAGL1, NEG3	Hypermethylation of PLAGL1 and hypomethylation of NEG3 in obese mothers 1.5–3.9% differences in DNA methylation between groups	✓/x	[153]
Children (n = 88)	Cord blood	MMP7, KCNK4, TRPM5, NFKB1	Association between GWG and hypermethylation of genes	✓/✓	[154]
BBC cohort (n = 309)	Blood leucocytes	Genome-wide	Association between maternal BMI and DNA methylation of 20 CpGs	✓/✓	[155]
Cardiogenic Consortium (n = 479), MARTHA (n = 339), KORA F4 (n = 1789), and MuTHER (n = 635) cohorts	Blood and adipose tissue	HIF3A	Association between DNA hypermethylation of gene and BMI 2.3–4.6% increase in BMI as DNA methylation changes in 0.1%	✓/✓	[130]
MZ twin pairs (n = 84)	PBC	SLC6A4	Association between DNA hypermethylation of gene and BMI, weight and WC Increase 0.3 kg/m2 BMI, 0.16 kg body weight, and 0.78 cm in WC per 1% of DNA methylation	✓/✓	[156]
ARIC cohort (n = 2097)	WBC	Genome-wide	Association between eight CpG methylation and BMI changes	✓/✓	[157]
Three cohorts: BPRHS, GOLDN and FHS (n = 1986)	CD4+, WBC	APOA2	Association between DNA methylation status of the gene with SFA intake and obesity risk	✓/✓	[158]

Abbreviations: CpG, cytosine linked by a phosphate to guanine; PBC, peripheral blood cells; BMI, body mass index; DMP, differentially methylated positions; GOLDN, Genetics of Lipid Lowering

Drugs and Diet Network; ARIC, Atherosclerosis Risk in Communities; FHS, Framingham Heart Study; PBMC, peripheral blood mononuclear cells; WC, waist circumference; RESMENA, Metabolic

Syndrome Reduction in Navarra; OBEPALIP, Effects of Lipoic Acid and Eicosapentaenoic Acid in Human Obesity; WBC, white blood cells; VAT, visceral adipose tissue; LDL-c, low-density

lipoprotein cholesterol; NEST, The Newborn Epigenetic Study; GWG, gestational weight gain; BBC, Boston Birth Cohort; MARTHA, MARseille THrombosis Association; KORA, Cooperative

Health Research in the Region of Augsburg; MuTHER, Multiple Tissue Human Expression Resource; MZ, monozygotic; BPRHS, Boston Puerto Rican Health Study; SFA, saturated fatty acid. ✓,

adjusted; x, no adjusted; -, no information

A study in three populations of diverse ancestries found that the methylation levels of the apolipoprotein A2 (APOA2) regulatory region was associated with the consumption of saturated fatty acids [50], a nutritional factor that has been associated with an increase in the risk of obesity in previous studies.

Not only is the methylation profile of blood cells associated with obesity, but also the methylation of other tissues and body fluids. An EWAS identified predominantly DNA hypermethylation in white adipose tissue (WAT) from obese subjects, which was related to gene expression of proinflammatory pathways. These findings suggest that DNA methylation may link dysfunctional adipocytes to WAT inflammation and IR in obesity [92]. Hypermethylation of the beta-3 adrenoceptor (ADRB3) gene was also reported in WAT of obese subjects, which was linked with an increased susceptibility to visceral obesity and altered body fat distribution [93]. Similar results were previously described in blood cells [93], revealing that the methylation changes of ADRB3 gene in blood may reflect obesity-related DNA methylation changes of this gene in WAT. In addition, a case-control study demonstrated an association between hypomethylation of leptin (LEP) gene in obese individuals, measured in saliva, and obesity-related parameters [94]. Likewise, an EWAS performed in 92 children’s saliva samples described 17 CpGs associated with maternal BMI [95]. The saliva sample analysis in 50 girls with and without obesity found two interesting genes, neuron navigator 3 (NAV3) and melanocortin 2 receptor (MC2R), whose methylation levels were associated with BMI [96]. These studies evidenced that saliva may be a viable probe for epigenetic testing in obesity. However, further studies would have to include both, saliva and blood samples, to demonstrate that saliva is consistent with whole blood findings. With this purpose, a recent study provided a novel application of noninvasive buccal samples for the identification of DNA methylation markers in relation to the management of overweight and insulin sensibility [97].

Most of the investigations analysing DNA methylation in relation to the pathophysiology of obesity are performed in a cross-section studies, where the data were collected by the observation of the variable at the same time point. In most of the populations, it has been difficult to rule out other effects such as the influence of nutritional status, nutrients, diet, or genetics. Moreover, in some of these cohorts, important covariates are missing. Even though all these studies back the DNA methylation-obesity link hypothesis, most of them are association studies and are far from causative and can only correlate DNA methylation with phenotypic outcomes. In this sense, few longitudinal studies have tried to investigate the causal effect of DNA methylation on obesity and related measures (Table 2).

Table 2.

Longitudinal studies analysing the association between DNA methylation and obesity.

Study population	Sample type	Target gene	Major finding	Age/Sex adjusted	Reference
Botnia prospective study: T2D (n = 129); controls (n = 129), 8 years	Blood	PHOSPHO1, ABCG1	Associated with BMI, insulin, TG.	✓/✓	[104]
Children (n = 438), 5 years	Blood	Genome-wide	Associated with obesity, BMI z-score, HOMA-IR, insulin and glucose	✓/✓	[98]
GUSTO cohort (n = 987), 4 years	Blood	CDKN2B/P4HA3	Associated with BMI	x/✓	[99]
CHAMACOS study (n = 373), 9 years	Blood	PPARγ	Associated with body size	x/✓	[159]
Two birth cohorts: ALSPAC (n = 178), 7 years, and PTBGS (n = 24), 11–13 years	Blood	CDKN1C, EPHA1, CASP10, HLA, NID1	2.08% and 0.80% increase in BMI per 1% increase in methylation at CDKN1C and EPHA1 respectively Increase of 5.16% and 1.84% in fat mass per 1% increase in methylation respectively.	✓/✓	[100]
Two prospective cohorts: PAH, (n = 78), 9 years and SWS (n = 239), 6 years	Blood	RXRA	Higher methylation of RXRA, was associated with lower maternal carbohydrate intake and neonatal adiposity and fat mass	✓/✓	[101]
Two cohort: NPBGS (n = 94), 11 years and ALSPAC (n = 161), 7 years	Blood	TACSTD2	Hypomethylation associated with higher fat mass	✓/✓	[102]
BSCC study (n = 600), 2.5 years	Blood	LINE-1	Hypomethylation associated with adiposity	-/-	[103]
POUNDS Lost Trail (n = 48)	Blood	NFATC2IP	A 52.8% impact on weight change mediated by the cis DNA methylation at NFATC2IP	-/-	[105]

Abbreviations: T2D, type 2 diabetes; BMI, body mass index; TG, triglycerides; HOMA-IR, homeostasis model assessment-insulin resistance; GUSTO, Growing Up in Singapore Towards healthy Outcomes; CHAMACOS, Center for the Health Assessment of Mothers and Children of Salinas; ALSPAC, Avon Longitudinal Study of Parents and Children; PTBGS, Preterm Birth Growth Study; PAH, Princess Anne Hospital; SWS, Southampton Women’s Survey; NPBGS, Newcastle Preterm Birth Growth Study; BSCC, Bogota School Children Cohort; POUNDS, The Preventing Overweight Using Novel Dietary Strategies. ✓, adjusted; x, no adjusted; -, no information.

Most of the longitudinal and prospective studies have been focused on the paediatric population, where obesity is not usually accompanied by other comorbidities that may confound the results. Prospective studies during the early life stage described DNA methylation of several weight-linked loci in new-borns that continued to show a longitudinal association with adiposity, fat mass, body size and other obesity parameters in childhood [98–103]. Interestingly, several prospective studies including adult life stage evidenced that these DNA methylation pattern changes were associated with the later risk of developing metabolic and other diseases [104,105]. These examples described an association between early cues (environmental stimulation in utero, in early infancy, and in adult stage) and the later development of disease. Actually, this is the basis of the ‘epigenotype model’ of DOHaD, in which the environment can modulate the epigenetic signature during human life.

DNA methylation and MetS

MetS is characterized by a cluster of risk factors associated with CVD, diabetes, obesity, and stroke, which include visceral adiposity, IR, hypertension, hypertriglyceridemia, and low high-density lipoprotein cholesterol (HDL-C) phenotypes [106]. There are several criteria for the diagnosis of MetS. According to the World Health Organization (WHO) definition, the diagnosis of MetS requires obesity, hyperinsulinemia plus two or more other clinical components. However, the most used definition is the revised Adult Treatment Panel-III, which requires at least three or more alterations.

The prevalence of MetS is rapidly increasing in most countries, affecting more than 20% of the adult population globally [107]. In 2010, the Spanish adult population presented higher prevalence of MetS, specifically, 38.37% in men and 29.62% in women, these percentages increasing in the elderly [108]. The high prevalence of MetS makes the implementation of new strategies for prevention necessary.

Numerous studies in humans have investigated the possible association between DNA methylation and MetS features, including visceral adiposity, IR, hypertension, HDL-C, and hypertriglyceridemia. Most of them observed that the DNA methylation patterns of different genes were associated with MetS markers, resulting in a worse prognosis or a higher risk of presenting other diseases (Table 3). Moreover, several studies have found evidence that MetS risk factors are likely to play a role in the DNA methylation process. Modification of gene methylation has been reported in an initial phase of the development of MetS.

Table 3.

Studies analysing the association between DNA methylation and MetS features.

Study Population	Sample type	Methylation strategy	Target gene	Major results	Age/Sex adjusted	Reference
GDM (n = 87) and healthy pregnant (n = 81)	Umbilical cord blood	Gene-specific	GNAS	GDM increases methylation level at foetal GNAS Hypermethylation of GNAS associated with high risk for diseases in offspring	-/-	[109]
GDM dietetically treated (n = 88), GDM insulin-dependent (n = 98), healthy pregnant (n = 65)	Cord blood and chorionic villi	Gene-specific	MEST	GDM decreases MEST methylation in new-born 7.8% differences in DNA methylation between groups MEST hypomethylation contributes to risk of development MetS	✓/✓	[110]
Controls (n = 70) and MetS subjects (n = 64)	VAT	Gene-specific	LPL	Association between LPL hypermethylation and aetiology of MetS	✓/✓	[19]
GDM (n = 82) and T1D in pregnancy (n = 67)	Skeletal muscle, SAT and blood	Gene-specific	TXNIP	Association between intrauterine hyperglycaemia and changes in TXNIP methylation 5.2% differences in DNA methylation between groups	✓/✓	[160]
Controls (n = 48) and T2D (n = 12)	Pancreatic islets	Gene-specific	PPARGC1A	PPARGC1A hypermethylation in T2D subjects Association between PPARGC1A hypermethylation and lower insulin secretion 55.2% differences in DNA methylation between groups	✓/-	[114]
FH (n = 61) and severely obese non-FH (n = 30)	Blood and VAT	Gene-specific	ADRB3	Association between ADRB3 hypomethylation and higher LDL-c, WC and apoB levels 28.8% differences in DNA methylation between groups	✓/-	[93]
Healthy men (n = 13)	Muscle biopsies	Gene-specific	ANGPTL4	Association between ANGPTL4 hypomethylation and low blood glucose and insulin sensitivity	-/-	[161]
Adults (n = 294)	SAT and blood	EWAS	2825 genes, 711 CpGs	Association between 2825 gene methylation and BMI Association between 711 CpGs methylation and HbA1c	✓/✓	[111]
Eight retrospective studies (n = 8,061), two case-control study (n = 687), and three cohorts (n = 1,490)	Blood	EWAS	187 CpGs	Association between 187 CpGs methylation and BMI Future risk for the development of T2D 6.3–40.2 kg/m2 change per 1 unit increase in DNA methylation	✓/✓	[25]
Adults (n = 185)	Blood	EWAS	CPT1A, TXNIP, ABCG1	Association between CPT1A and TXNIP hypomethylation and T2D and BMI 2.7–3.77 rate of change of T2D characteristics as methylation changes	✓/✓	[112]
LOLIPOP study (n = 25,372)	Blood	EWAS	ABCG1, PHOSPHO1, SOCS3, SREBF1	Association between gene methylation and BMI, WC, insulin and HOMA-IR 0.5–1.1% differences in DNA methylation between groups	✓/✓	[113]
Children (n = 438)	Neonatal blood	EWAS	69 CpGs	Association between 69 CpGs methylation and HOMA-IR, insulin, and glucose levels	✓/✓	[98]
Adults (three cohorts, n = 1,167)	Blood	EWAS	TXNIP	Association between DM and hypomethylation of TXNIP gene TXNIP hypomethylation is a consequence of hyperglycaemia levels 0.05–5% differences in DNA methylation between groups	✓/✓	[162]
LBW (n = 3), NBW (n = 3) and HBW (n = 3)	Cord blood and placenta	EWAS	360 DMR in LBW and 773 DMR in HBW	Genes encoding p38 MAPK signalling components are hypermethylated in LBW and HBW PI3K/AKT pathway and glucose transport components hypomethylated in HBW	✓/✓	[163]
ESTHER study (n = 527)	Blood	EWAS	TXNIP	Association between TXNIP hypomethylation and increasing fasting glucose and HbA1c	✓/✓	[115]
FH (n = 276) and controls (n = 26)	Blood	EWAS	TNNT1	TNNT1 methylation is related to 10% of the inter-individual variation in HDL-c levels > 5% differences in DNA methylation between groups	✓/✓	[164]
Adults (three cohorts, n = 99,994)	WBC	EWAS	IGFBP3,KCNK3, PDE3A,PRDM6, ARHGAP24,OSR1, SLC22A7,TBX2	Association between gene methylation and BP and hypertension	x/✓	[118]
Adults (n = 850)	Blood	EWAS	CPT1A, ABCG1	Association between gene methylation and TG storage	✓/✓	[116]
Controls (n = 59) and obese (n = 58)	SAT	EWAS	Genome-wide	Association between global DNA hypermethylation and WAT inflammation and IR	x/x	[92]
Botnia study: T2D (n = 129) and controls (n = 129)	Blood	Gene-specific	PHOSPHO1, ABCG1	Association between PHOSPHO1 methylation and HDL-c levels. Association between ABCG1 methylation and BMI, HbA1c, fasting insulin and TG levels	✓/✓	[104]
Human cell line (in vitro)	SW62 cell	Gene-specific	HSD11B2	Association between HSD11B2 methylation and blood pressure	-/-	[117]
Human cell line (in vitro)	NCI H295R	Gene-specific	AGT	Hypomethylation of AGT may result in hypertension and kidney injury.	-/-	[165]
C57BL/6J mice with IH	Endothelial cells	Gene-specific	Ace1, Agt	Association between gene methylation and systolic hypertension in mice exposed to IH	-/-	[166]

Abbreviations: GDM, gestational diabetes mellitus; MetS, metabolic syndrome; VAT, visceral adipose tissue; T2D, type 2 diabetes; SAT, subcutaneous adipose tissue; FH, familiar

hypercholesterolemia; LDL-c, low density lipoprotein cholesterol; WC, waist circumference; apoB, apolipoprotein B; EWAS, epigenome-wide association study; CpG, cytosine linked by a phosphate

to guanine; BMI, body mass index; HbA1c, haemoglobin A1C; LOLIPOP, London Life Sciences Prospective Population Study; HOMA-IR, homeostasis model assessment-insulin resistance; WAT,

white adipose tissue; IR, insulin resistance; LBW, low birth weight; HBW, high birth weight; DMR, differentially methylation region; ESTHER, Epidemiologische Studie zu Chancen der Verhütung,

Früherkennung und optimierten THerapie chronischer ERkrankungen in der älteren Bevölkerung; HDL-c, high density lipoprotein cholesterol; TG, triglycerides; BMI, body mass index, SW62, colon

carcinoma cell line; NCIH295R, adrenocortical cell line; IH, intermitent hipoxia. ✓, adjusted; x, no adjusted; -, no information.

Two prospective case-control studies in new-borns from mothers with gestational diabetes, observed changes in the DNA methylation levels of different genes, supporting the hypothesis that these epigenetic alterations contribute to the life-long risk of the development of obesity and other metabolic disorders [109,110]. In addition, an EWAS involving 8,165 participants integrating data from six independent cohorts, two case-control, and eight retrospective studies, showed a causal relationship between adiposity and DNA methylation alteration in blood cells and adipose tissue. This data supports an important function for altered DNA methylation, mediated by visceral adiposity, in the development of several diseases, such as T2D, obesity, CVD, and cancer [25,111]. Moreover, a case-control study with 64 subjects focused on hypertriglyceridemia found an association between lipoprotein lipase (LPL) gene promoter methylation, which was higher in these individuals, and poor metabolic profile and the development of MetS [19].

Moreover, an EWAS performed in 483 children identified a number of DMRs at birth that were associated with insulin sensitivity in childhood [98]. Many of these changes in DNA methylation were causally related to health outcomes [98]. In addition, the methylation patterns of different genes have been associated with increased risk of MetS and T2D [112–114]. Some of these genes were also involved in the dysregulation of glucose metabolism, hypertriglyceridemia, and decreased HDL-C levels, suggesting a role in the development of IR and CVD [104,115,116]. Other studies have associated the methylation patterns of different genes with alterations in blood pressure (BP) and endothelial vascular function resulting in hypertriglyceridemia [117,118].

DNA methylation markers of weight loss

DNA methylation has been associated with body weight regulation, since it has been involved in weight-related functions such as appetite, adiposity, adipogenesis, glucose, and lipid metabolism [119]. In addition, dietary factors, such as FAs, polyphenols or methyl donors, physical exercise, or pharmacological and surgical treatments, can induce changes in the DNA methylation pattern [23,75,120–122]. In this context, several studies have identified DNA methylation as a mechanism that could be implicated in the inter-individual variability to weight loss response (Table 4).

Table 4.

Studies analysing the association between DNA methylation and weight loss.

Weight loss strategy	Sample study	Sample type	Target gene	Main results	Age/Sex adjusted	Reference
Genotypic information	Obese and overweight (n = 95)	Blood	Global DNA	Association between global DNA hypomethylation and increased weight loss	✓/✓	[128]
Physical activity	Healthy women (n = 20)	Blood	NAMPT,RUNX3, BR,SLCO4C, WNT7A,RASGRP3,CYP2E,CA13, KANK4,SMOC2, SLIT3,GABRG3	Baseline DNA methylation was able to predict the percent body weight change over the six-month period	-/-	[126]
Behavioural and ER	NW (n = 20), overweight/obese (n = 20), and morbid obese women (n = 20)	Blood	CLOCK, BMAL1, PER2	Association between the baseline methylation of genes and the magnitude of weight loss. 10–60% differences in DNA methylation between groups at baseline	-/-	[75]
No treatment	Adults (n = 51)	Blood	POMC	Association between DNA hypermethylation and body weight Increase 10% of DNA methylation per 2.8 kg/m2	✓/✓	[167]
RYGB	Obese (n = 5) and control women (n = 6)	Skeletal muscle	409 DMR	409 DMR after weight loss >50% differences in DNA methylation between groups at baseline	✓/-	[120]
ER	Overweight and obese women (n = 14)	Adipose tissue	35 CpGs	Association between 35 loci methylation and weight control	✓/-	[123]
ER	Obese women (n = 27)	Adipose tissue	LEP, TNF	Association between hypomethylation of genes at baseline with better response to the dietary intervention.	-/-	[125]
ER	Overweight and obese adolescents (n = 24)	Blood	AQP9,DUSP22, HIPK3,TNNT1, TNNI3	Association between basal DNA methylation with changes in body weight, BMI-SDS, WC, and body fat mass after the weight loss intervention > 5% differences in DNA methylation between groups	-/-	[168]
ER	Obese men (n = 18)	Blood	POMC, NPY	Association between baseline NPY hypomethylation with weight-loss regain, and POMC hypomethylation with success in weight-loss maintenance	-/-	[124]
ER, BS	Control (n = 9), obese + ER (n = 22), obese + BS women (n = 14)	Blood	SERPINE1	Baseline DNA methylation may be a predictor of weight loss after BS	✓/-	[127]
ER	Obese and overweight men (n = 12)	Blood	ATP10A, CD44, WT1	Association between genes methylation at baseline and weight-loss outcome > 20% differences in DNA methylation between groups	-/-	[20]

Abbreviations: ER, energy-restriction; NW, normal weight; RYGB, Roux-en-Y gastric bypass; DMR, differentially methylated region; CpG, cytosine linked by a phosphate to guanine; BMI-SD, body mass index-standard deviation; WC, waist circumference; BS, bariatric surgery. ✓, adjusted; x, no adjusted; -, no information.

Investigations based on energy restriction have reported many CpGs whose methylation status was associated with the response to the weight loss intervention. For example, 35 different loci were differentially methylated between high and low responders to an energy-restricted diet in adipose tissue from overweight and obese women [123]. Many of these genes were associated with body weight control and insulin secretion [123].

In addition, the methylation status in adipose tissue and blood of several genes that participate in the regulation of BP, inflammation, lipid metabolism, appetite, and energy homeostasis has been related to differences in the weight loss response after a low-calorie diet (30% of energy-restriction) prescription [20,124,125]. On the other hand, DNA methylation of several clock genes implicated in the regulation of the circadian clockwork has been associated with body-weight loss in a women population under a behavioural and energy-restricted intervention [75].

Some studies have analysed other factors, such as physical activity or gastric surgery, in relation to the interaction between DNA methylation and weight loss. The analysis of blood samples from 20 healthy women within a physical activity programme identified significant associations between the methylation profile of 12 CpGs and weight loss [126]. A lifestyle and nutritional educational weight loss programme in 24 overweight and obese adolescents associated the methylation status of genes related to glucose metabolism, IR, inflammation, and CVD, with weight control and glucose and lipid metabolism [120,127].

Although genetic factors are involved in the regulation of body weight, genetic variants only partially explain the individual variation observed in the response to weight loss treatments. Thus, DNA methylation is thought to play a role in the environment-gene interaction in body weight regulation. In this context, Pirini et al. [128], studied the genetic and epigenetic alterations associated with weight loss within a personalized weight reduction programme designed on the basis of genotypic information, and identified an inverse association between global DNA methylation and weight loss depending on individual genetic variants for different genes.

Mechanisms linking DNA methylation and obesity

The present review has summarized the most relevant epidemiological studies analysing the association between DNA methylation and obesity during pregnancy, infancy, and adult life. Most of these analyses only describe a putative role of DNA methylation on the pathophysiology of obesity but are unable to deepen into the cause-and-effect relationship between epigenetics and obesity, which remains largely unknown.

Considering the epigenetic control of gene expression in the cellular activities, it is likely that the changes in DNA methylation contributed to the deregulation of important metabolic pathways and increased the susceptibility to develop obesity and its comorbidities. Despite the numerous epigenetic analyses performed in different populations and cohorts, they have not yet identified the precise mechanisms by which these changes in DNA methylation are related to the disease. However, some studies in rodents are helping to unveil the intricate mechanisms. For instance, DNA methyltransferase 3a (Dnmt3a) enzyme impaired insulin tolerance by increasing the DNA methylation of the direct target Ffg21 gene and reducing its expression in in vitro and in vivo models [129]. In the present review, multiple studies have shown changes in DNA methylation at key metabolic genes associated with obesity, such as HIF3A [130], IGFBP3 [118], SREBF1 [113], TNF [125], and CLOCK [75], pointing out to having important metabolic functions such as hypoxia, adipogenesis, inflammation, and circadian rhythm.

For example, HIF-system may play a role in mechanisms involved in the pathophysiology of adipose-tissue inflammation and IR. The abnormal expansion of WAT in obesity makes the correct blood perfusion on adipose tissue resulting in hypoxia impossible. HIF transcription factors are in charge of activating hypoxia-related gene expression, facilitating the adaptation to a hypoxic environment that contributes to inflammation and IR in WAT [65]. However, molecular mechanisms implicated in WAT dysfunction are still poorly characterized [131]. Therefore, HIF-system control genes related to glycolysis, such as pyruvate dehydrogenase kinase 1 (PDK1), lactate dehydrogenase A (LDHA), or glycogen phosphorylase L (PYGL), stimulating glucose uptake or increasing glucose production by activating phosphoenolpyruvate carboxykinase (PEPCK) in liver [132]; but HIFs also have a role in driving NF-κB and NFAT activation [131]. Interestingly, HIF activity has been reported to be regulated by DNA methylation [133,134]. This hypothesis has been demonstrated in the expression of the erythropoietin and class III beta-tubulin genes, where the HIF binding depends on the tissue-specific methylation status of HRE sites in the 3ʹUTR [135]. In addition to the role of DNA methylation in regulating HIF activity, hypoxia can induce epigenetic changes itself. Human colorectal and melanoma cell lines incubated in hypoxic conditions (< 0.1% oxygen) during 24 h suffered a 15–20% decrease in CpG methylation [135]. However, studies performed in conditions of hypobaric hypoxia (at high altitude) described progressive weight loss that was attributed to the reduction of appetite and food intake [65]. In addition, as a consequence of obesity and hypoxia, uncoupling protein one (UCP1) gene was stimulated, promoting thermogenesis by lipid oxidation and leading to browning of WAT [131]. These studies raise the controversial question: is hypoxia good or bad? In this context, recent investigations consider that the positive metabolic effects resulting from the activation of HIF signalling can be more beneficial than pathological against metabolic complications [65].

The process whereby fibroblast-like progenitors become mature adipocytes and accumulate nutrients and triglycerides is known as adipogenesis. Adipocyte expansion (hyperplasia) can improve the harmful effects of obesity [136]. Although many signalling pathways and ligands modulate the process of preadipocyte differentiation into mature adipocytes, the mechanisms and regulators of this adaptive mechanism are still under investigation [136]. Two of the main proteins involved in adipogenesis are peroxisome proliferator-activated receptor gamma (PPARγ) and co-activator CCAAT/enhancer-binding protein-α (C/EBPα). Also, the insulin signalling-cascade (including IGF receptors) plays an important role in this process [136]. In this context, in sows, maternal methyl donor supplementation throughout gestation has been reported to enhance the birth weight and postnatal growth rate of the offspring, associated with an increased expression of IGF1 and IGF1R, as well as an altered DNA methylation of IGF1 gene promoter [137]. During human adipogenesis, large epigenetic and transcriptomic changes occur in the transition from preadipocytes to mature adipocytes [138]; for example, the adipocytes present higher expression and lower methylation levels in genes involved in adipogenesis-related processes such as insulin signalling, PPAR signalling, and adipocytokine signalling [138]. These results suggest that DNA methylation changes may contribute to regulate adipogenesis. For instance, silencing of DNMT1 gene increases the rate of differentiation and modifies gene expression and lipid accumulation in mouse adipocytes [139].

In addition, other key physiological states involved in obesity, such as inflammation and circadian systems, can affect the adipogenic process. Obesity is characterized by a high expression of cytokines in WAT derived from the infiltrated macrophages. Macrophages correspond to about 40% of total WAT cells in obese subjects and 18% in lean individuals [140]. Macrophages secrete pro-inflammatory cytokines (i.e., TNF-α, IL-1β, and IL-6) and the inflammatory response stimulates Toll signalling pathways, which are able to phosphorylate insulin receptor substrate 1 (IRS-1) and, consequently, inhibit the insulin signal cascade [141], thus contributing to IR. The link between DNA methylation and inflammation in obesity-related comorbidities has been demonstrated. For example, epigenetic mechanisms play a key role in the regulation of macrophage activation and inflammation through the direct methylation of the PPARγ promoter by DNMT1 [142].

Recent investigations have revealed a wide interplay between the circadian system and metabolism. Disruptions in the physiological day-night cycles could be implicated in the development of obesity, as has been evidenced by studies of jet lag, shift work or disturbance in the timing of eating [143]. These conditions result in loss of internal rhythms and lead to weight gain and metabolic alterations [143]. Circadian rhythms are controlled by a set of core clock genes such as CLOCK, BMAL1, and NR1D1 (REV-ERBα). The CLOCK/BMAL1 heterodimer activates other core clock genes, such as PER and CRY [144]. These transcription factors drive the expression of many genes implicated in the regulation of lipid and glucose metabolism [145]. For instance, the circulating levels of lipids and the activity of enzymes involved in their metabolism, such as fatty acid synthase, fatty acyl-CoA synthetase 1, apolipoproteins, and low-density lipoprotein receptor, follow circadian rhythms in mammals [145]. CRY clock components also control gluconeogenesis by decreasing cAMP signalling, whereas CLOCK drives glycogen synthesis by enhancing Gys2 gene transcription [145]. In addition, the adipogenic factor PPARγ also oscillates with circadian rhythmicity in WAT [136]. PER activity is also cyclic in preadipocytes and, when it is activated, the adipogenic process is suppressed by the inhibition of PPARγ and KLF15 [146]. On the contrary, BMAL1, which oscillates in a contrary manner to PER, promotes adipogenesis by upregulating PPARγ [136]. Therefore, the abnormal DNA methylation of core clock genes could disrupt their oscillation causing metabolic disturbances.

Further studies are needed to elucidate the exact molecular mechanisms whereby DNA methylation and obesity are linked. However, transcriptional regulation of these genes, which participate in crucial pathways involved in the pathophysiology of obesity and its comorbidities, are partly mediated by DNA methylation. In this context, changes in the DNA methylation pattern of genes could contribute to modify the expression and disrupt the function of important metabolic pathways.

Critical analysis of the literature

Although there is increasing evidence of a link between epigenetic modifications and obesity and accompanying comorbidities in humans, yet several aspects remain obscure. The issue of causality remains the most pressing challenge because it is extremely difficult to establish and because most studies in humans only report correlations. To respond to this issue, it is necessary to follow big cohorts since conception, to compare their epigenetic marks (in different cell types) with those of their parents, to properly register the lifestyle and dietary patterns influences, to link DNA methylation results with gene expression in the same cells, and to analyse the interactions between the epigenetic marks, the genetic makeup and the environment. Currently, this task seems too vast and difficult.

Very few CpGs have repeatedly appeared in different human studies as significantly associated with obesity. This situation suggests that there is not probably a unique epigenetic pattern that is associated with the disease, but that the genetic, environmental, and physiological factors interact in a very personal way. On the other hand, one of the biggest limitations is that most studies in humans have been focused on peripheral blood cells, but very few have used liver, pancreas, muscle, adipose or other metabolically relevant tissues. As epigenetic marks are very dependent on cell type, it is imperative to broaden our knowledge of the epigenetic differences in these tissues, where most relevant biological functions (i.e., inflammation, hypoxia, IR,…) occur.

There are many studies focusing on individual CpGs that are differential between obese and non-obese subjects, but less have analysed DMR changes that could impact the regulation of the expression of metabolically relevant genes. On the other hand, the differences in DNA methylation are usually rather small (not higher than 15–20%) and very few studies have evidenced that these epigenetic changes are linked to changes in gene expression in normal conditions or in response to specific stressors. Theoretically, these small changes in methylation might be relevant when they are within a regulatory element of the gene, but evidence from human studies are lacking.

Another aspect that must be envisaged in the near future is the interplay between genetic, metagenomic and epigenetic characteristics of the individuals, and their environmental factors. Since epigenetic changes are a response to the environment, it can be hypothesized that each individual’s epigenetic pattern forms a kind of system shaped by his genetic background, perinatal history, long-term diet and lifestyle, pollutants, clinical history, drugs, usual gut microbiota, and many other factors. This extreme heterogeneity in the epigenetic patterns in the different cell types of the different individuals is one of the reasons that makes it difficult to find reliable epigenetic biomarkers associated with obesity.

As mentioned above, human studies are, currently, unable to deepen into the biological mechanisms that are linked to the epigenetic variations. There are so many environmental and biological factors that can mediate the relationships between epigenetic profile and health status, that it is very difficult to attribute specific epigenetic effects to different putative causative factors. The lack of adequate in vitro approaches and the inaccessibility to simulate the combination of several of these factors in a controlled manner adds additional difficulties to the analysis of possible causal nexi between all the variables simultaneously.

Nevertheless, in the last years considerable progress has been made in the identification and characterization of novel epigenetic signatures related to obesity and metabolic health, which could be relevant as risk biomarkers as well as causal factors. However, very few of them (if any) have been validated in independent cohorts. In this context, the use of large-scale population-based studies is necessary to untangle the complex interactions between epigenetics and environmental factors in the developmental origins of obesity and metabolic syndrome. To accomplish this objective, new instruments must be used to measure the nutritional, environmental, and parental interactions with the epigenetic modifications found in the individuals, including new indicators or markers of exposure, novel systems for collecting and processing nutritional data, and more control of the environmental conditions that may modify the epigenome.

Future perspectives

While these epigenetic studies allow understanding, at least partially, in the variability between obesity-related phenotypes and the interactions between genetics and environmental factors, it is still necessary to clarify how each specific environmental factor, such as diet or gut microbiota composition, targets specific genes to be methylated and demethylated. In addition, although all the studies found that DNA methylation is linked to obesity and obesity-related measures, until now, few studies have investigated the causal direction of phenotypic outcomes and DNA methylation profile. This question remains unclear and longitudinal analyses in other populations are required. It is also necessary to investigate how DNA methylation affects the pathophysiology of specific diseases, and the real effect on health and disease outcomes, in order to improve weight loss strategies. Finally, these studies could also help guide the use of DNA methylation regulation as a new therapeutic target for the prevention and treatment of obesity and other metabolic diseases.

Conclusion

The advances in the study of DNA methylation and the establishment of relevant longitudinal models in the study of obesity, may now allow to identify novel DNA methylation markers for obesity and related pathologies. Mainly, the first potential epigenetic markers at birth have been detected. This might help to predict obesity risk, adiposity and body size at a young age, and give opportunity for the development of prevention strategies. On the other hand, DNA methylation marks could improve the success of weight loss therapies in the context of precision nutrition. In summary, all these studies confirm that DNA methylation, in combination with genetic variants, gut microbiota and other biomarkers, could be useful in the assessment of metabolic risk and the personalization of the clinical management of obesity. However, it is necessary to conduct new studies in humans analysing the effects that nutritional, lifestyle and psychological factors can have on human health via epigenetic mechanisms.

Funding Statement

This work was supported by the Ministerio de Economía, Industria y Competitividad, Gobierno de España [AGL2013-4554-R]; Ministerio de Educación, Cultura y Deporte [BES-2014-068409].

Acknowledgments

The authors are grateful to Nora Goodwin (Technological University Dublin, Ireland) for English proofreading. The authors also acknowledge the financial support of MINECO (Nutrigenio Project reference AGL2013-45554-R), Spanish Centro de Investigación Biomédica en Red de la Fisiopatología de la Obesidad y Nutrición (CIBERobn) and University of Navarra. Mirian Samblas holds a FPI grant from the MECD (BES-2014-068409).

Disclosure statement

No potential conflict of interest was reported by the authors.