**3. Epigenomics**

Epigenomics is the study of alterations to the transcriptional activity of genes in response to environmental stimuli *without* altering the DNA sequence. These alterations include methylation, histone modification as well as other changes in the chromatin structure that may affect the transcriptional regulation of a given gene. Importantly, these epigenetic modifications are thought to be an important link in the "missing heritability" hypothesis [16-18]. The concept of epigenetic regulation of obesity has become of great interest over the last decade and several studies have demonstrated that epigenetic modifications can occur very early in life which may predispose individuals to obesity later in life.

#### **3.1. Early life epigenetic programming**

In the late 1980s, Barker and Osmond were the first to describe that early life exposure to certain adverse conditions including over or under nutrition and stress may increase susceptibility to disease later in life [19]. Since then, there have been many studies, particular animal studies that have verified this hypothesis and have demonstrated that malnutrition (both excess and deprivation) during pregnancy is associated with increased fat deposition in the offspring and may also directly impact the oocytes in females and primordial germ cells of male foetuses [20, 21]. Furthermore, there is evidence that *in utero* over or under nutrition can affect DNA methylation, histone post‐translational modification as well as gene expression of target genes involved in insulin signalling and fatty acid metabolism [22, 23]. Hence, this predisposes the offspring and the grand‐offspring to obesity and related disorders. Therefore, epigenetic modifications are hypothesised to be trans‐generational and have been shown to be reversible whereby specific epigenetic marks can be turned on or off depending on the stimuli (Bishop et al). Most of these studies have been performed in animal models including mice [24, 25], rats [26, 27], Macaque [28], drosophila [29] and sheep [23, 30, 31]. The effects of the *in utero* environment on foetal programming in humans have not been well explored. Two recent studies have investigated the effects of maternal nutrition before and during pregnancy on DNA methylation on the offspring. The first study, the Dutch Hunger Winter Study [22], examined the effects of prenatal exposure to famine on DNA methylation and was able to show an increased methylation in *LEP* and *IL10* and decreased methylation in *IGF2* and *INSIGF*; however, these subjects were obese, and therefore, it is not clear whether these changes occurred due to the maternal famine exposure or a response to increased BMI. The second study, the Gambian mother‐child cohort, was a large randomised control trial which examined the effects of DNA methylation in response to maternal nutrition at conception and have demonstrated altered methylation in the offspring of these mothers and that these epigenetic changes were sex specific [32, 33]. A more recent randomised controlled study examined the effects of double blind peri‐conceptional micronutrient supplementation in cord blood and observed sex‐specific methylation changes which further supports the importance of *in utero* nutrition and the potential effects it may have on adiposity later in life (Dominguez‐Salas, Moore et al. 2014). Another study, the AVON Longitudinal Study of parents and children, performed an epigenome‐wide association analysis in 1,108 patients and found that the maternal underweight condition may be more influential on the DNA methylation of the offspring compared to the offspring from the maternal overweight mothers, while weight gain during pregnancy did not have much effect on DNA methylation in the offspring [34]. Although there is growing evidence of the effects of maternal nutrition on fatty acid metabolism and insulin signalling in humans later in life, there still remains some uncertainty and replication in large‐scale population cohorts are warranted. The early detection of changes in the expression and epigenetic changes of candidate genes may provide biomarkers that will prevent or delay the onset of disease.

#### **3.2. Human epigenome‐wide association studies**

be explained, they have provided us with an emerging wealth of knowledge of the genomic localisation, frequency and effect sizes, and potential functional implications that these loci may have. More recently, a number of meta‐analyses of GWAS data have been performed and have identified new loci associated with body mass index (BMI) (**Table 1**). Surprisingly, these meta‐analyses also replicated previously identified loci, which validate that these loci may be contributing to obesity and underscore the importance of performing GWASs in the

Epigenomics is the study of alterations to the transcriptional activity of genes in response to environmental stimuli *without* altering the DNA sequence. These alterations include methylation, histone modification as well as other changes in the chromatin structure that may affect the transcriptional regulation of a given gene. Importantly, these epigenetic modifications are thought to be an important link in the "missing heritability" hypothesis [16-18]. The concept of epigenetic regulation of obesity has become of great interest over the last decade and several studies have demonstrated that epigenetic modifications can occur very early in life which may

In the late 1980s, Barker and Osmond were the first to describe that early life exposure to certain adverse conditions including over or under nutrition and stress may increase susceptibility to disease later in life [19]. Since then, there have been many studies, particular animal studies that have verified this hypothesis and have demonstrated that malnutrition (both excess and deprivation) during pregnancy is associated with increased fat deposition in the offspring and may also directly impact the oocytes in females and primordial germ cells of male foetuses [20, 21]. Furthermore, there is evidence that *in utero* over or under nutrition can affect DNA methylation, histone post‐translational modification as well as gene expression of target genes involved in insulin signalling and fatty acid metabolism [22, 23]. Hence, this predisposes the offspring and the grand‐offspring to obesity and related disorders. Therefore, epigenetic modifications are hypothesised to be trans‐generational and have been shown to be reversible whereby specific epigenetic marks can be turned on or off depending on the stimuli (Bishop et al). Most of these studies have been performed in animal models including mice [24, 25], rats [26, 27], Macaque [28], drosophila [29] and sheep [23, 30, 31]. The effects of the *in utero* environment on foetal programming in humans have not been well explored. Two recent studies have investigated the effects of maternal nutrition before and during pregnancy on DNA methylation on the offspring. The first study, the Dutch Hunger Winter Study [22], examined the effects of prenatal exposure to famine on DNA methylation and was able to show an increased methylation in *LEP* and *IL10* and decreased methylation in *IGF2* and *INSIGF*; however, these subjects were obese, and therefore, it is not clear whether these changes occurred due to the maternal famine exposure or a response to increased BMI. The second study, the Gambian mother‐child cohort, was a large randomised control trial

first place.

**3. Epigenomics**

26 Adiposity - Omics and Molecular Understanding

predispose individuals to obesity later in life.

**3.1. Early life epigenetic programming**

Stemming from the animal epigenetic studies, there has been a recent incline in epigenome‐ wide association studies (EWAS) across different human populations. Most studies have focused more on site‐specific differences in DNA methylation and its association with metabolic phenotypes, as it has been proposed to be a potential biomarker for clinical diagnosis and prognosis of disease [35]. A number of epigenetic marks related to obesity have been identified and replicated in other populations.

A study by Aslibeykan et al. (2015) performed an EWAS in CD4+ cells from frozen buffy coats from 991 healthy participants from the used Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) study which identified eight differentially methylated sites associated with BMI, including cg00574958 (*CPT1A*), cg04332373 (*CD38*), cg17287155 (*AHRR*), cg26164488 (NA), cg07504977 (NA), cg14476101 (*PHGDH*), cg26680760 (NA) and cg26140475 (NA). They also identified four differentially methylated sites associated with waist circumference Cg00574958 (*CPT1A*), Cg04332373 (*CD38*), Cg14476101 (*PHGDH*) and cg25349939 (*GTDC1*) [36]. They replicated these methylation sites in whole blood from two independent cohorts—the Framingham Heart Study (FHS) (*n* = 1,935 case: control study, and *n* = 442 random samples) and the Atherosclerosis Risk in Communities (ARIC) study (*n* = 2,015 samples). A follow‐up study in the ARIC study identified 76 BMI‐related and 164 WC‐related epigenetic marks mainly in *HIF3A*, *CPT1A* and *ABCG1* [37]. Thirty‐seven of the BMI and one waist circumference mark were replicated in the GOLDN (*n* = 991 samples) and FHS (*n* = 2,377 samples) cohort. Another study identified 982 individual differentially methylated CpG sites between the insulin sensitive (IS) and insulin resistant (IR) groups. Of these sites, 538 were associated with unique genes with functions in cell adhesion, signal transduction and regulation of transcription and include *ADAM2, IGF2BP1, TBX5, HDACM, CD44* and *ZNF711*. These data are supportive of the close link between obesity and T2D and suggest the existence of a methylome map within Visceral adipose tissue (VAT) that may predispose individuals to IR and T2D [38].

Family studies are beneficial as they have the potential to identify causative candidate regions of differential DNA methylation. A study by Ali et al. (2016) used a discovery approach in 192 subjects from seven families from the Take off Pounds Sensibly Family Study of Epigenetics (TFSE), and then validated the initial differential methylation marks in an extended cohort of 1,052 subjects [39]. They identified and replicated three loci where methylation status was associated with BMI%, and these were located in the body of suppressor of cytokine signalling (*SOCS3*), 3′ untranslated region of the zinc finger protein 771 (*ZNF771*) and at the transcription start site of LIM domain containing 2 (*LIMD2*) gene. Functional analyses were then performed in 330 subjects from the methylation analyses [39].

Examining promoter methylation patterns is thought to be a promising strategy for the use of early detection of disease (Laird 2003). Given that TNF*a* is pro‐inflammatory cytokine elevated in obese subjects, Campion et al. (2009) investigated methylation patterns in the promoter region of TNF*a*. Methylation was measured in baseline peripheral blood mononuclear cells (PBMCs) from 24 patients who were put on an 8‐week calorie restricted diet. Two differentially methylated promoter regions were assessed according to previous literature. The group of men who lost weight with the low calorie diet showed a significantly lower TNF*a* promoter methylation and that their baseline circulating TNF*a* levels were positively associated with total promoter methylation (Campion, Milagro et al. 2009). This demonstrates that examining promoter methylation may well be an important aspect in early detection and prognosis of disease.

#### **3.3. Can PBMCs be used as a surrogate for epigenetic studies?**

Despite progress in the epigenetic marks of obesity, there still remain some controversy in the use of peripheral blood mononuclear cells (PBMCs) as a surrogate for identification of epigenetic marks that may contribute to obesity development as there may be site‐specific marks that may be missed. However, it is important to consider that with human studies, we are limited to tissues that we can readily access and the most commonly and easily accessible tissue to examine epigenetic modifications in humans is whole blood or PBMCs. This has been thought to have limitations in itself as it may not explain tissue‐specific modifications that may be present. In light of this, many recent studies have examined epigenetic profiles in both PBMCs as well as subcutaneous adipose tissue (SAT). Once such study was from Arner et al. 2016, where they performed and EWAS using PBMCs, SAT and VAT in 80 obese women of which 40 insulin resistant and 40 insulin sensitive (Arner, Sahlqvist et al. 2016). They wanted to determine similarities of methylation status between the tissues and to determine whether PBMCs could be used as a marker of systemic IR. They were not able to show similar methylation sites between the PBMCs and adipose tissue and therefore concluded that CpG methylation in PBMCs does not reflect differential methylation sites in white adipose tissue (WAT). These data suggest that PBMCs may not be a suitable tissue for metabolic phenotyping of obese individuals. However, another study by Demarath et al. (2015) (described above) also measured methylation markers in SAT from their ARIC cohort and replicated 16 of these BMI markers (Demerath, Guan et al. 2015). Given that these associations remained in different ethnic cohorts as well as different tissue samples (whole blood vs. CD4 cells from buffy coat vs. adipose tissue) demonstrate that adiposity traits are associated with different methylation sites independent of tissue type or ethnicity. It is clear that more thorough investigations into tissue differences are required; however, this may not always be possible in some settings such as brain‐ or heart‐related disorders.
