Dietary fat and obesity as modulators of breast cancer risk: Focus on DNA methylation

Abstract

Breast cancer (BC) is the most common cancer and second leading cause of cancer mortality in women worldwide. Validated biomarkers enhance efforts for early detection and treatment, which reduce the risk of mortality. Epigenetic signatures have been suggested as good biomarkers for early detection, prognosis and targeted therapy of BC. Here, we highlight studies documenting the modifying effects of dietary fatty acids and obesity on BC biomarkers associated with DNA methylation. We focus our analysis on changes elicited in writers of DNA methylation (i.e., DNA methyltransferases), global DNA methylation and gene‐specific DNA methylation. To provide context, we precede this discussion with a review of the available evidence for an association between BC incidence and both dietary fat consumption and obesity. We also include a review of well‐vetted BC biomarkers related to cytosine‐guanine dinucleotides methylation and how they influence BC risk, prognosis, tumour characteristics and response to treatment.

Linked Articles

This article is part of a themed section on The Pharmacology of Nutraceuticals. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v177.6/issuetoc

Abbreviations

AA

arachidonic acid (20:4)

AHR

aryl hydrocarbon receptor

APC

adenomatous polyposis Coli

BMI

body mass index

BRCA1

BC 1 β‐value: Infinium HumanMethylation 450 k probes

COL18A1

collagen type XVIII α1 chain

CpG

cytosine‐guanine dinucleotides

CPT1B

cg01081346 CpG site

DFA

dietary fatty acids

DMBA

7,12‐dimethylbenz[a]anthracene

DNMT

DNA methyltransferases

ECAD

E‐cadherin 1

EEF2

eukaryotic translation elongation factor 2

ER

oestrogen receptor

EVOO

extra virgin olive oil

EZH2

enhancer of zeste 2 polycomb repressive complex 2 subunit

FOb

formerly obese

5‐mdC

5‐methyldeoxycytosine

GNAS

cg17071192 CpG site

HCO

high‐corn oil diet

HER2

human EGF receptor‐2

HIN1

hairpin‐induced 1

HOO

high‐OO diet

IHC

immunohistochemistry

IFRD1

IFN‐related developmental regulator 1

LA

linoleic acid (18:2)

LEPR

leptin receptor

LF

low fat

LUMA

luminometric methylation assay

MAPKAPK2

MAPK‐activated protein kinase 2

MBD

methyl‐CpG binding domain protein

MD

Mediterranean diet

MD + Nuts

Mediterranean diet plus nuts

MUFA

monounsaturated fatty acids

NACT

neoadjuvant chemotherapy

Ob

obese

OS

overall survival

PLAGL1

plag 1 zink finger 1

PPARGC1B

peroxisome‐activated proliferator receptor γ coactivator 1 β

PREDIMED

Prevencion con Dieta Mediterranea

PR

progesterone receptor

PUFA

polyunsaturated fatty acid

RARβ

retinoic acid receptor β

RASSF1A

Ras association domain family 1 isoform A

SCD1

stromal stearoyl‐CoA desaturase 1

SFA

saturated fatty acid

TIMP

metallopeptidase inhibitor 3

TNBC

triple‐negative breast cancer

TNM

tumour node metastasis

TSC22D3

TSC complex subunit 22 domain family member 3

TWIST

twist‐related protein 1

WBC

white blood cells

WC

waist circumference

WHR

waist to hip ratio

1. INTRODUCTION

Breast cancer (BC) is the most common cancer and second leading cause of cancer mortality in women (DeSantis, Ma, Goding Sauer, Newman, & Jemal, 2017). Prevention, detection and treatment of BC are optimized by using validated biomarkers that provide information regarding risk, prognosis, tumour characteristics and therapeutic response. Epigenetics is the functional regulation of gene expression without a change in DNA nucleotide sequence. Hence, epigenetics refers to methylation of cytosine‐guanine dinucleotides (CpG) in DNA, histone post‐translational modifications, changes in non‐coding RNA expression and recruitment of chromatin remodelling factors. Aberrant epigenetic regulation is characteristic of most cancers (Kanwal, Gupta, & Gupta, 2015) and considered an early event in mammary carcinogenesis (Pasculli, Barbano, & Parrella, 2018). For this reason, epigenetic signatures have been suggested as good biomarkers for early detection, prognosis and targeted therapy of BC (Bhat, Majid, Wani, & Rashid, 2019). Moreover, these modifications may be useful targets for pharmacological intervention in BC.

Positive associations between high‐fat diets (HFD) and BC incidence have been documented in observational and preclinical studies (Boyd et al., 2003; Cao, Hou, & Wang, 2016; Liu et al., 2014; Nguyen et al., 2017; Wu, Yu, Tseng, Stanczyk, & Pike, 2009). Moreover, obesity is considered a BC risk factor and may influence BC survival (Chen et al., 2016; Chen et al., 2017; Pierobon & Frankenfeld, 2013; Xia et al., 2014). Both dietary fats and obesity influence various epigenetic signatures associated with BC. Therefore, the purpose of this review is to highlight studies documenting modulating effects of dietary fatty acids (DFAs) and obesity/adiposity on BC biomarkers associated with DNA CpG methylation. We focus our review on DNA methylation given that detecting CpG methylation in haematological samples has shown promise as a non‐invasive and high‐throughput method for early detection and screening. To provide context, we precede this discussion with a review of the available evidence for an association between BC incidence and both dietary fat consumption and obesity. We also include a review of BC biomarkers related to CpG methylation and how they influence BC risk, prognosis, tumour characteristics and response to therapies. We close this review with a section describing examples of studies focusing on the pharmacological impact of various fatty acids (FA) on CpG methylation and BC risk.

2. DIETARY FAT AND BREAST CANCER (BC)

2.1. Total fat

2.1.1. Epidemiological studies

Results from epidemiological studies regarding an association between BC incidence and higher consumption of dietary fatty acid (DFA) are discordant (Table 1). For example, an analysis of 46 studies suggested a relative risk (RR) increase of ~13% (RR = 1.13) when comparing highest versus lowest levels of fat consumption (Boyd et al., 2003). In contrast, an analysis comprising 57 studies suggested no association, although when data were stratified by menopausal status, a small albeit statistically significant positive association was found for post‐menopausal BC (RR = 1.04; Turner, 2011). Conversely, no association was found for pre‐menopausal cases, although subjects were not stratified by BC subtype. When considering the relationship between DFA consumption and risk of BC, a meta‐analysis comprising 24 studies from the United States, Europe, Canada and Asia suggested an RR increase of ~10% (RR = 1.10; Cao et al., 2016). However, upon a subgroup analysis, this effect was specific to post‐menopausal women. Conversely, studies conducted with Chinese women reported a marked increase (RR = 1.36) in BC risk associated with higher fat consumption (Liu et al., 2014). Another analysis of studies conducted with women residing in China confirmed this positive association, although the magnitude of the effect was not as large (15% increase, RR = 1.15). Unfortunately, meta‐analyses that specifically examined this association in other ethnic groups are lacking.

Table 1.

Epidemiological studies for the association between dietary fat consumption and breast cancer incidence

Reference	Population	No. of cases (no. of studies)	Association	RR/HR/OR [95% CI]
Cao et al. (2016)	USA, Asia, Europe, Canada	38,262 (20)	Positive	1.10 [1.02, 1.19]
Wu, Zheng, Sun, Zou, and Ma (2015)	Chinese	4,302 (9)	Positive	1.15 [1.01, 1.30]
Liu et al. (2014)	Chinese	5,087 (6)	Positive	1.36 [1.13, 1.63]
Turner (2011)	Global	44,070 (52)	n/a (all cases)	1.011 [0.99, 1.03]
			Positive (post‐menopause)	1.043 [1.01 1.07]
			n/a (pre‐menopause)	0.973 [0.94, 1.01]
Boyd et al. (2003)	Global	25,015 (46)	Positive	1.13 [1.03, 1.25]

Abbreviations: CI, confidence interval; HR, hazard ratio; n/a, no association; No., number; OR, odds ratio; RR, relative risk. The use of bold emphasis in these tables denotes results that were reported as significant from these articles.

2.1.2. Clinical trials

Clinical trials investigating the effect of low‐fat diet interventions on BC incidence are not confirmatory of a change in BC risk. For example, a trial conducted across 40 clinics in the United States tested in healthy post‐menopausal women (n = 48,835, age 50–79) the effect of a low‐fat diet on BC incidence (Prentice et al., 2006). Women in the intervention group had intake targets of 20% of energy from total fat; 5 servings·day⁻¹ of fruits and vegetables; and 6 servings·day⁻¹ of grains. Compared with control subjects, women in the intervention group had statistically lower consumption of fat and higher consumption of fruits and vegetables. After ~8 years of follow up, there was no statistically significant reduction in the incidence rates of invasive BC between intervention (0.42%) and control (0.45%) groups (hazard ratio [HR]: 0.91; 95% CI [0.83, 1.01]). However, there was evidence for decreased occurrence of BC cases that were positive for both the https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=620 (https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=620) and the https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=627 (https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=627) (HR: 0.64; 95% CI [0.49, 0.84]). Authors reported that behavioural and lifestyle differences likely influenced the resulting incidence rates.

A clinical trial conducted in Canada investigated the effects of a low‐fat, high‐carbohydrate diet on BC incidence in a high‐risk population (Martin et al., 2011). Women (n = 4,960) with increased mammographic density were assigned to either intervention or control groups and followed on average for 10 years. Women in the intervention group were advised to decrease fat consumption to ~15% of calories and increase carbohydrate consumption to ~65%. At follow up, 118 BC cases were identified in the intervention group compared to 102 in the control group (HR: 1.19; 95% CI [0.91, 1.55]). Fat intake at baseline and after intervention were not associated with changes in BC risk overall. Interestingly, higher fat intake was associated with lower risk of ERα‐negative BC (IQOR: 0.18; 95% CI [0.05, 0.60]). The latter trend remained significant in analyses for polyunsaturated (PUFA; IQOR: 0.26; 95% CI [0.11, 0.63]) and monounsaturated (MUFA; IQOR: 0.21; 95% CI [0.07, 0.64]) FAs. No significant association was noted for saturated FAs (SFA) or in ER‐positive patients for consumption of total or specific FAs. Food records indicated that the intervention group had ~9–10% lower intake of calories from fat compared with controls. Authors reported that the intervention and control groups show similarity in consumption of carbohydrates and fat. Additionally, it is noted that the intervention group reported increased consumption of dietary fat in the final year of the study. Overall, these results suggest contrasting effects of total fat and fat type on the development of ERα‐positive versus ERα‐negative breast tumours. Whether or not different fats influence BC risk via pharmacological effects on epigenetic machinery will be discussed later.

2.2. Fat subtypes

2.2.1. Animal and saturated fat

Two meta‐analyses failed to demonstrate significant associations between BC incidence and animal fat consumption (Alexander, Morimoto, Mink, & Lowe, 2010; Cao et al., 2016). Moreover, epidemiological evidence provides an unclear association between BC risk and consumption of SFA. Boyd et al. (2003) demonstrated a positive association (RR = 1.19) when estimating risk from 46 studies. In contrast, a meta‐analysis with a larger number of studies (n = 57) failed to demonstrate a relationship between saturated fatty acid (FAC) consumption and BC incidence overall and in subgroup analyses stratified by menopause status (Turner, 2011). In line with the latter results, Cao et al. (2016) found no association between BC incidence overall and higher consumption of SFA. However, subgroup analysis revealed a statistically significant positive association with incidence of tumours positive for both https://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=96 and https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=627.

One meta‐analysis investigated the effect of SFA intake on BC risk as a primary endpoint from 28 case–control and 24 cohort studies (Xia, Ma, Wang, & Sun, 2015). Whereas case–control studies demonstrated a positive association between BC incidence and higher intake of SFA (RR = 1.18), no association was observed among cohort studies. However, when analysis was restricted to population‐based recruitment, both case–control (RR = 1.26; 95% CI [1.03, 1.53]) and cohort (RR = 1.11; 95% CI [1.01, 1.21]) studies showed a positive association. Moreover, SFA intake was reported to be positively associated with an RR increase of ~33% (RR = 1.33; 95% CI [1.08, 1.73]) among post‐menopausal women.

2.2.2. Vegetable fats

Meta‐analyses have failed to demonstrate a consistent association between BC incidence and consumption of vegetable oils. Cao et al. (2016) investigated total fat as the primary variable and derived an estimate of no association for vegetable consumption from subgroup analysis of five studies. An analysis of an even larger number of studies (n = 16) failed also to detect an association between BC risk and intake of vegetable oils (Xin, Li, Sun, Wang, & Huang, 2015). However, results from the latter study may have been confounded by the lack of controls for olive oil consumption. As discussed later in this work, there is significant evidence to suggest a protective effect of olive oil intake against BC development (OR = 0.74) (Xin et al., 2015).

2.2.3. Polyunsaturated fatty acids (PUFAs)

Epidemiological studies have investigated the association between BC incidence and consumption of total and specific PUFAs (Table 2). A meta‐analysis of 46 studies found no association for PUFAs consumption (Boyd et al., 2003), whereas another analysis, comprising a larger number of studies (n = 57), suggested a positive association between consumption of PUFAs and BC incidence (RR = 1.069; Turner, 2011). Subgroup analysis in the latter study revealed this association may be relevant for post‐menopausal (RR = 1.221), but not pre‐menopausal cases. However, this study did not differentiate between n‐6 and n‐3 PUFA when analysing risk. A more recent meta‐analysis, which did consider specific PUFA consumption, confirmed a positive association between overall PUFA intake and post‐menopausal BC (RR = 1.17); however, no statistical associations were noted for consumption of n‐3 https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1049 (18:3), https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=3362 (20:5), https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1051 (22:6), n‐6 https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1052 (https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1052, 18:2) and https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=2391 (https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=2391, 20:4; Cao et al., 2016).

Table 2.

Epidemiological studies for the association between PUFA consumption and breast cancer incidence

Reference	Variable	Population	Association	RR/HR/OR [95% CI]
Boyd et al. (2003)	PUFA	Global	n/a	0.94 [0.80, 1.10]
Turner (2011)	PUFA	Global	Positive (all cases)	1.069 [1.01, 1.14]
			Positive (post‐menopause)	1.221 [1.08, 1.38]
			n/a (pre‐menopause)	0.943 [0.81, 1.10]
Cao et al. (2016)	PUFA	USA, Asia, Europe, and Canada	n/a (all cases)	1.05 [0.96, 1.14]
			Positive (post‐menopause)	1.17 [1.00, 1.36]
			n/a (pre‐menopause)	0.98 [0.83, 1.15]
	n‐3 PUFA		n/a (all cases)	1.02 [0.89, 1.17]
	n‐6 PUFA		n/a (all cases)	1.10 [0.88, 1.38]
	EPA		n/a (all cases)	0.93 [0.79, 1.11]
	DHA		n/a (all cases)	0.94 [0.77, 1.16]
	ALA		n/a (all cases)	0.99 [0.92, 1.07]
	LA		n/a (all cases)	1.01 [0.94, 1.08]
	AA		n/a (all cases)	1.00 [0.93, 1.08]
Zheng, Hu, Zhao, Yang, and Li (2013)	Marine n‐3	USA, Europe, and Asia	Negative (all cases)	0.86 [0.78, 0.94]
			Negative (post‐menopause)	0.88 [0.76, 1.00]
			n/a (pre‐menopause)	0.96 [0.78, 1.18]
	EPA		n/a (all cases)	0.93 [0.85, 1.02]
			n/a (post‐menopause)	0.94 [0.86, 1.03]
			n/a (pre‐menopause)	0.87 [0.47, 1.60]
	DHA		n/a (all cases)	0.88 [0.75, 1.03]
			n/a (post‐menopause)	0.89 [0.75, 1.07]
			n/a (pre‐menopause)	0.66 [0.36, 1.22]
	DPA		n/a (all cases)	0.90 [0.69, 1.19]
			n/a (post‐menopause)	0.89 [0.67, 1.18]
			n/a (pre‐menopause)	1.20 [0.42, 3.45]
	ALA		n/a (all cases)	0.97 [0.90, 1.04]
			n/a (post‐menopause)	0.92 [0.82, 1.03]
			n/a (pre‐menopause)	0.74 [0.42, 1.32]
Nindrea, Aryandono, Lazuardi, and Dwiprahasto (2019)	Fish n‐3	East Asian	Negative	0.80 [0.73, 0.87]
Yang, Ren, Fu, Gao, and Li (2014)	High n‐3/n‐6 ratio	Global	Negative	0.90 [0.82, 0.99]
Zhou, Wang, Zhai, Li, and Meng (2016)	n‐6 linoleic acid	Global	n/a (all cases)	0.99 [0.92, 1.06]
			n/a (post‐menopause)	1.01 [0.94, 1.08]
			n/a (pre‐menopause)	0.64 [0.20, 2.09]

Abbreviations: AA, arachidonic acid; ALA, alpha linolenic acid; CI, confidence interval; DHA, docosahexaenoic acid; DPA, docosapentaenoic acid; EPA, eicosapentaenoic acid HR, hazard ratio; LA, linoleic acid; n/a, no association; n‐3, omega‐3; n‐6, omega‐6; OR, odds ratio; PUFA, polyunsaturated fatty acid; RR, relative risk.

Omega‐3 PUFA, which are abundant in fish, are usually considered beneficial compared to SFA from other animal food products. An analysis of 21 observational studies comprising a global population suggested a negative association for all BC cases (RR = 0.86; 95% CI [0.78, 0.94]) with higher consumption of marine n‐3 PUFA (Zheng et al., 2013). Interestingly, the results were significant for Asian, Western and U.S. subgroups, but not for women residing in European countries. Moreover, the protective effect of n‐3 PUFA was shown to be specific for post‐menopausal cases only (RR = 0.88). No associations were found between consumption of eicosapentaenoic acid, docosahexaenoic, or https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1052 and BC incidence. Studies with Eastern Asian populations confirmed the association between higher consumption of fish n‐3 PUFA and reduced BC incidence (RR = 0.80; Nindrea et al., 2019).

The ratio of n‐3 to n‐6 PUFA may influence BC risk. For example, in an analysis of 11 studies comprising a global population, Yang et al. (2014) estimated reduced BC incidence with higher ratios of n‐3/n‐6 PUFA (RR = 0.90). Women residing in the United States with a higher serum n‐3/n‐6 phospholipid ratio have a marked decrease (RR = 0.62; 95% CI [0.39, 0.97]) in BC risk. On the other hand, some meta‐analyses have not demonstrated a statistically significant association between BC incidence and consumption of the n‐6 LA (Zhou et al., 2016). Although, dietary and behavioural factors, such as exercise, may confound the influence of LA.

2.2.4. Monounsaturated fatty acids (MUFA)

Although studies investigating associations between BC incidence and DFA consumption (Table 3) have not identified a specific effect for MUFA intake (Boyd et al., 2003; Cao et al., 2016; Turner, 2011), epidemiological evidence (Xin et al., 2015) suggests a protective effect (RR = 0.74) for consumption of olive oil, which is rich (60–70%) in the MUFA https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1054 (https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1054, 18:1). Higher OO intake is inversely related to cancer incidence overall (logOR: −0.41; 95% CI [−0.53, −0.29]) and BC specifically (logOR: −0.45; 95% CI [−0.78, −0.12]; Psaltopoulou et al., 2011) in Mediterranean countries for which a Mediterranean Diet (MD) pattern has been shown to be inversely associated with BC risk. Although the MD does include considerable consumption of olive oil, it also encourages dietary changes beyond fat intake including increased consumption of plant‐based foods and limiting intake of red and processed meats and high‐fat dairy (Schwingshackl, Schwedhelm, Galbete, & Hoffmann, 2017).

Table 3.

Epidemiological studies for an association between MUFA consumption and breast cancer incidence

Reference	Variable	Association	RR/HR/OR [95% CI]
Boyd et al. (2003)	MUFA	n/a	1.11 [0.96, 1.28]
Turner (2011)	MUFA	n/a (all cases)	0.999 [0.95, 1.05]
		n/a (post‐menopause)	1.015 [0.93, 1.10]
		n/a (pre‐menopause)	0.962 [0.87, 1.06]
Cao et al. (2016)	MUFA	n/a	1.08 [0.97, 1.21]
Xin et al. (2015)	Olive oil	Negative	0.74 [0.60, 0.92]
Psaltopoulou, Kosti, Haidopoulos, Dimopoulos, and Panagiotakos (2011)	Olive oil	Negative	−0.45 [−0.78, −0.12] a
Toledo et al. (2015)b	MD + EVOO	Decreased risk	0.32 [0.13, 0.7]

Abbreviations: CI, confidence interval; EVOO, extra virgin olive oil; HR, hazard ratio; n/a, no association; MUFA, monounsaturated fatty acid; OR, odds ratio; RR, relative risk.

^aReported as logOR.

^bClinical trial.

Results of the Prevencion con Dieta Mediterranea (PREDIMED) study (Toledo et al., 2015) showed in post‐menopausal (age 67.7 ± 5.8 years) women that an MD supplemented with extra virgin olive oil (EVOO) (MD + EVOO) significantly decreased BC incidence when compared to a control low‐fat diet. In total, 4,282 women were randomized in a 1:1:1 fashion and followed for an average of 4.8 ± 1.7 years. The interventions included the MD + EVOO diet, an MD plus mixed nuts (MD + Nuts) or a low‐fat control diet. Subjects in the MD + EVOO group received 1 L·week⁻¹ of EVOO and the MD + Nuts participants received 30 g·day⁻¹ of mixed nuts (15‐g walnuts, 7.5‐g hazelnuts, and 7.5‐g almonds). Overall, there were 35 confirmed cases of BC upon follow up and the observed rate (per 1,000 person‐years) was 1.1 for MD with EVOO (HR: 0.32; 95% CI [0.13, 0.7]), 1.8 for the MD group with nuts (HR: 0.59; 95% CI [0.26, 1.35]), and 2.9 for the control group. These data point to the adoption of a MD rich in extra virgin olive oil as a strategy to lower BC risk.

3. ADIPOSITY AND BC

Normal weight is defined as body mass index (BMI) between 18.5 and 24.9 kg·m⁻²; overweight BMI between 25 and 29.9 kg·m⁻²; and obesity a BMI of ≥30 kg·m⁻². Although BMI alone is a poor index for total fat or fat distribution (Mackey, McTigue, & Kuller, 2013), a strong association between higher BMI and BC incidence has been observed (Blucher & Stadler, 2017). The combination of waist circumference (WC; Zhu et al., 2004) and high BMI is considered an independent prognostic factor for triple‐negative breast cancer (TNBC; Al Jarroudi, Abda, Seddik, Brahmi, & Afqir, 2017; Chen et al., 2016).

3.1. Pre‐menopausal versus post‐menopausal BC

In pre‐menopausal women, two meta‐analyses have identified inverse associations between BC risk and obesity (Chen et al., 2017; Munsell, Sprague, Berry, Chisholm, & Trentham‐Dietz, 2014). Stratification by hormone receptor status suggests that this inverse association is specific for ERα+/PR+ tumours (Munsell et al., 2014). On the other hand, a positive association has been shown between obesity and TNBC incidence (Pierobon & Frankenfeld, 2013). A study that analysed how central adiposity impacted BC risk suggested a positive association for WC but not for waist to hip ratio (WHR; Chen et al., 2016). Taken together, these results indicate that obesity may influence the risk of pre‐menopausal BC based on hormone receptor status. In post‐menopausal women, epidemiological studies consistently report a positive association between BC and obesity (Chen et al., 2016; Chen et al., 2017; Munsell et al., 2014; Xia et al., 2014). An overweight (>25 kg·m⁻²) and obese (>30 kg·m⁻²) BMI are positively associated with BC incidence (Xia et al., 2014). This positive association appears to be specific for ERα+/PR+ BC, but nor for ERα−/PR− breast tumours (Munsell et al., 2014). Consistent with the latter findings, a meta‐analysis found no association between obesity and risk of TNBC in post‐menopausal women (Pierobon & Frankenfeld, 2013).

3.2. BC survival

Obesity is associated with worse overall survival (OS), but not with BC‐specific survival (Chan et al., 2014; Mei et al., 2018; Niraula, Ocana, Ennis, & Goodwin, 2012; Playdon et al., 2015). Weight gain after diagnosis increases the risk of mortality; however, this association is not seen when analysing BC‐specific survival (Playdon et al., 2015). No associations between overall and TNBC specific survival, and obesity have been found (Mei et al., 2018). A meta‐analysis that stratified BC patients by menopausal and hormone receptor status confirmed a positive association between obesity and BC mortality (Niraula et al., 2012). A similar association has been documented between obesity and overall mortality in both pre‐ and post‐menopausal women (Chan et al., 2014). How interactions between obesity and hormone receptor status impact BC development and response to treatment remains unclear.

4. DNA METHYLATION BIOMARKERS OF BC

4.1. Writers of cytosine‐guanine dinucleotides (CpG) methylation

We have summarized clinical and molecular features associated with aberrant regulation of CpG methylation (global and gene specific) and modifiers of DNA methylation in Figure 1. DNA methylation is catalysed by DNA methyltransferases (DNMT), which transfer methyl groups from the cofactor https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=4786 to cytosine substrates in CpG (Jurkowska, Jurkowski, & Jeltsch, 2011). The https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2605 protein catalyses methylation of newly synthesized daughter strands of DNA and is responsible for the “maintenance” of the DNA CpG methylation pattern throughout cell division. On the other hand, https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2750 and DNMT3B are regarded as “de novo” methyltransferases, which are responsible for setting the initial methylation pattern and adding new CpG methylation marks under stimulating perturbations (Jurkowska et al., 2011).

Figure 1.

Clinical and molecular features associated with aberrant DNA methylation. (a) Inside the cell shows molecular tumour features. Up‐regulation of methylation modifiers is associated with ER and PR negativity and BRCA1 methylation. For gene‐specific methylation, tumours with BRCA1 methylation (BRCA1 ^METHYL) are usually TNBC phenotype; TIMP3 ^METHYL have ER, PR and HER2 overexpression; and RASSF1A ^METHYL tumours are more often non‐TNBC and associated with recurrence and relapse risk. (b) Clinical features including propensity for recurrence and relapse, high grade, large tumour size, high stage, propensity for lymph node (LN) metastasis and decreased survival are listed outside the cell with corresponding DNA methylation biomarkers. Filled black circles in the gene‐specific methylation section of the nucleus indicate methylated CpG sites. Open circles in the global methylation section indicate non‐methylated CpG sites

A study (Girault, Tozlu, Lidereau, & Bieche, 2003) with 130 cases of sporadic BC demonstrated that DNMT3B mRNA was overexpressed in 30% of BC samples, whereas DNMT1 and 3A mRNA were overexpressed in only 5.4% and 3.1% of tumours, respectively. The same study demonstrated that overexpression of DNMT3B was significantly associated with histological grade III, ERα negativity and expression of markers of proliferation (i.e., Ki‐67). No link was established between DNMT3B overexpression and age, menopausal status, lymph node involvement, tumour size or https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=621 (https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=621) status. When investigating the prognostic value in patients who received adjuvant therapy, DNMT3B overexpression was associated with a significantly shorter relapse‐free survival in patients that received hormonal therapy, however there was no such association for chemotherapy‐based treatment.

Investigations of the expression of DNMT1, 3A protein and 3B protein by immunohistochemistry (IHC) in breast tumours (n = 256) demonstrated that both DNMT1 and DNMT3A levels were increased compared with non‐malignant fibroadenoma samples (Yu et al., 2015). However, no significant differences in DNMT3B immunoreactivity were noted. Levels of DNMT1 and DNMT3A were associated with increased lymph node metastasis, decreased expression of ERα and BRCA1 proteins and hypermethylation of the ERα (ESR1) and BRCA1 genes. In patients with ER negative tumours, levels of both DNMT1 and DNMT3A were associated with shorter OS and disease‐free survival (DFS). Overall, a higher level of DNMT3A and DNMT3B, but not DNMT1, was associated with increased tumour node metastasis (TNM) stage. Increased expression of DNMT3A, but not DNMT3B or DNMT1, was associated with shorter OS. Elevated DNMT1 was also associated with shorter DFS in patients younger than 50 years of age and shorter OS and DFS in patients who received a combination of chemotherapy and endocrine therapy.

Another study utilizing an IHC‐based approach investigated associations with levels of DNMT1 protein in sporadic BC based on surrogate molecular subtype, clinicopathological features and stromal histology (Shin, Lee, & Koo, 2016). Cases (n = 348) were categorized as luminal A (42.8%), luminal B (23.7%), https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2019 (https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2019)‐enriched (9%) or TNBC (24%) based on IHC staining for ERα, PR, HER‐2 and Ki‐67. Authors reported that the frequency of DNMT1 immunoreactivity in tumour cells differed by molecular subtype, with TNBC being the most common (28%), followed by HER2‐enriched (11.1%), luminal B (8.3%), and luminal A (0.6%). DNMT1 expression in malignant cells was higher in tumours that had inflammatory type stromal cells and lowest in tumours with the sclerotic phenotype. Moreover, DNMT1 expression in tumour cells was associated with higher histological grade, ERα negativity, PR negativity, and higher Ki‐67.

4.2. Global CpG methylation

In cancer, alterations in CpG methylation patterns are characterized by global DNA hypomethylation and gene‐specific hypermethylation. Global hypomethylation contributes to cancer development and progression through activation of oncogenes (Van Tongelen, Loriot, & De Smet, 2017) and chromosomal instability (Gaudet et al., 2003). Changes in global CpG methylation can be monitored by measuring the methylation of repetitive elements in DNA (e.g., LINE‐1 and Alu) or the per cent of methylated DNA by luminometric methylation assay (LUMA), using mean methylation intensities of Infinium HumanMethylation 450 k probes (β‐value), or quantifying the concentration of 5‐methyldeoxycytosine (5‐mdC; Tang, Cheng, Cao, Surowy, & Burwinkel, 2016). Using the β‐value approach, a study reported an inverse association between BC incidence and global CpG methylation (OR = 0.42; 95% CI [0.20, 0.90]) in peripheral blood of BC patients (n = 420; Severi et al., 2014).

The usefulness of LUMA‐based approaches to predict BC incidence remains unclear. A study with a population of Japanese women reported global CpG methylation (measured by LUMA) in peripheral blood leukocyte DNA was decreased in BC patients compared with healthy controls (Kuchiba et al., 2014). Compared with subjects in the highest tertile of global CpG methylation, the ORs for BC incidence of women in the second and lowest tertile were 1.87 (95% CI [1.20, 2.91]) and 2.86 (95% CI [1.85, 4.44]), respectively. In contrast, Xu et al. (2012) reported that women in the highest quartile of global CpG methylation (measured by LUMA) had a 2.41‐fold increased risk of BC compared with women in the lowest quartile.

A study investigating global methylation in peripheral blood leukocytes from BC patients (n = 176) demonstrated a significant difference in 5‐mdC content, but not LINE‐1 methylation, compared with control (n = 173) samples (Choi et al., 2009). Women in the second (OR = 1.49; 95% CI [0.84, 2.65]) and lowest (OR = 2.86; 95% CI [1.65, 4.94]) tertile of 5‐mdC content had increased risk of BC compared to women in the highest tertile. Interestingly, there was no correlation between 5‐mdC content and LINE‐1 methylation. Furthermore, no associations have been found between BC incidence and peripheral blood LINE‐1 methylation (Brennan et al., 2012; Cho et al., 2010; Choi et al., 2009; Deroo et al., 2014; Kitkumthorn, Tuangsintanakul, Rattanatanyong, Tiwawech, & Mutirangura, 2012; Wu et al., 2012; Xu, Gammon, et al., 2012) or CpG methylation of Alu repeat elements (Cho et al., 2010; Wu et al., 2012).

Changes in global methylation may also be prognostic of BC development. A population‐based study that investigated the association between obesity, DNA methylation and BC mortality demonstrated lower LUMA values associated with increased all‐cause (HR = 1.81; 95% CI [1.19, 2.74]) and BC‐specific (HR = 2.61; 95% CI [1.45, 4.69]) mortality in obese patients (McCullough et al., 2016). Lower 5‐mdC content was also found to be significantly associated with poor differentiation (P < .001) and disease‐specific survival (HR = 1.59; 95% CI [0.99, 2.55]; P = .056) and DFS (HR = 1.54; 95% CI [0.93, 2.54]; P = .093; Tsai et al., 2015). Global DNA hypomethylation has also been shown to correlate with disease stage, tumour size, and histological grade in BC (Soares et al., 1999).

4.3. Gene‐specific CpG methylation

Hypermethylation at tumour suppressor loci is a common mechanism for somatic gene inactivation in cancer. Examples of hypermethylation at genes in BC include those encoding the proteins BRCA1 (Zhang & Long, 2015), https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?tab=biology&ligandId=5311 (TIMP3; Maleva Kostovska et al., 2018), Ras association domain family 1 isoform A (RASSF1A; Yadav et al., 2018) and https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=591 (RARβ; Xu, 2007) among several others. When discussing gene‐specific CpG methylation, it is important to consider that genes affected by methylation can harbour multiple CpG sites and methylation at a certain CpG does not necessarily cause transcriptional silencing. Moreover, studies investigating CpG methylation may analyse different CpG sites within the same gene.

The BRCA1 gene encodes a 220‐kDa nuclear phosphoprotein that functions as a tumour suppressor through maintenance of genomic stability via the DNA homologous recombination pathway and its involvement in cell cycle control, transcriptional regulation, apoptosis and mRNA splicing (Savage & Harkin, 2015). Women who inherit germline mutations in BRCA1 have ~72% lifetime risk of developing BC (Kuchenbaecker et al., 2017), the majority of which (~70%) are TNBC (Silver et al., 2010). Although somatic mutations in BRCA1 are rare, epigenetic inactivation of BRCA1 by DNA methylation is frequently observed (~25% of unselected cases; Xu et al., 2013) and most commonly associated with TNBC (Yamashita et al., 2015). Studies have demonstrated that BRCA1 promoter methylation in peripheral blood cells is associated with increased risk of breast tumours that have BRCA1 promoter methylation (Iwamoto, Yamamoto, Taguchi, Tamaki, & Noguchi, 2011). Due to this observation, various studies have assessed BRCA1 hypermethylation in breast tissue and blood samples (peripheral and cell‐free) and its association with BC incidence, prognosis and clinicopathological characteristics.

A meta‐analysis (n = 40 studies) that analysed BRCA1 methylation in both blood and tumour samples demonstrated that hypermethylation of BRCA1 was associated with an increased risk of BC (OR = 3.15; Zhang & Long, 2015). In subgroup analyses, BRCA1 methylation in tissue was even more predictive of BC risk (OR = 4.75) whereas methylation in peripheral blood was less predictive, albeit significant (OR = 1.87). BRCA1 hypermethylation was associated with lymph node metastasis (OR = 1.25), higher histological grade (OR = 2.29), ERα negativity (OR = 2.36), PR negativity (OR = 2.14), and TNBC phenotype (OR = 2.79). BRCA1 methylation is significantly associated with worse OS (OR = 1.38) and DFS (OR = 3.92; Wu et al., 2013). Studies have also shown that BRCA1 methylation modulates the response to therapy. For example, patients with tumours that had methylated BRCA1 were shown to be more sensitive to adjuvant chemotherapy than patients with non‐methylated BRCA1 tumours (Xu et al., 2013). Hypermethylated BRCA1 was associated with improved DFS in patients that received anthracycline‐based chemotherapy (Ignatov et al., 2013). Moreover, TNBC patients with hypermethylated BRCA1 showed better responses to https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=7624‐based neoadjuvant chemotherapy (NACT; Silver et al., 2010).

The TIMP3 gene encodes a 22‐kDa protein (TIMP3) that acts as a tumour suppressor due to its capacity to inhibit the activity of MMPs (Jackson, Defamie, Waterhouse, & Khokha, 2017), which have several described oncogenic functions including promoting tumour growth, angiogenesis, invasion and metastasis (Radisky & Radisky, 2015). The TIMP3 gene was found to be hypermethylated in ~21% of invasive BC samples (n = 173; Lui, Loo, Zhu, Cheung, & Chow, 2005). TIMP3 also emerged as one of the 10 hypermethylated genes from a panel of 22 tumour suppressor genes in BC tissue (n = 48) compared to matched normal control tissue (n = 48; Radpour et al., 2009). Methylation levels of TIMP3 are higher in cell‐free serum DNA (Radpour et al., 2011) and peripheral blood (Zmetakova et al., 2013) of BC patients compared with healthy controls. Studies that analysed the promoter methylation status of 22 tumour suppressor genes using tumour biopsies that contained invasive tissue concurrently with benign or in situ lesions and normal epithelium demonstrated that hypermethylation of TIMP3, among other genes, was an early event in mammary tumourigenesis (Chen, Stephen, Raju, & Worsham, 2011). This collective evidence indicates TIMP3 CpG methylation may be a predictive biomarker of BC risk. Moreover, a study in invasive ductal carcinoma patient samples (n = 173) suggested that TIMP3 methylation was associated with high tumour grade, lymph node metastasis and overexpression of ERα, PR and human EGF receptor‐2 (HER2; Lui et al., 2005).

The RASSF1 gene encodes for seven protein isoforms designated RASSF1A‐G, which are generated as a result of alternative splicing and transcription initiation from multiple promoters (Hesson, Cooper, & Latif, 2007). The RASSF1A promoter harbours a CpG island and is one of the most frequently hypermethylated genes in cancer. Tumour suppressive functions of RASSF1A have been linked to its involvement in regulation of apoptosis, cell cycle control and microtubule dynamics. In T47D BC cells, overexpression of RASSF1A alone significantly decreased cellular proliferation (Reeves, Firek, Chen, & Amaar, 2013). Moreover, RASSF1A overexpression was shown to enhance the pro‐proliferative and anti‐apoptotic effect of https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=1877 in these cells. Hypermethylation of RASSF1A has been reported in ~57–90% of primary lesions (Buhmeida et al., 2011; Hagrass, Pasha, Shaheen, Abdel Bary, & Kassem, 2014; Jezkova et al., 2016; Jezkova et al., 2017; Kajabova et al., 2013; Karray‐Chouayekh et al., 2010; Kioulafa, Kaklamanis, Mavroudis, Georgoulias, & Lianidou, 2009; Li, Wei, Cao, & Cao, 2008) and ~17–64% of serum (Hagrass et al., 2014; Kajabova et al., 2013) and 47% of peripheral blood mononuclear cell (Yadav et al., 2018) samples from BC patients. Interestingly, RASSF1A hypermethylation was detected in ~67% of hereditary BC cases in one study (Alvarez et al., 2013). In these hereditary cases, RASSF1A methylation was found in both ductal carcinoma in situ (DCIS) and grade I infiltrating ductal carcinoma, suggesting it may be a marker for pre‐malignant phenotypes. Studies investigating CpG methylation in malignant breast epithelial cell specimens obtained by fine‐needle aspiration biopsy from BC patients (n = 164) demonstrated that RASSF1A hypermethylation was associated with a significant increase in BC risk (OR = 5.28; 95% CI [1.95, 14.32]; Euhus et al., 2008).

Several studies have documented an association between RASSF1A hypermethylation and worse DFS and OS in BC (Buhmeida et al., 2011; Kioulafa et al., 2009; Martins et al., 2011; Xu et al., 2012). A meta‐analysis by Jiang, Cui, Chen, Shen, and Ding (2012) that investigated the link between RASSF1A hypermethylation and DFS and OS among 1,795 patients from eight studies showed an HR of 2.54 (95% CI [1.96, 3.66]) for DFS, and among 1,439 patients from five studies an HR of 3.47 (95% CI [1.44, 8.34]) for OS. Higher methylation of RASSF1A has also been linked to a worse time to first recurrence (HR = 1.93; Xu, Shetty, et al., 2012) and higher incidence of relapse (Kioulafa et al., 2009). Moreover, multivariate Cox regression analyses have demonstrated RASSF1A hypermethylation in tumour tissues to be an independent predictor of poor prognosis (HR for DFS = 5.64; 95% CI [1.23, 25.81]; Buhmeida et al., 2011). In terms of tumour characteristics, RASSF1A hypermethylation has been associated with stage (Hagrass et al., 2014; Jezkova et al., 2016; Karray‐Chouayekh et al., 2010), size (Kajabova et al., 2013), grade, invasive ductal carcinoma (Jezkova et al., 2016), invasive BC, lymph node metastasis (Hagrass et al., 2014), and non‐TNBC subtypes (Kajabova et al., 2013; Wang et al., 2012).

Studies have also demonstrated a positive correlation between RASSF1A hypermethylation and expression of both ERα and PR (Kajabova et al., 2013). Hypermethylation of RASSF1A has also been shown to modulate the response to chemotherapy. For example, patients that did not have a response to https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=6809‐based chemotherapy had a higher RASSF1A methylation compared with patients that achieved partial or complete pathological responses (Gil et al., 2012). in vitro experiments demonstrated an enhancing effect of RASSF1A on docetaxel‐induced cell cycle arrest. In another study that investigated the relationship between RASSF1A methylation and response to NACT, RASSF1A methylation was found to be higher in both tumour tissue and serum of patients that did not have an effective response compared to those that did (Han, Xu, Han, Wang, & Lin, 2017). However, treatment may influence RASSF1A methylation status, as it was reported that NACT‐treated tumours had a reduced RASSF1A methylation compared with NACT‐naïve tumours (Wang et al., 2012).

The RARβ gene encodes a member of the thyroid–steroid hormone receptor superfamily of nuclear transcriptional regulators involved in mediating cellular effects of retinoic acid. Studies have demonstrated that RARβ is an inhibitor of BC cell proliferation (Hayashi et al., 2003) and migration (Flamini et al., 2014). Re‐expression of RARβ in deficient BC cells has also been shown to significantly attenuate in vitro and in vivo growth (Sirchia et al., 2002). Hypermethylation of RARβ2 has been reported in ~37–66% of BC samples, depending on the population studied (Karray‐Chouayekh et al., 2010; Pirouzpanah, Taleban, Atri, Abadi, & Mehdipour, 2010). Meta‐analyses have demonstrated a significant association between RARβ2 hypermethylation and BC incidence (OR = 7.21), risk of lymph node metastasis (OR  =  2.13) and TNM stages III–IV (OR  =  1.85; Qi & Xiong, 2018). Studies in BC patients of Indian (Mirza et al., 2012) and African‐ and European‐American (Wang et al., 2012) ethnicity have demonstrated that RARβ2 hypermethylation is associated with worse OS. Moreover, hypermethylated RARβ2 was associated with younger age at diagnosis, cases with a negative family history of BC (Pirouzpanah et al., 2010), and lymph node metastasis (Marzese et al., 2012). The frequency of RARβ2 hypermethylation was also significantly higher in the bone, lung, and brain metastases compared with primary tumours, suggesting its potential utility as a predictive biomarker of distant metastasis (Mehrotra et al., 2004).

The ERα protein is encoded by the ESR1 gene. Given the putative role of oestrogen signalling in driving hormone‐sensitive BCs, ESR1 is not considered a tumour suppressor gene. However, the ERα status of breast tumours is a useful biomarker for the determination of prognosis and pharmacological approach. Compared to tumours that express ERα, ERα‐negative cases tend to be more aggressive, confer a worse prognosis, and have refractory responses to anti‐endocrine therapy (Li, Chen, Hardy, & Tollefsbol, 2013). The hypermethylation of ESR1 has been reported in ~40–85% of BC cases unselected for subtype or histological profile (Parrella et al., 2004; Pirouzpanah, Taleban, Mehdipour, Sabour, & Atri, 2018; Ramos et al., 2010) and may represent an early event in mammary tumourigenesis (Chen et al., 2011). Perhaps unsurprisingly, ESR1 methylation in breast tissue (Pirouzpanah et al., 2018) and serum samples (Martinez‐Galan et al., 2008; Martinez‐Galan et al., 2014) is associated with the ERα‐negative tumour phenotype. An analysis of ESR1 methylation in 136 primary breast tumours found hypermethylation in 51.5% of all samples; 83% of ERα‐negative cases, which was significantly greater than ERα‐positive; 67.9% of PR‐negative, which was significantly greater than PR‐positive; and 81.2% of ERα‐negative/PR‐negative cases, which was more than both ERα‐positive/ERα‐positive and ERα‐positive/PR‐negative (Pirouzpanah et al., 2018). A study testing ESR1 methylation in cell‐free DNA from non‐metastatic BC patients (n = 110) reported hypermethylation in 80% of TNBC, 60% of HER2‐enriched, 28% of luminal A and 36% of luminal B cases (Martinez‐Galan et al., 2014). The hypermethylation of ESR1 in circulating free DNA has also been associated with a lack of response to https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=7073 treatment (Mastoraki et al., 2018). In BC cells with hypermethylated ESR1, epigenetic reactivation of ERα imparts sensitivity to https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1016 in vitro (Li, Chen, et al., 2013). Moreover, a meta‐analysis of three studies comprising 227 patients demonstrated ESR1 promoter methylation was linked to a worse prognosis in terms of OS (HR = 1.55; 95% CI [1.06, 2.28]; Sheng, Guo, & Lu, 2017).

5. DIETARY FAT, OBESITY AND TARGETING OF DNA METHYLATION IN BC

A growing body of scientific evidence supports a modulating effect of both obesity and dietary fat consumption on the BC methylome. These interactions have been summarized in Figure 2. A population‐based study (McCullough et al., 2015b) analysing global DNA methylation in white blood cells (WBC) from post‐menopausal BC patients reported that non‐obese women in the highest quartile of LUMA score had an increased risk of BC (OR = 2.16; 95% CI [1.35, 3.57]), but no specific effect was found in obese patients. BC‐specific mortality was also shown to be higher in obese patients with decreased LUMA levels (HR = 2.61; 95% CI [1.45, 4.69]) and hypermethylation of adenomatous polyposis Coli (APC) (HR = 2.47; 95% CI [1.43, 4.27]) and TWIST1 (HR = 4.25; 95% CI [1.43, 12.70]) in DNA of WBC samples (McCullough et al., 2016). Lower LUMA levels in WBC of obese patients were associated with increased all‐cause mortality (HR = 1.81; 95% CI [1.19, 2.74]).

Figure 2.

Summary of how high‐fat diet/obesity influences changes in (1) expression of DNA methylation modifiers, (2) global CpG methylation, and (3) gene‐specific CpG methylation. For example, a diet high‐corn oil increased expression of methylation modifiers; a HFD was shown to increase global CpG methylation; and a supplementation with n‐3 FAs increased gene‐specific methylation. Closed arrowheads indicate stimulus whereas blunted lines indicate inhibition/suppression. In instances where there is both stimulus and suppression depicted for a given interaction (a–d), there is evidence in the literature to support both outcomes. (a) There is a positive association between global methylation and body fat percentage, and overweight mice have 511 CpG sites hypermethylated and 248 CpG sites hypomethylated, compared to mice with normal body weight. (b) Pregnant rats fed a HFD have increased Dnmt1 expression in the mammary glands. However, in utero exposure to a high butterfat diet represses tumour expression of Mbd2, Mbd3, and Dnmt3a in offspring. (c) Lactating mice fed a HFD produce male offspring with decrease Scd1 methylation in epidydimal white adipose tissue. RARβ is hypermethylated in clinical subjects that adhere to a HFD. (d) Oleic acid, the predominant FA in olive oil (OO), dose‐dependently decreases global methylation in vitro. Conversely, a diet rich in OO increases global methylation in rats

A study investigating global CpG methylation using an Illumina‐based approach in (n = 345) breast tumour samples from the Carolina BC Study demonstrated in 87% of probes analysed an increase in β‐value with increased obesity (BMI of ≥30 kg·m⁻²; Hair et al., 2015). Upon stratification by ERα status, differential methylation of 21 gene loci in ERα‐positive cases and genes modulated by obesity were reported to be involved in immune response, cell growth, and DNA repair. Interestingly, DNMT3B showed the most significant difference in β‐value between obese and normal weight patients. A small study (n = 24 patients) analysed associations between lifestyle modifications and global CpG methylation in peripheral blood from Hispanic, African American, and Afro‐Caribbean overweight and sedentary female BC survivors (Delgado‐Cruzata et al., 2015). Changes in body fat percentage were positively associated with LINE‐1 methylation values. In contrast to obesity, increased exercise was associated with a decreased risk of BC, and meta‐analyses have suggested that this may be due to a positive association between global CpG methylation and physical activity (Boyne et al., 2018)

A methylation study of RARβ, E‐cadherin 1 (ECAD) and P16 in (n = 803) breast tumour blocks from the Western New York Exposure and BC Study found that a greater WHR was associated with an increased probability (OR = 1.85) of hypermethylation of one or more of these genes (Tao et al., 2011). An analysis of non‐malignant breast tissue from healthy women (n = 120) who had undergone cosmetic mammoplasty reported hypermethylation of ESR1 in obese women (BMI of ≥30 kg·m⁻²) compared to both overweight (BMI of ~25–29 kg·m⁻²) and normal weight (BMI of <25 kg·m⁻²; Daraei et al., 2017). Multivariate linear regression models demonstrated a positive association between ESR1 methylation and increasing BMI. A positive correlation has also been established between BMI and ESR1 hypermethylation in primary breast tumours (n = 137; Pirouzpanah et al., 2010). Increased BMI has also shown a positive association with hypermethylation of both RASSF1A and BRCA1 in surgically excised breast tumours (n = 120; Naushad et al., 2014). Studies have also identified that obesity was associated with hypermethylation of hairpin‐induced 1 (HIN1) in post‐menopausal breast tumour tissue (n = 532) from the Long Island BC Study Project (McCullough et al., 2015a). The HIN1 gene encodes a protein that is a negative regulator of both cell growth and invasion (Krop, Parker, Qimron, Porter, & Polyak, 2005).

Analysis of DNA methylation changes in serum is gaining interest as a potential non‐invasive biomarker for screening and early detection of BC (Cheuk, Shin, & Kwong, 2017). Both clinical and preclinical evidence suggest a modifying effect of dietary fats on DNA methylation in haematological samples. An in vitro study revealed differential effects of FAs on the methylome using THP‐1 monocytes (Silva‐Martinez et al., 2016). Through quantification of 5‐mdC levels, studies reported that AA dose‐dependently increased, while oleic acid dose‐dependently decreased global CpG methylation. These methylation changes were predominately in gene bodies and corresponded to changes in gene expression. A clinical intervention trial also demonstrated that supplementation for 8 days with n‐3 PUFA (5.7 g·day⁻¹) and EVOO (6 g·days⁻¹) caused differential effects on the methylation of the gene encoding https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=4998 (IL‐6) and expression of DNMT1 in circulating leukocytes (Hunter et al., 2019). Specifically, methylation of IL‐6 was increased following n‐3 PUFA supplementation, whereas EVOO caused decreased methylation. Moreover, compared to n‐3 PUFA, supplementation with EVOO significantly decreased expression of DNMT1.

Earlier in this review, we described results from the PREDIMED study, a randomized control trial that demonstrated a protective effect of an MD supplemented with EVOO against the development of BC in high‐risk women (Toledo et al., 2015). Arpon et al. (2017) demonstrated that both the MD + EVOO and MD + Nuts diets induced significant DNA methylation changes in peripheral WBC (PWBC). Ingenuity pathway analysis revealed that genes with differentially methylated CpGs were related to factors involved in metabolism, diabetes, inflammation and signal transduction. Interestingly, the methylation status of two CpG sites (cg01081346 [CPT1B] and cg17071192 [GNAS]) was modulated based on dietary exposure. The cg01081346 CpG showed an increase in MD + Nuts specimens, whereas the methylation of cg17071192 was decreased by MD + EVOO. Changes in cg01081346 methylation were linked to n‐6 PUFA consumption. Furthermore, adherence to the MD correlated with methylation changes in https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2756 (EEF2), collagen type XVIII α1 chain (COL18A1), IL4I1, leptin receptor (LEPR), plag 1 zink finger 1 (PLAGL1), IFN‐related developmental regulator 1 (IFRD1), https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2063 (MAPKAPK2), and peroxisome‐activated proliferator receptor γ coactivator 1β (PPARGC1B; Arpon et al., 2017). A positive correlation was also shown between EEF2 methylation in PWBC and serum levels of https://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=906 and TNF‐α, which are associated with an increased risk of BC (Guo et al., 2015) and poor prognosis (Berberoglu, Yildirim, & Celen, 2004; Ma et al., 2017), respectively.

Hypermethylation of RARβ has been reported as more prevalent (45%) in subjects reporting adherence to an HFD compared with those consuming less fat (23%, P = .007; Brait et al., 2009). Interestingly, the short‐chain FA https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1059 (butyrate) has been shown to demethylate RARβ and up‐regulate its expression in colon cancer cells (Spurling et al., 2008). Although butyrate has demonstrated anti‐tumourigenic effects in BC cells (Salimi et al., 2017), its specific effect on methylation of RARβ in BC is unknown.

An investigation into potential differences between fat sources utilized the rat 7,12‐dimethylbenz[a]anthracene (DMBA)‐induced BC model to investigate the effect of diets high in corn (20%) or OO (17%) on DNA methylation in normal mammary glands and mammary tumours (Rodriguez‐Miguel et al., 2015). From weaning, rats were administered either a low‐fat (LF), high‐corn oil (HCO) or high‐OO (HOO) diet. After carcinogen administration, a subset of rats on the LF diet were switched to the HCO (LF‐HCO) or HOO (LF‐HOO) diets. Rats on the HOO diet had higher global DNA methylation in both mammary glands and tumours compared to those on the HCO and LF diets. Rats on both the HCO and LF‐HCO diets had higher methylation of Rassf1A and Timp3 in both mammary glands and tumours compared to rats on the LF diet, whereas Timp3 methylation was decreased in tumours of LF‐HOO rats. In normal mammary glands, HCO rats had increased expression of Dnmt1 and HOO rats had decreased expression of Dnmt3b compared with LF and HCO animals. Moreover, DNMT activity was shown to be higher in both tissue types of HCO‐fed rats.

Experiments in mouse mammary tumour virus‐neu mice demonstrated that both overweight‐ and obesity‐inducing diets enhanced the development of spontaneous mammary tumours and increased mammary gland expression of Dnmt1 (Rossi et al., 2017). In contrast, calorie‐restricted mice had stable mammary gland expression of Dnmt1 over time and had improved tumour‐related mortality. Analysis of global methylation showed that overweight mice (obese mice were not assessed) had 511 CpG sites with increased methylation and 248 CpG sites with decreased methylation compared to calorie‐restricted mice.

DNA methylation changes have been assessed in tumour‐distal mammary fat pads from orthotopic xenograft tumours from normal weight, obese (Ob), and formerly Ob (FOb) mice (Rossi et al., 2016). Analysis of global CpG methylation demonstrated that Ob tumours had 39 genes significantly hypermethylated compared to normal weight controls and normalization of weight in FOb mice did not reverse epigenetic reprogramming. Interestingly, FOb mice had 11 genes that were hypermethylated compared to controls. In Ob mice, hypermethylation was observed at TSC complex subunit 22 domain family member 3 (TSC22D3) and https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2049 loci, which were shown by ingenuity pathway analysis to regulate https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2654 (https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2654), a histone methyltransferase protein and key factor in epigenetic gene regulation. Analysis of EZH2 mRNA levels in mammary fat pad tissue showed higher expression in Ob and FOb mice compared to normal weight controls.

It has been hypothesized that adult disease susceptibility is influenced by perturbations in the gestational environment during embryonic and fetal development, which are mediated in part by epigenetic changes. For example, in Japanese macaques, it was demonstrated that maternal feeding of a high‐fat diet (35% of kcal from fat) resulted in a significant increase in fetal (gestational day 130) hepatic expression of DNMT1 (Aagaard‐Tillery et al., 2008). In rats, maternal intake of a high‐fat diet (43% of kcal from corn oil) during pregnancy caused a significant increase in the susceptibility of F1 offspring to DMBA‐induced mammary tumours, which was associated with increased expression of Dnmt1 in offspring mammary glands (de Assis et al., 2012). Studies utilizing the DMBA mammary tumour model in mice have also demonstrated a transgenerational (in F1–F3) enhancing effect of gestational exposure to a maternal high‐fat diet (38.9% of kcal from corn oil) on mammary tumourigenesis, although whether this is related to aberrant expression of Dnmt genes was not specifically addressed (Nguyen et al., 2017).

Methyl‐CpG binding domain proteins (MBD) recognize and bind CpG methylated DNA and facilitate its interactions with other histone and chromatin modifying complexes (Wood & Zhou, 2016). Studies investigating the effect of in utero exposure to a high‐fat diet (39%) on rat offspring susceptibility to DMBA‐induced mammary tumours have demonstrated different sources of fat exerted differential effects on the expression of Mbd2, Mbd3 and Dnmt3a (Govindarajah et al., 2016). All high‐fat diets, regardless of whether they were enriched with butter fat, olive oil and safflower oil, significantly increased offspring mammary tumour incidence compared with animals on a standard chow diet. Interestingly, compared to controls, the butter fat diet significantly repressed tumour expression of Mbd2, Mbd3,and Dnmt3a, whereas the OO and safflower oil diets had no effect on the epigenetic machinery.

The MBD_V2 alternative mRNA splicing variant of MBD2 has been shown to be a crucial factor in the regulation of cancer stem cell‐like cell maintenance and expansion in TNBC cell cultures, and its expression has been shown to be driven by ROS (Bao et al., 2017). Since obesity is associated with both ROS production and general inflammation, a study tested the effect of diet‐induced obesity (60% of kcal from fat) on TNBC xenograft development and MBD_V2 expression in mice (Teslow et al., 2019). Results showed that obesity was a driver of both tumourigenicity and tumour MBD_V2 expression. Moreover, analysis of TNBC data sets demonstrated a positive association between MBD_V2 expression and high BMI and relapse rates.

The mammary gland is composed of epithelial tissue organized into lobules and ducts that produce and excrete milk, adipose tissue, endothelial cells, fibroblasts and immune cells. Although the majority of invasive breast carcinomas develop in the ductal (80%) component of the mammary epithelial tree (Makki, 2015), a significant proportion of the breast is adipose tissue (35–56% depending on lactation status). Breast adipocytes may facilitate mammary tumourigenesis through mechanisms including, but not limited to, paracrine signalling of adipokines and inflammatory factors (e.g. TNF‐α and IL‐6), metabolic reprogramming, and remodelling of the extracellular matrix. For example, higher levels of TNF‐α are associated with poor BC prognosis (Berberoglu et al., 2004; Ma et al., 2017). A study in Sprague–Dawley rats demonstrated that TNFA promoter methylation was decreased in adipocytes of animals administered a diet with coconut oil as the principle fat source (10% of kcal from fat) compared to diets with isogenic amounts of fat from olive and sunflower oils (Garcia‐Escobar et al., 2017). Stromal stearoyl‐CoA desaturase 1 (SCD1) has been shown to drive mammary tumourigenicity in preclinical models (Angelucci et al., 2015). In male mice, maternal intake of a high‐fat diet during breastfeeding has been shown to increase Scd1 expression via CpG demethylation in epidydimal white adipose tissue in adulthood (Butruille et al., 2019). However, the relevance of SCD1 up‐regulation to BC development has not been clearly elucidated and should be the subject of future research.

A trial in healthy subjects investigated the effect of 7‐week interventions with diets excessive in SFA or PUFA on subcutaneous adipose tissue (Perfilyev et al., 2017). Although both diets increased global methylation in adipose tissue, the SFA affected methylation of 1,797 genes, whereas the PUFA diet only influenced 125 genes. Transcriptional analysis revealed aberrant expression of 28 genes induced by the SFA diet, whereas the PUFA intervention had no significant effects on gene expression. This study suggested that different FAs have differential effects on adipose tissue epigenetic reprogramming. These studies suggest that more research is needed regarding the influence of dietary fat on breast adipocyte epigenetics and the effect on mammary tumourigenesis.

6. DISCUSSION

The purpose of this work was to highlight research evidence regarding modulating effects of dietary fat and obesity on BC risk and epigenetic biomarkers. We first reviewed epidemiological and clinical studies that investigated the association between BC incidence and the consumption of dietary fat overall and specific FAs (e.g. PUFA and MUFA). Additionally, we discussed epidemiological evidence for the association between obesity and BC risk and survival. We highlighted a candidate list of DNA methylation changes in writers of CpG methylation (i.e. DNMT), global CpG methylation and gene‐specific CpG methylation and how they have been related to BC risk, prognosis, survival outcomes and responses to therapy. Finally, we highlighted results from preclinical, clinical and population‐based studies suggesting a modulating effect of obesity and dietary fat on DNA methylation biomarkers.

In general, epidemiological evidence suggests a positive association between BC incidence and higher consumption of dietary fat and specifically for post‐menopausal BC (Boyd et al., 2003; Cao et al., 2016; Liu et al., 2014; Turner, 2011; Wu et al., 2015). Surprisingly, none of the studies discussed in this review reported a clear association between consumption of animal fat and SFA and BC incidence (Alexander et al., 2010; Boyd et al., 2003; Cao et al., 2016; Turner, 2011; Xia et al., 2015; Zheng et al., 2013). Two studies included in this review demonstrated a positive association between higher consumption of PUFA and risk of post‐menopausal BC, however, these studies did not differentiate between n‐3 and n‐6 PUFA in the analysis. On the other hand, higher consumption of marine n‐3 PUFA has been shown to be inversely associated with BC incidence (Nindrea et al., 2019; Zheng et al., 2013). Moreover, a higher ratio of n‐3/n‐6 is also considered to be protective (Alarcon de la Lastra, Barranco, Motilva, & Herrerias, 2001; Yang et al., 2014).

Although none of the studies presented in this review found associations between BC incidence and MUFA consumption, strong evidence exists for a protective role of EVOO. Two meta‐analyses demonstrated a robust inverse association between higher consumption of EVOO and BC risk (OR = 0.74; Xin et al., 2015; logOR = −0.45; Psaltopoulou et al., 2011). Moreover, the PREDIMED trial demonstrated in women at high risk that adherence to an MD supplemented with EVOO significantly decreased BC incidence (HR = 0.32; Toledo et al., 2015), whereas an MD + Nuts and a low‐fat control diet did not significantly influence incidence. However, the MD also encourages other dietary habits that may also be contributing to the inverse association observed.

Research evidence linking obesity to BC incidence suggests a potential mediating effect of menopausal status and a dependence on BC subtype. Two meta‐analyses discussed in this review showed inverse associations between BMI and BC risk (Chen et al., 2017; Munsell et al., 2014). Interestingly, in one of these studies (Munsell et al., 2014), subgroup analyses by hormone receptor status demonstrated that obesity was associated with increased risk for hormone receptor‐positive post‐menopausal BC, but not for ERα−/PR− cases. On the other hand, studies have demonstrated a positive association between higher BMI and pre‐menopausal TNBC incidence (Pierobon & Frankenfeld, 2013) as well as higher WC and increased overall BC incidence (Chen et al., 2016).

With regard to reproductive stage, there appears to be convincing evidence of a positive association between obesity and increased risk of post‐menopausal BC (Chen et al., 2016; Chen et al., 2017; Munsell et al., 2014; Xia et al., 2014). Interestingly, stratification by hormone receptor status suggests a consistent positive association for ERα+/PR+ BC, but no trend for ERα−/PR− cases (Munsell et al., 2014). Consistent with these findings, no association has been reported between BMI and post‐menopausal TNBC (Pierobon & Frankenfeld, 2013). In terms of survival, several studies documented a positive association between obesity and worse OS related to BC (Chan et al., 2014; Mei et al., 2018; Niraula et al., 2012; Playdon et al., 2015). In contrast, studies found no association between obesity and either DFS or OS in TNBC (Mei et al., 2018) and that ERα and PR status did not influence the association between obesity and mortality (Niraula et al., 2012).

With respect to epigenetic influences on BC risk, overexpression of DNMT3B mRNA appears to be an indicator of poor prognosis in terms of lymph node metastasis, grade, and relapse‐free survival in patients that receive hormone therapy (Girault et al., 2003). Higher levels of DNMT1 and 3A protein in BC associate with lymph node metastasis, lack of ERα and BRCA1 expression, and poor OS in ERα‐negative patients (Yu et al., 2015). Increased DNMT1 also associates with decreased DFS in patients younger than 50 years of age and decreased OS and DFS in patients that receive a combination of chemotherapy and endocrine therapy. Higher levels of DNMT1 are common in TNBC compared with other molecular BC subtypes (Shin et al., 2016). Although increases in DNMT protein and mRNA does not necessarily lead to changes in CpG methylation, these data indicating poor prognoses associated with their elevation warrant consideration for their targeting.

Conflicting data have been published with regard to the association between global CpG methylation and BC incidence. Using a relatively large sample size (n = 420), one study suggested an inverse association between CpG methylation β‐value and BC incidence (Severi et al., 2014). Two studies included in this review assessed global CpG methylation by LUMA with conflicting results since one study reported a positive association between BC incidence and low LUMA score (Kuchiba et al., 2014), whereas a second investigation showed a positive association with high LUMA score (Xu, Gammon, et al., 2012). A positive association has been suggested between 5‐mdC content and BC incidence (Choi et al., 2009). However, studies repeatedly found no link between BC risk and measures of CpG methylation at LINE‐1 and Alu DNA elements (Brennan et al., 2012; Cho et al., 2010; Choi et al., 2009; Deroo et al., 2014; Kitkumthorn et al., 2012; Wu et al., 2012; Xu, Gammon, et al., 2012). We also presented the results of studies demonstrating an association between global CpG hypomethylation and decreased all‐cause mortality, BC‐specific mortality (McCullough et al., 2016), poor differentiation (Tsai et al., 2015), disease stage,and tumour size (Soares et al., 1999).

We discussed several examples of CpG hypermethylated genes in BC including BRCA1, RASSF1A, TIMP3, RARβ, and ESR1. When considering these studies, it is important to note that different CpG sites have been investigated for the same gene. Hypermethylation of BRCA1 is associated with increased risk of BC development, lymph node metastases, higher histological grade and worse survival in terms of OS and DFS (Zhang & Long, 2015). Studies have also linked BRCA1 hypermethylation with ERα and PR negativity and the development of the TNBC phenotype. Moreover, methylated BRCA1 has shown potential efficacy as a biomarker of therapeutic response as BRCA1 methylated tumours are more sensitive to anthracycline (Ignatov et al., 2013) and carboplatin (Silver et al., 2010) chemotherapies. The hypermethylation of TIMP3 has been suggested as an early event in mammary tumourigenesis (Chen et al., 2011). Breast tumours with methylated TIMP3 tend to be of higher tumour grade, develop lymph node metastases and overexpress ERα, PR and HER2 (Lui et al., 2005).

Hypermethylated RASSF1A is detected in the majority of breast tumours (Buhmeida et al., 2011; Hagrass et al., 2014; Jezkova et al., 2016; Jezkova et al., 2017; Kajabova et al., 2013; Karray‐Chouayekh et al., 2010; Kioulafa et al., 2009; Li et al., 2008), serum (Hagrass et al., 2014; Kajabova et al., 2013) and peripheral blood cells (Yadav et al., 2018). The hypermethylation of RASSF1A has also been reported in hereditary BCs, DCIS and grade I tumours, suggesting it may be a good marker of pre‐malignant neoplasia (Alvarez et al., 2013). The hypermethylation of RASSF1A is associated with an increased risk of BC (Euhus et al., 2008), worse DFS and OS (Buhmeida et al., 2011; Kioulafa et al., 2009; Martins et al., 2011; Xu, Shetty, et al., 2012), higher risk of relapse (Kioulafa et al., 2009), increased lymph node metastasis (Hagrass et al., 2014) and refractory response to https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=6809 chemotherapy (Gil et al., 2012). Breast tumour characteristics associated with RASSF1A hypermethylation include advanced stage (Hagrass et al., 2014; Jezkova et al., 2016; Karray‐Chouayekh et al., 2010), larger size (Kajabova et al., 2013) and higher grade (Jezkova et al., 2017).

The RARβ gene is also commonly hypermethylated in BC and meta‐analyses have reported an increased risk with this biomarker (Qi & Xiong, 2018). Studies also reported that RARβ methylation was associated with TNM stages III–IV, lymph node metastases (Marzese et al., 2012; Qi & Xiong, 2018) and worse OS (Mirza et al., 2012; Wang et al., 2012). Hypermethylation of RARβ may also be a marker of distant metastases because studies showed a higher frequency of methylated RARβ in metastatic tumours compared to primary tumours (Mehrotra et al., 2004). We also summarized evidence for the potential utility for using hypermethylated ESR1 as a marker of ERα negativity (Martinez‐Galan et al., 2008; Martinez‐Galan et al., 2014; Pirouzpanah et al., 2010), lack of response to anti‐endocrine therapy (Li et al., 2013; Mastoraki et al., 2018) and worse OS (Sheng et al., 2017).

Potential associations between global CpG hypomethylation and BC risk may be confounded by obesity. For example, a positive association has been described between global CpG methylation, measured by LUMA, and BC risk in non‐obese individuals (McCullough et al., 2016). However, in obese patients, no specific association has been noted. This discrepancy may be due to the fact that obesity positively associates with global CpG methylation since the average β‐value increases with obesity for ERα negative BCs (Hair et al., 2015). Moreover, in BC survivors, global CpG methylation, measured by LINE‐1 methylation, appears to increase as a function of increased body fat percentage (Delgado‐Cruzata et al., 2015).

The notion that obesity increases global CpG methylation is supported by preclinical evidence as measured by gene hypermethylation in tumour‐distal mammary fat pads of obese mice with orthotopically injected tumours (Rossi et al., 2016). Moreover, these hypermethylation changes were associated with dysregulation of EZH2, a histone methyltransferase protein and a key factor in epigenetic regulation. Clearly, more research is needed to investigate the effects of obesity on regulation of global CpG methylation and BC risk.

Population‐based studies in BC patients suggest that obesity is associated with hypermethylation of several tumour suppressor genes including RARβ, ECAD, P16 (Tao et al., 2011), HIN1 (Krop et al., 2005), RASSF1A and BRCA1 (Naushad et al., 2014). Increased BMI is also associated with ESR1 hypermethylation (Daraei et al., 2017). There is a large gap in knowledge regarding the mechanisms linking obesity and hypermethylation of tumour suppressor genes in BC development. However, this may be the result of obesity‐induced up‐regulation of factors that stimulate these epigenetic aberrations. For example, the BRCA1 gene is hypermethylated as a result of activation of the https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2951 (https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2951), a ligand‐activated transcription factor. The tryptophan metabolites https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=5799 and https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=2918 are strong endogenous ligands of the AHR (Novikov et al., 2016) and studies have shown that their elevation in serum is positively associated with increases in BMI (Favennec et al., 2015). Moreover, the n‐6 PUFA metabolite AA is also an activator of the AHR (Go, Hwang, & Choi, 2015) and serum n‐6 PUFA levels have been shown to positively associate with increased WHR in women (Aglago et al., 2017). Future studies should seek to determine the causal link between obesity and gene methylation.

Evidence discussed in this review supports a modulating effect of dietary fat on DNMT expression and activity in the mammary gland in preclinical models and breast tumours. Overweight‐ and obesity‐inducing diets enhance mammary tumourigenesis and increase Dnmt1 expression in the mammary gland (Rossi et al., 2017). Maternal intake of a high‐fat diet during pregnancy also enhances mammary tumourigenesis and tumour Dnmt1 expression in offspring mice (de Assis et al., 2012). Moreover, diets containing high levels of corn oil rich in LA augment Dnmt1 expression and DNMT1 activity in mammary glands of rats with DMBA‐induced mammary tumours (Rodriguez‐Miguel et al., 2015). In contrast, EVOO‐containing diets decreased Dnmt3b expression in mammary tumours. Conversely, supplementation with EVOO has been shown to decrease DNMT1 expression in human peripheral blood leukocytes (Hunter et al., 2019).

7. SUMMARY AND CONCLUSIONS

Preclinical studies and human intervention trials suggest differential effects of FAs on global CpG methylation. For example, in vitro studies with monocytes show that AA increases 5‐mdC content, whereas oleic acid decreases 5‐mdC levels. Results from PREDIMED participants also suggests that EVOO has a significant influence on global CpG methylation in haematological samples. SFA and to a lower extent PUFA significantly increase global CpG methylation in adipose tissue of healthy subjects. Finally, evidence presented in this review suggests that dietary fat affects CpG methylation of putative mammary tumour suppressor genes including RARβ, RASSF1A, and TIMP3. Future studies investigating the effects of various FAs on gene‐specific CpG methylation and BC development are warranted. Specifically, there is a lack of knowledge regarding the mechanisms of how obesity or obesogenic diets influence aberrant epigenetic regulation. Future research in this field could lead to a deeper understanding of how DFAs influence breast cell epigenetics and thus act as positive and negative modulators of BC risk. In addition, future research should investigate if obesity or HFDs modulate the pharmacological response to epigenetic therapies.

7.1. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 (Alexander et al., 2019).

CONFLICT OF INTEREST

The authors declare no conflict interests.

Donovan MG, Wren SN, Cenker M, Selmin OI, Romagnolo DF. Dietary fat and obesity as modulators of breast cancer risk: Focus on DNA methylation. Br J Pharmacol. 2020;177:1331–1350. 10.1111/bph.14891