Obesity-Associated Alterations in Inflammation, Epigenetics, and Mammary Tumor Growth Persist in Formerly Obese Mice

Obesity is highly prevalent in the United States and many other parts of the world and is an established risk and progression factor for several intrinsic subtypes of breast cancer, including basal-like breast cancer (BLBC), in postmenopausal women (1). Greater rates of metastatic breast cancer and increased mortality are seen in the obese breast cancer patient population, independent of disease stage at diagnosis (2, 3). These effects may be mediated in part by adipose tissue inflammation, as excess adipose tissue results in local as well as systemic effects. Obesity is typically accompanied by increased systemic levels of proinflammatory mediators, such as IL6, TNFα, matrix metalloproteinase-9 (MMP-9), and IL1β, and these can independently influence tumor growth. In addition, obesity enhances the secretion of monocyte chemoattractant protein-1 (MCP-1), which stimulates the recruitment of macrophages to adipose tissue, including the breast adipose. These tumor-associated macrophages likely contribute to tumor growth by increasing local and/or systemic inflammatory and angiogenic factors and generating reactive oxygen species (4, 5).

In obese mice, calorie restriction reduces mammary tissue expression of several mediators of inflammation, including TNFα and IL1β (6). Less is known about the effects of obesity or weight loss interventions on the epigenome. There is mounting evidence that DNA methylation can be affected by energy balance. For example, weight loss interventions (30% energy reduction) in obese male subjects have been shown to induce DNA methylation changes at the TNFα promoter after only 8 weeks of diet intervention (7). In another study, severe weight loss after gastric bypass surgery in obese humans normalized methylation in the promoter regions of 11 of the 14 metabolic genes that had been dysregulated by obesity (8). Expression of the EZH2 gene, which encodes the epigenetic regulator histone-lysine N-methyltransferase, has been linked to breast cancer growth and metastasis (9). It is also modulated in the offspring of obese mothers (10), suggesting that this enzyme may mediate some of the effects of energy balance on epigenetic regulation. Although DNA methylation patterns are energy responsive, the relationship between weight loss, aberrant DNA methylation, adipose tissue inflammation, and cancer progression are unknown.

Epidemiologic and preclinical studies have clearly shown that the metabolic perturbations typically accompanying obesity promote breast cancer progression (11). One study, in particular, illustrated how women who underwent bariatric surgery had a 5-fold reduction in incidence of breast cancer after 5 years of follow-up from surgery compared with women who did not undergo a weight-loss intervention, surgical or otherwise (12). However, with the exception of bariatric surgery interventions, the impact of intentional obesity reversal prior to diagnosis has not been well established. The epidemiologic data have been mixed regarding nonsurgically induced weight loss, reflecting the difficulties in achieving and maintaining meaningful weight loss. Despite their limitations, animal model studies may be informative in addressing this question, given their benefit of greater control of diet, physical activity, and numerous other confounding variables. Sundaram and colleagues found that a high-fat diet promotes mammary tumor progression in a transgenic mouse model of BLBC, and switching mice to a low-fat diet after a short-term high-fat diet exposure led to small, but statistically significant, weight loss and reversed many of the high-fat diet effects on tumor progression to that of control levels (13). However, the transgenic mouse model used in this study was resistant to obesity induction, and tumors arose early in life, so the question of whether the mammary tumor–promoting effects of chronic obesity are reversed by weight loss was not addressed. We have previously found in a mouse model of chronic obesity, as well as in clinical studies of obese women, that a severe weight-loss intervention may provide an anticancer benefit, but moderate weight loss had minimal effects on cancer-related biomarkers in mice and women and tumor growth in mice (14, 15).

Given the uncertainty regarding the benefits of obesity reversal, we sought to determine the effects of weight loss on the host microenvironment and tumor progression in an orthotopic transplant model of murine BLBC in syngeneic mice. Here, we demonstrate that tumor growth did not change with significant weight loss prior to orthotopic tumor transplant, despite the normalization of several obesity-related hormones. Moreover, we provide evidence that breast tumor growth after obesity reversal may be driven by epigenetic dysregulation in the microenvironment, as well as sustained elevations in inflammatory signaling. These findings suggest that short-term weight loss may not reverse the tumor-promoting epigenetic reprogramming and increased inflammatory signaling that accompanies chronic obesity.

All animal protocols were approved by the University of Texas (Houston, TX) at Austin Institutional Animal Care and Use Committee and carried out in compliance with all guidelines and regulations. Six-week-old female ovariectomized C57BL/6 mice were obtained from Charles River Laboratories, Inc., individually housed on a 12-hour light/dark cycle, and consumed food and water ad libitum. Food intake was measured twice a week and body weights were measured weekly. Mice were randomized to continue on the control diet (n = 17) or a 60% kcal from fat diet–induced obesity (DIO) diet regimen (n = 34; D12492, Research Diets, Inc). After 17 weeks on diet, the DIO mice were randomized to either continue on the DIO regimen (DIO, n = 17) or switch to the control diet to induce gradual weight loss; this latter group is referred to as formerly obese (FOb, n = 17). After 24 weeks on diet, 5 mice per group were fasted for 6 to 8 hours and then euthanized for pretumor tissue collection. The remaining mice (n = 15/group) were orthotopically injected with 5 × 10⁴ syngeneic viable MMTV-Wnt-1 mammary tumor cells as described previously (16). In vivo growth was measured twice per week with skinfold calipers, and in vivo tumor area was approximated using the formula πr². At study endpoint (week 36), mice were euthanized and mammary tumors, tumor-adjacent and tumor-distal mammary fat pad were excised and divided in portions to be formalin fixed or flash frozen in liquid nitrogen and stored at −80°C until further analysis. Mice were excluded if they had developed dermatitis during study (one DIO mouse and one FOb mouse). Ex vivo tumor volume was calculated using the formula 4/3π × R₁² × R₂, (with R₁ denoting the smaller radius of the ellipsoid; ref. 17).

Body composition was measured on all mice at weeks 17 and 24 of diet treatments by quantitative magnetic resonance (qMR; Echo Medical Systems). Measurements included lean mass, fat mass, and total water mass. Percent body fat was calculated by dividing fat mass by total body weight.

Serum was collected from mice fasted 6 to 8 hours prior to tumor cell injection (week 24) by retro-orbital bleed. Serum hormones, adipokines, and cytokines, including leptin, adiponectin, insulin, and IL6, were measured using mouse adipokine LINCOplex Multiplex Assays (Millipore, Inc) and analyzed on a Bio-Rad Bio-Plex 200 Analysis System (Bio-Rad, Inc). Insulin-like growth factor 1 (IGF1) concentrations were measured using a Millipore MILLIPLEX Rat/Mouse IGF-1 Single Plex Assay (Millipore, Inc).

Four-micron thick sections were prepared from formalin-fixed, paraffin-embedded mammary fat pad tissue and stained with hematoxylin and eosin. The total number of crown-like structures (CLS) per section was quantified by a pathologist, blind to the sample group, and the amount of adipose tissue present on each slide was determined using NIH ImageJ. Prevalence of CLS was quantified as CLS per cm² of adipose tissue.

Four-micron thick sections were prepared from formalin-fixed, paraffin-embedded tumor tissue and stained with hematoxylin and eosin. Tumors were chosen at random (4–5/group) and digitally imaged under 20× magnification. The total number of adipocytes per section (2–4 representative sections, each 830 μm × 580 μm) was quantified independently by 3 investigators who were blinded to experimental group.

Total RNA was extracted from the flash-frozen tumor-adjacent and tumor-distal mammary fat pad samples collected at the end of the study using TRI Reagent (Sigma-Aldrich) according to the manufacturer's instructions. RNA concentration was spectrophotometrically determined using NanoDrop (Thermo Scientific) and quality was confirmed using an Agilent 2100 Bioanalyzer. RNA was reverse transcribed with MultiScribe RT (Applied Biosystems). Resulting cDNAs from tissue samples were assayed in triplicate for PCR using TaqMan Gene Expression Assays for IL6, TNFα, MMP-9, IL1β, IGFBP6, CITED1, TFE3, JAK3, EZH2, SMYD3, and TSC22D3 (Applied Biosystems). PCR reactions were completed using a ViiA7 Real-Time PCR System (Applied Biosystems). Gene expression data were normalized to the housekeeping gene β-actin and analyzed using the delta delta cycle threshold method (ΔΔC_t; ref. 18).

DNA was extracted from the ninth tumor-distal mammary fat pad, at the end of the study (n = 3/group) using UltraPure Phenol:Chloroform:Isoamyl Alcohol per manufacturer's instructions (Life Technologies). Genome-wide methylation profiles for mammary fat pad from control, DIO, and FOb mice were determined by reduced representation bisulfite sequencing (RBBS). Among the available methods for genome-wide analysis of the DNA methylation, we selected RRBS based on its compatibility and validation with mouse DNA, its capacity to investigate individual sites by bisulfite sequencing, and the large number (above 400,000) of CpG sites that are evaluated in a single experiment. Library preparation and sequencing were completed at the UT MD Anderson Cancer Center's DNA Methylation Analysis Core and Science Park Next-Generation Sequencing Facility, according to the published protocols (refs. 19, 20; Supplementary Table S3). Briefly, 1 μg of genomic DNA was digested with MspI, a methylation-insensitive enzyme that cleaves the DNA at CCGG sequences, followed by end repair, A-tailing, and size selection. Enrichment for CpG sites in CG-dense and CpG islands is achieved by selective collection and PCR amplification of MspI fragments with sizes between 40 and 170 bp. Sequence reads were mapped to the mouse genome (mm10) using Bismark Bisulfite Read Mapper. The location of individual CpG sites in relationship to known genes was based on the mm10 RefSeq gene annotations and classified as promoter [CpG sites 1 kb upstream of the annotated transcription start site (TSS) and extending 500 base pairs (bp) downstream of TSS], exon, intron, or CpG island. CpG island coordinates were collected from the UCSC Genome Browser mm10 version, Genome Reference Consortium GRCm38. Differential methylation was calculated by filtering samples based on read coverage ≥20, then performed at the single base level independent of the CpG site location. TSS of transcripts smaller than 300 bp was removed for the analysis. MethylKit R package was used to apply logistic regression and the likelihood ratio test. Observed P values were adjusted with the success likelihood index method.

Selection criteria for the Normal Breast Study included having an invasive ductal carcinoma with a tissue specimen sampled ≥4 cm from tumor margins, undergoing reduction mammoplasty, or undergoing prophylactic surgery (21). There were 150 participants with invasive breast cancer who had tissue collected that was >4 cm from the tumor. Participants were randomly selected from each body mass index (BMI) group: 30 from the high BMI (>30), 20 from the normal BMI (22–27), and 22 from the low BMI (<25) group. All prophylactics and reduction participants were also included. In all, 96 participants were selected. All tissue specimens were snap frozen. A 50-mg sample was taken from the participants' tissue specimens and used for methylation assessment. The University of North Carolina at Chapel Hill Institutional Review Board approved this study.

DNA was isolated using the DNeasy Blood and Tissue Kit from Qiagen following the manufacturer's instructions. Sodium bisulfite modification of the DNA was conducted using EZ DNA Methylation-Gold Kit (Zymo Research), and 500 ng of DNA was used for the HumanMethylation450 BeadChip platform. The samples were processed for DNA methylation analysis by the University of North Carolina at Chapel Hill's Mammalian Genotyping Core (Chapel Hill, NC).

We compared our mouse methylation dataset with a dataset obtained from the University of North Carolina Normal Breast Study utilizing nontumorous breast tissue, subject to methylation array, and stratified by BMI. A methylation difference was calculated for each significant methylation change (obese vs. nonobese) with a positive coefficient signifying increasing methylation with increased adiposity. Intraspecies concordance was assayed only for genes with significant differences within species (Supplementary Table S2). Welch t test was used for the pairwise comparisons between obese and nonobese humans using R version 3.1.2 (http://www.r-project.org). Comparisons between both species were done by one-way ANOVA using GraphPad Prism V6 with Bonferroni correction for multiple comparisons.

Values are presented as mean ± SD. One-way ANOVA using Tukey multiple comparisons correction was used to assess the effects of diet treatment on body weight and fat percentage, mean tumor size, serum hormone and cytokine concentrations, and mammary fat pad gene expression. Student t test was used to compare differences in CLS between experimental groups and control. For all tests, GraphPad Prism software was used (GraphPad Software Inc), and P ≤ 0.05 was considered statistically significant.

Mice were fed a low-fat control or a DIO regimen for 17 weeks, at which time, percent body fat, assessed by quantitative resonance analysis, was higher in DIO mice (59.7 ± 1.8%) compared with control mice (39.5 ± 4.0%; P < 0.001; Fig. 1A). DIO mice also had a higher mean body weight (51.5 ± 0.7 g) than control mice (33.1 ± 0.7 g; Fig. 1B). After weight loss was initiated at week 17, FOb mice lost 27.6 ± 3.4% of their body weight (14.2 ± 1.7 g; P < 0.0001) relative to their last weight taken while on the DIO diet. By week 24, the body fat percentage in FOb mice (45.1 ± 3.5%) did not significantly differ from mice maintained on the control diet throughout study (42.6 ± 3.3%), with DIO mice (60.1 ± 1.5%) having significantly greater body fat percentage than control and FOb mice (P < 0.001; Fig. 1A). FOb and control mice maintained significantly lower body weights than DIO mice through the end of the study (P < 0.0001, Fig. 1B).

Serum was collected from mice at week 24 by retro-orbital bleed, prior to tumor cell injection and when the body weights of the FOb mice were statistically equivalent to the control mice. The reduction in adiposity in the FOb mice was accompanied by lower serum leptin (P < 0.0001), leptin:adiponectin ratio (P < 0.0001), and insulin (P < 0.0001) relative to DIO mice (Fig. 1C–E), with concentrations statistically equivalent to those found in the control mice. IGF1 levels were significantly higher in DIO mice relative to control (P < 0.05; Fig. 1G) but were intermediate in FOb mice and not statistically different from the levels seen in control or DIO mice. Furthermore, IL6 levels did not differ between FOb and DIO mice and were significantly elevated in both groups compared with control mice (P < 0.05; Fig. 1H).

The high-fat dietary regimen increased the prevalence of CLS (expressed as CLS/area of tissue in cm²) in the mammary fat pad of DIO mice relative to control mice (P < 0.003; Fig. 2). This was analyzed in tissue harvested at the interim (week 24) time point. FOb mice had a significant reduction in CLS compared with DIO mice (P < 0.03); however, prevalence of CLS remained slightly elevated relative to control, although this difference was not statistically significant (P = 0.4).

Ex vivo tumor volumes from DIO and FOb mice were larger (1,792.7 ± 1,132.9 mm³, 1,563.4 ± 1,673.3 mm³, respectively) than tumors from the control mice (51.6 ± 139.8 mm³; P < 0.05; Fig. 3A). Furthermore, a shared attribute of tumors from both the DIO and FOb groups was the presence of adipose cells infiltrating the tumor in greater number compared with control (Fig. 3B). Tumors from DIO mice contained a significantly higher number of adipocytes than control tumors (168.4 ± 83.4 adipocytes/mm² and 34.6 ± 32.2 adipocytes/mm², respectively; P < 0.05). Tumors in FOb mice had an intermediate level of adipocyte infiltration (127.1 ± 83.6 adipocytes/mm²) that did not significantly differ from either control (P = 0.19) or DIO (P = 0.66). Collectively, these findings suggest that significant weight loss does not completely reverse the effects of obesity on BLBC growth and adipocyte infiltration.

RT-PCR analysis was performed on tumor-distal mammary fat pad (n = 4–5 mice). Relative to control, FOb mice had significantly elevated IL6 expression (P < 0.001, Fig. 4A), with DIO mice having nonsignificantly elevated levels relative to control. TNFα expression was significantly elevated in DIO (P < 0.04) and FOb (P < 0.03; Fig. 4B) versus control mice. MMP-9 expression was greater in FOb mice (P < 0.05; Fig. 4C), with DIO mice experiencing an intermediate increase compared with control. IL1β expression was significantly higher in FOb and DIO mice (P < 0.01; Fig. 4D) compared with control.

To determine whether the sustained elevation in tumor growth observed in FOb mice may be mediated in part by epigenetic reprogramming in the microenvironment, we assessed the global DNA methylation pattern of the tumor-distal mammary fat pad at the end of the study. DIO mice had 39 genes significantly hypermethylated compared with control. Of these 39 differentially methylated genes, FOb mice generally possessed intermediate levels of methylation relative to control and DIO mice, suggesting that weight reduction back to control levels was insufficient to completely reverse the epigenetic reprogramming that occurs with chronic obesity (Fig. 5A). FOb mice had 11 genes significantly hypermethylated compared with control. Although the methylation levels of these genes were also elevated in DIO mice, they were not significantly different from control

Hypermethylated genes in DIO and FOb mice relative to control were entered into Qiagen's Ingenuity Pathway Analysis (IPA) to delineate pathologic signaling pathways. Gene expression was confirmed on a select number of genes (Fig. 5B–H). IPA biofunction analysis connected the genes from our methylation analysis to a disease or biologic function. Cancer was the primary disease with 23 genes hypermethylated in DIO (22 of the genes) or DIO and FOb (1 gene) implicated in cancer, specifically neoplasia of epithelial tissue (Supplementary Table S1). Another biofunction with 17 genes associated with it was cellular growth and proliferation. Four genes in our dataset were identified as playing a role in cell-mediated immune response. Significant overlap of differentially methylated genes in DIO or FOb mice with each biofunction was determined by Fisher exact test.

We moved forward with genes that appeared in the IPA biofunction of cancer with a function of neoplasia of epithelial tissue (IGFBP6, SMYD3, and JAK3) and transcription factor TSC22D3 and investigated their role as regulators in inflammatory signaling, with particular interest in the relationship between each gene and IL6 (Supplementary Fig. S1). As stated previously, IL6 serum (Fig. 1H) and tumor-distal mammary fat pad mRNA levels (Fig. 4A) were elevated in DIO and FOb mice relative to controls. Of note, inhibitors of IL6, including TSC22D3 (intron) and JAK3 (exon), were hypermethylated. EZH2 was also predicted by the IPA analysis to be regulated by these genes, and expression of EZH2 in tumor-distal mammary fat pad of endpoint animals was significantly higher in DIO (P < 0.05) and FOb (P < 0.005) mice compared with control mice (Fig. 5H). The gene for another methyltransferase enzyme, SET and MYND domain–containing protein 3 (SMYD3), was also identified as differentially methylated in our analysis, and its expression in tumor-distal mammary fat pad was significantly elevated in FOb mice (P < 0.02), with DIO mice having nonsignificantly elevated levels.

After we identified a panel of genes that were significantly hypermethylated in the mammary fat pad of DIO mice compared with controls, we assessed the methylation status of this gene panel in a methylation dataset from normal human breast tissue involving subjects stratified by BMI (Supplementary Table S2). Thirteen genes were identified with concordant hypermethylation (% methylation difference with the same directionality) in DNA from normal mammary tissue at the same genomic feature (i.e., intron vs. exon) in obese versus nonobese women and mice (Table 1). However, as some significant methylation changes mapped to the same gene, of the 13 concordant genes, 6 had additional discordant methylation sites. We also identified an additional 3 genes with only discordant methylation between species. We then compared methylation (as % methylated DNA) for CpGs related to the same gene in the same genomic feature (i.e., intron or exon) between women and mice. Nonobese women and control-fed mice had levels of DNA methylation that were statistically equivalent for 4 of the 13 concordant genes identified. Of the 13 hypermethylated genes in obese women and DIO mice, 6 genes reached statistical equivalence (CDK2AP1, CYFIP1, MAF, MAML3, SH3PXD2A, and SMYD3). Of those, 4 were statistically equivalent in obese women, DIO mice, and FOb mice. Seven genes were significantly different between obese humans and DIO mice due to varying degrees of hypermethylation.

In our orthotopic tumor transplant model of BLBC in C57BL/6 mice, we found that DIO, relative to control diet, elevated: (i) body fat percentage; (ii) serum leptin, leptin:adiponectin ratio, insulin, IGF1, and IL6; (iii) prevalence of CLS in mammary fat pad; (iv) mammary tumor growth; and (v) genome-wide mammary gland hypermethylation. Significant weight loss initiated after 17 weeks on DIO diet resulted in a normalization of percent body fat and serum leptin, leptin:adiponectin, insulin levels, and prevalence of CLS in mammary fat pad in the FOb mice. Serum IGF1 decreased to levels intermediate to DIO and control levels. Despite these beneficial metabolic changes and weight normalization, FOb mice had a final tumor volume nearly identical to DIO mice. In addition, we observed that obesity reversal in the FOb mice failed to normalize serum IL6 and mammary fat pad inflammatory gene expression, as well as aberrant methylation in the mammary gland, and these may contribute to the persistence of the procancer effects of obesity even after weight is normalized. We have identified several genes with concordant obesity-related hypermethylation in humans and mice, and the DNA methylation levels of many of these genes were unchanged after significant weight loss in the FOb mice, providing further support that obesity-associated epigenetic reprogramming may not be easily reversed with weight normalization alone.

To our knowledge, this is the first study to: (i) characterize the effects of obesity reversal on mammary gland microenvironment (including DNA methylation) and tumor growth; and (ii) identify genes with concordant and discordant obesity-related methylation patterns in normal mammary tissue from mice and humans. We and others have previously found that a severe weight loss intervention, or inhibiting the mTOR signaling pathway in obese mice using genetic or pharmacologic approaches (18, 22, 23), decreases murine mammary tumor progression. Severe weight loss in humans, such as that achieved by bariatric surgery, has also shown the ability to significantly decrease inflammatory signaling (24), DNA hypermethylation (8), and cancer risk (12, 25). However, the weight loss that was achieved in our study via an ad libitum low-fat diet, which models many human weight-loss plans, failed to reduce inflammatory signaling and elevated methylation levels in the FOb mice.

Inflammation is implicated in many of the deleterious effects of obesity (26, 27). In this study, weight loss in FOb mice normalized metabolic regulators that are elevated in obesity, such as serum leptin:adiponectin, insulin, and adipose tissue CLS. However, several measures of inflammation persisted after weight loss, including serum IL6 and gene expression of inflammatory cytokines in the mammary tissue. Thus, we are able to distinguish the procancer effects of obesity-related metabolic perturbations from obesity-related inflammation, which persist after weight loss.

We also found evidence suggesting that obesity-induced aberrant hypermethylation contributes to the maintenance of a proinflammatory, progrowth microenvironment in the mammary fat pad of mice, and this is not normalized by significant weight loss. Abnormal expression of EZH2 plays a role in cancer growth and metastasis in a variety of cancers, including breast (28). A high-density tissue microarray analysis of breast tumors from 280 patients revealed that EZH2 levels were increased in patients with invasive breast carcinoma relative to patients with no lesions or atypical hyperplasia (29). In addition, EZH2 increases expression of SMYD3, a histone H3 lysine-4–specific methyltransferase, which plays a key role in the proliferation and migration of several cancers. Cell culture studies have shown that silencing of SMYD3 by siRNA inhibits growth of breast cancer cells (30). We demonstrated here that both EZH2 and SMYD3 expression remained elevated in the mammary fat pad tissue of FOb mice, similar to the DIO group, indicating that these epigenetic regulators may be involved in mediating the mammary tumor–promoting effects of obesity.

Study limitations include the lack of a time course, as it is plausible that FOb mice could have had decreased tumor growth if challenged after a longer period of weight normalization. In addition, as serum IL6, mammary fat pad inflammatory gene expression, and mammary fat pad methylation all remained elevated in FOb mice compared with control, we are unable to delineate the individual contribution of each factor to the increased tumor burden. As expected, many, but not all, methylation changes investigated resulted in a corresponding change in gene expression. This is likely influenced by distance between the site of methylation change and promoter region, influence of chromatin features, or multiple systemic levels of regulation at play (31, 32).

The results of our study raise important questions concerning the relationship between obesity and the progression of more aggressive breast cancers, such as the basal type. Numerous studies have shown that obesity is associated with increased risk of mortality from breast cancer (33). As weight loss did not suppress tumor progression in mice that had at one time been obese, further research is needed to investigate a potential critical window regarding the timing and magnitude of weight loss to achieve an antitumor effect. Future studies should also explore weight loss in combination with adjuvant therapy, such as prophylactic NSAID use to combat inflammation. The hypermethylation observed in DIO and FOb mice may also provide a new druggable target for breaking the obesity–breast cancer link.

In summary, we found that MMTV-Wnt-1 mammary tumor growth was similar in DIO and FOb mice, with both groups exhibiting larger tumors than control mice. DNA methylation in the tumor-distal mammary fat pad at study endpoint was also significantly elevated in DIO mice relative to control, with FOb mice possessing intermediate DNA methylation levels. This is particularly surprising, considering that at the time of tumor cell injection, FOb mice did not differ from control mice in terms of body weight, body composition, and serum metabolic markers of obesity, with the exception of serum IL6. Taken together, our findings suggest a relationship between chronic obesity, the epigenome, inflammation, and mammary cancer. Consequently, a combination of weight loss and epigenetic or anti-inflammatory interventions may be required to effectively break the obesity–breast cancer link.

No potential conflicts of interest were disclosed.

Conception and design: E.L. Rossi, R.E. de Angel, D. Giri, S.D. Hursting

Development of methodology: R.E. de Angel, D. Giri, S.D. Hursting

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.E. de Angel, P. Bhardwaj, D. Giri, M.R. Estecio, M.A. Troester, B.Y. Hair, E.L. Kirk, J. Shen, A.J. Dannenberg

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.L. Rossi, R.E. de Angel, L.W. Bowers, S.A. Khatib, L.A. Smith, E. Van Buren, D. Giri, T. Gong, S.D. Hursting

Writing, review, and/or revision of the manuscript: E.L. Rossi, R.E. de Angel, L.W. Bowers, S.A. Khatib, L.A. Smith, D. Giri, M.A. Troester, B.Y. Hair, T. Gong, J. Shen, A.J. Dannenberg, S.D. Hursting

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.E. de Angel, S.A. Khatib, L.A. Smith, D. Giri, E.L. Kirk

Study supervision: R.E. de Angel, S.D. Hursting

Other (facilitated collaboration with investigators managing human methylation dataset): E.L. Rossi, S.D. Hursting

Other (managed animal study): R.E. de Angel

The animal study reported here was funded by a grant to S.D. Hursting from the Breast Cancer Research Foundation. The mouse methylation analysis was funded by a pilot grant from the MD Anderson Science Park NGS Core, supported by CPRIT Core Facility Support Grant RP120348. Fellowships to E.L. Rossi and L.W. Bowers were supported by a grant from the NCI (R25CA057726). S.D. Hursting is supported by an Outstanding Investigator Grant from the NCI (R35CA197627).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.