Pubertal high fat diet: effects on mammary cancer development

Diets were initiated at 3 weeks of age and maintained throughout the experimental period up to 45 weeks of age. Two diets (I and II) were used in these studies. Diet I: low fat diet 58G7 (12% kcal fat - LFD) and high fat diet 58G9 (60% kcal fat - HFD) were purchased from TestDiet (PMI Nutrition International LLC). Initial concern over the carbohydrate content of Diet I and the potential for the development of metabolic syndrome led us to also use Diet II, which contained maltodextrin instead of the dextrin and sucrose contained in Diet I. Diet II: low fat diet 12450B (10% kcal fat - LFD) and high fat diet 12492 (60% kcal fat - HFD), both with corn oil replacing soy oil, were purchased from Research Diets (New Brunswick, NJ, USA). For Diet I, 11% kcal fat came from corn oil, with the remainder of the fat (1 or 49% kcal) coming from lard. For Diet II, 5.5% kcal fat came from corn oil, with the remainder of the fat (4.5 or 54.5% kcal) coming from lard. Detailed composition of Diets I and II is described in Additional file 1. LFD and HFD Diets I and II did not differ significantly with regard to their effects on body weight, blood glucose, insulin levels, tumor incidence, or tumor latencies (see Additional files 2, 3, 4, 5). Therefore, results obtained on tumorigenesis with the two diets were combined.

Mice were treated with 7,12-dimethylbenz(a)anthracene (DMBA) (Sigma-Aldrich, St. Louis, MO, USA) prepared in vegetable oil and administered by oral gavage (50 mg/kg body weight/mouse) once per week for 4 weeks starting at 5 weeks of age. Body weights were monitored weekly, and animals were palpated for tumors once a week starting at 8 weeks post first DMBA dose (see Additional file 6). Tumor volume was measured twice per week, and harvested at 1 cm size. At 2 h prior to sacrifice, mice were injected with 5-bromo-2’-deoxyuridine (BrdU) (70 μg/g body weight; Sigma-Aldrich) for analysis of cellular proliferation. At termination of all feeding studies, portions of tumors and mammary tissues were either snap frozen for protein and RNA isolation, or formalin fixed and either processed as whole mounts [19] or paraffin embedded for H&E staining and immunohistochemistry [20]. Whole-mount preparations of glands and H&E sections were scored for the presence of hyperplasia and neoplasia [21]. There were no notable diet effects on mammary gland morphology as assessed by whole mounts or histological sections at the time points analyzed. All lesions and tumors were reviewed and classified, as previously described [22].

Plasma glucose and insulin levels were the metabolic parameters measured. Non-fasting, randomly sampled glucose and insulin levels were obtained from mice fed ad libitum, as an appropriate and acceptable method based on mouse feeding habits and the stress caused by fasting [23]. Plasma levels of glucose were determined by AccuChek Compact Glucometer (Roche, Nutley, NJ, USA) and insulin levels were determined with an insulin ELISA kit from EMD Millipore (Cat. number: EZRMI-13 K; St. Charles, MO, USA), following the manufacturer’s instructions.

Total serum E levels were determined using Delfia Estradiol time-resolved fluoroimmunoassay (Cat. number: 1244–056; PerkinElmer, Turku, Finland), following the manufacturer’s instructions. The levels of serum P were measured by ELISA (Cat. number: 11-PROHU-E01; Alpco Diagnostics, Salem, NH, USA), following the manufacturer's instructions.

Mammary gland or tumor sections (5 μm) were prepared and subjected to antigen retrieval and immunofluorescent staining, as previously described [20]. Briefly, sections were blocked with goat anti-mouse immunoglobulin G (IgG) Fab (antigen binding fragment; 1:100 with 1% BSA in PBS (PBSA)) for 1 h at room temperature (RT), followed by blocking with normal goat serum in PBS for 30 minutes at RT. Then, sections were incubated with mouse anti-ERα (1:10 in PBS-0.5% Triton X-100; Cat. number: NCL-ER-6 F11; Novocastra Laboratories Ltd, Newcastle upon Tyne, UK) at 4°C overnight. After a brief wash, sections were incubated with an Alexa 488-labeled goat anti-mouse secondary antibody (Ab) (1:200 in PBS; Invitrogen Molecular Probes, Grand Island, NY, USA) at RT for 30 minutes, and then counterstained with 4’,6-diamidino-2-phenylindole (DAPI) for 5 minutes. The stained sections were visualized with a Nikon Eclipse TE2000-U fluorescence microscope (Nikon, Inc., Melville, NY, USA) using a 40X objective lens, and the captured fluorescent images were analyzed using MetaMorph software (Molecular Devices, LLC, Sunnyvale, CA, USA). A minimum of 1,000 cells were counted for each tumor. Tumors were considered to be ERα-positive (ER+) if 10% or more of the total cells counted were ER + [24].

Mammary gland or tumor sections (5 μm) were prepared for immunofluorescent staining [20]. Briefly, after deparaffinization, mammary gland sections were subjected to antigen retrieval at 121°C and 15 pounds per square inch (psi) for 5 minutes; tumor sections did not receive antigen retrieval, as the integrity of tumor sections was compromised by the procedure, and staining was adequate in its absence. All sections were treated with proteinase K (20 μg/mL in Tris-EDTA buffer, pH 8.0) at 37°C for 3.5 minutes. After brief washes with PBS, sections were blocked with normal rabbit serum in PBS, followed by incubation with goat anti-Arginase 1 (Arg1) (1:200 in PBS-0.5% Triton X-100; Cat. number: sc-18354; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) at 4°C overnight. After brief washes, sections were incubated with Alexa 546-labeled rabbit anti-goat secondary Ab (1:100 in PBS; Cat. number: A21085; Invitrogen Molecular Probes) at RT for 30 minutes and then blocked with normal goat serum in PBS for 30 minutes. The sections were then incubated with rat monoclonal anti-F4/80 (1:75 in PBS-0.5% Triton X-100; Cat. number: MCA497R; AbD Serotec, Raleigh, NC, USA) at 4°C overnight, followed by incubation with Alexa 488-labeled goat anti-rat secondary Ab (1:100 in PBS; Invitrogen Molecular Probes) at RT for 30 minutes and counterstaining with DAPI for 5 minutes. The stained sections were visualized with a Nikon Eclipse TE2000-U fluorescence microscope (Nikon, Inc.) using a 40X objective lens, and the captured fluorescent images were analyzed using MetaMorph software (Molecular Devices, LLC). The number of F4/80 and/or Arg1-positive cells is expressed as cells per structure in the mammary gland peri-epithelial area, and cells per image in tumor samples.

Mammary gland or tumor sections (5 μm) were prepared and immunoperoxidase staining was performed [20]. Briefly, after deparaffinization, sections were incubated in 3% of hydrogen peroxide (H₂O₂) in methanol for 10 minutes. Then sections were subjected to antigen retrieval at 121°C and 15 psi for 15 minutes. Sections were blocked with goat anti-mouse IgG Fab (1:100 in 1% PBSA) for 1 h, followed by blocking with normal goat serum in PBS. Then, sections were incubated with mouse anti-BrdU (1:100 in PBS-0.5% Triton X-100; Cat. number: ab27958; Abcam plc, Cambridge, MA, USA) at 4°C overnight. After a brief wash, sections were incubated with goat anti-mouse biotin secondary Ab (1:400 in PBS; DAKO Denmark A/S, Glostrup, Denmark) at RT for 30 minutes, and then incubated with ABC reagent (PK-7100, Vector laboratories, Inc., Burlingame CA, USA) for 30 minutes. The sections were then incubated with metal-enhanced 3,3-diaminobenzidine (DAB) substrate solution (100 μL of DAB substrate + 900 μL stable peroxide substrate buffer; Thermo Scientific, Rockford, IL, USA) for 7 minutes and counterstained with hematoxylin for 2 minutes. The stained sections were visualized with a Nikon Eclipse E400 light microscope (Nikon, Inc.) using a 40X objective lens. A minimum of 1,000 cells were counted for each section, and a minimum of two to three tissue sections per animal were analyzed. The number of BrdU-positive cells is expressed as the percentage of total epithelial cells counted.

Mammary gland or tumor sections (5 μm) were deparaffinized and were incubated in 2% H₂O₂ in methanol/PBS (1:1 ratio) for 30 minutes followed by antigen retrieval by boiling at 95°C for 5 minutes. Sections were then treated for 2 h with rabbit anti-CD31 (1:50 in PBS-0.5% Triton X-100; Cat. number: AP15436PU-N; Acris Antibodies, Inc., San Diego, CA, USA) for endothelial cell staining and detection of blood vessels. After a brief wash, sections were incubated with secondary swine anti-rabbit Ab at RT for 30 minutes, and then incubated with ABC reagent (PK-7100, Vector laboratories, Inc.) for 30 minutes. The sections were then incubated with metal-enhanced DAB substrate solution (100 μL of DAB substrate + 900 μL stable peroxide substrate buffer; Thermo Scientific) for 7 minutes and counterstained with hematoxylin for 2 minutes. The stained sections were visualized with a Nikon Eclipse E400 light microscope (Nikon, Inc.) using a 40X objective lens. A minimum of 1,000 cells were counted for each section, and a minimum of two to three tissue sections per animal were analyzed. Digital micrographs were captured and the images were overlaid with grids containing 240 squares (324 μm²/square). Blood vessel density is expressed as the percentage of CD31-positive squares.

Deparaffinized 5-μm sections were stained with Astra Blue/Vital New Red [25]. Sections dehydrated in 95% ethanol for 5 sec were incubated in Astra Blue solution (5 mg/mL in 75% ethanol) for 30 minutes at RT. After rinsing in double-distilled water (H₂O), sections were again dehydrated in 95% ethanol for 5 sec, and then incubated in Vital New Red solution (0.2 g/L in 50% ethanol) for 1 h at RT and counterstained with hematoxylin for 2 minutes. The stained sections were visualized with a Nikon Eclipse E400 light microscope (Nikon, Inc.) using a 40X objective lens. The number of eosinophils and mast cells is expressed as cells per structure in the mammary gland peri-epithelial area.

Sections stained for macrophages, cellular proliferation, eosinophils, and mast cells were analyzed by mammary gland epithelial structure: small ducts, large ducts, terminal end buds (TEBs), or hyperplastic foci. Large ducts were qualitatively characterized by larger lumen diameter, comprising greater than 50 cells in a section, and by an extensive extracellular matrix and fibroblasts surrounding the epithelium. Conversely, small ducts were qualitatively characterized by smaller lumen diameter, comprising fewer than 50 cells, and by a limited extracellular matrix surrounding the epithelium. TEBs were characterized by location within the gland, presence of multiple epithelial cell layers, and direct apposition of epithelium to adipocytes without extracellular matrix. Hyperplastic structures had multiple cell layers of noticeably distorted epithelium compared to normal epithelial structures.

Total RNA was isolated from mouse mammary glands (intact or cleared fat pads) or tumors using TRIzol® reagent (Invitrogen, Carlsbad, CA, USA) and purified using the RT² qPCR-Grade RNA isolation kit (SABiosciences, Frederick, MD, USA). We used 3 μg of total RNA for first strand cDNA synthesis using the RT² First Strand kit (SABiosciences), following the manufacturer’s instructions. The cDNA (20 μL) were diluted to 150 μL with de-ionized H₂O. RNA expression was then analyzed using targeted PCR arrays (SABiosciences) for Breast Cancer (PAMM-131), Cancer Pathway Finder (PAMM-033), Growth Factors (PAMM-041), and Inflammatory Cytokines and Receptors(PAMM-011). In addition to the arrays, primers for the following selected RNA were purchased from SABiosciences: Tnfs11 (RANKL) (PPM03047E), 18S ribosomal RNA (rRNA) (PPM57735E), Ribosomal protein L32 (RPL32) (PPM03300B). For qRT-PCR analysis, each reaction (25 μL) included 12.5 μL of 2X SABiosciences RT² qPCR Master Mix (SYBR Green), 1 μL of diluted first-strand cDNA synthesis reaction, and 11.5 μL of de-ionized H₂O. qRT-PCR was performed with the ABI 7500 Fast Real-Time PCR System (Applied Biosystems, Inc., Foster City, CA, USA) using the following program: step 1: 95°C, 10 minutes; step 2: 40 cycles of 95°C, 15 s and 60°C, 1 minute; step 3: dissociation curve 95°C, 1 minute; 65°C, 2 minutes (optics off); 65 to 95°C at 2°C per minute (optics on). The data were analyzed using online software from SABiosciences [26]. For RANKL RNA, the comparative cycle threshold (CT) method was used to calculate the fold change in gene expression after normalization to the values for 18S rRNA and RPL32 RNAs. RNA from the number 4-inguinal mammary glands of three animals was analyzed from each treatment group. In the tumor analysis, the three HFD-E tumors were adenosquamous (ER-), ductal (ER+) and cribriform (ER-) adenocarcinomas with tumor onset at 10, 14 and 17 weeks post-DMBA, respectively. The three LFD tumors were adenosquamous (ER-), papillary (ER+) and cribriform (ER-) adenocarcinomas with tumor onset at 31, 34 and 38 weeks post-DMBA, respectively.

Genes that were significant (unadjusted p-value <0.05) in univariate analyses were evaluated for ontological enrichment using Ingenuity Pathway Analysis (IPA, Ingenuity® Systems, Redwood City, CA, USA), with Benjamini-Hochberg (B-H) multiple testing correction. The background was set to include only the genes on the qPCR arrays analyzed for each sample group. Significant functions and pathways were defined as those with at least two significant genes per pathway and with B-H P-values less than 0.05.

Plasma levels of 20 different cytokine proteins (AREG, basic fibroblast growth factor (bFGF), epidermal growth factor (EGF), hepatocyte growth factor (HGF), insulin-like growth factor 2 (IGF-2), IL-6, leptin, receptor activator of nuclear factor kappa-B ligand (RANKL), vascular endothelial growth factor (VEGF), resistin, TNFα, IL-1α, IL-1β, insulin-like growth factor binding protein (IGFBP)-2, IGFBP-3, prolactin, T-cell activation (TCA)-3 (CCL1), macrophage colony stimulating factor (M-CSF), epiregulin, and osteoprotegerin (OPG)) were determined by cytokine Ab arrays (Cat. number: AAM-CUS-G; RayBiotech, Norcross, GA, USA), following the manufacturer’s instructions. Briefly, the glass array slide was incubated with blocking buffer for 30 minutes at RT, and then incubated with diluted mouse plasma (30 μL plasma diluted to 120 μL with blocking buffer) overnight at 4°C with slow shaking. Then, after thorough washing with washing buffer, the array slide was incubated with biotin-labeled Ab for 2 h at RT with slow shaking. After thorough washing with washing buffer, the array slide was incubated with HiLytePlus™ 555 Fluor-conjugated streptavidin for 2 h at RT with slow shaking. Then, after thorough washing, the slide was dried and absolute fluorescent intensity measured by an Agilent G2505B laser scanner (Agilent Technologies, Inc., Santa Clara, CA, USA). The data were analyzed by software provided by the company. Additionally, plasma IGF-1 levels were measured using a mouse IGF-1 ELISA kit (ELM-IGFI-001; RayBiotech), according to the manufacturer’s directions.

The PCR arrays were statistically analyzed using proprietary software from SABiosciences. Correlations between cytokine mRNA levels and leukocyte populations were determined by Spearman’s method [27]. Otherwise, results are expressed as mean ± standard error of the mean (SEM). Differences were considered significant at P <0.05 using the Student t-test or analysis of variance (ANOVA) followed by the Tukey multiple comparison test, as appropriate. Tumor incidence was analyzed by the Chi-square test. Tumor latencies were determined from Kaplan-Meier plots.

A Kaplan-Meier plot (Figure 1A) shows that HFD-fed mice developed tumors significantly earlier than LFD-fed mice (P = 0.01) with a significantly reduced median time to tumor onset (204 days LFD versus 115 days HFD; P = 0.00075). Although HFD caused a 1.8-fold increase in tumor incidence (LFD, 14.95%, n = 87 mice versus HFD, 26.3%, n = 95 mice; P = 0.059), this did not reach statistical significance. Notably, there were no significant differences in body weight (see Additional file 2) or in parametrial fat pad weights (data not shown) between animals fed HFD and LFD over the course of the tumor study.

Since decreased latency was the major difference between the two diets, we focused on investigating the basis for the latency difference. To that end, we focused our analysis on comparing early developing tumors on HFD (HFD-E) tumors with LFD tumors. Representative HFD-E tumors were selected from tumors that developed before the earliest observed LFD tumors, at 19 weeks or less post-first DMBA treatment. Conversely, representative LFD tumors were selected after 19 weeks or more post-first DMBA treatment, when LFD tumors were first detected. We considered the possibility that reduced latency in HFD-fed mice might be due to accelerated development and increased numbers of atypical hyperplastic, pre-cancerous lesions. Time-course analysis showed that, indeed, HFD-fed mice developed significantly more hyperplastic lesions per mouse after 8 and 14 weeks post-first DMBA treatment (10 and 16 weeks on diet, respectively), times before palpable tumors were detected (Figure 1B). The hyperplastic lesions comprised ductal and alveolar hyperplasia for both diets (Figure 1C). There were no significant differences in tumor types or ER status (Table 1).

We examined mammary epithelial cell and tumor cell proliferation, as determined by BrdU incorporation, at 10 weeks on diet and in tumors (Figure 2A). These measurements were compared to our earlier report of HFD-enhanced proliferation at 4 weeks on diet [18]. At both 4 and 10 weeks on diet, there were significant increases in epithelial cell proliferation in HFD-fed mice, with the highest proliferation seen at 10 weeks in normal mammary structures and hyperplastic lesions. Proliferation was also significantly greater in HFD-E tumors, but was less than that seen after 10 weeks on HFD.

Since tissues and tumors require adequate vascularization to sustain proliferation, we also measured angiogenesis, as determined by blood vessel density (Figure 2B, C). At 3 and 4 weeks on diet, blood vessel density was low and independent of diet. At 10 weeks on diet, HFD-fed mice exhibited significantly increased blood vessel density around normal mammary gland structures and foci of atypical hyperplasia compared with LFD-fed mice. Notably, blood vessel density was highest around microscopic, non-palpable tumors in HFD-fed mice. HFD-E tumors also exhibited increased blood vessel density.

To gain further insight into the enhanced proliferative characteristics observed on HFD, we analyzed RNA expression by qRT-PCR using PCR arrays targeted to Growth Factors (Table 2). At 3 weeks, Fgf18 and Il4 were upregulated, and at 4 weeks, brain derived neurotrophic factor (Bdnf) and leukemia inhibitory factor (Lif) were upregulated, whereas Lefty2 was downregulated. Interestingly, at 4 weeks on HFD, only RANKL was robustly upregulated 17-fold. RANKL is a P-regulated paracrine growth factor in the mammary gland [28‐30], and HFD-induced RANKL upregulation could contribute to the increased proliferation observed at 4 weeks on diet. There was no specific overlap in mammary gland gene expression at 3 or 4 weeks on diet.

At 10 weeks on diet, a large number of growth factor genes were upregulated in mammary glands of HFD-fed mice. The one with most obvious relevance to mammary epithelial proliferation was TGFα[31], a ligand for the EGF receptor (EGFR). Other upregulated genes included several encoding TGFβ superfamily growth factors [32] (that is, Bmp2, Bmp3, growth differentiation factor (Gdf) 5, Gdf10, Nodal, and Tgfb1), as well as genes encoding IL-2, IL-7, FGF10, PGF, VEGF-A, Chemokine (C-X-C motif) ligand (CXCL) 12, IL-1α, and IL-1β. Interestingly, there was no specific overlap with growth factors found induced at 3 and 4 weeks on HFD.

Analysis of growth factor-related gene expression in the HFD-E tumors in comparison to LFD tumors showed upregulation of bone morphogenic protein (Bmp) 7, and inhibin α (Ihha), as well as a trend toward upregulation of Fgf15 (P = 0.09). Neurotrophin 3 (Ntf3) was most robustly upregulated (53-fold). Bmp10 was downregulated. It is striking that the HFD-E tumors display a unique gene expression profile that shows no overlap with the genes identified at 3, 4, and 10 weeks on diet.

Leukocytes (that is, macrophages, eosinophils and mast cells) play important roles in normal pubertal mammary gland proliferation and development [9‐11]. Since HFD promoted mammary epithelial cell proliferation, we also examined the effect of diet on leukocyte recruitment (Figure 3). At 3 weeks, there was a significant HFD-induced increase in recruitment of eosinophils to all mammary structures (Figure 3A) and of mast cells to large ducts (Figure 3B). At 4 weeks, this effect was no longer seen; however, high levels of eosinophils were observed around TEBs on both diets (Figure 3A). Macrophage recruitment to glandular structures was similar at 3 and 4 weeks, and was independent of diet (data not shown). The majority of macrophages were of the M2 phenotype (that is, Arg1+) for both diets.

At 10 weeks on diet, HFD-fed mice had significantly more macrophages recruited to all normal mammary gland structures (Figure 3C). With both diets, the vast majority of recruited macrophages were M2-polarized. The greater number of macrophages recruited at 10 weeks on HFD did not appear to be due to an interaction with DMBA because the same fold increases in macrophages were detected at 10 weeks on HFD diet without DMBA treatment (data not shown). There were exceedingly low numbers of eosinophils and mast cells on either diet (data not shown). There were significantly more M2 macrophages present within HFD-E tumors and within their tumor stroma (Figure 3D). There were no significant differences in the numbers of eosinophils and mast cells associated with HFD-E or LFD tumors (data not shown).

To gain a better understanding of how HFD increased leukocyte recruitment, we compared the expression of inflammation-related genes in LFD and HFD mammary glands by qRT-PCR using the Inflammatory Cytokines and Receptors PCR array (Table 3). There was no overlap in inflammation-related mRNA expressed at 3, 4, and 10 weeks on diet. Genes encoding CCL24 (eotaxin-2), CCL3 (MIP-1α), IL-4, and IL5Rα were transiently induced at 3 weeks. RANKL was the only upregulated inflammatory gene observed at 4 weeks.

At the 3- and 4-week diet time points, mice also had one inguinal mammary gland surgically cleared of endogenous mammary epithelium (cleared fat pad), whereas the contralateral gland was kept intact (intact gland). This was done to identify effects that can be mediated in the fat pad independently of epithelial cells. Analysis of mammary gland inflammatory gene expression showed that upregulated genes only overlapped with that in intact glands for Il5ra at 3 weeks (Table 4). At 4 weeks, however, the cleared fat pads showed robust upregulation of Il1f6 (Il36α) (6.5-fold) and Il1f8 (Il36β) (4.9-fold). These genes were not upregulated in intact glands.

The Inflammatory Cytokines and Receptors PCR array identified 10 genes that were upregulated by HFD at 10 weeks on diet (Table 3). The upregulated genes were Il1b, Il2rb, lymphotoxin α (Lta), and a number encoding immune chemokines and chemokine receptors: chemokine (C-C motif) ligand (Ccl) 1, Ccl17, Ccl19, Ccl20, Ccl22, chemokine (C-motif) receptor 1 (Xcr1), and chemokine (C-X-C motif) receptor (Cxcr) 2. Similar to the Growth Factor gene analysis, there was no overlap between the genes found to be modulated at 3, 4, and 10 weeks on diet.

To determine the properties of HFD-E tumors that could explain their reduced latency and their enhanced proliferative characteristics, we also analyzed RNA expression by qRT-PCR using Breast Cancer, and Cancer Pathway Finder PCR arrays. This analysis (Table 5) showed upregulation of cyclin D2 (Ccdn2), insulin-like growth factor 1 receptor (Ifg1r), telomerase reverse transcriptase (Tert), Slit2, and β-catenin (Ctnnb1) expression, whereas keratin 8 (Krt8) and keratin 18 (Krt18) were expressed at lower levels. Discordant with the pro-tumorigenic properties of many genes upregulated in HFD-E tumors, several upregulated genes are known tumor suppressors (Brca2[33], Slit2[34], Trp53[35], Apaf1[36], and Brca1[33]). Nonetheless, it is striking that the HFD-E tumors displayed a unique gene expression profile in comparison to LFD tumors.

In ontology analyses (Table 6), few canonical pathways were altered at 3 weeks and 4 weeks, and among those pathways that met statistical significance criteria, few genes per pathway were significantly altered (see Additional files 7, 8). By 10 weeks, a much broader range of genes were altered and several statistically significant pathways were identified, including Granulocyte and Agranulocyte Adhesion and Diapedesis (P = 5.66 × 10^-9), Hepatic Fibrosis/Hepatic Stellate Cell Activation (P = 1.55 × 10^-6), Altered T Cell and B Cell Signaling in Rheumatoid Arthritis (P = 3.24 × 10^-6), and Role of Macrophages, Fibroblast and Endothelial Cells in Rheumatoid Arthritis (P = 3.32 × 10^-6) (see Additional file 9). The ontology analyses also identified several statistically significant pathways altered in the HFD-E tumors. These include Molecular Mechanisms of Cancer (P = 7.1 × 10^-7), p53 Signaling (P = 7.1 × 10^-7), Basal Cell Carcinoma Signaling (P = 1.27 × 10^-5), GADD45 Signaling (P = 1.43 × 10^-5), and Role of NANOG in Mammalian Embryonic Cell Pluripotency (P = 4.54 × 10^-5) (see Additional file 10).

As shown in Additional file 3, there were no major differences in non-fasting blood insulin or blood glucose levels at 10 weeks on LFD or HFD, or in HFD-E and LFD tumor-bearing mice. Mice fed LFD and HFD had glucose levels in the normal non-fasting range for BALB/c mice (<319 mg/dL) (see Additional file 3: Figure S2A,C) [37]. Non-fasting serum insulin levels were similar at 10 weeks between LFD and HFD, with some values in the hyperinsulinemic range (normal non-fasting insulin = 0.77 ng/mL) (see Additional file 3: Figure S2B) [37]. Insulin levels in HFD-E and LFD tumor-bearing mice varied widely and generally were in the hyperinsulinemic range (see Additional file 3: Figure S2D). The high insulin values were most likely indicative of some degree of insulin resistance. There were no differences in E or P levels at 4 or 10 weeks on HFD or LFD, or in tumor bearing mice (data not shown).

Plasma levels of additional growth and inflammatory factors were measured by Ab arrays to investigate potential systemic effects of HFD exposure (Table 7). At 4 weeks on HFD, the levels of OPG and IGF-1 were significantly elevated (1.3-fold, P = 0.04 and 1.3-fold, P = 0.005, respectively); this increase was transient and not seen at later time points, or in tumor-bearing mice. At 10 weeks on HFD, the only significantly modulated systemic protein was prolactin, of which there was decreased plasma expression (0.68-fold, P = 0.03). However, there was a trend toward increased IL-1α levels (7.8-fold, P = 0.066). In HFD-E tumors, HGF was significantly increased (2.1-fold, P = 0.046). We compared mammary glands for mRNA expression of the factors detected in the plasma samples. The possible increase in systemic IL-1α parallels the significant increase in mammary gland Il1a RNA expression observed at 10 weeks on HFD (Table 3). Notably, mRNA encoding OPG, IGF-1, HGF, and prolactin were not detected at any time point in mammary gland or tumors, suggesting alternative systemic sources for their presence in plasma.

We previously reported that pubertal BALB/c mice fed HFD for 4 weeks showed increased mammary epithelial cell proliferation [18]. Herein, we observed that HFD induced elevated mammary gland expression of several growth factor genes as early as 3 and 4 weeks on diet. In particular, RANKL was robustly induced at 4 weeks. RANKL is a P-induced paracrine factor that has known mitogenic activity in the mammary gland [38‐40]. Plasma levels of OPG, a decoy receptor and RANKL antagonist [41], were also elevated at this time, and this may reflect a physiologic response to excess mammary RANKL levels. Additionally, plasma levels of IGF-1 were elevated at this time. Systemic IGF-1 and mammary gland expression of RANKL are plausible early sources of enhanced proliferation prior to the appearance of tumors in HFD-fed BALB/c mice [28‐30, 42, 43]. IGF-1 is an essential growth factor for TEB formation [42] that has been implicated in breast cancer progression [43].

By 10 weeks on HFD, a time point prior to the appearance of palpable tumors, mammary glands showed increased numbers of abnormal mammary epithelial lesions, and enhanced cellular proliferation. Increased levels of the RNA encoding growth factors associated with mammary epithelial cell proliferation and development were detected. Perhaps most provocative among these was the gene encoding TGFα, an EGFR ligand that has been associated with precocious alveologenesis, delayed involution, and mammary tumorigenesis [21]. Interestingly, it was recently reported that paracrine EGFR signaling between tumor-associated macrophages and murine breast cancer cells can promote a cancer stem cell-like phenotype [44]. Also provocative was the increased expression of a number of genes in the TGF-β superfamily (Bmp2 and 3, Gdf5 and 10, Nodal, and Tgfb1) in HFD-fed mice. Signaling by TGF-β1 family members mediate embryonic development, tissue homeostasis and regeneration, immune responses, tumor suppression, and metastasis, as well as govern the behavior of many stem cell populations [45‐48]. TGF-β1 has an established role in mammary gland ductal development [49‐51]. The Nodal signaling pathway is activated by Cripto-1, which encodes a growth factor with a role in mammary gland development that is capable of inducing ductal hyperplasia [52]. Lefty2, which was downregulated at 4 weeks, functions as an antagonist of the Nodal pathway [53], and this earlier downregulation may set the stage for activation of Nodal signaling at 10 weeks. Specific roles for Bmp3 and Gdf10 in the mammary gland have not been reported. Increased Tert expression is also associated with the immortalization of cells, and as such, it antagonizes apoptosis [54]. This may also contribute to a proliferative phenotype. Indeed, 95% of human cancers show increased Tert activity [55]. Upregulation of genes encoding IL-2 and IL-7 has been associated with some classes of breast tumors [56, 57]; these were also upregulated in mammary glands of 10-week HFD-fed mice.

Analysis of HFD-E tumor characteristics showed increased cellular proliferation that was associated with upregulation of genes that promote cellular proliferation such as Bmp7, Ccdn2, Inha, and Igf1r. BMPs are recognized as key regulators during the control of cell fate and cancer development [58], and BMP signaling can downregulate levels of mitotic checkpoint components in human breast cancer cells [59]. Bmp7 has also been implicated in increasing the metastatic potential of 4 T1 mouse mammary tumor cells [60]. There was also a trend toward upregulation of Fgf15 (P = 0.09). FGFs and their receptors control a wide range of biological functions operative in cancer including cellular proliferation [61]. On the other hand, Bmp10, which has been shown to decrease aggressiveness of breast cancer cells [62], was downregulated in HFD-E tumors. Thus, decreased Bmp10 expression is consistent with greater cellular proliferation of HFD-E tumors. Expression of IGF-1R is known to be increased in human breast cancers and associated with increased cellular proliferation [63‐66]. Inhibins are growth factors that are also involved in cell proliferation and differentiation, and inhibin α is expressed in normal mammary tissues and in human breast cancers; however, its specific role in tumors is not well understood [67]. The increased expression of Ccdn2 is discordant with enhanced tumorigenesis of HFD-E tumors, as loss of cyclin D2 expression is frequent in breast cancers [68]. However, transgenic overexpression of cyclin D2 does block lobuloalveolar development [69] and, perhaps, Ccdn2 overexpression in our system could be viewed as suppressing differentiation.

Several genes that are activated in response to DNA damage (Brca1, Brca2, and Trp53) were upregulated in HFD-E tumors. Trp53 is a downstream target of Brca1[70], and Apaf1, an important mediator of apoptosis that is a downstream target of Trp53, is also upregulated [71]. Thus, the upregulation of these genes appears to constitute the upregulation of a pathway associated with DNA damage, perhaps resulting from exposure to the mutagenic carcinogen DMBA, rather than the coincident overexpression of several mutated tumor suppressor genes. It is plausible that Trp53 upregulation of Apaf1 is related to removal of damaged cells by apoptosis [71]. While DMBA treatment was dissociated in time from the occurrence of tumors, DMBA is known to induce aneuploidy and unstable karyotypes that might sustain DNA damage long after exposure [72]. Additionally, expression of Krt8 and 18, associated with more differentiated mammary luminal cells [73, 74], was significantly decreased, suggesting a more aggressive phenotype of HFD-E tumors. In regard to increased Ctnnb1 expression, the Wnt/β catenin pathway is involved in normal mammary gland proliferation and development, and associated with poor prognosis in breast cancer [75]. Elevated Ctnnb1 (β-catenin) expression may activate this pathway.

HGF was elevated in the plasma of HFD-E tumor-bearing mice, and may play a role in driving HFD-E tumor growth. HGF regulates multiple cellular processes that stimulate cell proliferation, invasion, and angiogenesis, both in the normal mammary gland [76] and in breast cancer [77]. HGF is generally produced locally within the mammary gland/mammary cancer and acts in a paracrine manner. However, no HFD-induced increase in mammary gland or tumor expression of HGF was observed in the present study. Recently, HGF serum levels were reported to be significantly elevated with increasing tumor stage in breast cancer patients, raising the possibility of alternate sites of HGF production [78].

The integrated overview of the various aforementioned genes that is provided by pathway analysis tends to support the notion that genotoxic stress has activated p53 and Gadd45 signaling, two pathways associated with DNA damage [79, 80]. The altered expression of Trp53, Ccnd2, Apaf1, Brca1, and Ctnnb1 identified the p53 signaling and GADD45 signaling pathways to high significance. Provocatively, altered expression of a partially overlapping set of genes, Trp53, Bmp7, Ctnnb1, and Bmp10, identified the Basal Cell Carcinoma signaling and Role of NANOG in Mammalian Stem Cell Pluripotency pathways. This suggests similarity between the HFD-E tumors and breast tumors displaying stem and/or progenitor cell characteristics, such as basal-like breast cancer [81]. The downregulation of luminal epithelial makers, Krt8 and 18, in HFD-E tumors is also consistent with this. Although several of the HFD-E tumors were ER-positive, basal-like breast cancers can express ERα [82]. Further, epidemiological studies have associated basal-like breast cancer with increased abdominal adiposity [83]; although HFD-E tumor-bearing animals were not obese, obesity was associated with HFD.

It is noteworthy that a number of genes associated with angiogenesis (Gdf5, Nodal, Pgf, Vegfa, and Cxcl12) were upregulated by HFD at 10 weeks, concomitant with the observation of increased vascularity in the mammary glands of HFD-fed mice. Nodal expression is significantly elevated in malignant human breast cancers versus benign breast disease [84] and has been associated with vasculogenic mimicry, defined as functional plasticity of aggressive cancer cells forming de novo vascular networks [85]. GDF5 has also been shown to have angiogenic properties [86]. PGF and VEGF-A, angiogenic proteins of the VEGF family, are upregulated mainly in pathologic conditions, such as breast cancer, and are associated with poor prognosis [87]. CXCL12, produced by stromal fibroblasts within invasive breast cancers, promotes angiogenesis [88] and has additionally been associated with increased macrophage density in tumors [89]. Coincident with the induction of the above mentioned genes associated with angiogenesis, analysis of plasma at this time revealed decreased prolactin levels. Prolactin is reported to have anti-angiogenic effects [90] and its reduced expression may promote angiogenesis.

HFD-E tumors also exhibited increased blood vessel density and genes associated with angiogenesis, such as Ntf3, Bmp10, and Slit2, were also differentially regulated in these tumors. In fact, Ntf3 was the most highly upregulated growth factor gene (53-fold). Neurotrophins and the neurotrophin receptor p75 are expressed in human breast cancers and are implicated in promoting angiogenesis, tumor growth, invasion, and resistance to apoptosis [91]. Interestingly, in mice fed 60% HFD, neurotrophin expression is increased in the brain, suggesting its upregulation here in HFD-E tumors may also be diet-induced [92]. Further, we found that Bdnf, another member of the neurotrophin family of growth factor genes [93], was robustly induced at 4 weeks on HFD, perhaps suggesting an association between this family of growth factors and HFD.

The induction of Ccl24 (eotaxin-2) and Ccl3 (MIP-1α) RNAs at 3 weeks was concomitant with influx of eosinophils into the mammary peri-epithelial compartment. Both CCL24 [94] and CCL3 [95] are potent chemoattractants for eosinophils. Il5ra RNA is also elevated at this time; IL-5 is a differentiation factor for eosinophil progenitors [96]. Consistent with chemotaxis of eosinophils to adipose tissue, upregulation of Il4 RNA has also been observed [97]. Ccl24, Ccl3, and Il4 mRNAs all had high correlation with the level of eosinophils at 3 weeks on diet (Ccl24: r = 0.865 to 0.886; Ccl3: r = 0.667 to 0.812; Il4: r = 0.714 to 0.943 (depending upon mammary structural element); see Additional file 11). The induction of Il4 RNA has potentially important implications for the function of macrophages in the mammary gland, as eosinophil-derived IL-4 is essential for the maintenance of alternatively activated (that is, M2) macrophages in adipose tissue [97]. This is consistent with the high numbers of Arg1⁺ macrophages observed in the peri-epithelial stroma of the normal gland at 3 and 4 weeks on diet, and in the vicinity of early lesions and within tumors in DMBA/HFD-treated mice. While pathway analysis did not identify any canonical pathways to a high level of significance, chemotaxis of eosinophils was identified as a highly significant function (P = 6.4 × 10^-5) (see Additional file 7), and eosinophil levels correlated well with RNAs for all of the aforementioned chemokines and cytokine (see Additional file 11). Another inflammatory growth factor RNA increased in mice at 4 weeks was that encoding RANKL, a potent activator of NF-κB [98]. Thus, RANKL may play a role in the upregulation of pro-inflammatory factors at this time. In parallel to the analysis of whole intact mammary gland at 3 and 4 weeks of diet, a similar analysis of epithelium-devoid fat pads contralateral to the intact mammary glands was carried out. At 3 weeks, this revealed that Il5ra was similarly induced by HFD in both cleared fat pads and intact glands, suggesting that it is, in fact, associated with the stroma. At 4 weeks, this analysis also revealed HFD induction of the genes encoding IL-1f6 (IL-36α) and IL-1f8 (IL-36β), IL-1 family cytokines known to be expressed in adipocytes as inflammatory mediators [99]. This induction is not observed in intact glands. It may be that the mammary epithelium suppresses this induction in intact glands, or, alternatively, that the stromal induction of these cytokines is obscured in the intact gland by dilution or because the same cytokines are downregulated in the epithelium, while being upregulated in the stroma. This highlights the mammary stroma as an epithelium-independent source of inflammatory factors that can mediate the effects of HFD.

At 10 weeks on diet, the RNA encoding a number of cytokines and chemokines associated with immune function were modulated by HFD, and these may have profound implications for the status of the immune system at this point in time. Most interesting are the genes encoding the chemokines CCL1, CCL17, and CCL22, as well as the cytokine TGF-β1, which are all associated with the recruitment and function of immunosuppressive Treg cells [100‐102]. There is also one report of CCL20 recruiting Treg cells in an induced colorectal cancer model [103]. Further, these factors are all products of M2 macrophages, recruitment of which was induced by HFD at 10 weeks of diet. In fact, TGF-β1 additionally can function in the generation of immunosuppressive M2 macrophages [104], perhaps, in part, through the suppression of M1 macrophage activity [105]. Upregulation of Ccl19, Ccl20 and chemokine (C-X-C motif) receptor (Cxcr) 2 has also been associated with breast tumor growth and invasion [106‐108], Cxcr2 particularly with angiogenic processes [108]. Systemic levels of prolactin were also reduced at 10 weeks on HFD. As prolactin can suppress the function of suppressor T cells [109], a decrease in its expression would be expected to enhance Treg activity. Indeed, increased numbers of Treg cells have been associated with increased DMBA-induced mammary carcinogenesis in mice [110], suggesting the importance of Treg-associated chemokines in this experimental system. Characterization of the influence of dietary fat upon T cell populations awaits future studies.

Pathway analysis comparing mice on HFD to those on LFD at week 10 identified Granulocyte and Agranulocyte Adhesion and Diapedesis (identified by Il1a, Cxcr2, Ccl17, Ccl20, Il1b, Ccl22, Ccl19, and Ccl1), which is consistent with leukocyte migration into the mammary gland. Furthermore, immune cell trafficking (adhesion of immune cells and leukocyte migration) was among functions showing the highest significance at 10 weeks on HFD (P = 1.64 × 10^-13 and 2.03 × 10^-13, respectively). Pathway analysis also identified Altered T Cell and B Cell Signaling in Rheumatoid Arthritis (identified by altered expression of Il1a, Il2, Tgfb1, Lta, and Il1b) and Role of Macrophages, Fibroblasts and Endothelial cells in Rheumatoid Arthritis (identified by altered expression of Vegfa, Il1a, Tgfb1, Lta, Il1b, Il7, and Pgf) (see Additional file 9), pathways that can lead to macrophage-mediated pathology [111] and abnormal angiogenesis [112]. The identification of the Hepatic Fibrosis/Hepatic Stellate Cell Activation pathway (identified by Vegfa, Il1a, Tgfb1, Tgfa, Il1b, and Pgf) is interesting, as the polarization of M2 macrophages is a critical component of this pathology [113]. The levels of RNA encoding IL-1b, a common factor in all of these pathways, correlated well with the level of macrophages in the mammary gland (r = 0.828 to 0.886 (depending upon mammary structural element); see Additional file 11) suggesting the plausibility that these pathways may operate, at least in part, through macrophage recruitment. The genes included on the PCR arrays were highly selected for genes with established involvement in particular pathways (that is, growth factors, inflammation). Thus, it is also noteworthy to observe that the absolute number of significant genes on each targeted pathway array increased between week 4 and week 10 (Figure 4). The parallel increase in expression of inflammation and growth factor genes underscores that immune infiltrates can alter the expression of cytokines with regulatory effects on mammary epithelial proliferation.

The status of inflammatory processes also differed in HFD-E tumors compared to LFD tumors. There was a significant increase in intra-tumoral and stromal alternatively activated M2 (Arg1+) macrophages. The M2 phenotype is associated with tumor-associated macrophages that are known to promote the growth of tumors through support of angiogenic and tissue remodeling processes, as well as immune suppression [114]. The increase in M2 macrophages is consistent with the enhanced vascularity of HFD-E tumors in comparison to LFD tumors. While none of the cytokines identified as expressed in HFD-E tumors is known to generate M2 macrophages, Il4 is upregulated at 3 weeks and Tgfb1 is upregulated at 10 weeks on HFD, and these cytokines may be involved in the prior polarization of M2 macrophages [104]. There is compelling evidence at the molecular level that many cancers, including breast cancer, are linked to a dysregulated inflammatory response [115]. The present results that demonstrate HFD-induced inflammatory processes involving M2 macrophages are consistent with a role for these macrophages and their secreted chemokines/cytokines in the promotion of HFD-E tumors.

As for the role of HFD, there is evidence that it can induce low-grade inflammation after feeding [116]. It has been proposed that HFD induces post-prandial gut permeability allowing low levels of bacterial endotoxin from the gut to enter the circulation and, thus, induce low-grade inflammation [116, 117]. There is also evidence that saturated fatty acids can directly modulate inflammatory processes through toll-like receptor 4 (TLR4) [118]. Palmitic acid, an abundant component of lard [119], has particularly been implicated in TLR2 [120, 121] and TLR4 signaling [120‐123]. This stimulation has been associated with increased IL-1 signaling [120]. This is consistent with our observation of increased levels of Il1a and Il1b gene expression after 10 weeks on HFD.

The levels of metabolism-related factors such glucose, insulin, leptin, and resistin were not significantly altered by HFD. However, an inflammatory growth factor RNA that increased in HFD mice at 4 weeks was that encoding LIF, an IL-6-class cytokine that can stimulate lipolysis and fat loss [124‐126]. This places LIF as a plausible mediator contributing to the lack of weight gain in BALB/c mice placed on HFD. At 10 weeks on HFD, plasma IL-1α levels were increased. Since HFD also increased mammary gland IL-1α mRNA expression at 10 weeks, increased plasma levels might reflect spillover from the mammary gland or other tissues affected by HFD. Palmitic acid has been associated with increased IL-1 signaling [120], and elevation of serum IL-1α has been reported in HFD-induced obesity in mice [127]. Tumors showed a trend toward increased Fgf15 expression. FGF15-specific signaling is thought to control metabolic homeostasis associated with HFD to restore glucose tolerance and insulin sensitivity [128].

Other studies of the effect of dietary fat on normal mammary gland development and tumorigenesis have been carried out in rodents [129‐131]. However, they differ significantly from the current studies with regard to species studied (rat versus mouse), differences in fat formulations (corn oil versus lard), caloric content (isocaloric versus non-isocaloric), obesogenic versus non-obesogenic outcomes, development of metabolic syndrome, and age at diet initiation. More similar studies on the effect of HFD on tumor development without the confounding effects of obesity have been performed in other mouse mammary cancer models. Results vary by tumor model and age at diet initiation. In two studies of the effects of HFD initiated at 4 weeks of age in mice overexpressing HER2/Neu in the mammary gland [14, 15], HFD promoted tumor development by increasing tumor incidence, but without increasing tumor cell proliferation; there was only a small increase in body weight and no insulin resistance or hyperinsulinemia. In another study, HER2/Neu transgenic mice fed HFD starting in adulthood, at 10 weeks of age, showed no difference in tumor latency, incidence, or metastasis [132]. In a tumor transplant model, mice were started on HFD at 4 weeks of age, and after 16 weeks on diet, 4 T1 mammary carcinoma cells were transplanted into their mammary glands [133]. There were only slight increases in body weight with HFD. However, tumor weight and number of metastases were significantly increased by HFD. Thus, based on the results of the present study and of the previous studies of HFD initiated in pubertal mice (4 weeks of age), there was a significant promotional effect on tumor development with only a modest effect on weight gain or metabolic parameters. This is in contrast to the lack of a promotional effect of HFD when initiated in adult, 10-week-old mice. The most extensive study was carried out in the 4 T1 tumor transplant model with diet started at 4 weeks of age [133]. Similar to our present results, that study revealed an association of HFD with increased macrophage infiltration, angiogenesis, and cellular proliferation, as well as increased levels of a number of inflammatory factors.