Dietary polyunsaturated fatty acids modulate adipose secretome and is associated with changes in mammary epithelial stem cell self-renewal

Abstract

Chronic low-grade adipose inflammation, characterized by aberrant adipokine production and pro-inflammatory macrophage activation/polarization is associated with increased risk of breast cancer. Adipocyte fatty acid composition is influenced by dietary availability and may regulate adipokine secretion and adipose inflammation. After feeding F344 rats for 20 weeks with a Western diet or a fish oil-supplemented diet, we cultured primary rat adipose tissue in a three-dimensional explant culture and collected the conditioned medium. The rat adipose tissue secretome was assayed using the Proteome Profiler Cytokine XL Array, and adipose tissue macrophage polarization (M1/M2 ratio) was assessed using the iNOS/ARG1 ratio. We then assessed the adipokine’s effects upon stem cell self-renewal using primary human mammospheres from normal breast mammoplasty tissue. Adipose from rats fed the fish oil diet had an ω-3:ω-6 fatty acid ratio of 0.28 compared to 0.04 in Western diet rats. The adipokine profile from the fish oil-fed rats was shifted toward adipokines associated with reduced inflammation compared to the rats fed the Western diet. The M1/M2 macrophage ratio decreased by 50% in adipose of fish oil-fed rats compared to that from rats fed the Western diet. Conditioned media from rats fed the high ω-6 Western diet increased stem cell self-renewal by 62%±9% ($\overline{X}\%\pm \mathrm{SD}$) above baseline compared to only an 11%±11% increase with the fish oil rat adipose. Modulating the adipokine secretome with dietary interventions therefore may alter stromal-epithelial signaling that plays a role in controlling mammary stem cell self-renewal.

1. Introduction

In the past decade, chronic inflammation and dysregulated cytokine signaling have been implicated in the development of many cancers, including breast cancer [1–4]. Recently, adipose tissue has been identified as a major signaling organ that modulates inflammation [5]. Adipose tissue secretes cytokines and hormones, termed adipokines, that play a role in regulating inflammation and other metabolic functions [6–8]. Adipokines also have pleiotropic effects on both normal and tumor cells [9]. Therefore, understanding the mechanisms by which adipokines may promote carcinogenesis and how to modulate adipokine expression and secretion may lead to effective chemopreventive interventions.

Dietary ω-3 fatty acids, such as eicosapentaenoic acid (EPA; 20:5ω-3), docosahexaenoic acid (DHA; 22:6ω-3) and α-linolenic acid (ALA; 18:3ω-3), and their potential chemopreventive effects have been studied extensively, with mixed results [10–13]. Dietary ω-3 fatty acids facilitate weight loss, increase adiponectin synthesis and reduce adipose inflammation [14–18]. Furthermore, fatty acids directly and indirectly modulate macrophage activation and polarization [19,20]. Specifically, DHA shifts the adipose tissue macrophages from the proinflammatory M1-state into the anti-inflammatory M2-state by decreasing Notch1/Jagged1 signaling [20,21].

High proportions of saturated fatty acids and low proportions of ω-3 polyunsaturated and monounsaturated fatty acids in serum are associated with obesity and lead to low circulating and cell membrane ω-3:ω-6 fatty acid ratios [22]. High dietary ω-6 fatty acid intake alone decreases membrane ω-3 fatty incorporation into phospholipids [23,24]. Conversely, lowering ω-6 fatty acid intake alone increases membrane ω-3 fatty acids [25].

Membranous fatty acids are cleaved by phospholipase A2 and serve as substrates for local cellular cyclooxygenases (COX), lipoxygenases and cytochrome P450 that produce downstream eicosanoids, prostaglandins (PGs), hydroxyeicotetraenoic acids, leukotrienes and epoxyeicosatrienoic acids [26,27]. For example, arachidonic acid (AA; 20:4ω-6), an ω-6 polyunsaturated fatty acid (PUFA) commonly high in a Western diet, serves as the substrate for the catalytic cyclooxygenase monomer of prostaglandin H synthases-1 and 2 (commonly known as COX) to form PGH₂, a product upstream of the proinflammatory eicosanoid PGE₂, whereas the corresponding the ω-3 fatty acid EPA is catalyzed into the anti-inflammatory PGE₃ via COX and prostaglandin E synthase [28–30]. EPA reduces PGE₂ synthesis in COX-1 and competes for the COX-2 catalytic site. For both isoforms, EPA effectively reduces PGE₂ synthesis [31–34]. These dietary fatty acid substrates, especially AA and EPA, play an important role in controlling tissue eicosanoid concentrations and subsequent downstream effects, which include local generation of cytokines, chemokines and adipokines [35].

One potential target of adipokine signaling is the population of normal mammary stem cells that may give rise to the cancer cell of origin [36–40]. We have previously found that an increase in the leptin/adiponectin ratio increases normal breast stem cell self-renewal in vitro [41]. A larger stem cell pool may also be associated with increased breast cancer incidence by increasing the number of potential targets for mutation and tumor formation [42]. Here, we developed an adipose tissue explant culture system using an F344 rat model to test the hypotheses that dietary ω-3 fatty acid supplementation, rich in EPA and DHA, shifts the adipokine milieu toward a less inflammatory phenotype. We also utilized an in vitro mammosphere assay to test the hypothesis that this shift in adipokines has the potential to reduce mammary stem cell self-renewal compared to a Westernized diet.

2. Materials and Methods

2.1. Animals and diets

All animal protocols for this experiment were approved by the University Committee on Use and Care of Animals at the University of Michigan. Male F344 rats (5 weeks old) were purchased from Harlan Laboratories (Haslett, MI, USA). Pelleted Western blend control diet (ω-3:ω-6 fatty acid ratio=0.05) and menhaden fish oil diets (ω-3:ω-6 fatty acid ratio=0.4) were prepared by Dyets Inc. (Bethlehem, PA, USA). Supplemental Table 1 shows the composition of the two diet arms; diets were equivalent other than the PUFA ratio and included 34% of calories from fat. The Western blend consisted of 45% coconut oil, 30% olive oil, 15% corn oil and 10% soybean oil by weight and no EPA. Menhaden fish oil contains mainly monounsaturated and saturated fats, similar to the Western blend fat, but the polyunsaturated fatty acids are mainly ω3 (only 4% ω6 fatty acids). Menhaden fish oil was stabilized against oxidation with a combination of mixed tocopherols and combined with the Western blend to achieve a 0.4 ω-3: ω-6 ratio as previously described [43]. Fish oil dose was chosen for a larger study using these rats [44] and the based upon data from our published dose response of dietary ω-3: ω-6 ratios and their impact upon blood and colonic tissue PGE₂ reduction [43]. Pelleted food was stored at −80°C, and small amounts were thawed at 4°C prior to feeding. Food in the animal cages was replaced on a daily basis to avoid lipid oxidation. The rats were given water ad libitum throughout the experiment and maintained on a 12-h light/dark cycle per standard animal care protocols.

2.2. Study timeline

Twelve male 5-week-old rats were used in this experiment and given 1 week to acclimate. Rats were randomly divided into two groups (six animals per diet arm) and provided with either the control diet composed of Western blend oil or the experimental diet supplemented with menhaden fish oil for 20 consecutive weeks. Body weight was recorded at baseline and then weekly. Animals were euthanized using carbon dioxide inhalation after 26 weeks of age.

2.3. Tissue collection and adipose tissue explant culture

Retroperitoneal visceral adipose depots were removed from each animal. Tissue was visually inspected, and grossly visible blood vessels and fascia were removed with forceps. Tissue was minced with scalpels, and 500 mg of total adipose from each animal was immediately placed on dry ice for fatty acid extraction and analysis. For protein extraction, adipose tissue was homogenized in lysis buffer (HEPES, NaCl₂, glycerol, triton-X, protease inhibitor cocktail). Total protein concentration was measured with the Bradford Assay (Bio-Rad Laboratories Inc., Hercules, CA, USA), and all samples were standardized to 10 mg/ml total protein prior to storage at −80°C.

The remaining adipose tissue was placed in explant culture, as previously described [41]. Briefly, tissue was partially digested with 0.05% trypsin/EDTA and Dispase in HBSS for 10 min at 37°C with gentle mixing to partly disrupt the extracellular matrix and allow for nutrient and oxygen diffusion in culture while still maintaining normal three-dimensional tissue architecture (Supplemental Fig. 1A–B). Tissue was rinsed twice with HBSS and once with warm M199 media. The partially dissociated adipose was divided into 0.5 g aliquots and placed in 40-μm cell strainers and inserted into a 6-well tissue culture plate with 4 ml of adipose culture media (M199, 1 μg/ml insulin, 400 ng/ml dexamethasone, and 1× PCN/Strep). Cultures were maintained at 37°C with 5% CO₂ for 48 h prior to collecting conditioned media. Adipocyte viability was determined qualitatively by intracellular enzymatic conversion of MTT to formazan (ATCC, Manassas, VA, USA; Supplemental Fig. 1C). Subsequent follow-up determined that these tissue explants can be maintained in culture with periodic media changes for at least 10 days with >95% adipocyte viability. Adipose conditioned media was centrifuged at 1000g to pellet any cells or debris. Because we wanted to examine the adipokine secretome from different animals without confounding variables, we removed free fatty acids, free amino acids, exogenous insulin, glucose and other micronutrients supplied by the base culture medium by concentrating the conditioned media with Amicon Ultra-15 centrifugal filter units (10-kDa cutoff; Sigma-Aldrich, St. Louis, MO, USA). Media was then sterilized with a 0.2-μm syringe filter. Total protein was measured with the Bradford Assay and standardized to 10 mg/ml prior to storage at −80°C.

2.4. Confocal imaging

Freshly collected adipose tissue was fixed overnight in 10% formalin. Tissue was then washed three times in PBS/0.1 M glycine and permeabilized with 0.2% Triton-X for 15 min before blocking in PBS/1% BSA. Neutral fatty acids were stained with BODIPY^493/503 (2 μg/ml; Life Technologies, Carlsbad, CA, USA), and actin filaments were stained with phalloidin–AlexaFluor⁵⁴⁶, and nuclei stained with DAPI. Stained tissue was rinsed in PBS and mounted onto a coverslip using a Nunc Labtek II chamber system (Thermo Scientific, Waltham, MA, UAS) with Prolong Gold. The stained tissue was then visualized with a Leica Inverted SP5X Confocal Microscope System (University of Michigan Medical School Biomedical Research Core Facility).

2.5. Adipose tissue fatty acid analysis by gas chromatography–mass spectrometry

The adipose tissue was homogenized in 900 μl of PBS. Adipose homogenate (200 μl) was used for fatty acid analysis. To quantify the incorporated fatty acids in the adipose tissue, 10 μl of internal standard (17:0, 1 mg/ml in hexane) was added to the 200 μl of adipose homogenate, which equates to approximately 70 mg of adipose tissue. The samples were then extracted with 1.5 ml of Folch reagent (chloroform/methanol 2:1), vortexed for 2 min and centrifuged (200g for 5 min). The organic layer was removed to a 10×75-mm glass tube and dried in a SpeedVac. The samples were solubilized in 150 μl of hexane/chloroform (1:1) and vortexed. Fatty acid methyl esters were prepared by adding 10 μl of METH-PREP II derivatization reagent [0.2 N methanolic (m-trifluoromethylphenyl) trimethylammonium hydroxide; Alltech, Deerfield, IL, USA]. The samples were assayed using a gas chromatography-mass spectrometry (GC–MS) method as previously described using selected ion monitoring [45]. To quantify, the mass of each fatty acid peak was calculated against an internal standard. Total fatty acid mass was calculated as the sum of individual fatty acid masses, then normalized per milligram of adipose tissue and summed into a total fatty acid concentration. After determining the concentrations of the individual fatty acids, they were converted to a molecular % of total fatty acids measured.

2.6. Assessment of adipocyte hypertrophy

The degree of adipocyte hypertrophy was determined by light microscopy. Adipose tissue from animals in each diet was pooled into a Western diet plate and a fish oil plate. Images of fresh partially dissociated adipose tissue were taken at 20× magnification using an inverted tissue culture microscope (Fig. 1A). Five hundred eight individual adipocyte diameters, per diet group, were measured with ImageJ software to calculate cell volume $(V=\frac{4}{3}{r}^3)$.

Fig. 1.

Dietary fish oil structurally alters visceral adipocytes in rats. (A) Pooled adipose tissue examined by light microscopy and adipocyte volume quantified with ImageJ. (B) Left: Kernel density of adipocyte volume by frequency (n=508 adipocytes per diet group). Right: Cumulative distribution of adipocytes by diet group. Adipocyte volume distribution is shifted to smaller volumes in the fish oil diet group. (C) Adipose tissue fatty acids were quantified by GC–MS and displayed as a percentage of total by mass. The tissue ω-3:ω-6 PUFA ratio was 0.04 in the animals fed the Western diet (n=6) and increased to 0.28 in the animals fed the fish oil diet (n=6; P<.001).

2.7. Assessment of macrophage polarization

To assess the macrophage population, total adipose tissue lysate was assayed by ELISA for CD68 (pan-macrophage marker; LS Bio, Seattle, WA, USA), inducible nitric oxide synthase (iNOS; marker of M1 polarization; Neo Scientific, Cambridge, MA, USA), arginase-1 (Arg1; M2 polarization marker; LS Bio) and fatty acid binding protein-4 (FABP-4; control for total adipocyte mass; Neo Scientific). Original concentration was assayed as mass in ng per unit volume in microliters and converted to nanomolar concentration using known molecular weights of the proteins measured. The ratio of CD68 concentration/FABP4 concentration was used as a measure of the total macrophage population in adipose tissue. iNOS and Arg1 concentrations were ratioed to the measured CD68 concentration assayed via ELISA kit. The iNOS/Arg1 ratio has been used as a surrogate for M1/M2 polarization, as previously described [46,47].

2.8. Adipokine measurements

Adipose explant conditioned media was assayed with the Proteome Profiler Cytokine XL Array (R&D Systems) according the manufacturer’s protocol. Resulting blots were exposed to film, which were scanned with a flatbed scanner, and pixel intensity for each adipokine was measured with ImageJ. Each animal adipose conditioned media specimen was assayed on a separate array. Median pixel intensity of each cytokine was calculated for the rats grouped by diet (n=6 per diet) and then used to determine statistical differences between the two diet groups using the Wilcoxon rank sum test. Group median pixel intensity for each cytokine was used to calculate fold differences between each diet, which were then log₂ transformed. Data are presented as ${\log }_2[\frac{\text{Fish}\text{O}{\mathrm{il}}_i}{{\text{Western}}_i}]$.

2.9. Normal human mammary tissue collection and epithelial cell isolation

Human mammary tissue was obtained from healthy women undergoing elective reduction mammoplasty surgery at the University of Michigan after giving informed consent (University of Michigan IRBMED approved tissue collection protocol). The specimens were carefully reviewed by a pathologist through the Tissue Procurement Service and the Department of Pathology at the University of Michigan Medical School to ensure the absence of malignancy. The reduction mammoplasty tissue from 15 de-identified women ranging in age from 18 to 48 years $(\overline{X}=32.8\pm 9.5\text{years})$ was minced with scalpels and separately digested with collagenase overnight. The digested tissue was centrifuged at 40g to pellet the “organoids” containing intact mammary glands, and the top fat layer and fibroblasts were discarded as previously described [40]. The separated organoids were then stored in liquid nitrogen until the cells were to be pooled. To create the mammary epithelial cell pool, the organoids were thawed and dissociated enzymatically and mechanically to generate a single-cell suspension. Cells were counted with a hemocytometer, and viability was assessed by trypan blue exclusion. Because we had no demographic information other than age for our mammoplasty donors and to control for intersubject variability, 1.0×10⁷ live epithelial cells from each of the 15 individual specimens were then pooled to create a standardized mix for use in all subsequent experiments; aliquots were stored in liquid nitrogen.

2.10. Mammosphere assay

Using qualitative immunohistochemistry for the mammary stem cell marker, Aldh1a1, in primary human mammospheres, we previously demonstrated that the quantitative mammosphere formation assay is a valid tool to assess the rate of stem cell self-renewal [41]. To conduct the mammosphere assay, 1.8×10⁴ viable pooled epithelial cells were plated into each well of a 24-well ultralow attachment culture plate (Corning Inc., Corning, NY, USA) with 400 μl of Mammocult Media (Stem Cell Technologies, Vancouver, Canada). Triplicate wells were supplemented with adipose conditioned media from either a Western diet-fed (n=6) or a fish oil-supplemented (n=6) rat (standardized to 1 mg/ml final protein concentration) or nonconditioned (n=3) control media (M199 with 1 mg/ml BSA) during the period of primary mammosphere formation. Media was refreshed every 2–3 days, and the primary spheres were counted and collected after 10 days. Primary mammosphere formation efficiency (MFE) was calculated as the percentage of mammospheres counted over the total number of cells seeded in the culture dish, and mean MFE $(\overline{\text{MFE}})$ was calculated for both diet groups and control. MFEs in the two experimental diets were normalized to the control by subtracting the ${\overline{\text{MFE}}}_{\text{control}}$ from the ${\overline{\text{MFE}}}_{\text{Western}}$ and ${\overline{\text{MFE}}}_{\text{FO}}$. As we previously described [41], the primary mammospheres were dissociated with 0.05% trypsin/EDTA and FACSmax (Genlantis, San Diego, CA, USA); all viable cells were replated in a new ultralow adherent culture plate in Mammocult media. Media was refreshed every 2–3 days, and secondary mammospheres were counted after 10 days. The secondary mammosphere formation assay is a high-throughput and reliable in vitro assay that allows for the relatively quick and simple interrogation of changes in stem cell self-renewal activity in response to specific or novel agents. The percent difference between primary (1°) and secondary (2°) mammosphere number is a measure of changes in stem cell self-renewal activity: $\%\text{change}=[\frac{\left(\#2^{°}\text{spheres}-\#1^{°}\text{spheres}\right)}{\#1^{°}\text{spheres}}]\times 100\%$[48]. The percent change in the two experimental diets was normalized to the negative control by subtracting the ${\overline{\%\text{Change}}}_{\text{control}}$ from the ${\overline{\%\text{Change}}}_{\text{Western}}$ and ${\overline{\%\text{Change}}}_{\text{FO}}$.

2.11. Statistics and data analysis

All data, unless otherwise noted, are expressed as $\overline{X}\pm \text{SD}$ (n=6 animals per diet group). The predetermined upper limit of probability for statistical significance determined by the nonparametric Wilcoxon rank sum test was P<.05. Statistical differences in MFE and stem cell self-renewal were determined using the two-sample t test after confirmation of normality using the Shapiro-Wilk’s test (data not shown). Differences in adipocyte volume by group were determined using the nonparametric Wilcoxon rank sum test, and differences in adipocyte volume distribution by group were determined using the Kolmogorov–Smirnov test in R. Analysis of ELISA data and mammosphere data was conducted using GraphPad Prism and R (San Diego, CA, USA). Differences in mean pixel intensity from the cytokine arrays were assessed by nonparametric Wilcoxon rank sum tests, and adjustments for multiple comparisons were performed using the Benjamini–Hochberg method in R. Statistical significance cutoff for fatty acid molecular percentage was determined as P≤003 by Bonferroni correction, and cutoff for total fatty acid % and ω3:ω6 ratio was determined by P≤.01 by Bonferroni correction.

3. Results

3.1. Dietary fish oil supplementation reduces adipocyte hypertrophy

At the end of this 20-week study, animals fed the control Western Blend diet weighed an average of 441±35 g, whereas those fed the fish oil diet weighed 441±28 g (P=.88; data not shown). Adipocyte volume is an indicator of the degree of lipid storage and potential tissue hypoxia. It has been suggested to serve as one surrogate marker of tissue inflammation since it positively correlates with CRP, MCP-1 and IL-6 levels in healthy humans [49–51]. Images of pooled adipose tissue were taken (Fig. 1A), and adipocyte volume was assessed using ImageJ. Median adipocyte volume was reduced by 19% in the pooled fish oil group (median, 0.83) compared to the pooled Western Blend diet group (median, 1.03). Adipocyte volumes were shifted to a lower distribution in the fish oil fed diet compared to the Western blend diet (Fig. 1B).

3.2. The fish oil diet decreases total adipose fatty acids and modulates individual fatty acid composition compared to the Western blend diet

To confirm tissue incorporation of the dietary fatty acids, total fatty acids were extracted from fresh adipose tissue in each animal. Fifteen individual fatty acids were measured by GC–MS. After 20 weeks of dietary fish oil supplementation, the adipose tissue ω-3:ω-6 PUFA ratio increased to 0.28 from 0.04 in the rats fed the fish oil diet compared to rats fed the Western diet (P<.001; Fig. 1C and Table 1). The percentage of adipose tissue ω-3 fatty acids increased to 3.63%± 0.31% (median, 3.59%) after 20 weeks of dietary fish oil supplementation from 0.55%±0.06% (median, 0.56%) in the Western Blend diet animals (P<.001). Thirteen of 15 fatty acids measured were statistically different between the fish oil-supplemented and the Western diet groups, including EPA [0.72%±0.08% (median, 0.72%) vs. 0.03%± 0.02% (median, 0.03%), respectively] and DHA [1.48±0.16% (median, 1.46%) vs. 0.05%±0.01% (median, 0.05%), respectively; see Table 1 in Ref. [52]). The total fatty acid concentration in adipose was reduced to 39±15 μg/mg (median, 33.42 μg/mg) adipose tissue in the fish oil diet rats from 104±55 μg/mg (median, 104.55 μg/mg) in the control Western blend diet animals, although insignificantly after Bonferroni correction (P=.026).

Table 1.

Composition of rat adipose tissue ($\overline{\text{Molecular}\%}\pm \text{S}.\text{D}.$, median molecular %)

Fatty acid	Western blend	Fish oil
Total ω3 (%)	0.55±0.06, 0.56	3.63±0.31, 3.59 *
Total ω6 (%)	14.58±0.48, 14.56	13.01±0.14, 13.01 *
Total other (%)	84.87±0.47, 84.84	83.36±0.40, 83.51 *
Total (μg/mg)	104.03±54.64, 104.44	38.97±15.23, 33.42
ω3:ω6	0.038±0.004, 0.039	0.28±0.02, 0.28 *

^*P≤.01 by Bonferroni adjustment.

3.3. Adipose tissue macrophages from fish oil rats were shifted toward the anti-inflammatory M2 phenotype

Macrophage infiltration into adipose tissue may be considered a marker of tissue inflammation. Macrophages are recruited to necrotic adipocytes, form crown-like structures and alter the microenvironment [53]. Two distinct polarized states of macrophage activation have been described: the classically activated proinflammatory (M1) and the alternatively activated anti-inflammatory (M2) macrophage phenotype, although several subtypes of M2 have been described [54–56]. Typically, M1 polarized macrophages preferentially express iNOS, whereas M2 polarized macrophages are identified by upregulation of Arg1. Thus, the iNOS/Arg1 ratio in adipose tissue may be considered as a surrogate marker of the average macrophage polarization state [57–59].

As shown in Fig. 2A, the adipose tissue CD68/FABP4 ratio was the same in animals from both diet groups (P=.48) suggesting no change in the total tissue macrophage population. However, they were significantly shifted toward an anti-inflammatory M2 phenotype as measured by a significant increase in the median Arg1/CD68 ratio (Arg1/CD68=2.732 in control animals vs. 4.795 in experimental animals, P=.002; Fig. 2B) and subsequent decrease in the median ratio of iNOS/Arg1 in the adipose tissue from animals fed the high ω-3:ω-6 PUFA diet (iNOS/Arg1=0.808 in control animals vs. 0.357 in the experimental animals, P=.002; Fig. 2C). Taken together with the reduction in adipocyte hypertrophy, a lower M1/M2 macrophage polarization ratio suggests decreased adipose tissue inflammation [54].

Fig. 2.

Dietary fatty acids influence adipose tissue macrophage polarization. (A) The CD68/FABP4 ratio in total adipose tissue lysate. There was no statistical difference in macrophage population in the two diet groups (P=.31). (B) The macrophage M1/M2 polarization ratio was estimated by the iNOS/Arg1 ratio in total adipose tissue lysate. Animals in the fish oil diet had a 58% reduction in the M1/M2 polarization ratio compared to animals in the Western blend diet group (0.67 and 0.28, respectively). *P≤.05.

3.4. Replacing dietary ω-6 rich oils with ω-3 rich fish oil shifts the adipose secretome toward molecules associated with a less inflammatory state

Adipose conditioned media is enriched for a variety of known adipokines and cytokines as demonstrated by protein array. Dietary fish oil supplementation shifted the adipose secretome molecular products measured on this assay toward those products associated with an anti-inflammatory milieu (Fig. 3 and Table 2 in Ref. [52]). For example, the fish oil adipose conditioned media had a fourfold increase in the anti-inflammatory adiponectin and significant decreases in proinflammatory factors such as leptin, resistin, C-reactive protein, and the proinflammatory biomarker, lipocalin-2 [60–63]. These data suggest that increasing the dietary ω-3:ω-6 PUFA ratio skews the adipokine secretome toward products associated with reduced inflammation.

Fig. 3.

The dietary ω-3:ω-6 PUFA ratio modulates the adipose tissue secretome. (A) Representative Proteome Profiler Cytokine XL protein arrays (R&D Systems) for a rat from the Western diet arm showing relative adipokine concentrations in adipose conditioned media. (B) Representative Proteome Profiler Cytokine XL protein arrays (R&D Systems) for a rat from the fish oil diet arm showing relative adipokine concentrations in adipose conditioned media. (C) Volcano plot displaying log₂-transformed fold cytokine changes (x-axis) between the rats fed the fish oil diet and the Western diet and the negative log₁₀ Benjamini–Hochberg adjusted P values from the Wilcoxon rank sum tests (y-axis). A vertical line at x=0 indicates no change in cytokine concentration between diets. A dotted line at −log₁₀(0.05) represents the cutoff for statistical significance. Fold change was determined by dividing the median intensity of adipokine signal for the fish oil-fed rats (n=6) conditioned media sample by the median intensity of adipokine signal for Western diet fed rats (n=6).

3.5. Factors in adipose conditioned media differentially modulate human mammary stem cell self-renewal

Primary mammary epithelial cells from 15 separate human specimens were pooled in equal amounts and plated in nonadherent conditions and allowed to form mammospheres as an assay of stem cell activity (each grown in triplicate). Primary cultures were exposed to adipose conditioned media from animals fed either the Western Blend or fish oil diets (standardized to 1 mg/ml total protein), or nonconditioned control media (with 1 mg/ml BSA). Primary sphere formation was unchanged in the Western diet group ($\overline{X}=35.3\pm 1.86$, $\overline{\text{MFE}}=-0.004\%\pm 0.004\%$, P=.57) and in the fish oil group ($\overline{X}=33.9\pm 5.3$, $\overline{\text{MFE}}=-0.011\%\pm 0.012\%$, P=.54) when compared to the controls that were incubated in media that was not conditioned by adipose ($\overline{X}=36\pm 1$; Fig. 4A).

Fig. 4.

Adipose-derived factors differentially modulate human mammary stem cell self-renewal. (A) Pooled mammary epithelial cells were grown in nonadherent cultures to form mammospheres and treated with adipose tissue conditioned media from rats fed either fish oil or the Western diet, or negative control media. Primary MFE was unchanged between diet groups and the vehicle control, when normalized to the spheres treated with the negative control media (M199+1 mg/ml BSA). (B) Primary mammospheres were dissociated and replated to form secondary mammospheres as a measure of stem cell self-renewal. The change in secondary mammospheres over primary mammospheres was normalized to the negative control; graph shows the mean % change±S.E.M. Factors present an adipose conditioned media from animals fed the standard Western Blend diet increased the rate of stem cell self-renewal by 61.6%±9.03% $(\overline{X}\%\pm \text{S}.\text{D}.)$ compared to the negative control (P<.001), while factors in the adipose secretome from animals fed a fish oil-supplemented diet attenuated this effect with a mean increase in stem cell self-renewal of only 11.4%±10.59% compared to the negative control (P=.12).*P≤.05.

Primary mammospheres were then counted, dissociated and replated to form secondary mammospheres. Since a self-renewing stem cell is expected to form a mammosphere and each mammosphere is formed by a single stem cell, the difference between secondary and primary mammospheres may be considered a measure of stem cell self-renewal activity and is expressed as a percent change, compared to baseline negative control. When the mammary stem cells from 15 human samples were treated with adipose conditioned media from animals fed the low ω-3:ω-6 diet, there was a 50%–74% increase $(\overline{X}=62\%\pm 9\%)$ in the rate of stem cell self-renewal compared to the negative control (P<.001; Fig. 4B). In contrast, adipose conditioned media from animals fed the high ω-3:ω-6 diet promoted only an 11%±11% (range, 6.6%–24.6%; P=.12) increase in the rate of stem cell self-renewal. The mammary stem cell self-renewal was significantly higher in the adipose conditioned media from low ω-3:ω-6 compared to media conditioned from the adipose of the high ω-3:ω-6 diet (P<.001).

4. Discussion

Published data demonstrate the association between the obesity-associated metabolic syndrome, its changes in specific circulating adipokines and their association with cancer risk [64–66]. We used breast epithelial stem cells from normal human mammoplasty procedures as a model of a carcinogenesis target based upon our data and those of others suggesting that these stem cells may be the tumor initiating cells. Using this cellular end point, we used fatty acid dietary modification to assess shifts in the balance of a profile of adipokines associated with proinflammatory and anti-inflammatory activity and its effects on mammary epithelial stem cells.

Omega-3 fatty acids (EPA and DHA) from the dietary fish oil supplement incorporate into visceral adipose after 20 weeks and reduce the adipose mass while not changing the total body mass. The rats fed the dietary fish oil supplement had decreased total concentrations of 13 fatty acids per gross adipose tissue weight, which were associated with decreased adipocyte size compared to the Western diet rats. We infer that the adipokine profile and adipocyte size are consistent with the physiology of the obesity-associated metabolic syndrome. We found that the relationship between the total body mass of the adipose compartment, adipocyte volume and fatty acid concentrations is not linear. We suggest that total body mass represents shifts to other storage depots, for example, lipid and perhaps conversion to nonfatty tissues such as muscle.

Among the limitations of our study design were the lack of measurement of total lipid content. The discrepancies between weight of the rats, reduced adipocyte volume and reduced fatty acid concentrations could be due to unmeasured fatty acids or shifts in lipid profiles. Another limitation of the design was pooling the adipose and measurement of adipose diameter and volume on the pooled adipose as opposed to individually from each rat. As a preliminary assessment of culture viability and adipocyte size, we pooled a portion of adipose tissue into the different diet groups. We were still able to assess differences in adipocyte volume distributions between the two diet groups and report changes in the median; however, to properly assess the effects of fish oil on adipocyte size, independent measurements of individual rats in each diet group are needed.

Using the adipokine arrays, we demonstrate that dietary intervention with ω-3 fatty acids alters the adipose secretome, shifting it toward products associated an anti-inflammatory activity. Of the 111 adipokine proteins in the array, we interrogated in the adipose conditioned media, we identified important changes in 46. While we were able to identify some proteins known to enhance or delay carcinogenesis process (e.g., leptin, osteopontin, adiponectin, IL-10, Flt-3 ligand, CXCL-1), other proteins including some with large magnitudes of change after incubation of adipose conditioned media (e.g. RBP4) are not known to play a role in the carcinogenesis process. The data provided by adipokine arrays are semiquantitative and observational; yet, they provide important leads that are likely to identify new mechanisms by which key nutritional components such as fatty acids control inflammation at the tissue level. Elucidating such mechanisms may support new strategies to ameliorate the inflammation component of obesity associated metabolic syndrome and will be an important direction for future research.

Adipose conditioned media from animals fed a typical Western style diet with a low ω-3:ω-6 PUFA ratio increased the size of the stem cell pool compared to the a balanced rat chow diet control by promoting symmetric self-renewal. This effect was much less pronounced with the adipose conditioned media of rats fed the fish oil-supplemented diet. The factors in the adipose conditioned media from Western diet-fed rats that stimulate mammary stem cell self-renewal are likely less abundant in the conditioned media from the rats fed the fish oil. For example, the adiponectin concentration was fourfold higher in the fish oil group conditioned media compared to the Western diet conditioned media. Previously, we found that recombinant human adiponectin reduced human mammary stem cell self-renewal as measured by secondary sphere formation when compared to leptin or control media [41]. Conditioned media from adipose presents a complex mixture of both stimulatory and repressive factors that impact mammary stem cell self-renewal and sphere formation. The data presented here can be considered a balance of these complex mixtures that, in the case of fish oil conditioning, reduces mammary stem cell self-renewal compared to Western diet conditioning. Future work will dissect these factors and their mechanisms of action to elucidate the associations we observe here. While this in vitro effect was statistically significant, the clinical significance of a different rate of stem cell self-renewal needs more in-depth translational research.

Results from multiple in vivo studies suggest that ω-6 PUFAs accelerate tumorigenesis, whereas dietary ω-3 PUFAs may have anticancer effects [67,68]. Our observational data suggest that fatty acids derived from dietary sources play an important role in controlling local cellular stem cell homeostasis. In the human breast, data from our group [69,70] and of others [71,72] suggest that the size and regulation of mammary stem cell/progenitor pools are associated with cancer risk and progression. These data are also supported by the biological similarities between normal and cancer stem cells [73,74]. The ratios of ω-3 to ω-6 fatty acids may indirectly regulate the size of the self-renewing stem cell pool via shifts in local adipokines that serve as signaling molecules to the adjacent epithelium. Future identification of mechanisms by which fatty acids contribute to the regulation of the stem cell niche, the plasticity of breast stem cell self-renewing (ALDH⁺) and quiescent pools (CD44⁺/CD24⁻), or maturation to progenitor pools may ultimately be used to individualized dietary supplementation or modification to reduce cancer risk.

Obesity is associated with increased adipocyte size resulting in necrotic adipocytes that recruit local macrophages. This pathology has been recently identified by the detection of crown-like structures, macrophages that surround necrotic adipocytes, and are identified by the macrophage cell marker, CD68 [75,76]. Surprisingly, we did not see any difference in the CD68/FABP4 ratios between the two diet groups (Fig. 1A). Recent data suggest that ω-3 fatty acid diet decreases the number of crown-like structures per adipocyte in humans with metabolic syndrome [77]. Due to the limited adipocyte supply from the rats we studied, we were unable to further characterize the macrophage populations. Rather, we created an adipose tissue lysate to quantify macrophage phenotypes. This shift in ω-3:ω-6 fatty acid ratio in adipose was associated with a qualitative change in the types of macrophage infiltration—a statistically significant increase in the anti-inflammatory M2 macrophage population compared to the control Western blend fed group (Fig. 1C).

Current data on the effects of ω-3 fish oil supplementation on macrophage infiltration are conflicting suggesting either macrophage infiltration decreases [77,78] or that there is no change [79]. Our data may clarify this issue by suggesting that the reduction in adipose driven inflammation by ω-3 fatty acids may, in part, be driven by shifting the macrophage population toward the M2 phenotype as opposed to the quantitative estimates of macrophage infiltration. We recognize that these data require further exploration. For example, adipocytes also express both CD68 and FABP4. Expression of these proteins is altered in dysfunctional adipose tissue [80,81]. However, the CD68 macrophage marker gene expression is over 10-fold higher in the stroma-vascular fraction compared to the adipocyte [82]. Moreover, because the F344 rat model is not considered a fully developed model of obesity, we suggest that the expression of CD68 and FABP4 we detected is more likely signal from macrophage as opposed to dysfunctional adipose.

Given the increased cancer risk associated with obesity and obesity-associated inflammation [83,84], these and similar studies may lead to new potential therapeutic interventions that reduce the inflammatory component of the metabolic syndrome. This is of particular importance given the recent findings that the metabolic disturbances associated with obesity also occur in normal-weight people [85]. Studies show that use of nonsteroidal anti-inflammatory drugs, such as aspirin, decreases the risk of developing colorectal cancer; however, the risks of gastrointestinal toxicity and other side effects limit the benefits of nonsteroidal anti-inflammatory drug use [86]. In our study, omega-3 fatty acids composed of 3.7% of the total calories in the fish oil diet. This equates to about 8 g of fish oil daily in a diet consisting of 2000 kcal/day. Such a large dose would be difficult for humans to consume on a daily basis. In a recent publication, we developed a dosing model using serum EPA/AA ratios as a surrogate biomarker for PGE₂ tissue responses [87]. Using a mathematical dosing model generated initially with rodent data and then enriching with human data as the trial progressed using dynamic dosing adjustments, we found that ω-3 fatty acid dosing in humans is more potent than in rats in reducing colonic mucosal PGE₂. We found that a mean daily ω-3 fatty acid dose of 5.5 g, a dose that is clinically well tolerated, in normal-body-weight humans (body mass index <25 kg/m²) is sufficient to modulate important colonic mucosal eicosanoids during small cancer prevention clinical trials [87]. An appropriately designed clinical trial that addresses the obese population directly and identifies and quantifies biomarker end points reflective of obesity-associated metabolic syndrome would be useful to personalize ω-3 doses required to resolve human adipose inflammation and adipokine secretion. Dietary manipulation to enhance ω-3 fatty acids pools alone or in combination with ω-3 fatty acids supplementation provides an alternative, nontoxic anti-inflammatory therapeutic strategy that can be generalized to large at-risk populations at minimal cost. Moreover, such strategies may be personalized in the future to individual doses using biomarkers of the local tissue inflammatory status such as PGE₂ or circulating adipokine profiles.

In summary, we demonstrate that secreted adipokines from rats fed a Western style diet increases the pool of self-renewing breast epithelial stem cells, a likely carcinogenesis target, while adipokines from rats fed a fish oil-supplemented diet did not. Moreover, the omega-3-rich diet shifts the macrophage population in rats as well as secreted adipokines toward a less inflammatory profile, suggesting that a fish oil diet decreased adipose inflammation. These data are releavant to the adipokine microenvironment that is anatomically contiguous in the breast stroma. This environment likely plays an important paracrine, as well as endocrine, role in stromal-epithelial signaling in the breast epithelium that is crucial in carcinogenesis.

Abbreviations

EPA

eicosapentaenoic acid

DHA

docosahexaenoic acid

ALA

α-linolenic acid

COX

cyclooxygenase

AA

arachidonic acid

PUFA

polyunsaturated fatty acid

GC–MS

gas chromatography–mass spectrometry

iNOS

inducible nitric oxide synthase

Arg1

arginase-1

FABP-4

fatty acid binding protein-4