Interactions between Adipocytes and Breast Cancer Cells Stimulate Cytokine Production and Drive Src/Sox2/miR-302b–Mediated Malignant Progression

Obesity is associated with a more aggressive course and greater mortality for several cancers (1–3). Breast cancers in obese individuals show increased inflammation, hypoxia, angiogenesis, and changes in endocrine factors, including insulin, IGF-1, leptin, adiponectin, and estrogen (4). Obese tissue adipocytes upregulate chemokines and leptin, which attract macrophages that produce proinflammatory proteins, including TNFα and IL1β, and proangiogenic factors, providing a microenvironment favorable for adipose tissue expansion, but also propitious for tumor progression (2).

Obese individuals frequently have elevated proinflammatory cytokine levels in serum (3, 5). High-serum cytokines are also associated with poor breast cancer patient outcome (6–8). Because autocrine and paracrine cytokine pathway activation can promote oncogenesis and tumor metastasis (9), we investigated how interactions between invading breast cancer and adipose cells could alter the potential for aggressive tumor behavior and metastasis.

The cancer stem cell (CSC) model posits that tumors are generated and maintained by a subpopulation of cells that self-renew and yield heterogeneous progeny with reduced proliferative potential (10). Increasing evidence indicates that CSC underlie drug resistance tumor recurrence and metastasis (10). Deregulation of pathways that control normal embryonic stem cells (ES) may contribute to breast cancer–initiating cell expansion (11, 12). ES-transcription factors (ES-TF) Oct4, Sox2, Klf4, Nanog, and Myc are master regulators of ES self-renewal whose enforced expression in somatic cells yields induced pluripotent stem cells (13). ES-TFs govern expression of short non-coding miRNAs that modulate gene expression programs (14, 15). miR-302 drives ES self-renewal and its loss predicates differentiation (16). miR-302b overexpression on its own can induce pluripotency in somatic cells (17, 18). Mechanisms governing ES-self-renewal are often subverted by cancer cells. Src kinase is not only critical for maintaining ES pluripotency and proliferation, but is also a key mediator of oncogenesis (11, 19). Src is activated by proinflammatory cytokines and can mediate malignant transformation (9).

Although tissue niche influences ES self-renewal (11), little is known of how cancer cell interactions with the tissues they invade promote cancer progression. When breast cancer cells transgress the mammary duct basement membrane, they encounter fibroblasts, inflammatory cells, and adipocytes. Activated fibroblasts and tumor-associated macrophages (TAM) are known to drive breast cancer invasion (20, 21). This work indicates that the interaction between breast cancer cells and adipocytes, which occurs early in breast cancer invasion, markedly upregulates proinflammatory cytokines. These increase the abundance of stem cell–like cells that initiate highly metastatic breast cancers. Furthermore, these cytokines activate Src to upregulate Sox2 and induce miR-302b, generating feed-forward loops that contribute to mammary CSC self-renewal and drive metastatic tumor progression.

MDA-MB-231, MCF-7, and T47D were mycoplasma free, authenticated by short tandem repeat profiling, and passaged for <6 months. HMLER3 and BPLER3 were cultured as in ref. 22. Dissociated tumor cells, DT28, and carcinoma-associated fibroblasts (CAF23) were isolated from primary triple-negative breast cancers (23) and used at early passage. Human mammary fibroblasts (HMF; ScienCell) were cultured as per the manufacturer's instructions. Human adipocyte stem cells (hASC) and mature adipocytes were isolated from human abdominal fat from cytoreduction surgery and breast fat obtained at mastectomy (three samples each), collected with informed consent and Institutional Review Board approval (see Supplementary Methods). Immature adipocytes were differentiated from hASC or MIAMI cells (bone marrow derived, human mesenchymal stem cells) using adipogenic differentiation media for 4 weeks as in refs. 24 and 25. hASC, in vitro differentiated (immature), and mature adipocytes were cocultured with breast cancer cells over 7 days (details in Supplementary Methods). For conditioned medium assays, MDA-MB-231 was cultured in immature adipocyte conditioned media over 1 week. Cytokines were used at 10 ng/mL (100 ng/mL for IL8; R&D Systems) for 7 days, unless otherwise indicated, and media renewed every 2 days.

After 7 days of monoculture or coculture, media were collected and cytokine concentrations assayed by an xMAPBio-Plex cytokine array (Bio-Rad Life Sciences) using a Luminex 200 plate reader (Biosource, Invitrogen; ref. 26).

For orthotopic limiting dilution assays, MDA-MB-231-luc-GFP cells were flow sorted from immature adipocytes after 7 days coculture, and 10, 100, or 1,000 viable GFP⁺ cells were suspended in 100 μL Matrigel and injected into the mammary fat pad of 4-week-old Balb/c nude mice (n = 8–10/group).

To analyze metastasis from primary tumors, tumors formed from 10⁴ MDA-MB-231-luc-GFP cells were excised at 1 cm and metastatic tumor bioluminescence monitored by Xenogen IVIS weekly (27).

For tail vein injections, 2 × 10⁵ MDA-MB-231-luc-GFP cells were injected and monitored by IVIS weekly (27). Lungs were photographed and surface tumors counted.

Primary tumors, metastatic lymph nodes, or lung tumors were excised and pooled for each experimental group and single-cell suspensions seeded for sphere assays as in ref. 27. All animal research was University of Miami Animal Care Committee approved.

Circulating tumor cells (CTC) were recovered and quantitated as in ref. 28.

Src inhibitors, 4-amino-5-(4-methylphenyl)-7-(tbutyl) pyrazolo[3,4-d]pyrimidine (PP1, Sigma) or saracatinib/AZD0530 (from Astra Zeneca), each at 1 μmol/L, were added for 48 hours after 5 days of cytokine or adipocyte exposure, then RNA was extracted or spheres were seeded. For Western blot analyses, see Supplementary Methods. Src inhibitors were added 20 minutes before cytokine treatment, and proteins extracted at intervals for Western blot analyses.

At least three different siRNA oligonucleotides to SOX2 and scrambled controls (Cell Signaling Technology) were used per the manufacturer's instructions. SOX2 knockdown was confirmed by Western blot analysis 48 to 72 hours after transfection.

RNA was isolated from MDA-MB-231 with or without exposure to 10 ng/mL IL6, CCL5, or IP10 for 7 days. RNA integrity number was 9.3 to 10. miRNA array analysis assayed 742 miRNA targets and 6 reference genes (mirBase 13) using Human microRNA Ready-to-Use PCR Panels I and II. Details are in Supplementary Methods.

Cells were transduced with lentiviral hsa-miR-302b construct, anti–miR-302b construct, or GFP control vectors (System Biosciences) as described in Supplementary Methods.

All graphs present mean ± SEM from ≥3 assays. The Student t test was used for experiments with two groups. Comparisons of more than two groups used one-way ANOVA followed by Dunnett or Tukey post hoc analysis. Some experiments used two-way ANOVA followed by Tukey post hoc tests. Tumor-initiating cell frequency was calculated by L-Calc Limiting Dilution Software (http://www.stemcell.com/en/Products/All-Products/LCalc-Software.aspx) from STEMCELL Technologies Inc., as given in ref. 29. Statistical differences between tumor growth curves (orthotopic or after tail vein) used the “Compare Growth Curves” function statmod software package (http://bioinf.wehi.edu.au/software/compareCurves/).

Primary human breast cancers invade surrounding fat and contact adipocytes, inflammatory infiltrate, and fibrous stroma (Fig. 1A). To investigate consequences of breast cancer/adipocyte interactions, human adipocyte stem cells (hASC) or mature adipocytes were isolated from human abdominal or breast fat. Adipocytes were differentiated from either hASC or marrow-isolated adult multilineage inducible (MIAMI) cells (Fig. 1B). Fat-derived hASC were cultured for more than three passages and did not contain contaminating immune cells (see Supplementary Fig. S1A for hASC characterization; refs. 24, 25). In vitro differentiated adipocytes formed adherent cultures containing abundant lipid droplets (see Fig. 1B), and expressed fat lineage markers that were undetectable in hASC. In vitro differentiated adipocytes showed intermediate maturation, expressing preadipocyte factor 1 (PREF-1), which was absent in mature adipocytes, and lower levels of aromatase, PPARγ, and other markers than mature adipocytes (Fig. 1C). These are referred to hereafter as immature adipocytes.

hASC, immature adipocytes, and mature adipocytes were cocultured with estrogen receptor–negative (ER⁻) MDA-MB-231 breast cancer cells. Over 200 secreted cytokines and growth factors were assayed by Luminex 200 in media from monocultures or after 1 week of adipose cell–cancer cell coculture. Of these, concentrations of proinflammatory cytokines IL6, IL8, IFNγ-inducible protein-10 (IP10, also called CXCL10), CCL2 (MCP1), and CCL5 (RANTES) were each synergistically increased after breast cancer cell coculture with hASC or immature or mature adipocytes (Fig. 1D). GFP-tagged MDA-MB-231 cells were sorted from hASC, immature adipocyte, or adipocyte cocultures and cancer cell cytokine expression assayed by qPCR (Fig. 1E). Although all adipose cells significantly upregulated cytokine expression by MDA-MB-231, these five cytokines increased most consistently and strongly after immature adipocyte coculture (Fig. 1E).

In obese fat, the immature adipocyte:mature adipocyte ratio is increased and immature adipocytes are a major source of proinflammatory cytokines (4, 30, 31). Thus, subsequent investigation focused on immature adipocyte–cancer cell interaction. Comparison of relative cytokine expression before and after coculture showed immature adipocytes were the major producer of proinflammatory cytokines (Supplementary Fig. S2A). Coculture of immature adipocytes separated by a Transwell also increased MDA-MB-231 cytokine expression but less than after direct coculture (Supplementary Fig. S2B). Exposure to immature adipocyte conditioned medium modestly increased IL6 and CCL2 (Supplementary Fig. S2C).

To test the generalizability of these findings, immature adipocytes were also cocultured with oncogene-transformed ER⁻ human mammary epithelial HMLER3 (22), and with ER⁺ MCF-7 and T47D. All cell lines showed a significant increase in expression (Fig. 1F) and in secretion of IL6, IL8, IP10, CCL2, and CCL5 after 1 week of coculture (Supplementary Fig. S2D).

GFP-tagged MDA-MB-231 were flow sorted after 7 days coculture with hASC, immature adipocytes, or mature adipocytes from abdominal or breast fat, and seeded into limiting-dilution mammosphere assays as in ref. 32. Exposure to all adipose cell types significantly increased MDA-MB-231 sphere formation, but this was most notable with immature adipocytes (Fig. 2A). Immature adipocyte coculture also increased sphere formation by HMLER3, MCF-7, and T47D (Fig. 2B and Supplementary Fig. S3A). Results were validated using an early passage, primary dissociated breast cancer culture, DT28, whose tumor marker and gene expression profiles reflect that of its primary cancer (23). Immature adipocyte coculture significantly increased sphere formation (Fig. 2C).

Because secretion of these five cytokines, among 200 cytokines assayed, was synergistically increased following cancer cell–fat cell coculture (Fig. 1D), these were further investigated. A 7-day exposure to IL6, IL8, IP10, CCL2, or CCL5 significantly increased sphere-generating cells in breast cancer lines and DT28 cultures (Fig. 2D–F and Supplementary Fig. S3A–S3C). Notably, all cytokines caused a sustained increase in sphere-forming cells when primary MDA-MB-231 spheres were dissociated and re-seeded without cytokine over two serial sphere assays (Fig. 2D). Increased sphere formation was not a global mitogenic effect. These cytokines did not increase short-term population growth or stimulate cell cycle (representative data, Supplementary Fig. S3D and S3E).

Prolonged cytokine exposure also increased ALDH1⁺ cells (shown for MDA-MB-231 and T47D, Fig. 2G) and clonogenicity in soft agar (Fig. 2H).

Breast cancer cells invading local fat also encounter fibroblasts. Thus, effects of breast cancer cell coculture with normal HMFs and cancer-associated fibroblasts (CAF) were compared with those of immature adipocytes. Both HMF and CAF coculture upregulated cancer cell cytokine expression, but generally to lower levels than with immature adipocytes (Fig. 3A). Baseline cytokine expression in HMFs and CAFs was lower than in adipocytes and, after coculture, increased to levels at least a log lower in magnitude than observed in cocultured immature adipocytes (Fig. 3B). Notably, HMF and CAF only modestly upregulate CCL2 after coculture, and neither increased CCL2 in cocultured MDA-MB-231. The increase in sphere formation by MDA-MB-231 after coculture with HMF or CAF was also less than after immature adipocyte coculture (Fig. 3C).

Effects of immature adipocyte coculture on tumor-initiating cell abundance were assayed in vivo. GFP+ MDA-MB-231-luc was grown with or without immature adipocytes for 1 week, then 10, 100, or 1,000 sorted cancer cells were injected orthotopically. All 1,000 cell injections yielded tumors. Tumor initiating cell frequency increased 3-fold and tumor latency decreased after prior adipocyte exposure (Fig. 4A and B, P < 0.0278). Despite the more rapid tumor growth from fat exposed cells (data for 10⁴ injected cells, Fig. 4C), Ki-67 staining was similar (n = 7–10 tumors/group; Fig. 4D). Thus, immature adipocyte exposure did not increase global tumor proliferation, further supporting a stem cell effect.

Increasing data suggest that CSC may initiate metastasis (10). Prior MDA-MB-231 exposure to immature adipocytes or cytokines for 1 week increased subsequent Matrigel invasion in vitro (Supplementary Fig. S4A and S4B). To test whether prior immature adipocyte exposure increased metastasis, orthotopic tumors were excised at 1 cm and subsequent metastasis documented by IVIS. Notably, immature adipocyte-exposed cells generated more deeply invasive tumors, and 2 of 7 mice died at excision. Metastases arose in all remaining animals. Only 3 of 7 control mice developed metastasis. Although metastasis-bearing animal numbers did not differ significantly, adipocyte-exposed cells generated a greater volume of metastases than controls (Fig. 4E). Dissociated primary tumors and nodal metastasis generated by adipocyte-exposed cells yielded significantly more spheres than controls (Fig. 4F). Thus, exposure to immature adipocytes appears to cause a stable increase in the stem cell population, sustained during primary tumor formation and subsequent metastasis.

To further investigate effects on metastasis, GFP +MDA-MB-231-luc cells were exposed for 1 week to immature adipocytes or cytokines before tail vein injection. Cells exposed ex vivo yielded a significant increase in lung tumor bioluminescence and numbers compared with controls (Fig. 4G–I and Supplementary Fig. S4C) and more circulating tumor cells (28) at 6 weeks (Fig. 4J). Notably tail vein injection of cytokine-exposed cells led to the formation of distant nodal and myocardial metastasis in 40% to 60% of animals, not seen in controls (representative data, Supplementary Fig. S4D).

Several ES-TFs maintain ES self-renewal (13) and may regulate CSC self-renewal. Cytokine exposures increased c-MYC, SOX2, and NANOG expression in both ER⁻ and ER⁺ breast cancer lines (Fig. 5A). Because Src promotes ES self-renewal (13, 15), we tested its role in CSC upregulation by immature adipocyte exposure. All cytokines increased activated pSrc within 1 hour, with no change in kinase level (Fig. 5B, top). Sox2, c-Myc, Nanog, Klf4, and Oct4 increased 12 to 18 hours after Src activation and this was abolished by Src family kinase (SFK) inhibitors (PP1, 1 μmol/L) or saracatinib (AZD0530, 1 μmol/L; PP1 in Fig. 5B and Supplementary Fig. S5A). Immature adipocyte coculture for 7 days also caused an Src-dependent increase in c-MYC, SOX2, and NANOG expression in ER⁻ and ER⁺ lines (Fig. 5C).

SFK inhibition also impaired adipocyte- and cytokine-stimulated increases in mammosphere and soft agar colony formation (Fig. 5D and E and Supplementary Fig. S5B), without anti-proliferative effects (Supplementary Fig. S5C).

Src is not only activated by immature adipocyte and cytokine exposures, but also appears to promote/sustain cytokine induction following coculture. SFK inhibition during the last 48 hours of immature adipocyte–MDA-MB-231 coculture dramatically decreased all cytokine levels (Fig. 5F). Thus, Src is not only activated by cytokine signaling, but Src and/or SFK members appear to drive further cytokine upregulation, creating a feed-forward loop to promote malignant stem cell expansion.

Certain miRNAs govern ES self-renewal (15). miRNA array analysis after 7-day cytokine treatments showed miR-302b was increased by IL6 and IP10, whereas Let7d*, miR34c-3p, and miR34b levels decreased (Fig. 6A). miR-302b is a Sox2 target (16). Since this potent ES self-renewal mediator was so strongly upregulated, its role was pursued further. qPCR showed immature adipocyte coculture also significantly increased miR-302b in MDA-MB-231, HMLER3, MCF-7, and T47D (Fig. 6B).

MDA-MB-231 and MCF-7 treatment with Src inhibitors or SOX2 silencing (Supplementary Fig. S6A) over 48 hours before plating of spheres decreased miR-302b compared with controls, and decreased miR-302b upregulation by IP10 (Fig. 6C).

Stable miR-302b overexpression in both MDA-MB-231 and MCF-7 (Supplementary Fig. S6B) increased sphere formation over parental (P < 0.0004 and P < 0.0005, respectively) and made both lines resistant to loss of sphere formation by SFK inhibitors or SOX2 knockdown (Fig. 6D). SFK inhibition and SOX2 knockdown abolished the IP10-driven increase in sphere formation in parental cells, but not in cells overexpressing miR-302b (Fig. 6E and Supplementary Fig. S6C). Notably, SOX2 and c-MYC were upregulated by miR-302b overexpression and decreased by miR-302b knockdown (Fig. 6F and Supplementary Fig. S6D). Thus, SFK activation and SOX2 both drive cytokine-mediated miR-302b induction, which further stimulates c-MYC and SOX2 expression to feed forward and sustain self-renewal.

Obesity affected over 600 million worldwide in 2014 and numbers are rising (33). Obesity is associated with greater morbidity and mortality for many cancers (2). Obese women have a higher risk of postmenopausal breast cancer and worse disease outcome at any age (1, 3), but mechanisms thereof remain obscure. This study informs the more aggressive course of breast cancer in obese individuals.

Obese adipose tissue is a mediator of chronic inflammation (5, 30). In obesity, adipocyte maturation and adiponectin production are both reduced. Leptin is elevated and promotes differentiation of macrophages that produce TNFα, FGF, EGF, and cytokines that stimulate immature adipocyte proliferation and vasculogenesis. High local TNFα and IL1β oppose adipocyte maturation, increasing immature adipocytes, which, in turn, produce further proinflammatory cytokines and chemokines to yield a chronic inflammatory milieu (4, 30, 31). IL6 and CCL2 attract monocytes and CCL5 reduces anti-tumor immunity (4, 34). Thus, tumor-adjacent fat, particularly in obese individuals, is rich in factors that promote tumor progression.

Present data indicate that close proximity or contact between breast cancer cells and hASC, immature adipocytes, and the mature adipocytes present in breast or central fat can increase cytokines that stimulate mammosphere formation in breast cancer models. Of over 200 peptide growth factors and cytokines assayed, secretion of IL6, IL8, IP10, CCL2, and CCL5 was synergistically increased upon breast cancer cell–fat cell coculture. Because immature adipocytes are increased and act as proinflammatory mediators in obese adipose tissue, their effects were assayed in both ER⁺ and ER⁻ breast cancer models. Immature adipocytes express higher basal cytokine levels and these were increased by coculture to considerably higher levels than in cancer cells. While cytokine upregulation did not require direct contact between the two cell types, it was significantly less upon coculture across a Transwell membrane or using immature adipocyte–conditioned medium, suggesting a concentration effect that would be greatest with the two populations in close contact. Adipocyte or cytokine exposures increased ALDH1 activity and mammosphere formation in both breast cancer lines and in an early passage primary dissociated breast cancer culture. Prolonged immature adipocyte exposure also significantly increased tumor-initiating cell abundance and the volume of metastasis generated from orthotopic primary cancers. That tumor growth increased without stimulation of global proliferation is compatible with increased cancer stem cell abundance. Experimental lung metastases were also increased, and metastatic tumors contained more sphere-forming cells and yielded more circulating tumor cells than controls, suggesting a stable increase in these properties.

The relevance of these cytokines to tumor progression is supported by clinical observations. Increased IL8, IL6, CCL2, and CCL5 in primary breast cancers are each correlated with advanced stage, reduced differentiation, and poor patient outcome, and increased circulating cytokines are associated with worse prognosis (6–8). Although obese women also have higher circulating cytokines (4), the possibility that increased circulating cytokines may in part drive poor cancer outcome in obese patients has not been fully appreciated.

These cytokines are known mediators of tumorigenesis and metastasis. IL6 and IL8 are vasculogenic in adipose tissue and angiogenic in cancers (4, 34). Tumor-derived CCL2 drives tumor-associated macrophage (TAM) recruitment and metastasis in breast cancer models (35), while CCL5, produced by invading mesenchymal stem cells (MSC; ref. 26), also increases breast cancer xenograft invasion and metastasis (8). Ras upregulates IP10 and decreases the growth-inhibitory IP10 receptor, CXCR3-B (36), and IL6 and IL8 are both required for Ras-transformation (37). Notably, IL6 induction by cancer-associated adipocytes promotes tumor invasion (38) and metastasis (39). Although these cytokines have been shown individually to promote xenograft growth and metastasis (8, 35, 39), present data reveal that their simultaneous upregulation upon cancer cell–adipose cell/stromal tissue interaction that generates tumor-driving loops that promote CSC expansion and metastasis. That some of these cytokines were also upregulated upon cancer cell coculture with mammary fibroblasts and CAFs raises the possibility that cytokine induction may be stimulated by aberrant heterotypic cell–cell interactions in many ectopic tissue locations during metastasis.

Several cytokines have been implicated in CSC self-renewal. IL8 increases ALDH1, sphere, and breast cancer xenograft formation, and IL8 blockade increased chemotherapy response (29). IL6 stimulates Notch3-dependent mammosphere formation (40) and MSCs drive breast CSC self-renewal via IL6 and CXCL7 (41, 42). Src was shown to activate NF-κB via pSTAT3 to induce IL6 and oppose Let7-mediated IL6 repression (19). Stromal CCL2 and CCL5 were recently implicated in CSC self-renewal (43, 44). This study adds IP10 to the cytokine network that drives mammary CSC self-renewal and reveals that adipocyte–cancer cell interaction engenders feed-forward mechanisms in which CCL2, CCL5, IL6, IL8, and IP10 all activate Src. SFK action, in turn, sustains cytokine induction and promotes Sox2-dependent miR-302b upregulation, to further induce MYC and SOX2 expression and increase CSC (see model, Fig. 7). Oncogenic Src activation, frequently observed in human breast cancers (45), would prime this process. Although we posit that adipocyte exposure enhances CSC self-renewal, methods utilized cannot exclude the possibility that adipocytes may also induce a “dedifferentiation” of more differentiated cancer cells.

Immature adipocyte or cytokine exposures increased subsequent serial mammosphere and xenograft tumor formation over months, long after stimulus withdrawal. Feed-forward mechanisms established by Src-driven SOX2 and miR-302b induction would sustain subsequent cytokine production after initial exposures. SOX2 is amplified in certain cancers and can drive clonogenic tumor growth (46). Moreover, miR-302b expression in breast cancers was recently associated with expression of stem cell markers, nodal metastasis, and poor patient outcome (47). In ES, Sox2 transactivates OCT4, NANOG, MYC, and KLF4 and these, in turn, transactivate SOX2 (13). Sox2 maintains the undifferentiated state in ES, in part by directly inducing miR302 and miR17-92 clusters (16). miR-302b powerfully mediates ES self-renewal and can efficiently reprogram somatic cells to iPSC (17, 48). Present data reveal a role for miR-302b in the effects of an adipocyte niche on breast CSC. miR-302b was induced in both ER⁺ and ER⁻ breast lines after adipocyte coculture, and IP10 and IL6 both upregulate miR-302b and miR19a (part of the miR17-92 cluster). miR-302b overexpression significantly increased sphere formation and its upregulation by IP10 required both SFK action and Sox2. Notably, miR-302b overexpression did not make cells fully resistant to either SOX2 knockdown or SFK inhibition, suggesting that miR-302b-independent mechanisms also drive cytokine-activated CSC self-renewal. miR-302b also upregulated both SOX2 and c-MYC expression; thus, cytokine-driven SOX2 induction may feed forward to maintain high ES-TF levels in part via miR-302b.

Present and prior observations suggest two additional feed-forward mechanisms that would sustain cytokine production and tumor progression. IP10 and CCL5 both downregulated Let7d*. The Let7 family governs CSC and oncogenesis by targeting RAS and HMG2a (49, 50). Ras drives cytokine secretion (36, 37, 51); thus, Let7 repression would feed forward to upregulate Ras and drive further IL6, IL8, and IP10 expression (19). IP10 upregulates CCL2 (51) and CCL2, in turn, stimulates CCL5 production by breast cancer cells (8); thus, coordinated upregulation of IP10, IL6, IL8, CCL2, and CCL5 following cancer cell invasion into fat would generate a self-perpetuating mechanism to support CSC expansion and metastasis even after cancer cell egress to other sites. Finally, coculture and cytokine exposures all upregulated c-Myc, potentially due to IL6-, IP10-, and CCL5-dependent suppression of miRs 34c-3p and/or 34-b, which target c-MYC, and also due to miR-302b-dependent c-MYC induction. c-Myc transactivates SOX2, providing yet another feed-forward loop to sustain self-renewal (13, 15).

Epithelial–stromal and inflammatory cell interactions contribute importantly to breast cancer progression (20, 21) but tumor cell–adipocyte interactions are less well understood. This study provides mechanistic insight into the aggressive behavior of breast cancer in obesity and supports a model in which cancer cell invasion into local fat would stimulate secretion of IL6, IL8, CCL2, CCL5, and IP10 by stromal cells, including mammary fibroblasts, which become CAFs, adipose cells, and cancer cells, leading to Src activation. It is tempting to speculate that the higher cytokine secretion by immature adipocytes in obese fat may initiate CAF activation. Src, in turn, would drive further cytokine induction, upregulate Sox2 and other ES-TFs to induce miR-302b, creating feed-forward loops to sustain cytokine and ES-TF induction, causing a sustained increase in tumor-initiating cell abundance and competence for metastasis. The increased inflammatory milieu generated as cancer cells migrate though mammary fat and upregulate circulating cytokines may also act systemically to promote metastasis. This mechanism may have relevance to the tropism of many malignancies to highly fatty bone marrow and may also increase metastatic potential of colon and ovarian cancers that directly infiltrate mesenteric fat. Up to 20% of cancer deaths in the United States may be attributable to obesity (52). Although the chronic inflammation of obesity is known to promote metabolic syndrome and heart disease (5), this study reveals novel feedforward mechanisms whereby tumor invasion into fat would drive inflammation and tumor progression. The obesity epidemic in the developed world may be a harbinger of an unprecedented increase in morbidity and mortality from cancer in decades to come.

No potential conflicts of interest were disclosed.

Conception and design: M. Picon-Ruiz, C. Pan, K. Drews-Elger, K. Jang, M. Kim, R.J. Cote, J.M. Slingerland

Development of methodology: M. Picon-Ruiz, C. Pan, K. Drews-Elger, K. Jang, D. Zhao, C. Morata-Tarifa, D.J. Azzam, R.H. Datar, R.J. Cote, G.A. Howard

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Picon-Ruiz, C. Pan, K. Jang, A.H. Besser, D. Zhao, M. Kim, T.A. Ince, D.J. Azzam, S.A. Wander, B. Wang, B. Ergonul, R.J. Cote, G.A. Howard, D. El-Ashry, P. Torné-Poyatos, J.A. Marchal

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Picon-Ruiz, C. Pan, K. Drews-Elger, C. Morata-Tarifa, D.J. Azzam, S.A. Wander, B. Wang, B. Ergonul, R.H. Datar, R.J. Cote, J.M. Slingerland

Writing, review, and/or revision of the manuscript: M. Picon-Ruiz, C. Pan, S.A. Wander, R.J. Cote, G.A. Howard, D. El-Ashry, J.M. Slingerland

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Pan

Study supervision: R.J. Cote, G.A. Howard, J.A. Marchal, J.M. Slingerland

The authors thank their lab, M. Lippman, and M. Abreu for useful discussions.

This work was supported by grants from the Breast Cancer Research Foundation (J.M. Slingerland and T.A. Ince) and Doris Duke Charitable Foundation (J.M. Slingerland, C. Pan, S.A. Wander, A.H. Besser, and B. Ergonul).

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.