Tumor-Associated Macrophages Regulate Murine Breast Cancer Stem Cells Through a Novel Paracrine EGFR/Stat3/Sox-2 Signaling Pathway§


  • Author contributions: Y.L.: conception, design and manuscript writing, and final approval of manuscript; R.A.R.: manuscript writing and final approval of manuscript; J.Y. and D.L.: collection of data and data analysis; C.C., Y.L., T.-H.C., R.X., and D.M.: collection of data. J.Y. and D.L. contributed equally to this article.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS November 21, 2012.


The cancer stem cell (CSC) hypothesis has gained significant recognition as a descriptor of tumorigenesis. Additionally, tumor-associated macrophages (TAMs) are known to promote growth and metastasis of breast cancer. However, it is not known whether TAMs mediate tumorigenesis through regulation of breast CSCs. Here, we report that TAMs promote CSC-like phenotypes in murine breast cancer cells by upregulating their expression of Sox-2. These CSC-like phenotypes were characterized by increased Sox-2, Oct-4, Nanog, AbcG2, and Sca-1 gene expression, in addition to increased drug-efflux capacity, resistance to chemotherapy, and increased tumorigenicity in vivo. Downregulation of Sox-2 in tumor cells by siRNA blocked the ability of TAMs to induce these CSC-like phenotypes and inhibited tumor growth in vivo. Furthermore, we identified a novel epidermal growth factor receptor (EGFR)/signal transducers and activators of transcription 3 (Stat3)/Sox-2 paracrine signaling pathway between macrophages and mouse breast cancer cells that is required for macrophage-induced upregulation of Sox-2 and CSC phenotypes in tumor cells. We showed that this crosstalk was effectively blocked by the small molecule inhibitors AG1478 or CDDO-Im against EGFR and Stat3, respectively. Therefore, our report identifies a novel role for TAMs in breast CSC regulation and establishes a rationale for targeting the EGFR/Stat3/Sox-2 signaling pathway for CSC therapy. STEM CELLS2013;31:248–258


Transcriptional regulatory networks consist of functional interactions between regulatory genes and a much larger set of downstream target genes. The network of Sox-2/Oct-4 target genes includes Sox-2, Oct-4, Nanog, FGF-4, UTF1, Fbx15, and Lefty1 [1–3]. Each of these genes is expressed in embryonic stem cells (ESCs) and was shown to be substantially downregulated upon differentiation of both ESCs and embryonic carcinoma cells due to decrease of Sox-2 and Oct-4 expression [1, 2]. Additionally, Sox-2 was one of four transcription factors shown to reprogram mammalian somatic cells into induced pluripotent stem (iPS) cells [4]. In the context of cancer, Sox-2 overexpression has been demonstrated in cervical, ovarian, lung, liver, and breast carcinomas and was a predictor of increased malignancy and poor prognostic outcome [5, 6]. Although the importance of Sox-2 for ESC maintenance and induction of iPS cells has been clearly demonstrated [7, 8], it is unknown whether Sox-2 plays a role in regulating breast cancer stem cells (CSCs).

The CSC hypothesis postulates that neoplastic clones are maintained by a small subpopulation of cells that possess the capacity for self-renewal and differentiation potential, thus giving rise to cancer cells that comprise the tumor bulk [9, 10]. Furthermore, CSCs have been suggested to be the root cause of cancer recurrence and disease relapse due to their resistance to chemotherapy and radiotherapy [11]. CSCs were first identified in hematologic tumors [12] and were subsequently found in various solid malignancies including breast, brain, prostate, and lung cancers [13–16]. CSCs have also been discovered in various tumor-derived cell lines based on their expression, or lack thereof, of such surface markers as CD44, CD24, CD29, Lin, CD133, and Sca-1, and by their ability to form spheres in nonadherent three-dimensional cultures in vitro [13, 16].

Another useful approach for identifying CSCs, particularly in the absence of suitable surface marker expression, uses the phenomenon of stem cells' unique ability to efflux lipophilic fluorescent dyes, including Hoechst 33342 [17]. This efflux capacity was shown to correlate with expression of ATP-binding cassette (ABC) transporters and could be specifically inhibited with Ca++-channel blockers [18]. The dye-effluxing population was given the designation side population (SP) based on their low dye retention characteristic [17, 19]. The dye-efflux method has been used to successfully identify SP cells possessing stem cell-like properties from a variety of tissues including mammary gland, skin, brain, and liver [20–23]. Additionally, previous work in our laboratory has demonstrated that this method can be used to isolate SP cells from murine breast carcinoma and neuroblastoma cell lines [24]. Importantly, compared to Non-SP, SP cells exhibited increased expression of stem cell-associated markers and showed enhanced tumorigenicity and resistance to chemotherapy when transplanted into immune-competent mice [24]. Furthermore, these SP cells also possessed a markedly higher repopulation potential in vivo, when compared with non-SP cells [24].

In vivo, stem cells are concentrated in areas rich in blood vessels and stromal cells, in regions referred to as the “stem cell niche.” This niche is thought to protect stem cells from apoptotic stimuli and to enable a proper balance between self-renewal and differentiation. In fact, it was suggested that CSCs in the tumor microenvironment (TME) reside in a niche that is critical for their maintenance, similar to normal stem cells [25]. This theory implies that disruption of the CSC niche, and thus stem cell maintenance, may provide an avenue of attack for eliminating CSCs [26].

Tumor-associated macrophages (TAMs) constitute a major cell population in the breast TME [27]. Importantly, it has been suggested that macrophages secrete growth and other factors that permeate the breast stem cell niche to promote survival and self-renewal of stem cells [28]. Intriguingly, Ohno et al. showed that TAMs have varying responses to regulatory signals, depending on the area of the TME in which the signaling occurs [29]. Additionally, Pollard and coworker suggested that TAMs may be educated to carry out specific functions in support of maintaining CSC activities within perivascular and hypoxia/necrotic areas in the TME [30]. However, the precise role of TAMs in influencing CSC niches in vivo and the molecular mechanisms underlying TAM/CSC interactions remain unclear.

Therefore, in this study, we critically investigated whether TAMs regulate breast CSC maintenance. To this end, we identified a novel epidermal growth factor receptor (EGFR)/signal transducers and activators of transcription 3 (Stat3)/Sox-2 paracrine signaling pathway between macrophages and breast cancer cells that is required for upregulation of Sox-2 in tumor cells and maintenance of CSC phenotypes. Importantly, we show here that this crosstalk was effectively blocked by inhibition of either EGFR tyrosine kinase activity or Stat3 activation using the small molecule inhibitors AG1478 or CDDO-Im, respectively. To our knowledge, this is the first report identifying a mechanism by which TAMs regulate expression of Sox-2 in breast CSCs that establishes a rationale for targeting the EGFR/Stat3/Sox-2 signaling pathway for CSC-directed cancer therapy.


Animal Cell Lines and Reagents

Female BALB/c mice, 6–8 weeks of age, were purchased from the Scripps Research Institute Rodent Breeding Facility (La Jolla, CA). All animal experiments were performed according to National Institutes of Health guidelines. The 4T1 and 4T07 mammary carcinoma cell lines were generously provided by Dr. Suzanne Ostrand-Rosenberg (University of Maryland, Baltimore). The RAW 264.7 macrophage cell line was purchased from ATCC (Manassas, VA). Our animal protocol was approved by The Scripps Research Institute Institutional Animal Care and Use Committee. AG1478, HOECHST 33342, and Verapamil hydrochloride were purchased from Sigma, Inc. (Sigma, St Louis, MO, http://www.sigmaaldrich.com). CDDO-Im was a gift from Dr. Michael Sporn (Dartmouth Medical School, Hanover, NH). Recombinant epidermal growth factor (EGF) was purchased from (BD Bioscience, Franklin Lakes, NJ, http://www.BD.com).

Flow Cytometry and Cell Sorting

We stained 1 × 106 tumor cells per milliliter with either 5 μg/ml HOECHST 33342 dye or HOECHST dye plus 100 mM Verapamil hydrochloride, to block dye efflux at 37°C for 30 minutes, as described previously [24]. These cells were then further stained with antibodies against Sca-1, c-Kit, ABCB1b, or ABCG2 (BD Pharmingen, San Diego, CA, http://www.BD.com) with appropriate isotype controls (BD Pharmingen). Cells were analyzed and sorted by an EPICS ALTRA flow cytometer (Beckman Coulter, Fullerton, CA, https://www.beckmancoulter.com). The HOECHST dye was excited at 350 nm, and its fluorescence measured at two wavelengths (450/20 nm band-pass filter and 675LP optical filter). Fluorescence-activated cell sorting data were analyzed with FlowJo software (Tree Star, Ashland, OR, http://www.treestar.com).

RT-PCR Analysis

Total RNA from three independently sorted 4T1 or 4T07 SP and non-SP cell samples was isolated using the RNAeasy Mini Kit (Qiagen, Valencia, CA, http://www.qiagen.com). RNA was then reverse transcribed applying the SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) and cDNA was analyzed by PCR. Alternatively, total RNA was analyzed using SuperScript One-Step RT-PCR with Platinum Taq (Invitrogen).

Western Blotting

Western blotting was performed with either total protein from cell lysate homogenates and cytoplasmic or nuclear extracts with polyclonal rabbit anti-mouse Sox-2, Oct-4, Stat3 or Nanog antibody, monoclonal rat anti-mouse E-cadherin antibody, and goat anti-mouse Actin antibody (Santa Cruz Biotechnology, Santa Cruz, CA, http://www.scbt.com). We also used polyclonal rabbit anti-mouse N-cadherin and beta catenin antibodies (Abcam, Cambridge, MA, http://www.abcam.com), or monoclonal rabbit anti-mouse phospho-Stat3 antibody (Cell Signaling Technology, Danvers, MA, http://www.cellsignal.com). Primary antibodies were detected with appropriate species-specific horseradish peroxidase-conjugated IgG antibodies (Santa Cruz Biotechnology).

Immunohistochemical Fluorescence Staining

4T1 or 4T07 cultured tumor cells were initially stained with HOECHST 33342 dye for 30 minutes at 37°C. The cells were then washed twice with cold phosphate-buffered saline (PBS) and stained with rabbit anti-mouse Sox-2 or Oct-4, rat anti-mouse EGFR (AbD Serotec, Raleigh, NC, http://www.abdserotec.com/), or rabbit anti-mouse Nanog (Santa Cruz Biotechnology) antibodies and detected with either goat anti-rabbit IgG Alexa 588, goat anti-rabbit IgG Alexa 568, or goat anti-rat IgG Alexa 488 secondary antibody (Life Sciences, Carlsbad, CA, http://www.lifesci.com). Macrophages were stained with a monoclonal rat anti-mouse F4/80 antibody (AbD Serotec) and detected by goat anti-rat IgG Alexa 568 secondary antibody (Life Sciences). Fluorescence images were captured with a Zeiss Axiovert 100 microscope (Carl Zeiss MicroImaging, Thornwood, NY, http://microscopy.zeiss.com).

Coculture Assay

4T1 breast carcinoma cells were cocultured with either RAW macrophage cells or macrophages derived from normal mouse spleen or from breast tumor tissue. Briefly, for coculture without cell-cell contact, 3 × 105 macrophages were seeded in Boyden Transwell inserts (5-μm pores; Nunc, Rochester, NY, http://www.nuncbrand.com) permeable for soluble factors but not cells. Transwells containing macrophages were then inserted into a six-well plate and seeded with 8 × 105 4T1 tumor cells in each well. For coculture with cell-cell contact, 8 × 105/ml 4T1 tumor cells were mixed with 2 × 105/ml macrophages in a flask and then cultured together for different time points at 37°C, 5% CO2, prior to analysis by flow cytometry.

In Vivo Macrophage Depletion

Chlodronate liposome nanoparticles were a gift from Roche Diagnostics GmbH (Roche, Mannheim, Germany, http://www.roche.com). Mononuclear phagocytes ingest the clodronate-containing liposomes and the liposomal bilayer is disrupted with phospholipase, resulting in intracellular release of clodronate and subsequent cell death [31]. A dose of 100 ml/10 g of mouse body weight of clodronate liposomes diluted in PBS was administered intravenously in BALB/c mice every week. The first injection was administered 2 days before the start of the experiment. Control mice were administered with either PBS as a control for unstimulated macrophages or liposomes in saline to control for any nonspecific effects of liposome administration. Macrophage depletion was maintained throughout the experimental period.

siRNA Gene Silencing

We used the siRNA gene Silencer system (Santa Cruz Biotechnology, Inc.) to perform the Sox-2 gene silencing in 4T07 tumor cells according to the manufacturer's protocol. Sox-2 knockdown was verified by RT-PCR or Western blot, as described above.

Migration Assay

Transwell migration assays were performed with 5 × 103 tumor cells seeded in Boyden Transwell chambers (8 μm pore size; Corning, Lowell, MA, http://www.corning.com). After 4-hour incubation at 37°C, cells that had migrated to the lower surface of the membrane were fixed with 1% paraformaldehyde, stained with 1% crystal violet, and counted.

Apoptosis Assay

Apoptosis was analyzed by flow cytometry using the Annexin V assay kit (eBioscience, San Diego, CA, http://www.ebioscience.com) according to the manufacturer's protocol. HOECHST 33342 dye was added 25 minutes before incubation with Annexin V to identify the SP or non-SP cell of 4T1 or 4T07 populations.

In Vivo Tumor Cell Challenge

Female age-matched BALB/c mice (n = 4 mice/group) were injected i.v. with either 1 × 104 4T07 wild-type or 4T07 Sox-2 knockdown breast carcinoma cells. All mice were sacrificed 25 days after tumor cell challenge.

Statistical Analysis

The statistical significance of different findings between experimental groups was determined by the Student's t test using Microsoft Excel (Microsoft, Redmond, WA, http://www. microsoft.com) or GraphPad Prism (GraphPad Software, LA Jolla, CA, http://www.graphpad.com) software. Results were considered significant if two-tailed p values were < .05.


TAM-Associated Increases in Breast Cancer Tumorigenicity Correlates with Increased Percentage of SP Cells In Vivo

Interactions between CSCs and cells within their niche in the TME appear to be important for their maintenance and function [26]. In order to assess the impact of TAMs on breast CSC maintenance and tumorigenicity in vivo, we first isolated 4T07 SP cells by flow cytometry and then injected 1 × 103 of these cells into the cleared fat pads of syngeneic BALB/c mice that were either depleted or not of macrophages by clodronate liposomes (Fig. 1A). The depletion of macrophages in both blood and tumor tissue was confirmed by flow cytometry (Fig. 1B). Analysis of mice 25 days after tumor cell implantation revealed significantly reduced primary tumor and lung weights in animals treated with clodronate liposomes, indicating reduced tumorigenicity and metastatic potential, respectively, in mice depleted of macrophages (Fig. 1C, 1D). Importantly, this reduction in tumorigenicity and metastasis in macrophage-depleted mice correlated with a significant decrease in the percentage of HOECHST 33342 dye-effluxing SP cells in primary tumors (Fig. 1E). Taken together, these results indicate that TAMs mediate maintenance of SP cell populations in vivo. Furthermore, these findings also imply that TAMs within the stem cell niche may play a critical role in CSC maintenance.

Figure 1.

Tumor-associated macrophages mediate SP cell maintenance in vivo. (A): 4TO7-SP cells were isolated by HOECHST 33342 dye staining and flow cytometry cell sorting. These SP cells (1 × 103) were then injected i.v. into BALB/c mice that had previously been depleted of macrophages by treatment with chlorinate liposome nanoparticles (M φ KO). Control animals were treated with saline and thus not depleted of macrophages (WT). (n = 5 mice/group), (B) Macrophage populations (CD45+/F4/80+) in blood and primary tumors of WT or M φ KO mice were measured by flow cytometry. Mice were sacrificed 25 days after SP cell challenge and tumor (C) and lung (D) weights were measured. Data represent means ± SEM. (E): The percentages of SP cells in primary tumors from WT and M φ KO mice were measured by HOECHST 33342 dye staining and flow cytometry. Data represent means ± SEM. *, p < .05; **, p < .005. One of at least three independent experiments is shown. Abbreviations: SP, side population; WT, wild type.

TAM Coculture Enriches the SP Population of 4TO7 Tumor Cells In Vitro

To further elucidate the mechanism of TAM-mediated regulation of breast CSCs, we performed a series of ex vivo experiments involving coculture of murine breast cancer cells with the murine RAW 264.7 macrophage cell line. This macrophage cell line was originally established from a tumor induced by Abelson murine leukemia virus [32] and we showed that these macrophages possess an M2/TAM phenotype characterized by expression of CD206high, IL-10high, CD86low, and Major histocompatibility complex Class IIlow (supporting information Fig. S1). Importantly, this phenotype was enhanced by either cocultured with tumor cell or tumor cell conditional medium with increasing release of Th2 cytokines and growth factors, which suppressed antitumor immunity and enhanced tumor growth (supporting information Fig. S2). We first analyzed the effects of RAW and tumor cell coculture on the SPs of 4TO7 tumor cells by the HOECHST 33342 dye exclusion assay. Flow cytometry analysis revealed that coculture of tumor cells with RAW macrophages resulted in a 2.8-fold increase in percentage of SP cells (Fig. 2A), which correlated with a marked increase in expression of Sca-1 and AbcG2 on the surface of these cells (Fig. 2B). At the mRNA and protein levels, macrophage coculture resulted in increased mRNA and protein expression of Sox-2, Oct-4, and Nanog by SP cells (Fig. 2C, 2D, respectively). Further analysis of the expression of Sox-2, Oct-4, and Nanog in primary tumors by immunohistochemistry revealed that these key transcription factors were expressed by tumor cells proximal to TAMs (Fig. 2E, 2F). Together, these results suggest that TAMs actively communicate with tumor cells to enhance their expression of Sox-2 and other key regulators of CSC phenotypes. These findings further imply that TAMs regulate breast CSCs by mediating their expression of Sox-2 through paracrine signaling within the stem cell niche.

Figure 2.

Tumor-associated macrophages enrich SP cells and enhance expression of Sox-2, Oct-4, and Nanog in breast cancer cells. (A): The SP of 4T07 breast cancer cells was obtained by Hoechst staining and flow cytometry after 96 hours of coculture with either TAM derived from 4T07 tumor tissue or RAW macrophages. One of at least five independent experiments is shown. (B): The expression of Sca-1 and ABCG2 was also detected in this same population. (C): Expression of Sox-2, Oct 4, and Nanog was determined by RT-PCR and (D) Western blot. One of three independent experiments is shown. (E, F): Expression of Sox-2, Oct 4, and Nanog in 4T07 tumor tissue was confirmed by immunofluorescence histology staining. Data represent means ± SEM. **, p<0.005, One of at least three independent experiments is shown. Scale bars = 100 μm; 150 μm on lower panel of (E). Abbreviations: DAPI, 4',6-diamidino-2-phenylindole; SP, side population; WT, wild type.

The Sox-2 Transcription Factor Regulates the Tumorigenicity of SP Cells

Transcriptional regulation plays a key role in maintenance of CSC properties [33–35] and the Sox-2 transcription factor was reported to be important for regulating ESCs [33, 36]. Therefore, we postulated that the Sox-2 signaling pathway might be involved in networks controlling breast CSC maintenance. To determine whether Sox-2 is important for maintenance of CSC properties in murine breast cancer cells, we used siRNA to silence Sox-2 gene expression in 4TO7 cells, which was confirmed by RT-PCR (Fig. 3A). Intriguingly, we found that Sox-2 silencing also resulted in decreased Oct-4 and Nanog mRNA expression (Fig. 3B). Furthermore, expression of Sca-1 and AbcG2 stem cell markers was also markedly suppressed in the SP population of 4TO7 cells treated with Sox-2 siRNA (Fig. 3C). In vivo, Sox-2 knockdown in 4TO7 cells markedly suppressed tumor growth (Fig. 3D), reduced tumor weights (Fig. 3E), and decreased lung metastasis, as indicated by decreased lung weight (Fig. 3F). Ex vivo transwell migration assays showed that Sox-2 silencing also suppressed tumor cell motility (Fig. 3G). Additionally, Sox-2 silencing in SP cells not only increased tumor cell apoptosis but also increased the sensitivity of these cells to mitoxantrone chemotherapy (Fig. 3H). Taken together, these data confirm that the transcription factor Sox-2 is important for the maintenance of CSC phenotypes in murine breast cancer cells.

Figure 3.

Transcription factor Sox-2 regulates maintenance of cancer stem cell-like properties of SP cells. Sox-2 gene expression in 4TO7 cells was silenced by siRNA. (A): Downregulation of Sox-2 was confirmed by RT-PCR. (B): Gene expression of Sox-2, Oct-4, and Nanog by SP cells was determined by RT-PCR after Sox-2 silencing. (C): Expression of surface makers Sca-1 and ABCG2 was assessed by flow cytometry in 4TO7 SP cells after Sox-2 silencing. One of at least three independent experiments is shown. (D): BALB/c mice were challenged with 4TO7-SP WT or 4TO7-SP cells subjected to Sox-2 silencing (Sox-2 siRNA), and tumor volumes were measured every 3–4 days. (n = 5 mice/group). Twenty-five days after tumor cell challenge, tumor (E), and lung (F) weights were measured. One of three independent experiments is shown. (G): Migration assays were performed on 4TO7 WT or Sox-2 siRNA-treated SP cells using Boyden transwell chambers (n = 3 wells/group). (H): Effects of Sox-2 silencing on apoptosis of non-SP and SP cells were determined by Annexin V staining and flow cytometry. Additionally, the sensitivity of SP and non-SP cells to mitoxantrone chemotherapy was also assessed after Sox-2 knockdown. Data represent means ± SEM. **, p<0.005. One of three independent experiments is shown. Scale bars = 500 mm. Abbreviations: SP, side population; WT, wild type.

TAM Activation of EGFRs on CSCs Results in Increased Sox-2 Expression by Tumor Cells Which Are Inhibited by EGF Neutralizing Antibody

We next aimed to identify the signaling molecules responsible for TAM-mediated upregulation of Sox-2 expression by breast cancer cells. TAMs are known to secrete many different growth factors and chemokines that promote tumor progression, including EGF [37]. Importantly in this regard, aberrant activation of the EGFR signaling pathway has been observed in many human cancers [38]. To determine whether soluble EGF released by TAMs mediates acquisition of CSC phenotypes by breast cancer cells, we examined expression of EGFR and phosphorylated EGFR (pEGFR) on 4T1 cells cocultured with RAW macrophages. Interestingly, analysis by flow cytometry showed overexpression of both EGFR and pEGFR by SP cells, compared with non-SP cells, in pure 4T1 cultures (Fig. 4A). Moreover, this expression of EGFR and pEGFR by SP cells was further enhanced by coculture with RAW macrophages, and correlated with an increase in percentage of SP cells, compared with tumor cells cultured alone (Fig. 4D). Similarly, culture of tumor cells with recombinant mouse EGF (mEGF) also induced an increase in the percentage of SP cells and a concordant increase in EGFR and pEGFR expression by these cells (Fig. 4E). Importantly, mEGF treatment resulted in increased expression of Sox-2 mRNA and protein by 4T1 tumor cells (Fig. 4B). This increase in Sox-2 expression also correlated with increased tumor cell migration upon mEGF stimulation (Fig. 4C). Together, these results demonstrate that TAM/tumor cell crosstalk via EGF/EGFR functions as an upstream activator of Sox-2 expression in murine breast cancer cells.

Figure 4.

TAMs and EGF induce overexpression of EGFR and pEGFR on SP cells that correlates with increased Sox-2 expression and cell motility, which are inhibited by EGF neutralizing antibody. (A): The expression of EGF in F4-80+ cells from either 4T1 tumor tissue or normal spleen of BALB/c mice was determined by flow cytometry. EGF expression in RAW cells after cocultured with 4T1 tumor cells was also evaluated simultaneously. (B): Expression of Sox-2 at the mRNA (left panels) and protein levels (right panels) in 4T1 cells was assessed after treatment with mEGF. (C): EGF-induced migration of 4T1 cells as determined using Boyden transwell chambers. Expression of EGFR or phosphorylated EGFR (pEGFR) was detected by HOECHST 33342 dye and antibody staining, followed by flow cytometry analysis of 4T1 SP and non-SP cells after coculture with RAW macrophages (D) or recombinant mouse EGF (mEGF) (E). Data represent means ± SEM. (F): Expression of Sca-1 and ABCG2 in SP population of 4T1 cells was analyzed after coculture with RWA cells and treatment with EGF neutralizing antibody. (G): Either TAMs from 4T1 tumor tissue or RAW cells induced overexpression of Stat 3, pStat3, and Sox-2, which was inhibited by EGF neutralizing antibody in 4T1 cells. (H): The migration of 4T1 tumor cell after coculture with RAW cell was also inhibited by EGF-Ab. Data represent means ± SEM (three wells per group). , **, p<0.005 Scale bars = 500 mm. Abbreviations: mEGF, mouse epidermal growth factor; RAW, RAW264.7 cell line; SP, side population; TAM, tumor-associated macrophage; WT, wild type.

It was well known that TAMs release many kinds of cytokines and growth factors, including EGF, that promote tumor progress [39]. To determine whether EGF released by TAMs mediated maintenance of a CSC-like phenotype on SP of 4T1 breast cancer cell, we performed an experiment to detect expression of Sca-1 and ABCG2 on SP cells isolated from 4T1 tumor cells which were cocultured with TAMs after treatment with EGF neutralizing antibody. These results clearly indicated that expression of both Sca-1 and ABCG2 was downregulated by treatment of TAMs with EGF neutralizing antibody. The SP population of 4T1 cells was shown to be inhibited as well (Fig. 4F). Moreover, analysis of protein expression by 4T1 cells that had been cocultured, with either fresh isolated TAMs or RAW cells, demonstrated that expression of Stat3, pStat3, or Sox-2 was downregulated after treatment with EGF neutralizing antibody (Fig. 4G). Furthermore, migration of 4T1 cells was suppressed by treatment with EGF neutralizing antibody (Fig. 4H). Together, these data show that TAMs mediate effects that promote maintenance of SP CSC-like phenotypes supported by EGF release.

The EGFR Inhibitor AG1478 Blocks TAM-Mediated Upregulation of Sox-2 in Breast Cancer Cells

The requirement of EGFR activation for TAM/CSCs crosstalk and upregulation of Sox-2 expression by tumor cells was confirmed by blocking EGFR kinase activity with the inhibitor, AG1478. Treatment of 4T1 cells with AG1478 effectively blocked mEGF-induced upregulation of Sox-2 at both the mRNA and protein levels (Fig. 5A). Importantly, the increase in percentage of SP cells and expression of EGFR and pEGFR observed when 4T1 cells were cultured with RAW macrophage conditioned medium (MCM) was also completely blocked by AG1478 (Fig. 5B). Furthermore, analysis of mRNA and protein from these cells showed that AG1478 inhibited MCM-induced upregulation of both Sox-2 mRNA and protein expression (Fig. 5C). The inhibition of Sox-2 and EGFR expression by AG1478 was also confirmed by immunohistochemistry in 4T1 cells cocultured with either RAW macrophages or with RAW MCM (Fig. 5D, 5E, respectively). Together, these data show that TAM-mediated upregulation of Sox-2 expression in breast cancer cells is dependent on EGF/EGFR paracrine signaling.

Figure 5.

Macrophage-mediated effects on breast cancer cells are inhibited by EGFR inhibitor AG1478. (A): EGF-induced overexpression of Sox-2 was found to be inhibited in 4T1 cells by AG1478, at both the mRNA (left panels) and protein levels (right panels). (B): Culture of 4T1 cells with RAW MCM increased the percentage of side population cells and was correlated with increased EGFR and pEGFR expression as determined by HOECHST 33342, antibody staining, and flow cytometry. (C): MCM-induced increases in Sox-2 expression are inhibited by AG1478 at both the mRNA (left panels) and protein levels (right panels). Inhibition of Sox-2 and EGFR expression after AG1478 treatment was confirmed by immunohistochemistry in 4T1 cells cocultured with either RAW macrophages (D) or MCM (E). One of three independent experiments is shown. Scale bar = 100 μm. Abbreviations: DAPI, 4',6-diamidino-2-phenylindole; mEGF, mouse epidermal growth factor; MCM, macrophage conditioned medium; RAW, RAW264.7 cell line; WT, wild type.

EGFR-Mediated Upregulation of Sox-2 in Breast Cancer Cells Is Stat-3 Dependent

EGF is a known ligand for the Stat3 signaling pathway [40]. Upon binding of EGF to EGFR, Stat3 is phosphorylated which allows dimerization and translocation of the Stat3 protein into the nucleus where it activates transcription of downstream target genes [40]. Intriguingly, a recent study by Foshay et al. showed that Stat3 signaling could promote development of neural stem cells through regulation of Sox-2 expression [41]. Therefore, we hypothesized here that TAMs might use a similar paracrine signaling pathway via EGFR/Stat3 to regulate Sox-2 expression in breast cancer cells. Indeed, treatment of 4T1 cells with EGF induced phosphorylation of Stat3 and nuclear accumulation of Stat3 in a dose-dependent manner (Fig. 6A). In contrast, treatment of tumor cells with the Stat3 inhibitor, CDDO-Im, blocked EGF-induced phosphorylation of Stat3 (Fig. 6B). Critically, EGF treatment significantly increased the percentage of SP cells in 4T1 tumor cell cultures, and this increase was also effectively blocked by CDDO-Im (Fig. 6C). Furthermore, RT-PCR and Western blot analysis indicated that treatment of 4T1 cells with CDDO-Im also blocked EGF-induced upregulation of Sox-2 mRNA and protein expression (Fig. 6D, 6E). Additionally, partial inhibitions of Oct-4 and Nanog mRNA and protein expression were also observed with CDDO-Im treatment (Fig. 6D, 6E).

Figure 6.

EGF-induced upregulation of SP cells is Stat3 dependent. (A): Western blot analysis of cell lysates from 4T1 cells treated for 15 minutes with EGF at varying concentrations. (B): Western blot analysis of cell lysates from 4TO7 cells treated for 10 minutes with EGF (20 ng/ml) plus CDDO-Im at varying concentrations. (C): 4T1 cells treated with EGF, with or without CDDO-Im, were analyzed by HOECHST 33342 dye staining and flow cytometry to quantify the percentage of SP cells. Data represent means ± SEM. *, p < .05. (D, E): Treatment of 4T1 cells with CDDO-Im (100 nM) suppresses EGF-induced upregulation of Sox-2, Oct-4, and Nanog mRNA (D) and protein (E) expression, as evaluated by RT-PCR and Western blotting, respectively. (F): Western blotting of lysates from 4T1 cells, treated with either IL-6 or EGF in combination with CDDO-Im. (G): Western blotting of lysates from 4T1 cells cultured in MCM, with or without CDDO-Im. (H, I): Migration of 4T1 and 4TO7 breast cancer cells in response to incubation with MCM, either with or without CDDO-Im, was determined in Boyden transwell chambers and quantified with ImageJ software (three wells per group). *, p < .05; **, p < .005. One of three independent experiments is shown. Abbreviations: EGF, epidermal growth factor; MCM, macrophage conditioned medium; SP, side population.

Since we found that CSC phenotypes can be induced by MCM and blocked by inhibiting EGFR tyrosine kinase activity via AG1478, we next determined whether these phenotypes could also be blocked through Stat3 inhibition by CDDO-Im. To this end, we treated 4T1 cells with EGF or IL-6 in combination with CDDO-Im. Western blot analysis revealed that only EGF treatment resulted in an increase in nuclear Stat3 and Sox-2 protein expression and was inhibited by CDDO-Im (Fig. 6F). Interestingly, EGF treatment also induced a decrease in E-cadherin and an increase in N-cadherin expression in 4T1 cells (Fig. 6F). This observation is significant since changes in cadherin expression have been linked to dedifferentiation of breast cancer cells and are associated with acquisition of CSC phenotypes [42]. Similar to EGF, incubation of breast cancer cells with MCM also resulted in nuclear accumulation of Sox-2 protein (Fig. 6G). Additionally, nuclear accumulation of β-catenin, and a concomitant decrease in cytoplasmic β-catenin, protein expression was also observed in response to culture with MCM (Fig. 6G). Importantly, both upregulation of Sox-2 expression and nuclear translocation of β-catenin in response to MCM could be inhibited by treatment with CDDO-Im (Fig. 6G).

Because these changes in E- and N-cadherin expression and β-catenin localization are associated with increased cell motility, we investigated the effects of MCM on tumor cell motility. We found that transwell migration of both 4T1 and 4TO7 breast cancer cells was significantly increased when cultured with MCM (Fig. 6H). Importantly, this MCM-induced migration of tumor cells was significantly inhibited by CDDO-Im (Fig. 6I).


It has been approximately 150 years since the idea of CSCs was first proposed. During this time, advances in stem cell biology have redefined the CSC hypothesis. This hypothesis now centers on the concept that a small fraction of tumor cells retains a self-renewing capacity that drives the tumorigenic potential. CSCs have been identified and isolated as a SP fraction from various human tumors on the basis of this self-renewing property [43]. Additionally, a better understanding of the TME has lead to the supposition that CSCs exist within a stem cell niche that regulates transcriptional networks important for the self-renewal capacity and tumorigenic potential of CSCs [25].

In this study, we provide further insight into the regulatory mechanisms that govern the stem cell niche, and its effects on Sox-2 expression in CSCs. We first confirmed the presence of CSC-like cells by identifying a HOECHST 33342 dye-effluxing SP in the 4T1 and 4T07 murine breast cancer cell lines (supporting information Fig. S2). In contrast to the dye-retaining non-SP, these SP cells possessed CSC phenotypes, including overexpression of Sox-2 and Oct-4 transcription factors and Sca-1 and AbcG2 cell surface markers. Association of these transcription factors and surface markers with stem cells is well established [44]. Additionally, SP cells isolated from murine breast cancer cell lines showed markedly increased tumorigenicity in vivo when implanted into syngeneic mice, in comparison to their non-SP counterparts. These findings are in agreement with previous work in our laboratory that identified SP cells in human breast cancer cell lines possessing CSC phenotypes, including increased tumorigenic potential in vivo [24]. Importantly, we found here that expression of the Sox-2 transcription factor was essential for maintenance of CSC properties in murine breast cancer cells. Thus, knockdown of Sox-2 in these cells by siRNA not only decreased expression of Sox-2 but also decreased expression of its partner genes Oct-4 and Nanog and Sca-1 and AbcG2 CSC surface markers by SP cells. Significantly, these reductions in CSC-associated gene expression and surface markers upon Sox-2 knockdown coincided with a concomitant decrease in the percentage of SP cells within the tumor cell population.

Increasing evidence emerging in the stem cell field suggests that acquired resistance to chemotherapy can be attributed to CSCs [45]. It is therefore intriguing that Sox-2 knockdown in SP cells resulted in decreased expression of ABC transporters which correlated with increased sensitivity of SP cells to mitoxantrone chemotherapy. These results may provide a potential avenue of attack for sensitizing chemo-resistant CSCs to chemotherapy through downregulation of Sox-2 expression in these cells.

Tissue niches play a critical role in the self-renewal and differentiation capacity of stem cells. For example, circulating hematopoietic stem cells have little function outside their tissue-specific niches and trafficking to specific tissue niches within the bone marrow is critical for maintaining normal hematopoiesis [46]. Considering that SP cells isolated from tumors share many characteristics with their normal stem cell counterparts, it is not implausible to predict that niches in which CSCs reside also play important roles in preserving their functions. Therefore, crosstalk between stromal and tumor cells within the CSC niche may regulate cytokine and growth factor expression levels to promote tumor cell migration or drug resistance.

In this regard, TAMs have been identified as a major cellular component of the breast TME and their trafficking from the circulation to the tumor stroma is a key phenomenon associated with breast cancer progression [27]. Additionally, TAMs are known to secrete various factors that promote tumor growth [47]. However, whether TAMs regulate CSC functions is not known. Due to their association with increased tumorigenicity and their physical proximity to tumor cells, we hypothesized that TAMs play a central role in CSC maintenance. Indeed, in support of our hypothesis, we found that depletion of macrophages in vivo suppressed the tumorigenicity of murine breast cancer cells implanted into syngeneic mice. Critically, suppressed tumor growth in these mice was associated with a decrease in the percentage of SP cells in their primary tumors. In contrast, ex vivo coculture of such tumor cells with RAW macrophages lead to an increase in percentage of SP cells and correlated with increased expression of CSC-associated genes, including Sox-2. These results demonstrate that TAMs mediate Sox-2 expression in tumor cells and thus represent a stromal cell responsible for regulating and maintaining CSCs in the stem cell niche of the breast TME.

Even though the existence of crosstalk between CSCs and niche cells is well-established [48, 49], the cellular mechanisms mediating this crosstalk and the signaling molecules involved remain a complex web waiting to be untangled. To this end, we identified EGF as a key signaling molecule that mediates crosstalk between TAMs and tumor cells. We also showed that EGF enriches the percentage of SP cells in murine breast cancer cell cultures and increases their expression of Sox-2. Additionally, EGF also increased tumor cell motility. Critically, all these effects were effectively blocked by AG1478, an EGFR tyrosine kinase inhibitor, confirming the essential role of EGF in TAM/tumor cell crosstalk. These results are intriguing given that constitutive activation of EGFR is commonly observed in epithelial tumors and is a validated target for therapy of solid cancers, including breast cancer [38, 50].

Since EGF is a known ligand for the Stat3 signaling pathway, and Stat3 has recently been shown to mediate Sox-2 expression in neural stem cells [41, 51], we hypothesized that this transcription factor could be a downstream target of EGFR signaling that mediates CSC phenotypes in breast cancer cells. Importantly, phosphorylation of Stat3 at Y705 is critical for homodimerization and nuclear translocation of the protein, thus mediating its transcriptional activity [51]. Synthetic triterpenoids are a novel class of multifunctional inhibitors that have been shown to inhibit the growth of many solid tumors. In particular, an imidazole derivative of synthetic oleanane triterpenoid (CDDO-Im) has been shown to be a potent inhibitor of Stat3 phosphorylation at Y705 [52]. Additionally, we showed here that blocking Stat3 phosphorylation with CDDO-Im effectively inhibited EGF-induced increases in breast cancer SP cells and upregulation of Sox-2, Oct-4, and Nanog expression. Furthermore, CDDO-Im also blocked MCM- and EGF-induced decreases in E-cadherin and increases in N-cadherin expression as well as β-catenin nuclear localization. These results are significant since formation of CSCs has been linked to epithelial-to-mesenchymal transitions, which is characterized by loss of E-cadherin, a gain of N-cadherin expression, and nuclear translocation of β-catenin. In turn, these changes are associated with increased tumor cell motility and tumorigenicity. Importantly, the increase in tumor cell motility we observed upon culture with MCM was also inhibited by CDDO-Im. These results are of particular interest since CDDO-Im effectively blocked these TAM-induced CSC phenotypes at nM concentrations. Therefore, due to its potency, CDDO-Im may be a useful chemotherapeutic drug to be developed for cancer therapy targeting CSCs


As summarized schematically in Figure 7, our collective results demonstrate that TAMs play a key role in promoting maintenance of CSC phenotypes in murine breast cancer cells. Importantly, we have identified a novel mechanism of regulation that is achieved by paracrine EGF signaling between TAMs and tumor cells. We further demonstrated that this signaling involves activation of the EGFR/Stat3 signaling pathway and the downstream upregulation of the Sox-2 transcription factor. Critically, this TAM/tumor cell crosstalk required EGFR and Stat3 activity and could be blocked by inhibitors of either EGF or Stat3. To our knowledge, we are first to describe this unique interaction between TAMs and breast cancer cells via EGF/EGFR/Stat3 signaling that is critical for Sox-2 expression and breast CSC maintenance. Importantly, our report identifies a novel role for macrophages in breast CSC regulation and establishes a rationale for targeting the EGFR/Stat3/Sox-2 signaling pathway for CSC therapy.

Figure 7.

Schematic of interactions between macrophages and tumor cells in the breast tumor microenvironment. EGF released from TAMs binds to and activates EGFRs on tumor cells resulting in downstream phosphorylation of Stat3. Activation of Stat3 promotes transcription and expression of Sox-2 and other target genes involved in maintenance of CSC phenotypes. Acquisition of these phenotypes by tumor cells results in increased tumor growth, metastasis, and chemoresistance. Induction of such TAM-driven CSC phenotypes in tumor cells can be inhibited by anti-EGF neutralizing antibody, AG1478—a small molecule inhibitor of EGFR protein tyrosine kinase activity or CDDO-Im—a small molecule inhibitor of Stat3 phosphorylation and activation. Abbreviations: CSC, cancer stem cell; EGF, epidermal growth factor; TAM, tumor-associated macrophage.


This work was supported by grants from the Major State Basic Research development Program of China (973 program): Grant No. 2013CB967202 (to Y.L.); the National Science Foundation of China (NSFC): Grant Nos. 91029734 and 81071711 (to Y.L.); Grant Nos. 30830096 and 09ZCZDSF04000 (to R.X.), Grant SFP 1645 from the EMD Serono Research Center (to R.A.R.).


Partial funding for this work provided by Merck Serono.