|Home | About | Journals | Submit | Contact Us | Français|
Colony stimulating factor-1 (CSF-1) and its receptor (CSF-1R) have been implicated in the pathogenesis and progression of various types of cancer, including breast cancer. This is based on high levels of circulating CSF-1 in patient sera with aggressive disease and increased CSF-1R staining in the tumor tissues. However, there have been no direct in vivo studies to determine whether a CSF-1 autocrine signaling loop functions in human breast cancer cells in vivo and whether it contributes to invasion. Recently, in mouse and rat models, it has been shown that invasion and metastasis are driven by an EGF/CSF-1 paracrine loop between tumor cells and host macrophages. In this macrophage-dependent invasion, tumor cells secrete CSF-1 and sense EGF, while the macrophages secrete EGF and sense CSF-1. Here we test the hypothesis that, in human breast tumors the expression of both the CSF-1 ligand and its receptor in tumor cells leads to a CSF-1/CSF-1R autocrine loop which contributes to the aggressive phenotype of human breast tumors. Using MDA-MB-231 cell derived mammary tumors in SCID mice we show here for the first time in vivo that invasion in a human mammary tumor model is dependent on both paracrine signaling with host macrophages as well as autocrine signaling involving the tumor cells themselves. In particular we show that, the autocrine contribution to invasion is specifically amplified in vivo through a tumor microenvironment induced upregulation of CSF-1R expression via the transforming growth factor-β1.
Metastasis is a multi-step process that requires increased motility of the tumor cells inside the primary tumor and invasion of surrounding tissues and blood vessels. The tumor microenvironment has an essential role in promoting these steps of motility and invasion in tumor cells, through either secretion of chemotactic factors or direct interactions with stromal cells. In rat and mouse mammary tumors, tumor- associated macrophages are essential for promoting angiogenesis, matrix remodeling and chemotactic motility of the tumor cells (1). In particular, a paracrine interaction between macrophages and tumor cells, that involves epidermal growth factor (EGF) and colony stimulating factor 1 (CSF-1), is the driving force for relay chemotaxis supporting macrophage-mediated invasion in both transgenic mouse and rat mammary tumors. During relay chemotaxis, tumor cells secrete CSF-1 and sense EGF, while the macrophages secrete EGF and sense CSF-1. This phenomenon has been studied extensively both by intravital imaging in living animals as well as by reconstituting the interaction of the two cell types in vitro (2–4). Additional studies support the importance of macrophages in invasion and metastasis of the tumor cells. Absence of CSF-1 in the mammary cancer-susceptible PyMT mice retarded tumor progression and metastasis but did not affect primary tumor development (5), directly implicating macrophages and CSF-1 as important regulators of invasion and metastasis. In later work with xenogeneic tumors in mice, blockade of CSF-1 through antisense oligonucleotides or neutralizing anti-CSF-1 antibodies also reduced primary tumor growth and angiogenesis and prolonged long-term survival (6–7).
In humans, patient sample data has suggested that CSF-1 and its receptor might play critical roles during progression of tumors of the female reproductive system and other solid tumors. Expression of the CSF-1 receptor (CSF-1R) has been associated with adverse clinicopathological prognostic outcome in ovarian, endometrial and breast carcinomas (8–10). Expression of CSF-1R has also been detected in prostate cancer cells in tumors with elevated Gleason scores (11). Interestingly, CSF-1 is also expressed in ovarian and endometrial tumors and cell lines (12) and in breast cancer (10, 13) and CSF-1 and the CSF-1R are co-expressed in greater than 50% of mammary tumors (14). In addition, increased circulating CSF-1 levels are a prognostic marker for epithelial ovarian cancer (15, 16) and are elevated in a large proportion of endometrial cancers (16, 17). Elevated circulating CSF-1 was also suggested to be an indicator of early metastatic relapse in breast cancer patients (13). These observations have led to the hypothesis that there may be an autocrine loop involving tumor cells expressing both CSF-1 and CSF-1R, which contributes to invasion and metastasis in human tumors (14). Several in vitro studies now support this hypothesis. Human lung cell lines and breast cell lines that express CSF-1R, but not the ligand, show increased invasion in vitro into an amniotic basement membrane upon stimulation with CSF-1 (18). Similarly, when the normal non-invasive mouse mammary cell line HC11, that expresses CSF-1 but not the receptor, is forced to overexpress CSF-1R, it shows rapid growth and colony formation in soft agar, increased invasion through matrigel, and a higher incidence of lung tumors after tail vein injections into BALB/c mice, compared to the parental line (19). Additionally, when CSF-1 and its receptor are both stably overexpressed in the MCF10A human mammary cell line, the cells exhibit abnormal acinar morphogenesis and increased motility in an in vitro wound-healing assay (20).
These latter in vitro studies were performed with cells in which CSF-1 or its receptor, or both, were overexpressed, in order to mimic hypothesized autocrine signaling and they did not assess the contribution of the tumor microenvironment to the onset of autocrine signaling. Indeed, there have been no direct in vivo studies to determine whether a CSF-1 autocrine signaling loop is induced in the microenvironment of the primary tumor and if it contributes to invasion in human breast cancer. Moreover, the contribution of autocrine CSF-1 signaling to invasion cannot be investigated in the popular transgenic mouse models, because neither the mouse epithelial cells nor the mouse tumor cells express the CSF-1R. Epithelial expression of the CSF-1R transcript in human mammary epithelium is hormonally regulated via a glucocorticoid response element that is absent from the murine locus (21). Therefore, additional models and studies are needed to test the hypothesis that CSF-1/CSF-1R autocrine signaling contributes to invasion of human breast cancer cells in vivo.
Based on the rodent models and the patient data described above, the CSF-1R could be involved in human breast tumor invasion in at least two ways. The rodent studies point to macrophage-assisted invasion involving an EGF/CSF-1 paracrine loop and the patient studies suggest that tumor cell invasion is regulated in an autocrine manner by CSF-1, provided the tumor cells express both CSF-1 and CSF-1R. In this study, we directly address these two possibilities in vivo.
We used the human breast tumor cell line MDA-MB-231 as a model for our analysis, because its triple-negative status (estrogen/progesterone receptor negative, HER2 negative) categorizes it in the basal-like subtype of breast cancers (22). These are generally the most aggressive and highly metastatic breast tumors. As we show that MDA-MB-231 cells express both CSF-1 and the CSF-1R, this cell line, without further manipulation, is an appropriate one in which to study the role of autocrine signaling in aggressive breast tumors.
MDA-MB-231 cells (ATCC) were cultured in DMEM (Invitrogen, Carlsbad, CA) with 10% fetal bovine serum (FBS). A stable GFP-expressing line was made by transfecting the MDA-MB-231 cells with the pEGFP-C1 vector (Clontech, Mountain View, CA) using the Amaxa nucleofector (Amaxa, Gaithersburg, MD), selecting in 500 μg/ml G418 for 2 weeks and sorting for the GFP-positive cells. Mouse BAC1.2F5 macrophages (23) were cultured in α-MEM with 10% FBS and 36 ng/mL of Human recombinant CSF-1 (a gift from Chiron Corp., Emeryville, CA).
All procedures were conducted in accordance with the National Institutes of Health regulations, and approved by the Albert Einstein College of Medicine animal usecommittee. A total of 2×106 MDA-MB-231-GFP cells per animal were resuspended in sterile PBS with 20% collagen I (BD Biosciences, Franklin Lakes, NJ) and injected into the lower left mammary fat pad of severe combined immunodeficiency mice (SCID) (NCI, Frederick, MD). All experiments were performed on tumors that were 1.2 – 1.5 cm in diameter. SB431542 (Tocris), where indicated, was injected intraperitoneally in the xenografts at 100mg/kg 3 hours prior to experiments. FVB/NJ mice transgenic for the Polyoma Middle T oncogene (PyMT) under the MMTV-LTR, generated as previously described in (24), were used for in vivo invasion assays at 12–14 weeks of age.
Cell collection into needles placed into live anesthetized animals was carried out as described previously (25). Cells can only enter the needles by active migration since a block is used to prevent passive collection of cells and tissue during insertion of the needle into the tissue. Cell migration has been demonstrated to be required for cell collection (3). After 4 hours, the needles were removed and the total number of cells collected was determined by 4′,6-diamidino-2-phenylindole (DAPI) staining. The chemoattractants used include mouse EGF (Invitrogen), human recombinant EGF (Invitrogen) and human recombinant CSF-1 (Chiron Corp.), at final concentration 25nM. To inhibit EGFR, Iressa (AstraZeneca), a tyrosine kinase inhibitor specific for EGFR, was used in the needles in final concentration 5 μM, as described previously (3). To inhibit CSF-1R, a monoclonal rat anti-mouse antibody (AFS98, (26)) or a monoclonal mouse anti-human antibody (MAB3291, R&D Systems, Minneapolis, MN) were used in the needles at a final concentration of 125 μg/ml. Controls were same quantities of DMSO (Sigma), rat IgG (BD Biosciences) or mouse IgG same isotype antibodies (R&D Systems).
Typing of the collected cells was done as described previously (3), using cell type-specific antibodies rabbit anti-pancytokeratin (Santa Cruz Biotechnology Inc, Santa Cruz, CA) for carcinoma cells and rat anti-F4/80 (27) for macrophages. DAPI was used for counting total cells.
The assay was performed as described in (4). Briefly, MDA-MB-231-GFP cells were plated in the presenceor absence of BAC1.2F5 macrophages. The next day they were overlaidwith collagenI (BD) and incubated for 24 hours after which the assay was fixed with 4% formaldehyde. Where indicated,cells were pretreated with Iressa (1 μM final), or the CSF-1R blocking antibodies described above (final 10–25 μg/mL) for 2 hours. The fixed assay was analyzed by confocal microscopy. Invasion of the tumor cells was quantified as the proportion of fluorescence more than 20 μm into the collagen.
For the in vitro stimulations, the cells were starved overnight in DMEM/0.35% BSA, prior to addition of cytokines for 1, 4, 7, 24 and 75 hours. The results shown are for the 4 hour stimulation, when the effect of TGFβ1 was maximal (EGF and CSF-1 were identical for all timepoints). RNA was extracted with the RNeasy Mini kit (Qiagen, Valencia, CA), and 1 μg of total RNA was reverse transcribed using SuperScript II (Invitrogen). 2–5 ng of total cDNA was used per real-time PCR reaction or standard PCR (30 cycles) with specific primers (see Supplementary Table for primer sequences). Electrophoresis of the PCR productswas performed on 1.5% agarose gel and visualized by ethidium bromidestaining.
For the comparison of the average primary tumor cells (APTC) with the cultured MDA-MB-231 cells, amplified total cDNA was used as input in the real-time PCR. The detailed protocol and validation of the technique has been published elsewhere (28, 29). Briefly, for the APTC sample, tumors from MDA-MB-231-GFP xenograft animals were excised, mechanically dissociated into single cell suspensions on ice, sorted for the GFP-positive cells and lysed with the RNeasy Micro kit (Qiagen). Total RNA from APTC and from cultures of MDA-MB-231 cells was converted to cDNA and amplified with the SMART amplification kit (Clontech). The final amplified cDNA was purified with Qiagen MinElute columns and 2 ng was used in the real-time PCR per reaction.
Quantitative PCR analysis was performed as described previously (29), using the Power SYBR Green PCR Core Reagents system (Applied Biosystems). Each PCR reaction was performed in triplicate, and the mean threshold cycle (CT) values were used for analysis. All the genes tested were compared with two housekeeping genes (β-2 microglobulin and GAPDH) for the analysis. Resultswere evaluated with the ABI Prism SDS 2.1 software.
For the western blot analysis whole cell lysates were prepared by washing the cells with cold PBS, followed by direct addition of SDS-PAGE sample buffer and sonication and boiling of the lysate. Western blots were carried out as follows: the samples were resolved by SDS-PAGE, transferred to nitrocellulose, blocked in odyssey blocking solution (LiCor), incubated in primary antibodies overnight at 4°C (MAB3291 anti-human CSF-1R antibody and anti-β-actin antibody AC15 from Sigma-Aldrich for control), secondary antibodies for 1 hour at RT (Mouse 680 and Rabbit 800 from Licor and chicken 800 from Rockland), and finally analyzed using the Odyssey (LiCor). Visualization and processing of images was performed with ImageJ (NIH).
Results shown are representative of at least 3 experiments with duplicate plates of cells for the in vitro experiments and at least 5 different mice per point for the in vivo experiments. All statistical analyses were assessed using a two-tailed Student’s t test.
In previous studies we have characterized a paracrine interaction between breast carcinoma cells and macrophages involving EGF and CSF-1, using both xenograft (MTLn3 rat breast tumor cells) and transgenic (MMTV-PyMT) mouse mammary tumors (3, 4). In an effort to test whether similar interactions exist in a human breast cancer cell - derived mammary tumor we tested the expression of these growth factors and their ligands in the human breast cancer line MDA-MB-231 (Fig. 1A). Similar to the rodent breast carcinoma cells (4), the MDA-MB-231 cells express the mRNA for the EGF receptor (EGFR) but not for EGF. The MDA-MB-231 human breast tumor cells, like their rodent counterparts, also express mRNA for CSF-1, raising the possibility of an EGF/CSF-1 paracrine loop akin to the paracrine loop we have observed with rodent tumors, in which EGF is produced by tumor-associated macrophages (Fig. 1A). Based on this gene expression pattern, we used a 3D in vitro invasion assay where the tumor cells, either alone or in the presence of macrophages, are monitored for their invasive ability through a 3D collagen matrix (4). The invasion of the MDA-MB-231 cells was enhanced in the presence of macrophages (Fig. 1C, and also in (4)), demonstrating that the EGF/CSF-1 paracrine interaction with the macrophages exists for the human breast tumor cell-derived mammary tumors as seen in rat and mouse mammary tumors.
However, as described above, patient data have also shown concomitant expression of CSF-1 and its receptor in tumor cells of aggressive breast cancers (14) and unlike their rodent counterparts, the human MDA-MB-231 cells also express the CSF-1R as well as the mRNA for CSF-1 (Fig. 1A and 1B). Thus, the MDA-MB-231 cell line is a good model to test the potential of a CSF-1/CSF-1R autocrine contribution to invasion and metastasis in human breast cancer, without forced and artificial overexpression of either molecule.
To investigate the relative contributions of autocrine and paracrine signaling to invasion and intravasation in vivo, we made mammary tumors using MDA-MB-231 cells in mice and we used the in vivo invasion assay (25) to measure the invasive potential of these cells. In this assay, as described in detail in the materials and methods, microneedles containing matrigel and a chemoattractant are inserted into the primary tumors while the animal is alive and under anesthesia. This assay mimics natural blood vessels inside the primary tumor where EGF secreted by macrophages attracts the invasive tumor cells to these blood vessels, a movement that eventually will lead to intravasation and hematogenous metastasis (2). Using this assay in the MDA-MB-231 xenografts, we collected the invasive cells that responded to EGF (Fig. 2A), extruded them out of the needles and identified the cell types with cell type-specific antibodies. As expected from previous studies with rat MTLn3 xenografts and the MMTV-PyMT transgenic mouse models (3), the invasive subpopulation in MDA-MB-231 tumors was comprised of macrophages and tumor cells (Fig. 2B). However, the proportion of macrophages was less than observed with rodent tumors (approximately 6% versus 25% respectively, Figs. 2B and 2C), implying that MDA-MB-231 cell-derived mammary tumors are less dependent on macrophages for invasion and hematogenous dissemination. The possibility that the interaction between tumor cells and macrophages is hindered in this xenograft model due to species difference is unlikely; MDA-MB-231 cells respond to mouse EGF as well as to human EGF both in vitro and in vivo (Supplementary Fig. 1) and mouse macrophages respond similarly to both mouse and human CSF-1 (30).
To further investigate the relative contributions of paracrine and autocrine signaling in vivo, we performed a series of experiments in the MDA-MB-231 derived primary mammary tumor using the in vivo invasion assay. First, we evaluated the contribution of the EGFR to the invasion in primary MDA-MB-231 tumors (Fig. 3A). There was a decrease in the number of invasive cells collected in response to EGF in the presence of the EGFR inhibitor, Iressa (Fig. 3B). A similar inhibition by Iressa was observed when the cells were collected in response to CSF-1, suggesting a CSF-1/EGF paracrine loop between the human tumor cells and the macrophages in vivo. Interestingly, for either EGF- or CSF-1- driven invasion, the inhibition by Iressa was incomplete (Fig. 3B). In contrast, in an identical previous experiment in a rodent mammary tumor (3), where there was complete inhibition by Iressa, consistent with a more significant involvement of macrophage-supplied EGF in the mouse tumors.
We next examined the contribution of the CSF-1R to the paracrine and autocrine interactions in these human cell line-derived mammary tumors. To investigate the role of this receptor in tumor cells versus macrophages, we used an antibody that specifically blocks the mouse CSF-1R ((3) and supplemental Fig. 2A) to block the paracrine interaction between the macrophages and the tumor cells (Fig. 3A). In contrast, to block only autocrine signaling in the human tumor cells (Fig. 3A), we used an antibody that specifically blocks the human (but not the mouse) CSF-1R (supplemental Fig. 2A). In the MDA-MB-231 mammary tumors, blocking either the human or the mouse CSF-1R only partially decreased the number of total cells collected in response to EGF (Fig. 3C). However, a combination of both antibodies completely inhibited invasion to background levels (Fig. 3C, matrigel alone control bar). This is not an effect of doubling the total amount of antibody, as using twice the amount of each antibody alone did not produce this synergistic result (Supplementary Fig. 2B). Therefore, in MDA-MB-231 mammary tumors, a mix of paracrine signaling between tumor cells and macrophages and autocrine signaling in the tumor cells is involved in invasion in vivo.
We used the 3D in vitro invasion assay, where MDA-MB-231 cells alone or in co-culture with macrophages are monitored for their invasion ability through a collagen matrix (Fig. 4A), with the specific inhibitors and the blocking antibodies described above, to investigate autocrine and paracrine contributions of EGFR and CSF-1R to in vitro invasion. In co-cultures of MDA-MB-231 cells and macrophages, pre-treatment with the EGFR specific inhibitor, Iressa dramatically decreased the ability of MDA-MB-231 cells to invade (Fig. 4B). Additionally, incubation with the blocking antibody to the mouse (macrophage) CSF-1R substantially reduced invasion through the collagen matrix as well, almost to the background level of the MDA-MB-231 cells alone (Fig. 4C), while it had no effect on the migration of the tumor cells when alone in the collagen matrix (Supplemental Fig. 3). The requirement for the tumor cells’ EGFR function (Fig. 4B) together with the requirement for the mouse macrophage CSF-1R function (Fig. 4C), further supports the CSF-1/EGF paracrine interaction between the tumor cells and macrophages. However, incubation with the blocking antibody to the human tumor cell CSF-1R had no effect on the in vitro invasion of the tumor cells in the presence or absence of macrophages; i.e., invasion by the tumor cells was similar in the presence of the blocking or control antibodies (Fig. 4C and Supplemental Fig. 3). This result was unexpected, given that this same antibody inhibited invasion in vivo (Fig. 3C) and it suggests that autocrine CSF-1R signaling contributes more to invasion in vivo in the primary tumors than in vitro.
In Fig. 1A, we examined the gene expression pattern shown for the MDA-MB-231 cells, in cell culture. In order to test for gene expression changes that could account for the difference in behavior of these cells in vitro and in vivo, we also isolated RNA from the average primary tumor cells after their growth in the microenvironment of the primary tumor. Interestingly, when comparing the average primary tumor cells to the cultured MDA-MB-231, we found that the mRNA expression of CSF-1R in the MDA-MB-231 cells from tumors was 2.5 to 5-fold upregulated compared to the same cells in culture (Fig. 5A). We also tested expression of the mRNAs for EGFR, CSF-1 and TGFα, but none showed such striking upregulation as did the CSF-1R mRNA (Fig. 5A). These results indicate that the CSF-1R mRNA upregulation is a phenotype specific to the growth of tumor cells in vivo in the primary tumor and not a direct result of the stimulation by the growth factors involved in the paracrine/autocrine loops tested in this study. In fact, stimulation of MDA-MB-231 cells with either EGF or CSF-1 was not sufficient to produce the CSF-1R mRNA upregulation seen in the invasive tumor cells isolated from the primary tumor (Fig. 5B and 5C).
To explore the microenvironmental signal that causes the upregulation of CSF-1R expression in tumor cells in vivo, we investigated the involvement of TGFβ1. Induction of the CSF-1R mRNA by TGFβ1 has been reported in vascular smooth muscle cells and in cervical cancer cells lines (31, 32). MDA-MB-231 cells express the receptors for TGFβ1 (33) and thus could be responsive to it. Considering the established role of TGFβ signaling in tumor progression in breast cancer (34, 35), we hypothesized that TGFβ1 is the microenvironmental signal that prompts the upregulation of CSF-1R in human breast tumor cells. Indeed, when we stimulated MDA-MB-231 cells in vitro with TGFβ1, we found a 2.5–3 -fold upregulation of their CSF-1R mRNA expression. To address the role of TGFβ1 in the upregulation of CSF-1R in vivo, we treated MDA-MB-231 xenografts with SB431542, a specific inhibitor to the TGFβ receptor (TBRI) (32, 36). Treatment with this inhibitor significantly decreased the CSF-1R mRNA expression of the primary tumor cells (Fig. 6B). Additionally, invasion in the treated animals in vivo was significantly reduced compared to the animals injected with DMSO vehicle control (Fig. 6C), and similar to numbers observed when the human CSF-1R blocking antibody was used in the in vivo invasion assay (Fig. 3C). Our results suggest that TGFβ1 is the signal inside the primary tumor that promotes the in vivo autocrine CSF-1R mediated invasion of the human breast tumor cells.
Increased tumor expression of CSF-1 and its receptor has been observed in several types of solid tumors, including breast cancer (14). The correlation of expression of both the receptor and its ligand with cancer aggressiveness led to the hypothesis of autocrine signaling through CSF-1R in these tumors. In addition to this potential autocrine role, the elevated expression of CSF-1 by some tumors suggested that CSF-1 might contribute to increased infiltration of macrophages into the primary tumor and their interaction with the tumor cells, a phenomenon also linked to tumor aggressiveness (13, 37). We have used the highly metastatic human breast tumor line MDA-MB-231 to address the mechanism of contribution of CSF-1R in invasion by breast cancer cells. Here we show that invasion of MDA-MB-231 cells in vitro is paracrine requiring the presence of macrophages for optimum invasion. However, using MDA-MB-231 derived mammary tumors,, we show that invasion in these human cell line-derived tumors in vivo involves both autocrine CSF-1R signaling, as well as macrophage-associated signaling through the EGF/CSF-1 paracrine loop. Interestingly, the autocrine contribution in human breast tumor cell invasion is greatly enhanced in vivo and this is associated with a transforming growth factor-β1 (TGFβ1)-mediated increased expression of the CSF-1R mRNA by the carcinoma cells in the microenvironment of the primary tumor.
As described in the introduction, transgenic mouse models have failed to address the autocrine role of the CSF-1R in mammary tumor cells, because of the lack of CSF-1R expression in murine mammary epithelia. Notably, in a study by Kirma et al., two transgenic mouse models were prepared where MMTV promoter-driven transgenic mice were forced to overexpress either human CSF-1 or the mouse CSF-1R in the mammary epithelium (38). Both mouse models exhibited an increased rate of glandular dysplasia and ductal hyperplasia and an increased tissue macrophage infiltration. This model further illustrated the importance of CSF-1 and the CSF-1R in cancer progression, but did not evaluate the effects of autocrine regulation through their co-expression in the same tumor cells. In another study, Sapi et al. used tail vein injection in mice of CSF-1R overexpressing breast tumor cell lines to address a potential role of autocrine CSF-1R signaling in metastasis in vivo (19). The mice that received the CSF-1R overexpressing cells showed increased lung metastasis compared to those receiving the parental cells not expressing the CSF-1R. However, this experimental metastasis system does not address the invasiveness of tumor cells in the primary tumor. In our study, we show a direct role for autocrine CSF-1R signaling in invasion in primary mammary tumors, resulting from spontaneous expression of CSF-1/CSF-1R in human breast cancer cells. Considering that invasion is an early step in the metastatic cascade, our results indicate that both autocrine and paracrine CSF-1 loops significantly contribute to dissemination of cancer cells and progression in human breast cancer. In addition, it is possible that different microenvironments of the same tumor have different contributions to paracrine and autocrine invasion depending on local TGFβ levels and local macrophage densities. The important clinical implication of our findings is that since the mammary tumor microenvironment is inducing the elevated expression of CSF-1R, blocking CSF-1 or its receptor in patients would not only suppress macrophage infiltration in breast tumors and the EGF/CSF-1 paracrine loop, but also the CSF-1/CSF-1R autocrine-mediated invasion by carcinoma cells.
While MDA-MB-231 cells do not express significant amounts of EGF mRNA, they do express the mRNA for TGFα, another EGFR ligand (Fig. 1A). However, invasion by MDA-MB-231 cells does not appear to be regulated in an autocrine fashion by TGFα. In particular, the EGFR inhibitor, Iressa, did not affect the ability of the cells to move by themselves in the in vitro invasion assay (data not shown), but decreased their invasive ability when they were in the presence of macrophages, due to the effect on the paracrine loop (Fig. 4B). Furthermore, simultaneously blocking both the tumor cell and macrophage CSF-1 receptors brings in vivo invasion to background levels (Fig. 3C), showing that there is no additional contribution from autocrine activation of the EGF receptor.
Interestingly, the human MDA-MB-231 breast tumor cells are more autocrine for CSF-1 in vivo than in vitro, a finding explained by the upregulation of the CSF-1R mRNA in tumor cells within the microenvironment of the primary tumor. We also show that TGFβ1 signaling is likely the microenvironmental signal that causes this in vivo CSF-1R upregulation in the human tumor cells. Tumors are regularly infiltrated by leukocytes, macrophages, myeloid precursors and other cells, which are the source of TGFβ secretion and accumulation in primary tumors (39). TGFβ is associated with tumor progression (40), through several mechanisms such as induction of epithelial-mesenchymal transition, myofibroblast generation, inhibition of host immunosurveillance and others (35). Our study suggests an additional role for TGFβ signaling in invasion and metastasis of human breast cancer cells, through an upregulation of the CSF-1R and induction of the CSF-1/CSF-1R autocrine loop in human primary tumor cells.
We wish to acknowledge Drs Jeffrey Segall, Dianne Cox and Anne Bresnick for their help in discussions. For technical help, we wish to thank the AECOM Genomics facility and the AECOM Flow Cytometry Facility, especially Drs Jinhang Zhang, Lydia Tesfa and Siu-Hong Ho (supported by NCI P30 CA 013330).
Financial support: NIH RO1 CA 113395 (AP, JC), NIH RO1 CA 26504 (ERS), CA126511 (YW, JC) and NIH PO1 CA 100324 (JW, YW, SG, ERS). SG is the recipient of Young Investigator Award from Breast Cancer Alliance Inc.