|Home | About | Journals | Submit | Contact Us | Français|
Tumor associated macrophages (TAMs) are implicated in breast cancer progression and metastasis, but relatively little is known about the genes pathways in these cells that contribute to malignant phenotypes. The transcription factor Ets2 is a direct target of signaling pathways involved in regulating macrophage functions during inflammation. To test whether Ets2 in TAMs modulated mouse mammary tumor growth and metastasis a genetic approach was used to conditionally delete Ets2 in TAMs. Ets2 deletion in TAMs decreased the frequency and size of mammary tumor metastases to lung in three different metastatic models. Expression profiling and chromatin immunoprecipitation assays with isolated TAMs established that Ets2 repressed several well characterized inhibitors of angiogenesis. Consistent with these results, Ets2 ablation in TAMs led to decreased tumor angiogenesis and growth. An Ets2-TAM expression signature was identified within human breast cancer expression data and this signature could retrospectively predict overall survival of breast cancer patients in two independent sets of human breast cancer microarray data. In summary, we have identified Ets2 as a critical factor that acts to enhance mammary tumor growth and metastasis by regulating a transcriptional program in TAMs.
Sporadic human cancer results from somatic gene mutations that lead to aberrant growth, survival, genetic instability and increased motility of tumor cells (1). In addition to genetic complexity, it is increasingly apparent that cellular complexity inherent in the tumor stroma plays an active role in promoting all stages of tumor progression (2). Among the many cell types in the tumor stroma, the tumor associated macrophage (TAM) is a broadly defined myeloid cell type that has been implicated in tumor progression (2). TAMs are thought to be a polarized M2 subtype of macrophage that promote tumor growth, invasion, and angiogenesis (3). Alternatively, the pleiotropic effects of macrophages within the tumor microenvironment may be mediated by distinct subpopulations of TAMs that can selectively affect distinct processes such as tumor angiogenesis or invasion (4, 5).
The link between TAMs and tumor progression is especially well established in breast cancer. Human clinical studies have demonstrated that a high focal infiltration of TAMs in primary human breast tumors directly correlates with tumor cell invasion, increased vascularization, axillary lymph node involvement and reduced relapse-free survival of patients (6-9). In a mouse mammary tumor model, genetic ablation of Colony Stimulating Factor-1 (Csf-1), a growth factor critical for macrophage growth, differentiation and survival, results in a reduction in mammary TAMs and a lower incidence of lung metastasis (10).
Ets2, a member of Ets family of transcription factors, is a direct effector of CSF-1 signaling pathways that modulates macrophage functions and survival during inflammation (11, 12). ETS2 activates or represses transcription of target genes in a context-dependent manner (13, 14). Elevated expression of ETS2 has been correlated with human breast cancer (15). However, in mouse mammary tumor models, Ets2 promotes tumor progression from the stroma and not the tumor epithelial cell (16).
In the current study, a genetic approach was used to define the action of Ets2 in mouse mammary TAMs. The results demonstrate that Ets2 in TAMs decreased the growth rate of the primary tumor and tumor metastases and the mechanism involved repressing genes that are inhibitors of angiogenesis. 133 human genes orthologous to the Ets2-TAMs profile could retrospectively predicted disease free survival among patients present in two human breast cancer microarray data sets (17, 18). These results identify an Ets2-regulated transcriptional program in TAMs that regulates growth and spread of breast tumors.
The Ets2LoxP allele, Ets2db knockout allele, MMTV-PYMT transgenic mice and Lys-Cre knockin allele were previously described (19-22). The c-fms-YFP construct was identical to the published c-fms-EGFP construct except for the substitution of YFP (23). Transgenic mice were produced by standard DNA microinjection procedures. All alleles utilized were >10 generations FVB/N background. Use and care of mice were approved by the Ohio State University Institutional Animal Care and Use Committee.
The two breast cancer cell lines used, Met-1 (MMTV-PyMT)) and MVT-1 (MMTV-c-Myc;MMTV-VEGF) were used (24, 25). The cell lines were cultured in Dulbecco’s Modified Eagle Medium containing 10% Fetal Bovine Serum at 37°C in 5% CO2 incubator. Cultured tumor cells were harvested at 80-90% confluence and suspended in filtered cold 0.9% NaCl. Three million Met-1 cells or 200,000 MVT-1 cells were injected into the tail vein or mammary gland, respectively. Tail-vein and orthotopic injected animals were dissected 18 and 35 days post-injection, respectively.
Minced mammary glands or lungs with metastatic tumors were digested with 20mg Collagenase Type 2 (Worthington), 480 Units DNAseI (Boehringer) and 1mM MgCl2 at 37°C and stroma was enriched by gravity separation (26). The YFP-positive cell population was sorted using Fluorescent Activated Cell Sorting with the BD FACSAria™.
RNA extraction and cDNA preparation were done as described previously (27). For samples used in microarray analysis, RNA was extracted with the RNeasy Stratagene micro-prep column (Stratagene) as per the manufacturer’s instructions. Two independent sets of RNA isolated from different TAMs/mice than RNA used for the microarrays were used for verification.
Real-time quantitative RT-PCR was conducted using the Roche Universal Probe Library system (Roche Diagnostics) in an Icycler iQ Real-Time Detection system (Bio-Rad) as described previously (27). Primer-probe combinations are available on request.
Tumor tissues were fixed in formalin overnight, processed, paraffin-embedded and 5μM sections were prepared. For immunostaining, rat α-mouse F4/80 (1:40 dilution; Caltag Labs), rat α-mouse CD31 (1:50 dilution, Abcam), mouse α-human THBS1 (1:50 dilution, Abcam), mouse α-mouse THBS2 (1:50 dilution, BD Biosciences), goat α-mouse SPARC (1:100 dilution, B.D Biosciences), and mouse α-mouse BrDU (1:50 dilution, DAKO) primary antibodies were used. Biotinylated goat α-rat, goat α-mouse, or donkey α-goat (BD Biosciences) were the secondary antibodies used for immunohistochemical analysis. Images were acquired using an Axioscope 40 microscope (Zeiss) equipped with an Axiocam HRc camera (Zeiss). Immunohistochemical data was quantified by calculating the area of antibody staining per unit area of tumor using Metamorph 6.0 software. Wholemount hematoxylin staining of lungs was performed as described (28).
For co-localization studies, frozen sections of mammary tumors fixed in 4% paraformaldehyde were double immunostained with α-F4/80 antibody (Alexa-594 secondary antibody, Invitrogen) and either α-THBS2, α-THBS1 or α-SPARC antibody (Alexa-488 secondary antibody, Invitrogen). Nuclei were stained with DRAQ5. Images of stained mammary tumors were acquired using a Zeiss 510 META laser scanning confocal microscope Results are presented as the percentage of F4/80-positive or negative cells that had co-localized staining in or around (extracellular space) for α-THBS2, α-THBS1 or α-SPARC, respectively,.
ChIP assays were performed as described (27). Immunoprecipitation was carried out with 2.5μg of antibodies. The ETS2 antibody has been previously described (19). Rabbit α-mouse HDAC1 and rabbit-IgG were purchased from SantaCruz Biotechnology and Upstate, respectively. For lung TAMs, the immunoprecipitated chromatin was amplified using an unbiased genome amplification kit (Sigma Aldrich). Samples were analyzed by real-time PCR using the Roche Universal Probe Library (Roche Diagnostics) and the Faststart TaqMan master kit (Roche Diagnostics).
Microarrays were performed on the Mouse Affymetrix 130A.2 platform. The primary data was analyzed by a modified RMA method to yield an average gene expression value (29, 30). The detailed description of the experiment and subsequent data analysis is presented in Supplementary Table 1A.
A high confidence 142 probe set (p<0.05) human Ets2-TAM signature was generated by comparing 407 mouse probe sets (357 genes, absolute INT > 1.5) to the 98 lymph-node negative Rosetta cohort (www.rii.com/publications/2002; divided in to 2 groups based on lymphocyte/leukocyte infiltration status (31). For survival analysis, the 142 probe set Ets2-TAM signature was used as a query to retrieve gene-expression data from Stockholm (GSE1456) breast cancer microarrays (downloaded from NCBI-GEO webpage). Similarly, gene-expression data was also extracted from total and lymph node-negative Rosetta microarrays. The resultant data sets were loaded onto BRB-Array Tools as described in Supplementary Table 3. Briefly, unsupervised K-means clustering of each dataset was performed by using Cluster 3.0 (32) and samples were assigned into two groups. Kaplan-Meier survival analysis was performed by using the Survival Analysis module of the software package StatsDirect (StatsDirect Ltd). Significance of survival analyses was performed by using the Log-Rank (Peto) test.
For lung metastases data, a non-parametric Kruskal-Wallis test with no multiplicity adjustment was used to compare medians between experimental and control groups. A repeated measures ANOVA model was used to analyze mammary tumor progression between the genetic groups over a period of 42 days. This approach takes into consideration longitudinal data, and the following terms were included in the model: genetic group, time and interaction (genotype* time). For the statistical analysis of imaging data, an unpaired Student’s t-test was used. All the tests were two sided.
Cre/LoxP technology was used to conditionally delete Ets2 in TAMs in the PyMT model, a penetrant breast cancer model with a high frequency of lung metastasis (21). The conditional Ets2LoxP allele used for this study contained LoxP sites flanking exon3-exon5 so that Cremediated recombination of the region resulted in the generation of a null allele (19). The well-characterized Lys-Cre knockin allele was used to delete Ets2 specifically in the macrophage compartment (22). However, initial studies revealed that Cre-recombination in Lys-Cre;Ets2LoxP/LoxP mice was only 30-50% efficient (data not shown). To circumvent this problem, we adopted a strategy whereby mice contained one conditional Ets2LoxP allele and one conventional knockout allele, Ets2db (20). In the final cross, PyMT;Lys-Cre;Ets2db/+ males were crossed with Ets2LoxP/LoxP females to generate both the experimental genotype, PyMT;Lys-Cre;Ets2LoxP/db, and the control genotype, PyMT;Ets2LoxP/db (Supplementary Figure 1A). The frequency of Ets2 rearrangement in isolated mammary tumor macrophages varied between 70-90% with this allele configuration (Supplementary Figure 1B).
Tumor progression was monitored in females of the two genotypes. Tumor initiation was identical between experimental and control mice (data not shown). A small, but statistically significant, decrease in overall tumor growth was observed in the experimental group (Supplementary Figure 1C). This difference in tumor growth was not significant in the early carcinoma stage of progression, but was more pronounced during the late carcinoma stage (days 21-35 post-initiation; Supplementary Figure 1C). However, the final tumor burden and tumor volume were similar in both PyMT;Lys-Cre;Ets2LoxP/db and PyMT; Ets2LoxP/db mice (Supplementary Figure 1D).
Lung metastasis in both genetic groups was studied by whole-mount analysis (Figure 1A; Supplementary Figure 1E). After image acquisition, the size of the tumors relative to total lung area and the total number of metastases in PyMT;Lys-Cre;Ets2LoxP/db versus PyMT;Ets2LoxP/db mice were quantified. The results showed that both the size and number of lung metastases were significantly reduced in PyMT;Lys-Cre;Ets2LoxP/db mice compared to controls (Figure 1A right panel, size decreased 3-4 fold, p=0.001; Supplementary Figure 2A, number decreased 2-fold, p=0.02).
To confirm and extend the results obtained in the genetic PyMT model, a syngeneic model was employed. The highly metastatic cell line, MVT-1, derived from mice doubly transgenic for MMTV-c-Myc and MMTV-VEGF (25), was injected into mammary fat pads of Lys-Cre;Ets2LoxP/db and control Ets2LoxP/db female mice. After 35 days mice were euthanized and examined. While there was no difference in final tumor burden for the primary tumors (data not shown), the size of metastases per total lung area was three-fold reduced in the experimental Lys-Cre;Ets2LoxP/db group compared to the control group (Figure 1B). These results indicate that the effect of Ets2 is independent of the PyMT oncogene and also demonstrate that haploinsufficieny of Ets2 in the PyMT model is not a confounding factor.
To firmly establish that the effect of Ets2 in TAMs on metastasis was independent of effects at the primary mammary tumors, a tail-vein injection model was employed. A metastatic PyMT cell line, Met-1 (24), was injected into the circulation via the tail vein in the same two genetic groups as above. After 18 days mice were euthanized and metastases to lungs were quantified in H&E stained sections (Figure 1C). The results demonstrated that the size of lung metastases were significantly reduced more than three-fold in the Lys-Cre;Ets2LoxP/db mice compared to controls.
A potential explanation for the lower levels of metastasis observed in all three models might be that Ets2 regulated genes were required for macrophage survival and/or motility (11, 12). Immunostaining of tumor sections with F4/80 antibody, a marker for mature macrophages, revealed that Ets2 deletion did not result in a decrease in F4/80 positive macrophages associated with either primary or metastatic tumors (Supplementary Figure 2B-C, respectively).
To address the mechanism of Ets2 function in TAMs, mammary TAMs were isolated and subjected to gene expression profiling using the Affymetrix platform. To accomplish this, mammary TAMs were tagged using a c-fms-YFP transgene ((23); Supplementary Figure 3A). This transgene was incorporated into the breeding scheme outlined above to produce experimental PyMT;Lys-Cre;Ets2LoxP/db;c-fms-YFP and control PyMT;Ets2LoxP/db;c-fms-YFP mice. YFP-positive cells isolated from collegenase digested tissue by digital high-speed fluorescence activated cell sorting (FACS) represented approximately 10-15% of the total cells from the primary mammary tumor-site (Supplementary Figure 3B). Greater than 90% of these YFP-positive cells co-expressed macrophage markers like F4/80 (Supplementary Figure 3C). Typically, 3-5×105 YFP-positive TAMs could be isolated from a single mouse.
YFP-positive TAMs were isolated from both genetic groups at the stage when early carcinoma was initially detected in the PyMT model (21). The percentage of YFP-positive cells per mammary gland isolated by FACS was similar in both genetic groups, supporting the conclusion that a reduction in tumor macrophages was not responsible for the observed effects (Supplementary Figure 3B). Since macrophages have also been shown to play a central role in tissue remodeling during mammary gland development (33), YFP-positive macrophages were extracted from the mammary gland of Lys-Cre;Ets2LoxP/db;c-fms-YFP and Ets2LoxP/db;c-fms-YFP females approximately 14 days after the onset of puberty. We reasoned that the role of macrophages in tissue remodeling during mammary gland development would provide a useful comparison to unmask the tumor-specific effects of Ets2.
Expression profiling was performed on the resulting four sets of RNA samples. Comparisons between all four sets of expression data were used to identify 357 genes (407 probe sets) whose expression depended on both loss of Ets2 and presence of tumor (see Supplementary Table 1 for details). Approximately 25% of these genes were negatively regulated in the tumor microenvironment and the expression of these genes increased when Ets2 was deleted in TAMs. Gene ontology indicated that genes encoding extracellular components were principally affected by Ets2 deletion (Figure 2A). The major biological process represented was angiogenesis, with 34% of the genes annotated as having a role in this process (Figure 2A). Many of the genes in the angiogenesis class were classified as inhibitors of angiogenesis.
Quantitative RT-PCR using RNA from independently isolated mammary TAMs representing early (first palpable tumor) and late (6 weeks after tumor initiation) carcinoma stages were used to verify the microarray results (Figure 2B). Of 31 genes tested, 25 were confirmed to be differentially expressed in TAMs with or without Ets2 (Supplementary Table 2). Data for fourteen of the genes classified as encoding inhibitors of angiogenesis are shown (Figure 2B and Supplementary Figure 3D). Expression of these genes in both early and late tumors was increased when Ets2 was deleted. In contrast, potential ETS2 targets known to be involved in inflammation like Mmp9 and Tnfα (12), and other genes associated with inflammation like Il6 were not significantly affected by Ets2 deletion in TAMs, emphasizing that the analysis identified tumor-specific targets of ETS2 (Supplementary Figure 3D).
The same 31 genes were also studied in lung TAMs isolated following tail-vein injection of Met-1 cells (bottom panel in Figure 2B and Supplementary Table 2). In these TAMs, 25/31 genes were differentially expressed when Ets2 was deleted, including the angiogenic gene set, indicating the Ets2 targets were similar in mammary or lung TAMs.
Examination of 1kb of the proximal promoter regions of four candidate genes not previously reported as ETS2 targets (Thbs1, Thbs2, Timp1, and Tim) revealed conserved ETS binding motifs in their proximal promoter regions (Supplementary Figure 4A). Based on these conserved sequences, chromatin immunoprecipitation (ChIP) experiments were performed on lung TAMs from mice with or without Ets2. For the experiments, approximately 50,000 YFP-tagged, F4/80 positive cells were isolated from lungs containing metastases following tail vein injection of Met-1 cells. Antibodies against ETS2 and its co-repressor HDAC1 (14) were used in the ChIP assays (Figure 3).
The ChIP experiments revealed that in wild-type cells ETS2 and HDAC1 were both enriched at all four of these promoter sequences (Figure 3). In contrast, when Ets2 was conditionally deleted both the levels of ETS2 and HDAC1 were significantly reduced at each of the four promoters. Similar results were obtained for the Thbs1 promoter in TAMs isolated from the primary mammary tumor (Supplementary Figure 4B).
To verify the expression of ETS2 targets in situ, we performed immunohistochemical staining on paraffin-imbedded samples prepared from metastatic lung tumors using commercially available antibodies. This analysis demonstrated robust expression of THBS2, THBS1 and SPARC within tumors from mice with Ets2 deletion in TAMs compared to Ets2+ controls (Figure 4A and Supplementary Figures 5A-B).
To confirm that the tumor macrophages were expressing these proteins, frozen mammary tumor sections were analyzed by double immunofluorescent staining using F4/80 to identify TAMs. The MVT-1 orthotopic mammary fat pad injection model was used for this analysis. Staining with α-F4/80-and α-THBS2 showed extensive overlap between the two proteins in sections obtained from tumors with Ets2 deletion (Figure 4B, top panels, Supplementary Movie 1). Since THBS2 is an extracellular protein, expression was found both intracellularly and in the adjacent extracellular space in approximately 75% of F4/80 positive cells, as clearly evident in confocal reconstructions of 15μM sections (see Supplementary Movie 1). In contrast, co-expression of THBS2 in F4/80 positive cells was 10-fold lower in tumors with Ets2 (Figure 4B, bottom panels). Importantly, expression of THBS2 in F4/80 negative cells was not affected by deletion of Ets2 in TAMs (Figure 4B, lower bar graph). Identical results were obtained for THBS1 and SPARC (Supplementary Figure 6A-B respectively and Supplementary Movies 2-3 respectively).
Since many of the tumor-specific Ets2 targets detected, including THBS1, THBS2 and SPARC, have been implicated in angiogenesis, blood vessel density was analyzed in experimental and control mice using α-CD31 immunostaining of paraffin imbedded tumor sections. For these experiments, both primary MVT-1 tumors and lung tumors formed by tail-vein injection of Met-1 cells were studied (Figure 5A). A significant 2-3 fold reduction in tumor vasculature was observed in both primary mammary tumors and lung metastases (Figure 5A).
BrDU incorporation was used to measure cell proliferation in lung metastases in the Met-1 tail vein injection model (Figure 5B). The analysis demonstrated a significant 2.5-fold decrease in BrDU-labeled tumor cells in mice with Ets2-deficient TAMs compared to controls. Tumor cell apoptosis, measured by staining with activated caspase-3 antibody, was not significantly affected by Ets2 deletion (Supplementary Figure 2D).
In order to determine if the mouse genetic studies were relevant to human disease, the mouse expression data was compared to the Rosetta human breast cancer data set (31). Initially, 407 mouse probe sets that were differentially expressed in mouse TAMs with or without Ets2 were compared to the Rosetta array platform and 341 homologous human probe sets were identified (see Supplementary Table 3 for details). These 341 probe sets were compared to 2856 probe sets that represented genes differentially expressed in 117 human samples annotated as with or without lymphocyte/leukocyte infiltration (31). This comparison showed that 142 of the mouse Ets2-TAM probe sets, representing 133 genes, were significantly differentially expressed in lymphocyte/leukocyte infiltration-positive versus -negative human breast cancers (p<0.05, see Supplementary Table 3 and Supplementary Figure 7A). Gene ontology analysis of these human genes showed that extracellular matrix components and angiogenesis were predominantly affected, just as for the mouse Ets2-TAMs genes (Supplementary Figure 7B). A subset of 70 genes differentially expressed with high significance (p<0.001) is represented in the heat map presented in Figure 6A. Interestingly, Ets2 expression itself was on average 8-fold higher inlymphocyte/leukocyte infiltration-positive patients when compared to the negative group (Figure 6A, bar graph, p = 0.0002).
To determine if the TAM gene signature correlated with clinical outcome of patients, the 133 human Ets2-TAM gene signature was used for unsupervised clustering of expression data obtained from 159 sporadic breast cancer patients in the Stockholm data set ((18), see Supplementary Table 3). Expression of the Ets2-TAM signature predicted overall survival in this group with high confidence (Figure 6B, p = 0.0007, Hazard ratio of 3.1). Similar results were obtained with the entire Rosetta 295 patient sample set (Figure 6C, p=.0.0003, Hazard ratio of 2.31).
The influence of the microenvironment, particularly macrophages, on tumor growth and metastasis have long been recognized, but relatively little is known of the gene pathways and mechanisms macrophages use to promote tumor malignancy (34). The results presented here demonstrate that in mouse models Ets2 in tumor macrophages promotes angiogenesis and growth of both primary tumors and lung metastases. The mechanism of action of ETS2 in TAMS involved direct repression of genes encoding predominantly extracellular products, including well-characterized inhibitors of angiogenesis. Recently, an independent report of global gene profiling in TAMs also observed expression of several anti-angiogenic genes along with well-known positive regulators like Vegf-a, results consistent with our data (35). However, the anti-angiogenic effect of TAMs lacking Ets2 is dominant even in the context of MVT-1 tumor cells that overexpress VEGF-A. Additionally, the presumed role of VEGF-A produced by TAMs in triggering the angiogenic switch have been challenged by recent finding demonstrating that deletion of VEGF-A in TAMs actually results in increased tumor growth (36, 37). Thus, Ets2 has previously unappreciated role in TAMs in controlling the balance between positive and negative regulators of angiogenesis necessary for tumor metastasis.
Ets2 in TAMs increased the growth of primary and metastatic tumors. Ets2 could indirectly effect tumor growth by modulating angiogenesis, or directly through paracrine mechanisms. The Ets2 targets identified would favor the former possibility, as obvious paracrine candidates like Il6 or Egf were not differentially expressed. In either case, the results are consistent with the Ets2 pathway playing a role in some activities associated with the alternatively-activated M2 macrophage population (2, 3). M2 macrophages are believed to modulate inflammatory response and to promote tissue remodeling and angiogenesis; in the context of tumor progression, M2-like cells are believed to promote immune suppression, as well as tumor angiogenesis, invasion and metastasis (2, 3). Extracellular function and angiogenesis are the major Ets2 targets identified in our studies, providing a molecular mechanism by which M2-like tumor macrophages modulate the extracellular microenvironment to promote tumor growth and angiogensis at both primary and tumor sites.
A key finding is that a portion of the mouse Ets2-TAM gene expression signature was present in human breast cancer expression data and that it could retrospectively predict overall survival in two independent cohorts of sporadic breast cancer patients. This 133 gene signature is independent of other breast tumor signatures capable of predicting patient outcome, including stromal gene signatures (38, 39). While further efforts will be required to fully implement these findings and determine their significance to human disease, the results validate the relevance of our hypothesis-driven mouse modeling approach for dissecting TAM functions in tumor growth and metastasis.
Dispersed tumor cells are present in many breast cancer patients and may be the mediators of tumor recurrence (40). Breast tumor micrometastases are genetically distinct from the primary tumor indicating that they are disseminated early in tumor progression (41, 42). Results obtained in the PyMT and Her2/Neu mouse models demonstrate an early spread of mammary epithelial cells before the carcinoma stage, providing experimental verification of the human data (42). Thus, understanding how dispersed dormant cells progress to growing metastases is a problem with considerable clinical relevance. Further studies on Ets2 and its downstream targets could provide unique insights in understanding how the microenvironment modulates the growth of tumor cells at metastatic sites.
We thank Alexander Borowsky and Michael Johnson for the Met-1 and MVT-1 cell lines, respectively, Robert Oshima for the Ets2db mice, Karl Kornacker for microarray data analysis, Kartic Krishnamurthy for image analysis software, and Lisa Rawahneh for histology support. We acknowledge the Cancer Center Microscopy/Imaging, Microarray, Genomic, Transgenic/Knockout, Histology, and Flow Cytometry Shared Resources. TZ was supported by a DOD Pre-doctoral Fellowship. This work was supported by NCI Grants P01 CA097189 (MCO, GL, TJR) and R01 CA053271 (MCO) and the Evelyn Simmer’s Charitable Trust (MCO).