Breast cancer is the second leading cause of cancer-related deaths in western countries. Colony-Stimulating Factor-1 (CSF-1) and its receptor (CSF-1R) regulate macrophage and osteoclast production, trophoblast implantation and mammary gland development. The expression of CSF-1R and/or CSF-1 strongly correlates with poor prognosis in several human epithelial tumors, including breast carcinomas. We demonstrate that CSF-1 and CSF-1R are expressed, although at different levels, in 16/17 breast cancer cell lines tested with no differences among molecular subtypes. The role of CSF-1/CSF-1R in the proliferation of breast cancer cells was then studied in MDAMB468 and SKBR3 cells belonging to different subtypes. CSF-1 administration induced ERK1/2 phosphorylation and enhanced cell proliferation in both cell lines. Furthermore, the inhibition of CSF-1/CSF-1R signaling, by CSF-1R siRNA or imatinib treatment, impaired CSF-1 induced ERK1/2 activation and cell proliferation. We also demonstrate that c-Jun, cyclin D1 and c-Myc, known for their involvement in cell proliferation, are downstream CSF-1R in breast cancer cells. The presence of a proliferative CSF-1/CSF-1R autocrine loop involving ERK1/2 was also found. The wide expression of the CSF-1/CSF-1R pair across breast cancer cell subtypes supports CSF-1/CSF-1R targeting in breast cancer therapy.
Citation: Morandi A, Barbetti V, Riverso M, Dello Sbarba P, Rovida E (2011) The Colony-Stimulating Factor-1 (CSF-1) Receptor Sustains ERK1/2 Activation and Proliferation in Breast Cancer Cell Lines. PLoS ONE 6(11): e27450. https://doi.org/10.1371/journal.pone.0027450
Editor: Ilya Ulasov, University of Chicago, United States of America
Received: January 3, 2011; Accepted: October 17, 2011; Published: November 9, 2011
Copyright: © 2011 Morandi et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by Istituto Toscano Tumori (PDS), Associazione Italiana per la Ricerca sul Cancro (AIRC, contr.n. IG5220; PDS), Ministero della Salute (Ricerca Finalizzata; PDS), Regione Toscana (Programma per la Ricerca in Materia di Salute; PDS), Associazione Italiana per la lotta contro le Leucemie e i Linfomi (AIL) sezione di Prato (PDS), Fondazione Cassa di Risparmio di Volterra (PDS). VB was recipient of a fellowship from Federazione Italiana per la Ricerca sul Cancro (FIRC/AIRC). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The c-fms proto-oncogene encodes the only known receptor (CSF-1R) for Colony Stimulating Factor 1 (CSF-1 or M-CSF) , . CSF-1R is a class III transmembrane tyrosine kinase receptor and its ligand CSF-1 has secreted glycoprotein, secreted proteoglycan and membrane-bound isoforms , . The CSF-1/CSF-1R pair has essential physiological functions in the generation of osteoclasts and macrophages  and, via its action on macrophages and other CSF-1R-expressing cells, in female and male fertility , . Activation of CSF-1R by its ligand triggers a series of rapid events, including receptor dimerization and tyrosine phosphorylation of its intracellular domain. Phosphorylation at particular CSF-1R tyrosines creates binding sites for a variety of cytoplasmic proteins that activate signal transduction pathways including that of ERK1/2 and PI3K .
CSF-1 and CSF-1R are expressed in normal breast tissue during puberty, pregnancy and lactation. However, the expression of CSF-1R and/or CSF-1 has been documented in several human cancers, including carcinomas of breast, female reproductive tract, prostate and kidney –. Data reported in literature for solid tumors indicate that the oncogenic potential of CSF-1/CSF-1R is due to the co-expression of this growth factor/receptor pair, rather than CSF-1R overexpression or mutations activating CSF-1R independently of ligand . This is supported by the fact that the expression of normal c-fms into CSF-1-expressing non-transformed fibroblasts and epithelial cells can be sufficient to induce a fully transformed phenotype , . In this respect, activation of CSF-1R by its ligand is likely to occur in tumor cells in which CSF-1R and CSF-1 are co-expressed (i.e. autocrine activation), or when CSF-1R is stimulated by CSF-1 released by cancer associated fibroblasts (i.e. paracrine activation). Consistent with this, in breast cancer patients, the expression of both CSF-1 and its receptor in neoplastic epithelial cells strongly correlates with poor prognosis and is predictive of ipsilateral recurrence –. In addition, the presence of tumor associated macrophages in breast tumors also correlates with poor prognosis ,  and, in mouse models, CSF-1 promotes metastasis , stimulates angiogenesis ,  and is involved in a paracrine loop with EGF to promote tumor cell invasion . While previous studies indicated that CSF-1R and CSF-1 are expressed in breast cancer cell lines and tumors and demonstrated the relevance of CSF-1/CSF-1R signaling in the invasiveness of breast cancer cells –, few studies have focused on the biological role of CSF-1/CSF-1R signaling in the proliferation of breast cancer cells.
Targeting receptor tyrosine kinases with kinase inhibitors (e.g. imatinib, dasatinib or nilotinib) has recently opened a new era in the treatment of hematologic malignancies and solid tumors such as gastrointestinal stromal tumors , . These drugs are effective on CSF-1R ,  and other CSF-1R-specific inhibitors have been developed –. More importantly, several drugs targeting CSF-1 and CSF-1R are currently in Phase I/II trial (www.clinicaltrials.org). Elucidation of the involvement of CSF-1R in breast cancer cell proliferation would strengthen the rationale of CSF-1R targeting in CSF-1R expressing cancers.
In this work, we characterized the role of CSF-1R in the proliferation of breast cancer cells and found that CSF-1R is widely expressed in breast cancer cell lines at both mRNA and protein levels. Interfering with the CSF-1/CSF-1R signaling pathway, either by CSF-1R inhibition or by inhibition of autocrine CSF-1, impaired MDAMB468 and SKBR3 cell proliferation. In addition, exposure to ectopic CSF-1 stimulated MDAMB468 and SKBR3 growth. We found ERK1/2, c-Jun, cyclin D1 and c-Myc, known for their involvement in cell proliferation, to be downstream CSF-1R in breast cancer cells. The wide expression of CSF-1/CSF-1R pair across breast cancer cell subtypes supports CSF-1/CSF-1R targeting in breast cancer therapy.
Materials and Methods
Cells and cell culture
NIH/3T3 murine fibroblasts expressing ectopic human CSF-1R (kind gift of MF Roussel, St. Jude Children's Research Hospital, Memphis, TN, USA)  and HepG2 human hepatoblastoma cells (www.lgcstandards-atcc.org) periodically tested in our laboratory by western blotting for the presence of EGFR protein) were cultured in DMEM, while human chronic myeloid leukemia K562 cells (www.lgcstandards-atcc.org); periodically tested in our laboratory by western blotting or Q-PCR for the expression of BCR/Abl) in RPMI, supplemented with 4 mM glutamine and 10% fetal bovine serum (FBS). Human primary macrophages were obtained after informed consent as previously described . The breast cancer cell lines (MCF7, T47D, MDAMB175VII, ZR751, 734B, MDAMB361, BT474, SKBR3, MDAMB453, HCC1954, MDAMB468, BT20, SUM149PT, HCC1500, MDAMB231) and MCF10A and MCF12A cells (two immortal, non transformed cell lines showing basal B molecular pattern) ,  were a kind gift of Dr MG Daidone and Dr E Tagliabue, Istituto Nazionale Tumori, Milano, Italy; Dr D Lerouge, Institut Gustave-Roussy, Villejuif, France; Prof CM Isacke, Institute of Cancer Research, London, UK. Laboratories of origin have tested all cell lines by microsatellite analysis or microarray. However, cells have been tested upon arrival and periodically in our lab by western blotting for HER2, EGFR and by PCR for estrogen receptor expression. Cells were cultured as previously described . Cells were incubated in the presence or the absence of 100 mg/ml streptomycin and 100 I.U. penicillin, at 37°C in humidified atmosphere containing 5% CO2.
Analysis of Gene Expression and cGH Datasets
Data were obtained mapping CSF-1 and CSF-1R gene symbols to GSE2603 (probes 2078082_at and 203104_at) , NKI (probes NM_000757 and NM_000971)  and Neve (probes 2078082_at and 203104_at)  datasets. Cell lines molecular subtypes were reported as classified by the authors in the Neve dataset . Breast cancer subtypes of tumor samples were predicted using centroid Spearman correlation to the PAM50 classifier in NKI and GSE2603 datasets. The PAM50 gene expression predictor classifies breast cancers into molecular intrinsic subtypes (Luminal A, Luminal B, HER2-enriched, Basal-like) and provides a risk of recurrence (ROR) score based on the similarity of an individual sample to prototypic subtypes . GSE2603 data were GCRMA normalized, NKI data were mean centered normalized and the Neve data were RMA normalized as reported by the authors in their studies using the ROCK database . The CGH array data for CSF-1R and CSF-1 were obtained from Neve  and Fridlyand  (RP1-141L3 and CTD-2050A15). Statistical analysis was performed using Prism software (GraphPad Software, La Jolla, CA, USA). Means of the normalized data of each subtype were compared using one-way ANOVA and the Bonferroni test. Differences were considered statistically significant when p<0.05. We did not submit our research to the local ethical committee nor did we obtain informed consent for the use of tumor explants because our analyses were performed on publicly available datasets (see below).
Total cell lysates
Culture plates were placed on ice, cell monolayers rapidly washed 3 times with ice-cold PBS containing 100 mM orthovanadate and cells lysed by scraping in Laemmli buffer (Tris/HCl 62.5 mM, pH 6.8, 10% glycerol, 0.005% blue bromophenol, 2% SDS) and incubating at 95°C for 10 minutes in the presence of 100 mM 2-mercaptoethanol. Lysates were then clarified by centrifugation (20000 g, 10 minutes, RT).
Western Blotting and Immunoblotting
30–60 µg of total proteins was separated by SDS-PAGE in 9–15% polyacrylamide gel and transferred onto PVDF membranes (Millipore) by electroblotting. Membranes were incubated (1 hour, RT) in Odyssey Blocking Buffer diluted 1∶1 with PBS, and then in the same buffer containing 0.1% Tween-20 and the primary antibody (16–18 hours, 4°C). After extensive washing with PBS/0.1% Tween-20, membranes were incubated in Odyssey Blocking Buffer diluted 1∶1 with PBS containing IRDye®800CW- or IRDye®680-conjugated secondary antibody (1 hour, 4°C). Bands were visualized by infrared imaging (Licor, Odissey) and images recorded as TIFF files for quantification with Adobe Photoshop software. Rabbit α-phospho-T202/Y204-ERK1/2 (Cell Signaling, # 9101); rabbit α-ERK1 (Santa Cruz, sc-93); rabbit polyclonal α-CSF-1R C-20 (Santa Cruz; sc-692); goat α-CSF-1 antibody N-16 (Santa Cruz, sc-1324); α-phospho-723-CSF-1R (Cell Signaling, # 3151); mouse α-vinculin (sigma; V9131); rabbit α-phospho-S63/73-c-Jun (Santa Cruz; sc16312); mouse monoclonal α-cyclin D1 (Santa Cruz, sc-8396); mouse monoclonal α-myc (Santa Cruz, sc-8396).
Cells were detached by incubation in PBS containing 0.2% EDTA (pH 7.2), washed with PBS, pelleted and incubated in 20 µl of an anti-CSF-1R antibody (24A4, PE-conjugated; sc-02PE; Santa Cruz Biotechnology)  or isotype control antibody (IgG2b, sc-2873, Santa Cruz Biotechnology) for 45 minutes in the dark. After two washes with PBS followed by a 5 minutes centrifugation, cells were resuspended in 500 µl PBS and analyzed with a FACSCanto (Becton Dickinson). The percentage of positive cells was calculated by subtracting values obtained with isotype control antibody from those obtained with anti-CSF-1R antibody.
Enzyme-Linked ImmunoSorbent Assay
CSF-1 was measured by ELISA (RayBio ELISA KIT Human MCSF), according to the manufacturer's instructions. The manufacturer claims that the minimum detectable dose of CSF-1 is typically less than 5 pg/ml (sensitivity of the methods). Supernatants were collected at cell confluency from cultures in complete medium. Each sample (100 µl) was assessed in duplicate.
CSF-1R silencing with siRNA and measurement of DNA synthesis by [3H]thymidine uptake
Silencing was performed as previously described  with 100 nM SMART-pool siRNA for CSF-1R (NM_005211 mRNA, Dharmacon, cat. No M-003109-03), 100 nM SMART-pool siRNA for CSF-1 (NM_000757, Dharmacon, cat. No M-017514-00) or 100 nM siCONTROL non-targeting pool (Dharmacon, cat. n. D-001206-13) following the manufacturer's instructions. Transfection efficiency was 90%, as assessed by cotransfection with Cy3-labelled siGLO RISC-free siRNA (Dharmacon, cat. n. D-001600-01). One day after transfection, cells were serum starved for 24 hours before treatment with CSF-1 for 24 h. [3H]thymidine uptake analysis was performed as previously described .
Measurement of DNA synthesis by bromodeoxyuridine uptake and immunofluorescence
Cells were seeded onto glass coverslips in complete medium for 24 hours and then incubated in the absence of FBS for 24 hours before being incubated in DMEM with or without 25 ng/ml CSF-1 or a 1∶50 dilution of a goat α-CSF-1 blocking or pre-immune serum  for further 24 hours. During the last 4 hours of incubation, bromodeoxyuridine (BrdU) was added to the culture (final concentration 10 nM) and then BrdU uptake was stopped by incubating cells in 4% formaldehyde/PBS for 10 minutes at RT. After washing in PBS, cells were incubated for 20 minutes with 2M HCl and then 0.1 M Na2B4O7 was added. Cells were washed with PBS and permeabilized by a 5 minute incubation in PBS containing 0.2% Triton X-100. After three washes in PBS cells were incubated with 10% horse serum in PBS/1%BSA for 45 minutes and then washed in PBS and incubated overnight at 4°C in a 1∶250 dilution of a mouse monoclonal α-BrdU antibody (Millipore, MAB3222) in PBS/1%BSA, washed with PBS and incubated with a 1∶800 dilution of an α-mouse Cy3-labelled secondary antibodies (Chemicon, AP192C). Cells were washed in PBS and incubated with 5 µg/ml Hoechst 33258 (Sigma) nuclear dye in PBS for 10 minutes. Following two washes in PBS, coverslips were mounted with propyl-thiogallate on glass slides and cells observed with a Leica DC200 microscope. Pictures were taken from 6 different field/sample (>600 cells were scored for each treatment) and the percentage of cells undergoing DNA synthesis was calculated by the ratio of the number of BrdU-positive cells to the total number of cells determined by Hoechst 33258 staining. Incubation with secondary antibody alone did not produce any significant fluorescence.
Measurement of cell number by Crystal Violet staining
Cells were seeded in 12 multi-well plates and incubated for 24 hours in complete medium before serum starvation for further 24 hours. Cells were then treated in serum-free medium with imatinib (Gleevec, Glivec, Novartis, Basel, Switzerland) for 45 minutes and then with CSF-1 for further 48 hours. Cell were then washed twice with PBS and incubated for 10 minutes at RT with a Crystal Violet solution (0.5% Crystal violet, SIGMA, 30% ethanol and 3% formaldehyde). After extensive water washing, plates were allowed to dry and dye extracted by incubating with 1% SDS. Densitometric measurement was then performed at 550 nm.
Quantitative Real-Time PCR (Q-PCR)
After total RNA extraction by TRIzol (Invitrogen) as specified by the manufacturer, 1 µg of total RNA/sample was submitted to reverse transcription with SuperScriptVILO-Reverse Transcriptase (Invitrogen) for 10 minutes at 25°C, 1 hour at 42°C and 5 minutes at 85°C utilizing 50 pmol random hexameric primers. The primers used were as follows: CSF-1R, C-term, for: 5′-CCTCGCTTCCAAGAATTGCA-3′, rev: 5′-CCCAATCTTGGCCACATGA-3′ (amplicon size 60 bp, designed to span the intron between exons 16 and 17); CSF-1R, N-term, for: 5′-GGAGGCTGCCCAGATCGT-3′ rev: 5′-GCGAGCTTGGTGTTGTTGTG-3′ (amplicon size 60 bp, designed to span the intron between exons 4 and 5); cyclin D1 (CCND1), for: 5′-GAGAGGAAGCGTGTGAGGCGGTAG-3′, rev: 5′-GGATGCTGGAGGTGCGAGGA-3′; cyclin G1 (CCNG1), for: 5′-CCAGCTGAATGCCCTGTTG-3′, rev: 5′-AGTCTCAAACCACAGACCTTTGG-3′; cyclin H (CCNH), for: 5′-GAGGAGCAGCTGGCAAGACT-3′, rev: 5′-ACGGCTTTGCATCTGAATTTG-3′; cyclin I (CCNI), for: 5′-CATCTCAACATTTGGCAGTCCTT-3′, rev: 5′-GAAGTTGGTTGCAGGCCATAC-3′; rRNA 18S, for: 5′-CGGCTACCACATCCAAGGAA-3′, rev: 5′-GCTGGAATTACCGCGGCT-3′ (amplicon size 180 bp); GAPDH, for: 5′-AACAGCC TCAAGATCATCAGCAA-3′, rev: 5′-CAGTCTGGGTGGCAGTGAT-3′. CSF-1R mRNA expression was assessed by Q-PCR (2 minutes 50°C, 5 minutes 95°C, 40 cycles at 95° C for 15 seconds and 60°C for 1 minute) with the ABI Prism 7500 Sequence Detection System (Applied Biosystem) using Power SYBR® Green PCR master mix (Applied Biosystem). A melting curve analysis was performed to discriminate between specific and non-specific PCR products. Alternatively, amplified product size was verified by running in 3–4% agarose gels. The housekeeping 18S rRNA and/or GAPDH genes were used as internal references for normalization. The relative expression of CSF-1R, with respect to SKBR3 cells chosen as calibrator, was calculated by using a comparative threshold cycle method and the formula 2(−ΔΔCt) .
Analysis of StellARray Gene Expression
SKBR3 cells were transfected with the indicated siRNA and cultured in complete medium for 72 hours. Cells were then lysed and total RNA extracted. Gene expression analysis was performed with Q-PCR Arrays (Human Cell Cycle Tox and Cancer 96 StellARray™ qPCR array 00188263 Fast 96 Well, Lonza). A list of these 96 genes is available online at http://array.lonza.com/plate/00188263/. The housekeeping rRNA 18S gene was used as internal reference for normalization. Fold change was calculated as described above. Data reported (+SEM) were obtained from four independent experiments.
The expression of CSF-1R and/or CSF-1 in human breast carcinomas has been documented in both cell lines and tumors samples –, , . However, whether their expression is restricted to one molecular subtype has not been studied. To address this issue, we performed Q-PCR for CSF-1R mRNA on 17 cell lines endowed with different molecular profiles and referred to as luminal, basal A and basal B subtypes (Figure 1A) , , , . These experiments indicated that CSF1R is expressed, although at different levels, in all cell lines tested.
(A) Routinely cultured cells were lysed and total RNA extracted. Q-PCR was performed using two different sets of primers (C-terminus, light gray; N-terminus, dark grey). The 18S rRNA was used for normalization and SKBR3 were chosen as calibrator. Data represent the mean (± SD) of three independent experiments. Asterisks indicate the HER2-overexpressing cell lines. (B) Routinely cultured cells were processed and subjected to flow cytometry with a rat monoclonal anti-CSF-1R antibody. Columns represent the percentage of CSF-1R-positive cells within the bulk population analyzed (gated in order to exclude debris and cellular aggregates). Data represent the mean (± SEM) of 3 independent experiments. Asterisks indicate the HER2-overexpressing cell lines. One-way ANOVA among different subtypes showed no differences.
The presence of cell surface CSF-1R protein was then verified by flow cytometry (Figure 1B, S1A). All the cell lines tested, except HCC1500, MCF10A and MCF12A, expressed appreciable cell surface CSF-1R. However, MCF10A and MCF12A cells expressed high levels of CSF-1R mRNA. We hypothesized that these differences are due to ligand-induced down-regulation of the receptor (see below). Furthermore, ectopic CSF-1 activates ERK1/2 in these cells pointing to the presence of functional CSF-1R. In contrast, HCC1500 cells express very low CSF-1R mRNA levels and did not respond to CSF-1 as monitored by activation of ERK1/2 (Figure S3), PI3K or ERK5 (not shown). HCC1500 is, therefore, the only breast cancer cell line of the panel analyzed that we consider CSF-1R negative. The analysis of mean fluorescence intensity produced similar results (not shown). No significant differences in cell surface CSF-1R expression among subtypes were found by one-way ANOVA (not shown).
The cell lines were also tested for CSF-1 expression by measurement of secreted CSF-1 (by ELISA and western blotting) and of CSF-1 in cell lysates (by western blotting). CSF-1 was detected in all the cell lines tested (Table 1). It is of note that we were unable to detect CSF-1 in SKBR3 supernatants by ELISA, a finding that is at variance with previous reports . The detection of CSF-1 in total cell lysates of SKBR3 indicates that these cells, which are known to express CSF-1 , may express membrane-bound CSF-1.
The data reported in Figure 1 and Table 1 indicated that all the cell lines tested express CSF-1 and that 16/17 express CSF-1R. In addition, we performed in silico analysis on publicly available gene expression profiling datasets (Figure 2) , , . This analysis indicated that the mean expression of CSF1 and CSF1R genes did not vary significantly (assessed by One-way ANOVA analysis) among breast cancer subtypes in either cell lines or in tumors samples. Moreover, in silico analysis of comparative genomic hybridization datasets relative to 49 breast cancer cell lines  and 67 primary tumors  indicated the absence of CSF-1R or CSF-1 gene amplifications (Figure S2).
CSF1 and CSF1R transcripts levels are shown for breast cancer cell lines dataset (blue) from Neve et al.  and for two independent breast tumor datasets (red) from Minn et al.  and van de Vijver et al. . One-way ANOVA and the Bonferroni test did not show any statistical significant difference among the subtypes within the dataset.
To determine whether CSF-1R transduces proliferative signals in breast cancer cells we chose two cell lines from different subtypes (Figure 3). When basal MDAMB468 cells were exposed to exogenous CSF-1, their proliferation increased by 40% and ERK1/2 and c-Jun phosphorylation were markedly increased. In addition CSF-1 administration increased cell proliferation and ERK1/2 and c-Jun phosphorylation of luminal SKBR3 cells. It is of note that this cell line is characterized by a high basal ERK1/2 phosphorylation suggesting an autocrine CSF-1/CSF-1R loop and/or activation of CSF-1R-independent signaling pathways. The relevance of ERK1/2 in CSF-1R signaling was highlighted by the fact that enhancement of ERK1/2 phosphorylation upon CSF-1 treatment was found in 10 out of 17 breast cancer cell lines (Figure S2).
MDAMB468 (A) or SKBR3 (B) cells were cultured in DMEM without serum for 24 hours and then for additional 24 hours with DMEM with or without 25 ng/ml CSF-1. Cells were scored for (A) BrdU or (B) tritiated thymidine uptake or lysed and protein subjected to immunoblotting with the indicated antibodies. (A) Values reported under the pictures are percentages (± SEM) of BrdU-positive nuclei, normalized to total nuclei labeled by Hoechst 33258, from 5 independent experiments; **, Student's t test: p<0.01. (B) Data represent mean (± SEM) of one out of 3 representative experiments; *, Student's t test: p<0.05. Densitometric values of bands (normalized for loading control) are reported as ratios between the CSF-1-treated and untreated value, set as 1.
To further characterize CSF-1R involvement in breast cancer cell proliferation we treated breast cancer cells (Figure 4) with CSF-1R siRNA (Figure S1A and B). CSF-1R silencing in SKBR3 cells markedly impaired cell proliferation in the presence of ectopic CSF-1 (Figure 4A), reduced the phosphorylation of ERK1/2 and c-Jun as well as the expression of c-Myc, cyclin D1 (Figure 4B) and several other cyclins (Figure 4C and Table 2).
(A–C, F) SKBR3 or (F) MDAMB468 cells were transfected with the indicated siRNA and incubated for 24 hours. (A) Cells were then serum-starved for a further 24 hours and treated with 25 ng/ml CSF-1 for 24 hours, and tritiated thymidine uptake measured. Data represent mean (± SEM) of one out of 3 representative experiments; **, Student's t test: p<0.01. (B, C, F) 72 hours post transfection cells were lysed and total protein or RNA extracted. (B, F) Protein lysates were subjected to immunoblotting with the indicated antibodies. Densitometric values of bands (normalized for loading control) are reported as ratios between the siCSF-1R and the siNT value, set as 1. (C) Q-PCR was performed for the indicated genes. Data were normalized against GAPDH. siNT-treated samples (dark green) were chosen as calibrator over the siCSF1R-treated samples (light green). Data represent the mean (± SEM) from three independent experiments performed in triplicates. (D, E) SKBR3 cells were cultured in DMEM without serum for 24 hours and then for 24 hours with 25 ng/ml CSF-1, pre-incubated for 1 hour with a 1∶50 dilution of CSF-1-blocking anti-serum (I) or control pre-immune serum (PI). Cells were then scored for BrdU uptake or lysed. (D) Values reported under the pictures, from 5 independent experiments, are percentages (± SEM) of BrdU-positive nuclei, normalized to total nuclei labeled by Hoechst 33258; ***, Student's t test: p = 0.0004. (E) Protein lysates were subjected to immunoblotting with the indicated antibodies. Densitometric values of bands (normalized for loading control) are reported as ratios between the I and the PI value, set as 1.
To test whether an autocrine CSF-1/CSF-1R loop exists in breast cancer cells (Figure 4D) we used a CSF-1-blocking antiserum (Figure S1C) . SKBR3 cells treated for 24 hours with the CSF-1-blocking antiserum showed a 35% reduction in proliferation when compared to the cells treated with the control serum (pre-immune) (Figure 4D). In the same experiments, the CSF-1-blocking antiserum markedly inhibited c-Jun phosphorylation and reduced cyclin D1 and c-Myc expression (Figure 4E). This autocrine signaling sustained ERK1/2 activation in both SKBR3 and MDAMB468 cells (Figure 4F) as demonstrated by the fact that CSF-1 silencing partially impaired basal ERK1/2 phosphorylation. CSF-1 silencing was confirmed by Q-PCR (not shown). The results shown in Figure 3 and 4 indicated the involvement of ERK1/2 and its downstream targets (i.e. c-Jun, cyclin D1 and c-Myc) in CSF-1-induced cell proliferation in breast cancer cells.
The fact that CSF-1 promotes proliferation in breast cancer cells prompted us to determine whether their growth is sensitive to tyrosine kinase inhibitors used in the clinic. Imatinib (IM) is known to inhibit CSF-1R activity . We found, in keeping with previous data , that 10 µM IM was necessary to prevent CSF-1-induced CSF-1R phosphorylation in NIH/3T3-Fms cells (Figure S1D). When 10 µM IM was given to SKBR3 and MDAMB468 breast cancer cell lines for 48 hours in the presence of CSF-1, cell number was markedly reduced, as determined by crystal violet staining (Figure 5A). Further, IM was able to prevent in part the activation of ERK1/2 induced by CSF-1 confirming the involvement of ERK1/2 in the proliferative effect of CSF-1 in breast cancer cells (Figure 5B). IM treatment alone did not decrease ERK1/2 phosphorylation (not shown).
24 hours serum deprived SKBR3 or MDAMB468 cells were treated for 45 minutes with (light gray columns, right pictures) or without (dark gray columns, left pictures) 10 µM imatinib (IM) and then with 25 ng/ml CSF-1 for (A) 48 hours or (B) 10 minutes. (A) Cells were scored for cell viability by crystal violet staining. Data represent percentages (± SEM) of crystal violet staining normalized for IM-untreated cells from 4 independent experiments; *, Student's t test: p<0.05. (B) Cells were lysed and protein lysates subjected to immunoblotting with the indicated antibodies. Densitometric values of bands (normalized for loading control) are reported as ratios between treated and untreated value, set as 1.
Although the expression of CSF-1/CSF-1R has been previously documented in breast cancer and shown to correlate with poor prognosis, few studies have been performed to understand the role of CSF-1R-dependent signaling in the proliferation of breast cancer cells or other solid tumors , , . In the present study we found that: i) breast cancer cell lines consistently express CSF-1 and CSF-1R; ii) the CSF-1/CSF-1R pair sustains the proliferation of breast cancer cell lines; iii) ERK1/2 is downstream CSF-1R in proliferating breast cancer cells.
CSF-1R sustains breast cancer cells proliferation, as highlighted in two cell lines of different molecular subtypes. Indeed, interfering with CSF-1/CSF-1R signaling, either by targeting the receptor or by blocking the ligand binding, impacts on breast cancer cell proliferation. This proliferation was induced either by ectopic CSF-1, that mimic CSF-1 produced by fibroblasts and monocytes/macrophages associated with the tumor, or autocrine CSF-1. We found that CSF-1R activation sustained the expression of cyclin D1 and c-Myc and activated c-Jun, which are established CSF-1 downstream targets in other cell types . Interestingly, the expression of several cyclins was decreased following CSF-1R silencing in SKBR3 cells, indicating that in these cells CSF-1R-dependent signaling sustains the progression across different phases of the cell cycle.
We also found that CSF-1R-induced proliferation involves ERK1/2 activation in SKBR3 and MDAMB468 cells. Our data are in line with previous studies in which breast cancer cell lines expressing ectopic CSF-1R showed increased expression of cyclin D1 as a consequence of ERK1/2 activation upon CSF-1 administration . Our results therefore support the existence of a proliferative pathway elicited by CSF-1/CSF-1R that acts through ERK1/2 thereby inducing c-Jun activation as well as c-myc and cyclin D1 expression. We also found that ERK5 ,  is activated in a restricted number of cell lines following CSF-1 treatment (not shown). PI3K  was not activated by CSF-1 in any of the cell lines tested. In this respect, it is to note, however, that both ERK5 and PI3K pathways are often constitutively active in breast cancer cells , , possibly masking a response to CSF-1.
Our data show that 16 cell lines among 17 tested express CSF-1 and CSF-1R, although at different levels. Moreover, gene expression profiling datasets show that CSF1 and CSF1R expression is a general feature of breast cancer cells. These findings support the possibility to target CSF-1R signaling in a large proportion of breast cancers, independently of their molecular subtype. This is particularly relevant for tumors that are classify as of the basal-like subtype. Triple-negative (ER-, PR-, HER2-negative) breast cancers make up the majority of this subgroup ,  and are generally unresponsive to standard treatments, i.e. tamoxifen, aromatase inhibitors and herceptin. Although triple-negative breast cancers can be treated with chemotherapy, early relapse and metastasis is common and therefore the need of potential targets for this tumors is of high priority.
The data obtained from our experiments and in silico analysis indicate that CSF-1R is not overexpressed (when compared to human monocytes) nor amplified  and therefore support previous reports  that the oncogenic potential of CSF-1R is due to its co-expression with CSF-1. Furthermore, this conclusion is further established by the fact that CSF-1R-dependent proliferation of SKBR3 cells is impaired when the CSF-1/CSF-1R interaction is prevented. ERK1/2 is apparently involved in autocrine-induced proliferation as silencing CSF-1 or CSF-1R decreased ERK1/2 activation in both SKBR3 and MDAMB468. Moreover, ERK1/2 was constitutively active in several cell lines where CSF-1 exposure had a minor effect on the basal ERK1/2 phosphorylation. The low level or absent responsiveness to CSF-1 in these cell lines may be the consequence of autocrine CSF-1 production (as demonstrated for SKBR3) and/or of activation of CSF-1R-independent signaling pathways.
IM is a tyrosine kinase inhibitor in use for the clinical management of chronic myeloid leukemia and has been shown to inhibit several different kinases, including CSF-1R. We found that IM impairs the proliferation of breast cancer cell lines in the presence of CSF-1. As IM targets kinases other than CSF-1R our experiments do not prove that the effects shown are only due to CSF-1R inhibition. However, they support the concept that targeting CSF-1R with tyrosine kinase inhibitors can effectively block CSF-1-dependent growth stimulation. This is particularly relevant when considering strategies to interfere with autocrine CSF-1-dependent proliferation.
In conclusion, the evidence we provided for the expression and functional role of the CSF-1/CSF-1R pair in breast cancer, together with the established role of CSF-1/CSF-1R in breast cancer motility and invasiveness –, indicate that CSF-1R targeting may be pursued therapeutically, irrespective of breast cancer subtype, at either early or late stages of tumor progression.
Settlement of CSF-1 or CSF-1R targeting using NIH/3T3 cells expressing ectopic CSF-1R. (A) NIH/3T3-Fms cells were transfected with the indicated siRNA. Total protein lysates obtained at the indicated times were subjected to immunoblotting with the indicated antibodies. Densitometric values of bands (normalized for loading control) are reported as ratios between the siCSF1R and the siNT value, set as 1. 72 hours post-transfection cells were analyzed by flow cytometry. Percentages of CSF-1R-positive cells are reported. (B) NIH/3T3-Fms cells were transfected with the indicated siRNA and incubated for 24 hours. Cells were then serum-starved for further 24 hours and treated with (CSF-1) or without (−) 25 ng/ml CSF-1 for 24 hours, and tritiated thymidine uptake measured. Data represent mean (± SEM) of one of 3 representative experiments; ** and ***, Student's t test: p<0.01, p<0.001, respectively. (C) NIH/3T3-Fms cells were incubated with or without 25 ng/ml CSF-1 for 10 minutes. Before cell treatment, CSF-1 had been incubated for 1 hour at 37°C in the absence (−) or the presence of a 1∶50 dilution of a CSF-1-blocking anti-serum (I) or pre-immune serum (PI). Total protein lysates were subjected to immunoblotting with the indicated antibodies. Densitometric values of bands (normalized for loading control) are reported as ratios between the siCSF-1R and the correspondent siNT value, set as 1. (D) NIH/3T3-Fms cells were cultured for 24 hours without serum and then for 45 minutes with the indicated doses of imatinib (IM) before treatment with or without 25 ng/ml CSF-1 for 10 minutes. Cells were then lysed and protein subjected to immunoblotting with the indicated antibodies.
CSF-1 and CSF-1R gene copy number in breast cancer cell lines and tumor samples. Data have been collected from CGH experiments performed by others with breast cancer cell lines (left)  or tumor samples (right) . Whisker graphs represent median, 25- and 75- percentile, min and max values.
CSF-1R signaling induces ERK1/2 phosphorylation in breast cancer cell lines. Serum-deprived cells (24 hours) were incubated with or without CSF-1 (25 ng/ml) for 10 minutes and lysed in RIPA buffer. Total protein lysates were subjected to immunoblotting with the indicated antibodies. N3F: NIH/3T3-Fms cells. Densitometric values of bands (normalized for loading control) are reported as ratios between the CSF-1-treated and the untreated value, set as 1. Threshold for activation was arbitrary set at ≥1.2.
We thank Dr Betti Giusti (University of Florence, Italy) and Qiong Gao (Breakthrough Breast Cancer Research, UK) for the helpful discussion on the microArray and array CGH data; Dr Richard Stanley (Albert Einstein College of Medicine, NY, USA) and Dr Helen Yarwood (Breakthrough Breast Cancer Research Centre, UK) for their suggestions; and Prof Clare M Isacke and Matthew Ashenden (Breakthrough Breast Cancer Research Centre, UK) for revising of the manuscript.
Funded the research: PDS. Conceived and designed the experiments: ER VB AM. Performed the experiments: ER VB AM MR. Analyzed the data: ER VB AM. Wrote the paper: ER AM PDS.
- 1. Sherr CJ, Rettenmier CW, Sacca R, Roussel MF, Look AT, et al. (1985) The c-fms proto-oncogene product is related to the receptor for the mononuclear phagocyte growth factor, CSF-1. Cell 41: 665–676.
- 2. Dai XM, Ryan GR, Hapel AJ, Dominguez MG, Russell RG, et al. (2002) Targeted disruption of the mouse colony-stimulating factor 1 receptor gene results in osteopetrosis, mononuclear phagocyte deficiency, increased primitive progenitor cell frequencies, and reproductive defects. Blood 99: 111–120.
- 3. Sherr CJ (1990) Colony-stimulating factor-1 receptor. Blood 75: 1–12. Review.
- 4. Pixley FJ, Stanley ER (2004) CSF-1 regulation of the wandering macrophage: complexity in action. Trends Cell Biol 14: 628–638. Review.
- 5. Cohen PE, Nishimura K, Zhu L, Pollard JW (1999) Macrophages: important accessory cells for reproductive function. J Leukoc Biol 66: 765–772. Review.
- 6. Sapi E (2004) The role of CSF-1 in normal physiology of mammary gland and breast cancer: an update. Exp Biol Med Maywood 229: 1–11. Review.
- 7. Hamilton JA (1997) CSF-1 signal transduction. J Leukoc Biol 62: 145–155. Review.
- 8. Kacinski BM (1997) CSF-1 and its receptor in breast carcinomas and neoplasms of the female reproductive tract. Mol Reprod Dev 46: 71–74. Review.
- 9. Flidel-Rimon O, Roth P (1997) Effects of milk-borne colony stimulating factor-1 on circulating growth factor levels in the newborn infant. J Pediatr 131: 748–750.
- 10. Tang R, Beuvon F, Ojeda M, Mosseri V, Pouillart P, et al. (1992) M-CSF (monocyte colony stimulating factor) and M-CSF receptor expression by breast tumour cells: M-CSF mediated recruitment of tumour infiltrating monocytes? J Cell Biochem 50: 350–356.
- 11. Kirma N, Hammes LS, Liu YG, Nair HB, Valente PT, et al. (2007) Elevated expression of the oncogene c-fms and its ligand, the macrophage colony-stimulating factor-1, in cervical cancer and the role of transforming growth factor-beta1 in inducing c-fms expression. Cancer Res 67: 1918–1926.
- 12. Tamimi RM, Brugge JS, Freedman ML, Miron A, Iglehart JD, et al. (2008) Circulating colony stimulating factor-1 and breast cancer risk. Cancer Res 68: 18–21.
- 13. Ide H, Seligson DB, Memarzadeh S, Xin L, Horvath S, et al. (2002) Expression of colony-stimulating factor 1 receptor during prostate development and prostate cancer progression. Proc NatlAcadSci U S A 99: 14404–14409.
- 14. Espinosa I, Beck AH, Lee CH, Zhu S, Montgomery KD, et al. (2009) Coordinate expression of colony-stimulating factor-1 and colony-stimulating factor-1-related proteins is associated with poor prognosis in gynecological and nongynecologicalleiomyosarcoma. Am J Pathol 174: 2347–2356.
- 15. Kacinski BM, Scata KA, Carter D, Yee LD, Sapi E, et al. (1991) FMS (CSF-1 receptor) and CSF-1 transcripts and protein are expressed by human breast carcinomas in vivo and in vitro. Oncogene 6: 941–952.
- 16. Sapi E, Flick MB, Rodov S, Gilmore-Hebert M, Kelley M, et al. (1996) Independent regulation of invasion and anchorage-independent growth by different autophosphorylation sites of the macrophage colony-stimulating factor 1 receptor. Cancer Res 56: 5704–5712.
- 17. Rohrschneider LR, Rothwell VM, Nicola NA (1989) Transformation of murine fibroblasts by a retrovirus encoding the murine c-fms proto-oncogene. Oncogene 4: 1015–1022.
- 18. Maher MG, Sapi E, Turner B, Gumbs A, Perrotta PL, et al. (1998) Prognostic significance of colony-stimulating factor receptor expression in ipsilateral breast cancer recurrence. Clin Cancer Res 4: 1851–1856.
- 19. Scholl SM, Pallud C, Beuvon F, Hacene K, Stanley ER, et al. (1994) Anti-colony-stimulating factor-1 antibody staining in primary breast adenocarcinomas correlates with marked inflammatory cell infiltrates and prognosis. J Natl Cancer Inst 86: 120–126.
- 20. Kluger HM, Dolled-Filhart M, Rodov S, Kacinski BM, Camp RL, et al. (2004) Macrophage colony-stimulating factor-1 receptor expression is associated with poor outcome in breast cancer by large cohort tissue microarray analysis. Clin Cancer Res 10: 173–177.
- 21. Bingle L, Brown NJ, Lewis CE (2002) The role of tumour-associated macrophages in tumour progression: implications for new anticancer therapies. J Pathol 196: 254–265.
- 22. Lin EY, Nguyen AV, Russell RG, Pollard JW (2001) Colony-stimulating factor 1 promotes progression of mammary tumors to malignancy. J Exp Med 193: 727–740.
- 23. Aharinejad S, Abraham D, Paulus P, Abri H, Hofmann M, et al. (2002) Colony-stimulating factor-1 antisense treatment suppresses growth of human tumor xenografts in mice. Cancer Res 62: 5317–5324.
- 24. Aharinejad S, Paulus P, Sioud M, Hofmann M, Zins K, et al. (2004) Colony-stimulating factor-1 blockade by antisense oligonucleotides and small interfering RNAs suppresses growth of human mammary tumor xenografts in mice. Cancer Res 64: 5378–5384.
- 25. Wyckoff J, Wang W, Lin EY, Wang Y, Pixley F, et al. (2004) A paracrine loop between tumor cells and macrophages is required for tumor cell migration in mammary tumors. Cancer Res 64: 7022–7029.
- 26. Horiguchi J, Sherman ML, Sampson-Johannes A, Weber BL, Kufe DW (1988) CSF-1 and C-FMS gene expression in human carcinoma cell lines. Biochem Biophys Res Commun 157: 395–401.
- 27. Ramakrishnan S, Xu FJ, Brandt SJ, Niedel JE, Bast RC Jr, et al. (1989) Constitutive production of macrophage colony-stimulating factor by human ovarian and breast cancer cell lines. J Clin Invest 83: 921–926.
- 28. Yee LD, Liu L (2000) The constitutive production of colony stimulating factor 1 by invasive human breast cancer cells. Anticancer Res 20: 4379–4383.
- 29. Kluger HM, Kluger Y, Gilmore-Hebert M, DiVito K, Chang JT, et al. (2004) cDNA microarray analysis of invasive and tumorigenic phenotypes in a breast cancer model. Lab Invest 84: 320–331.
- 30. Patsialou A, Wyckoff J, Wang Y, Goswami S, Stanley ER, et al. (2009) Invasion of human breast cancer cells in vivo requires both paracrine and autocrine loops involving the colony-stimulating factor-1 receptor. Cancer Res 69: 9498–9506.
- 31. Filderman AE, Bruckner A, Kacinski BM, Deng N, Remold HG (1992) Macrophage colony-stimulating factor (CSF-1) enhances invasiveness in CSF-1 receptor-positive carcinoma cell lines. Cancer Res 52: 3661–3666.
- 32. Druker BJ (2004) Imatinib as a paradigm of targeted therapies. Adv Cancer Res 91: 1–30. Review.
- 33. Quintás-Cardama A, Kantarjian H, Cortes J (2009) Imatinib and beyond–exploring the full potential of targeted therapy for CML. Nat Rev ClinOncol 6: 535–543.
- 34. Dewar AL, Cambareri AC, Zannettino AC, Miller BL, Doherty KV, Hughes TP, Lyons AB (2005) Macrophage colony-stimulating factor receptor c-fms is a novel target of imatinib. Blood 105: 3127–3132.
- 35. Brownlow N, Russell AE, Saravanapavan H, Wiesmann M, Murray JM, et al. (2008) Comparison of nilotinib and imatinib inhibition of FMS receptor signaling, macrophage production and osteoclastogenesis. Leukemia 22: 649–652.
- 36. Mashkani B, Griffith R, Ashman LK (2010) Colony stimulating factor-1 receptor as a target for small molecule inhibitors. Bioorg Med Chem 18: 1789–1797.
- 37. Meegalla SK, Wall MJ, Chen J, Wilson KJ, Ballentine SK, et al. (2008) Structure-based optimization of a potent class of arylamide FMS inhibitors. Bioorg Med ChemLett 18: 3632–3637.
- 38. Ohno H, Kubo K, Murooka H, Kobayashi Y, Nishitoba T, et al. (2006) A c-fms tyrosine kinase inhibitor, Ki20227, suppresses osteoclast differentiation and osteolytic bone destruction in a bone metastasis model. Mol Cancer Ther 5: 2634–2643.
- 39. Roussel MF, Sherr CJ (1989) Mouse NIH 3T3 cells expressing human colony-stimulating factor 1 (CSF-1) receptors overgrow in serum-free medium containing human CSF-1 as their only growth factor. ProcNatlAcad Sci U S A 86: 7924–7927.
- 40. Rovida E, Spinelli E, Sdelci S, Barbetti V, Morandi A, et al. (2008) ERK5/BMK1 is indispensable for optimal colony-stimulating factor 1 (CSF-1)-induced proliferation in macrophages in a Src-dependent fashion. J Immunol 180: 166–4172.
- 41. Neve RM, Chin K, Fridlyand J, Yeh J, Baehner FL, et al. (2006) A collection of breast cancer cell lines for the study of functionally distinct cancer subtypes. Cancer Cell 10: 515–527.
- 42. Kao J, Salari K, Bocanegra M, Choi YL, Girard L, et al. (2009) Molecular profiling of breast cancer cell lines defines relevant tumor models and provides a resource for cancer gene discovery. PLoS One 4: e6146.
- 43. Minn AJ, Gupta GP, Siegel PM, Bos PD, Shu W, et al. (2005) Genes that mediate breast cancer metastasis to lung. Nature 436: 518–24.
- 44. van de Vijver MJ, He YD, van't Veer LJ, Dai H, Hart AA, Voskuil DW, et al. (2002) A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 347: 1999–2009.
- 45. Parker JS, Mullins M, Cheang MCU, Leung S, Voduc D, et al. (2009) Supervised risk predictor of breast cancer based on intrinsic subtypes. J ClinOncol 27: 1160–7.
- 46. Sims D, Bursteinas B, Gao Q, Jain E, MacKay A, et al. (2010) ROCK: a breast cancer functional genomics resource. Breast Cancer Res Treat 124: 567–572.
- 47. Fridlyand J, Snijders AM, Ylstra B, Li H, Olshen A, et al. (2006) Breast tumor copy number aberration phenotypes and genomic instability. BMC Cancer 6: 96–109.
- 48. Ashmun RA, Look AT, Roberts WM, Roussel MF, Seremetis SF, et al. (1989) Monoclonal antibodies to the human CSF-1 receptor (c-FMS proto-oncogene product) detectepitopes on normal mononuclear phagocytes and on leukemic blast cells. Blood 73: 827–837.
- 49. Stanley ER (1985) The macrophage colony-stimulating factor, CSF-1. Methods in enzymology 116: 564–587.
- 50. Livak KJ, Schmittgen TD (2001) Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) method. Methods 25: 402–408.
- 51. Blick T, Widodo E, Hugo H, Waltham M, Lenburg ME, et al. (2008) Epithelial mesenchymal transition traits in human breast cancer cell lines. Clin Exp Metastasis 25: 629–642. Review.
- 52. Perou CM, Sørlie T, Eisen MB, van de Rijn M, Jeffrey SS, et al. (2000) Molecular portraits of human breast tumours. Nature 406: 747–752.
- 53. Gallet M, Mentaverri R, Sévenet N, Brazier M, Kamel S (2006) Ability of breast cancer cell lines to stimulate bone resorbing activity of mature osteoclasts correlates with an anti-apoptotic effect mediated by macrophage colony stimulating factor. Apoptosis 11: 1909–1921.
- 54. Kirma N, Luthra R, Jones J, Liu YG, Nair HB, et al. (2004) Overexpression of the colony-stimulating factor (CSF-1) and/or its receptor c-fms in mammary glands of transgenic mice results in hyperplasia and tumor formation. Cancer Res 64: 4162–4170.
- 55. Paulus P, Stanley ER, Schäfer R, Abraham D, Aharinejad S (2006) Colony-stimulating factor-1 antibody reverses chemoresistance in human MCF-7 breast cancer xenografts. Cancer Res 66: 4349–4356.
- 56. Roussel MF (1997) Regulation of cell cycle entry and G1 progression by CSF-1. Mol Reprod Dev 46: 11–8. Review.
- 57. Lee AW, Nambirajan S, Moffat JGCSF-1 activates MAPK-dependent andp53-independent pathways to induce growth arrest of hormone-dependent humanbreast cancer cells. Oncogene 1999 Dec 9; 18(52): 7477–94.
- 58. Esparís-Ogando A, Díaz-Rodríguez E, Montero JC, Yuste L, Crespo P, et al. (2002) Erk5 participates in neuregulin signal transduction and is constitutively active in breast cancer cells overexpressing ErbB2. Mol Cell Biol 22: 270–285.
- 59. Lee AW, States DJ (2006) Colony-stimulating factor-1 requires PI3-kinase-mediated metabolism for proliferation and survival in myeloid cells. Cell Death Differ 13: 1900–1914.
- 60. Cully M, You H, Levine AJ, Mak TW (2006) Beyond PTEN mutations: the PI3K pathway as an integrator of multiple inputs during tumorigenesis. Nat Rev Cancer 6: 184–192. Review.
- 61. Sotiriou C, Pusztai L (2009) Gene-expression signatures in breast cancer. N Engl J Med 360: 790–800. Review.
- 62. Snijders AM, Nowak N, Segraves R, Blackwood S, Brown N, et al. (2001) Assembly of microarrays for genome-wide measurement of DNA copy number. Nat Genet 29: 263–264.