Transforming growth factor-β-induced epithelial-mesenchymal transition facilitates epidermal growth factor-dependent breast cancer progression

Abstract

Transforming growth factor-β (TGF-β) and epidermal growth factor (EGF) have critical roles in regulating the metastasis of aggressive breast cancers, yet the impact of epithelial-mesenchymal transition (EMT) induced by TGF-β in altering the response of breast cancer cells to EGF remains unknown. We show in this study that murine metastatic 4T1 breast cancer cells formed compact and dense spheroids when cultured under three-dimensional (3D) conditions, which was in sharp contrast to the branching phenotypes exhibited by their nonmetastatic counterparts. Using the human MCF10A series, we show that epithelial-type and nonmetastatic breast cancer cells were unable to invade to EGF, whereas their mesenchymal-type and metastatic counterparts readily invaded to EGF. Furthermore, EMT induced by TGF-β was sufficient to manifest dense spheroid morphologies, a phenotype that increased primary tumor exit and invasion to EGF. Post-EMT invasion to EGF was dependent on increased activation of EGF receptor (EGFR) and p38 mitogen-activated protein kinase, all of which could be abrogated either by pharmacologic (PF-562271) or by genetic (shRNA) targeting of focal adhesion kinase (FAK). Mechanistically, EMT induced by TGF-β increased cell-surface levels of EGFR and prevented its physical interaction with E-cadherin, leading instead to the formation of oncogenic signaling complexes with TβR-II. Elevated EGFR expression was sufficient to transform normal mammary epithelial cells, and to progress their 3D morphology from that of hollow acini to branched structures characteristic of nonmetastatic breast cancer cells. Importantly, we show that TGF-β stimulation of EMT enabled this EGFR-driven breast cancer model to abandon their inherent branching architecture and form large, undifferentiated masses that were hyperinvasive to EGF and showed increased pulmonary tumor growth upon tail vein injection. Finally, chemotherapeutic targeting of FAK was sufficient to revert the aggressive behaviors of these structures. Collectively, this investigation has identified a novel EMT-based approach to neutralize the oncogenic activities of EGF and TGF-β in aggressive and invasive forms of breast cancer.

Figure 1

Loss of mammary branching characterizes increasing metastatic potential and is induced by TGF-β1. A) 4T1 (highly metastatic), 4T07 (invasive, but nonmetastatic), and 67NR (noninvasive and nonmetastatic) mammary carcinoma cells were grown in 3D culture for 5 days and representative structures are shown. Data indicate that branching mammary structures are characteristic of non-metastatic cells, a phenotype that is abandon by fully metastatic cells. B) The cells described in panel A expressing firefly luciferase were engrafted onto the mammary fat pad of 4 week old female Balb/C mice (4T1=10,000 cells; 4T07 and 67NR= 100,000 cells). Mean pulmonary luminescence (Area flux) is shown at various time points as a measure of metastasis from the primary mammary tumor to the lungs (n=5 mice per group, ±SE,**P<0.01). C) Pre- and post-EMT 4T07 cells were cultured in 3D-organotypic conditions for 5 days and representative structures are shown. D) The Pre- and Post-EMT 4T07 cells were analyzed by immunoblot for EGFR, E-cadherin (E-cad), Vimentin (Vim) and β-Actin (Actin) as a loading control. Shown are representative immunoblots that were completed three times with similar results. E) Pre- and Post-EMT 4t07 cells were allowed to invade through Matrigel-coated synthetic membranes in response to 2% serum or EGF. Data are normalized to a serum-free control (solid line at 100%) and are the mean (±SE) of three independent experiments completed in triplicate (***P<0.001). F) The Pre- and Post-EMT 4T07 cells as in panel C were engrafted onto the mammary fat pad of Balb/C mice as in Panel B. Mean pulmonary luminescence (Area flux) is shown at various time points following engraftment (n=5 mice per group, ±SE, ***P<0.001, **P<0.01).

Figure 2

TGF-β stimulation of EMT results in the generation of highly invasive spheroids that possess elevated EGFR cell surface expression. A) NMuMG cells were serum-deprived (0.5% FBS) for 6 h in the absence or presence of the Src inhibitor, PP2 (10μM) or the EGFR inhibitor, AG1478 (1 μM), at which point they were stimulated for 30 min with EGF and analyzed for the phosphorylation of EGFR [pEGFR(Y845)], [pEGFR(Y1116)] or ERK1/2 (pErk1/2). The resulting immunoblots were stripped and reprobed with antibodies against EGFR or ERK1/2, to monitor differences in protein loading. Images are from a representative experiment that was performed at least three times with similar results (NS, no stimulation). B) Pre-EMT and post-EMT NMuMG (48 hour pretreatment with TGF-β1 (5ng/ml)) cells were induced to invade synthetic basement membranes by either serum (2%) or EGF (50 ng/ml) as indicated. Data are the mean (± SE) invasion relative to serum free media for both Pre- and Post-EMT cells (solid line) observed in three independent experiments completed in triplicate (**P < 0.01). Accompanying photomicrographs depict the morphology of pre-EMT (lower left) and post-EMT (lower right) NMuMG cells when propagated for 24 h on Matrigel cushions. C) Pre-EMT and post-EMT NMuMG cells were incubated with either TGF-β1 (5 ng/ml), EGF (50 ng/ml), or both cytokines for 24 h prior to labeling cellular DNA by administration of [³H]thymidine. Data are the mean (± SE) quantities of incorporated [³H]thymidine normalized to unstimulated controls (solid line) observed in three independent experiments completed in triplicate (**P < 0.01). D) Quiescent NMuMG cells were stimulated with TGF-β1 (5 ng/ml) for indicated times over a period of 48 h, at which point alterations in EGFR expression were monitored by immunoblotting. Stripped membranes were reprobed with anti-β-tubulin (β-Tub) to monitor differences in protein loading. Images are from a representative experiment that was performed at least four times with similar results. E) Quiescent NMuMG cells were stimulated with TGF-β1 (5 ng/ml) as in panel D. Afterward, total RNA was isolated and subjected to semi-quantitative real-time PCR to monitor the expression of EGFR or β3 integrin, which served as a marker for EMT induced by TGF-β. Data are the mean (±SD) fold changes in gene expression relative to untreated control cells observed in at least three independent experiments. F) Quiescent NMuMG cells were stimulated with TGF-β1 (5 ng/ml) for 18 h in the absence or presence the Src inhibitor, PP2 (10 μM). Alterations in EGFR expression were monitored by immunoblotting as in panel D.G) Cell surface expression of EGFR in pre-EMT and post-EMT NMuMG cells was determined by flow-cytometric analysis of bound Alexa488-labeled EGF. The presented histogram is representative of three independent experiments.

Figure 3

EMT increases the coupling of EGFR to p38 MAPK activation via FAK. A) Control (i.e., scrambled shRNA; scram) and FAK-depleted (shFAK) NMuMG cells were incubated in the absence (Pre-EMT) or presence of TGF-β1 (5 ng/ml; Post-EMT) for 48 h, at which point they were washed, serum deprived for 6 h and subsequently stimulated with EGF (50 ng/ml) for 30 min and analyzed for phospho-p38 MAPK (p-p38) or phospho-Y845-EGFR [p-EGFR(Y845)] as indicated. The resulting immunoblots were stripped and reprobed with antibodies against p38 MAPK, EGFR, FAK, and β-actin (Actin) to monitor differences in protein loading. Images are from a representative experiment that was performed four times with similar results (NS, no stimulation). B) Quiescent control (i.e., YFP) and EGFR-expressing NMuMG cells were stimulated for 30 min with either TGF-β1 (5 ng/ml) or EGF (50 ng/ml) and analyzed to monitor the phosphorylation status of Smad2 (pSmad2), Smad3 (pSmad3), p38 MAPK (p-p38), and ERK1/2 (pErk1/2). The resulting immunoblots were stripped and reprobed for total Smad2/3 (tSmad2/3), p38 MAPK (t-p38), ERK1/2 (tErk1/2), EGFR and β-actin (Actin) as loading controls. Images are from a representative experiment that was performed at least four times with similar results. C) Control (i.e., scrambled shRNA; scram) and FAK-depleted pre- and post-EMT NMuMG cells were allowed to invade synthetic basement membranes in response to EGF (50 ng/ml). Data are the mean (± SE) invasion relative to unstimulated MECs (i.e. serum-free media placed in the bottom chamber = solid line set to 100%) observed in three independent experiments completed in triplicate. D) Pre- and post-EMT control (i.e., YFP) NMuMG cells were induced to invade to EGF (50 ng/ml) in the absence or presence of the following pharmacological inhibitors a FAK inhibitor, PF-562271 (271, 1 μM), TβR-I inhibitor, SB431542 (T1-I, 10 μg/ml), p38 MAPK inhibitor, SB203580 (p38-I, 10 μM), EGFR inhibitor, AG1478 (AG, 1 μM). E) Data consists of the identical invasion experiments as in panelD completed with the EGFR-expressing (EGFR) NMuMG cells. Presented data in panels D and E are the mean (±SE) invasion relative to an EGF free control (i.e. serum free media placed in the bottom chamber = NS, set to 100%) observed in three independent experiments completed in triplicate. (*P<0.05; **P<0.01).

Figure 4

EGFR overexpression transforms NMuMG cells and sensitizes them to EMT by altering EGFR complexes. A) Control (i.e., YFP), PyMT-, and EGFR-expressing NMuMG cells were engrafted onto the mammary fat pads of Nu/Nu mice, whose development of mammary tumors was monitored over 32 days. Data are the mean (± SE) tumor volumes measured for indicated NMuMG tumor variants. (*P<0.05, n=6 mice/group). Inset depicts PyMT expression in NMuMG cells, which served as a positive control for tumor formation. B) Control (i.e., YFP) and EGFR-expressing NMuMG cells were stimulated with TGF-β1 (5 ng/ml), EGF (50 ng/ml), or both cytokines for 48 h as indicated. Afterward, detergent-solubilized whole-cell extracts were prepared and subjected to immunoblot analyses to monitor changes in the expression of E-cadherin (E-cad), N-cadherin (N-cad), cycloxygenase-2 (Cox2), EGFR, and β-actin (Actin), which served as a loading control. Images are from a representative experiment that was performed at least three times in its entirety with similar results. C) Control (i.e., YFP) and EGFR-expressing NMuMG cells were stimulated as described in panelB prior to visualizing alterations in the actin cytoskeleton by direct phalloidin fluorescence. Images are representative photomicrographs (600x) from a single experiment that was performed two times with identical results. D) Control (i.e., YFP) and EGFR-expressing NMuMG cells were allowed to reach confluence to normalize E-cad expression levels and incubated in the absence or presence of TGF-β1 (5 ng/ml) for 24 h prior to isolating EGFR complexes by immunoprecipitation. The resulting EGFR immunocomplexes (EGFR I.P.) were immunoblotted with antibodies against TβR-II, E-cadherin (E-cad), FAK, and β-actin (Actin; top panel). Aliquots of the original whole-cell extract (input) was also immunoblotted with antibodies against TβR-II, E-cadherin (E-cad), FAK, β-actin (Actin), EGFR, and β3 integrin to control for differences in protein loading (bottom panel). Images are from a representative experiment that was performed three times with similar results.

Figure 5

EGFR expression enhanced the delocalization of E-cadherin induced by EMT. A) Control (i.e., YFP) and EGFR-expressing (EGFR) NMuMG cells were incubated in the absence (Pre-EMT) or presence (Post-EMT) of TGF-β1 (5 ng/ml) 24 h, at which point they were processed for E-cadherin (E-cad) and DAPI immunofluorescence (400x). Junctional localization of E-cadherin was slightly disrupted in EGFR-expressing NMuMG cells as compared to their control counterparts, except for regional pockets of cells designated by white outline. Delocalized and degradation of E-cadherin in response to TGF-β was enhanced in EGFR expressing cells. B) EGFR-expressing cells were stimulated to undergo EMT as in panel A, at which point they were subjected to dual immunofluorescent staining to visualize E-cadherin (E-cad, red) and EGFR (green). Regions lacking EGFR expression (white outline) exhibit normal junctional localization of E-cadherin, while EMT induction resulted in the appearance two populations of NMuMG cells: one that was EGFR-positive and E-cadherin-negative (*) and a second that lacked expression of both EGFR and E-cadherin (arrows). Photomicrographs (400x) presented in panels A and B are representative of three independent experiments.

Figure 6

MEC branching induced by EGFR is dependent upon TGF-β:FAK signaling. A) Control (i.e., YFP) and EGFR-expressing (EGFR) NMuMG cells were propagated for 10 days in 3D-ogranotypic cultures, and subsequently were processed for direct phalloidin and DAPI fluorescence to visualize the actin cytoskeleton and nuclei, respectively. Shown are representative photomicrographs (YFP=400x; EGFR=100x) from a single experiment that was performed more than five times with identical results. B) Control (i.e., YFP) and EGFR-expressing cells were propagated as in panel A in the absence (NS) or presence of either a) EGF (50 ng/ml); b) the TβR-I inhibitor, SB431542 (TβR-I inh, 10 μM); c) the FAK inhibitor, PF-562271 (PF-271, 1 μM), d) the EGFR inhibitor, AG1478 (1 μM). Shown are representative photomicrographs (small bar = 100x; large bar = 400x) from a single experiment that was performed at least 3 times with similar results. C) Parental NMuMG cells were incubated in the absence (Pre-EMT) or presence of TGF-β1 (5 ng/ml; Post-EMT) for 48 h prior to their isolation and propagation for 10 days in 3D-organotypic cultures. Afterward, the resulting organoids were stained with DAPI to visualize the nuclei and the percentage of hollowed acini was quantified. Data are the mean (±SE; n=3) percent of hollowed acinar structures. Representative acini are shown.

Figure 7

EMT prevents EGF-induced mammary branching and increases pulmonary outgrowth. A) Control (i.e., YFP) and EGFR-expressing (EGFR) NMuMG cells were incubated in the absence (Pre-EMT) or presence of TGF-β1 (5 ng/ml; Post-EMT) for 48 h prior to their isolation and propagation for 10 days in 3D-organotypic cultures supplemented with either EGF (50 ng/ml) or the EGFR antagonist, AG1478 (1 μM) as indicated. Differences in organoid morphology were monitored by phase-contrast microscopy (100x). B) Alterations in organoid branching were quantified and presented as the mean (±SE; n=3) percentage of branched structures. C) NMuMG-EGFR cells were transduced with firefly luciferase, treated with TGF-β1 as described in panel A and injected into the lateral tail vein of 6 week old, female, Nu/Nu mice. Shown is a longitudinal study of a representative mouse from each group imaged at the indicated time points (n=5 mice per group). D) A survival curve of mice injected with NMuMG-EGFR cells as in panel C, indicating that induction of EMT in NMuMG-EGFR cells decreases the time in which lethal pulmonary tumor burden is reached. E) Schematic depicts the relationship between MEC invasion, their EMT status induced by TGF-β, and their 3D-culture morphologies. In particular, EMT stimulated by TGF-β suppressed the branching of developing organoids including that induced by EGF/EGFR and instead resulted in the formation of large, dense spheroids that were hyper-invasive to EGF. This unique invasive morphology and phenotype is metastable (double-sided arrow) in normal mammary epithelial cells and required autocrine TGF-β signaling for its manifestation. In stark contrast, post-EMT invasospheres in mammary carcinoma cells are stabilized in their hyper-invasive phenotype (single-sided arrow) and hence rendered independent of the need for continued TGF-β stimulation. Our findings suggest that these novel invasive spheroids likely represent the post-EMT subpopulation of pathologically invasive and metastatic breast cancer cells.