Activation of FOXO3a by the Green Tea Polyphenol Epigallocatechin-3-Gallate Induces Estrogen Receptor α Expression Reversing Invasive Phenotype of Breast Cancer Cells

Consumption of tea around the world is believed second only to water, although the levels vary widely by country. Green tea is rich in polyphenols, which have potent antioxidant properties. The most abundant polyphenol in green tea is epigallocatechin-3 gallate (EGCG). Many cancer epidemiologic studies have shown an inverse association between green tea consumption and cancer incidence, of which breast cancer risk reduction and tea consumption are among the best characterized (1). For example, consumption of green tea was closely associated with decreased numbers of axillary lymph node metastases among premenopausal patients with stage I and II breast cancer and with increased expression of progesterone receptor and estrogen receptor α (ERα) among postmenopausal women (2). An inverse correlation between regular green tea consumption before breast cancer diagnosis and the subsequent risk of recurrence has also been shown (3).

Studies using animal models have similarly shown a powerful protective role of green tea against mammary gland malignancies. Initial studies were done with chemical carcinogens. For example, diets containing green tea catechins significantly increased survival of female Sprague-Dawley (S-D) rats treated with 7,12-dimethylbenz(a)anthracene (DMBA) compared with those given regular chow diet (4). In work from our group, S-D rats were given either 0.3% green tea or water as their sole fluid source and then treated with DMBA. We observed a significant increase in mean latency to first tumor, an approximate 70% decrease in tumor burden, and an 87% reduction in the number of invasive tumors per tumor-bearing animal in the rat group drinking green tea (5). Similar protection has been seen in other animal models of cancer, including those of the prostate, skin, and lung (1). For example, oral infusion of a polyphenolic fraction isolated from green tea significantly inhibited prostate cancer development and increased survival in the transgenic adenocarcinoma of mouse prostate (TRAMP) mouse model (6). Furthermore, oral application of green tea polyphenols resulted in markedly reduced tumor size in UV B–induced DMBA-initiated skin tumors in mice (7).

To begin to examine the molecular mechanisms of green tea action, we have turned to cells in culture transformed by specific oncogenes, e.g., Her-2/neu or ErbB2, the second member of the epidermal growth factor (EGF) receptor family or the Nuclear Factor-κB (NF-κB) c-Rel subunit. Her-2/neu is overexpressed in about 30% of human breast cancers and indicates poor prognosis. Her-2/neu overexpression confers anchorage-independent growth and invasion properties in vitro (8), and correlates with axillary lymph node positivity and the presence of vascular invasion and metastases in breast cancer patients (9). The gene expression signature of Her-2/neu–transformed tumor cells shows high expression of genes promoting proliferation, migration, invasion, and metastasis and low expression of those promoting apoptosis and immune response (10). Constitutive activation of the phosphoinositide-3-kinase (PI3K) to Akt/protein kinase B to NF-κB signaling pathway has been implicated in the induction of growth and transformed phenotype of Her-2/neu–overexpressing cells (11–13). Of note, we observed that treatment of Her-2/neu–overexpressing breast cancer cells with EGCG reduced PI3K, Akt, and NF-κB activity and anchorage-independent growth (14).

Epithelial to mesenchymal transition (EMT), which has been recognized for several decades as a critical feature of embryogenesis, has more recently been shown to be relevant for cancer. During EMT, cancer cells lose expression of proteins that promote cell-cell contact such as E-cadherin and γ-catenin and acquire mesenchymal markers such as the zinc-finger transcription factor Snail, vimentin, fibronectin, and N-cadherin (15, 16), which promote cell migration and invasion (17). Recent work by several groups has uncovered a signaling pathway that inhibits EMT in cancer cells linked to ERα signaling. Transcription of MTA3, a component of the histone deacetylase NuRD, is regulated by ERα (16). In turn, NuRD represses transcription of the gene encoding Snail, which represses E-cadherin gene transcription (18). Thus, ERα plays a central role in the control of EMT. The expression status of Her-2/neu and ERα correlate inversely in breast cancer (19, 20). Reduced ERα activity in breast tumors is frequently associated with a switch from an epithelial architecture to induction of invasive growth (15, 16). We recently showed that ERα synthesis is controlled by the Forkhead family transcription factor FOXO3a, and that the activity of FOXO3a is greatly reduced in Her-2/neu–driven cancers due to constitutive Akt kinase activity (21). Because Her-2/neu–mediated activation of Akt was repressed upon EGCG treatment (14), here we tested the hypothesis that activation of FOXO3a by EGCG can revert invasive phenotype of breast cancer cells that are ERα low. We show that EGCG represses invasive phenotype of breast cancer cells driven by Her-2/neu, as well by the NF-κB c-Rel subunit and protein kinase CK2, via inducing FOXO3a, which activates the ERα to E-cadherin pathway leading to an epithelial expression profile. Thus, our studies elucidate, for the first time, a novel functional role of FOXO3a as an important upstream mediator of the anti-invasive effects of ERα in breast cancer cells.

Cell growth and treatment conditions. The mouse mammary tumor virus (MMTV)–Her-2/neu cell line NF639 (kindly provided by P. Leder, Harvard Medical School, Boston, MA) was derived from a mammary gland tumor and cultured as described previously (22). The rel-3875 line was derived from a mammary adenocarcinoma with poorly differentiated large cells that developed in the MMTV–c-rel mouse (23). The rel/CK2-5839 cell line was derived from a mammary adenocarcinoma in the MMTV–c-rel × MMTV-CK2α bitransgenic mouse (24). The NF639, rel-3875, and rel/CK2-5839 cells are invasive and express low levels of ERα (ref. 21 and see below). NMuMG, which is an untransformed, immortalized mouse mammary epithelial cell line, was cultured as described previously (25). The MDA-MB-231 ERα-negative breast cancer cell line was grown as described previously (24). EGCG, purchased from LKT Laboratories Inc., was dissolved in sterile DMSO. ICI 182,780 and transforming growth factor-β1 (TGF-β1) were purchased from Tocris Cookson and R&D Systems, respectively.

Plasmid constructs and transfection analysis. The expression vectors for wild-type (WT) FOXO3a, constitutively active A3 FOXO3a mutant, dominant negative FOXO3a (FOXO DN), and parental pECE vectors were kindly provided by M. Greenberg and A. Brunet (Harvard Medical School, Boston, MA; ref. 26). In the A3 mutant, three sites of Akt phosphorylation of FOXO3a (T32, S253, S315) were mutated to alanine residues. The dominant negative TM ΔDB FOXO3a mutant has a deletion in the DNA binding domain and no longer binds to DNA; however, it still localizes in the nucleus and may titer transcriptional coactivators away from endogenous FOXO, thereby exerting a dominant-negative effect (27). The E-cadherin promoter reporter vector contained −21 bp of the E-cadherin gene (28) subcloned into the pGL3-Basic luciferase (Luc) reporter construct. The SV40 β-galactosidase (β-gal) reporter vector (23) was used to normalize transfection efficiency. For reporter assays, cells were plated in 6- or 12-well plates, and 24 h later, cells were transfected, in triplicate, with the indicated amounts of DNA using Fugene6 transfection reagent (Roche Diagnostics Co.). Forty-eight hours after transfection, luciferase assays were done as described previously (23).

Matrigel assay. Matrigel was diluted to a concentration of 6.3 mg/mL with serum free medium (DMEM) and stored at −80°C. Before use, diluted Matrigel was thawed on ice overnight. For the bottom layer, 200 μL of Matrigel solution was added into a 24-well tissue culture plate and incubated at 37°C for 30 min to allow the Matrigel to solidify. A single cell suspension (5 × 10³ cells/10 μL) was then mixed with 190 μL of Matrigel at 4°C and added to the solidified bottom layer. The plates were incubated at 37°C for 30 min to allow the Matrigel to solidify, and 500 μL complete medium containing the appropriated concentration of EGCG or DMSO was then added. Following incubation at 37°C for 5 to 7 days, the cell growth was analyzed using a Zeiss Axiovert 200 M microscope.

Migration and invasion assays. Suspensions of 5×10⁵ cells were layered in the upper compartment of a Transwell (Costar) on an 8-mm-diameter polycarbonate filter (8 μm pore size) and incubated at 37°C for the indicated times. For invasion assays, filters were precoated with 30 μg of Matrigel. Migration of the cells to the lower side of the filter was evaluated with the acid phosphatase enzymatic assay using p-nitrophenyl phosphate and OD_{410 nm} determination.

Immunoblot analysis. Cells were rinsed with cold PBS and harvested in lysis buffer [50 mmol/L Tris-HCl (pH 8.0), 5 mmol/L EDTA (pH 8.0), 150 mmol/L NaCl, 0.5 mmol/L DTT, 2 μg/mL aprotinin, 2 μg/mL LP, 0.5 mmol/L phenylmethylsulfonyl fluoride, 0.5% NP40]. Whole cell extracts were obtained by sonication, followed by centrifugation at 14,000 rpm for 30 min. Samples were subjected to electrophoresis in an 8% or 10% polyacrylamide-SDS gel and immunoblotting, as previously described (13). Antibodies against ERα and vimentin were purchased from NeoMarker. Antibodies against Snail (sc-10432) were purchased from Santa Cruz Biotechnology. The E-cadherin and γ-catenin antibodies were purchased from BD Transduction Lab, and the β-actin (AC-15) and FOXO3a antibodies were from Sigma and Upstate Biotechnology, respectively.

RNA preparation and reverse transcription-PCR analysis. Cultures were washed twice with PBS, and total RNA was harvested using the UltraspecII RNA isolation kit (Biotecx), following the manufacturer's instructions. The primers and annealing temperatures used for PCR are as follows: E-cadherin 5′-CAGCGTCAACTGGACCATTG-3′ and 5′-CCACCGTTCTCCTCCGTAGA-3′ at 65°C; MTA3 5′-CAACCCAGCCTTCACATGAGT-3′ and 5′-TTCCACGGGAGAAAGTCCTG-3′ at 65°C and glyceraldehyde-3-phosphate dehydrogenase (Gapdh) 5′-TCACCATCTTCCAGGAG-3′ and 5′-GCTTCACCACCTTCTTG-3′ at 60°C. The cycle numbers for each gene and RNA preparation from different cell types were determined to be in the linear and quantitative range in pilot experiments. PCR products were analyzed on 5% acrylamide gels run with 0.5× Tris-borate EDTA buffer and stained with GelStar nucleic acid stain (Bio Whittaker Molecular Applications).

Retroviral gene delivery. The cDNAs of human WT FOXO3a and the A3 mutant and dominant negative FOXO3a (FOXO DN) were excised from the pECE expression vector by HindIII and XbaI digestion and end-filled by Klenow enzyme. The blunt-end fragment was then ligated into the SnaBI site of the pBabe-puro ecotropic retroviral vector (29), yielding pBabe-puro-A3FOXO3a and pBabe-puro-dnFOXO3a, respectively. Phoenix packaging cells were used for the generation of retrovirus, which were selected with 500 μg/mL hygromycin to increase Gag and Pol viral protein expression. Briefly, P100 dishes of 80% confluent Phoenix cells were transfected with 15 μg of pBabe-puro-FOXO3a (A3 or DN) or pBabe-puro along with 5 μg of an Env-expressing vector. After 24 h, the medium was changed, and cells were incubated for another 24 h at 32°C to increase retrovirus half-life. Supernatants containing retrovirus were then harvested, filtered, and transferred onto target cells in the presence of 2 μg/mL polybrene (Sigma). After 24 h, infected cells were washed, selected with complete medium plus 4 μg/mL puromycin for 4 days, and expanded in medium containing 1 μg/mL puromycin. As a positive control, cells were infected with the pBabe-puro-GFP retrovirus, indicating more than 90% efficiency in retroviral infection (data not shown).

Green tea polyphenol EGCG promotes a less invasive profile of gene expression in Her-2/neu–overexpressing cells. To further elucidate the molecular mechanisms whereby EGCG alters Her-2/neu–mediated transformation, microarray analysis was done with samples from two independent experiments (Affymetrix Mouse 430A 2.0 chips). RNA samples from NF639 cells treated for 24 h with 60 μg/mL EGCG showed an up-regulation of E-cadherin, α-catenin, ERα, and MTA3 and a down-regulation of the proinvasive master regulators Snail1 and Snail3 compared with control DMSO-treated cells; whereas no differences were seen in the expression of control housekeeping genes β-actin, glyceraldehyde 3-phosphate, peptidylprolyl isomerase A, or phosphoglycerate kinase 1 (data not shown). To verify the findings with microarray, NF639 cells were treated with EGCG or DMSO and total RNA harvested and subjected to analysis for expression of E-cadherin and MTA3 by reverse transcription-PCR (RT-PCR). EGCG treatment induced the expression of these two anti-invasive genes (Fig. 1A). Furthermore, when protein extracts were probed for the expression of E-cadherin, a substantial induction was observed (Fig. 1B). Similarly, the epithelial marker γ-catenin was induced, whereas expression of Snail was decreased. EGCG treatment also induced expression of FOXO3a and its target gene ERα (Fig. 1B), consistent with our previous findings (21). Thus, EGCG induces an anti-invasive gene expression signature in the highly malignant Her-2/neu–overexpressing NF639 breast cancer cells.

EGCG inhibits invasive phenotype of NF639 cells. One of the hallmarks of invasive mesenchymal phenotype is the branching colony formation in Matrigel. The Her-2/neu–overexpressing NF639 cells, which grow effectively in soft agar (14), formed branched colony structures with a highly invasive phenotype in Matrigel (Fig. 1C, DMSO and data not shown). When 40 μg/mL of EGCG was added to the NF639 cells in Matrigel, the branched structures were dramatically reduced (Fig. 1C,, left). Similar data were obtained upon treatment with 20 μg/mL EGCG (data not shown), which has little effect on growth of these cells (14). Furthermore, the cells took on a more flattened, normal morphology when grown on plastic (Fig. 1C,, right). Lastly, incubation with EGCG profoundly reduced the ability of NF639 cells to invade through Matrigel (Fig. 1D). Together, these data indicate EGCG can reverse the mesenchymal characteristics of Her-2–driven mammary cancer cells, promoting a less invasive phenotype.

EGCG inhibits invasive phenotype of c-Rel– and CK2-driven mammary cancer cells. Almost 90% of human breast cancers contain active, nuclear c-Rel (30), which can play a causal role in mammary tumorigenesis (23). The serine/threonine protein kinase CK2, which has been implicated in basal NF-κB activity (31), is overexpressed in most human breast cancers (32). We next tested the effectiveness of EGCG on invasive tumor lines rel-3875 and rel/CK2-5839 cells, which were derived from mammary tumors in MMTV–c-rel transgenic, and MMTV–c-rel × MMTV-CK2α bitransgenic mice, respectively. EGCG inhibited the ability of rel-3875 and rel/CK2-5839 cells to form invasive colonies in Matrigel (Fig. 2A,, top) and induced a more normal, i.e., less elongated, epithelial cell morphology (Fig. 2A,, bottom). Similarly, invasive colony formation of rel/CK2-5839 cells in Matrigel was inhibited upon treatment with 20 μg/mL EGCG for 1 day, which has no effect on growth rate (data not shown). The effects of EGCG treatment on expression of EMT genes were examined next. Consistent with the morphologic changes, EGCG treatment very effectively induced expression of epithelial markers E-cadherin and γ-catenin and reduced the level of Snail, the repressor of E-cadherin gene transcription, in rel-3875 and rel/CK2-5839 cells (Fig. 2B). Furthermore, EGCG treatment led to increased levels of ERα in both lines (Fig. 2B). EGCG also profoundly inhibited the ability of rel-3875 cells to invade through Matrigel (Fig. 2C), as seen above with NF639 cells. Lastly, we assessed the ability of EGCG to reverse the formation of invasive colonies by rel-3875 cells. After an initial 4 days in Matrigel, substantial invasive colony formation was noted (Fig. 2D, day 0). At this time, either EGCG or carrier DMSO was added, and the plates were incubated for an additional 10 days. EGCG completely reverted invasive colony formation, whereas continued growth of invasive colonies was observed in the presence of carrier DMSO (Fig. 2D, day 10). Thus, EGCG potently inhibits the invasive properties of breast cancer cells driven by c-rel and c-relxCK2, reverting mesenchymal and promoting a more epithelial phenotype.

EGCG induces FOXO3a, which reverses invasive phenotype. Given the pivotal role of ERα in EMT and our findings that FOXO3a is a major regulator of ERα gene transcription (21), we next examined the effects of EGCG treatment on FOXO3a expression in rel-3875 and rel/CK2-5839 cells. EGCG caused a substantial increase in FOXO3a in the two lines (Fig. 2B). To directly test the role of FOXO3a in preventing invasive growth in Matrigel, NF639 cells were infected with either a retrovirus expressing a constitutively active form of FOXO3a (A3; ref. 26) or the empty pBabe vector. The exogenous expression of FOXO3a was confirmed by immunoblotting and led to an induction of E-cadherin protein levels (Fig. 3A). Moreover, active FOXO3a reduced the level of Snail protein, the transcriptional repressor of the E-cadherin gene (Fig. 3A). When plated in Matrigel, the presence of pBabe had minimal effects on the cellular phenotype as judged by the extensive branched structures (Fig. 3B,, top). Importantly, when A3 FOXO3a-expressing NF639 cells were grown in Matrigel, the branching growth was completely blocked, resulting in the appearance of tightly clustered cells. When grown on plastic, A3 FOXO3a expression caused the cells to have a more epithelial morphology (Fig. 3B,, bottom), but the growth of the cells was only modestly affected, suggesting that the reduced size of colonies in Matrigel was not simply a result of inhibition of cell cycle progression. Furthermore, ectopic A3 FOXO3a expression caused a substantial decrease in migration of NF639 cells (Fig. 3C). These findings could be further extended to the rel/CK2-5839 (Fig. 3D), consistent with the observed decrease in levels of the master regulator of EMT Snail (Fig. 3D , inset). Thus, sustained FOXO3a activity resulting from the A3 constitutively active mutant reverts the invasiveness conferred by the oncogenic inducers Her-2/neu and c-Rel/CK2.

FOXO3a prevents TGF-β1–induced EMT of NMuMG cells. Another commonly used cell model to study induction of invasive phenotype is the mouse mammary epithelial NMuMG cells, which form a monolayer under normal growth conditions, but change to a fibroblastic phenotype when treated with TGF-β1 (33). This model was used to further assess the role of FOXO3a in EMT. NMuMG cells, infected with a retrovirus expressing either WT FOXO3a, A3 FOXO3a, or the empty pBabe vector, were treated with 1 ng/mL TGF-β1 for 48 h. The expression of the exogenous FOXO3a was confirmed by probing the whole cell extracts with a FOXO3a antibody (Fig. 4A). The expression of pBabe, WT, or A3 FOXO3a had minimal effects on the appearance of the cells in the absence of TGF-β1 (Fig. 4B,, left), although an ∼20% reduction of growth rate was observed with cells expressing A3 FOXO3a (data not shown). However, when the cells were grown in the presence of TGF-β1, the pBabe-expressing NMuMG cells appeared spindle shaped, a hallmark of mesenchymal phenotype (Fig. 4B,, right). Expression of WT FOXO3a resulted in an increased number of flattened cells, indicative of an epithelial phenotype; whereas almost all the cells expressing A3 FOXO3a appeared flattened and epithelial in morphology (Fig. 4B,, right). The higher potency of A3 FOXO3a in improving the epithelial appearance is consistent with the inability of this FOXO3a mutant to be phosphorylated and thereby inactivated. In particular, treatment of NMuMG cells with TGF-β1 for 48 h resulted in an increase in inactive (phosphorylated) WT FOXO3a (data not shown). Given the ability of A3 FOXO3a to promote a more epithelial phenotype, we assessed its effects on E-cadherin expression. TGF-β1 treatment resulted in a decrease in E-cadherin levels in the pBabe cells (Fig. 4C), consistent with the altered cellular morphology (Fig. 4B). In contrast, E-cadherin expression was higher in cells expressing A3 FOXO3a, and the levels were maintained following TGF-β1 treatment (Fig. 4C). Hence, functional FOXO3a reduces TGF-β1–mediated EMT of NMuMG cells.

FOXO3a induces E-cadherin expression via ERα signaling. We next hypothesized that ectopic expression of FOXO3a in NMuMG cells activates the ERα to E-cadherin pathway. NMuMG cells were transfected with vectors expressing the empty vector pECE, WT, or constitutively active A3 FOXO3a. Whole cell protein extracts were probed for the expression of E-cadherin (Fig. 5A). Expression of both WT and A3 FOXO3a induced E-cadherin protein levels, with the constitutively active A3 FOXO3a again being more effective, consistent with our previous findings, demonstrating that it more effectively induces ERα promoter activity and expression than WT FOXO3a (21). Immunoblotting confirmed ectopic expression of WT, and A3 FOXO3a was essentially equivalent, and Coomassie blue staining of proteins showed equal loading (Fig. 5A). In agreement with the increase in protein levels, RT-PCR showed both WT and A3 FOXO3a increased E-cadherin mRNA levels in transfected cells, with the A3 FOXO3a again leading to a more substantial effect (Fig. 5B,, inset). Next, we tested the ability of the A3 FOXO3a protein to induce E-cadherin promoter activity. Expression of the A3 FOXO3a led to a potent activation of the E-cadherin promoter in both NMuMG (Fig. 5B) and rel-3875 cells (Fig. 5C).

We next asked whether a dominant negative FOXO3a (FOXO DN) can induce mesenchymal gene expression. Untransformed NMuMG cells were infected with either pBabe or FOXO DN, and whole cell extracts were analyzed. The dominant negative FOXO3a reduced E-cadherin and γ-catenin levels and increased the expression of Snail (Fig. 6A). Expression of the dominant negative FOXO3a similarly repressed E-cadherin protein level in rel-3875 and rel/CK2-5839 cells (Fig. 6B). Expression of ERα, a known target of FOXO3a, was potently reduced by expression of the dominant negative FOXO3a, confirming the efficiency of this mutant to block FOXO3a activity (Fig. 6B). Thus, FOXO3a induces a gene expression profile associated with a more epithelial phenotype in rel-3875 and rel/CK2-5839 cells, which express low level of ERα, similar to NF639 cells (ref. 21 and data not shown).

To test the role of ERα in FOXO3a signaling to E-cadherin expression, two approaches were used. We first subjected the prototypical ERα-negative MDA-MB-231 cancer cell line to similar analysis. A3 FOXO3a failed to induce E-cadherin expression in MDA-MB-231 cells (Fig. 6C). Immunoblotting confirmed FOXO3a expression. We next tested whether inhibition of ERα blocks the effects of FOXO3a on EMT in ERα-positive rel/CK2-5839 cells. Ectopic expression of A3 FOXO3a induced the levels of E-cadherin and γ-catenin, as well as ERα, as expected (Fig. 6D). The addition of the antiestrogen ICI 182,780, which inhibits estrogen-dependent ERα function (34) and decreases ERα protein levels (35), blocked the activation of the E-cadherin and γ-catenin genes by FOXO3a (Fig. 6D). Taken together, these data indicate that FOXO3a confers anti-invasive properties in breast cancer cells via an ERα-mediated pathway.

Here, we show for the first time a role for the Forkhead family member FOXO3a in control of EMT of ERα-low breast cancer cells. Recently, we identified two forkhead elements upstream of ERα promoter B and showed that FOXO3a can potently activate ERα gene transcription (21), which was confirmed by Madureira et al. (36). Activation of FOXO3a and resulting enhanced ERα signaling by EGCG promoted a less malignant phenotype of breast cancer cells. ERα signaling has been shown to promote a more epithelial phenotype via inhibiting synthesis of Snail, a repressor of E-cadherin expression (16). ERα can induce synthesis of MTA3, a component of the HDAC NuRD, which inhibits Snail gene transcription. In the presence of active ERα signaling, we noted that EGCG induced a more epithelial gene expression profile, i.e., increased expression of MTA3, E-cadherin, γ-catenin while repressing Snail. Of note, FOXO3a inhibited the formation of invasive colonies in Matrigel, invasive phenotype induced by TGF-β1 signaling, and reduced the ability of cells to migrate. In the absence of ERα signaling, this movement of gene expression toward an epithelial phenotype did not occur. Hence, our studies place FOXO3a upstream of ERα in the MTA3/NuRD pathway that represses the expression Snail and thereby promotes a less invasive, more epithelial phenotype.

The ability of active FOXO3a to functionally promote a less invasive phenotype via ERα signaling defines a novel aspect for FOXO proteins as potential tumor suppressors. FOXO proteins have been shown to induce cell cycle arrest, maintain genome stability, and promote apoptosis largely by inducing p27, GADD45, and BIM gene expression, respectively (27, 37, 38). In breast cancer patients, inactive cytoplasmic FOXO3a expression correlates with poor survival (39). Inhibition of FOXO3a increases tumorigenesis in nude mice (39). Although the overall antitumor effect of FOXO3a is likely mediated through the combination of mechanisms, our data suggest that FOXO3a contributes through the activation of the ERα/MTA3/E-cadherin pathway. Recently, it was reported that FOXM1, another forkhead protein, interacts with FOXO3a and cooperatively regulates ERα expression (36). It remains to be determined how this interaction affects invasive phenotype.

The mechanism we report here extends the known ability of EGCG to promote a more epithelial phenotype via its antioxidative activity. Tumor cells produce large amounts of reactive oxygen species (ROS), which leads to the induction of matrix metalloproteinases that promote invasion and metastasis (40, 41). The antioxidative activity of green tea polyphenols was found to potently inhibit many matrix proteinases (42–44). ROS scavengers have been seen to inhibit FOXO3a induction (45), suggesting ROS is capable of inducing FOXO3a. The observed activation of FOXO3a by EGCG likely occurs via a mechanism that is independent of its antioxidative activity. Interestingly, FOXO proteins have been shown to protect from oxidative damage–induced apoptosis by inducing the antioxidant enzymes MnSOD and catalase (46). Whether induction of MnSOD by FOXO3a contributes to the inhibition of an invasive phenotype remains to be determined.

In summary, our study offers a novel molecular understanding of the observed anticarcinogenic effect of green tea. EGCG possesses potent anti-invasive activity by activating the FOXO3a/ERα/MTA3/E-cadherin pathway. Substantial anticarcinogenic effects have been seen in studies in animal drinking doses comparable to 3 to 10 cups daily (reviewed in ref. 1). Green tea extracts potently reduce (87%) invasive phenotype of mammary tumors induced by DMBA in S-D rats (5). Oral infusion of a polyphenolic fraction isolated from green tea significantly inhibited prostate cancer development and increased survival of TRAMP mice (6). A study in Japan has found that drinking green tea, at the level used in these animal studies, reduced the risk of recurrence in women with stage I and II breast cancer (3). Thus, green tea and EGCG offer attractive antitumorigenic agents that can be easily and safely incorporated into daily lives.

Grant support: NIH PO1 ES11624 and a grant from the Avon Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank M. Greenberg and A. Brunet for providing cloned FOXO DNAs, P. Leder for the NF639 cell line, C. Taylor for assistance with the microarray analysis, and K. Kirsch for generously allowing access to the microscope and camera. We also thank Junfeng Wu for preparing the E-cad-Luc construct.