Mesenchyme Forkhead 1 (FOXC2) plays a key role in metastasis and is associated with aggressive basal-like breast cancers

FOXC2 Is Overexpressed in Metastatic Cancer Cells.

To identify genes capable of programming cancer invasion and metastasis, we performed gene expression profiling by using primary tumors generated by four murine mammary carcinoma cells lines (67NR, 168FARN, 4TO7, and 4T1) that exhibit differing metastatic abilities (31). The properties of these lines are summarized in Fig. 1A and in earlier reports (25, 31). The first of these (67NR) is capable of forming only primary carcinomas, whereas the second (168FARN) can release cells into the blood stream as well; the third (4TO7) can complete the aforementioned steps and generate micrometastases, and the fourth (4T1) can execute all of the steps required for the formation of macroscopic metastases. We performed gene expression profiling (25) and identified 38 genes as being up-regulated and 59 as being down-regulated in 4T1 cells compared with the other three cell lines (67NR, 168FARN, and 4TO7). The up-regulated genes included FOXC2, a winged helix/Forkhead domain-containing transcription factor, which is expressed mainly in mesoderm and mesoderm-derived tissues (32, 33) during embryogenesis and specifies mesenchymal cell fates (34). It is not expressed in any adult tissues except those containing adipocytes (35). We found FOXC2 expression to be up-regulated in the 4T1 cells, the most aggressive of the four mouse mammary tumor cell lines and the only one capable of forming macroscopic lung metastases (Fig. 1 B–D).

We examined FOXC2 expression in a panel of established metastatic and nonmetastatic mouse and human tumor cell lines. Five of six metastatic cell lines expressed FOXC2 protein, whereas only two of seven nonmetastatic lines did so. In addition, expression of FOXC2 was not detected in immortalized normal human mammary epithelial cells (HMLE) (Fig. 1D). These associations suggested that the expression of FOXC2 was often induced during the course of tumor progression and that it might play a causal role in enabling metastatic dissemination.

FOXC2 Is Required for the Metastatic Ability of Breast Cancer Cells.

We set out to determine whether expression of FOXC2 contributes to the metastatic ability of the 4T1 mouse mammary carcinoma cells. To do so, we undertook to suppress FOXC2 expression in these cells by constructing a series of short hairpin RNA (shRNA) oligonucleotides targeting mouse FOXC2 mRNA. These shRNAs were stably expressed in the 4T1 cells by using lentiviral vectors (36). Of these, the FOXC2 shRNA14 vector reduced the level of FOXC2 protein expression by >90% (Fig. 2A).

To gauge the contribution of FOXC2 to primary tumor formation and metastatic dissemination, we injected 4T1 cells expressing FOXC2 shRNA14 into the mouse mammary fat pad and examined the resulting primary tumors and lung nodules 28 days later. The FOXC2 shRNA14-expressing cells formed primary tumors (3.32 ± 0.40 g) similar in size to those arising from 4T1 cells expressing a nonspecific control shRNA (3.66 ± 0.20 g) (Fig. 2B). However, suppression of FOXC2 expression reduced the number of metastatic nodules in the lung by 75% (7 ± 1 versus 29 ± 5; P <0.001; Fig. 2 C and D). A second FOXC2 shRNA (FOXC2 shRNA7) suppressed the expression of endogenous FOXC2 partially (by 50%) and also reduced the number of lung nodules (by 40%) (data not shown). Importantly, the few metastatic nodules that eventually did form in the lungs of mice carrying FOXC2 shRNA-expressing primary tumors retained FOXC2 expression, indicating that they arose from cells within the primary tumors in which FOXC2 expression had never been inhibited (Fig. 2E). These results indicated that expression of FOXC2 is required for the full metastatic ability of the 4T1 mouse mammary carcinoma cells.

We next tested whether ectopic FOXC2 expression in weakly metastatic cells could promote metastatic nodule formation. To do so, we overexpressed FOXC2 in a weakly metastatic EpRas mouse mammary carcinoma cell line (37) (Fig. 3A). Ectopic expression of FOXC2 in EpRas cells did not affect either their proliferation rate in vitro or the growth kinetics of the resulting primary tumors at s.c. sites in nude mice (Fig. 3B). However, the tumors formed by FOXC2-expressing EpRas cells gave rise to 33 ± 6 metastatic nodules per lung compared with the control cells, which formed only 12 ± 1 nodules per lung in nude mice (P < 0.02) (Fig. 3C).

We proceeded to examine the expression level of FOXC2 protein in individual primary tumors derived from the control and EpRas FOXC2 cells. Interestingly, the FOXC2 protein in individual primary tumors correlated directly with the number of metastatic nodules generated by such tumors: primary tumors expressing higher levels of FOXC2 protein formed greater numbers of metastatic nodules in the lung (Fig. 3D). These results show that FOXC2 can contribute significantly to the metastatic potential of the EpRas mammary carcinoma cells.

EMT-Promoting Signals Induce FOXC2 Expression.

Previous work of others (37) had shown that a collaboration of TGF-β1 signals and the activated Ras oncoprotein in the EpRas cells described above causes them to undergo an EMT and thereby promotes their ability to invade and form distant metastases. This result raised the possibility that FOXC2 might operate downstream of TGF-β1 to mediate induction of the EMT program.

Accordingly, we tested whether induction of an EMT in EpRas cells by TGF-β1 treatment affected endogenous FOXC2 expression, which is normally not detected in these cells. Indeed, we found that concomitant with EMT induction, TGF-β1 activated expression of FOXC2 mRNA in EpRas cells (Fig. 4A) by as much as 13-fold. Similarly, treatment of HMLE-immortalized HMLE (25, 38) with TGF-β1 also resulted in an EMT (Fig. 4B Center), and concomitant induction of FOXC2 protein synthesis (Fig. 4C, 12 days). Expression of FOXC2 was first detected after 3 days of TGF-β1 treatment and increased in parallel with acquisition of the mesenchymal morphology. Moreover, withdrawal of TGF-β1 from the cells that had undergone EMT resulted in the reversion of most cells to their original epithelial morphology (Fig. 4B Right) concomitant reversion of FOXC2 protein expression to its basal level of expression (Fig. 4C, 24 days). The slow kinetics of induction of FOXC2 expression by TGF-β1 and subsequent shutdown of FOXC2 expression after removal of TGF-β1 suggests that the FOXC2 gene is not a direct target of TGF-β signaling pathways.

We also found ectopic expression of any one of the three EMT-inducing transcription factors (Twist, Snail, or Goosecoid) led, as anticipated, to an EMT in the HMLE (24, 25) (Fig. 4D) and, interestingly, to induction of both FOXC2 mRNA (Fig. 4E) and protein expression (Fig. 4F). Together, these results demonstrate that FOXC2 expression is induced by a variety of EMT-inducing agents and suggest that FOXC2 is commonly involved in a diverse set of EMT programs, possibly by organizing the mesenchymal state of cells, as suggested by its expression in mesoderm-derived tissues during embryogenesis (32, 33).

FOXC2 Promotes Mesenchymal Differentiation as Part of the EMT Program.

To test whether FOXC2, on its own, is sufficient to induce the EMT program, we ectopically expressed FOXC2 in Maden-Darby canine kidney (MDCK) epithelial cells, which have been widely used to study epithelial cell biology (Fig. 5A). In fact, ectopic FOXC2 expression in these cells caused them to shed their cuboidal epithelial morphology and acquire instead a morphology reminiscent of mesenchymal cells (Fig. 5B).

At the biochemical level, FOXC2 expression elicited expression of mesenchymal markers typically induced during an EMT (Fig. 5 E and F). Significantly, however, the preexisting expression of epithelial cell-specific proteins was only partially reduced (Fig. 5 C and D). Moreover, expression of the mRNA encoding E-cadherin, a central effector of the epithelial cell phenotype, was only partially reduced by the expression of FOXC2 (Fig. 5 G and H). This behavior contrasted starkly with the known behavior of Snail, Slug, SIP1, Goosecoid, and Twist, all of which strongly reduce E-cadherin expression (16, 17, 21, 24, 25, 39). In fact, the E-cadherin protein that continued to be synthesized in the FOXC2-expressing cells was delocalized from the plasma membrane to the cytoplasm, ostensibly the cytosol (Fig. 5I). Hence, unlike other embryonic transcription factors studied to date, FOXC2 is able to effectively program only the second half of the EMT transdifferentiation program, that involving the induction of mesenchymal traits.

We also performed Boyden-chamber transwell assays and determined that ectopic expression of FOXC2 in MDCK cells rendered them to become highly migratory (Fig. 5J) and invasive (Fig. 5J). Cancer cells use secreted matrix metalloproteinases (MMPs) to degrade the nearby extracellular matrix (40). As gauged by ELISA, we found MMP2 and MMP9 to be induced to ≈31 and 26 ng/ml, respectively, in response to the expression of FOXC2 in MDCK cells (Fig. 5K) in comparison with control MDCK cells, which do not express any detectable MMP2 and express very low levels of MMP9 (≈2.5 ng/ml). These observations reinforced the notion that FOXC2 can serve as an organizer of mesenchymal differentiation during an EMT and can potentially program malignancy-associated cellular traits.

FOXC2 Is Specifically Overexpressed in Highly Aggressive Basal-Like Breast Cancers.

We undertook to determine whether FOXC2 expression plays a role in the development of various subtypes of human mammary carcinomas. To do so, we performed immunohistochemistry with a FOXC2-specific antibody on tissue microarrays containing 117 samples of human invasive breast carcinomas of various subtypes and a set of normal human breast tissue sections. In fact, FOXC2 protein was not detected in most cells of normal breast ducts and lobules obtained from reduction mammoplasties, with the exception of a small percentage of basal epithelial cells (Fig. 6A). In stark contrast, FOXC2 expression was readily detectable by immunohistochemistry in 85% of invasive breast carcinomas. Among these tumors, the immunoreactivity pattern ranged from absent (Fig. 6B), to faint cytoplasmic staining (low) in 52% of the cases (Fig. 6C), moderate cytoplasmic staining in 37% (Fig. 6D), to strong cytoplasmic and/or nuclear staining (high) in 10% (Fig. 6 E and F). Further analysis of these data revealed that FOXC2 expression was associated with several adverse prognostic markers, including estrogen receptor (ER) negativity (P < 0.001) and high tumor grade (P < 0.002) (Table 1).

Most interestingly, high levels of FOXC2 expression were associated with the aggressive, basal-like subtype of invasive ductal breast cancers (P < 0.0001) (Fig. 6G and Table 1) (41–46). As shown in Fig. 6G, 44% of basal-like tumors exhibited high levels of FOXC2 expression, whereas only 9% of HER2+ tumors and 2% of the far more common luminal ER+ subtype of tumors showed high expression of FOXC2 protein (Fig. 6G and Table 1). Only a few distinct molecular markers have been identified to date that are uniquely associated with basal-like breast cancers (45, 46). FOXC2 expression therefore may prove to be a valuable diagnostic marker for this subtype.

Cell Cultures.

The mouse mammary tumor cell lines (67NR, 168FARN, 4TO7, and 4T1), human mammary epithelial cell line, immortalized with SV40 large T antigen and catalytic subunit of telomerase (HMLE), and MDCK cell lines were maintained as described (25). MCF7, LoVo, NMuMG, T24, BT474, MDA MB 435, and MDA MB 231 were obtained from the American Type Culture Collection (Manassas, VA) and grown per American Type Culture Collection recommendations. A375P and A375M cells (a gift from Isaiah J. Fidler, M.D. Anderson Cancer Center, Houston, TX) were maintained in MEM supplemented with vitamins, sodium pyruvate, l-glutamine, nonessential amino acid, and 10% FBS. EpRas cells (a gift from Ernst Reichmann, University of Zurich, Zurich, Switzerland) were maintained in DME supplemented with 8% FCS, pen/strep, and buffered Hepes, pH 7.0 and 500 μg/ml G418. HMLEs expressing Snail and Twist have been described (25). For TGF-β1 treatment, the HMLEs were cultured in DME/F12 media (1:1) supplemented with insulin, EGF, hydrocortisone, and 5% calf serum and treated with 5 ng/ml TGF-β1 for the indicated time. The EpRas cells were treated with 1 ng/ml of TGF-β1 for the indicated period.

Plasmids.

The human FOXC2 cDNA was subcloned into the pBabe-Puro vector. The pBp-Snail, pBp-Twist, and pWZL-GSC have been described (24, 25). The shRNA-expressing U6 lentivirus system was provided by Richard Iggo (Swiss Institute for Experimental Cancer Research, Epalinges, Switzerland) (36). The FOXC2 shRNA14-targeting sequence is CCACACGTTTGCAACCCAA; FOXC2 shRNA7-targeting sequence is GGACGAGGCTCTGTCGGAC. The U6 promoter with FOXC2 shRNA14 and FOXC2 shRNA7 insert was subcloned into pSP108-PURO. A control shRNA oligo, which does not match any known mouse coding cDNA, was used as control.

Metastasis Assay.

EpRas-pBp and pBp-FOXC2 cells were generated by infecting EpRas cells with a retroviral vector expressing FOXC2 or the control vector and selected in 6 μg/ml of puromycin. Cells (5 × 10⁵) were injected s.c. into female nude mice. The primary tumor weight was monitored periodically. Five to six weeks later, the mice were killed and analyzed for the number of metastatic nodules in the lung under the dissection microscope. All mouse experiments were performed in accordance with relevant guidelines and regulations. See supporting information (SI) Text for additional methods.

Statistical Analysis.

Data are presented as mean ± SEM. When two groups were compared, we used Student's t test. P < 0.05 was considered significant. Expression of FOXC2 in a subset of human breast cancer samples was analyzed by using the cumulative hypergeometric distribution and χ² test.