FOXD3 Is a Mutant B-RAF–Regulated Inhibitor of G₁-S Progression in Melanoma Cells

Forkhead box (FOX) transcription factors represent a closely related family of proteins that mediate cell cycle progression, survival, and differentiation (1, 2). The defining feature of FOX transcription factors is a conserved DNA-binding region known as the FOX or winged helix domain. FoxD3 (formerly Genesis/HFH2) was originally identified due to its expression in embryonic stem (ES) cells (3), and it serves numerous indispensable roles during development (4). Foxd3 is required for maintaining pluripotent cells in the early mouse embryo, and Foxd3 knockout mice die early during mouse development because it functions to maintain cells of the inner cell mass (5), trophoblast progenitors (6), and neural crest cell precursors (7). Use of a conditional mouse knockout model shows a requirement for Foxd3 in the establishment of murine ES cell lines (8). FoxD3 regulates specification of the neural crest lineage. Its expression is temporally downregulated in neural crest cells during the wave of cell emigration that gives rise to melanocytes (9). In vitro and in vivo inhibition of FoxD3 in early-migrating neural crest cells results in an expansion of the melanoblasts at the expense of other lineages. Conversely, overexpression of FoxD3 in late-migrating neural crest cells prevents melanoblast formation in favor of the glial and neuronal lineages (9). Reexpression of FoxD3, as well as other early neural crest markers, has been reported in quail pigment cells that have undergone dedifferentiation in culture (10). However, the exact mechanism by which FoxD3 directs neural crest development remains unclear.

Melanoma, the deadliest form of skin cancer, arises from the transformation of melanocytes. Because treatment options for advanced melanoma remain limited, identifying regulators of aberrant melanoma growth is critical. Recent developments have led to the identification of B-RAF mutations, which hyperactivate the extracellular signal-regulated kinase (ERK) 1/2 pathway, in approximately two thirds of melanomas (11). The ERK1/2 pathway may also be activated through mutations in N-Ras, autocrine growth factor action, or upregulation of G protein–coupled receptors (12, 13), underscoring the importance of ERK1/2 pathway to melanoma progression. Additionally, B-RAF mutations have been identified in ∼30% of thyroid carcinomas and in 14% of colorectal cancer (11). Inhibitors to RAF and mitogen-activated protein/ERK kinase (MEK) are currently being investigated in clinical trials (14).

FOX transcription factors are known to both promote and inhibit the onset and progression of tumors; however, their role in human melanoma is poorly understood. In zebrafish, FoxD3 represses expression of microphthalmia-associated transcription factor (MITF; refs. 15, 16), a regulator of melanocyte development. MITF is amplified in a subset of human melanomas, is regulated by B-RAF signaling (17), and, in some contexts, cooperates with B-RAF^V600E to promote the transformation of human melanocytes (18). Together, with evidence of FoxD3 functioning as a potent antagonist of expansion of the melanoblast lineage, these studies led us to examine the regulation and actions of FOXD3 in human melanoma cells. Here, we show that FOXD3 expression is enhanced following depletion of B-RAF or inhibition of MEK in human mutant B-RAF melanoma cells. Expression of Foxd3 results in a potent inhibition of proliferation in multiple mutant B-RAF melanoma lines by inducing a G₁ arrest. Mechanistically, Foxd3/FOXD3 acts by causing upregulation of the cyclin-dependent kinase (CDK) inhibitor p21^Cip1 and repression of cyclin A. Regulation of p21^Cip1 is p53 dependent, and FOXD3-dependent cell cycle arrest is prevent by p53 depletion and partially by p21^Cip1 depletion. These findings identify FOXD3 as a B-RAF target and novel cell cycle repressor in melanoma and may give insight to the mechanism by which Foxd3 prevents melanoblast differentiation during development.

Human melanoma cell lines WM793, WM115, and WM3211 were kindly donated by Dr. Meenhard Herlyn (Wistar Institute, Philadelphia, PA). SK-MEL-28 and A375 cells were purchased from the American Type Culture Collection. WM793, WM115, WM3211, and SK-MEL-28 cells were cultured in MCDB 153 medium containing 20% Leibovitz L-15 medium, 2% fetal bovine serum (FBS), 0.2% sodium bicarbonate, and 5 μg/mL insulin. A375 cells were cultured in DMEM with 10% FBS. Neonatal foreskins were isolated and cultured, as previously described (19). Thyroid cancer cell lines WRO, NPA, and ARO were maintained in RPMI 1640 supplemented with 10% FBS. NPA and ARO cells harbor mutant B-RAF, and WRO cells are wild-type for B-RAF (20).

U0126 was purchased from Cell Signaling Technology. AZD6244/ARRY-142866 was kindly provided by Dr. Paul Smith (AstraZeneca UK Ltd.; ref. 21).

WM793 and WM115 cells were transfected for 4 h with chemically synthesized short interfering RNAs (siRNA; Dharmacon) at a final concentration of 25 nmol/L using Oligofectamine (Invitrogen). The siRNA sequences for B-RAF #1, p21 #1, cyclin D1 #10, and p53 #5 were ACAGAGACCUCAAGAGUAAUU, GAUGGAACUUCGACUUUGU, ACAACUUCCUGUCCUACUAUU, and GAGGUUGGCUCUGACUGUAUU, respectively. The nontargeting siRNA, UAGCGACUAAACACAUCAAUU, was used as a control.

Cells were lysed and analyzed by Western blotting as previously described (22). Primary antibodies used were ERK1/2 (K-23), B-RAF (F-7), p53 (DO-1), and cyclin A (H-432) from Santa Cruz Biotechnology, Inc.; cyclin D1 (DCS-6) from BD Biosciences; p21 (DCS60) and phospho-ERK1/2 (E10) from Cell Signaling Technology; β-galactosidase (Z378A) from Promega; FOXD3 (Poly6317) from BioLegend; and V5 epitope from Invitrogen. Chemiluminescence was visualized on Fluor-S and Versadoc MultiImagers and quantitated using Quantity One software (Bio-Rad).

pLenti6/TR and pLenti4/TO/V5-GW/LacZ were purchased from Invitrogen. Murine Foxd3 (generously provided by Dr. Robert Hromas, Indiana University Medical Center, Indianapolis, IN) and human FOXD3 were cloned into pENTR/D-TOPO (Invitrogen) and LR recombined into pLenti4/TO/V5-DEST. Expression constructs and packaging plasmids pLP1, pLP2, and pLP/VSVG were cotransfected into HEK293FT cells to generate viral particles. Cells were transduced with particles for 72 h and then selected with blasticidin (pLenti6/TR) or zeocin (pLenti4 constructs), as described previously (22, 23). WM793TR, WM115TR, SK-MEL-28TR, and A375TR cells are clonal isolates selected for high expression of the Tet repressor (TR) and used for sequential transductions. Transgene expression was induced with either tetracycline or doxycycline (0.1 μg/mL). Similar results were obtained with either inducing agent.

DNA oligonucleotides (listed in Supplementary Table S1) were annealed and ligated into pENTR/H1/TO using the manufacturer's kit and protocol. Short hairpin RNA (shRNA) cassettes were recombined into pLenti4/BLOCK-iT-DEST. Lentivirus particles were packaged and used as above.

Total cellular RNA was extracted using the PerfectPure RNA Cultured Cell kit (5 Prime). cDNA was made using the iScript cDNA Synthesis kit (Bio-Rad). Quantitative PCR was performed using iQ SYBR Green supermix (Bio-Rad), 0.8 μmol/L oligonucleotide primers, and 0.1 μg cDNA. The primers used are listed in Supplementary Table S2. Primer specificity was confirmed by melt curve analysis and TAE gel electrophoresis. Reaction conditions were as follows: denaturation at 94°C for 30 s, annealing at 50°C for 30 s, and elongation at 72°C for 30 s (50 cycles in total). PCR was performed on an iCycler with MyiQ version 1.0 software (Bio-Rad). Relative mRNA levels were calculated using the comparative C_t method (ΔC_t; ref. 24).

For cell growth, 2 × 10⁵ cells were plated in complete medium ± inducing agent. Cells were trypsinized, counted, and then replated every 3 d. For growth in soft agar, 2% agar was overlaid with bottom agar (0.3%), which in turn was overlaid with top agar (0.3%) containing 3 × 10³ cells/mL. Polymerized soft agar was overlaid with another layer of acellular soft agar and finally by medium. Doxycycline was added to a final concentration of 0.2 μg/mL. Cells were grown for 14 d, replacing the medium every 3 d. Photos were taken on an Olympus IX70 microscope with Image-Pro Plus software.

Cells were cultured in complete medium containing 10 μmol/L 5-ethynyl-2′-deoxyuridine (EdU) for 8 h. Cells were trypsinized and processed using the Click-iT EdU Flow Cytometry Assay kit and protocol (Invitrogen). Incorporated EdU was stained with Alexa Fluor 647 azide. Cells were analyzed by flow cytometry on a BD FACSCanto.

Cells on coverslips were fixed in 3.7% formaldehyde, permeabilized with 0.5% Triton X-100, and stained with Hoechst 33342 (Molecular Probes, Inc.), anti-RhoGDI (K-21; Santa Cruz Biotechnology), and anti-V5 epitope (Invitrogen). Goat anti-rabbit IgG-Alexa Fluor 594 and anti-mouse-Alexa Fluor 488 (Molecular Probes) were used as secondary antibodies. Photos were taken on an Olympus BX61 microscope with IPLab software.

Analysis was performed using Minitab software. For ANOVA, Tukey's and Dunnett's post hoc tests were performed.

We examined whether B-RAF regulates FOXD3 levels in vertical growth phase (WM793 and WM115) and metastatic melanoma (A375) cells that harbor activating B-RAF mutations (11, 12). In WM793 and WM115 cells, transient siRNA-mediated knockdown of B-RAF led to an inhibition of ERK1/2 activation but increased protein expression of FOXD3 (Fig. 1A). To exclude concerns about off-target effects of RNA interference, we engineered three independent shRNA sequences targeting different regions of the B-RAF mRNA in an inducible system. Efficient B-RAF knockdown with all three shRNAs was associated with upregulation of FOXD3 and inhibition of ERK1/2 (Fig. 1B). We have previously reported that C-RAF and A-RAF expression levels are not altered by B-RAF siRNAs (25), and similarly, no effect on C-RAF and A-RAF was observed in the B-RAF shRNA systems (Supplementary Fig. S1). No effect on FOXD3 expression was observed with nontargeting control siRNA or LacZ shRNA. FOXD3 levels were also enhanced following inducible expression of B-RAF shRNA but not control shRNA in A375 cells (Fig. 1B). Upregulation of FOXD3 occurred at the mRNA level because inducible knockdown of B-RAF led to an 11-fold increase in FOXD3 mRNA levels, as measured by quantitative reverse transcription-PCR (RT-PCR; Fig. 1C). FOXD3 upregulation could be uncoupled from cell cycle progression because cyclin D1 knockdown, which induces a potent G₁ arrest (26), did not lead to upregulation of FOXD3 (Fig. 1D). These data show that mutant B-RAF signaling represses the expression of FOXD3 in human melanoma cells.

Inhibitors to the B-RAF–MEK pathway are in phase I/II trials for melanoma (27). We used the clinical grade MEK inhibitor AZD6244 (21) and, in parallel, the commonly used nonclinical grade MEK inhibitor U0126 (28). Inhibition of MEK-ERK1/2 signaling with either AZD6244 or U0126 upregulated FOXD3 levels in WM793 and A375 cells (Fig. 2A). FOXD3 levels were low/undetectable in human melanocytes. Furthermore, we did not observe upregulation of FOXD3 following AZD6244 treatment of melanocytes (Fig. 2B) and WM3211 cells (Fig. 2C) that are wild-type for B-RAF. Finally, inhibition of MEK in thyroid cancer lines, two of which harbor B-RAF mutations, also failed to upregulate FOXD3 (Fig. 2D). Thus, MEK activity is required to suppress FOXD3 expression selectively in mutant B-RAF melanoma cells.

To determine the role of Foxd3, we engineered inducible expression of V5-tagged mouse Foxd3 in multiple mutant B-RAF lines. As a control, we generated V5-tagged β-galactosidase–expressing cells. Expression of Foxd3 in WM793 cells was induced in response to addition of doxycycline (Fig. 3A) and localized to the nucleus (Fig. 3B). Similarly, inducible expression of Foxd3 in WM115, SK-MEL-28, and A375 cells was confirmed (Supplementary Fig. S2). Noninduced WM793TR/Foxd3 cells grew at a similar rate to noninduced and induced LacZ cells (Fig. 3C, left graph). However, expression of Foxd3 dramatically inhibited cell growth. Expression of Foxd3 also effectively inhibited growth of WM115 cells (Fig. 3C, right graph). Growth in soft agar is a well-established assay for determining the malignant behavior of tumor cell lines. Expression of Foxd3 potently inhibited growth in soft agar of metastatic melanoma cell lines SK-MEL-28 and A375 (Fig. 3D). FOXD3 expression in SK-MEL-28 cells decreased the number of colonies by 93%, whereas LacZ expression caused a 15% decrease. In A375 cells, FOXD3 expression did not alter the number of colonies but rather reduced colony size by an average of 56%. Expression of LacZ in A375 cells did not affect the size of colonies. These data represent a novel finding that Foxd3 is an inhibitor of melanoma cell growth.

To determine whether Foxd3 expression inhibited cell cycle progression, we performed propidium iodide/flow cytometry experiments. Expression of Foxd3 caused a reduction in the percent of S-phase cells and increased the G₀-G₁ population, although not as dramatically as MEK inhibition with AZD6244 (Supplementary Fig. S3A and B). We did not observe an increase in a sub-G₀-G₁ peak or Annexin V staining (data not shown), indicating that expression of Foxd3 does not induce apoptosis. In DNA synthesis labeling experiments, WM793 cell entry into S phase was inhibited by expression of Foxd3 (Fig. 4A). Additionally, expression of cyclin A was reduced in Foxd3-expressing cells (Fig. 4B). We further defined the effects of Foxd3 expression by analyzing the levels of G₁-S cell cycle regulators. Notably, expression of a negative cell cycle regulator, p21^Cip1, was enhanced following expression of Foxd3 (Fig. 4B). Because it is difficult to distinguish cause versus effect from a single time point (72 hours), we performed time course experiments. Upregulation of p21^Cip1 occurred rapidly (24–48 hours) following induction of Foxd3 and preceded the efficient inhibition of cyclin A expression (Fig. 4C). Upregulated p21^Cip1 in FOXD3-expressing cells coimmunoprecipitated with CDK4 and cyclin D1/cyclin D3, in which p21^Cip1 is thought to stabilize complexes (29), but also with CDK2, consistent with it playing an inhibitory role in G₁-S progression (data not shown). Upregulation of p21^Cip1 and downregulation of cyclin A occurred at the mRNA level, as shown by quantitative RT-PCR (Fig. 4D). Together, these data show that expression of Foxd3 inhibits G₁-S progression, which is associated with upregulation of p21^Cip1.

We expanded our analysis on cell cycle regulators to additional melanoma cell lines and to nonmelanoma cell types. Inducible expression of Foxd3 resulted in downregulation of cyclin A and upregulation of p21^Cip1 in WM115, SK-MEL-28, and A375 cells (Fig. 5A). Foxd3 expression did not lead to upregulation of other CDK inhibitors (i.e., p27^Kip1 and p57^Kip2; Supplementary Fig. S4). To determine whether Foxd3 elicited similar effects in nonmelanoma tumor types, we generated inducible Foxd3 lines for MDA-MB-231, a human breast cancer line, and HT-1080, derived from a human fibrosarcoma. Expression of Foxd3 did not upregulate p21^Cip1 levels in either MDA-MB-231 or HT-1080 cells (Fig. 5B). Thus, we observe effects of Foxd3 on p21^Cip1 in melanoma cells but not in two other cell types.

The above experiments were performed using mouse Foxd3. Although the mouse and human forms of this protein are >88% identical, we wanted to confirm our findings with human FOXD3. To this end, we cloned human FOXD3 and expressed it using the inducible lentiviral system as either a nontagged or a V5-tagged protein. Western blotting showed expression of FOXD3 and FOXD3-V5 in WM793 cells in response to doxycycline (Fig. 5C). Importantly, FOXD3 or FOXD3-V5 expression led to upregulation of p21^Cip1 and downregulation of cyclin A. These results are consistent with our findings with mouse Foxd3, although more dramatic effects were seen with the latter likely due to enhanced transgene expression. Levels of cell cycle proteins are regulated by B-RAF–MEK–ERK1/2 signaling in melanoma cells (30, 31). Phospho-ERK1/2 levels were elevated and unaltered following FOXD3 expression in WM793 cells (Fig. 5C).

Because p21^Cip1 is a known cell cycle inhibitor, we tested its role in the FOXD3-induced cell cycle arrest. We also analyzed a role for p53 because it is known to regulate p21^Cip1 expression and is wild-type in the majority of melanoma cells, including WM793 and WM115 (32–34). We used siRNA to knock down either p21^Cip1 or p53 in FOXD3-expressing cells. Notably, FOXD3-induced expression of p21^Cip1 was dependent on p53 (Fig. 6A). Similar to mouse Foxd3, expression of human FOXD3 in WM793 inhibited entry into S phase (Fig. 6B). Transfection with nontargeting siRNA did not affect S-phase entry in either nonexpressing or FOXD3-expressing cells. By contrast, knockdown of p21^Cip1 resulted in a partial restoration of S-phase entry in FOXD3-expressing cells. Interestingly, knockdown of p53 restored S-phase entry to a level that was not statistically different from the noninduced control knockdown. Western blot analysis confirmed that p21^Cip1 or p53 knockdowns were persistent through the course of the experiment, and cyclin A levels in FOXD3-expressing cells were enhanced following p21^Cip1 or p53 knockdown correlating with S-phase entry (data not shown). Additionally, in WM115 cells, knockdown of p53 prevented FOXD3-induced inhibition of cyclin A expression and p21^Cip1 knockdown elicited a partial reversal of cyclin A levels (Fig. 6C). These results suggest a critical role for p53 in FOXD3-dependent p21^Cip1 induction and cell cycle arrest.

In this study, we report novel findings on a stem cell transcription factor, FOXD3, in melanoma cells. We show that expression of FOXD3 is regulated by mutant B-RAF and that FOXD3 is a potent melanoma cell cycle inhibitor. Mechanistically, we show that FOXD3 upregulates the CDK inhibitor p21^Cip1 in a p53-dependent manner and that FOXD3-mediated cell cycle arrest shows dependence on p21^Cip1 and p53.

B-RAF is mutated in approximately two thirds of melanoma. Mutant B-RAF and MEK signaling is required for aberrant cell proliferation (26, 31, 35, 36), but the mechanisms remain poorly understood. Here, we show that depletion of B-RAF expression and inhibition of MEK upregulates FOXD3 expression. ERK1/2 regulation of other forkhead transcription factors has been reported. For example, ERK phosphorylation promotes FOXO3a proteasomal degradation (37) and FOXM1c nuclear translocation (38). We show that mutant B-RAF represses FOXD3 at the mRNA level; thus, the mechanism is distinct from the aforementioned posttranslational modification of forkhead proteins. The promoter region of FOXD3 contains multiple potential SP-1, AP-1, CRE, E-box, and Ets family binding sites, any of which could serve as a portal for ERK1/2-dependent actions. Regulation may also occur via activation of the transcription factor E2F4, which has been shown to bind to the FOXD3 promoter region (39). Although FOXD3 is suppressed by B-RAF and its expression is sufficient to induce a cell cycle arrest, initial experiments show that shRNA-mediated depletion of FOXD3 in melanoma cells does not reverse AZD6244-mediated inhibition of S-phase entry (Supplementary Fig. S5). These data suggest that additional factors likely contribute to cell cycle arrest following B-RAF inhibition. This is not surprising because B-RAF–MEK signaling regulates multiple cell cycle events, including levels of cyclin D1, Cks1/SKP2, and the CDK inhibitor p27^Kip1 (26, 30), as well as the Brn2 and MITF transcription factors, which are required for melanoma proliferation (17, 40). The interplay between FOXD3 and these other cell cycle regulators in melanoma requires further analysis.

Most reports on Foxd3 focus on its role in development because Foxd3^−/− mouse embryos die shortly after implantation due to the loss of progenitor cells needed to form the epiblast and trophoblast (5, 6). We are the first to show a role for FOXD3 as a regulator of aberrant tumor cell proliferation. Such findings are consistent with FOXD3 acting as a suppressor of melanoblasts during development (4). FOXD3-induced cell cycle arrest was associated with upregulation of p21^Cip1 in a manner dependent on p53 in melanoma cells but not in either of the two nonmelanoma cells that we examined. FOXD3 may alter posttranslational modification of p53, which occurs at multiple sites via several distinct mechanisms (41), and/or the recruitment of coactivators and corepressors to p53. Interesting recent findings suggest that FoxD3 may act indirectly by sequestering the transcription factor PAX3 from binding its DNA targets (16), and PAX3 has been shown to control p53-dependent functions in the developing neural tube and in lung carcinoma and mouse fibroblast lines (42, 43). FOXD3-dependent inhibition of S-phase entry and cyclin A expression was prevented by p53 and p21^Cip1 depletion. With regard to p21^Cip1, the effect was incomplete, indicating that additional p53-dependent mechanisms may play a role or that residual p21^Cip1 following the knockdown elicits effects. An additional consequence of FOXD3 expression was upregulation of cyclin D1 (data not shown), although this may represent an effect rather than a cause because cyclin D1 is required for S-phase entry of melanoma cells (30) and, in some experiments, was reversed following p21^Cip1 and p53 knockdown.

There is increasing interest in the role of tumor-initiating cells with stem cell–like properties in melanoma and other malignancies. Recent studies suggest that the ability of individual cells to form tumors may be more widespread in melanoma compared with other malignancies (44). Foxd3 was initially identified as a transcriptional repressor expressed in normal and malignant ES cells (3), and ES cells or teratocarcinoma cells cannot be established from Foxd3 knockout embryos (5). Furthermore, FoxD3 is believed to contribute to the pluripotency and self-renewal of ES cells through a complex negative feedback loop with the transcription factors Oct4 and Nanog (45). An interesting possibility raised from upregulation of FOXD3 following B-RAF–MEK targeting is that melanoma cells are reprogrammed toward a less well-differentiated state following B-RAF inhibition. Because treatment options for melanoma are extremely limited in part due to the high resistance of melanoma to chemotherapy, elucidation of FOXD3 function in subgroups of tumor cells as well as the general tumor mass population may be extremely important.

No potential conflicts of interest were disclosed.

We thank Dr. Paul Smith for providing AZD6244/ARRY-142866, Dr. Meenhard Herlyn for WM melanoma cell lines, Dr. Jeffrey Knauf (Memorial Sloan-Kettering Cancer Center) and Dr. James Fagin (University of Cincinnati) for the thyroid cancer cell lines, Dr. Robert Hromas for providing the Foxd3 cDNA, and Whitney Longmate for technical assistance. We thank Diane Colello (Albany Medical College) for the HT1080TR and MDA-MB-231TR cells. Diane Colello was supported by NIH grant GM51540 to Dr. Susan LaFlamme (Albany Medical College).

Grant Support: American Cancer Society grant RSG-08-03-01-CSM and NIH grants R01-GM067893 and R01-CA125103 (A.E. Aplin). E.V. Abel was supported, in part, by T32-HL-07194.

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.