Oncogene-induced senescence (OIS) is a potent tumor-suppressive mechanism that is thought to come at the cost of aging. The Forkhead box O (FOXO) transcription factors are regulators of life span and tumor suppression. However, whether and how FOXOs function in OIS have been unclear. Here, we show a role for FOXO4 in mediating senescence by the human BRAFV600E oncogene, which arises commonly in melanoma. BRAFV600E signaling through mitogen-activated protein kinase/extracellular signal-regulated kinase kinase resulted in increased reactive oxygen species levels and c-Jun NH2 terminal kinase–mediated activation of FOXO4 via its phosphorylation on Thr223, Ser226, Thr447, and Thr451. BRAFV600E-induced FOXO4 phosphorylation resulted in p21cip1-mediated cell senescence independent of p16ink4a or p27kip1. Importantly, melanocyte-specific activation of BRAFV600E in vivo resulted in the formation of skin nevi expressing Thr223/Ser226-phosphorylated FOXO4 and elevated p21cip1. Together, these findings support a model in which FOXOs mediate a trade-off between cancer and aging. Cancer Res; 70(21); 8526–36. ©2010 AACR.
Activating mutations in the Ser/Thr kinase BRAF are observed in ∼7% of all human tumors with high occurrence in thyroid carcinoma, colorectal cancer, ovarian cancer (1), and especially melanoma (∼70%; ref. 2). The predominant BRAF mutation present in these cases is a substitution of Val600 for Glu (BRAFV600E), which causes increased downstream signaling toward mitogen-activated protein kinase/extracellular signal-regulated kinase (ERK) kinase (MEK; ref. 2). Although BRAF-activating mutations initially stimulate proliferation, cell cycle progression is ultimately arrested through induction of senescence (3–5). Oncogene-induced senescence (OIS) can be facilitated through the individual activities of p16ink4a and p21cip1 (6, 7), and also in case of BRAFV600E, these cell cycle inhibitors are thought to regulate senescence (4, 8, 9).
Reactive oxygen species (ROS) propagate cellular signaling induced by growth factors and thereby regulate a variety of cellular processes including proliferation (10, 11). However, when ROS levels increase above a certain threshold, sometimes called oxidative stress, ROS react with and damage the cellular interior. Additionally, excessive ROS can induce cellular senescence (12), and as such, they are considered to accelerate aging and age-related pathologies (13, 14). ROS are known to signal to a plethora of downstream targets, and it is currently elusive which of these regulate the induction of senescence.
Forkhead box O (FOXO) transcription factors are the mammalian orthologues of the Caenorhabditis elegans protein DAF-16, which functions as an important determinant of life span (15). FOXOs were originally identified as downstream components of insulin/insulin-like growth factor signaling through phosphoinositide 3-kinase (PI3K) and protein kinase B (PKB/AKT; refs. 16, 17). In mice, FOXOs act as functionally redundant tumor suppressors (18), and in cell systems, FOXOs can either mediate apoptosis or quiescence in response to growth factor deprivation (19). In contrast to insulin signaling, which represses FOXO activity, cellular ROS can activate FOXOs (20, 21). Regulation of FOXOs by ROS occurs through numerous posttranslational modifications (22), rendering FOXOs sensors of cellular ROS (23). Consequently, FOXO activation increases resistance to oxidative stress through transcription of enzymes as MnSOD (24) and catalase (25) through a negative feedback loop. Increased FOXO activity is associated with longevity in model organisms (15) and humans (26), which lends credit to the hypothesis that excessive ROS accelerate aging. Thus, FOXOs are regulated by ROS, play a role in both tumor suppression and aging, and thereby provide an important paradigm to understanding the relation between aging and disease such as cancer.
Materials and Methods
Additional information is available in Supplementary Materials and Methods.
The antibodies against FOXO4 (834), HA (12CA5), phosphorylated Thr447, and phosphorylated Thr451 have been described before (21, 27). The following antibodies were purchased: phosphorylated Thr183/Tyr185–c-Jun NH2 terminal kinase (JNK) and phosphorylated Thr202/Tyr204-ERK (Cell Signaling); Thr28–phosphorylated FOXO4 (Upstate); MnSOD (Stressgen); trimethyl-H3K9 and FOXO3a (Upstate); p27kip1 and p21cip1 (BD Pharmingen); p16ink4a (ab-2; Neomarkers); p21cip1 (M19), BRAF (C19), FOXO4 (N19), FOXO1 (N18), proliferating cell nuclear antigen (PCNA; PC10), and p53 (DO-1; Santa Cruz); and tubulin (Sigma). Antibodies against phosphorylated Thr223 and phosphorylated Thr223/Ser226 were generated by immunizing rabbits with the keyhole limpet hemocyanin–conjugated peptides CKAPKKKPSVLPAPPEGA-pT-PTSPVG and CKAPKKKPSVLPAPPEGA-pT-PT-pS-PVG, respectively, wherein pT and pS present phosphorylated threonine and serine. Produced antibodies were subjected to positive and negative affinity purification according to manufacturer's protocol (Covance).
Constructs and RNA interference
The following constructs have been described before: pbabe-puro, pMT2-HA-FOXO4, and pRP261-GST-FOXO4-ΔDB (16); 6× FOXO DNA-binding element (DBE)–firefly luciferase, MnSOD-firefly luciferase, and TK-Renilla luciferase (24); pEFm-BRAFV600E (2); and p21cip1-luciferase (28). pSuper-p21cip1 was a kind gift from Mathijs Voorhoeve (29). A detailed explanation on the generation of HA-FOXO4-4A/E and pSuperior-shFOXO1/3 and 4 is available in Supplementary Materials and Methods. Smartpool oligos against FOXO1, 3a, and 4; BRAF; or scrambled oligos (Dharmacon) were transfected at a final concentration of 100 nmol/L each (300 nmol/L for scrambled) using oligofectamine according to the manufacturer's protocol (Invitrogen).
Immunofluorescence, terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling staining, and BrdUrd incorporation
Immunofluorescence was performed as described (27) using antisera against FOXO4 (834 and monoclonal antibody), HA (12CA5), PCNA, H3K9-Me(III), and pT223/S226. BrdUrd incorporation and terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) staining were performed according to the manufacturer's protocols (Roche). For the mouse sections, anti-p21cip1 M19 was used.
Cellular ROS measurements with H2DCFDA
HEK293T cells were transfected with pcDNA3 or a plasmid encoding BRAFV600E (2 μg), in parallel with pbabe-puro (500 ng). At 16 hours posttransfection, cells were selected with 2 μg/mL puromycin for 36 hours and subsequently left untreated or pretreated for 24 hours with 4 mmol/L N-acetyl cysteine (NAC) or 10 μmol/L U0126, washed with PBS, and incubated for 10 minutes with 1 mL 10 μmol/L H2DCFDA (Invitrogen). Following recovery for 4 hours in medium with or without NAC or U0126, cells were pretreated with or without 200 μmol/L H2O2 for 45 minutes and collected by trypsinization. Centrifuged cells were incubated with 0.02 mg/mL propidium iodide (PI), and live cells were analyzed by FACS for DCF fluorescence. CHL and WM266.4 cells were treated similarly, but without puromycin and PI selection.
Colony formation assay and senescence-associated β-galactosidase staining
A14 or U2OS cells were transfected as indicated together with pbabe-puro (500 ng). At 24 hours posttransfection, cells were subjected to puromycin selection (2 μg/mL). Following 2.5 days of selection, one set of cells was lysed and analyzed by immunoblotting for protein expression. At 10 days posttransfection, cells were fixed in methanol and stained with 0.5% crystal violet in 25% methanol. Plates were dried, and colony formation was quantified by destaining in 10% acetic acid and measuring absorbance at 560 nm. CHL, PMWK, Colo829, and A375 cells were treated similarly but transfected with 500 ng FOXO4 and 250 ng pbabe-puro. Senescence-associated β-galactosidase (SA-β-GAL) staining was performed 9 days posttransfection as described (30).
Ectopic introduction of FOXO4 induces cellular senescence in BRAFV600E-expressing Colo829, A375, and SK-mel28 melanoma cells
To study the involvement of FOXOs in BRAFV600E-dependent cellular responses, we ectopically expressed FOXO4 in the human melanoma-derived cell line, Colo829, harboring an endogenous BRAFV600E mutation. This resulted in reduced colony formation along with diminished PCNA and BrdUrd positivity (Fig. 1A) but without significant TUNEL staining (Supplementary Fig. S1).
FOXOs repress oxidative stress (21), and increased oxidative stress is suggested to cause cellular senescence (12). Surprisingly, however, ectopic FOXO4 expression rendered Colo829 cells positive for SA-β-GAL activity (Fig. 1B). Also detection of two other independent markers of senescence (4, 31), senescence-associated heterochromatin foci (SAHF) and H3K9-trimethylation, was significantly enhanced by FOXO4 (Fig. 1C), suggesting that this indeed is a senescence response.
To exclude artifacts of a single cell type, we also expressed FOXO4 in other melanoma cell lines that express endogenous BRAFV600E, A375, and SK-Mel28, or wild-type BRAF, CHL, and PMWK. Whereas FOXO4 induced SA-β-GAL expression in A375 and SK-Mel28, no positivity was observed in CHL or PMWK cells (Supplementary Fig. S2 and data not shown). Thus, in endogenous BRAFV600E-expressing Colo829, A375, and SK-Mel28 melanoma cells, expression of FOXO4 induces a growth arrest through cellular senescence.
BRAFV600E induces phosphorylation of FOXO4 on JNK target sites
The MEK-ERK pathway is a primary signaling output for normal and oncogenic BRAF. In addition to MEK-ERK signaling, BRAFV600E expression is reported to promote activation of JNK (32) which we confirmed (Supplementary Fig. S3). Previously, we showed that FOXO4 is a JNK target and identified Thr447 and Thr451 through mutation analysis as a subset of the phosphorylated acceptor sites (21). We therefore wondered whether BRAFV600E could signal through JNK toward FOXO4 to promote senescence. To fully address this question, we first determined all possible JNK sites of in vitro phosphorylated FOXO4 by liquid chromatography–tandem mass spectrometry analysis (Supplementary Data). In addition to the previously characterized Thr447 and Thr451, this revealed two novel residues, Thr223 and Ser226 (Supplementary Fig. S4). We generated phosphospecific antisera against these sites, including dually phosphorylated Thr223/Ser226. In vitro phosphorylation by JNK significantly increased detection of wild-type FOXO4 by these respective antisera, especially the newly discovered Thr223 and Ser226, whereas FOXO4-4A in which these residues are mutated to Ala (Fig. 2A) was not detected. This indicates that Thr223, Ser226, Thr447, and Thr451 are JNK-phosphorylated acceptor sites.
Because BRAFV600E signaling induces activation of both ERK and JNK, we next determined whether phosphorylation of the identified sites in cultured cells is mediated by either of these kinases. H2O2, which activates both, indeed resulted in the phosphorylation of Thr223 of FOXO4. Additionally, stimuli that exclusively activate either ERK [12-O-tetradecanoylphorbol-13-acetate (TPA) and epidermal growth factor] or JNK (anisomycin) showed that Thr223 phosphorylation correlates with activation of JNK, not ERK (Fig. 2B; Supplementary Fig. S5).
In agreement with JNK activation, BRAFV600E induced a significant increase in phosphorylation on all JNK sites, but not on the PKB/AKT site Thr28 (Fig. 2C). Furthermore, treatment of cells with the JNK inhibitor SP600125 not only inhibited BRAFV600E-induced JNK autophosphorylation in a dose-dependent manner, but also Thr223 phosphorylation of FOXO4 (Supplementary Fig. S6). Together, these results indicate that BRAFV600E promotes JNK-mediated phosphorylation of FOXO4.
To address whether phosphorylation of FOXO4 on the JNK sites is required for FOXO4 to be able to induce senescence in BRAFV600E-expressing melanoma cells, we expressed the FOXO4-4A mutant next to wild-type FOXO4. FOXO4-4A neither significantly repressed colony formation nor induced SA-β-GAL positivity (Fig. 2D). In contrast, a mutant of FOXO4 that mimics phosphorylation on JNK sites, FOXO4-4E, induced a senescence response similar to wild-type FOXO4 (Fig. 2D). Altogether, these data indicate that FOXO4 is a downstream target of BRAFV600E through JNK-mediated phosphorylation and that phosphorylation on the JNK target sites is required for FOXO4 to promote senescence in response to BRAFV600E.
BRAFV600E signaling elevates cellular ROS levels, which promote FOXO4 phosphorylation by JNK
JNK activity is regulated through a large variety of signaling pathways, and we therefore next addressed the molecular mechanism through which BRAFV600E regulates JNK and thereby FOXO4 activity. Elevations in cellular ROS generated through H2O2 treatment of cells can directly invoke senescence (12), and senescence induction in, for instance, melanocytes has recently been correlated with increased ROS (33). Moreover, OIS can be bypassed by ROS scavenging compounds such as NAC (34, 35). Hence, we investigated the possibility that BRAFV600E signaling affects cellular ROS levels by loading cells with the ROS detecting probe H2DCFDA (DCF). BRAFV600E expression significantly increased cellular ROS levels as detected by DCF fluorescence (Fig. 3A). The BRAFV600E-induced increase in cellular ROS could be further increased by treatment with H2O2 (45 minutes, 200 μmol/L) but was impaired on preincubation with NAC. Downstream signaling through MEK seems at least partially required, because preincubation with the MEK inhibitor U0126 reduced DCF fluorescence. These data indicate that ectopic BRAFV600E expression leads to the generation of cellular ROS through downstream MEK signaling. In agreement herewith, melanoma cells expressing BRAFV600E showed higher basal ROS levels compared with wild-type BRAF-expressing cells (Supplementary Fig. S7). Elevations in ROS are sufficient for phosphorylation of FOXO4 by JNK, as treatment of cells with H2O2 resulted in a time-dependent increase of both JNK activation and Thr223 phosphorylation (Supplementary Fig. S8; Fig. 2B). Moreover, BRAFV600E-mediated JNK activation and FOXO4 phosphorylation were repressed on pretreatment of cells with NAC or U0126 (Fig. 3B).
Prolonged treatment with U0126 induces apoptosis in Colo829 cells (36), making it impossible to interpret the effect of this inhibitor on FOXO4-induced senescence. Therefore, Colo829 cells were treated with NAC to reduce cellular ROS. This resulted in reduced colony formation of Colo829 cells (Fig. 3C), most likely due to the fact that proliferation per se requires low amounts of ROS (14). Importantly, however, NAC impaired the ability of FOXO4 to induce senescence in these cells. Altogether, these data point to a pathway in which BRAFV600E induces FOXO4 phosphorylation by JNK through a MEK-regulated elevation of intracellular ROS, and in line with that, ROS are essential for FOXO4 to induce senescence in the presence of BRAFV600E.
p21cip1 mediates the cell cycle arrest and senescence response by BRAFV600E-FOXO4 signaling
Next, we addressed the mechanism downstream of how FOXO4 promotes BRAFV600E-induced senescence. p27kip1 is an important mediator of FOXO-induced G1 arrest and subsequent quiescence response in the absence of growth factors (19). Therefore, we reasoned a role for p27kip1. FOXO4-induced p27kip1 expression, however, was counteracted rather than enhanced by coexpression of BRAFV600E (Fig. 4A; data not shown). Thus, we conclude that the FOXO4-mediated cell cycle arrest, in response to BRAFV600E signaling, is unlikely to be regulated through p27kip1.
Next, we addressed the importance of another cyclin-dependent kinase (CDK) inhibitor, p16ink4a, which has been implicated in senescence. p16ink4a levels do not seem to increase on FOXO4 and BRAFV600E coexpression (Fig. 4A; data not shown). Also, in Colo829 cells in which FOXO4 induces senescence (Fig. 1), a premature stop mutation is present in the CDKN2A gene resulting in loss of p16ink4a expression (37). These data also argue against involvement of p16ink4a in FOXO4-mediated OIS driven by BRAFV600E.
Because p21cip1 and p16ink4a seem functionally redundant in OIS (6, 7, 9), we next analyzed a role for p21cip1. Interestingly, BRAFV600E cooperated with FOXO4 to induce p21cip1 expression (Fig. 4A), and in correlation with the induction of senescence, FOXO4 expression increased p21cip1 expression in Colo829 cells (Supplementary Fig. S9). Similar effects were observed on p21cip1 mRNA expression determined by quantitative real-time PCR. Moreover, BRAFV600E and FOXO4 expression resulted in a synergistic activation of a luciferase-reporter gene driven by the p21cip1 promoter (Fig. 4B). This level of synergy was also observed using a construct under a different FOXO-responsive promoter (i.e., MnSOD) and a synthetic promoter encompassing six optimal FOXO DBEs (6× DBE; Supplementary Fig. S10), suggesting that the cooperative induction indeed reflects increased FOXO activity.
As HA-FOXO4-4A did not induce senescence in Colo829 cells, whereas HA-FOXO4-4E did, we also determined the ability of these mutants to induce p21cip1 transcription. In line with the lack of senescence induction, HA-FOXO4-4A, but not HA-FOXO4-4E, was significantly less capable of driving p21cip1 transcription (Fig. 4C). These data indicate that BRAFV600E activates FOXO4 through JNK-mediated phosphorylation to promote p21cip1 transcription, which in Colo829 cells correlates with the induction of senescence.
To address to what extent p21cip1 is required for the FOXO4-induced cell cycle arrest and senescence in response to BRAFV600E signaling, we used short hairpin RNA (shRNA)– mediated knockdown of p21cip1. This impaired p21cip1 expression induced by BRAFV600E FOXO4 coexpression (Supplementary Fig. S11). Whereas BRAFV600E and FOXO4 together induced a strong G1 arrest as determined by FACS analysis, this effect was abolished on knockdown of p21cip1 (Fig. 4D). Because p21cip1 expression is elevated in FOXO4-induced senescence in Colo829 cells, we also addressed the effect of p21cip1 knockdown on the induction of senescence. Strikingly, FOXO4 expression did not induce SA-β-GAL staining in Colo829 cells on p21cip1 knockdown (Fig. 4D), indicating that p21cip1 is required in FOXO4-induced senescence in these cells. Altogether, these data show that FOXO4 is a downstream target of BRAFV600E that can facilitate a cell cycle arrest and OIS through regulation of p21cip1.
BRAFV600E regulates p21cip1 expression through MEK and ROS-dependent phosphorylation of FOXOs
Following our observations that suggest BRAFV600E-mediated JNK/FOXO4 activation runs through MEK-ROS signaling, we addressed the involvement of MEK and ROS in the regulation of p21cip1 and cell cycle arrest by BRAFV600E and FOXO4. Pretreatment of cells with either NAC, to reduce ROS (Fig. 5A), or U0126, to inhibit MEK (Fig. 5B), repressed JNK activation by BRAFV600E, phosphorylation of FOXO4 on the JNK target site Thr223, and the cooperative induction of p21cip1. Furthermore, whereas ectopic expression of FOXO4 in Colo829 cells significantly enhanced p21cip1 promoter activity, pretreatment of these cells with U0126 or NAC reduced this effect (Fig. 5C). This shows that JNK-mediated phosphorylation of FOXO4 and the concomitant activation of p21cip1 transcription are dependent on MEK activity and elevations in cellular ROS.
Next, we investigated the role of endogenous FOXOs in signaling from BRAFV600E toward p21cip1 transcription. High ectopic expression of BRAFV600E strongly induced p21cip1 promoter activity (ref. 8; Fig. 5D). This induction was abrogated on shRNA-mediated simultaneous depletion of endogenous FOXO1, 3a, and 4, whereas add-back of a FOXO4 mutant insensitive to shRNA-mediated knockdown (FOXO4-SM) was sufficient to rescue BRAFV600E-induced transactivation of the p21cip1 promoter (Fig. 5D; Supplementary Fig. S12 and S13). Thus, endogenous FOXOs are essential for ectopic BRAFV600E to induce p21cip1 transcription.
Endogenous BRAFV600E regulates FOXO4 phosphorylation and p21cip1 expression in cultured melanoma cells and in vivo
To further investigate the endogenous regulation of FOXO4 by oncogenic BRAF, we used a distinct human melanoma-derived cell line WM266.4 (BRAFV600D; Fig. 6A). WM266.4 cells are tumorigenic yet express very high levels of p21cip1. This, we reasoned, made them suitable to investigate the entire endogenous signaling cascade from oncogenic BRAF toward p21cip1. Like Colo829 and in agreement with hyperactive BRAF signaling, WM266.4 cells expressed a significant amount of active ERK and JNK. As for Colo829 cells, expression of p16ink4a was not detectable in this cell line (Fig. 6A; ref. 38). Small interfering RNA (siRNA)–mediated knockdown of BRAF in WM226.4 cells reduced ERK and JNK activity and, importantly, resulted in diminished p21cip1 expression (Supplementary Fig. S14), arguing that the high p21cip1 level in WM266.4 cells is indeed driven by the oncogenic BRAF. Treatment of WM266.4 cells with U0126 inhibited MEK activity and subsequent JNK activation, indicating that, indeed also in these cells, MEK signaling is essential for JNK activation by oncogenic BRAF (Fig. 6B). Interestingly, next to impaired p21cip1 expression, the U0126-mediated repression of JNK reduced phosphorylation of endogenous FOXO4 on the JNK sites Thr223 + Ser226 and also siRNA-mediated knockdown of endogenous FOXOs reduced the p21cip1 expression (Fig. 6B). U0126 further enhanced this reduction, probably reflecting incomplete knockdown of FOXOs by these siRNAs. Together, these experiments indicate that oncogenic BRAF can regulate p21cip1 expression through phosphorylation of endogenous FOXOs by JNK, confirming the results we obtained in our overexpression studies.
Ultimately, to study the biological relevance of our observations in vivo, we used a Braf+/LSL-V600E; Tyr::CreERT2+/o mouse model, which expresses BRAFV600E in melanocytes off the endogenous Braf gene in a tamoxifen-inducible manner (39). As reported before, activation of BRAFV600E signaling induced melanocytic nevi within the dermis, composed of nests of pigmented epitheloid cells intermingled with whorls of lightly pigmented and amelanotic spindle cells (Supplementary Fig. S15). At the periphery of these melanocytic nevi, we observed multiple patches of darkly pigmented, large polygonal cells, interpreted as neoplastic melanocytes. p21cip1 expression was significantly expressed within these neoplastic melanocytes at the periphery of the BRAFV600E-induced nevi (Fig. 6C), and minor p21cip1 expression was detected in the less pigmented regions of the nevi and within epidermal layers. To investigate endogenous FOXO4 expression in the mouse skin, we developed novel monoclonal antisera. The antisera could immunostain ectopically expressed mouse HA-FOXO4 in Colo829 cells (Supplementary Fig. S16). When applied to the mouse skin sections, the antisera showed expression of endogenous FOXO4 in the mouse skin (Fig. 6C). To determine the phosphorylation status of FOXO4 on the JNK target sites in the BRAFV600E-expressing skin samples, we used pT223/S226 antisera. Detection with the pT223/S226 antisera showed nuclear staining in unstimulated cells, including Colo829 (Supplementary Fig. S17; data not shown). Knockdown of endogenous FOXO4 reduced, although not abolished, the signal, showing the extent of specificity of this antisera for endogenous FOXO4. Importantly, endogenous Thr223/Ser226 phosphorylation of FOXO4 was specifically enriched in the areas of the nevi that also showed p21cip1 staining (Fig. 6C). Thus, in line with the cell culture data, in vivo activation of oncogenic BRAF promotes nevi formation, i.e., senescence in vivo, which harbor phosphorylation of FOXO4 on the JNK target sites Thr223/Ser226 and elevated p21cip1 expression within similar compartments.
Here, we describe a role for FOXO4 in BRAFV600E-induced senescence. BRAFV600E activates FOXO4 through a MEK-ROS-JNK signaling cascade to induce p21cip1 expression and senescence (Fig. 6D). Senescence represents a barrier for tumor formation, and consequently, the melanoma-derived cells we have used de facto have bypassed this barrier. Irrespectively, in cell culture active FOXO reimposes this barrier, suggesting that FOXO inactivation is one of the requirements for senescence bypass. This conclusion is supported by data showing that, in mice, loss of PTEN and consequently reduced FOXO activity synergize with BRAFV600E to induce melanoma (40). Despite limitations in studying senescence in melanoma cell lines in culture, our histochemical analysis of lesions from BRAFV600E mice clearly suggests that in vivo FOXO and p21cip1 indeed function in the senescence response induced by BRAFV600E.
Oncogenes induce senescence through various mechanisms. Although HRAS is an upstream regulator of RAF, HRASG12V expression in primary melanocytes induces senescence through the ER-associated unfolded protein response, whereas oncogenic (B)RAF does not (32). This difference between RAS and RAF is also reflected in mice models in which BRAFV600E induces both melanocyte senescence and melanoma (39), whereas HRASG12V, but not NRASQ61K, induces senescence and only melanoma if combined with loss of tumor suppressors p16ink4a or p19Arf (41). Interestingly, senescence in general, including melanocyte senescence (33), frequently correlates with elevated levels of ROS and OIS can be bypassed by ROS scavenging compounds (34, 35). BRAFV600E chronically increases cellular ROS, which, as we showed, is required for activation of FOXO4, p21cip1 transcription, and subsequent senescence. Together, these and our data suggest that, besides oncogene-specific pathways, increased ROS direct part of the senescence program, which may be more generic. Recently, RAS-induced senescence was shown to require a RAS-dependent negative feedback loop repressing PI3K-PKB/AKT activity (42). As ROS, reduced PKB/AKT activity also activates FOXO, suggesting that activation of FOXO is the general event in senescence rather than the ROS/JNK signaling mechanism. Interestingly, the idea that FOXO activation will be a general component of senescence onset is in agreement with the current notion that the reverse, i.e., FOXO inactivation, represents a general component of tumor onset (18).
Mechanisms of senescence induction also greatly differ between cell types. In cell culture, melanocyte senescence differs from fibroblast senescence (discussed in ref. 43). Human melanocytes deficient in INK4a show an impaired senescence response, but INK4a-deficient human fibroblasts senesce normally. Because a number of families with inherited predisposition to melanoma showed loss of p16ink4a (44, 45), these and other data suggest that INK4a-dependent senescence is especially important in melanocytes. However, loss of p16ink4a is not very common in early-stage melanomas (46), and in oncogenic BRAF-positive human and mouse nevi, examples of cellular senescence in vivo p16ink4a expression are mosaic (47, 48). Also recently, we showed in the Braf+/LSL-V600E; Tyr::CreERT2+/o mouse model that loss of p16ink4a does not affect BRAFV600E-induced nevus formation (39). Furthermore, in these mice, BRAFV600E-induced melanoma showed nuclear p16ink4a staining in agreement with clinical data showing significant nuclear p16ink4a expression in primary melanoma (30%–85%) as well as metastatic melanoma (15%; ref. 48). Thus, although p16ink4a fulfills an important role in the suppression of melanoma progression, it seems not to be essential for establishing senescence (see also ref. 4). Here, we show firstly that, in the absence of p16ink4a, FOXO4 can induce senescence, and secondly, this requires p21cip1. This confirms the earlier suggestion that p21cip1 may facilitate melanocyte senescence in the absence of p16ink4a (5). Thus, in BRAFV600E signaling, p21cip1 and p16ink4a seem to regulate two independent cell cycle inhibitory responses that are functionally redundant to the induction of BRAFV600E-induced senescence.
Besides INK4a, the requirement for FOXO further defines differences between fibroblasts and melanocytes in senescence induction. In contrast to our observations with respect to OIS in melanoma cells and melanocytes in vivo, in fibroblasts loss of FOXO3a rather than activation of FOXO has been implicated in replicative senescence (49). Recently, a differential requirement for FOXO has been suggested in tumor progression (50), and it will be of interest to see whether a similar differential requirement applies to the various stages at which cellular senescence can be induced.
FOXOs function as tumor suppressors (18), and senescence induction by FOXO as shown here provides one mechanism for this function of FOXO. Importantly, although a mechanism of tumor suppression, it is argued that cellular senescence is also causative to organismal aging (51, 52). OIS may therefore represent a trade-off between tumor suppression and life span. Interestingly, both lack of growth factor signaling and increased ROS result in FOXO activation. However, the absence of growth factor signaling can impose a reversible p27kip1-mediated G1 cell cycle arrest and/or quiescence, which may be used to repair, for example, cellular damage (24). In this manner FOXO may positively affect life span and importantly with little cost to the organism. However, in response to BRAFV600E-induced ROS, FOXOs protect against tumorigenesis through induction of senescence, and unlike the former, this protection is not without cost. Our findings underline the pivotal role that FOXOs play in mediating the role of ROS in normal signaling as well as aging, and it will be of interest to see whether, for example, age in return affects the ability of FOXO to mediate senescence.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant Support: The Netherlands Science Organization (NWO, Vici), Dutch Cancer Society (KWF Kankerbestrijding), Center for Biomedical Genetics, Cancer Genomics Center, Cancer Research UK (Ref C107/A10433), and the Harry J Lloyd Charitable Trust.
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.