The APC/C E3 Ligase Complex Activator FZR1 Restricts BRAF Oncogenic Function

The anaphase-promoting complex/cyclosome (APC/C, also named APC) ubiquitin E3 ligase is essential for cell-cycle progression through the M–G₁ phases, largely by controlling the proteolytic degradation of key cell-cycle regulators, including mitotic cyclins and DNA replication factors (1). Notably, unlike CDC20, which has restricted function in M phase, its close homolog, Fizzy-related protein 1 (FZR1, also named CDH1), associates with the APC core complex in late M phase and early G₁ phase and determines G₁ phase cell-cycle fate decisions (2). During the remainder of the cell cycle, phosphorylation of FZR1 by cyclin-dependent kinases (CDK) abolishes the interaction between FZR1 and the APC core complex, therefore inhibiting APC-dependent functions of FZR1 (3–6). Although the APC-dependent functions of FZR1 have been well documented (7), how APC-free FZR1 participates in various cellular processes is just beginning to be uncovered (8). To this end, we have recently demonstrated an APC-independent function of FZR1 in positive regulation of the ubiquitin E3 ligase activity of SMURF1 to influence osteoblast differentiation (9). However, APC-independent functions of FZR1 in controlling tumorigenesis, as well as in other cellular or tissue contexts, remain largely elusive.

There is mounting evidence indicating a tumor-suppressive role for FZR1 (8). Consistent with this notion, most FZR1 substrates, including mitotic and S-phase cyclins (10), mitotic kinases (11), and DNA replication factors (12), are frequently overexpressed in a wide spectrum of human malignancies (8). Moreover, FZR1 deletions, reduced expression, and mutations are found in various human tumor tissues (13, 14). Furthermore, whereas Fzr1 homozygous deletion results in embryonic lethality, Fzr1 heterozygous mice are more susceptible to developing epithelial tumors (15). In addition, our recent studies revealed a crucial role of APC^FZR1 in controlling melanocyte differentiation and pigmentation (16). However, the molecular mechanisms underlying how loss of FZR1 induces tumorigenesis still remain largely unclear. Hence, it is important to define the major downstream oncogenic signaling pathways that are negatively regulated by the FZR1 tumor suppressor, which will further define the critical role of FZR1 in tumorigenesis.

The RAF family of protein kinases consists of ARAF, BRAF, and CRAF isoforms, which play a central role in driving tumorigenesis through activation of the MEK/ERK oncogenic signaling cascade (17). Notably, although cancer-associated BRAF mutants were found in over 60% of patients with melanoma and thyroid cancer (18), in other types of cancers, BRAF genetic status is largely wild-type (WT; ref. 19). Therefore, it is important to understand mechanistically how BRAF is aberrantly upregulated or hyperactivated in human cancers with WT-BRAF. This valuable information will provide further insights to guide targeted therapy strategies to efficiently treat patients with cancer carrying WT-BRAF.

On the other hand, the BRAF^V600E oncogenic mutation has become a major drug target in developing targeted therapeutics against BRAF^V600E-driven cancers, including melanoma (20, 21). Although anti-BRAF^V600E inhibitors, including vemurafenib (PLX4032; ref. 22) and dabrafenib (23), were approved in treating patients with melanoma harboring this mutation, drug resistance is frequently reported (24), suggesting a limitation of single-agent treatment. Recent clinical trials have adopted combinational strategies by using a BRAF^V600E inhibitor together with another compound targeting either MEK, CDK4/6, PI3K, or HSP90 to increase efficiency and to improve survival (25, 26). However, mechanistically how these combinational therapies suppress tumor growth remains poorly defined. Our findings reported here illustrate an inverse correlation between the ability of FZR1 to suppress BRAF activity and the aggressiveness of tumor developmental stages.

We and others have previously demonstrated that acute depletion of FZR1 in human primary fibroblasts leads to premature senescence (27, 28). Consistent with a critical role for BRAF/ERK signaling in triggering senescence in melanocytes (29), we found that BRAF and phosphorylated ERK (pERK), but not other RAF proteins, were significantly upregulated in FZR1-depleted human primary fibroblasts (Supplementary Fig. S1A), human primary melanocytes (HPM; Fig. 1A) and mouse melanocyte melan-a cells (Fig. 1B). Inhibiting APC^FZR1 using a specific APC inhibitor, Apcin (30), which blocks the FZR1 substrate binding pocket, also resulted in an upregulation of BRAF and its downstream pERK levels (Supplementary Fig. S1B). In keeping with the oscillating nature of APC^FZR1 E3 ligase activity during cell-cycle progression (1), in primary melanocyte-derived melan-a cells, both endogenous BRAF and pERK levels decreased in the G₁ phase where APC^FZR1 is most active (Fig. 1C). Notably, depletion of endogenous Fzr1 in melan-a cells resulted in stabilization of BRAF coupled with elevated pERK across the cell cycle, supporting the notion that the BRAF signaling pathway is negatively regulated by APC^FZR1 in a cell cycle–dependent manner (Fig. 1C and D; Supplementary Fig. S1C). In further support of BRAF being a putative APC^FZR1 downstream ubiquitin substrate, ubiquitination of endogenous BRAF was suppressed in FZR1-depleted melanocytes (Fig. 1E), resulting in an extended half-life of endogenous BRAF (Fig. 1F and G).

Additional depletion of endogenous BRAF in FZR1-knockdown HPM (Fig. 1H) and melan-a cells (Supplementary Fig. S1D) largely suppressed FZR1 depletion–triggered pERK elevation, suggesting that FZR1 inhibits ERK activation primarily through BRAF. Furthermore, depletion of other APC complex subunits, such as CDC27 or APC10, also resulted in BRAF accumulation and ERK activation (Fig. 1I and J). In contrast, depletion of endogenous CDC20 failed to elevate either BRAF or pERK levels (Fig. 1K). Together, these data suggest that in primary cells, FZR1, but not CDC20, negatively regulates BRAF abundance and subsequent ERK activation largely in an APC-dependent manner.

As hyperactivation of the RAF/MEK/ERK signaling cascade has been closely linked to nevi formation composed of senescent melanocytes (29), we sought to investigate whether acute depletion of FZR1 could also result in a similar senescent phenotype in primary melanocytes as it does in human primary fibroblasts (27). Notably, compared with control cells, a marked increase of senescence-associated β-galactosidase (SA-β-gal)–positive cells was observed in FZR1-depleted HPM (Fig. 2A and B) and melan-a cells (Supplementary Fig. S2A-B), indicating an accumulation of senescent cells 14 days after FZR1 depletion. Depletion of FZR1 also led to elevated expression of CDK inhibitors, p16^INK4A (also named CDKN2A), p15^INK4B (also named CDKN2B), and p21^WAF1 (also named CDKN1A) in HPM (Fig. 2C), and p21^WAF1 in melan-a cells (Supplementary Fig. S2C), driving cell-cycle arrest upon FZR1 depletion (Fig. 2D; Supplementary Fig. S2D). As hyperactivation of the RAS/RAF/ERK pathway has been reported to cause premature senescence (31, 32), our results suggest that loss of a negative regulatory mechanism of the BRAF/ERK signaling pathway by FZR1 may lead to the onset of premature senescence. In support of this notion, we found that compared with normal human skin, human nevi, which are composed of largely senescent melanocytes (33), exhibited relatively lower FZR1 expression (ref. 34; Supplementary Fig. S2E).

To gain further insight into FZR1-regulated melanocyte senescence, we found that in FZR1-depleted human primary melanocytes, reintroducing WT-FZR1 largely rescued the senescent phenotype (Fig. 2E and F), leading to escape from the growth arrest phenotype (Supplementary Fig. S2F), in part by suppressing the elevated expression of p16^INK4A and p21^WAF1 that is associated with depleting FZR1 (Fig. 2G). Notably, this effect was not observed in the shFZR1 cells reintroduced with the E3 ligase–deficient ΔC-box-FZR1 mutant (Fig. 2E–G; Supplementary Fig. S2F–G), which is unable to associate with the APC core complex (35). This result indicates an APC-dependent function of FZR1 in controlling BRAF signaling in primary melanocytes. Moreover, additional depletion of BRAF largely reversed the senescence phenotype in FZR1-depleted melanocytes (Fig. 2H–J; Supplementary Fig. S2H). Consistently, the MEK inhibitor PD0325901 (36) could also largely reverse FZR1 depletion–induced senescence (Fig. 2K and L; Supplementary Fig. S2I), in part by suppressing FZR1 loss–induced upregulation of p16^INK4A and p21^WAF1 (Supplementary Fig. S2J). These results suggest that depletion of FZR1 triggers the onset of premature senescence in primary melanocytes mainly through activation of the BRAF/MEK/ERK signaling cascade.

In support of APC^FZR1 as a negative upstream regulator for BRAF, BRAF interacted with FZR1, the substrate recruiting subunit for the APC^FZR1 E3 ligase complex, in cells (Fig. 3A and B; Supplementary Fig. S3A) and in vitro (Fig. 3C). Notably, BRAF was able to specifically interact with FZR1, but not CDC20 (Fig. 3A). Furthermore, FZR1 was coimmunoprecipitated with BRAF, but not CRAF, in cells (Supplementary Fig. S3B). Although FZR1 could also interact with ARAF in an ectopic expression condition, FZR1 only promoted the degradation of BRAF, but not ARAF (Supplementary Fig. S3C), supporting BRAF as a specific APC^FZR1 ubiquitin substrate. Similar to other FZR1 substrates, BRAF specifically bound to the WD40 domain of FZR1 (ref. 37; Supplementary Fig. S3D–S3E). Ectopic expression of FZR1, but not CDC20, led to BRAF downregulation (Fig. 3D), which could be largely abolished by the 26S proteasome inhibitor MG132 (Fig. 3E). However, the E3 ligase activity–deficient mutant ΔC-box-FZR1 failed to promote BRAF degradation (Fig. 3F), indicating that APC^FZR1 controls BRAF abundance probably through ubiquitination-mediated proteolysis.

As most APC^FZR1 substrates contain a destruction box (D-box) motif [RxxLx(2-5)N/D/E; ref. 38], examination of the BRAF sequence revealed four putative D-boxes within its coding region (Fig. 3G; Supplementary Fig. S3F). Deletion of D-box 4 (ΔD4), and to a much lesser extent D-box 3 (ΔD3), but not D-box 1 or D-box 2, conferred a moderate resistance to FZR1-mediated BRAF degradation in cells (Supplementary Fig. S3G), indicating that D-box 4 is the primary degron motif. To minimize the impact of altering amino acids in the BRAF protein, we generated a D-box 4–mutated version of BRAF (D4-RLAA; Fig. 3G). Similar to ΔD4-BRAF, D4-RLAA-BRAF exhibited a noticeable resistance to FZR1-mediated degradation (Fig. 3H), in part due to impaired interaction with FZR1 (Fig. 3I and J).

In addition, we identified a cancer patient–derived mutation within D-box 4 (R671Q, mutation ID COSM159405, cancer.sanger.ac.uk; and Fig. 3G), and further showed that analogous to D4-RLAA, the BRAF^R671Q mutant failed to interact with FZR1 (Fig. 3I and J), which might allow it to escape FZR1-mediated ubiquitination and subsequent degradation to favor tumorigenesis. In keeping with a critical role of D-box 4 in mediating BRAF ubiquitination by APC^FZR1, compared with WT-BRAF, D4-RLAA-BRAF was insensitive to APC^FZR1-promoted ubiquitination (Fig. 3K and L), thereby displaying an extended half-life (Fig. 3M and N) and relatively elevated abundance in the G₁ phase (Supplementary Fig. S3H).

Notably, UV irradiation is a well-characterized risk factor for developing melanoma (39, 40). To this end, previous reports showed that acute UV radiation leads to FZR1 degradation and accumulation of APC^FZR1 substrates (41, 42). Consistently, we found that UV exposure resulted in a reduction of FZR1 protein abundance in melan-a, HBL, and A375 cells (Supplementary Fig. S3I–S3K). Of note, in normal melanocyte melan-a cells, UV-triggered FZR1 downregulation led to the accumulation of both BRAF and PLK1 (Supplementary Fig. S3I). In contrast, in melanoma cells HBL and A375 (Supplementary Fig. S3J–S3K), although FZR1 protein abundance decreased upon UV treatment, BRAF and PLK1 protein levels were largely unaffected. This result indicates that APC^FZR1 might be more active in primary cells to target BRAF for proteolysis. Altogether, these data demonstrate that in the primary melanocyte setting, APC^FZR1 controls the ubiquitination and subsequent degradation of BRAF in a D-box–dependent manner (Supplementary Fig. S3L).

Consistent with results obtained in primary cells (Fig. 1A; Supplementary Fig. S1A), depletion of FZR1 in OVCAR8, HBL, WM3670, U2OS, HEK293, and H1755 cell lines led to pERK upregulation (Fig. 4A–C; Supplementary Fig. S4A–S4C). However, unlike in normal cells, depletion of FZR1 did not significantly elevate BRAF abundance in these cancer or transformed cell lines (Fig. 4A–C; Supplementary Fig. S4A–S4C), which may be in part due to elevated CDK activity that has been shown to inhibit APC^FZR1 E3 ligase activity (3–6). Additionally, compared with well-characterized APC^FZR1 substrates such as PLK1, BRAF displayed reduced binding to both FZR1 and APC10 (Supplementary Fig. S4D–S4E), presumably due to lack of a canonical APC10 binding motif (ref. 43; Supplementary Fig. S4F). Given that APC10 also participates in recognizing APC substrates (ref. 44; Supplementary Fig. S4G), the lack of a strong interaction with APC10 indicates that BRAF might be a relatively weak or distributive rather than processive substrate for APC^FZR1 (45). On the other hand, ARAF, which binds to FZR1 (Supplementary Fig. S3B) but not APC10 (Supplementary Fig. S4H), was resistant to FZR1-mediated degradation (Supplementary Fig. S3C), presumably due to its inaccessibility to the APC core complex. These results suggest that compared with other well-characterized strong APC^FZR1 substrates, such as HSL1, CYCLIN B1, and PLK1, BRAF is a weak APC^FZR1 substrate, likely due to the observation that recruitment of BRAF to the APC core complex for proteolysis is relatively inefficient, in part owing to its weaker interaction with APC10.

Interestingly, further depletion of endogenous BRAF in FZR1 knockdown OVCAR8 cells largely attenuated the upregulation of pERK levels (Fig. 4D), indicating an indispensable role for BRAF in mediating ERK activation upon loss of FZR1. Given that in tumor cells, BRAF is largely refractive to APC^FZR1-mediated degradation, FZR1 might utilize an alternative mechanism to harness BRAF kinase activity, thereby restraining its downstream MEK/ERK oncogenic signaling. In support of this notion, we found that reintroducing WT-FZR1 or the APC binding–deficient ΔC-box-FZR1, but not the non–BRAF-interacting N-terminal-FZR1, could effectively suppress pMEK and pERK in FZR1-depleted OVCAR8 cells (Fig. 4E). Given that the APC core complex is indispensable for FZR1-mediated substrate ubiquitination and degradation (7), these results suggest that FZR1 might govern MEK/ERK activation independent of APC E3 ubiquitin ligase activity.

In keeping with this notion, bacterially purified His-WT- and ΔC-box-FZR1, both of which are APC-free, could efficiently inhibit the phosphorylation of GST-MEK1 by immunopurified BRAF kinase in vitro (Fig. 4F). In addition, depletion of the core APC subunit APC10 failed to influence the BRAF/MEK/ERK signaling cascade in HBL cells (Supplementary Fig. S4I). These results directed our efforts to further define the molecular mechanism by which FZR1 could possibly regulate BRAF activation through an APC-independent manner in the tumor cell setting (Supplementary Fig. S4J), where FZR1 mainly exists in an APC-free mode, in part due to elevated CDK activities that promote phosphorylation of FZR1, blocking its association with the APC core complex (ref. 38; Supplementary Fig. S4K).

BRAF activation is under tight control by numerous mechanisms such as phosphorylation and dimerization (46). As our previous report suggested a scaffolding role for FZR1 in disrupting SMURF1 dimerization independent of its APC E3 ligase activity (9), we next sought to explore whether FZR1 could also regulate BRAF dimerization to control its activation. Notably, ectopic expression of FZR1 abolished BRAF dimerization both in cells (Fig. 4G–J) and in vitro (Fig. 4K), supporting a pivotal role for FZR1 in regulating BRAF dimerization independent of APC. In addition, utilizing gel filtration chromatography, single-molecule kinetic analysis (47, 48), and chemical crosslinking, we further demonstrated a potent role for FZR1 in disrupting BRAF dimerization in vitro (Fig. 4L; Supplementary Fig. S4L) and in cells (Supplementary Fig. S4M). Furthermore, two FZR1 mutants, ΔC-box-FZR1 and ΔFizzy-FZR1 (Supplementary Fig. S2G), both of which are deficient in interacting with the APC core complex (37), could disrupt BRAF dimer as effectively as WT-FZR1 (Fig. 4M). These results advocate a model that FZR1 largely disrupts the BRAF dimerization process in an APC-independent manner.

Unlike WT-BRAF, the dimerization of FZR1 interaction–deficient mutants D4-RLAA-BRAF and R671Q-BRAF (Fig. 3G; Supplementary Fig. 4N and O) could not be disrupted by FZR1 (Fig. 4N). In line with this finding, ectopic expression of FZR1 significantly suppressed pERK levels induced by WT-BRAF, but not CRAF, ARAF (Supplementary Fig. S4P), or the dimerization-deficient BRAF^R509H mutant that is relatively weak in activating ERK (ref. 49; Supplementary Fig. S4Q). More importantly, depletion of FZR1 in T98G cells resulted in a stabilization of pERK levels across the cell cycle without a significant impact on BRAF protein abundance (Fig. 4O). Together, these data reveal a possible mechanism through which FZR1 suppresses BRAF activity by disrupting BRAF dimerization in cancer cells (Fig. 4P).

Examination of the crystal structure of BRAF kinase domain (50) revealed that the D-box 4 motif of BRAF is located on a surface loop region of the C-lobe of BRAF kinase domain (Supplementary Fig. S4R). By docking the BRAF kinase domain onto the WD40 domain of yeast Fzr1 (Supplementary Fig. S4S; ref. 51), we found that R671 of BRAF D-box 4, which extrudes outward the kinase domain, could potentially form hydrogen bonds with acidic residues of Fzr1–WD40 (Supplementary Fig. S4T-U). In addition, Y673 of BRAF D-box 4 might interact with F286 of Fzr1–WD40 by aromatic stacking (Supplementary Fig. S4T). However, different from the Acm1 D-box in the resolved yeast Fzr1–WD40 structure (Supplementary Fig. S4S; ref. 51), the side chain of L674 in the BRAF D-box 4 faces backward from Fzr1–WD40, indicating that to properly insert the BRAF D-box into the D-box binding pocket on the Fzr1–WD40, a conformational change on BRAF might be required. This hypothesis might partly explain why FZR1 could disrupt BRAF dimerization while the D-box 4 motif is not located within the BRAF dimerization interface.

In further support of this notion, the D-box 4 motif of BRAF that mediates interaction with FZR1 does not overlap with the BRAF–BRAF interacting surface (Supplementary Fig. S4R), indicating a possible mechanism that FZR1 interferes with BRAF dimerization through an interaction-induced conformational change, i.e., FZR1 might function as an allosteric inhibitor for BRAF dimerization (52). In line with this notion, the constitutive dimerization of the BRAF^E586K mutant (53) could still be disrupted by FZR1 (Supplementary Fig. S4V–S4W). This observation thereby suggests an allosteric rather than a competitive regulation of BRAF dimerization by FZR1, which warrants further in-depth investigation.

In addition to V600E, many other BRAF oncogenic mutations or splicing variants have been identified in human cancers (49, 50, 54). Among them, mutations in the P-loop of the BRAF kinase domain (residues 464–469) have been categorized as dimerization-dependent mutations (21). In contrast to BRAF^V600E-expressing A375 cells, depletion of FZR1 in BRAF^G469-mutated melanoma and lung cancer cell lines led to an elevation of BRAF activity (Fig. 4C; Supplementary Fig. S4C), supporting a critical function for FZR1 in disrupting BRAF dimers (Supplementary Fig. S4X–S4Y).

It has been previously shown that RAF proteins utilize either homodimerization or heterodimerization to prime their activation (46, 55). However, the oncogenic BRAF^V600E mutation at the activation segment of the BRAF kinase domain adopts an altered constitutive active conformation, which allows for full activation of the kinase without dimerization (24). Consistent with the dimerization-independent activation of BRAF^V600E (24), we found that unlike its role in suppressing WT-BRAF activity, FZR1 failed to suppress pERK levels when coexpressed with BRAF^V600E in vivo (Supplementary Fig. S5A) or in an in vitro kinase assay (Supplementary Fig. S5B). Furthermore, unlike depleting FZR1 in melanoma cells harboring WT-BRAF (Fig. 4B), depletion of FZR1 in BRAF^V600E melanoma cells did not induce ERK activation (Supplementary Fig. S5C), although FZR1 could still efficiently bind to and disrupt dimer formation of the BRAF^V600E mutant (Supplementary Fig. S5D–S5E).

Although monomeric BRAF^V600E is fully active, recent studies have demonstrated critical roles for BRAF^V600E dimerization in contributing to vemurafenib (PLX4032) resistance (49, 56). For instance, several BRAF splicing variants have been identified in vemurafenib-resistant melanoma cells that lack exons encoding the RAS-binding domain. As such, these BRAF^V600E variants exhibit increased dimerization in the absence of RAS binding to confer PLX4032 resistance (49). Additionally, vemurafenib has been found to induce the transactivation of WT-BRAF or CRAF through heterodimerization between vemurafenib-bound BRAF^V600E and WT-BRAF or CRAF (53, 56). These studies suggest that further understanding of the molecular mechanisms governing BRAF^V600E dimerization could have great clinical significance in overcoming PLX4032 resistance.

Intriguingly, we found that in keeping with its ability to disrupt WT-BRAF or BRAF^V600E homodimers, FZR1 could also inhibit the heterodimerization of BRAF^V600E with CRAF (Supplementary Fig. S5F) or with ARAF (Supplementary Fig. S5G). Moreover, depletion of FZR1 in A375 and HBL cells resulted in increased BRAF–CRAF heterodimerization (Supplementary Fig. S5H–S5I). This finding prompted us to further explore the role of FZR1 in regulating vemurafenib resistance in BRAF^V600E melanoma cells. Notably, in A375 melanoma cells harboring homozygous BRAF^V600E, depletion of FZR1 led to a moderate resistance to PLX4032 treatment as evidenced by elevated pMEK and pERK signals (Supplementary Fig. S5J) as well as increased cell viability under drug challenge (Supplementary Fig. S5K). These findings therefore indicate that FZR1 loss or reduced FZR1 abundance might contribute to vemurafenib resistance in patients with melanoma.

In contrast to WT-BRAF–expressing cancer cells (Fig. 4A), in the A375 melanoma cell line harboring the BRAF^V600E oncogenic mutation, protein levels of most APC^FZR1 substrates examined, including PLK1, AURORA A, GEMININ, CYCLIN B and CDC20, were barely affected by depletion of endogenous FZR1 (Supplementary Fig. S5C), suggesting that in these BRAF-hyperactive melanoma cells, the ubiquitin E3 ligase activity of APC^FZR1 is largely attenuated.

We and others have previously pinpointed serine/threonine residues within the N-terminal domain of FZR1 as CYCLIN A2 (also named CCNA2)/CDK2 or CYCLIN E1 (also named CCNE1)/CDK2 target sites (Fig. 5A; refs. 3–6), phosphorylation of which abolishes interaction between FZR1 and the APC core complex (3, 43). Because elevation of RAS/RAF/MEK/ERK and its downstream CYCLIN D1 (also named CCND1)/CDK4 signaling pathways has been found in most patients with melanoma (57), we next sought to examine whether these CDK2 target sites within FZR1 could also be phosphorylated by ERK and/or CYCLIN D1/CDK4. Notably, in vitro kinase assays revealed that purified ERK or CYCLIN D1/CDK4 could phosphorylate WT-FZR1, but not 4A-FZR1 or 6A-FZR1, in which the previous identified CDK2 sites were mutated to alanines (Fig. 5B and C). Our finding is consistent with a recent report that FZR1 could be phosphorylated by CDK4 at its N-terminus (58).

To further evaluate the function of this phosphorylation on FZR1, compared with empty vector (EV)–infected parental cells, increased phosphorylation of FZR1 (5) and elevation of various FZR1 substrates were observed in BRAF^V600E-expressing IHPM (Fig. 5D) and melan-a cells (Fig. 5E), indicating that in BRAF^V600E-expressing cells, increased phosphorylation of FZR1 might lead to inactivation of its E3 ligase activity. In support of this notion, depletion of BRAF in A375 (Fig. 5F) and HBL cells (Fig. 5G) led to a significant reduction of FZR1 phosphorylation as well as steady-state level of APC^FZR1 substrates, including PLK1 and CDC6. Furthermore, induced expression of CYCLIN D1 led to an elevation in protein abundance of known APC^FZR1 substrates (Fig. 5H), whereas depletion of CYCLIN D1 gradually suppressed the levels of various APC^FZR1 substrates, including BRAF (Fig. 5I).

To further determine a causal relationship between hyperactivation of ERK and/or CDK4 in BRAF^V600E-expressing cells and phosphorylation-dependent inactivation of APC^FZR1 (Fig. 5A), we demonstrated that compared with EV-FZR1– or WT-FZR1–expressing cells, ectopic expression of the nonphosphorylatable 6A-FZR1 mutant led to a more dramatic decrease in APC^FZR1 substrates such as PLK1 and, to a lesser extent, BRAF (Fig. 5J–L). Furthermore, the number of senescent cells was decreased when introducing 6A-FZR1 in BRAF^V600E-expressing melan-a cells (Supplementary Fig. S5L-M), suggesting that the nonphosphorylatable FZR1 mutants were more efficient in suppressing BRAF^V600E-mediated premature senescence (29). These findings indicate that 4A-FZR1 or 6A-FZR1, which are deficient in ERK- or CDK-mediated phosphorylation (Fig. 5B and C), display elevated APC^FZR1 E3 ligase activity even in BRAF^V600E-expressing cells with hyperactive ERK, presumably due to their higher affinity toward the APC core complex regardless of CDK and ERK activation status (43).

Interestingly, although N-terminal phosphorylation suppressed the E3 ligase activity of FZR1 (8), it barely affected FZR1′s function in disrupting BRAF dimerization (Supplementary Fig. S5N), a process that mainly requires the WD40 domain of FZR1 to bind BRAF (Supplementary Fig. S3D–S3E). However, how APC^FZR1-mediated BRAF ubiquitination is suppressed in different cancer settings warrants further in-depth studies.

Given the pivotal role of the BRAF and CYCLIN D1/CDK4 signaling pathways in driving melanomagenesis (57), inhibitors targeting BRAF^V600E, MEK, or CDK4/6 are either approved or under clinical trials (59), which include the combinational treatment of patients with melanoma with a CDK4/6 inhibitor (LEE011) and a BRAF inhibitor (LGX818; ClinicalTrials.gov identifiers: 01777776 and 01820364). However, mechanistic studies to reveal the benefit of CDK4/6 inhibitors are largely absent. As our findings pinpointed both ERK and CYCLIN D1/CDK4 as FZR1 upstream kinases to inhibit APC^FZR1 E3 ligase activity, we next challenged multiple melanoma cells with the BRAF^V600E inhibitor PLX4032 (60), the MEK inhibitor PD0325901, the pan-CDK inhibitor mimosine (61), or the CDK4/6 inhibitor PD0332991 (62). Consistent with frequent aberrancies in CDK4, but not CDK2, signaling in the melanoma disease setting, we found that a CDK4/6 inhibitor is more potent than a pan-CDK inhibitor, which mainly targets CDK2, in restoring APC^FZR1 activity in both BRAF^V600E and WT-BRAF genetic backgrounds (Fig. 6A and B; Supplementary Fig. S6A–S6B). Furthermore, we also observed a cooperative role for PLX4032 and PD0332991 in reactivating APC^FZR1 E3 ligase activity in BRAF^V600E-expressing cells, as evidenced by the downregulation of various APC^FZR1 substrates, including BRAF (Fig. 6A; Supplementary Fig. S6A–S6B).

In supporting our hypothesis that inhibition of BRAF/MEK and/or CDK activities in cancer cells reactivates APC^FZR1, we further found that although depletion of FZR1 did not affect BRAF or CDC6 protein abundance in A375 cells, further PLX4032 treatment resulted in an apparent upregulation of BRAF and CDC6 levels compared with PLX4032-treated shScr-A375 cells (Fig. 6C). Similarly, in WT-BRAF–expressing OVCAR8 cells, combinational treatment of PD0325901 with mimosine led to an increase in BRAF protein levels in shFZR1-OVCAR8 cells compared with control cells (Supplementary Fig. S6C).

Moreover, BRAF^V600E or MEK inhibitor treatment led to an increase in ubiquitinated species of endogenous CDC20, a well-characterized APC^FZR1 substrate (ref. 63; Supplementary Fig. S6D–S6E), supporting the notion that the observed downregulation of various APC^FZR1 substrates upon BRAF^V600E or MEK inhibitor treatment is at least in part through restoring APC^FZR1 E3 ligase activity. Although CDK4/6 inhibitor treatment of melanoma cells led to an enrichment in G₁ phase, MEK or BRAF^V600E inhibitor has a minimal effect on cell-cycle profiles of various melanoma cell lines we examined (Supplementary Fig. S6F–S6H). These results indicate that the observed downregulation of APC^FZR1 substrates upon MEK or BRAF^V600E inhibitor treatment (Fig. 6A and B; Supplementary Fig. S6A–S6B) are mainly through restoring APC^FZR1 activity rather than simply altering the cell-cycle progression status. Moreover, the elevated APC^FZR1 activity might be partly due to the increased interaction between FZR1 and the APC core complex (Fig. 6D; Supplementary Fig. S6I).

To gain further genetic insight into the critical role of APC^FZR1 deficiency in promoting melanoma development, we found that in the Skin Cutaneous Melanoma (The Cancer Genome Atlas, Provisional) dataset with 287 cases (cbioportal.org; refs. 64, 65), there is a 23.0% mutation or deletion rate for FZR1 and 14 APC subunits (Supplementary Table S1 and Supplementary Fig. S6J), 13% mutation or amplification rate for CYCLIN D1 and CDK4, and 80% mutation or amplification rate for BRAF and NRAS (Supplementary Fig. S6J). Notably, the genetic alterations of the FZR1 and APC subunits do not significantly overlap with the genetic alterations of CCND1 and CDK4 (Supplementary Fig. S6J), whereas the wild-type BRAF/NRAS genetic status displays a significant association with the amplification of CCND1/CDK4 (χ² = 6.877, P < 0.01 by Pearson χ² test). Consistent with the genetic analysis, we found that melanoma-derived FZR1 mutants displayed a reduced binding with BRAF and subsequently attenuated activity in promoting BRAF ubiquitination (Supplementary Fig. S6K–S6M). These analyses suggest that the genetic inactivation of APC^FZR1 via mutation or deletion, together with phosphorylation-dependent inactivation of APC^FZR1, leads to the deficiency of APC^FZR1 in a vast majority of patients with melanoma, which in turn elevates a cohort of oncogenic substrates of APC^FZR1 to drive melanomagenesis.

Having pinpointed that FZR1 deficiency leads to hyperactivation of the BRAF/ERK signaling pathway via either an APC-dependent or APC-independent mechanism in different cellular contexts (Fig. 4P), we next sought to explore whether depletion of FZR1 could promote melanomagenesis, in which BRAF activation plays a pivotal role (66). To this end, we took advantage of a widely utilized in vitro melanomagenesis model (67) to evaluate the contribution of FZR1 deficiency in transforming melanocytes.

Activation of both ERK and AKT signaling pathways has been shown to facilitate melanoma development (68, 69). Notably, simultaneous depletion of FZR1 and PTEN in IHPM cells led to the upregulation of both ERK and AKT oncogenic signaling (Fig. 7A). Furthermore, consistent with previous reports (67, 70), complete withdrawal of growth factors such as TPA abolished the proliferation of the parental IHPM cells (Fig. 7B and C), whereas codepletion of FZR1 and PTEN conferred growth factor–independent growth to IHPM cells (Fig. 7B and C; Supplementary Fig. S7A). Moreover, codepletion of FZR1 and PTEN favored anchorage-independent growth of IHPM cells in soft agar (Supplementary Fig. S7B). These results support the synergetic role of FZR1 depletion–induced ERK activation and PTEN loss–induced AKT activation in transforming primary melanocytes to drive melanomagenesis in vitro.

Although complete deletion of both Fzr1 (Fzr1^−/−) alleles in mice led to embryonic lethality (71), mouse embryonic fibroblasts (MEF) obtained from Fzr1^−/− mouse embryos displayed elevated BRAF and pERK levels compared with WT-MEFs (Fig. 7D; Supplementary Fig. S7B–S7C), which further strengthens our finding that BRAF is an APC^FZR1 substrate. Moreover, in keeping with the fact that the RAF/MEK/ERK signaling pathway is rapidly stimulated by growth factor treatment, we found that, compared with the WT-MEFs, Fzr1^−/− MEFs were more responsive to EGF treatment and displayed extended activation kinetics (Fig. 7D).

Given that the BRAF/MEK/ERK oncogenic pathway plays a pivotal role in governing melanocyte proliferation and differentiation, as well as melanomagenesis (17), to further examine the physiologic role of the FZR1–BRAF signaling axis in the melanocyte setting, melanocyte-specific Fzr1 and Pten conditional knockout mice (Tyr::CreER;Pten^lox/lox;Fzr1^lox/lox; refs. 68, 72) were generated. Notably, we found that topical application of 4-hydroxytamoxifen (4-OHT) on one side of the mouse flank (ethanol was used on the other side as a negative control), which induced the expression of the Cre-recombinase specifically in melanocytes to delete endogenous Fzr1 and Pten, led to a marked increase of BRAF and its downstream signals as evidenced by both immunoblot (Fig. 7E) and immunohistochemistry (Fig. 7F; Supplementary Fig. S7D) analyses.

More importantly, we found that the elevation of both pAKT and pERK mainly occurred near hair follicles (Fig. 7F; Supplementary Fig. S7D), where most melanocytes reside in the postnatal mice (26). This observation is in agreement with previous studies demonstrating that Tyr::CreER^T2 mainly targets melanoblasts and melanocytes within hair follicles (73), which might be the major reason accounting for the phenotype of pigment observed in 4-OHT–treated mouse skin (Supplementary Fig. S7E). Although it will be critical to further determine whether the elevated BRAF and AKT signaling could eventually drive melanoma development after long-term 4-OHT treatment, these in vivo mouse genetic data, however, coherently demonstrate that FZR1 might function as a tumor suppressor in vivo in part by suppressing the activation of the BRAF/MEK/ERK oncogenic pathway. As such, loss of the FZR1 tumor suppressor synergizes with PTEN deficiency, leading to a concomitant elevation of AKT and ERK signaling in vivo.

Deregulation of the RAS/RAF/MEK/ERK oncogenic signaling cascade is considered a hallmark for driving tumorigenesis in various human cancers (19). However, how different isoforms of RAF proteins are restrained from becoming hyperactive in normal tissue and how these negative regulations are attenuated during cancer progression still remain largely undefined. To date, several proteins have been identified to negatively regulate this important signaling pathway. For instance, binding of 14-3-3 to phospho-serines at both the N-terminus and the C-terminus of RAF proteins locked their closed inhibitory conformation (74), and phosphorylation of CRAF by PKA at Ser43 inhibited CRAF activation (75). Ubiquitination-mediated proteolysis has also been shown to negatively regulate RAF kinases. SEL-10, a C. elegans homolog of human FBW7, controls the turnover of LIN-45, the C. elegans homolog of RAF kinases (76). Furthermore, in human cells, ring finger protein 149 (RNF149) targets RAF proteins for ubiquitination and subsequent destruction (77). These studies suggest that RAF/MEK signaling is tightly regulated through various posttranslational modifications, including phosphorylation, ubiquitination, and scaffolding.

Our findings here have demonstrated that FZR1, which is a key cell-cycle regulator with a potential tumor suppressor role (8), could negatively regulate BRAF protein abundance and kinase activation through both APC-dependent BRAF proteolysis and APC-independent disruption of BRAF dimerization. Importantly, we have found that the mechanisms by which FZR1 regulates BRAF are different in normal cells and cancer cells (Supplementary Fig. S7F). In nontransformed cells, such as human primary fibroblasts, melanocytes, or MEFs, APC^FZR1 governs BRAF proteolysis in a cell cycle– and D-box–dependent manner, which retains BRAF abundance to prevent hyperactivation of MEK and ERK downstream of BRAF. Our finding therefore identifies APC^FZR1 as an upstream regulator of BRAF abundance in human somatic cells, serving to suppress unscheduled cell proliferation, an important step toward neoplasia.

In cancer cells, however, our data suggest that APC^FZR1-mediated BRAF degradation was largely attenuated (Fig. 4A–C; Supplementary Fig. S4A–S4C), which might be attributed to several mechanisms that lead to reduced APC^FZR1 E3 ligase activity as well as escape from APC^FZR1-mediated ubiquitination of various known APC^FZR1 substrates. Previous studies have revealed that phosphorylation of FZR1 by CYCLIN A/E-CDK2 in late G₁–S phases or by CYCLIN B1/CDK1 in G₂–M phases dissociates FZR1 from the APC core complex (1) to partly inactivate APC^FZR1 in cancer cells. Furthermore, numerous reports have identified that phosphorylation could disrupt the interaction between FZR1 and its substrates, hence facilitating the escape from APC^FZR1-mediated ubiquitination and degradation in cancer cells (78–80). Notably, inhibiting MEK/ERK signaling by PLX4032 or PD0325901, or suppressing CDK activity by mimosine or PD0332991, could repress BRAF abundance in various cancer cell lines (Fig. 6A and B; Supplementary Fig. S6A–S6B).

Recent structural and biochemical studies have shed light on the molecular mechanisms by which APC catalyzes polyubiquitination of its substrates (81). It has been demonstrated that polyubiquitination of substrates by APC displays distinct processivities due to different binding affinities between FZR1 and its substrates (45). In support of these reports, compared with well-characterized APC^FZR1 substrate PLK1, BRAF displayed attenuated affinity to both FZR1 and APC10, both of which are required for substrate interaction (Supplementary Fig. S4D–S4E; ref. 44), presumably due to the lack of the APC10 interacting motif within the D-box 4 degron of BRAF (Supplementary Fig. S4F–S4G; ref. 44). In support of this notion, a replacement of BRAF D-box 4 (RGYLSPDLSK) with HSL1 D-box sequence (RAALSDITN; termed Opti-D4; Supplementary Fig. S4F) resulted in an increased interaction between APC10 and the generated BRAF chimera protein (Supplementary Fig. S4N), leading to more efficient degradation of the chimera protein by APC^FZR1 (Supplementary Fig. S4O).

Although APC^FZR1 displays marginal effect in controlling BRAF turnover in cancer cells, our studies have revealed that APC-free FZR1 in WT-BRAF cancer cells could still suppress BRAF/ERK signaling, largely by disrupting BRAF dimerization without affecting BRAF protein abundance. This finding further advocates for the tumor-suppressive role of FZR1 toward BRAF, especially when it is largely APC-free in most cancer cells. FZR1-mediated disruption of BRAF dimerization was also observed in immortalized MEFs (Fig. 4J), in which deletion of FZR1 led to the accumulation of BRAF (Fig. 7D). These results suggest a model that FZR1 might harness BRAF oncogenic function via both APC-dependent and APC-independent mechanisms in normal cells, similar to the reported dual suppressive role of SOCS proteins by inhibiting the JAK kinases and targeting them for degradation (82).

Taken together, our findings reveal a reciprocal suppression mechanism between FZR1 and BRAF in controlling tumorigenesis, and further suggest that FZR1 might exert its tumor suppressor role in part by functioning as an upstream inhibitor of the BRAF oncogenic signaling pathway (Fig. 4P; Supplementary Fig. S7F).

Cell lines and their culturing conditions, plasmids, and antibodies, and experimental procedures for cell synchronization and FACS analysis can be found in the Extended Methods section of the Supplementary Information available online.

HeLa, HEK293, HEK293T, U2OS, and T98G cells were cultured in DMEM containing 10% FBS (HyClone), 100 mg/mL penicillin–streptomycin as described previously (9), which were obtained from Dr. William G. Kaelin, Jr. (Dana-Farber Cancer Institute), in June 2006. The OVCAR8 ovarian cancer cell line was obtained from Dr. Marc W. Kirschner (Harvard Medical School) in March 2011. WM266.4, SK-MEL-28, A375, B16 melanoma cell lines, HPMs, mouse primary melanocytes (melan-a), and hTERT/p53DD/R24C-CDK4 immortalized human melanocytes (IHPM) have been described previously (83). HPM cells were isolated from normal discarded foreskins, as described previously (84), and were cultured in Medium 254 (Gibco). Melan-a cells were obtained from the Wellcome Trust Functional Genomics Cell Bank at University of London in June 2013. IHPM cells were engineered with stable expression of human telomerase reverse transcriptase (hTERT), dominant-negative p53 (p53DD), and a constitutively active CDK4 mutant (R24C-CDK4), allowing the bypass of premature senescence under oncogenic stresses (67), and were obtained from Dr. Hans Widlund in September 2012. A375 and HBL melanoma cell lines have been described previously (85), and were obtained from Dr. David Fisher in June 2008. WM266.4 and SK-MEL-28 cells were obtained from Dr. Richard Marais in August 2012. The WM3670 melanoma cell line was obtained from Dr. Keiran S. Smalley in November 2016. The H1755 NSCLC cell line was obtained from Dr. Eric B. Haura in November 2016. All cell lines were obtained between 2006 and 2016 and routinely tested negative for Mycoplasma. Cell line authentication was not routinely performed.

Cell culture transfection, lentiviral shRNA virus packaging, and subsequent infection of various cell lines were performed according to the protocol described previously (86).

The purposes of using different cell lines can be found in the Extended Methods section of the Supplementary Information available online.

Lentivirus-infected human primary melanocytes or murine melanocytes melan-a were subjected to SA-β-gal staining, BrdUrd labeling or crystal violet staining assays 14 days after viral infection. The experimental procedures for BrdUrd labeling and SA-β-Gal and crystal violet staining were described previously (83, 87).

Cells were lysed in EBC buffer (50 mmol/L Tris pH 7.5, 120 mmol/L NaCl, 0.5% NP-40) supplemented with protease inhibitors (Complete Mini, Roche) and phosphatase inhibitors (phosphatase inhibitor cocktail set I and II, Calbiochem). The protein concentrations of the lysates were measured using the Bio-Rad protein assay reagent on a Beckman Coulter DU-800 spectrophotometer. The lysates were then resolved by SDS-PAGE and immunoblotted with indicated antibodies. For immunoprecipitation, 800 μg lysates were incubated with the appropriate antibody (1–2 μg) for 3 to 4 hours at 4°C followed by 1-hour incubation with Protein A sepharose beads (GE Healthcare). Immunocomplexes were washed five times with NETN buffer (20 mmol/L Tris, pH 8.0, 100 mmol/L NaCl, 1 mmol/L EDTA and 0.5% NP-40) before being resolved by SDS-PAGE and immunoblotted with indicated antibodies.

BRAF in vitro kinase assays were performed as described (42). Briefly, BRAF was immune-purified from 293T cells transfected with FLAG-BRAF constructs. GST-MEK1 and His-FZR1 were expressed in BL21 E. coli and purified using Glutathione Sepharose 4B media (GE Healthcare Life Sciences) and Ni-NTA agarose (Qiagen), respectively. BRAF kinase was incubated with 0.2 μg of GST-MEK in the absence or presence of His-FZR1 in kinase assay buffer (10 mmol/L HEPES pH 8.0, 10 mmol/L MgCl₂, 1 mmol/L dithiothreitol, 0.1 mmol/L ATP). The reaction was initiated by the addition of GST-MEK in a volume of 30 μL for 15 minutes at 30°C followed by the addition of SDS-PAGE sample buffer to stop the reaction before resolution by SDS-PAGE.

Denatured in vivo ubiquitination assays were performed as described (88). Briefly, 293 cells were transfected with FLAG-BRAF, His-ubiquitin, and HA-FZR1. Thirty-six hours after transfection, 10 μmol/L MG132 was added to block proteasome degradation, and cells were harvested in denatured buffer (6 mmol/L guanidine-HCl, 0.1 mol/L Na₂HPO₄/NaH₂PO₄, 10 mmol/L imidazole), followed by Ni-NTA (Ni-nitrilotriacetic acid) purification and immunoblot analysis.

APC^FZR1 in vitro ubiquitination assays were performed as described previously (89). Briefly, 8 μg of anti-CDC27 antibodies coupled to 80 μL of protein A-agarose (Sigma) was incubated with 0.8 mL extracts from nocodazole arrested and released G1 (3 hours post release) HeLa cells, and mixed for 2 hours at 4°C. The beads were washed three times with 1 mL swelling buffer (SB; 25 mmol/L HEPES, pH7.5, 1.5 mmol/L MgCl₂, 5 mmol/L KCl) supplemented with 0.05% Tween 20 and twice with SB. Finally, the beads were resuspended in 40 μL of SB and aliquoted into 8 tubes (5 μL for each tube). E1 (0.15 μmol/L), 1.5 μmol/L E2, 1 mg/mL ubiquitin, energy mix (7.5 mmol/L creatine phosphate, 1 mmol/L ATP, 1 mmol/L MgCl₂, 0.1 mmol/L EGTA), and 1 U creatine phosphokinase were mixed in UBAB buffer (25 mmol/L Tris/HCl, pH7.5, 50 mmol/L NaCl, 10 mmol/L MgCl₂) in a final volume of 8 μL, i.e., reaction mix. Reactions were started by the addition of bacterially purified various GST-BRAF proteins as substrates, incubated for 60 minutes at 30°C, resolved by SDS-PAGE, and immunoblotted with the indicated antibodies.

For gel filtration experiment using purified His-BRAF and His-FZR1, recombinant His-tagged proteins were purified using Ni-NTA Agarose (Cat. No. 30210, QIAGEN) according to the manufacturer's instructions. His-tagged proteins were further dialyzed with PBS supplemented with 0.1 mol/L NaHCO₃ (PBSC) and subjected to Superdex 200 10/300 GL column (GE Lifesciences Cat. No. 17-5175-01). Chromatography was performed on the AKTA-FPLC (GE Lifesciences Cat. No. 18-1900-26) with EBC buffer as described previously (9). One-column volume of elutes were fractionated with 500 mL in each fraction, at the elution speed of 0.5 mL/minutes. Aliquots (30 μL) of each fraction were loaded onto SDS-PAGE gels and detected with indicated antibodies.

For gel filtration experiment using cell lysates, cells were washed with phosphate-buffered saline, lysed in 0.5 mL of EBC buffer (50 mmol/L Tris pH 7.5, 120 mmol/L NaCl, 0.5% NP-40) containing protease inhibitors (Complete Mini, Roche) and phosphatase inhibitors (phosphatase inhibitor cocktail set I and II, Calbiochem), and filtered through a 0.45-μm syringe filter. Total protein concentration was then adjusted to 8 mg/mL with EBC buffer and 500 μL of the lysate was loaded onto a Superdex 200 10/300 GL column as described above.

Purification of GST- or His-tag–fused recombinant proteins and GST pulldown analyses were performed as described previously (27, 90).

The clonogenic survival and soft-agar assays for hTERT/p53DD/R24C-CDK4 melanocytes (IHPM) were performed as described previously (83). Briefly, for growth factor–independent clonogenic survival experiments, IHPM cells were cultured in 10% FBS containing RPMI-1640 media before plating into a 6-well plate at 3,000 cells per well. Three weeks later, cells were stained with crystal violet and the colony numbers were counted.

For soft-agar assays, IHPM cells (30,000 per well) were seeded in 0.5% low-melting-point agarose in RPMI-1640 with 10% FBS, and layered onto 0.8% agarose in RPMI-1640 with 10% FBS. The plates were kept in the cell culture incubator for 80 days, after which the colonies >50 μm were counted under a light microscope.

Coverslip passivation, TIRF microscope configuration, and image analysis were described previously (47, 48). Biotinylated anti-FLAG antibody (M2; 50 nmol/L) was added to cell lysate expressing FLAG-BRAF and incubated at room temperature for 20 minutes. Cell lysate–antibody mix was applied to streptavidin-functionalized coverslip right before the experiment to immobilize FLAG-BRAF. After 3-minute incubation, cell lysate on coverslip was washed off, replaced with buffer A (25 mmol/L Tris-HCl pH 7.5, 100 mmol/L NaCl, 20 mmol/L imidazole, 5 mg/mL BSA) containing 10 nmol/L BRAF or FZR1 labeled with Dylight550-NHS, and various concentrations of unlabeled BRAF or FZR1 as a competitor. Time series were acquired at 5 frames/second for 30 seconds. Binding constants of the competitor (K_I ∼ K_d) were calculated from the titration curve of the total number of binding events.

where N is the number of binding events; N_max is the number of binding events in the absence of a competitor; [C] is the concentration of the competitor.

All animal experiments were approved by the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee on Animal Research. The Tyr::Cre-ER^T2 transgenic mice, Pten^flox/flox mice, and Fzr1^flox/flox mice have been described previously (72, 91). Pten^flox/flox mice were first crossed with Tyr::Cre-ER^T2 mice. The resulting compound mice or Tyr::Cre-ER^T2 transgenic mice were then crossed with Fzr1^flox/flox mice to generate conditional knockout mouse models of Pten and/or Fzr1. To activate the Tyr::Cre-ER^T2 transgene to delete Pten and/or Fzr1 gene in the mouse melanocyte, the adult mice (6–8 weeks) were treated topically with 20 mg/mL4-OHT in 100% ethanol for up to 7 months at the right ear, flank, paw, and tail. Mouse tissues were fixed in 4% paraformaldehyde (PFA). Normal and tumor tissues were embedded in paraffin, sectioned, and hematoxylin and eosin (H&E) stained for pathologic evaluation.

Mouse skin tissues from EtOH- or 4-OHT–treated flanks of Tyr::CreER;Pten^lox/lox;Fzr1^lox/lox mice were lysed with RIPA buffer following sonication. The lysates were subjected to immunoblot analysis with the indicated antibodies.

Mouse skins were dissected and fixed in 4% PFA for histology and IHC. For staining, the tissues were embedded in paraffin in accordance with standard procedures. Sections (5 μm) were cut and processed for H&E staining or stained for FZR1 (34-2000, 1:100), pERK (20G11, 1:100), and pAKT (D9E, 1:100). The stained slides were visualized by a bright-field microscope.

All quantitative data were presented as the mean ± SEM or the mean ± SD as indicated of at least three independent experiments by the Student t test for between group differences. The P < 0.05 was considered statistically significant.

No potential conflicts of interest were disclosed.

Conception and design: L. Wan, M. Chen, J. Cao, H. Inuzuka, R. Cui, P.P. Pandolfi, W. Wei

Development of methodology: L. Wan, M. Chen, J. Cao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Wan, M. Chen, J. Cao, X. Dai, Q. Yin, J. Zhang, S.-J. Song, J. Liu, J.M. Katon, K. Berry, J. Fung, C. Ng, P. Liu, M.S. Song, L. Xue, M.W. Kirschner

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Wan, M. Chen, J. Cao, X. Dai, L. Xue, R.T. Bronson, M.W. Kirschner, R. Cui, P.P. Pandolfi, W. Wei

Writing, review, and/or revision of the manuscript: L. Wan, M. Chen, J. Cao, X. Dai, Q. Yin, J. Zhang, P.P. Pandolfi, W. Wei

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Cao, Y. Lu, L. Xue, R. Cui, W. Wei

Study supervision: L. Wan, J. Cao, W. Wei

Other (performed pathology): R.T. Bronson

We thank Brian North, Alan W. Lau, Jianping Guo, Naoe Nihira, and Wenjian Gan for critical reading of the manuscript, William C. Hahn, Peter K. Jackson, Hans R. Widlund, Keiran S. Smalley, and Eric B. Haura for providing reagents, and members of the Wei, Kirschner, Cui, and Pandolfi labs for useful discussions.

W. Wei is a Leukemia and Lymphoma Society Scholar and ACS Research Scholar. This work was supported in part by NIH grants (W. Wei, GM089763, GM094777, and CA177910; L. Wan, CA183914) and in part by the Skin SPORE (1P50CA168536) Developmental Research Program (L. Wan).