Loss of ARF sensitizes transgenic BRAF^V600E mice to UV-induced melanoma via suppression of XPC

Abstract

Both genetic mutations and ultraviolet (UV) irradiation can predispose individuals to melanoma. Although BRAF^V600E is the most prevalent oncogene in melanoma, the BRAF^V600E mutant is not sufficient to induce tumors in vivo. Mutation at the CDKN2A locus is another melanoma-predisposing event that can disrupt the function of both p16^INK4a and ARF. Numerous studies have focused on the role of p16^INK4a in melanoma, but the involvement of ARF, a well-known p53 activator, is still controversial. Using a transgenic BRAF^V600E mouse model previously generated in our laboratory, we report that loss of ARF is able to enhance spontaneous melanoma formation and cause profound sensitivity to neonatal UVB exposure. Mechanistically, BRAF^V600E and ARF deletion synergize to inhibit nucleotide excision repair by epigenetically repressing XPC and inhibiting the E2F4/DP1 complex. We suggest that the deletion of ARF promotes melanomagenesis not by abrogating p53 activation but by acting in concert with BRAF^V600E to increase the load of DNA damage caused by UV irradiation.

Introduction

Derived from the professional pigment-producing melanocytes, melanoma is considered the most lethal form of skin cancer (1). Mutation in the BRAF protein, especially the V600E point mutation, is the most prevalent genetic alteration in human melanoma (2). Interestingly, the BRAF^V600E mutation on its own is not sufficient to induce tumor formation in vitro and in vivo, consistent with the fact that expression of BRAF^V600E in melanocytes either in the culture dish or in animals results in senescence (3–6). Therefore, additional genetic alterations are required for the progression of BRAF^V600E-expressing melanocytes to melanoma, at least in part to achieve suppression of the senescence response.

In addition to BRAF mutation, the CDKN2A locus is frequently targeted in melanoma (7). Two distinct tumor suppressors, p16^INK4a and p14^ARF (p19^ARF in mice, hereafter referred to as ARF), are encoded by this locus via alternative reading frames (8). In contrast to p16^INK4a, the importance of ARF in melanomas has been debated, since germline mutations in ARF fail to associate with melanoma susceptibility (9) and CDKN2A mutations that target only p16^INK4a but not ARF for inactivation have been described (10). However, the recent discovery of mutations specific to the ARF-coding region and ARF-specific deletions (11), as well as the observation that ARF ablation facilitates oncogenic induction of murine melanoma (12–15), highlight the significance of ARF in melanomagenesis. Therefore, it is increasingly accepted that p16^INK4a-independent inactivation of ARF can also predispose individuals to melanoma.

The mechanistic basis of ARF’s ability to suppress melanoma may lie in ARF’s canonical role in the MDM2-p53 pathway, and indeed p53 is mostly wild-type in human melanomas (16), suggesting loss of ARF may abrogate the need for p53 mutation. However, ARF also has p53-independent functions (17, 18), which explains the fact that it can act as a melanoma suppressor in the absence of p53 (14). For example, ARF has been shown to bind to DRTF polypeptide 1 (DP1), a coactivator of E2F transcription factors, and inhibit the formation of active DP1-E2F complexes, thereby regulating the expression of E2F target genes (19–21). Therefore, the mechanism of melanoma suppression by ARF remains unclear. Understanding this mechanism may have impact on the design of future therapies for melanoma.

The ability of ARF to modulate E2F activity also extends ARF’s function to activation of the nucleotide excision repair (NER) pathway (21). NER is a DNA repair mechanism that functions to remove bulky DNA adducts such as cyclobutane pyrimidine dimers (CPD) and 6-4-photoproducts mostly caused by ultraviolet irradiation (UVR) (22). Interestingly, solar UVR is the major etiological factor for skin cancers including melanoma, although the relationship between UVR and melanoma is less evident than that of squamous and basal cell carcinomas (23). Recent studies demonstrated neonatal UVB irradiation (275–320nm) can exacerbate melanoma penetrance in various genetic mouse models (12, 13, 24); however, the mechanism leading to UV sensitivity in melanocytes is not well understood. Although BRAF mutations are not overrepresented in melanomas arising from chronic sun-exposed areas, it is unclear whether BRAF^V600E alone or in combination with other genetic events increases the sensitivity to UVR.

In the present study, we used a transgenic BRAF^V600E mouse model (3) to show that loss of ARF is able to shorten the latency of melanoma formation and causes profound sensitivity to neonatal UVB exposure. Oncogenic BRAF^V600E instigates transient DNA damage in melanocytes which triggers senescence that cannot be bypassed by ARF deletion in vivo. However, loss of ARF cooperates with BRAF mutation to suppress the NER pathway, enhancing UVB-induced DNA damage and melanomagenesis.

Materials and Methods

Mice

All mouse experiments were performed with the approval of Tufts University/Tufts Medical Center IACUC. The generation of melanocyte-specific BRAF^V600E transgenic mice was described previously (3). Founder mice were backcrossed to C57BL/6 mice for more than 10 generations. The exon 1b-specific p19^ARF knockout mouse in C57BL/6 background used in this study has been described elsewhere (25) and was a generous gift from N. Rosenberg (Tufts University School of Medicine, Boston, MA). All mice were maintained in a pathogen-free mouse facility at Tufts University School of Medicine.

Cell Culture and Recombinant Vectors

All cells were cultured in a 5% CO₂ humidified incubator at 37°C. Primary mouse melanocytes were isolated from neonatal mouse skins as described (26) and maintained in F12 medium supplemented with 3% fetal bovine serum (FBS, Gibco), 1% antibiotic-antimycotic (Invitrogen), 48 nM 12-O-tetradecanoyl-phorbol-13-acetate (Sigma), 0.1 mM isobutylmethyl xanthine (Sigma), 10 μg/ml bovine pituitary extract (Invitrogen) and 0.1 mM dibutyryladenosine 3′,5′-cyclic monophosphate (Sigma). Contaminating fibroblasts and keratinocytes were eliminated by treatment with 100 μg/ml geneticin (Invitrogen) for 24h. Primary mouse melanoma cells were isolated by collagenase/hyaluronidase digestion of tumor fragments for 30 min and grown in RPMI 1640 (Gibco) supplemented with 10% FBS and 1% antibiotic-antimycotic (27). Human melanoma cell line CHL1 was cultured in DMEM (Gibco) with 10% FBS and 1% penicillin-streptomycin.

The retroviral vector pBabe-puro-p19^ARF was generously provided by N. Sharpless (University of North Carolina, Chapel Hill, NC), pBabe-hygro-dominant-negative-p53 was obtained from Addgene (#9058) (28). The lentiviral construct FG12-HA-BRAF^E600-eGFP was a gift from D. Peeper (The Netherlands Cancer Institute, Amsterdam, The Netherlands) (5). shRNA targeting mouse p53 sequence 5′-GTACTCTCCTCCCCTCAAT-3′ was generated in the pMKO.1 retroviral vector.

UV Exposure

For irradiation of cultured cells, cells were washed with PBS and subsequently exposed to UV radiation using UVB lamps (280–314nm; UVP, Inc.) at the indicated dose. UV emittance was measured with the use of a UV photometer (model ILT1400A; UV Products).

For in vivo tumor induction, animals were irradiated under the UVB bulbs at a dose of 750mJ/cm² at neonatal day 3.5. Following UV exposure, mice were carefully monitored for the degree of erythema and/or desquamation. Severe erythema or desquamation resulted in sacrifice of the animal but was rarely observed. The majority of animals displayed mild to moderate erythema and was monitored for tumor formation as indicated.

Histology

Tissue samples were fixed in 10% buffered formalin overnight and stored in 70% ethanol prior to paraffin embedding, sectioning and hematoxylin/eosin (H&E) staining (by the Rodent Histopathology Core, Harvard Medical School). Immunohistochemistry was performed with the following antibodies: Ki-67 (SP6, Thermo Scientific), p16 (M-156, Santa Cruz), p53 (Thermo Scientific), interleukin 6 (ab6672, Abcam), Trp2 (ab74073, Abcam), S100 (Ab-2, Thermo Scientific), Melan A (C-20, Santa Cruz) and HMB45 (ab732, Abcam). Staining was performed with Vector NovaRED Substrate Kit (Vector Laboratories). Negative control was done by replacing the primary antibody with species-matched total IgG.

Senescence-associated β-Galactosidase Staining

Skin samples were fixed in 4% buffered paraformaldehyde at room temperature for 30 min, followed by a wash with 50 mM glycine in PBS. Dehydration was performed by subsequently soaking the samples in 20% sucrose overnight and 30% sucrose until the samples settle at the bottom. Dehydrated samples were embedded in OCT, frozen and sectioned. The staining for perinuclear SA-β-gal activity was performed according to protocol described previously (29, 30).

Immunoblotting and Immunoprecipitation

For immunoblotting, tissue samples or cultured cells were homogenized and/or lysed with RIPA buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, 0.1% SDS, 1% NP-40, 1% sodium deoxycholate, 1 mM sodium pyrophosphate, 10 mM β-glycerophosphate, 0.2 mM Na₃VO₄, 10 mM NaF, 1 mM DTT and eComplete-Mini protease inhibitor cocktail (Roche). Primary antibodies used included p19 (ab80, Abcam), p16 (F-12, Santa Cruz), GAPDH (MAB374, Chemicon), p53 (Thermo Scientific), p21 (F-5, Santa Cruz), phosphor-p53-Ser15 (9284, Cell Signaling), pERK1/2 (4376, Cell Signaling) and ERK2 (C-14, Santa Cruz).

Co-immunoprecipitation was carried out as described by Dominguez-Brauer et al. (21). The antibodies used were p19 (ab80, Abcam), E2F4 (C-20, Santa Cruz) and DP1 (hybridoma supernatant kindly provided by J. Lees, MIT, Cambridge, MA).

Quantitative Real-time PCR

Total RNA was extracted by TRIzol (Invitrogen) and converted into cDNA using either iScript (Bio-Rad) or SMARTScribe (Clontech) cDNA synthesis kit. Gene expression was quantified using QuantiTect SYBR Green (Qiagen) or SYBR Advantage (Clontech).

Genomic DNA was extracted by phenol-chloroform, followed by methylation sensitive enzyme HpaII digestion for 3h. Quantitative PCR was carried out with 1 μl of the HpaII-digested heat-inactivated gDNA and with the XPC promoter-specific and control primers (Supplementary Table 1). The sequences between XPC forward and reverse primers harbor multiple HpaII recognition sites, thereby will be digested if not methylated; the sequences between the control primers do not have any HpaII sites, and thus served as an internal control.

CPD Immunoblot

Heat-denatured genomic DNA was dot-blotted onto a nitrocellulose membrane and was blocked with 5% milk overnight. The membrane was then incubated with anti-CPD antibody (provided by R. Cui, Boston University, Boston, MA) for 30 min at room temperature.

Growth Curve and Soft Agar Colony-formation Assay

Primary mouse melanoma cells were stably infected with dominant-negative p53 (DNp53) or empty vector and seeded in 6-well plates at a density of 1X10⁴/well in triplicate. Cell number was counted at indicated time points.

For anchorage-independent growth assays, 2X10⁴ cells were suspended in medium containing 0.4% agarose and plated onto a solidified layer of medium-containing 0.8% agarose in 6-well plates. Colonies were quantified in five fields in each well 3 weeks post seeding.

Chromatin Immunoprecipitation

ChIP was carried out according to the protocol described previously (31) using E2F4 antibody (C-20, Santa Cruz). Real-time PCR was performed according to Dominguez-Brauer et al.(21). ChIP data were normalized to input chromatin.

Results

ARF is a suppressor of BRAF^V600E-driven murine melanoma

ARF can be induced by oncogenes such as c-Myc and Ras in cultured cells (17). To determine whether oncogenic BRAF^V600E can trigger ARF expression in vivo, we performed immunoblots on skin lysates from either wild-type (WT) or melanocyte-specific BRAF^V600E transgenic mice (3). ARF, but not p16^INK4a, was detectable in the BRAF^V600E skin from adult mice (Figure 1A), suggesting that ARF induction is a significant cellular response to chronic BRAF activation. Therefore, we introduced the germline ARF-specific deletion allele (p16^INK4a is intact) into BRAF^V600E mice (strain 476)(3) and monitored tumor development. Neither BRAF^V600E nor BRAF^V600E;ARF^+/− mice developed tumors for up to 5 months, whereas BRAF^V600E;ARF^−/− animals formed tumors relatively rapidly, with a mean latency for survival of 85 days of age (Figure 1B). 40% of the BRAF^V600E;ARF^−/− mice were sacrificed because of progressive melanoma, as indicated by asterisks in the curve (Figure 1B & S1A); the remaining mice were sacrificed as a result of sarcoma and lymphoma, the prominent tumor types in germline ARF-null animals (32). Surprisingly, during the experimental period, sarcomas and lymphomas were also observed in the BRAF^V600E;ARF^−/− mice but not in the control (except in one BRAF^V600E;ARF^+/− mouse, Figure 1B), suggestive of reduced latency for these ARF-null tumors in the tyrosinase-driven BRAF^V600E background. Given that expression of the BRAF^V600E transgene was not readily detected in the sarcomas and lymphomas examined (Figure S1B), it is unlikely that these tumors resulted from expression of BRAF^V600E in the affected tissues. Considering the existence of elevated amounts of inflammatory cytokines (Figure 3B & C), it is possible that alteration of the microenvironment by the BRAF^V600E transgene facilitates the development of other tumors.

Figure 1. Loss of ARF accelerates melanoma induced by BRAF^V600E in vivo.

(A) Immunoblots of total skin lysate show that ARF but not p16^INK4a is detected in the BRAF^V600E transgenic skin. pERK1/2 was blotted to confirm the presence of active BRAF in the skin. Spontaneous melanoma derived from a BRAF^V600E (470 strain) mouse was used as a positive control for both ARF and p16^INK4a.

(B) Kaplan-Meier plot of total mortality of mice with different genotypes. Asterisk indicates mouse died of melanoma. The melanoma-free survival rate of the same cohort of mice is shown in Supplementary Figure 1A.

(C) H&E-stained sections from representative skins of different genotypes, primary BRAF^V600E;ARF^−/− tumor and lung. Triangles in BRAF skin image indicate the large nests of epithelioid cells; arrows in BRAF;ARF^−/− skin image show the appearance of the Schwannian differentiated phenotype; arrowheads in BRAF;ARF^−/− tumor image show tumor cells invading the muscle layer; inset in the tumor image shows mitotic figure, and inset in the lung image shows pigmented micrometastasis.

(D) Immunohistochemical (IHC) staining of Ki-67, p16^INK4a and p19^Arf (Arf) in representative sections from BRAF^V600E;ARF^−/− tumors. Arf staining in tumor section confirms loss of Arf expression. Sections were counterstained with hematoxylin.

Figure 3. Loss of ARF cannot suppress BRAF^V600E-induced senescence in vivo.

(A) SA-β-gal staining on skin sections from age-matched mice of the indicated genotypes. Hematoxylin was used to counterstain the nucleus. Red boxed areas in the upper panel are enlarged in the lower panel.

(B) qRT-PCR showing the mRNA expression of selected members of the senescence-associated secretome in skins of different genotypes. Student t-test, compared to WT skin, * p<0.05, ** p<0.005.

(C) IHC of interleukin-6 (IL-6) performed on representative skin sections from BRAF^V600E and BRAF^V600E;ARF^−/− mice. Boxed areas are enlarged as indicated and arrowheads represent the positive staining cells. Sections were counterstained with hematoxylin.

(D) IHC of p16^INK4a and p53 on representative skin sections from BRAF^V600E;ARF^−/− mice. IgG is negative control and nucleus was counterstained by hematoxylin.

Histologically, ARF^−/− skin was comparable to WT skin (Figure 1C). Expression of BRAF^V600E caused melanocytic proliferation presenting as large nests of epithelioid cells in the deep dermis and subcutis (Figure 1C, triangles)(3). Knockout of ARF in BRAF^V600E mice increased the apparent hyperplasia of melanocytes, with increased deposits in the dermis and subcutis and the appearance of a Schwannian differentiated phenotype (Figure 1C, arrows)(3). Melanomas that formed in the BRAF^V600E;ARF^−/− mice showed a transition from pigmented to unpigmented, suggesting progression from a differentiated to an undifferentiated phenotype (Figure 1C). The tumors were often very invasive, penetrating through the muscle layer underneath the skin (Figure 1C, arrowheads), and showed a high proliferation index seen in the form of mitotic figures (Figure 1C, insert) and Ki-67 staining (Figure 1D). Interestingly, this proliferation was observed despite the fact that a substantial number of cells positive for nuclear p16^INK4a were detected in the tumor sections (Figure 1D). Analysis of lungs from 6 melanoma-bearing mice revealed pigmented micrometastasis in 2 cases (Figure 1C). Taken together, these data indicate that deletion of ARF accelerates invasive melanoma formation driven by BRAF^V600E and thus acts as a melanoma suppressor in this context.

ARF loss does not abrogate p53 activation in BRAF^V600E;ARF^−/− murine melanomas

ARF is a well-characterized p53 stabilizer, therefore abrogation of ARF is expected to dampen p53 activation. Surprisingly, we detected strong p53 expression in the BRAF^V600E;ARF^−/− precancerous skin and tumors (Figure 2A & S2A). p53 was active based on its phosphorylation at Ser15 site and expression of the downstream target p21 (Figure 2B), suggesting that loss of ARF might not affect the function of p53 induced by chronic BRAF^V600E stimulation in this model. To confirm that p53 in BRAF^V600E;ARF^−/− tumor cells is functional, we knocked down p53 and found that p21 level was reduced (Figure 2B). In addition, we restored ARF expression in BRAF^V600E;ARF^−/− tumor cells (Figure S2B), and challenged the cells with UV irradiation. Upon UV induction, p53 was rapidly phosphorylated regardless of ARF status (Figure 2D), further supporting the contention that p53 in ARF-null melanoma cells is functional.

Figure 2. p53 is functional in BRAF^V600E;ARF^−/− melanocytes and melanoma cells despite ARF deficiency.

(A) Immunoblots detecting p53 from total skin and tumor lysates.

(B) Immunoblots detecting p21 and phospho-p53 at Ser 15 (p-p53) from total BRAF^V600E;ARF^−/− tumor lysates. Control lane was lysate from a spontaneous BRAF^V600E melanoma (470 strain, the same sample with that used in Figure 1A) serving as a positive control for ARF.

(C) Immunoblots demonstrating p53 knockdown efficiency and reduced level of p21, a downstream target of p53, in isolated primary BRAF^V600E;ARF^−/− melanoma cells. Control sample for ARF was the same as in (B).

(D) Immunoblots demonstrating induction of phospho-p53 in primary BRAF^V600E;ARF^−/− melanoma cells with or without ARF restoration. p53 can be rapidly activated upon UV irradiation, regardless of ARF status, confirming that the p53 present in the cells is functional.

(E–F) Growth curve and soft agar colony-formation ability of BRAF^V600E;ARF^−/− primary melanoma cells. Cells (#4228) with dominant-negative p53 (DNp53) grew faster (D) and formed more colonies (E) than control. Colonies that were bigger than 30μm were quantified. Student t-test, * p<0.05, ** p<0.005.

Given that deletion of ARF does not alter p53 functionality, we reasoned that repressing the function of p53 would have an additive effect on ARF loss to enhance tumor formation. To test this, we generated BRAF^V600E;ARF^−/− tumor cells stably expressing either dominant-negative p53 (DNp53, Figure S2C) or mouse p53-specific shRNA. Proliferation assays showed that DNp53-expressing cells grew significantly faster than cells with vector control (Figure 2D). Soft agar colony formation assays confirmed that DNp53-expressing cells produced more and larger colonies than control cells (Figure 2E). Results using p53-shRNA were similar (Figure S2D–E). Taken together, the data suggest that loss of ARF affects targets other than p53 to increase melanoma formation, since p53 is active despite ARF deficiency.

Loss of ARF cannot suppress BRAF^V600E-induced senescence in vivo

It has been shown that loss of ARF bypasses senescence in cultured primary melanocytes (14). To examine whether ARF deficiency can repress oncogene-induced senescence in vivo, we performed senescence-associated (SA) β-gal staining on skin sections from age-matched mice. This approach indicated the presence of senescence in BRAF^V600E skin; however, to our surprise, strong SA-β-gal staining can also be detected in the double mutant skin (Figure 3A), suggesting senescent cells are still abundant in the ARF-null background. Quantitation of senescent cells in the skin based on SA-β-gal staining is complicated by dye retention in hair follicles and sebaceous glands. Therefore, we analyzed the expression of genes encoding members of the senescence-associated secretome as an additional marker. Quantitative RT-PCR revealed that most of the secretome factors tested were upregulated in BRAF^V600E skin and considerably more highly expressed in BRAF^V600E;ARF^−/− skin (Figure 3B). Further, IHC staining confirmed increased expression of IL-6 adjacent to putative senescent cells (Figure 3C). The presence of an equal or increased number of senescent cells observed in the BRAF^V600E;ARF^−/− skins is consistent with the observation that both p16^INK4a and active p53 are present in situ (Figure 3D). Therefore, the persistent expression of senescence activators and the presence of a copious number of senescent cells despite the lack of ARF strongly argue that direct evasion of senescence is not the mechanism by which loss of ARF accelerates BRAF^V600E-driven melanoma formation.

Loss of ARF sensitizes BRAF^V600E mice to UV-induced melanomagenesis

The data described above are consistent with the notion that increased and/or persistent cellular stress in BRAF^V600E;ARF^−/− melanocytes leads to genetic mutations that bypass senescence and lead to melanomagenesis. To determine if cellular changes elicited by combined BRAF and ARF mutation render melanocytes sensitive to DNA damage, we studied the response of mutant cells and animals to UV irradiation, since solar UV irradiation is the main etiological factor for melanoma. To this end, we irradiated neonatal mice of different genotypes with a single dose of UVB (750mJ/cm², Figure 4A). During the period of study (100 days), none of the WT, BRAF^V600E or ARF^−/− mice succumbed to melanoma following UVB exposure (Figure 4B). Extending the observation period to 300 days also failed to reveal UV-induced tumors in WT and BRAF^V600E mice, in agreement with a previous report focusing on Tyr-NRas mice (13). Significantly, every UVB-irradiated BRAF^V600E;ARF^−/− mouse developed multiple melanomas, with 50% of mice developing tumors by day 70 post-UV (73 days of age; Figure 4B). Thus, UVB irradiation promotes melanomagenesis in BRAF^V600E;ARF^−/− mice. Histopathological analysis of the tumors confirmed most of them were amelanotic melanomas, as they did not secrete melanin, but stained positive for melanoma markers Trp2, S100, Melan A and HMB45 (Figure 4C). Consistent with our observation in spontaneous melanomas, UVB-induced melanomas also preserve active p53 (Figure 4C, lower panel). Therefore, although UVB irradiation is not sufficient to induce melanomas in mice expressing melanocytic BRAF^V600E, loss of ARF sensitizes BRAF^V600E mice to UVB-induced melanomagenesis.

Figure 4. Loss of ARF sensitizes BRAF^V600E mice to UV-induced melanoma.

(A) Schematic representation of the neonatal UVB irradiation protocol. Mice of different genotypes were irradiated with 750mJ/cm² UVB at postnatal day 3.5 (P3.5).

(B) Kaplan-Meier curve showing melanoma-free survival of mice receiving neonatal UVB irradiation. Triangle in the ARF^−/− curve indicates mouse died of sarcoma.

(C) H&E and immunostaining of melanoma markers Trp2, S100, Melan A and HMB45 on representative sections from UVB-induced BRAF^V600E;ARF^−/− tumors. Arf staining confirms the absence of the protein in tumors. Arrowheads in the p53 and phospho-p53 (Ser15) images indicate the positive staining cells.

Loss of ARF impairs DNA damage repair by transcriptionally repressing XPC

UV irradiation results in bulky DNA adducts such as cyclobutane pyrimidine dimers (CPD) and 6,4-photoproducts, which can be repaired by nucleotide excision. In the absence of repair, mutagenic events may become fixed in the genome and lead to cellular transformation. Because UVB can increase melanoma development in BRAF^V600E;ARF^−/− mice, we wanted to know whether the combination of those two mutations has any effect on DNA repair capability. Since the xeroderma pigmentosum, complementation group C gene (XPC) is a key component in the NER pathway responsible for recognizing DNA adducts, we tested the expression of XPC in primary melanocytes with different genotypes by RT-qPCR. Single or double mutant melanocytes expressed a significantly lower level of XPC mRNA, with the lowest level seen in the double mutant cells (Figure 5A). The reduced XPC mRNA level correlated with impaired DNA repair capability as demonstrated by a CPD removal assay (Figure 5B): CPD accumulated dramatically 3h following UVB irradiation and was successfully cleared in WT melanocytes by 24h post-UVB, but persisted in cells with BRAF^V600E and ARF loss. It has been reported that in mouse fibroblasts, ARF enhances XPC expression by inhibiting the transcriptional repressor activity of E2F4 via disruption of the E2F4-DP1 interaction (21). To determine whether this mechanism might be active in melanomas, we subjected BRAF^V600E;ARF^−/− melanoma cells with or without ARF reconstitution to UVB irradiation and measured the XPC mRNA level. Upon UVB, XPC mRNA decreased dramatically in the absence of ARF, while the presence of ARF maintained the expression of XPC mRNA (Figure 5C). Further, we confirmed that ARF is able to bind DP1, depleting the fraction of DP1 bound to E2F4 (Figure 5D). Chromatin immunoprecipitation of E2F4 confirmed the association of this transcriptional repressor with the XPC promoter in both unirradiated and irradiated melanoma cells (Figure 5E). Interestingly, E2F4 association with the XPC promoter decreased following UVB irradiation (Figure 5E) concomitant with reduction in XPC mRNA (Figure 5C), suggesting that UVB irradiation may increase the association of E2F4 with corepressive factors such as p130 (21), thus repressing promoter activity more efficiently despite reduced promoter occupancy. Importantly, reconstitution of melanoma cells with ARF further decreases the association of E2F4 with the XPC promoter and results in increased transcription, suggesting that ARF acts to counteract E2F4-mediated repression of XPC in UVB-irradiated melanoma cells, consistent with the reported role of ARF in MEFs (Figure 5E)(21).

Figure 5. Loss of ARF impairs DNA damage repair pathway by transcriptionally repressing XPC.

(A) mRNA expression of XPC in primary mouse melanocytes of different genotypes measured by qRT-PCR. Student t-test, * p<0.05, ** p<0.005.

(B) Immunodot blot and quantification of the level of CPD in genomic DNA at various time points after UVB irradiation. Primary mouse melanocytes of different genotypes were subjected to UVB exposure at 25mJ/cm², followed by genomic DNA extraction at indicated time points.

(C) Expression of XPC mRNA in primary mouse melanoma cells after UVB exposure. BRAF^V600E;ARF^−/− melanoma cells with or without p19^ARF restoration were subjected to UVB irradiation for about 3 min to reach 100 mJ/cm², followed by RNA isolation at various time points. 0h represents RNA isolated immediately after UVB exposure.

(D) Co-immunoprecipitation showing the interaction between p19^ARF/E2F4 and DP1. The binding of p19^ARF and DP1 can be detected in BRAF^V600E;ARF^−/− melanoma cells expressing ectopic p19^ARF. In the presence of p19^ARF, the amount of DP1 bound to E2F4 was decreased.

(E) ChIP demonstrates XPC promoter occupancy by E2F4. Immunoprecipitation was performed using lysates of BRAF^V600E;ARF^−/− melanoma cells with or without p19^ARF restoration receiving 25mJ/cm² UVB; qPCR was performed using the primers described previously (22).

BRAF^V600E inhibits the NER pathway by promoting methylation of the XPC promoter

The RT-qPCR result in Figure 5A shows that BRAF^V600E mutation alone can reduce the level of XPC mRNA. It has been reported that in lung cancer, the promoter region of XPC is highly methylated (33). Further, perturbation of Ras signaling can regulate DNA methylation (34). Since BRAF functions downstream of Ras, it’s reasonable to postulate that activation of BRAF could also influence promoter methylation. To test the relevance of this mechanism to XPC regulation, we first analyzed the CpG island in the mouse XPC promoter that spans -1000 to -1 relative to the transcriptional start site. As is the case in the human XPC promoter, this region contains multiple CpG islands, as predicted by CpG island searcher software (35) (Figure 6A). Three HpaII restriction enzyme recognition sites are located within the XPC promoter. Therefore we performed HpaII-based quantitative PCR to determine XPC methylation in melanocytes (33). PCR products can be detected in HpaII-digested genomic DNA from BRAF^V600E and BRAF^V600E;ARF^−/− cells, with a higher amount in the BRAF^V600E;ARF^−/− DNA (Figure 6B), indicating that the XPC promoter is hypermethylated in cells of these genotypes. Consistent with this, treatment of melanocytes with the DNA demethylating agent 5’-aza-2’-deoxycytidine can increase the XPC mRNA level only in cells harboring the BRAF^V600E mutation (Figure 6C), suggesting a link between BRAF mutation and XPC hypermethylation.

Figure 6. Mutant BRAF^V600E caused XPC promoter methylation.

(A) Diagram showing CpG islands in promoter region of the mouse XPC gene.

(B) Detection of XPC promoter methylation in mouse primary melanocytes of different genotypes. Genomic DNA was subjected to methylation-sensitive HpaII treatment, following by quantitative PCR using primers flanking the HpaII-recognition sites. Primers flanking genomic sequences that do not contain any HpaII site, thereby resistant to HpaII treatment, were used as internal control. Increased recovery of PCR product indicates increased methylation in the corresponding region. Student t-test, * p<0.05, ** p<0.005.

(C) Expression of XPC following treatment with the demethylating agent 5′-aza-2′-deoxycytidine (aza-dC). Mouse primary melanocytes of different genotypes were treated with 5μM aza-dC for 48h, followed by RNA preparation and quantitative real-time PCR.

(D–E) Acute expression of BRAF^V600E increases XPC promoter methylation and inhibits XPC mRNA expression. BRAF wild-type, NRAS wild-type CHL1 melanoma cells were infected with either empty vector or HA-BRAF^V600E virus, followed by treatment with mutant BRAF inhibitor PLX-4032 at indicated concentration for 48h. Protein, RNA and DNA were collected and subjected to immunoblot to confirm the activation of BRAF pathway (D), real-time quantitative PCR to determine XPC mRNA level (E) and HpaII-based qPCR to examine promoter methylation (F).

To examine whether BRAF^V600E directly stimulates XPC promoter hypermethylation, we acutely expressed BRAF^V600E in BRAF wild-type CHL1 melanoma cells (Figure 6D & S3A). Consistent with a previous report (36), introduction of BRAF^V600E instigated senescence, presumably due to activation of the DNA damage response mainly in the form of double-strand breaks (Figure S3B). The expression of XPC was significantly inhibited in response to BRAF^V600E addition (Figure 6E), correlating with a dramatic induction of methylation in the XPC promoter region (Figure 6F). Abrogating mutant BRAF activity by selective inhibition with PLX-4032 was able to not only restore XPC expression but also impaired promoter methylation (Figure 6E & F), suggesting that BRAF kinase activity is required for maintaining XPC promoter hypermethylation. Interestingly, PLX-4032 caused ERK activation in CHL1 cells with wild-type BRAF (Figure 6D) (37), which also resulted in XPC reduction and promoter methylation (Figure 6E & F), further supporting the involvement of BRAF-mediated signaling in XPC promoter hypermethylation.

Discussion

In this study, we confirmed ARF is a melanoma suppressor that can cooperate with activated BRAF and demonstrated that loss of ARF is able to sensitize BRAF^V600E mice to neonatal UVB-induced melanoma. Oncogenic BRAF^V600E instigates senescence in melanocytes, which cannot be bypassed by ARF deletion in vivo. Mechanistically, we showed that ARF loss did not abrogate p53 activity, but instead reduced nucleotide excision repair (NER) by elevating the inhibitory effect of E2F4-DP1 on XPC expression, as first identified in fibroblasts (21). In addition, BRAF^V600E also impaired XPC expression by increasing promoter hypermethylation. Interestingly, XPC has been shown to play a role in melanoma photocarcinogenesis (38). Therefore, XPC expression appears to be a nexus for both oncogenic events studied here, and the combined effects of BRAF mutation and ARF deletion act synergistically to inhibit DNA repair and enhance melanomagenesis in the presence of UVB irradiation. We surmise that increased melanomagenesis in BRAF^V600E;ARF^−/− mice that have not been exposed to UV irradiation also results from decreased DNA repair capacity, rendering mutant melanocytes sensitive to the increase in double-strand breaks (DSB) that accompanies BRAF^V600E expression. Interestingly, chronically reduced XPC expression has been linked to a reduced ability to repair DSBs (39), but effects of BRAF^V600E and ARF loss on other forms of DNA repair may also contribute to the high rate of mutations observed in melanoma cells (40).

The efficient formation of melanomas in BRAF^V600E;ARF^−/− mice despite the persistence of high levels of senescence in the skin strongly suggests that the elevated mutation rate leads to selection for events that bypass senescence in individual cells that then emerge as tumor foci. The gene encoding p16^INK4a is an obvious candidate, yet BRAF^V600E;ARF^−/− tumors retained p16^INK4a expression, excluding the possibility of concurrent loss of p16^INK4a as a cause. Indeed, although deletion of p16^INK4a undoubtedly can predispose to melanoma formation, preservation of this protein in established melanoma is, nevertheless, not uncommon (4). Identification of mutational events that cooperate with BRAF^V600E to bypass senescence and drive melanoma formation or progression is of obvious importance therapeutically, and the system described here could be useful for studying these mutations. In addition to unbiased screens for cooperating alleles, approaches to this problem include further analysis of likely candidates such as components of the Rb pathway (12), as well as several genes newly identified as susceptible to UV signature mutations in human tumors (41).

One candidate that seems surprisingly excluded from alteration in these melanomas is p53, whose UV-signature mutants are commonly linked to non-melanoma skin tumors (42). Further, although ARF has been long appreciated as a p53 stabilizer, in our model, we did not observe p53 inactivation in spite of ARF deficiency. To the contrary, persistent p53 expression and activation were detected in precancerous skins, tumor tissues and primary cell lines derived from this model. This seemingly contradictory result is actually in accord with human studies concluding that p53 is commonly preserved in melanomas (43), and at the same time undermines the argument that the lack of TP53 mutations in melanoma is compensated by functional loss of p53 activity owing to ARF deletion (7, 43). Indeed, contrary to other human malignancies, p53 in melanomas commonly remains wild-type at the genomic level (16), and also tends to be overexpressed at the protein level (43). Although p53 targeting, in combination with lineage-specific oncogenes activation, has been proven to drastically induce melanoma formation (3, 6, 13, 14), BRAF^V600E;TP53^−/− murine melanomas show a distinct transcript profile from that of BRAF^V600E;CDKN2A^−/− tumors which closely resemble human melanoma (unpublished data, FG Haluska). This in turn suggests that the TP53-null melanoma might derive from a distinct population from that giving rise to human melanoma. It is well established that ARF possesses functions independent of p53 (17), and selection against these functions of ARF appear to predominate in melanoma. However, it is likely that alterations in the p53-independent targets of ARF, the actions of hyperactivated BRAF and selected genetic and epigenetic events combine to offset the effects of activated p53 in nascent melanomas.

The presence of p53 in precancerous BRAF^V600E;ARF^−/− skin correlates well with our observation that senescence persists in the same tissues, although a previous report indicates ARF ablation bypasses senescence in cultured melanocytes (14). A potential explanation for this discrepancy lies in the difference of senescence triggers: oncogenic BRAF^V600E stimulation in melanocytes in situ versus primary melanocytes passage in culture. Since we have not tested the stability of p53 in the context of ARF loss in this model, it’s also possible that without ARF, the turnover of p53 is augmented; however, constitutive oncogenic signaling may effectively counterbalance this, leading to the observed elevation of active p53. On the other hand, the fact that senescence persists in the skin of BRAF^V600E;ARF^−/− animals that eventually succumb to tumors does not challenge the tumor-suppressive effect of senescence intrinsically, but instead, supports the notion that massive senescence might be protumorigenic extrinsically. Senescent cells secrete numerous factors including inflammatory cytokines, growth factors, and extracellular matrix remodeling enzymes that are collectively called the senescence-associated secretome (44). Such secretome factors make the surrounding milieu protumorigenic, facilitating malignant progression of individual cells that have bypassed senescence due to further mutations gained spontaneously or UVB-induced. This could also potentially explain why BRAF^V600E;ARF^−/− mice display a decrease in the latency of sarcoma and lymphoma when compared to ARF^−/− counterparts.

In addition to increased mutation events resulting from impaired XPC expression, a broader array of epigenetic changes may also lead to bypass of tumor suppressive influences such as p53 function and senescence. Indeed, epigenetic alterations have emerged as an important mechanism underlying BRAF^V600E-driven tumorigenesis. For example, BRAF^V600E can contribute to methylation alterations in single genes (45, 46) as well as to broader changes in methylation patterns in the genomes of melanomas and papillary thyroid cancers (47, 48). Interestingly, one report failed to associate BRAF mutations with hypermethylation of 15 cancer-linked genes in melanomas (49), suggesting that important BRAF methylation targets in melanomas remain to be discovered. Here, we provide evidence that promoter hypermethylation of XPC is such an event, but this gene is unlikely to be solely responsible for BRAF^V600E-mediated changes in mutation sensitivity. Based on bioinformatic analysis, CpG islands are clearly evident in other NER genes, such as XPA and ERCC5, suggesting BRAF^V600E-mediated methylation of those promoters may also occur. Additionally, the mechanism behind BRAF^V600E-induced methylation remains to be elucidated. Upregulation of DNA methyltransferase 1 by BRAF^V600E is a promising candidate, as has been reported (47). Interestingly, Ras-mediated methylation requires numerous factors including E2F1 (50), and this is likely to be the case for BRAF^V600E as well. Because the activity of E2F1, like E2F4, requires the association of DP1, and this association can also be influenced by ARF (20), it would be intriguing to investigate whether ARF loss facilitates BRAF^V600E-mediated promoter methylation, in addition to altering the expression of individual genes in an E2F-depedent manner. Clearly, the ability to reverse BRAF^V600E-induced epigenetic changes that then lead to an increased incidence of genetic changes could have an important impact on both tumor progression and development of resistance to therapy. The induction of XPC observed here upon PLX4032 treatment offers a glimmer of hope that such processes remain targetable in malignant melanomas, and are therefore an important area for future study.