Wnt signaling potentiates nevogenesis

Significance

Human benign nevi (moles) are clonal neoplasms that rarely progress to melanoma because their cells (melanocytes) are arrested in a viable but nonproliferating state (senescence). However, at low frequency, nevus melanocytes do progress to melanoma. Consequently, it is important to understand the factors that determine nevus formation and progression to melanoma. We present evidence that repression of a proliferation-promoting cell signaling pathway (Wnt signaling pathway) contributes to senescence of melanocytes in vitro. However, Wnt signaling remains active in some senescent human melanocytes in nevi, and activation of Wnt signaling leads to a delay in melanocyte senescence in a mouse model. We suggest that activated Wnt signaling in human nevi delays senescence to promote nevus formation, and thereafter, persistent Wnt signaling might undermine senescence-mediated tumor suppression.

Abstract

Cellular senescence is a stable proliferation arrest associated with an altered secretory pathway (senescence-associated secretory phenotype). Cellular senescence is also a tumor suppressor mechanism, to which both proliferation arrest and senescence-associated secretory phenotype are thought to contribute. The melanocytes within benign human nevi are a paradigm for tumor-suppressive senescent cells in a premalignant neoplasm. Here a comparison of proliferating and senescent melanocytes and melanoma cell lines by RNA sequencing emphasizes the importance of senescence-associated proliferation arrest in suppression of transformation. Previous studies showed that activation of the Wnt signaling pathway can delay or bypass senescence. Consistent with this, we present evidence that repression of Wnt signaling contributes to melanocyte senescence in vitro. Surprisingly, Wnt signaling is active in many senescent human melanocytes in nevi, and this is linked to histological indicators of higher proliferative and malignant potential. In a mouse, activated Wnt signaling delays senescence-associated proliferation arrest to expand the population of senescent oncogene-expressing melanocytes. These results suggest that Wnt signaling can potentiate nevogenesis in vivo by delaying senescence. Further, we suggest that activated Wnt signaling in human nevi undermines senescence-mediated tumor suppression and enhances the probability of malignancy.

Senescence is a stable proliferation arrest that is activated by various molecular triggers, including activation of protooncogenes, such as BRAF and N-RAS, in primary human cells (1–3). Cellular senescence is also accompanied by an altered secretory program [senescence-associated secretory phenotype (SASP)], composed of increased expression of immune regulators and matrix proteases (4). Cellular senescence is an important tumor suppression mechanism, and proliferation arrest and SASP are thought to act in concert to achieve tumor suppression (3, 5–8). Human benign nevi (commonly known as moles) form due to a clonal hyperproliferation of melanocytes harboring an activated BRAF or N-RAS oncogene (9–11). However, melanocyte hyperproliferation is ultimately arrested, and progression toward melanoma is avoided due to oncogene-induced senescence (OIS) (3, 12, 13). Nevus melanocytes are senescent, as judged by expression of several hallmarks and effectors of senescence, namely, tumor suppressor p16INK4a (p16), histone variant macroH2A, DNA damage signaling effector 53BP1, and senescence-associated β-galactosidase (SA β-gal) (3, 12, 13). Even so, about 25% of melanomas are thought to arise in association with a preexisting nevus (14, 15). Inactivation of PTEN and p16 and activation of Wnt signaling are each thought to contribute to nevus to melanoma progression (16–19). Here we investigated senescence in melanocytes in vitro and in nevi, with a view to better understand both nevus formation and progression to melanoma.

The Wnt signaling pathway controls gene expression programs that regulate cell fate and morphogenesis and is often deregulated in cancer. In the canonical Wnt signaling pathway, extracellular Wnt proteins bind to their cognate transmembrane receptors, composed of Frizzled proteins and LRP5/6 coreceptors (20). This indirectly stabilizes β-catenin, which enters the nucleus to complex with the DNA-binding protein TCF4 and drive expression of proliferative genes, such as C-MYC and CYCLIN D1. Two proteins, Axin and the Adenomatous Polyposis Coli (APC) protein, antagonize the signaling pathway by promoting phosphorylation and destruction of β-catenin. Wnt signaling is frequently activated in human cancers by mutations in the protooncogene; β-catenin; or the tumor suppressor genes, APC and AXIN.

Wnt signals are important for proliferative expansion and differentiation of melanocyte precursors (21–23). Roughly 30% of melanoma cell lines and melanomas have activated Wnt signaling, as judged by accumulation of active nuclear β-catenin (24, 25). However, mutations of APC or β-catenin are relatively uncommon in melanoma, so it is hypothesized that the activation of canonical Wnt signaling in melanoma is driven by methylation and silencing of APC, overexpression of the protooncoprotein SKI, or the increased presence of Wnt ligands, presented either by neighboring cells or tumor cells themselves (24, 26). Forced expression of β-catenin in melanoma cells drives cell proliferation (27), and expression of β-catenin and/or β-catenin transcriptional targets is required for proliferation of melanoma cell lines (28). In the mouse, activated Wnt signaling promotes melanoma metastasis to the lymph nodes and lungs (29, 30). However, many human benign nevi are positive for nuclear β-catenin, and expression decreases on progression to metastatic melanoma (31). Other studies also suggest that canonical Wnt signaling antagonizes proliferation and metastasis of advanced melanoma, and patient survival is improved in those cases with elevated canonical Wnt signaling (32). In sum, the role of Wnt signaling as oncogenic or tumor-suppressive in melanoma is likely disease stage- and context-dependent (32).

Previously, we reported that activated Wnt signaling can delay senescence of primary human fibroblasts in culture (33). Moreover, Larue and coworkers (18) demonstrated that activation of Wnt signaling in mouse melanocytes is able to bypass senescence by β-catenin–mediated repression of the tumor suppressor, p16. Here we have further analyzed the regulation and function of Wnt signaling in senescent human melanocytes in vitro, in human tissues, and in a mouse model. Our results suggest that Wnt signaling can enhance nevogenesis in vivo by delaying the onset of melanocyte senescence.

Results

To better understand senescence of primary human melanocytes, we compared gene expression of control-infected primary human melanocytes and melanocytes made senescent by lentiviral transduction of an activated BRAF^V600E mutation commonly found in nevi (11) (Fig. 1A). Senescence of BRAF^V600E-expressing cells was confirmed by flattened morphology, cell cycle exit (repression of cyclin A, suppression of 5′-BrdU incorporation, and pRB hypophosphorylation), expression of Dec1 (6), appearance of senescence-associated heterochromatin foci (SAHF) (34), and expression of SA β-gal (Fig. 1 A–C). Microarray-based gene expression analysis showed that in senescent cells, expression of 4,370 and 5,215 probes were significantly up- and down-regulated, respectively (>1.5-fold change compared with empty vector, P value < 0.001 Benjamini-Hochberg false discovery rate (BH-fdr); Dataset S1). Consistent with previous studies in other cell types, gene set enrichment analysis (GSEA) showed that the most down-regulated gene sets were those involved in cell cycle progression and the most up-regulated sets were those involved in extracellular and inflammatory signaling, collectively composing the SASP (4) (Fig. S1 A and B). Hierarchical clustering of specific gene sets and families confirmed global repression of proliferation-associated genes (35) and up-regulation of many genes in Gene Ontology (GO) group “Inflammatory Response” and documented SASP genes (4) (Fig. 1 D and E, Fig. S1C, and Tables S1–S3). To verify these data, we performed RNA sequencing (RNA-seq). Globally, there was a strong correlation between microarray and RNA-seq data (Fig. S1D), and analysis of selected genes in the RNA-seq data confirmed repression and activation of proliferation and growth arrest genes, respectively, e.g., CYCLIN A2 and P15INK4B (Fig. 1F).

Fig. 1.

Expression of proliferation genes, but not repression of inflammation genes, is common to malignant melanoma cells that have evaded senescence. (A–C) Human epidermal melanocytes were infected with a lentivirus encoding BRAF^V600E and equivalent empty vector. Two weeks postinfection, assays were performed: (A) Western blotting. (B) DAPI staining to detect SAHF (including magnified outset), immunofluorescence to measure S phase after a 16-h pulse with 5′-BrdU, and SA β-gal activity. (Scale bar, 50 μm.) (C) Quantitation of B. (D and E) Microarray expression analysis of empty vector and BRAF^V600E-infected melanocytes: (D) heat map (red, up-regulated genes; green, down-regulated genes) of entire proliferation gene set extracted from (35) and (E) heat map of all changed genes (fold change > 1.5, BH-fdr < 0.05) from GO term Inflammatory Response. (F–H) RNA-seq analysis of uninfected melanocytes, BRAF^V600E-infected melanocytes, and seven uninfected melanoma cell lines: M1–M7 defined in Fig. S1E. (F) Aligned RNA-seq reads from uninfected and BRAF^V600E-infected melanocytes at indicated genes, (G) heat map of proliferation genes vertically ordered identical to D, and (H) heat map of all genes from GO term Inflammatory Response. Bracket indicates genes that show increased expression in both senescent cells (BRAF^V600E) and melanoma cell lines compared with uninfected melanocytes.

Both proliferation arrest and the SASP have been suggested to contribute to senescence-mediated tumor suppression (4–8). Therefore, we set out to ask whether malignant melanomas typically evade proliferation arrest and/or the SASP, as judged by these gene expression signatures. We used RNA-seq to compare gene expression in proliferating and senescent melanocytes with a panel of seven human immortal melanoma cell lines each harboring an activated BRAF^V600E or N-RAS^Q61K oncogene (Fig. S1E). Because these lines are immortal, they have, by definition, evaded senescence (either by escape or bypass). Virtually all of the proliferation-promoting genes that were down-regulated in senescence were comparably expressed in proliferating melanocytes (Unin) and melanoma cell lines (Fig. 1G). In contrast, many of the inflammatory response and SASP genes that were up-regulated in senescence were also more highly expressed in the melanoma cell lines than the proliferating melanocytes (Fig. 1H and Fig. S1F). Documented SASP genes expressed in the melanoma cell lines included IGFBP3, IL8, CCL20, CXCL2, and IL1β (4) (Table S3 and Fig. S1F). Based on these results, we conclude that primary human melanocytes expressing BRAF^V600E in vitro undergo proliferation arrest and express the SASP. Activation of cell proliferation, but not suppression of the SASP, is common to malignant melanoma cells that have evaded senescence.

In light of the complex and likely stage-dependent role of Wnt signaling in melanomagenesis (32), we next considered the role of Wnt in control of proliferation genes and SASP genes in senescent melanocytes. Many direct Wnt target genes were activated in BRAF^V600E-expressing senescent melanocytes, including designated SASP or SASP-like factors, such as FGF family members, FST (a TGFβ regulator), MMP7, and VEGF (4) (Fig. 2A, Fig. S1G, and Table S4). However, these changes are likely due to other signaling pathways because the levels of both activated (unphosphorylated) β-catenin (ABC) and total β-catenin protein and β-catenin mRNA were decreased in OIS (Fig. 2B and Fig. S1H) (36). Moreover, expression of key proliferation-promoting Wnt target genes, such as CYCLIN D1 and C-MYC (Fig. 2 A, C, and D, Fig. S1I, and Table S4), was also decreased in senescent cells. Expression of MITF was also repressed in senescent melanocytes (Fig. 2 A, C, and D, Fig. S1I, and Table S4). MITF has been reported to have both positive and negative effects on melanocyte proliferation, yet it is often amplified or a target of gain-of-function mutation in melanomas, suggesting that its proliferation-promoting effects can be dominant (37–39). Based on these data, we conclude that β-catenin–dependent Wnt signaling and expression of key proliferation-promoting Wnt targets, cyclin D1, c-myc, and MITF, are repressed in BRAF^V600E-induced senescence in vitro.

Fig. 2.

Repression of proliferation-promoting Wnt target genes is tightly linked to senescence in vitro. (A) Heat map (red, up-regulated genes; green, down-regulated genes) of direct Wnt target genes after microarray analysis of empty vector and BRAF^V600E-infected melanocytes. Gene list extracted from www.stanford.edu/group/nusselab/cgi-bin/wnt/target_genes and filtered through primary literature search for genes with functional TCF/LEF binding sites in gene promoter (Table S4). (B) Wnt signaling is decreased upon expression of BRAF^V600E. Melanocytes were lentivirally infected and analyzed at indicated days postinfection. ABC, activated (unphosphorylated) β-catenin; B, BRAF^V600E; E, empty vector. (C and D) Validation of down-regulated Wnt targets (cyclin D1, c-myc, and MITF) upon BRAF^V600E-induced senescence: (C) by Western analysis and (D) by quantitative real-time PCR (qRT-PCR). (E) Down-regulated Wnt targets upon NRAS^Q61K-induced senescence. (F–I) Melanocytes lentivirally transfected with dnTCF and control empty vector: (F) dnTCF expression by Western; (G) dnTCF inhibits expression of cyclin D1, c-myc, and MITF, measured by qRT-PCR; (H) quantitation of BrdU incorporation and cyclin A immunofluorescence (representative images in Fig. S1 J and K); and (I) SA β-gal activity after inhibition of Wnt signaling by dnTCF. (Scale bar, 100 μm.)

Mutated N-RAS^Q61K is also commonly found in benign nevi and melanoma (10, 11). BRAF and N-RAS mutations are typically mutually exclusive, and N-RAS is an upstream activator of BRAF, suggesting that BRAF and N-RAS mutations mediate similar, but not necessarily identical, effects to promote nevus formation. As expected, cyclin D1 and c-myc were also repressed in N-RAS^Q61K–induced senescence in vitro (Fig. 2E).

Consistent with a causative role for repression of Wnt signaling driving a senescence-associated proliferation arrest, expression of an inhibitor of Wnt-mediated transcription, dominant negative (dn) TCF, repressed expression of cyclin D1, c-myc, and MITF and induced a proliferation arrest in proliferating human melanocytes (Fig. 2 F–H and Fig. S1 J and K). Moreover, these cells showed flattened morphology and expressed SA β-gal, two hallmarks of senescence (Fig. 2I). Together, these results suggest a role for repression of proliferation-promoting Wnt target genes in enforcing the senescent phenotype.

Given these data and the fact that human benign nevi contain senescent melanocytes (3, 12, 13), we were surprised by previous reports suggesting that human nevi express proliferation-promoting Wnt targets, such as cyclin D1, c-myc, and MITF (40–42). However, we confirmed that many melanocytic nests of human benign nevi exhibited coincident expression of nuclear β-catenin and its proliferation-promoting targets, cyclin D1 and c-myc (Fig. 3A and Fig. S2A). Localization of activated Wnt signaling and expression of cyclin D1 and c-myc was linked to nevus maturation, an important histological parameter, reflected in decreased size and more spindle-shaped morphology of melanocytes deeper within a nevus (Fig. 3B) (43, 44). Maturation is associated with enhanced proliferation arrest because nevi can contain some mitotic cells, typically in the least mature upper portion of the nevus (45). Most important, maturation is an indicator of reduced malignant potential (43, 44). Although less mature, the upper part of the nevus is judged to be in a senescent-like state, based on widespread expression of p16 (3, 12, 13) (Fig. S2B). Expression of nuclear β-catenin, c-myc, and cyclin D1 was more marked in the upper, less mature portion of the nevus (Fig. 3 B–D). Together, these data indicate that incomplete maturation of nevus melanocytes correlates with activated β-catenin and expression of proliferation-promoting Wnt target genes whose repression is tightly linked to senescence-associated proliferation arrest in vitro.

Fig. 3.

Human benign nevi express proliferation-promoting Wnt targets inverse to nevus maturation. (A) Serial sections of a nevus stained for melan A, β-catenin, cyclin D1, and c-myc. (Scale bar, 100 μm.) (B) Cyclin D1, c-myc, and nuclear β-catenin preferentially stain the upper least mature portion of the nevus. Basement membrane is marked with a yellow dotted line in Nevus 3. (Scale bar, 30 μm.) (C) Quantitation of nevus staining comparing the topmost (least mature) and bottommost (most mature) 20% areas of nevus. Adjacent sections were scored in five nevi. Nevus tissue was defined based on melan A stain. P values (Cochran-Mantel-Haenszel test) are as follows: *P < 1 × 10⁻⁷⁰ and **P < 1 × 10⁻²⁰⁰. (D) Magnified image of boxed areas in B reveals a clear loss of nuclear β-catenin with maturation. In merge, DAPI is shown in blue, and β-catenin is shown in red. (Scale bar, 30 μm.)

Surprisingly, closer analysis of normal human skin lacking nevi also revealed that a subset of nonnevus melanocytes, in their usual location scattered along the basement membrane of the skin interfollicular epidermis, also stained positive for p16. This was initially suggested by p16-expressing cells adjacent to the basement membrane, exhibiting morphology characteristic of melanocytes (Fig. 4A), and was confirmed by two-color immunofluorescence staining of p16 and S100 (also expressed in melanocytes) in the same sections (Fig. 4B) and staining adjacent FFPE tissue sections with p16 and melan A (a marker of melanocytes) (Fig. 4C). These epidermal p16-positive melanocytes invariably failed to express Wnt targets, cyclin D1 and c-myc (Fig. 4C and Fig. S2 C and D). This was confirmed by scoring p16-, cyclin D1-, and c-myc–positive epidermal cells adjacent to basement membrane of normal skin, as a percentage of melan A-positive cells. We consistently detected p16-positive cells but never detected cyclin D1- or c-myc–positive cells in this location in normal skin (Fig. 4D). This phenomenon was particularly vivid in epidermal melanocytes overlaying dermal nevus melanocytes: the latter expressed p16 and cyclin D1, whereas the former expressed only p16 (Fig. S2E). We conclude that proliferation-promoting Wnt target genes are expressed in p16-positive nevus melanocytes with a prior history of proliferative expansion but not in isolated interfollicular epidermal p16-positive melanocytes.

Fig. 4.

Isolated p16-expressing epidermal melanocytes do not express proliferation-promoting Wnt targets. (A) Normal human skin stained for p16 reveals cells that resemble melanocytes [cells with dendritic projections adjacent to the basement membrane (BM)]. (Scale bar, 10 μm.) (B) Immunofluorescence of normal human skin reveals that some S100-postive epidermal cells adjacent to BM (melanocytes) are p16-positive. Basement membrane is marked with a yellow dotted line. (Scale bar, 10 μm.) (C) Serial sections of normal human skin show that many epidermal isolated interfollicular melan A-positive melanocytes along the BM are positive for p16 but not cyclin D1. Red arrows in melan A stain indicate melanocytes. (Scale bar, 50 μm.) (D) The number of p16-, cyclin D1-, and c-myc–positive epidermal cells adjacent to the BM was scored in three adjacent sections of normal human skin. Values are expressed as the percent relative to the number of melan A-positive cells in a fourth section of skin. Note that this method of scoring shows that p16-expressing cells can be found in the epidermis adjacent to the basement membrane, but cyclin D1- and c-myc–positive cells are never found in this location in normal skin. This method is essentially a frequency comparison and does not assume that the same cells are being scored in each section. p16 compared with cyclin D1 and c-myc, P < 0.001.

Accordingly, we wanted to test the impact of activated Wnt signaling on proliferation of oncogene-expressing melanocytes in vivo. To do this, we used mice expressing an activated N-RAS oncogene in melanocytes, under control of a tyrosinase promoter [Tyr-N-RAS^Q61K (46)]. Although human nevi are not thought to contain activating mutations in the Wnt pathway, activation of Wnt signaling was modeled here by homozygous inactivation of APC using Tyr-Cre and homozygous APC^fl.fl alleles (26, 47).

As reported previously (46), Tyr-N-RAS^Q61K mice exhibited an excess of S100-positive melanocytes in the dermis and became hyperpigmented with melanin within a few days of birth, compared with wild-type (WT) mice (and also Tyr-Cre APC^fl.fl mice) (Fig. 5 A and B and Fig. S3A). Consequently, Tyr-N-RAS^Q61K mice also showed darkened coat color and extremities, notably ears, nose, feet, and tail (Fig. 5C). Activation of Wnt signaling, by genetic inactivation of APC in Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K mice, exacerbated the N-RAS^Q61K–induced proliferative expansion of S100-positive melanocytes and skin melanization (Fig. 5 A and B and Fig. S3A). The percent melanization of the dermis in both genotypes eventually reached a plateau (Fig. 5B); however, the plateau was significantly higher in the Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K mice compared with the Tyr-N-RAS^Q61K mice. This suggested that activation of Wnt signaling delayed OIS-associated proliferation arrest.

Fig. 5.

Activation of Wnt signaling delays senescence of melanocytes in vivo. (A) Haematoxylin and eosin staining of the indicated mouse skin shows the presence of melanin (brown) produced by melanocytes. Immunofluorescence shows S100-expressing melanocytes. Representative mice aged 55–82 d are shown. BM, basement membrane; PC, panniculus carnosus. (Scale bar, 100 μm.) (B) The proportion of melanin-containing dermis over time. Color-coded lines display the logarithmic trend of each genotype. Pairwise comparison P values, Bonferroni adjusted, are *P < 1.5 × 10⁻², **P < 4.8 × 10⁻¹⁶, and ***P < 9.1 × 10⁻⁵. Logarithmic R² values are WT = 0.484, Tyr-Cre APC^fl.fl = 0.144, Tyr-N-RAS^Q61K = 0.682, and Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K = 0.644. (C) Coat color of representative mice aged 55–82 d. (D) Immunofluorescence of young (2–7 d) and adult (>70 d) mice for markers of proliferation (Ki67) and senescence (p16). (Scale bar, 25 μm.)

Consistent with ultimate proliferation arrest in both genotypes, although the dermis of young (2–7 d) Tyr-N-RAS^Q61K and Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K mice contained an excess of Ki67-expressing cells compared with WT mice, in adult (>70 d) mice of both genotypes the Ki67-positive cells were largely depleted (Fig. 5D). Also consistent with a finite proliferative capacity of melanocytic cells in the Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K mice, these mice showed marked hair greying and reduction in melanin in the hair follicles (Fig. 5C and Fig. S3B). Given that melanocytes of these mice can differentiate normally, based on expression and production of S100 and melanin (Fig. 5 A and B), this can be explained by depletion of pigment-producing melanocytes and/or their stem or progenitor cells in the hair follicle, and/or failure of these cells to migrate into or persist in the hair follicle. Either way, the Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K melanocytes do not appear to be immortal in vivo. In fact, nonproliferating melanocytes of adult mice of both genotypes were p16-positive (Fig. 5D), indicative of a senescent-like phenotype.

Based on histological analysis, the dermal melanocytes of Tyr-N-RAS^Q61K and Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K were not transformed, malignant, or invasive. The proliferating melanocytes in young mice were restricted to the dermal area between the basement membrane and the panniculus carnosus (the layer of skeletal muscle underlying the papillary dermis) (Fig. 5 A and D). In adult mice, the melanocytes were restricted to the dermis and showed no signs of anaplasia (dedifferentiation) (Fig. 5A). There were no unusual masses nor signs of ulceration (Fig. 5C). There was no disruption, invasion, nor sign of pressure on the basement membrane and panniculus carnosus (Fig. 5A). Hair follicles were normal in overall structure (Fig. S3B). To date, no melanomas have been observed in Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K mice of up to 4.5 mo of age. Based on these results, we conclude that activated Wnt signaling can promote transient proliferative expansion of oncogene-expressing melanocytes in vivo, thereby generating a larger population of, ultimately proliferation-arrested, senescent melanocytes. In these respects, these melanocytes are similar to those found in nevi (3, 12, 13, 19).

Discussion

We have characterized OIS (N-RAS^Q61K and BRAF^V600E) of primary human melanocytes in vitro and compared it to transformed melanoma cell lines in vitro and senescent melanocytes in vivo. Not surprisingly, compared with senescent cells, all melanoma cell lines expressed relatively higher levels of genes associated with cell proliferation. In contrast, senescent cells and melanoma cell lines expressed comparable levels of some inflammatory genes that compose the SASP. It is tempting to speculate that the different behavior of proliferation genes and SASP genes in this analysis stems from the fact that proliferation arrest is likely always tumor-suppressive, whereas an inflammatory phenotype can be tumor-suppressive or oncogenic depending on cell and tissue context (4, 8, 48).

By this view, it was not surprising to observe repression of Wnt signaling and proliferation-promoting Wnt target genes in senescent melanocytes in vitro. Moreover, the ability of dnTCF to inhibit expression of these genes and induce a senescence-like phenotype is consistent with the idea that repression of these genes contributes to the OIS-associated proliferation arrest. However, the view that cell proliferation arrest, and hence repression of Wnt signaling, is tightly linked to senescence-mediated tumor suppression is more difficult to reconcile with the expression of β-catenin and proliferation-promoting Wnt target genes in human benign nevi, tumor-suppressive neoplasms containing senescent melanocytes (3, 12, 13). Therefore, we went on to investigate the impact of activated Wnt signaling (modeled through genetic inactivation of APC) on OIS in a mouse model. In this model, we found that activated Wnt signaling delayed, but did not acutely bypass, OIS. Consequently, activated Wnt signaling ultimately resulted in a large expansion of the number of oncogene-expressing nonproliferating senescent melanocytes. Thus, Wnt signaling is active and proliferation-promoting Wnt targets are expressed in at least some parts of human benign nevi. This activated Wnt signaling is able to delay onset of OIS to allow for nevogenesis (see model, Fig. S4).

These findings can help to explain some aspects of nevus biology. For example, a nevus can contain tens to hundreds of thousands of clonal melanocytes (9). Although OIS in vitro is preceded by a brief oncogene-induced proliferative burst (49, 50), Wnt-dependent transient bypass of senescence might facilitate this proliferative expansion in vivo, leading to the relatively large size of human nevi. This could be particularly important in congenital nevi that form during development in utero. Specifically, Wnt signals that melanoblasts receive during their migration out of the neural crest may contribute to the delay of senescence and growth of nevi in utero (21–23). By extension, developmentally linked Wnt signals might contribute to preferential nevus formation in young children and adolescents (51, 52), although other factors, such as declining telomere length, are also likely to be responsible for decreased proliferative potential and nevogenesis with age (13, 52).

We observed another population of isolated, nonnevus, interfolicular p16-expressing melanocytes in their canonical location along the basement membrane of the epidermis. In contrast to nevus melanocytes, these cells always appeared to lack expression of proliferation-promoting Wnt target genes. Significantly, previous studies have identified similarly positioned p16-expressing cells in the epidermis, morphologically resembling melanocytes, whose abundance increases with age (53, 54). The trigger of p16 expression in these epidermal melanocytes, be it age-associated acquisition of an activated oncogene or another mode of stress, and the relationship, if any, of these cells to melanoma remains to be investigated. Regardless, the existence of these isolated p16-expressing epidermal melanocytes underscores the link between expression of proliferation-promoting Wnt targets and expansion of oncogene-expressing cells to form a p16-positive nevus.

Many human melanomas exhibit activated Wnt signaling (18, 24, 25), and Larue and coworkers previously showed that a stabilized β-catenin allele drives escape from N-RAS^Q61K–induced senescence and ultimately progression to melanoma (18). To date, we have not observed melanoma in any of our Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K mice. However, due to debilitating behavioral issues, perhaps due to Tyr-Cre–mediated inactivation of APC in other neural crest-derived cells, Tyr-Cre APC^fl.fl/Tyr-N-RAS^Q61K mice must typically be culled by 4 mo of age. This is likely insufficient time to acquire additional genetic and/or epigenetic alterations that are presumably required to fully escape senescence, such as inactivation of p16 or PTEN (18, 30). However, it seems likely that persistent activated Wnt signaling and expression of proliferation-promoting Wnt targets in nevi will potentiate their malignant progression. In humans, this may contribute to the recognized risk of nevi for progression to melanoma and reports that ∼25% of melanomas arise from a preexisting nevus (14, 15). Underscoring the possible link between Wnt and neoplastic progression of nevi, we find that Wnt signaling is preferentially active in the least mature, more proliferative component of the nevus that appears to be of greater malignant potential (43–45). Recently, we also reported that nevus maturation is associated with marked changes in nuclear chromatin structure (55). In aggregate, we suggest that nevus maturation is associated with repression of Wnt signaling and chromatin changes, perhaps indicative of a more complete or “deeper” senescence program and thus lower malignant potential.

In sum, our data, together with previously published works, suggest a dual role for Wnt signaling in nevogenesis and melanoma. On one hand, activated Wnt signaling can delay senescence in vivo to promote growth of a nevus-like lesion containing oncogene-expressing senescent melanocytes. On the other hand, activated Wnt signaling can promote nevus progression to melanoma. Surprisingly, most human benign nevi appear to contain some melanocytes harboring activated Wnt signaling. Activated Wnt signaling in human nevi is expected to undermine senescence-mediated tumor suppression and enhance the chance of malignancy. Human benign nevi are the paradigm for OIS as a tumor suppressor mechanism (3, 12, 13). However, further studies are required to achieve a full understanding of effective tumor suppression versus progression to malignancy in these neoplasms.

Materials and Methods

Details of materials and methods are available in SI Materials and Methods.

Microarray and RNA-seq.

Microarray samples were processed as described (SI Materials and Methods). RNA-seq samples were prepared according to manufacturer instructions, sequenced using the Illumina GAIIX sequencer, and aligned to the human genome. The sequence was submitted to Gene Expression Omnibus (GEO) as GSE46818.

Cell Culture, Plasmids, and Lentiviral Transduction.

Human epidermal neonatal melanocytes were grown in Medium 254 supplemented with human melanocyte growth supplement. HEK293T cells and all melanoma cell lines were maintained in DMEM supplemented with 10% FBS (vol/vol), l-glutamine, and antibiotics. Vector HIV-CS-CG-BRAF^V600E-puro and its equivalent empty vector, HIV-CS-CG-puro (3), and vector EF1alpha-dnhTCF4/SV40-PuroR (Addgene) and its equivalent empty vector, EF1alpha/SV40-PuroR, were used for lentiviral gene transfer.