Accumulation of distinct sets of genetic/epigenetic alterations is thought to contribute to stepwise progression of human cutaneous melanomas. We found evidence of frequent tumor cell autonomous transforming growth factor-β (TGF-β) signal activation in both premalignant and malignant stages of human cutaneous melanoma histogenesis and investigated its potential causative roles using human organotypic skin cultures. PTEN deficiency and Braf activation, two common coincident genetic alterations found in primary cutaneous melanomas, were first introduced into human melanocytes previously immortalized by the SV40 large T antigen and telomerase. These changes individually supported anchorage-independent growth and conferred benign, hyperplastic growth in a skin-like environment. In addition, PTEN deficiency combined with Braf activation together induced a melanoma in situ–like phenotype without dermal invasion. Further addition of cell autonomous TGF-β activation in the context of PTEN deficiency and Braf activation promoted dermal invasion in skin cultures without significantly promoting proliferation in vitro and in vivo. This proinvasive phenotype of cell autonomous TGF-β activation is genetic context–dependent, as hyperactivating the TGF-β type I receptor without PTEN deficiency and Braf activation failed to induce an invasive behavior. Evidence of genetic interactions among PTEN deficiency, Braf activation, and cell autonomous TGF-β activation shows that distinct stages of human melanoma are genetically tractable in the proper tissue architecture. [Cancer Res 2008;68(11):4248–57]
- Melanoma/skin cancers
- Premalignant lesions
- Signal transduction pathways
- Tumor suppressor genes
Human cutaneous melanoma is a heterogeneous malignancy, the incidence of which has been increasing over the last three decades ( 1). Human melanoma progresses through distinct stages of premalignancy and malignancy. Premalignant lesions include clinicopathologic entities, termed lentigo, nevi, or dysplastic nevi. The key malignant transition is thought to occur from radial growth phase melanoma (i.e., melanoma in situ) to vertical growth phase melanoma (i.e., invasive melanoma; refs. 2, 3). Radial growth phase melanoma is characterized by the loss of normal melanocytic growth restriction to the basal layer of the epidermis, resulting in intraepidermal growth known as pagetoid spread. Further clonogenic evolution can lead to vertical growth characterized by expansile infiltration of the dermis and its constituent structures, such as blood and lymphatic vessels, presaging distant metastasis.
Heterogeneous patterns of DNA copy number alterations and gene mutations group human primary cutaneous melanomas into genetic and clinical subtypes ( 4– 6). For instance, DNA amplifications and/or activating mutations in NRas, Braf, and c-Kit are generally found in primary human melanomas in a mutually exclusive fashion. Of these genetic abnormalities, Braf mutations are significantly associated with focal PTEN chromosomal loss. Another deletion, which is found commonly in both sporadic and familial melanomas, involves the CDKN2A locus, which encodes two tumor suppressors, p16INK4A and p14ARF, that respectively participate in the Rb and p53 pathways. Other altered pathways, such as up-regulation of transforming growth factor-β (TGF-β)–related ligands (e.g., Nodal), are selected during melanoma progression without an overt or known genetic event ( 7). How combinations of these abnormalities contribute to the progressive development of primary melanoma de novo remains largely unanswered. Addressing this important question requires understanding the behaviors of altered melanocytes in their proper tissue architectural context.
Most human melanomas arise in the skin, which provides important heterotypic cellular and matricellular interactions exerting non–cell autonomous homeostatic growth control over melanocytes ( 8). Traditional assays, such as soft agar growth or subcutaneous growth in immune-defective mice, have important roles in dissecting distinct aspects of cellular transformation but lack human skin tissue architecture. Genetically engineered murine models of human melanoma have provided key insights into melanoma pathogenesis ( 2) but suffer from one major drawback: adult murine melanocytes are dermal and follicular in location in contrast to normal adult human melanocytes, which reside in the follicular and interfollicular epidermis ( 9).
In the present work, we found evidence of frequent Smad2 pathway activation in clinical specimens of benign and malignant cutaneous melanocytic neoplasms, suggesting that this tumor cell autonomous pathway is hyperactivated in response to autocrine and/or paracrine ligand activity. The presence of activated nuclear Smad2 in the benign, hyperplastic stage of human melanocytic neoplasia suggests that its potential pathogenic role(s) may require other genetic or epigenetic alterations. Using a model system of reconstructing human melanoma initiation/progression in a human skin organotypic culture assay, we explored the functional consequences of constitutively activating TGF-β signaling in genetically engineered human epidermal melanocytes.
Materials and Methods
Tumor specimen analysis. Formalin-fixed, paraffin-embedded benign nevi, cutaneous melanoma in situ, and primary invasive cutaneous melanomas were obtained from diagnostic and therapeutic procedures at our institution. Sections containing tumor cells and stroma are used for immunohistochemical analysis of phosphorylated Smad2 using an anti–phosphorylated Smad2 (Ser465/467) antibody (Cell Signaling Technology). Studies were conducted under a protocol approved by the University of California at Los Angeles Institutional Review Board.
Lentiviral vectors and gene transfers. Braf (V600E) was subcloned into the pENTR1A entry vector (Gateway, Invitrogen) and transferred into the pLenti6/UbC/V5-DEST vector (Gateway, Invitrogen) by recombination following the manufacturer's recommendations. Five putative human PTEN short hairpin RNAs (shRNA) in pLK0.1 (Open Biosystems) were tested for their knockdown efficacy by Lipofectamine (Invitrogen)–mediated transient transfection into Mel-ST (parental) immortal melanocytes. The top two candidates were selected for virus production and stable knockdown. TβRI AAD-HA was subcloned into FUCRW. Virus stocks were produced by cotransfecting 293T HEK cells with the lentiviral vectors, along with three packaging vectors. Virus-infected melanocytes were negatively selected either by antibiotic killing (blasticidine for pLenti6 and puromycin for pLK0.1) or positively selected by fluorescence-activated cell sorting (FACS) based on RFP fluorescence (FUCRW).
Anchorage-independent growth. Soft agar growth was performed in 48-well flat-bottomed plates in quadriplicates. Each well was layered first with 0.8% acellular agarose in Iscove's media followed by 0.4% agarose in Iscove's media containing 1,000 cells per well. Serial photographs of the wells were then taken over 18 d (2.5×). At 18 d, the cultures were treated with Cell Quantification Solution (Chemicon International) according to the manufacturer's recommendations. The resultant quantifications of absorbance at A490 were shown as the means ± SEs of the quadriplicates. The soft agar experiment was repeated twice with comparable results.
Cell proliferation. One thousand cells per well in 96-well plates were plated in quadriplicates and incubated in either DMEM with 10% fetal bovine serum (FBS) or 0.2% FBS for 5 or 6 d. Each day after seeding, the cells were quantified using CellTiter 96 Aqueous One Solution (Promega) following the manufacturer's recommendations.
Cell cycle and apoptosis analysis. Cell cycle profiles and apoptosis were analyzed by FACS analysis. Briefly, cells were fixed in 90% ETOH and stained with propidium iodide. The sub-G1 fraction represented a subpopulation of apoptotic cells. Staurosporine (1 mmol/L) in DMEM–10% FBS was used to induce apoptosis acutely (150 min), and incubation of cells in DMEM–0.2% FBS for 72 h was used to induce apoptosis chronically.
Cell culture and organotypic human skin culture. All polyclonal melanocyte cell lines and human dermal fibroblasts (Cell Applications) were maintained in DMEM with 10% FBS. Human epidermal keratinocytes (Cell Applications and Cambrex) were maintained in fully supplemented KGM media (Cambrex). For organotypic human skin reconstruction based on a human dermal fibroblast–contracted collagen I matrix, an acellular layer of bovine type I collagen (1 mg/mL; Organogenesis) was pipetted into each insert of tissue culture tray (Organogenesis) followed by human dermal fibroblast–containing bovine type I collage (8 × 104 cells/mL). The collagen matrix is then allowed to contract in DMEM–10% FBS for 7 d at 37°C. Human epidermal keratinocytes were mixed with engineered melanocytes at a ratio of 10:1 in keratinocyte serum-free medium (Invitrogen) containing 2% dialyzed FCS, 60 mg/mL bovine pituitary extract, 5 ng/mL human basic fibroblast growth factor, 100 nmol/L human endothelin-3, and 10 ng/mL human stem cell factor. A total of 1 × 106 cells were seeded onto each contracted collagen gel. Cultures were covered in medium containing 1 ng/mL human epidermal growth factor (EGF) for 2 d and 0.2 ng/mL human EGF for two additional days and were then raised to the air-liquid interface via feeding from below with high calcium (2.4 mmol/L) medium. After 14 to 15 d, skin reconstructs were fixed in 10% buffered formalin, embedded in paraffin, and subjected to sectioning and staining. For organotypic human skin cultures based on devitalized intact human dermis (Alloderm), frozen-dried dermis is first rehydrated in PBS at 37°C for 1 h, cut into squared pieces to fit inserts of the skin culture tray, and placed into the inserts with the basement membrane side facing up. Subsequent seeding of primary epidermal keratinocytes and engineered melanocytes were performed as described earlier, except that cultures were harvested 11 to 12 d after reconstruction. All skin culture experiments were repeated twice with similar results.
Matrigel/collagen invasion assay. Cell were trypsinized, washed twice with PBS, resuspended in DMEM–2% FBS, and seeded at 5 × 105 per well in 24-well–sized Matrigel invasion chambers or collagen I/collagen III (85%:15%) invasion chambers in triplicates (Chemicon). DMEM–10% FBS served as the chemoattractant in the bottom wells. Seeding on regular 24-well culture plates in duplicates served as seeding input controls, which were quantified by incubation with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide solutions followed by measuring optical absorbance at 490 nm (Promega). After 20 to 24 h of incubation at 37°C, cells migrated through the chambers were fixed and stained according to the manufacturer's recommendations (Chemicon) and counted under the microscope (average of five representative fields).
ELISA. Conditioned media were prepared by plating 1 × 106 cells in DMEM–10% FBS into six-well plates in duplicates. At >90% culture confluence, the original medium was switched to DMEM–1% FBS for an additional culture of 24 h. The resultant conditioned medium was then collected and, after centrifugation to removed cellular contaminants, assayed for levels of extracellular factors by Searchlight Arrays (Pierce Biotechnology) in triplicates of each sample. The results are representative of two separate collections of conditioned media.
Gelatin zymography. Conditioned medium was mixed with Novex Tris-glycine SDS sample buffer (Invitrogen) and subjected to electrophoresis on 10% gelatin gel (Invitrogen). After electrophoresis, the gel was incubated sequentially in Zymogram Renaturing Buffer (Invitrogen) and Zymogram Developing Buffer (Invitrogen) and stained with SimplyBlue Safestain (Invitrogen).
Spheroid invasion assay. Spheroids derived from engineered melanocytes seeded in ultra low attachment plates (Sigma) were sandwiched in two sequential layers of collagen I (1 mg/mL; Organogenesis) matrix in DMEM–5% FBS. Invasion of cells from the spheroids were then followed by phase contrast microscopy over time.
Image acquisition/processing and data processing. Statistical analyses were performed using InStat 3 Version 3.0b (GraphPad Software), and graphical representations of results were performed using DeltaGraph (SPSS, Inc. and Red Rock Software). An Optronics camera system was used in conjunction with MagnaFire 2.1C software for image acquisition and Adobe Photoshop 7.0 for processing.
TGF-β–related factors, including TGF-β itself, Nodal, and activin, are commonly overexpressed in melanomas ( 7, 10). These factors activate specific cell surface receptors that funnel signaling into the shared Smad2/Smad3 pathway. To characterize the intracellular activation state of the TGF-β pathway during primary human melanoma progression, we surveyed a collection of human benign and malignant cutaneous melanocytic neoplasms for the presence of receptor-phosphorylated Smad2 or p-Smad2. Using an anti–p-Smad2 antibody previously validated in immunohistochemistry ( 11, 12), we found strong, heterogeneous nuclear p-Smad2 in all benign and malignant melanocytic neoplasms ( Fig. 1 and Supplementary Figs. S1 and S2). Interestingly, nuclear p-Smad2 was present in the epidermal keratinocytes overlying the tumors ( Fig. 1C3) but absent in the distally located epidermal keratinocytes ( Fig. 1C4), suggesting the presence of a tumor-centric gradient of TGF-β and related factors that may act on susceptible, nonneoplastic cell types.
To evaluate the role(s) of cell autonomous TGF-β signaling during human melanoma progression, we first sought to introduce two common, coincidental genetic alterations found in primary disease, namely mutational activation of Braf and PTEN deficiency, into primary human epidermal melanocytes. To this end, we took advantage of sequential lentiviral gene transductions into primary human epidermal melanocytes immortalized by coexpression of the SV40 large T-antigen (attenuating the Rb and p53 tumor suppressor pathways) and the catalytic subunit of human telomerase (hTERT; designated as the parental cell line hereafter; a kind gift of Robert Weinberg; ref. 13). This parental line has been shown previously to be anchorage-dependent and nontumorigenic in immune-defective mice ( 13). Importantly, immortalization of human melanocytes in vitro by attenuating the Rb and p53 pathways and overexpression of hTERT functionally mimics loss of the CDKN2A locus and gain of hTERT commonly seen in human melanomas ( 9).
We engineered the parental line to express a constitutively active Braf (the most common mutant allele, V600E; Braf cell line), a shRNA against PTEN [PTEN(low) cell line], or both [PTEN(low)/Braf cell line; Fig. 2A ]. Stable expression of BrafV600E increased the level of phosphorylated p44/42 mitogen-activated protein kinase (MAPK; Thr202/Tyr204) compared with that of parental melanocytes (Supplementary Fig. S3A), whereas PTEN knockdown enhanced the level of phosphorylated AKT (Ser473) compared with that of parental melanocytes only after serum starvation (Supplementary Fig. S3A and B). Curiously, BrafV600E expression specifically reduced the upper form of phosphorylated AKT (Ser473; Supplementary Fig. S3A). Parental melanocytes expressing constitutively active Ras or the c-Met receptor (hereafter termed Ras and c-Met melanocytes, respectively) served as functional comparisons ( 13). Autocrine and paracrine activation of the c-Met receptor has been implicated in several experimental models of melanoma metastasis ( 14, 15). The parental line transduced with empty vectors served as viral transduction, antibiotic selection, and FACS controls.
The coexistence of Braf activation and PTEN deficiency in primary cutaneous melanomas suggests these two genetic alterations might functionally interact. A basic panel of functional characterization failed to reveal evidence for functional interactions between Braf activation and PTEN deficiency in engineered human melanocyes (anchorage-independent growth, Fig. 2B; monolayer growth kinetic, Fig. 2C; cell cycle analysis with and without acute and chronic apoptotic induction, Fig. 2D). However, evidence that PTEN(low)/Braf melanocytes are not functionally equivalent to either PTEN(low) or Braf melanocytes was suggested by the behaviors of these genetically engineered melanocytes in the context of skin architecture ( Fig. 3A ). Human organotypic skin cultures were reconstructed by layering primary human epidermal keratinocytes and genetically engineered melanocytes on a human primary dermal fibroblast–contracted collagen I matrix. Previously, it has been shown that normal human epidermal melanocytes home to the basement membrane in skin culture and cannot survive in the upper epidermal layers unless genetically altered ( 16, 17). Importantly, cell lines derived from in situ melanomas do not invade into the dermis-like compartment, whereas those derived from vertical growth phase or invasive melanomas recapitulate their in vivo behaviors in the skin culture assay ( 18).
As shown in Fig. 3A, primary human keratinocytes (large eosinophilic cells with ample cytoplasm) formed a stratified epidermis-like layer, showing cornification with upward differentiation. Ras melanocytes displayed invasive growth with scattered cells in both the epidermis-like and dermis-like layers. In contrast, the parental melanocytes proliferated along the epidermal-dermal junction (EDJ) largely as single cells in a so-called lentiginous pattern, without evidence of dermal invasion. Braf or PTEN(low) melanocytes grew as well-defined, hyperplastic cellular nests or aggregates (so-called nevoid) along the EDJ without evidence of infiltrative dermal invasion. Interestingly, PTEN(low)/Braf cells displayed little to no dermal invasion but grew upward into the epidermal compartment (disrupting and displacing the epidermis), reminiscent of radial growth phase or in situ melanoma. All engineered melanocytes express the general melanocytic marker S-100 heterogeneously in skin cultures (Supplementary Fig. S4). This lack of significant invasive capacity in PTEN(low)/Braf melancoytes was corroborated by studies using a modified Boyden chamber assay to quantify invasion through Matrigel ( Fig. 3B). These data suggest PTEN deficiency and constitutive Braf activation, in the context of Rb/p53 attenuation and hTERT overexpression, collaborate to achieve stepwise, preinvasive growth alterations of human epidermal melanocytes in the skin.
As the levels of TGF-β2 expression positively correlate with primary melanoma thickness, a reflection of vertical/invasive growth and the single most important prognostic factor for early disease ( 19– 22), we hypothesized a potential role for melanocyte cell autonomous TGF-β signaling in promoting dermal invasion. We transduced Ras and PTEN(low)/Braf melanocytes with a constitutively activated TGF-β type I receptor (TβRI AAD), resulting in Ras/TGF-β and PTEN(low)/Braf/TGF-β melanocytes (Supplementary Fig. S3A). This mutant TβRI harbors three missense mutations: T204D that activates the kinase domain and L193A/P194A that prevent binding of the TβRI inhibitor FKBP-12 ( 23). In both Ras and PTEN(low)/Braf genetic contexts, constitutive TGF-β receptor activation led to similarly elevated levels of p-Smad2 while not affecting levels of phosphorylated p44/42 MAPK, phosphorylated AKT (Ser473), and phosphorylated AKT (Thr308; Supplementary Fig. S3A). In addition to the elevation of p-Smad2 levels, expression of TβRI AAD in Ras and PTEN(low)/Braf melanocytes resulted in a dramatic increase in the levels of plasminogen activator inhibitor-1, a well-known TGF-β target gene, in the conditioned media (data not shown).
We then incorporated these engineered melanocytes in the skin culture, Matrigel invasion, and collagen I invasion assays ( Fig. 4A and B, left and right , respectively). Importantly, whereas PTEN(low)/Braf melanocytes lacked significant dermal invasion in skin culture, constitutive TGF-β receptor I activation induced dermal invasion of these cells ( Fig. 4A). Ras and Ras/TGF-β melanocytes displayed a similar invasive phenotype indistinguishable using the skin culture assay (data not shown). Additionally, PTEN(low)/Braf/TGF-β melanocytes displayed a gain in both Matrigel and collagen I gel invasion compared with parental and PTEN(low)/Braf melanocytes, whereas Ras/TGF-β melanocytes did not seem to gain further invasive capacity compared with the Ras melanocytes under the assay conditions ( Fig. 4B). These data suggest that cell autonomous TGF-β signal hyperactivation of genetically aberrant human melanocytes can facilitate dermal invasion.
This TGF-β–dependent induction of dermal invasion may require both PTEN deficiency and Braf activation, as expression of TβRI AAD in Braf melanocytes (Supplementary Fig. S5A) failed to induce significant dermal (Supplementary Fig. S5B) or Matrigel (Supplementary Fig. S5C) invasion. We were unable to assess the individual contribution of PTEN deficiency to this TGF-β–dependent phenotype, because lentiviral transduction of TβRI AAD into PTEN(low) melanocytes failed to generate a stable line in two independent experiments. This counterselection in vitro suggests that the acquisition of PTEN deficiency in situ may follow concurrent Braf and melanocyte autonomous TGF-β activation.
Induction of tissue invasion by cell autonomous TGF-β activation correlated with enhanced levels and activities of specific matrix metalloproteinases (MMP; Fig. 4C and D, respectively). MMP-2 and MMP-9, in particular, have been implicated in melanoma invasion and metastasis ( 24– 27). We measured the levels of MMPs by ELISA in the conditioned media from Ras and PTEN(low)/Braf melanocytes with and without constitutive TGF-β activation compared with the parental melanocytes. As shown in Fig. 4C (left), TGF-β signaling significantly (P < 0.001, comparison with parental level) enhanced MMP-2 levels in both the Ras and PTEN(low)/Braf genetic contexts. However, TGF-β signaling significantly (P < 0.001) up-regulated the extracellular MMP-9 level only in the Ras genetic context ( Fig. 4C, right), highlighting the importance of signaling context on gene expression. Constitutive Ras activation alone led to significantly higher levels of MMP-9 (P < 0.001; Fig. 4C, right). In contrast, MMP-1 and MMP-3 levels were not significantly altered (data not shown). Importantly, the increased levels of MMP-2 and MMP-9 correlated with enhanced enzymatic activities on a gelatin gel ( Fig. 4D). TGF-β signaling induced invasion ( Fig. 3), while correlating with specific MMP up-regulation, did not correlate with a gain in the general growth properties of the engineered melanocytes in vitro on monolayer culture (Supplementary Fig. S6A) and in vivo as subcutaneous tumors in immune-defective mice (Supplementary Fig. S6B and C). In fact, PTEN(low)/Braf/TGF-β melanocytes displayed a delayed onset of subcutaneous growth compared with PTEN(low)/Braf melanocytes.
To address further whether this gain in invasive phenotype induced by a constitutively active TGF-β receptor I may be genetic context–dependent, we introduced TβRI AAD into parental melanocytes (TβRI AAD-HA and p-Smad2 expression shown in Supplementary Fig. S3A). We also adopted a modified human organotypic skin culture assay using devitalized human dermis to mimic better the extracellular environment of human dermal invasion. As shown in Fig. 4A, TGF-β melanocytes behaved similarly to parental melanocytes, showing no evidence of dermal invasion on H&E ( Fig. 5A ) and grew as single cells along the basal layer of the epidermis as distinguished from epidermal keratinocytes by positive S-100 staining ( Fig. 5B, top) and on top of an intact basement membrane as delineated by laminin staining ( Fig. 5B, bottom). In contrast, PTEN(low)/Braf/TGF-β melanocytes displayed invasive growth both in the superficial dermis and deep dermis ( Fig. 5A, bottom left and bottom right, respectively), as well as enhanced levels of p-Smad2 (Supplementary Fig. S7) compared with parental melanocytes. Lastly, the genetic context–dependent effect of cell autonomous TGF-β signaling on invasion was further corroborated by both Matrigel ( Fig. 5C, left) and collagen I matrix invasion assays using single, monolayer cells ( Fig. 5C, right) or spheroidal, three-dimensional “tumors” ( Fig. 5D).
Relatively, little is known about how genetic/epigenetic aberrations associated with human cutaneous melanomas contribute to the histogenesis of primary melanomas. The data presented in this report show that a pathway altered (i.e., hyperactivation of Smad2 signaling) at the premalignant, nevoid stage may be rewired into a proinvasive pathway by the acquisition of other genetic alterations (i.e., PTEN deficiency and Braf activation). In the pathogenesis of cutaneous melanoma, another early alteration, Braf activation, found commonly at the premalignant, nevoid stage ( 28), seems to have therapeutically relevant functional importance later during tumorigenesis (e.g., tumor maintenance and metastasis; 29). In the context of Rb/p53 attenuation and hTERT overexpression, a defined set of genetic alterations, including PTEN deficiency, constitutive Braf activation, and cell autonomous TGF-β activation, can suffice in driving melanocytic tumorigenesis.
The role(s) of cancer cell autonomous TGF-β signaling in tumor progression is complex ( 30). It is thought that early in tumorigenesis, the TGF-β–Smad pathway functions as a tumor suppressive pathway, whereas, in advanced cancers that have become resistant to TGF-β-induced growth arrest or apoptosis, the TGF-β–Smad pathway can direct proinvasive and prometastatic functions. For instance, it has been shown that transgenic expression of TβRI(AAD) in the breast epithelia retards Neu-driven primary tumor growth while promoting invasion and metastasis ( 23). Human colon cancers with inactivating mutations in the TGF-β type II receptor show reduced metastatic potential and correlate with increased patient survival ( 31). Smad2/Smad3 has been found in its active, COOH terminally phosphorylated, and nuclear-localized form in advanced human cancers, including metastatic breast cancers and high-grade gliomas ( 11, 12). Smad4, a common partner of receptor-regulated Smads and a bona fide human tumor suppressor, has been found to be an active mediator of breast cancer bone metastasis in xenograft models ( 12, 32). How this tumor suppressive pathway is rendered into a pro-oncogenic pathway remains an intriguing question. Recently, it has been proposed that the epigenetic status of a cancer cell can potentially corrupt the TGF-β/Smad pathway into a pro-oncogenic pathway ( 11).
In our survey of human melanocytic neoplasms, we have found high levels of nuclear p-Smad2 by immunohistochemistry in benign melanocytic nevi, melanomas in situ, and primary invasive melanomas. Consistent with this, normal melanocytes in situ lack TGF-β expression, whereas nevi and melanomas up-regulate TGF-β1, TGF-β2, and TGF-β3 expressions ( 19, 22). In contrast, the related ligand Nodal, which also signals through Smad2, seems to be absent in normal skin but becomes selectively expressed in the invasive portion of primary cutaneous melanomas and further up-regulated in 60% of metastatic melanomas ( 7). Thus, human melanocytic tumors as early as benign proliferations are microenvironments rich in TGF-β and related factors.
The present work shows that tumor cell autonomous hyperstimulation of the TGF-β-Smad2 pathway can be causally related to melanocytic oncogenic progression in the skin and may be, at least in part, responsible for the critical switch from radial to vertical growth during human melanoma histogenesis. The invasive switch may also be mediated by a TGF-β/Smad-independent pathway, as Ras melanocytes do not show elevated levels of p-Smad2 (Supplementary Fig. S3A) or elevated levels of secreted, active TGF-β1. 7 Furthermore, the invasive switch provided by constitutive TGF-β receptor I activation does not likely solely rely on Smad activation per se but rather on the collaboration of activated Smads with an altered genetic or epigenetic cellular context. Hence, the precise roles of TGF-β signaling in melanoma metastasis, tumor maintenance, and patient profile merit further investigation.
A recent study implicated genes coregulated by TGF-β and related factors to be associated with highly invasive/metastatic melanomas, whereas those coregulated by β-catenin are associated with less aggressive melanomas ( 10). Interestingly, this TGF-β transcriptional profile can coexist with either NRas or Braf activating mutations, two mutually exclusive events in primary and metastatic melanomas. In our study, constitutive TGF-β activation differentially up-regulated the levels of active MMP-2 and MMP-9 pending on the genetic context ( Fig. 4C and D).
During human melanoma progression, PTEN loss, either by deletion or promoter hypermethylation, commonly occurs in association with Braf mutational activation ( 5, 33, 34). Although PTEN antagonizes phosphatidylinositol 3-kinase (PI3K), it is thought to participate also in PI3K-Akt–independent pathways. PI3K itself does not seem to be overexpressed in situ with melanoma progression ( 35), but Akt3 has been observed as a target of amplification in melanoma ( 36). In the context of a dominant-negative p53, a p16-resistant CDK4, and hTERT overexpression, constitutive PI3K activation, with or without constitutive Braf activation, resulted in invasive melanoma in a regenerated skin model, whereas constitutive Braf activation in the same genetic background gave rise to hyperplastic growth ( 37). More recently, Akt overexpression in one melanoma cell line, WM35, derived from a radial growth phase melanoma was shown to enable subcutaneous growth in nude mice, suggesting a role of activated Akt in the invasive switch of primary melanoma ( 38). Although PTEN deficiency have well-documented Akt-independent effects on tumorigenesis ( 39), we found that, whereas PTEN-deficient, Braf-activated melanocytes are clearly tumorigenic in NOD-SCID mice (Supplementary Fig. S6B), these engineered melanocytes are poorly invasive in vitro, raising the possibility that subcutaneous xenograft growth derived from human cancer cell lines may not always reflect a locally invasive phenotype.
Several limitations of our experimental system are noteworthy. The serial genetic alterations are built on a single genetic background. During monolayer culture, introduced genetic alterations may impose on human epidermal melanocytes and its engineered derivatives survival requirements that may be distinct in vivo. We also cannot rule out potential genetic alterations that may have occurred during the obligate passage of cell cultures in vitro. Despite these limitations, we suggest that the functional genetic interactions described here may apply to autochthonously arising human cutaneous melanomas.
The present work has shown that a proper skin environment provides an important platform to investigate the potential interactions of distinct genetic alterations. The model system of reconstructing distinct molecular pathways of melanoma progression de novo in a human skin–like environment can be extended to interrogate the rapidly growing number of concurrent genetic/epigenetic abnormalities found in genetic subsets of human melanoma.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: American Skin Association, American Society for Dermatologic Surgery, Dermatology Foundation, and California Institute for Regenerative Medicine (R.S. Lo). R.S. Lo holds a Career Award for Medical Scientists from Burroughs Wellcome Fund. O.N. Witte is an Investigator of Howard Hughes Medical Institute.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank M. Herlyn (Wistar Institute) and L. Li (Herlyn Laboratory) for generously sharing the skin organotypic culture technology; D. Chen (Witte Laboratory) for FACS analysis and sorting; S. Binder (University of California at Los Angeles Department of Dermatopathology) for histologic assistance; R. Weinberg (Whitehead Institute) for Mel-ST, Mel-STR, and Mel-STM cell lines; Y. Chen (Tsinghua University) for TβRI AAD cDNA; and M. Kolodney (Harbor-University of California at Los Angeles) for Braf V600E cDNA. R.S. Lo is grateful to R. Modlin (University of California at Los Angeles Division of Dermatology) for mentorship.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
↵7 R.S. Lo and O.N. Witte, unpublished observation.
- Received October 1, 2007.
- Revision received March 10, 2008.
- Accepted March 25, 2008.
- ©2008 American Association for Cancer Research.