Recent studies suggest that cancer can arise from a cancer stem cell (CSC), a tumor-initiating cell that has properties similar to those of stem cells. CSCs have been identified in several malignancies, including those of blood, brain, and breast. Here, we test whether stem cell–like populations exist in human melanomas. In ∼20% of the metastatic melanomas cultured in growth medium suitable for human embryonic stem cells, we found a subpopulation of cells propagating as nonadherent spheres, whereas in standard medium, adherent monolayer cultures were established. Individual cells from melanoma spheres (melanoma spheroid cells) could differentiate under appropriate conditions into multiple cell lineages, such as melanocytic, adipocytic, osteocytic, and chondrocytic lineages, which recapitulates the plasticity of neural crest stem cells. Multipotent melanoma spheroid cells persisted after serial cloning in vitro and transplantation in vivo, indicating their ability to self-renew. Furthermore, they were more tumorigenic than adherent cells when grafted to mice. We identified similar multipotent spheroid cells in melanoma cell lines and found that the stem cell population was enriched in a CD20+ fraction of melanoma cells. Based on these findings, we propose that melanomas can contain a subpopulation of stem cells that contribute to heterogeneity and tumorigenesis. Targeting this population may lead to effective treatments for melanomas.

Due to their resistance to current therapies, melanomas remain a significant cause of mortality in Caucasians. Tumors consist of heterogeneous populations whose biological properties remain poorly characterized. Melanomas are believed to arise from a mature, differentiated melanocyte. However, mounting evidence suggests that cancer may in fact arise from a transformed stem cell, which is able to self-renew, differentiate into diverse progenies, and drive continuous growth (1). Cancer stem cells (CSC) have been identified in leukemias, and tumors of the breast and brain by tumor type–specific cell surface markers often associated with stem cells (25). Stem-like cells were also found in an established glioma cell line by their inability to incorporate a nuclear dye (6, 7). CSCs from brain tumors could be isolated, because they show in vitro growth characteristics similar to those of neural stem cells, which proliferate as nonadherent cell aggregates termed spheres or spheroids (8, 9).

Indirect evidence supports the presence of melanoma stem-like cells. First, melanomas show phenotypic heterogeneity both in vivo and in vitro, suggesting an origin from a cell with multilineage differentiation abilities. Melanoma cells retain their morphologic and biological plasticity despite repeated cloning (10). Second, melanoma cells often express developmental genes (11). Third, melanoma cells can differentiate into a wide range of cell lineages, including neural, mesenchymal, and endothelial cells. They frequently exhibit characteristics of neural lineages (1214). Melanomas from aggressive lesions can develop vessel-like structures and share with endothelial cells many matrix adhesion receptors such as β3 integrin or the cell-cell adhesion molecules MCAM, which are important for invasion and metastasis (1517). They can also acquire characteristics of stromal fibroblasts by constricting collagen type I (18) and expressing fibroblast-associated markers such as fibroblast activating protein (19). In some cases, melanoma lesions contain areas of adipogenic or osteocartilaginous differentiation patterns (2024).

Given this evidence, we asked whether melanomas contain a tumorigenic stem cell–like population.

Primary culture, propagation, and separation of melanoma cells. Metastatic melanomas were obtained in accordance with consent procedures approved by the Internal Review Boards of the University of Pennsylvania School of Medicine and The Wistar Institute. They were obtained within 1 to 2 hours after surgical removal. Tumors were rinsed, trimmed to remove connective tissues, and subjected to enzymatic dissociation in 1 mg/mL collagenase IV (Invitrogen, Carlsbad, CA) in DMEM for 4 to 6 hours at 4°C. Single cells were washed with HBSS, resuspended in culture medium, and plated onto noncoated flasks.

For growing melanoma cells, we used two types of media: (a) mouse embryonic fibroblast (MEF)–conditioned human embryonic stem cell (hESC) medium (25, 26). Before use, we mixed MEF conditioned with fresh hESC medium at a 3:1 ratio and supplemented with basic fibroblast growth factor (bFGF) at 4 ng/mL. (b) Mel 2% melanoma growth medium, which was used to establish permanent melanoma cell lines (27). It consisted of MCDB 153 medium (Sigma, St. Louis, MO; 4 parts), L15 medium (Invitrogen; 1 part), 2% FCS, 5 μg/mL insulin (Sigma), 15 μg/mL bovine pituitary extract (Cambrex, East Rutherford, NJ), 1.68 mmol/L calcium chloride, and 5 ng/mL epidermal growth factor (EGF, Sigma). Established melanoma cell lines WM115 and WM239A were cultured in Mel 2% medium without pituitary extract and EGF (27). EBV-transformed B-cell lines from melanoma patients were kindly provided by Dr. D. Herlyn of The Wistar Institute and cultured in RMPI 1640 with 10% FCS (28).

To isolate melanoma cells from heterogeneous primary cultures, individual cells derived from mechanically or enzymatically dissociated primary spheres were cloned by limiting dilution assay in hESC medium. Although one cell showed limited potential for proliferation, new spheroid cultures were regenerated from two single primary cells. For subculture, spheroid cells were dissociated and replated every 7 to 10 days at a clonal density of 1,000 cells/mL (29).

Evaluation of tumorigenicity and histologic staining. Tumorigenicity was determined by s.c. injecting single cells into the right flank of severe combined immunodeficient (SCID) mice at 1.7 × 106 cells per mouse. To compare tumorigenicity of melanoma spheroid cells and their adherent counterparts with limiting amount of cells, mice were treated with 200 μg cyclophosphamide monohybrate (Cytoxan, Sigma), which further suppressed the immune response. Four days later, 2 × 105 tumor cells were injected. H&E, melanin, and hemosiderin staining were done on 5-μm paraffin-embedded sections following standard protocols.

Flow cytometry and fluorescence-activated cell sorting. Adhesive cells were removed with 0.02% EDTA in HBSS. Cells were washed, suspended in buffer [0.1% bovine serum albumin (BSA), 0.1% NaN3 in PBS], and incubated with primary antibodies for 60 minutes at 4°C with constant agitation. Cells were washed twice with buffer then incubated with Alexa Fluor 488–conjugated goat anti-mouse secondary antibodies (Molecular Probes, Eugene, OR) for 60 minutes when unconjugated primary antibodies were used. Suspended spheroid cells were dissociated mechanically into single cells and stained. Approximately 5 × 103 cells were analyzed in an EPICS XL instrument (Beckman-Coulter, Inc., Miami, FL). Monoclonal antibodies (mAb) against CD3 (PE conjugated), CD4 (PE−), CD8 (PE−), CD20 (FITC−), MCAM, CD117, and CD26 were purchased from BD PharMingen (San Diego, CA). We purchased mAbs against CD45 (PeliCluster, Amsterdam, The Netherlands), CD34 (PeliCluster), CD133 (Miltenyi Biotech, Auburn, CA), CD31 (Dako, Carpinteria, CA), neural cell adhesion molecule (CD56/NCAM, NeoMarkers, Fremont, CA), E-cadherin (Zymed Laboratories, South San Francisco, CA), N-cadherin (Sigma), von Willebrand factor (vWF, NeoMarkers), vascular endothelial growth factor-2 (VEGFR2, Imclone, New York, NY), and growth-associated phosphoprotein-43 (GAP-43, Calbiochem, San Diego, CA). mAbs against the surface markers of hESCs, including stage-specific embryonic antigen (SSEA)-1, SSEA-3, SSEA-4, TRA-1-60, and TRA-1-81 (25), were obtained from where they were originally developed at The Wistar Institute. mAbs against GD2, chondroitin sulfate proteoglycan (CSPG), β3 integrin, EGF receptor (EGFR), HLA-DR, and p70NGFR were described previously (30). Isotype-matched mouse pure, PE- or FITC-conjugated antibodies (BD PharMingen) were used as controls.

Fluorescence-activated cell sorting (FACS) was done on a Cytomation MoFlo cytometer (DakoCytomation, Fort Collins, CO). FITC-conjugated mAb against human CD20 was used for separation. For sorting double-stained cells, a polyclonal antibody against human CD20 antigen (NeoMarkers) was used to costain with mAb against MCAM or β3 integrin. Polyclonal antibodies were labeled by PE-conjugated goat antibodies against rabbit IgG (Sigma) and mAbs by FITC-conjugated rabbit antibodies against mouse IgG (Sigma).

Differentiation assays of melanoma spheroid cells. Melanoma spheres were allowed to sediment, individual cells were removed with the supernatant. Spheres were dissociated into individual cells before plating onto tissue culture-grade plastic coated with 10 ng/mL fibronectin at the density of 2 × 103/cm2. Melanogenic differentiation medium, which was based on melanocyte growth medium (31) and exclusively differentiated hESCs into melanocytic lineages,3

3

Fang et al., unpublished.

contains dexamethasone (0.05 μmol/L, Sigma), insulin-transferrin-selenium (ITS, 1×, Sigma), linoleic acid-BSA (1 mg/mL, Sigma), low-glucose DMEM (30%, Life Technologies, Rockville, MD), MCDB 201 (20%, Sigma), l-ascorbic acid (10−4 mol/L, Sigma), conditioned media of mouse L-Wnt3a cells (American Type Culture Collection, Manassas, VA, 50%), stem cell factor (100 ng/mL, R&D System, Minneapolis, MN), endothelin-3 (100 nmol/L, American Peptide, Sunnyvale, CA), cholera toxin (20 pmol/L, Sigma), the phorbol ester 12-O-tetradecanoylphorbol-13-acetate (50 nmol/L, Sigma), and bFGF (4 ng/mL). Differentiation media for mesenchymal lineages were used as described with modifications (32). Osteogenesis medium consisted of 90% DMEM, 10% FCS, 1× ITS (Sigma), 1 mg/mL LA-BSA (Sigma), 0.1 μmol/L dexamethasone (Sigma), 0.05 mmol/L l-ascorbic acid-2-phosphate (Sigma), and 10 mmol/L β-glycerophosphate (Sigma). Alkaline phosphatase was detected after 14 days with the Vector Blue alkaline phosphatase substrate kit III (Vector Laboratories, Burlingame, CA). Adipogenic medium contained 90% DMEM, 10% horse serum (Invitrogen), 1× ITS, 1 mg/mL LA-BSA, 1 μmol/L hydrocortisone (Sigma), 0.5 mmol/L isobutylmethylxanthine (Sigma), and 60 μmol/L indomethacin (Sigma) for 10 to 14 days. Accumulation of lipid vacuoles was visualized by staining with Oil Red O (Sigma). Chondrogenic medium contained 90% DMEM, 10% FCS, 1× ITS, 1 mg/mL LA-BSA, 10 ng/mL transforming growth factor-β1 (R&D System), and 1 mmol/L pyruvate (Sigma) for 10 to 14 days. Chondrogenic differentiation was assayed by expression of type II collagen. As controls, undifferentiated spheres were stained in suspension and then centrifuged onto slides.

Immunocytochemical staining. Cells were fixed with 4% paraformaldehyde and stained with primary antibodies specific for microphthalmia-associated transcription factor (MITF, monoclonal, NeoMarkers), sex-determining region Y (SRY)–related transcription factor SOX10 (polyclonal, Abcam, Cambridge, MA), tyrosinase (TYR, monoclonal, Novocasta, Newcastle upon Tyne, United Kingdom), anti-human nuclei MAB1281 (Chemicon, Temecula, CA), and type II collagen (monoclonal, Chemicon). Isotype-matched mouse antibodies or normal rabbit IgG were used as controls. After washings, primary antibody binding was detected using the corresponding Alexa Fluor 488–conjugated secondary antibodies (Molecular Probes). Staining was observed by fluorescence microscopy.

RNA isolation, reverse transcription-PCR, and Western blotting. RNA isolation, reverse transcription-PCR (RT-PCR), and Western blot analyses were done using primers and antibodies against dopachrome tautomerase (DCT/DCT), MITF/MITF, and TYR as described (33, 34). TYR primers were TGCCAACGATCCTATCTTCC (sense) and TGAGGAGTGGCTGCTTTTCT (antisense). A mAb against β-actin (Sigma) was used as a control.

Statistical analysis. To assess the statistical significance of differences, a one-sided Student's t test was done. Ps <0.05 were considered significant, as indicated by asterisks.

Tumor cells form nonadherent spheres in mouse embryonic fibroblast–conditioned human embryonic stem cell medium. Seventeen metastatic melanoma lesions were used in this study, including 15 lymph nodes, one brain, and one parotid. To isolate a putative stem cell population from metastatic melanoma lesions, we used growth medium for totipotent hESCs and standard melanoma medium, Mel 2%, to culture dissociated single tumor cells (Fig. 1A,, a). Within 14 days, Mel 2% yielded adhesive, spindle, or epithelioid melanocytic cells from all specimens (Fig. 1A,, b). In three metastatic lesions (∼20%), hESC medium supported cells growing as nonadhesive spheres (Fig. 1A,, c). Spheres derived from lesions WM3517 and WM3539 were pigmented, as were their adhesive cultures established in Mel 2% (data not shown). Spheres of WM3523 lesion were nonpigmented, whereas the adherent monolayer cultures grown in Mel 2% from the same specimen were pigmented (Fig. 1A , d).

To characterize the cells obtained in stem cell medium, we tested for expression of cell surface markers. Spheres derived from WM3523, WM3517, and WM3539 were negative for endothelial markers vWF, CD31, CD34, and VEGFR2; stem cell markers SSEA-1, SSEA-3, SSEA-4, TRA-1-60, TRA-1-81, and CD133; and neural markers GAP-43 and CD56/NCAM (data not shown). Consistent with their pigmented phenotypes, both WM3517 and WM3539 spheres contained homogeneous melanoma cells expressing melanoma-associated cell surface markers CSPG (98.20% and 99.00%, respectively), β3 integrin (98.70% and 99.40%), and MCAM (99.90% and 99.20%, respectively). However, only a subpopulation of WM3523 spheres expressed melanoma markers GD2 (21.30%), CSPG (27.10%), β3 integrin (62.00%), MCAM (47.70%), and p70NGFR (26.00%; Fig. 1B). In contrast, adherent monolayer WM3523 cells established in Mel 2% medium were more homogenously positive for GD2 (78.40%), CSPG (96.80%), β3 integrin (98.60%), MCAM (98.40%), and p70NGFR (94.80%). Tumor-infiltrating B cells isolated from melanoma lesions were included as negative controls. Only a small fraction of the B cells expressed typical melanoma markers (Fig. 1B).

Both WM3517 and WM3539 spheres were negative for hematopoietic markers (data not shown), whereas WM3523 spheres contained hematopoietic cells. Very few of WM3523 primary spheroid cells expressed CD3 (0.99%), CD4 (1.21%), and CD8 (1.04%), but a significant population of WM3523 spheroid cells expressed CD45 (38.90%) and CD20 (41.70%; Fig. 1C). Whereas B cells uniformly expressed CD45 (99.80%) and CD20 (98.10%), adhesive melanoma cells isolated from the same lesion were homogeneously negative for all hematopoietic markers tested (Fig. 1C). These results suggest that the WM3523 spheres contained both melanoma and hematopoietic cells.

Sphere formation of melanoma cells isolated from heterogeneous populations. To ensure the purity of cell population, melanoma cells were clonally isolated from mixed cultures as described in Materials and Methods. All 14 clones of WM3523 were pigmented and were able to produce proliferating melanoma spheres (Fig. 2A). The spheres were grown for >8 months by continuous passage. A minor population, ranging from 5% to 10%, grew adherent and subsequently differentiated into small, oval melanocytic cells with short dendrites (data not shown), whereas the major population grew as melanoma spheres. Flow cytometry analyses of five clones showed that spheroid cells homogeneously expressed melanoma markers CSPG (95.30%), β3 integrin (96.00%), and MCAM (97.50%; Fig. 2B). A significant fraction of cells expressed E-cadherin (89.60%) but not N-cadherin (2.54%). Hematopoietic markers CD3 (0.76%) and CD4 (0.90%) remained negative, with a small portion expressing CD8 (1.22%), CD45 (1.10%), and CD20 (3.34%; Fig. 2B). The CD20-positive subpopulation coexpressed melanoma-associated MCAM and β3 integrin and was enriched by FACS (Fig. 2C-D). This indicates that a subpopulation of melanoma spheroid cells express the hematopoietic marker CD20.

Self-renewal and melanocytic differentiation of melanoma spheroid cells. After separating WM3523 melanoma spheres into single cells and reseeding at a clonal density of 1,000 cells/mL, they remained in stem cell growth medium as individual cells without aggregate formation for 5 hours (Fig. 3A,, a). New spheres developed from individual cells 4 days after seeding (Fig. 3A,, b). When dissociated spheroid cells were treated with melanocyte differentiation medium, a large proportion of cells adhered to flasks precoated with fibronectin (Fig. 3A,, c). Within 4 days, they differentiated into melanocytic cells (Fig. 3A,, d), whose pellets displayed increased pigmentation (Fig. 3A,, e, right) when compared with pellets from parental cells (Fig. 3A , e, left). The capacity for melanocytic differentiation of melanoma spheroid cells persisted in long-term cultures for up to 8 months with no significant decrease in efficiency. Melanoma spheroid cells in stem cell growth medium maintained their growth potential; whereas under differentiation conditions, cell growth slowed dramatically and stopped after 18 days (data not shown). Similarly, standard melanoma growth medium, Mel 2%, could differentiate melanoma spheroid cells into melanocytic cells but with lower efficiency (data not shown). These results suggest that melanoma spheroid cells can self-renew and differentiate into a melanocytic phenotype.

We then characterized melanocytic differentiation by immunostaining, RT-PCR, and Western blotting. A fraction of melanoma spheroid cells expressed melanocyte-specific MITF (52.8%) and melanogenic enzyme TYR (30.8%; Fig. 3B-C). After differentiation, the populations expressing both melanocytic markers increased significantly. In contrast, adherent monolayer melanoma cultures isolated in Mel 2% showed only marginal increases in MITF- or TYR-expressing cells after differentiation (Fig. 3B-C). These results suggest that melanoma spheroid cells contain a larger population of progenitor cells with differentiation potential for melanocytic cells lineage than the adherent cells. RT-PCR and Western blotting detected expression of melanocyte-specific genes MITF/MITF, TYR/TYR, and DCT/DCT. Significant increases in TYR mRNA and TYR protein levels were observed after melanocyte differentiation (Fig. 3D). The DCT protein was undetectable in undifferentiated and differentiated populations; however, its transcriptional expression was greatly enhanced after differentiation. Expression of two phosphorylated MITF proteins (54 and 60 kDa) was increased after differentiation regardless of a slight decrease in MITF mRNA. These results show increases in DCT transcription and MITF and TYR proteins correlate with enhanced pigmentation after differentiation.

Mesenchymal differentiation of melanoma spheroid cells. Melanocytes are derived from the neural crest during embryonic development. The transient neural crest consists of pluripotent stem cells that give rise to a wide array of lineages, including neurosecretory cells, peripheral neurons, glia, and the cephalic mesenchyme (bone and cartilage; ref. 35). To determine whether WM3523 melanoma spheroid cells can differentiate similarly to neural crest stem cells, we examined neural and mesenchymal differentiation. The cells failed to differentiate into neural lineages in defined medium containing 60% DMEM, 40% MCDB 201, 0.05 μmol/L dexamethasone, 1× ITS, 1 mg/mL LA-BSA, 10−4 mol/L l-ascorbic acid, and 100 ng/mL bFGF (ref. 36; data not shown). On the other hand, WM3523 melanoma spheroid cells readily differentiated into mesenchymal lineages with varying efficiencies: adipogenic (31.0%), chondrogenic (18.6%), and osteogenic (5.6%; Fig. 4A-B). Few or no undifferentiated melanoma spheroid cells stained for Oil Red O (5.6%), type II collagen (1.6%), or alkaline phosphatase (0.0%). Adherent monolayer melanoma cells established in Mel 2% showed significant potential for adipogenesis only (4.0%; Fig. 4B). These results suggest that multipotent stem cells are enriched in melanoma spheroid populations isolated in stem cell growth medium.

To verify the common origin of melanocytic lineages and differentiated mesenchymal cells, we double stained with melanocyte-associated transcription factors MITF or SOX10 and mesenchymal markers Oil Red O or alkaline phosphatase (Fig. 4C). In adipogenic medium, differentiated melanoma cells containing multiple MITF- or SOX10-positive nuclei also displayed Oil Red O staining (Fig. 4C). In osteogenic medium, differentiated melanoma cells expressing alkaline phosphatase also exhibited diffuse MITF or SOX10 expression. These data indicate a common origin for differentiated mesenchymal cell lineages and melanocytic cells, and that mesenchymal differentiation is not due to contaminating non-tumor stem cells.

The percentage of differentiated mesenchymal cell types varied considerably among clones. Of six tested, clones 4, 6, and 11 could differentiate into melanogenic, adipogenic, chondrogenic, and osteogenic lineages. Other clones were either tripotent (melanogenic, adipogenic, and chondrogenic lineages), bipotent (melanogenic and adipogenic), or unipotent (melanogenic). The mesenchymal differentiation capacity of melanoma spheroid cells persisted in long-term cultures up to 8 months, albeit with decreased efficiency (particularly in osteogenic differentiation). Melanoma spheroid cell lines WM3517 and WM3539 showed differentiation capacity for the melanocytic lineage only (data not shown).

Multipotent melanoma spheroid cells derived from established melanoma cell lines. We explored whether a stem cell population could be isolated from long-established melanoma cell lines. A total of 18 cell lines were tested in stem cell culture conditions using hESC medium. Spheroid phenotypes developed in WM115 and WM239A, a pair of primary and metastatic melanoma cell lines derived from the same patient (37). When adherent cultures of WM115 cells (Fig. 5A,, a) were exposed to hESC medium, spheres appeared after ∼7 days (Fig. 5A,, b). These spheroid cells consistently formed new spheres when separated into single cells (Fig. 5A,, c). Adipogenic, osteogenic, and chondrogenic differentiation was extensively induced under appropriate conditions in WM115 spheroid cells but not in the adherent population (Fig. 5A , d-f and B). Similar differentiation capacities were observed in WM239A spheroid cells established in stem cell medium (data not shown).

Flow cytometry analysis showed that, although the majority of hematopoietic and melanoma cell surface markers on WM115 spheroid cells retained expression levels similar to those of WM115 adhesive cells, a subpopulation of WM115 spheroid cells expressed the hematopoietic marker CD20 and showed reduced expression of EGFR (a marker for epithelial differentiation; Fig. 5C). Melanoma cells coexpressing MCAM or β3 integrin and CD20 were confirmed by FACS and fluorescence microscopy (Fig. 5D-E). These data suggest that a multipotent stem cell population can be isolated from established melanoma cell lines and that this population grows and differentiates similarly to fresh isolates. Moreover, a subpopulation of melanoma spheroid cells derived from melanoma cell lines is also similar to freshly isolated cells in that it coexpresses the hematopoietic marker CD20.

Multipotent melanoma spheroid cells are capable of forming tumors in vivo and self-renewing after transplantation. We examined the tumorigenic capacity of WM115 and WM3523 spheroid cells by s.c. injection in SCID mice. All mice injected with WM115 spheroid cells (n = 5), or WM3523 spheroid cells (clone 4, n = 5; clone 6, n = 3) developed tumors (Fig. 6A). The majority of tumors were observed between 28 and 40 days. Tumors consisted of large cells with abundant eosinophilic cytoplasm, oval to irregular nuclei, and dominant nucleoli. Most tumors also contained giant tumor cells (data not shown). Their melanocytic origin was confirmed by positive staining for melanin and not for hemosiderin (Fig. 6A; data not shown). These results suggest that melanoma spheroid cells isolated in stem cell medium are tumorigenic.

We then isolated and cultured melanoma cells from tumors grown in mice. Typical melanoma spheres were formed within 2 to 3 weeks in stem cell medium (Fig. 6A). These tumor cells could be stained by a specific monoclonal antibody against human nuclei, confirming their human origin (data not shown). Melanogenic, adipogenic, chondrogenic, and osteogenic differentiation of these spheroid cells were observed under appropriate differentiation conditions (data not shown). These results indicate that a stem cell population exists in melanomas after in vivo transplantation. Therefore, a self-renewing stem cell population persists not only in long-term cultures but also in tumors transplanted in vivo.

To address whether multipotent melanoma spheroid cells differed in tumorigenicity from adherent monolayer melanoma cells, we implanted melanoma spheroid cells or adherent monolayer cells into Cytoxan-treated SCID mice at 2 × 105 cells per mouse, an amount that usually does not induce tumors. After ∼35 days, mice transplanted with spheroid cells had developed pigmented tumors, whereas tumors were not observed in mice injected with adherent cells. When mice were sacrificed 70 days after injection, three animals in the spheroid group (n = 5) had small to large tumors compared with only one small tumor in one of the adherent groups (n = 5). Total tumor weight and volume from the spheroid group were significantly larger than the adherent group (Fig. 6B). This suggests that melanoma spheroid cells isolated in hESC culture conditions are more tumorigenic.

Multipotent stem cells are enriched in the CD20+ fraction of melanoma spheres. We consistently observed CD20+ populations in WM3523 and WM115 melanoma spheroid cells but not in the corresponding adherent populations. Individual CD20+ tumor cells could be detected by immunohistochemistry in ∼20% of metastatic melanoma lesions, supporting our in vitro observations (data not shown). We did cell sorting to determine whether the stem cell population was within CD20+ fractions. Both positive and negative fractions from WM3523 and WM115 proliferated extensively after sorting. Their phenotypes were confirmed by flow cytometry (data not shown). The CD20+ fractions of WM3523 cells formed larger spheres compared with the CD20 fraction. In WM115 cells, only the CD20+ fraction proliferated as spheres, whereas the CD20 fraction remained as single cells (data not shown). These data suggest that the stem cell population may reside within the CD20+ fraction.

The CD20+ or CD20 fractions of WM3523 and WM115 were then subjected to differentiation. Whereas WM115 CD20 fraction remained in suspension, the other three populations adhered to substrate, indicating their differentiation potentials (data not shown). Although similar percentages showed immunoreactivity for MITF in both CD20+ and CD20 fractions from both spheroid cell lines after melanocytic differentiation, only CD20+ fractions showed substantial potential for mesenchymal differentiation (Fig. 6C-D). These data confirm that multipotent stem cells are enriched in the CD20+ fraction of melanoma spheroid cells isolated from fresh tumor lesions and established cell lines.

We report a subpopulation of melanoma cells with characteristics of primitive progenitors for melanocytes, neural crest cells, that give rise to a broad range of cell types. When cultured in medium used for hESCs, melanoma cells proliferated as nonadherent spheres. When initially isolated from the lesion, the spheres contained both melanoma and hematopoietic cells. After cloning, a subpopulation of melanoma cells could be isolated that maintained stem cell characteristics; they were able to self-renew and differentiate into melanogenic, adipogenic, chondrogenic, and osteogenic lineages. These melanoma spheroid cells were also found in established cell lines, suggesting that a subpopulation of melanoma cells in long-term culture can maintain stem cell properties.

Our studies reveal that melanoma lesions can contain a subpopulation with stem cell properties and a fraction of more differentiated tumor cells. The adherent population established in vitro under standard conditions displays spindle to epithelioid morphology. In general, these cells reflect the stage of tumor progression from which they were initially derived, and those cells from metastatic lesions are tumorigenic in mice (27). They are characterized for their expression of melanoma-associated antigens such as MCAM or β3 integrin, which facilitate invasion and metastasis (16, 17). The second population, described for the first time here, can differentiate and self-renew. This population is characterized by its ability to form spheres. Sphere formation was initially observed in cultured neural stem cells (38). Cells within neural spheres have stem cell properties that manifest as self-renewal and multilineage differentiation potential (39). Recently, sphere formation was found when stem cells from a variety of normal and tumor tissues were isolated (3, 5, 40, 41), suggesting that sphere formation may be a common growth characteristic of stem cells.

It is not surprising that melanoma spheroid cells are capable of mesenchymal differentiation, because mesenchymal and melanocytic cells may originate from the same embryonic tissue, the neural crest (35). Differentiation of human melanomas into mesenchymal lineages has also been described in fat and osteocartilaginous differentiation (2024). Differentiating cells were often multinucleated, their cultures could not be maintained for extended periods of time, and their differentiation was irreversible. Furthermore, melanoma spheroid cells gave rise to cells that expressed both melanocytic and mesenchymal markers. These data confirmed that the mesenchymal progenitors are not derived from contaminating normal mesenchymal stem cells but rather are a property of the tumor itself.

In sphere-forming melanoma cell lines, WM3517 and WM3539, spheroid cells were able to undergo only melanogenic differentiation, suggesting they might arise from a cell at a later stage of differentiation, such as a lineage-committed melanocyte precursor cell. Multipotent WM3523 melanoma spheroid cells had a stable capacity for self-renewal over prolonged culture periods. Clones of WM3523 spheroid cells showed differences in their stability of differentiation potential. Over time in culture, some clones could no longer differentiate into all four cell lineages but only into three, two, or one suggesting the control mechanisms for each differentiation lineage is unique.

Neoplastic melanocytic cells frequently exhibit characteristics of other neural crest derivatives both in vitro and in vivo. Differentiation of melanoma cells into neural lineages has been well shown (1214); however, in this study, melanoma spheroid cells failed to differentiate into neural lineages. The differentiation medium contains bFGF, which has been used to induce neural differentiation of hESCs and bone marrow–derived stem cells (36, 42, 43). In this case, the plasticity of melanoma spheroid cells was limited.

Multipotent melanoma spheroid cells isolated from fresh tumors persist in culture for a long time (8 months) when passaged at clonal densities and as tumor xenografts. Similar multipotent melanoma spheroid cells can also be isolated from established cell lines. Importantly, multipotent melanoma spheroid cells can be serially recloned from a minimum of two cells during 8 months in culture. We have provided evidence that this subpopulation of melanoma cells is able to self-renew and meet other criteria of stem cells. Moreover, melanoma spheroid cells have differentiation potential similar to neural crest stem cells, supporting the notion that melanomas may arise from a transformed melanocyte stem cell. Recent studies suggest that neural crest stem cells persist in the skin in adult animals and are capable of differentiating into neurons, smooth muscle cells, Schwann cells, adipocytes, and melanocytes (41, 44, 45). Lineage-committed melanocyte stem cells have also been reported in mouse hair follicles (46). In humans, equivalent stem cells have not yet been identified but potentially exist in adult skin to maintain homeostasis and to repair damage after injury. These long-lived stem cells may accumulate mutations and become a candidate for transformation. With its capacity for self-renewal, this subpopulation of transformed stem cells may subsequently initiate and sustain malignancy.

Increasing evidence supports this scenario. A link between epithelial cancer and bone marrow–derived cells (most likely mesenchymal stem cells) has been shown in a mouse model of gastric cancer. Stem cells were recruited into gastric glands, differentiated into gastric epithelial cells, and developed into intraepithelial cancer (47). Moreover, mesenchymal stem cells transduced with the telomerase hTERT gene display neoplastic potential and contributes to mesenchymal tumor formation (48). These results show that stem cells can be transformed under appropriate conditions and give rise to malignancy. Alternatively, a transformed, differentiated melanocyte may have undergone a dedifferentiation process and regained stem cell properties such as self-renewal (1). Within neural crest derivatives, dedifferentiation of one lineage into a precursor pool that then gives rise to other lineages has been well shown (49, 50). In addition, results from chronic myelogenous leukemia indicate that a lineage-restricted progenitor or mature cell can acquire stem cell privileges after oncogenic transformation (51, 52).

The loss of E-cadherin seems important for epithelial-mesenchymal transitions, which facilitate morphogenesis during development and tumor progression (melanomas; refs. 53, 54). In this study, the sustained expression of E-cadherin was observed in both WM3523 melanoma spheroid cells (89.60%) and its adherent melanoma counterparts (99.6%; data not shown). Only a minor fraction of both spheroid and adherent populations of WM115 express E-cadherin. These results suggest that E-cadherin expression is not critical for stem cell properties of melanoma spheroid cells.

We have discovered, however, that a small subpopulation of CD20+ melanoma cells harbors multipotent stem cells, as in the case of WM115 melanoma spheroid cells. WM3523 melanoma spheroid cells could maintain sphere-forming capacity even among the CD20 population; however, these negative cells had limited capacity to undergo mesenchymal differentiation. Most interestingly, CD20 has been identified by gene expression profiling as one of the top 22 genes that define aggressive melanomas (55). In metastatic melanomas, we have identified individual CD20+ tumor cells (data not shown). Monoclonal antibodies against CD20 have become a standard treatment for non-Hodgkin's lymphoma (56). CD20 seems a potential target for melanoma as well, although a correlation between differentiation ability and tumorigenicity is still under investigation by comparing CD20+ with CD20 fractions.

In summary, our studies clearly show the presence of a stem cell population in melanomas. Like all stem cells, melanoma spheroid cells are also capable of proliferation, differentiation, and self-renewal. In addition, melanoma spheroid cells possess higher tumorigenicity. Understanding the highly tumorigenic, stem cell origin of melanomas has important implications for efficient therapy.

Grant support: NIH grants CA25874, CA80999, CA10815, and CA76674 and Commonwealth Universal Research Enhancement Program, Pennsylvania Department of Health.

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 James Hayden for support with photography pigmentation of cell pellets, Gian Ascione for editorial assistance, and the staff at the Flow Cytometry Facility for their analyses.

1
Reya T, Morrison SJ, Clarke MF, Weissman IL. Stem cells, cancer, and cancer stem cells.
Nature
2001
;
414
:
105
–11.
2
Bonnet D, Dick JE. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell.
Nat Med
1997
;
3
:
730
–7.
3
Singh SK, Clarke ID, Terasaki M, et al. Identification of a cancer stem cell in human brain tumors.
Cancer Res
2003
;
63
:
5821
–8.
4
Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF. Prospective identification of tumorigenic breast cancer cells.
Proc Natl Acad Sci U S A
2003
;
100
:
3983
–8.
5
Singh SK, Hawkins C, Clarke ID, et al. Identification of human brain tumour initiating cells.
Nature
2004
;
432
:
396
–401.
6
Kondo T, Setoguchi T, Taga T. Persistence of a small subpopulation of cancer stem-like cells in the C6 glioma cell line.
Proc Natl Acad Sci U S A
2004
;
101
:
781
–6.
7
Setoguchi T, Taga T, Kondo T. Cancer stem cells persist in many cancer cell lines.
Cell Cycle
2004
;
3
:
414
–5.
8
Galli R, Binda E, Orfanelli U, et al. Isolation and characterization of tumorigenic, stem-like neural precursors from human glioblastoma.
Cancer Res
2004
;
64
:
7011
–21.
9
Yuan X, Curtin J, Xiong Y, et al. Isolation of cancer stem cells from adult glioblastoma multiforme.
Oncogene
2004
;
23
:
9392
–400.
10
Kath R, Jambrosic JA, Holland L, Rodeck U, Herlyn M. Development of invasive and growth factor-independent cell variants from primary human melanomas.
Cancer Res
1991
;
51
:
2205
–11.
11
Hendrix M, Seftor EA, Meltzer PS, et al. The stem cell plasticity of aggressive melanoma tumor cells. In: S Sell. Stem cells handbook. Totowa (NJ): Humana Press, Inc.; 2003. p. 297–306.
12
Reed JA, Finnerty B, Albino AP. Divergent cellular differentiation pathways during the invasive stage of cutaneous malignant melanoma progression.
Am J Pathol
1999
;
155
:
549
–55.
13
Fang D, Hallman J, Sangha N, et al. Expression of microtubule-associated protein 2 in benign and malignant melanocytes: implications for differentiation and progression of cutaneous melanoma.
Am J Pathol
2001
;
158
:
2107
–15.
14
Brocker EB, Magiera H, Herlyn M. Nerve growth and expression of receptors for nerve growth factor in tumors of melanocyte origin.
J Invest Dermatol
1991
;
96
:
662
–5.
15
Hendrix MJ, Seftor EA, Hess AR, Seftor RE. Vasculogenic mimicry and tumour-cell plasticity: lessons from melanoma.
Nat Rev Cancer
2003
;
3
:
411
–21.
16
Hsu MY, Shih DT, Meier FE, et al. Adenoviral gene transfer of β3 integrin subunit induces conversion from radial to vertical growth phase in primary human melanoma.
Am J Pathol
1998
;
153
:
1435
–42.
17
Satyamoorthy K, Muyrers J, Meier F, Patel D, Herlyn M. Mel-CAM-specific genetic suppressor elements inhibit melanoma growth and invasion through loss of gap junctional communication.
Oncogene
2001
;
20
:
4676
–84.
18
Klein CE, Steinmayer T, Kaufmann D, Weber L, Brocker EB. Identification of a melanoma progression antigen as integrin VLA-2.
J Invest Dermatol
1991
;
96
:
281
–4.
19
Huber MA, Kraut N, Park JE, et al. Fibroblast activation protein: differential expression and serine protease activity in reactive stromal fibroblasts of melanocytic skin tumors.
J Invest Dermatol
2003
;
120
:
182
–8.
20
Banerjee SS, Coyne JD, Menasce LP, Lobo CJ, Hirsch PJ. Diagnostic lessons of mucosal melanoma with osteocartilaginous differentiation.
Histopathology
1998
;
33
:
255
–60.
21
Cruz J, Reis-Filho JS, Lopes JM. Primary cutaneous malignant melanoma with lipoblast-like cells.
Arch Pathol Lab Med
2003
;
127
:
370
–1.
22
Giele H, Hollowood K, Gibbons CL, Wilson DJ, Athanasou NA. Subungual melanoma with osteocartilaginous differentiation.
Skeletal Radiol
2003
;
32
:
724
–7.
23
Hoorweg JJ, Loftus BM, Hilgers FJ. Osteoid and bone formation in a nasal mucosal melanoma and its metastasis.
Histopathology
1997
;
31
:
465
–8.
24
Grunwald MH, Rothem A. Desmoplastic cartilaginous formation in malignant melanoma.
J Cutan Pathol
1987
;
14
:
255
.
25
Thomson JA, Itskovitz-Eldor J, Shapiro SS, et al. Embryonic stem cell lines derived from human blastocysts.
Science
1998
;
282
:
1145
–7.
26
Xu C, Inokuma MS, Denham J, et al. Feeder-free growth of undifferentiated human embryonic stem cells.
Nat Biotechnol
2001
;
19
:
971
–4.
27
Hsu MY, Elder DE, Herlyn M. Melanoma: the Wistar (WM) melanoma cell lines. In: Masters JRW, Palsson B, editors. Human cell culture. London: Kluwer Academic Publisher; 1999. p. 259–74.
28
Somasundaram R, Swoboda R, Caputo L, et al. A CD4+, HLA-DR7-restricted T-helper lymphocyte clone recognizes an antigen shared by human malignant melanoma and glioma.
Int J Cancer
2003
;
104
:
362
–8.
29
Groszer M, Erickson R, Scripture-Adams DD, et al. Negative regulation of neural stem/progenitor cell proliferation by the Pten tumor suppressor gene in vivo.
Science
2001
;
294
:
2186
–9.
30
Satyamoorthy K, DeJesus E, Linnenbach AJ, et al. Melanoma cell lines from different stages of progression and their biological and molecular analyses.
Melanoma Res
1997
;
7
Suppl 2:
S35
–42.
31
Hsu MY, Li L, Herlyn M. Cultivation of normal human epidermal melanocytes in the absence of phorbol esters.
Methods Mol Med
2004
;
107
:
13
–28.
32
Pittenger MF, Mackay AM, Beck SC, et al. Multilineage potential of adult human mesenchymal stem cells.
Science
1999
;
284
:
143
–7.
33
Fang D, Setaluri V. Role of microphthalmia transcription factor in regulation of melanocyte differentiation marker TRP-1.
Biochem Biophys Res Commun
1999
;
256
:
657
–63.
34
Fang D, Kute T, Setaluri V. Regulation of tyrosinase-related protein-2 (TYRP2) in human melanocytes: relationship to growth and morphology.
Pigment Cell Res
2001
;
14
:
132
–9.
35
Le Douarin NM, Kalcheim C. The neural crest. In: Le Douarin NM, Kalcheim C, editors. Cambridge (UK): Cambridge University Press; 1999.
36
Jiang Y, Jahagirdar BN, Reinhardt RL, et al. Pluripotency of mesenchymal stem cells derived from adult marrow.
Nature
2002
;
418
:
41
–9.
37
Herlyn M, Balaban G, Bennicelli J, et al. Primary melanoma cells of the vertical growth phase: similarities to metastatic cells.
J Natl Cancer Inst
1985
;
74
:
283
–9.
38
Weiss S, Reynolds BA, Vescovi AL, Morshead C, Craig CG, van der Kooy D. Is there a neural stem cell in the mammalian forebrain?
Trends Neurosci
1996
;
19
:
387
–93.
39
Reynolds BA, Weiss S. Clonal and population analyses demonstrate that an EGF-responsive mammalian embryonic CNS precursor is a stem cell.
Dev Biol
1996
;
175
:
1
–13.
40
Dontu G, Abdallah WM, Foley JM, et al. In vitro propagation and transcriptional profiling of human mammary stem/progenitor cells.
Genes Dev
2003
;
17
:
1253
–70.
41
Toma JG, Akhavan M, Fernandes KJ, et al. Isolation of multipotent adult stem cells from the dermis of mammalian skin.
Nat Cell Biol
2001
;
3
:
778
–84.
42
Zhang SC, Wernig M, Duncan ID, Brustle O, Thomson JA. In vitro differentiation of transplantable neural precursors from human embryonic stem cells.
Nat Biotechnol
2001
;
19
:
1129
–33.
43
Reubinoff BE, Itsykson P, Turetsky T, et al. Neural progenitors from human embryonic stem cells.
Nat Biotechnol
2001
;
19
:
1134
–40.
44
Sieber-Blum M, Grim M, Hu YF, Szeder V. Pluripotent neural crest stem cells in the adult hair follicle.
Dev Dyn
2004
;
231
:
258
–69.
45
Fernandes KJ, McKenzie IA, Mill P, et al. A dermal niche for multipotent adult skin-derived precursor cells.
Nat Cell Biol
2004
;
6
:
1082
–93.
46
Nishimura EK, Jordan SA, Oshima H, et al. Dominant role of the niche in melanocyte stem-cell fate determination.
Nature
2002
;
416
:
854
–60.
47
Houghton J, Stoicov C, Nomura S, et al. Gastric cancer originating from bone marrow-derived cells.
Science
2004
;
306
:
1568
–71.
48
Serakinci N, Guldberg P, Burns JS, et al. Adult human mesenchymal stem cell as a target for neoplastic transformation.
Oncogene
2004
;
23
:
5095
–8.
49
Dupin E, Glavieux C, Vaigot P, Le Douarin NM. Endothelin 3 induces the reversion of melanocytes to glia through a neural crest-derived glial-melanocytic progenitor.
Proc Natl Acad Sci U S A
2000
;
97
:
7882
–7.
50
Dupin E, Real C, Glavieux-Pardanaud C, Vaigot P, Le Douarin NM. Reversal of developmental restrictions in neural crest lineages: transition from Schwann cells to glial-melanocytic precursors in vitro.
Proc Natl Acad Sci U S A
2003
;
100
:
5229
–33.
51
Jamieson CH, Ailles LE, Dylla SJ, et al. Granulocyte-macrophage progenitors as candidate leukemic stem cells in blast-crisis CML.
N Engl J Med
2004
;
351
:
657
–67.
52
Manz MG, Miyamoto T, Akashi K, Weissman IL. Prospective isolation of human clonogenic common myeloid progenitors.
Proc Natl Acad Sci U S A
2002
;
99
:
11872
–7.
53
Thiery JP. Epithelial-mesenchymal transitions in development and pathologies.
Curr Opin Cell Biol
2003
;
15
:
740
–6.
54
Hsu MY, Wheelock MJ, Johnson KR, Herlyn M. Shifts in cadherin profiles between human normal melanocytes and melanomas.
J Investig Dermatol Symp Proc
1996
;
1
:
188
–94.
55
Bittner M, Meltzer P, Chen Y, et al. Molecular classification of cutaneous malignant melanoma by gene expression profiling.
Nature
2000
;
406
:
536
–40.
56
Coiffier B, Lepage E, Briere J, et al. CHOP chemotherapy plus rituximab compared with CHOP alone in elderly patients with diffuse large-B-cell lymphoma.
N Engl J Med
2002
;
346
:
235
–42.