VOLUME 26 ⅐ NUMBER 17 ⅐ JUNE 10 2008

JOURNAL OF CLINICAL ONCOLOGY REVIEW ARTICLE

Melanoma Stem Cells: The Dark Seed of Susan E. Zabierowski and Meenhard Herlyn

From The Wistar Institute, Philadelphia, PA. ABSTRACT Submitted November 27, 2007; accepted March 4, 2008. Cells with stem-cell markers and features have recently been identified in melanoma tissues and Authors’ disclosures of potential con- cell lines. Melanoma stem-like cells possess many traits of tumor-initiating or tumor stem cells flicts of interest and author contribu- including self-renewal capacity, high tumorigenicity, and differentiation into various mesenchymal tions are found at the end of this lineages, including melanocytic cells. Four subpopulations of melanoma-initiating cells have been article. distinguished: CD20ϩ, CD133ϩ, label-retaining or slow-cycling cells, and side-population cells with Corresponding author: Meenhard Herlyn, high efflux activities. Whether these are distinct or overlapping populations is currently under DVM, DSc, The Wistar Institute, 3601 investigation. Ongoing studies are dissecting and characterizing the hierarchy of these subpopu- Spruce St, Philadelphia, PA 19104; e-mail: [email protected]. lations within a malignant lesion. Understanding these and the dynamics of clonal dominance will aid in the development of novel therapeutic strategies. © 2008 by American Society of Clinical Oncology J Clin Oncol 26:2890-2894. © 2008 by American Society of Clinical Oncology 0732-183X/08/2617-2890/$20.00

DOI: 10.1200/JCO.2007.15.5465 and the isolation of stem-like cells has recently been INTRODUCTION described in melanoma.5-9 Here, we briefly summa- Melanoma development has classically been per- rize our current knowledge of melanoma stem cells. To date, four subpopulations including, CD20ϩ, ceived as a stepwise process in which mature mela- ϩ nocytes in the epidermis progressively acquire genetic CD133 , label-retaining cells, and side-population mutations in oncogenes or tumor-suppressor (SP) cells (Fig 2) have been identified in melanoma that provide survival and growth advantages allowing tissues and cell that possess stem-cell characteristics, uninhibited proliferation, ultimately leading to meta- defined as self-renewal capacity, differentiation po- static melanoma (Fig 1A). Yet clinically, only 26% tential, and high tumorigenicity. of observed evolve from nevi, and less 1 than 50% are associated with dysplastic nevi. This in- MELANOMA SPHERES dicates that the majority of melanomas arise from normal-appearing skin and not from dysplastic nevi, Utilizing an embryonic –based media, our suggesting that melanoma development may not fol- laboratory has propagated cells from both meta- low the classical linear mode of progression (Fig 1). static melanomas specimens and cell lines that har- Thus, although the apparent target of transformation is bor properties of cancer stem cells.9 Melanoma differentiated melanocytes, melanomas may also be cells capable of surviving and proliferating in this derived from transformed melanocytic stem or pro- selective media grew as nonadherent spheres, genitor cells (Fig 1B).2 Yet on the basis of clinical and termed melanoma spheres, similar to the mammo- pathological evidence, melanocyte stem cells the in and neurospheres generated from breast and brain hair follicles of mice and humans3,4 are not the targets cancer, respectively. Distinct from their bipolar, of transformation, suggesting that an alternative source spindle-shaped adherent counterparts grown in of melanocyte-producing cells may be hit. Preliminary conventional media, melanoma spheres could dif- studies from our laboratory indicate the existence of a ferentiate along multiple mesenchymal lineages, in- neural-crest stem cell with differentiation potential for cluding adipocytes, chondrocytes, and osteoblasts. melanocytes in the dermis. Whether these cells migrate Spheres were more than 10-fold more tumorigenic through the basement membrane to disburse among than adherent cells, could be reisolated after in vivo the basal-layer keratinocytes, and whether they are transplantation, and could persist despite serial more prone to transformation than mature melano- cloning, demonstrating self-renewal capacity. cytes, remains to be determined. Cancer stem cells, also known as tumor-initiating CD20 cells, have been defined in a variety of human and mouse tumors through known and novel stem-cell Although maintaining expression of melanoma- markers, as well as stringent self-renewal and tumor- associated markers MCAM and ␤3 integrin, indica- igenicity assays. Expression of stem-cell markers tive of their melanocytic origin, a subpopulation of

2890 © 2008 by American Society of Clinical Oncology Information downloaded from jco.ascopubs.org and provided by UNIV PENNSYLVANIA on October 28, 2008 from 130.91.8.12. Copyright © 2008 by the American Society of Clinical Oncology. All rights reserved. Melanoma Stem Cells

A Linear tumor progression model

Dysplastic RGP VGP Fig 1. Linear melanoma progression model Normal Nevus Metastatic Nevus Primary Primary versus the melanoma stem-cell and tumor microenvironment model. (A) Melanoma development is commonly perceived as a gradual process in which normal mature melanocytes acquire mutations that lead from benign to dysplastic nevi progressing B Endothelial toward radial growth phase (RGP) to vertical Cells growth phase (VGP), ultimately followed by Stem metastatic melanoma. (B) The melanoma Stem-cell and tumor Inflammatory Fibroblasts stem-cell and tumor microenvironment microenvironment model Cell Cells model posits that melanoma stem cells lead directly to the observed clinical outcomes without progressing through intermediates. Additionally, toward the late stages, the tu- mor microenvironment composed of sur- rounding and/or recruited fibroblasts and endothelial and inflammatory cells contrib- utes to and supports metastases. Normal Benign Intermediate Primary Metastatic

melanoma spheres was found to consistently express CD20, a cell and differentiate into adipogenic, chondrogenic, and osteogenic surface marker normally associated with B cells, that was completely lineages. Moreover, CD20ϩ cells could be identified in melanoma absent in the corresponding adherent cells grown in conventional specimens.9 The frequency of CD20 expression among metastatic media.9 CD20ϩ sphere cells could rapidly proliferate, reform spheres, melanomas and whether its expression in melanomas can be

CD20 (Fang, et al.) CD133 (Frank, et al., Mozani, et al., Klein, et al.) Side population Melanoma stem (Grichnik, et al.) cell Label retaining (unpublished) ? Progenitor cells

Fig 2. Tumor heterogeneity is supported by melanoma stem cells. Known and novel stem-cell markers, such as CD20 and CD133, and stem-cell characteristics, includ- ing side-population and label-retaining cells, that have been distinguished in melanoma. Trans-amplifying Self-renewing melanoma stem cells sustain cells the stem-cell population while giving rise to progenitor cells that may or may attain the capacity to self-renewing. These progenitor and trans-amplifying cells can give rise to differentiating clones of varying dominance that contribute to the overall tumor hetero- geneity found in melanomas.

Differentiated cells

www.jco.org © 2008 by American Society of Clinical Oncology 2891 Information downloaded from jco.ascopubs.org and provided by UNIV PENNSYLVANIA on October 28, 2008 from 130.91.8.12. Copyright © 2008 by the American Society of Clinical Oncology. All rights reserved. Zabierowski and Herlyn correlated with clinical outcome are currently under investigation. Ongo- pioneered by Goodell et al13 and found to enrich for hematopoietic stem ing studies are also in progress to assess whether elimination of CD20ϩ cells, has been adapted to identify and isolate stem/progenitor cells from cells from melanomas can abolish tumorigenic potential. multiple tissues including umbilical cord blood, skeletal muscle, mam- Highly specific anti-CD20 treatments that currently exist may be mary glands, lung, liver, epidermis, forebrain, testis, heart, kidney, and used to distinctly target these cells. Rituximab, a highly effective anti- prostate. Additionally, cancers such as neuroblastoma, breast, and lung, CD20 monoclonal antibody, has been successfully used to treat non- and tumor cell lines including melanomas, have also been shown to Hodgkin’s lymphomas and B-cell chronic lymphocytic leukemias.10 harbor SP with stem-cell characteristics.14 Melanoma SP cells have been The high efficacy of rituximab has led to the development of other characterized as small in size and less proliferative than their larger, anti-CD20 radioimmunoconjugates including ibritumomab tiuxetan non-SP counterparts.15 Despite the slower rate of proliferation, SP cells and tositumomab, approved for use in relapsed or refractory B-cell stained positive for the neural-crest stem-cell marker nestin and lymphomas.11 Thus, with such highly selective and effective treat- retained a greater capacity than non-SP cells to expand in vitro, ments to CD20 readily available, it is possible to target CD20- giving rise to larger cells. Although this study did not further assess expressing melanoma stem cells. the capacity of SP cells to give rise to both SP and non-SP cells, nor whether size was linked to tumorigenicity, it does suggest that a CD133 hierarchy exists within melanoma, in which heterogeneous popu- lations compete with each other for dominance. CD133, a transmembrane also known as prominin-1, is The heighten efflux capacity of the SP is mediated primarily through normally expressed on undifferentiated cells including endothelial pro- the expression of the ABC transporters, a superfamily of with genitor cells, hematopoietic stem cells, fetal brainstem cells, and prostate more than 30 members that translocates both hydrophilic and hydropho- epithelial cells, but has also been exploited to identify and purify cancer bic molecules across the , thought to provide protection 16 stem cells from various solid tumors including brain, prostate, and pan- from harmful exogenous substances. The major representative of the SP 17 creatic cancers. Thus, despite its unknown function in stem-cell biology, phenotype is the ABC transporter ABCG2, which has also been shown ϩ 5 CD133 appears to mark both normal and cancer stem cells. In addition, to be coexpressed on a subpopulation of CD133 melanoma cells. How- ever, whether ABCG2 expression coincides with SP has not been exam- expression of CD133 has been identified in melanomas and primary ϩ human melanocytes.5,7,8,12 Frank et al8,12 concluded that 0.2% to 0.5% of ined, nor have the properties of ABCG2 melanoma cells been explored. primary human-cultured melanocytes and 0.5% to 2% of G3361 mela- Contrary to this study, however, others have found that ABCG2 is not 18 noma cells comprise CD133 cells, demonstrating that a distinct subset of expressed in melanoma specimens. It should be noted that, in some melanocytic cells are marked by the CD133 progenitor phenotype. Klein systems, only a fraction of SP cells express ABCG2, suggesting that other et al7 observed significant increases in the expression of stem-cell markers transporters such as ABCB1, ABCC1, or ABCA2 may be involved in CD133, CD166, and nestin in primary and metastatic melanomas com- mediating the SP phenotype. pared with banal nevi. Increased melanoma aggressiveness corresponded Likely, the SP is not solely composed of melanoma tumor– ϩ with greater expression of these markers,7 suggesting that during mela- initiating cells. Despite the coexpression of ABCG2 with CD133 noma progression, stem-cell markers become more evident due to the melanoma cells, it is not known whether this transporter contributes ϩ ϩ increased dysregulation of stem-/progenitor-cell function and prolifera- to SP, nor has it been determined whether ABCG2 /CD133 mela- tion. Similarly, Mozani et al,5 determined that less than 1% of cells within noma cells are tumor-initiating cells. Thus, further in-depth analysis is metastatic melanomas express CD133. Moreover, only the CD133ϩ cells required to determine the constituents of the SP and to assess the role collected from these biopsies induced tumors in nonobese diabetic/severe of these transporters in melanoma establishment and resistance. combined immunodeficiency mice, whereas the CD133– fraction failed ABCB5, a novel energy-dependent drug efflux transporter highly to regenerate tumors, indicating that only a small percentage of cells are similar to ABCB1, was identified and characterized by Frank et al8,12 as capable of recapitulating tumor growth. Surprisingly, long-term cultured a determinant of membrane potential and a regulator of cell fusion. cell lines, such as WM115, expressed CD133 at nearly 100%, whereas only This transporter was also found to mediate doxorubicin drug resis- a subpopulation of these cells coexpressed adenosine triphosphate–bind- tance in melanoma, yet is speculated to have an even broader range of ing cassette (ABC) G2, a transporter commonly associated with stem chemotherapeutic activity. ABCB5 marked 11% of primary melano- cells.5 CD133 levels significantly decreased in tumors propagated in mice; cytes, 3% of melanoma cells, and 13% of cisplatin-resistant melanoma ϩ yet on in vitro reculturing, these cells readily regained high levels of CD133 cells. Moreover, ABCB5 cells highly coexpressed CD133 and could ϩ ϩ ϩ expression, suggesting that in vivo tumor growth conditions can signifi- enrich for CD133 cells, indicating that ABCB5 /CD133 may mark cantly differ from in vitro culture conditions and can influence observed melanoma stem cells.8,12 More recently, this ABC-transporter alone expression patterns. Combined, these studies establish the expression of has been shown to mark melanoma cells, harvested from fresh speci- CD133 in melanomas, supporting the notion that melanomas contain a mens as well as from melanoma cell lines, that are capable of initiating stem-cell component. However, comprehensive studies to correlate tumors in xenotransplantation models.19 Whether this transporter CD133 expression to self-renewal, differentiation, and tumorigenicity also contributes to the SP phenotype is not yet known. remains to be established.

SP LABEL RETAINING

SP is a widely utilized flow cytometric technique based on the heightened Stem cells are typically undifferentiated, infrequently dividing cells efflux capacity of stem cells to exclude fluorescent dyes. This technique, that normally remain quiescent until their participation is required

2892 © 2008 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY Information downloaded from jco.ascopubs.org and provided by UNIV PENNSYLVANIA on October 28, 2008 from 130.91.8.12. Copyright © 2008 by the American Society of Clinical Oncology. All rights reserved. Melanoma Stem Cells during tissue regeneration. Many have exploited this property to iden- purification and characterization of these cells from the bulk tumor tify stem cells via label-retaining experiments. Through pulse-chase population. For clinical applications, melanoma cancer stem-cell sur- experiments with a DNA-traceable agent such as BrdU, which is face molecules would allow their diagnostic identification in melanoma rapidly titrated out as cells divide, only those cells that have infre- biopsiesandmaybeusedasprognosticindicators.Furthermore,theymay quently divided retain BrdU. These slow-cycling, label-retaining cells pose as targets for antibody-based therapies that would selectively elimi- have in many instances defined the location of stem cells from tissues nate these cells. Knowledge of the mechanisms important for survival and as diverse as skin and lung.2,3,20 Others have utilized this approach to maintenance of melanoma stem cells will benefit in the development of prospectively seek out cancer stem cells within tissues.2 Meanwhile, strategies aimed at destroying these cells. Pathways and signaling mole- others have adapted this technique to in vitro tissue culture systems.21 cules involved in stem-cell homeostasis may be used to determine Although the ability of cells to retain label is indicative of slow-cycling whether and how they contribute to de-differentiation and transforma- stem cells, this method is not full proof given that recent evidence tion to melanoma stem cells and what the consequences of blocking these suggests that not all stem cells can be identified on the basis of BrdU pathways are. Many of the overactive signaling pathways and mutations label retention.22 It may, however, provide important clues regarding identified to date through expression profiling are largely markers of tumor heterogeneity (Fig 2). This type of adaptation has been applied heterogeneous tumor cells. However, because cancer stem cells typically to in vitro cultured melanoma cells utilizing the fluorescent dye CTO constitute only a minority of the total tumor population,9 the gene expres- to segregate and monitor the proliferation of phenotypically different sion profiles of melanoma stem cells are in all probability lost to the bulk of cells. In preliminary studies, we have identified a label-retaining mel- the tumor population. As a result, most treatment strategies are currently anoma cell population that can self-renew and differentiate, and that aimed at the bulk tumor population. Thus, combinations of markers will are highly tumorigenic, fulfilling the basic stem-cell criteria. Such cells be required to define the ultimate stem cells because the current popula- also have high efflux activities. These combined properties make label- tions most likely represent a variety of cells and not true stem cells. retaining cells key in the search for drug-resistant cells that are not only The notorious resistance of melanoma cells against all major resistant to chemotherapeutic agents but that may also be unaffected chemotherapeutic drugs strongly suggests that this tumor type pos- by drugs that block proliferation pathways. sesses major properties for drug resistance. Whether this drug resis- tance resides in cells with high ABC transporter activities or in the nonproliferative label-retaining cells remains to be determined. Clin- CONCLUSIONS AND PERSPECTIVES ically, it can be predicted that melanomas need to be treated with dual or even triple agents. Treatment strategies targeting overactive signal- The biologic and molecular characterization of melanoma stem cells is ing pathways should have a major immediate impact on the large, still in the early phase. The origin of melanoma stem cells has yet to be rapidly proliferating population of melanomas cells, resulting in tu- determined. Whether melanoma stem cells are derived from melano- mor debulking. Yet, without targeting the minor hidden fraction of cyte stem cells, melanocyte progenitors, or more mature melanocytes perhaps quiescent or drug-resistant melanoma stem cells, the goal of that have de-differentiated remains unclear at this point. Also, curing this disease may be unattainable. Despite initial remissive re- whether any of the four subpopulations isolated from cultures and (in sponses, failure to abolish self-renewing cancer stem cells sets the scene part) from specimens overlap or are indeed the true tumor-initiating for melanoma regrowth. Thus, multimodal treatment strategies that cells requires further characterization. With the exception of the label- combine therapeutic agents aimed at melanoma stem cells simulta- retaining cells, all subpopulations have been detected in freshly iso- neously with regimens intended for the bulk tumor population should lated tumors, yet how these correlate with clinical outcome remains to effectively eliminate melanomas and reduce the risk of relapse. be determined. Although each subpopulation can self-renew, it is not known yet whether any of these subpopulations will exhaust their self-renewal potential over time. Extensive genomewide gene expres- AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS sion analyses suggest that each population has a distinct profile, with OF INTEREST the SP having the least “stemness” (unpublished data), confirming data in other malignancies. Melanoma sphere cells, in contrast to The author(s) indicated no potential conflicts of interest. other malignancies, can be readily isolated and cultured from patient lesions at a success rate of more than 80% (unpublished data); how- AUTHOR CONTRIBUTIONS ever it is unclear at this point which subpopulation is being lost in culture. Thus, titration studies are ongoing in the nonobese diabetic/ Conception and design: Meenhard Herlyn, Susan E. Zabierowski severe combined immunodeficiency mouse model to determine the Provision of study materials or patients: Meenhard Herlyn minimum number of cells from which population is required for Data analysis and interpretation: Susan E. Zabierowski tumor formation. Further classification of cell-surface molecules spe- Manuscript writing: Meenhard Herlyn, Susan E. Zabierowski cific to melanoma stem cells, such as CD20 and CD133, will allow for Final approval of manuscript: Meenhard Herlyn, Susan E. Zabierowski

2. Kim CF, Jackson EL, Woolfenden AE, et al: 4. Yu H, Fang D, Kumar SM, et al: Isolation of a REFERENCES Identification of bronchioalveolar stem cells in nor- novel population of multipotent adult stem cells mal lung and . Cell 121:823-835, 2005 from human hair follicles. Am J Pathol 168:1879- 1. Bevona C, Goggins W, Quinn T, et al: Cuta- 3. Nishimura EK, Jordan SA, Oshima H, et al: 1888, 2006 neous melanomas associated with nevi. Arch Der- Dominant role of the niche in melanocyte stem-cell 5. Monzani E, Facchetti F, Galmozzi E, et al: matol 139:1620-1624, 2003 fate determination. Nature 416:854-860, 2002 Melanoma contains CD133 and ABCG2 positive

www.jco.org © 2008 by American Society of Clinical Oncology 2893 Information downloaded from jco.ascopubs.org and provided by UNIV PENNSYLVANIA on October 28, 2008 from 130.91.8.12. Copyright © 2008 by the American Society of Clinical Oncology. All rights reserved. Zabierowski and Herlyn cells with enhanced tumourigenic potential. Eur J 11. Fanale MA, Younes A: Monoclonal antibodies 33342 efflux pump and is preferentially expressed Cancer 43:935-946, 2007 in the treatment of non-Hodgkin’s lymphoma. Drugs by immature human hematopoietic progenitors. 6. Mihic-Probst D, Kuster A, Kilgus S, et al: 67:333-350, 2007 Blood 99:507-512, 2002 Consistent expression of the stem cell renewal 12. Frank NY, Pendse SS, Lapchak PH, et al: 18. Deichmann M, Thome M, Egner U, et al: The factor BMI-1 in primary and metastatic melanoma. Regulation of progenitor cell fusion by ABCB5 chemoresistance gene ABCG2 (MXR/BCRP1/ Int J Cancer 121:1764-1770, 2007 P-glycoprotein, a novel human ATP-binding cassette ABCP1) is not expressed in melanomas but in single 7. Klein WM, Wu BP, Zhao S, et al: Increased transporter. J Biol Chem 278:47156-47165, 2003 neuroendocrine carcinomas of the skin. J Cutan expression of stem cell markers in malignant mela- 13. Goodell MA, Brose K, Paradis G, et al: Isola- Pathol 32:467-473, 2005 tion and functional properties of murine hematopoi- 19. Schatton T, Murphy GT, Frank NY, et al: noma. Mod Pathol 20:102-107, 2007 etic stem cells that are replicating in vivo. J Exp Med Identification of cells initiating human melanomas. 8. Frank NY, Margaryan A, Huang Y, et al: 183:1797-1806, 1996 Nature 451:345-349, 2008 ABCB5-mediated doxorubicin transport and che- 14. Hadnagy A, Gaboury L, Beaulieu R, et al: SP 20. Tumbar T, Guasch G, Greco V, et al: Defining moresistance in human malignant melanoma. Can- analysis may be used to identify the epithelial stem cell niche in skin. Science 303: cer Res 65:4320-4333, 2005 populations. Exp Cell Res 312:3701-3710, 2006 359-363, 2004 9. Fang D, Nguyen TK, Leishear K, et al: A 15. Grichnik JM, Burch JA, Schulteis RD, et al: 21. Shinin V, Gayraud-Morel B, Gomes D, et al: tumorigenic subpopulation with stem cell properties Melanoma, a tumor based on a mutant stem cell? Asymmetric division and cosegregation of template in melanomas. Cancer Res 65:9328-9337, 2005 J Invest Dermatol 126:142-153, 2006 DNA strands in adult muscle satellite cells. Nat Cell 10. Cvetkovic´ RS, Perry CM: Rituximab: A re- 16. Dean M: The genetics of ATP-binding cassette Biol 8:677-687, 2006 view of its use in non-Hodgkin’s lymphoma and transporters. Methods Enzymol 400:409-429, 2005 22. Kiel MJ, He S, Ashkenazi R, et al: Haematopoi- chronic lymphocytic leukaemia. Drugs 66:791- 17. Scharenberg CW, Harkey MA, Torok-Storb B: etic stem cells do not asymmetrically segregate chro- 820, 2006 The ABCG2 transporter is an efficient Hoechst mosomes or retain BrdU. Nature 449:238-242, 2007

■■■

2894 © 2008 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY Information downloaded from jco.ascopubs.org and provided by UNIV PENNSYLVANIA on October 28, 2008 from 130.91.8.12. Copyright © 2008 by the American Society of Clinical Oncology. All rights reserved.