The field of melanoma cancer stem cells has generated both great interest and confusion. The therapeutic implications of cancer stem cells are vast in that a subset of melanoma cells may be responsible for tumor progression, relapse, and chemotherapeutic resistance. However, it has been difficult to reach conclusions to date, as several studies have generated differing results depending on the markers and assays used (Monzani et al., 2007; Schatton et al., 2008; Quintana et al., 2008; Held et al., 2010; Prasmickaite et al., 2010). In these various studies, one finding shared by every study has been that cell surface markers are expressed in a heterogeneous manner, allowing for functional evaluation of cellular subsets. These markers have varied from melanoma to melanoma and have shown clonigenic or tumorigenic differences in subpopulations in some instances, but not others. While it is not clear why many of these differences exist from study to study, it has become clear that the assays used to define cancer stem cells have a large effect. The gold standard in the field has been determination of the number of cells required for tumor formation in xenografting studies. The use of NOD-SCID IL2Rγ−/− immunodeficient mice in place of NOD-SCID mice and incorporation of co-injection of Matrigel increases the measured frequency of melanoma cancer stem cells by >10 000 fold (Quintana et al., 2008; Schatton et al., 2008). Based on findings that an average of 27% of unselected human melanoma cells can form tumors in these optimized assays and the inability to define markers that distinguish tumorigenic from non-tumorigenic melanoma subsets, it has been suggested that human malignant melanoma may be a relatively homogeneous cancer with regard to tumorigenic potential, despite cell-to-cell differences in surface antigen expression. This also implies that melanoma may not follow the cancer stem cell (CSC) model that suggests that a lower proportion of cells may be responsible for tumor formation and reestablishment of the original tumor cell hierarchy (Quintana et al., 2008). A recent study by Roesch et al. confirms that a high proportion of human melanoma cells can be tumorigenic and further demonstrates some important functional consequences in human melanoma cells with reduced expression of the intracellular protein JARID1B.
JARID1B belongs to the highly conserved family of jumonji/ARID1 histone demethylases recently discovered to be capable of removing mono-, di-, or trimethyl groups from histone lysine residues. Demethylation of these residues can lead to alteration of chromatin and transcription of previously silenced genes. Interestingly, in normal tissue, JARID1B is highly expressed in the testes and stem cell populations of the bone marrow. In melanocytic tumors, JARID1B is highly expressed in many cells within benign nevi, but is expressed in a smaller fraction of advanced-stage melanoma cells (∼5–10%). These findings and the demonstration that JARID1B expression correlates with reduced proliferation led to the suggestion that JARID1B is a tumor suppressor in melanoma. However, an alternative interpretation is that the slow-cycling, minor cell population defined by JARID1B expression could also represent cancer stem cells. Therefore, given the expression of JARID1B in stem cell populations and its variable expression in melanoma, the authors addressed two questions: (i) does JARID1B define a melanoma cancer stem cell subset, and (ii) is JARID1B expression required for continuous melanoma growth?
In order to do this, Roesch et al. examined established melanoma cell lines for JARID1B expression in two different culture conditions. Under adherent cell culture conditions, JARID1B expression was relatively constant but variable in intensity, whereas cells grown in embryonic stem cell medium (hESCM4) as spherical colonies resulted in high JARID1B expression, but only in a minor proportion of relatively smaller-sized cells. JARID1B+ cells did not exhibit co-expression of the proliferation marker Ki-67, suggesting that JARID1B+ cells cycle slowly. To enrich for JARID1B+ cells, the authors transduced melanoma lines with a lentiviral-based vector that drives EGFP expression from a JARID1B promoter construct (J/EGFP). The 2% of WM3734 melanoma cells that showed high expression of EGFP (and hence, JARID1B) also retained membrane dye, confirming relatively slow cell cycling. Interestingly, most of the membrane dye-negative J/EGFP− cells died after 3 weeks in culture in hESCM4.
Growth as non-adherent cellular spheres is a phenotype associated with certain stem cell populations. Differences in this property were evaluated in J/EGFP+ and J/EGFP− fractions purified by FACS. Interestingly, isolated J/EGFP+ cells showed less BrdU incorporation compared to JARID1B− cells after the first few days, but after day 10, J/EGFP+ progeny proliferated more rapidly than J/EGFP− cells. Over time, J/EGFP+ cells grew more viable spheres compared to the J/EGFP− cells. These findings suggested that, in vitro, the JARID1B+ population may have greater stem-like characteristics. However, xenografts of 100, 10 or 1 purified cell showed no difference between J/EGFP+ or J/EGFP− subsets with respect to tumorigenic frequency or growth, in contrast to the in vitro sphere-formation data. Furthermore, upon analysis of the tumors that formed, similar fractions of JARID1B-positive and -negative cells were present in the various injection conditions, suggesting that the ability to regenerate marker heterogeneity was not unique to either subset.
To address this paradox, the authors asked whether JARID1B is needed for tumor maintenance (i.e. continuous tumor growth) by performing serial xenografting experiments. The authors chose shRNA knockdown of JARID1B as the experimental methodology to address this. Using three melanoma lines, JARID1B protein expression was shown to vary considerably and knockdown efficiencies were variable (from ∼25 to 60%), yet showed convincing stable protein loss in one line (WM3734). Initial growth of JARID1B knockdown lines in conventional culture conditions showed a higher proliferative rate from days 7–10 relative to vector control lines, yet at days 30–33, no significant proliferative differences were observed. The authors then evaluated the effects of JARID1B knockdown in sphere-growth conditions with hESCM4 media. Knockdown of JARID1B in sphere-growth conditions led to proliferative exhaustion during days 28–39 and a dramatic increase in cell death. Therefore, one might expect lack of JARID1B expression (JARID1B− cells) to also produce lower tumor volumes over time. The authors further performed serial xenotransplantations in NOD-SCID IL2Rγ−/− mice with 10 000 JARID1B knockdown cells and vector control cells. They showed that the first round of inoculated JARID1B knockdown cells form significantly larger tumors by weeks 8–9 compared to controls. However, tumor volumes were significantly lower in serial xenograft rounds 2–4 with JARID1B knockdown cells compared to vector control cells, analogous to the decreased proliferation seen in later time points in the sphere culture conditions. These findings are in contrast to published experiments involving serial transplantation of melanoma cells in which <1% of serially transplanted tumors fail to reform tumors (Quintana et al., 2008).
These findings identify JARID1B as a target for therapeutic intervention in melanoma. The prediction is that inhibition of JARID1B may induce melanoma exhaustion following a proliferative phase (Figure 1). This unique mechanism would need to be taken into account when evaluating clinical responses and determining the timing of possible biomarker correlates of response. The findings also demonstrate that JARID1B expression does not define melanoma cancer stem cell subsets in the cell lines examined. The high rates of tumor formation and inability to define cancer stem cell subsets are similar to the findings of Quintana and co-workers. This also raises the question of whether the ∼73% of non-tumorigenic human melanoma cells identified by Quintana et al. and roughly 50% non-tumorigenic cells identified by Roesch et al. can be distinguished from tumorigenic cells using cell surface markers, as has been accomplished in mouse models of melanoma (Held et al., 2010).
With the rapid progress in this field, it is not surprising that the concepts associated with melanoma cancer stem cells have evolved to incorporate observations that cells capable of tumor formation need not be rare, that many cancers do not necessarily originate from tissue stem cells, that exceptions to this model in which cells capable of tumor formation do not always generate the phenotypic heterogeneity present in the starting tumor material, and that more than one population of phenotypically distinct cells may be capable of tumor formation. The recent manuscript by Roesch et al. further adds to the complexity of this field by demonstrating the requirement for a stem cell-associated gene, JARID1B, for the continuous growth of a melanoma cell line upon serial transplantation in immunodeficient mice. In addition, the cellular populations defined by the presence or absence of this gene product can readily interconvert, demonstrating the plasticity of subsets defined by this marker. The hope remains that further progress in the study of melanoma cancer stem cells will result in improvements in melanoma patient care.