This is the authors' abstract. We don't add key points for this paper.
In describing the perils of modelling, the statistician George Box famously claimed that ‘All models are wrong, but some are useful’. Over the last few decades, several useful conceptual models of tumor progression have been proposed to describe cancer initiation, metastasis and therapeutic resistance. These include the clonal evolution (CE), cancer stem cell (CSC) and phenotype switching (PS) models. Recently, the PS model has become increasingly invoked by melanoma researchers and clinicians as it seems to best explain how a dynamic and adaptive reprogramming of melanoma cells, independent of hierarchical tumor organization, can drive tumor progression. When developing this model, we and others incorporated aspects of CSC and CE to explain how plastic phenotypic states could allow melanoma cells to respond to microenvironmental cues by downregulating a proliferative, melanocytic program, and by activating a neural crest stem cell module more conducive to invasion and metastasis (Fig. 1). Despite the popularity of PS, each of the other models still has useful explanatory power. For instance, according to a selection-based model such as CE, microenvironmental forces act on genetic subclonal populations, which then evolve according to their differential responses to a selective pressure. This may be particularly relevant in a therapeutic setting where targeted inhibitors might select for subclones that remain proliferative in the presence of the drug, and where less-fit subclones cannot compete. In contrast, the CSC model was the preferred model in melanoma until it was shown that most cells in a tumor are capable of acting as tumor initiating cells, ruling out the possibility that just a few stem cells were necessary 1. However, the reactivation of neural crest, stem cell-like developmental programs during tumor progression in invasive phenotype melanoma cells may be seen as a variation on the CSC theme 2. A recently published study provides further support for the PS model 3. Marconi and colleagues used Reflectance Confocal Microscopy to group 12 melanoma biopsies into three subtypes. All samples were stained by immunohistochemistry and analysed for the expression of HIF1alpha and CD271. They observed an anticorrelation between HIF1alpha and CD271 expression. A correlation of the stainings with phases of progression led them to conclude that HIF1alpha is expressed in more aggressive melanoma subtypes, whereas CD271 was expressed in the slow growing, superficial melanoma and was progressively lost during melanoma progression 3. The two proteins that Marconi et al. investigated are well-known in melanoma biology. Both have been reported to be important players in PS whereby proliferative melanoma cells acquire mesenchymal, invasive characteristics 4, 5. HIF1alpha is a mediator of the hypoxic response, which we and others have shown can induce a switch from proliferative to invasive phenotype cells, thereby increasing melanoma invasion 6 (S1,S2). CD271 on the other hand is a marker of stem-like cells for which the data are contradictory (S3,S4,S5). However, their postulation that CD271 is progressively lost during melanoma progression does not agree with previous reports that CD271 expression in melanoma is associated with invasiveness (S4,S6). Hypoxia is not the only factor able to induce a phenotype switch. Many studies have observed an (epithelial to mesenchymal transition) EMT-like transformation of melanoma cells from a melanocytic, differentiated state to a more dedifferentiated, mesenchymal state. During this process, the typical melanocytic markers are downregulated and mesenchymal markers are upregulated. In addition to hypoxia, TGFbeta has been shown to induce an EMT equivalent in melanoma, thereby pushing the cells towards an invasive phenotype 7. These processes are probably not so much involved in the initial melanoma formation but seem to play a key role in tumor progression and even therapeutic resistance 8. Generally, tumors are composed of numerous genetic and epigenetic subclones which show a high transcriptional diversity and adaptability. Tissue-dependent microenvironments further complicate the situation which makes targeting melanoma a difficult endeavour. Some studies comparing tumors within the same patients have found multiple different mechanisms pointing towards the presence of a high number of genetic subclones. Newly developed single-agent drugs are likely to target only one subset of melanoma cells explaining the common development of resistance. Evidence suggests that the majority of genetic resistance mechanisms lead to a reactivation of the MAPK pathway (S7,S8). Future therapies will be composed of multiple drugs in order to find a way to cope with this heterogeneity in melanoma. Despite an enormous effort and thousands of sequenced tumors in hundreds of patients, there is still a large proportion of cases where no mutation could be found to explain treatment resistance. In addition to a long tail of potential oncogenic drivers, among which PPP6C, RAC1 and MAP2K1 may be the most prominent ones (S9), the modulation of epigenetic and transcriptional states may also play an important role in progression and resistance mechanisms, as predicted by the PS model (S10). We have just started to understand the different factors driving melanoma, but the accumulating evidence suggests that PS contributes to the strategies that melanoma cells use to escape microenvironmental and therapeutic pressures 8 (S11,S12). The systematic application of increasingly more powerful panomics technologies on better curated biobank collections will improve our ability to more accurately study tumor biology in vitro and in vivo and will contribute to more sophisticated and useful conceptual models 9. Ultimately, of course, the usefulness of any tumor model will be measured by how well it can generate hypotheses and melanoma-targeting treatment strategies that directly improve the well-being of cancer patients. Daniel S. Widmer and Mitchell P. Levesque wrote the manuscript and created the illustration. Ossia M. Eichhoff contributed to creating the illustration and helped in writing the manuscript. Reinhard Dummer helped in writing the manuscript. The authors of this study do not have any conflict of interests to declare. Data S1. References. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Widmer et al. (2015) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: