The biology of uveal melanoma – next challenges

Abstract

Uveal melanoma (UM), a rare cancer of the eye, has been deeply characterized for its molecular lesions in terms of chromosomal copy number alterations (CNAs), gene expression, somatic mutations and DNA methylation (for reviews see [1, 2]). It shows a very limited number of somatic mutations, very few of which are recurrent [3] (probable initiator mutations in GNAQ [4], GNA11 [5], CYSLTR2 [6] and PLCB4 [7], all acting in the same G-protein coupled receptor signaling pathway, mutations in BAP1 [8] and SF3B1 [9] that drive metastasis and mutations in EIF1AX [10] that apparently are involved in tumor formation but not progression). A few CNAs (monosomy of chromosome 3 [11] chr8q gain [12] and chr6p gain [13]), global gene expression profiles or an expression analysis of a number of genes that have been included in a prognostic signature [14] as well as whole genome DNA methylation similarly distinguish two to four classes of UM [15]. It is possible to predict the propension to develop metastases based on each of these molecular domains. Approaches to fuse these data in order to develop a combined molecular predictor have not significantly improved prognostic assessment [16]. Our present knowledge on the mutational landscape of UM indicates that a single mutation in one of the four known “initiator” genes (GNAQ, GNA11, CYSLTR2, and PLCB4) is enough to form a tumor and a single further mutation in either BAP1 [8] or SF3B1 [9] is enough to drive metastasis. These mutations are almost perfectly segregated from the classes defined by gene expression profiling or by CNA. All these approaches yield two clearly distinct classes with each two subclasses with different metastatic potential. This clear distinction can be taken for evidence of non-continuous risk distribution, yet a recent single cell transcriptomics-based analysis hints at a mixture of class-1 low risk and class-2 high risk cells within a single tumor whereby the proportion of these two cell types finally determines the real risk of metastasis [17]. It is not clear how this cell admixture model can explain the clearly distinct risk-associated molecular classes and further research is needed to clear that point. The few driver mutations, even if assisted by secondary drivers [18], are best compatible with a linear tumor evolution model, but recent evidence introduced the punctuated equilibrium model (or the big bang model) to UM [19]. This model postulates a phase of high genomic instability followed by the outgrowth of stabilized clones into a heterogenous tumor [20, 21]. Tumor heterogeneity has not systematically been addressed for UM. Given the paucity of mutations, heterogeneous subpopulations are unlikely to be traceable by exome sequencing but CNA analyses might help. A recent large-scale analysis of CNA revealed much more cytogenetic events with a discrete frequency than heretofore believed [22]. Still we do not know the deletions of which genes on chromosome 3 except for BAP1 are important for UM metastasis. Early work trying to define the minimal critical interval could not single out specific genes [23]. Chr3 monosomy can come about in a single step by losing one copy during mitosis due to non-disjunction although cases with partial deletion of one copy of chr3 have been reported [24]. Alternatively, several genes including noncoding genes on chr3 can cooperate in determining the metastatic risk.