Genomic and Transcriptomic Analysis Reveals Incremental Disruption of Key Signaling Pathways during Melanoma Evolution

Abstract

We elucidated genomic and transcriptomic changes that accompany the evolution of melanoma from pre-malignant lesions by sequencing DNA and RNA from primary melanomas and their adjacent precursors, as well as matched primary tumors and regional metastases. In total, we analyzed 230 histopathologically distinct areas of melanocytic neoplasia from 82 patients. Somatic alterations sequentially induced mitogen-activated protein kinase (MAPK) pathway activation, upregulation of telomerase, modulation of the chromatin landscape, G1/S checkpoint override, ramp-up of MAPK signaling, disruption of the p53 pathway, and activation of the PI3K pathway; no mutations were specifically associated with metastatic progression, as these pathways were perturbed during the evolution of primary melanomas. UV radiation-induced point mutations steadily increased until melanoma invasion, at which point copy-number alterations also became prevalent.

Keywords: DNA-seq; RNA-seq; dysplastic nevus; genomic; melanoma; metastasis; nevus; transcriptomic; tumor evolution; tumor progression.

Conflict of interest statement

DECLARATION OF INTERESTS

B.C.B. is a consultant for Lilly Inc.

Copyright © 2018 Elsevier Inc. All rights reserved.

Figures

Figure 1.. Genetic and Transcriptomic Data Implicate Activation of MAPK Signaling at Initiation and Subsequent Amplification of Signaling during Melanoma Progression

For a Figure360 author presentation of Figure 1, see http//dx.doi:10.1016/j.ccell.2018.06.005#mmc7. (A) The fraction of mutations (y axis) predicted to activate the MAPK signaling pathway at each stage of melanoma progression. The green bars denote multiple mutations in the same sample with the specific combinations iterated to the right. Strong and weak activators of MAPK signaling are separately annotated (see the STAR Methods for details on classifying strong and weak mutations). (B) The mutant allele fraction (MAF) of oncogenic MAPK mutations from RNA-seq data is plotted as a function of tumor purity (cellularity). The regression line indicates the expected relationship under a model in which the transcript level from the mutant allele is proportional to tumor purity. (C) Proportions of oncogenic transcript after accounting for tumor cell content (melanoma versus nevus, p = 10 3, t test). The specific driver mutations and their allelic status (loss-of-heterozygosity [LOH] or not) are also annotated for each neoplasm. (D) MAPK signaling output was inferred from an established MAPK gene expression signature (Joseph et al., 2010). Red and blue bars, respectively, denote a relatively more/less intense signature. The number of mutations in the MAPK pathway are indicated for each sample.See also Figure S2 and Table S3.

Figure 2.. Genetic and Transcriptomic Data Implicate Upregulation of Telomerase Early during Melanoma Progression

(A) The fraction of genetic alterations (y axis) affecting TERT at each phase of melanoma progression. (B) TERT expression was inferred from RNA-seq data and plotted from highest to lowest (left to right) with stage and mutation status designated. p values were calculated by comparing TERT expression between groups with two-tailed t tests: melanoma versus nevus, p = 6.7 × 10 3; mutant versus wild-type, p = 10 3. See also Figure S3.

Figure 3.. Genetic and Transcriptomic Data Indicate a Shift toward a PRC2-Modulated Chromatin Landscape at the Transition to Melanoma

(A) The fraction of pathogenic mutations in components of the SWI/SNF and PRC2 chromatin remodeling complexes (y axis) at each phase of melanoma progression. To avoid obscuring our analysis with passenger mutations, we only considered bona fide pathogenic alterations (see the STAR Methods). (B) Unsupervised clustering of samples (columns) and genes (rows) from RNA-seq data. The progression phase of each area and relative expression level of each gene are indicated. Two gene expression clusters are highlighted here (black bars). Gene sets significantly overlapping with the highlighted gene clusters are annotated alongside their q values (see the STAR Methods). (C) A model summarizing the balance between SWI/SNF and PRC2 during melanoma evolution. See also Figures S4 and S5.

Figure 4.. Genetic and Transcriptomic Data Implicates Impairment of the G1/S Checkpoint at the Transition to Invasive Melanoma

(A) The fraction of genetic alterations affecting genes involved in cell-cycle regulation (y axis) at each phase of melanoma progression. The green bars denote multiple mutations in the same sample with the specific combinations iterated to the right. (B and C) p16INK4A (B) and p14ARF (C) expression levels were inferred from junctional read counts specific to each transcript and are rank ordered from highest to lowest (left to right). The stage of each neoplasm is indicated (x axis) along with the mutation status of p16INK4A or p14ARF. The upper range of stromal expression (dotted line) was inferred from expression in tumors with loss of both alleles, a scenario in which all wild-type expression must derive from stromal cells. The asterisk (*) denotes samples with a point mutation in the CDKN2A, and proportion of mutant transcript is indicated by the striped bars. See also Figure S6.

Figure 5.. p53 and PI3K Pathway Mutations Appear Comparatively Later during the Evolution of Melanoma

(A) The fraction of genetic alterations affecting genes involved in the p53 pathway (y axis) at each phase of melanoma progression. (B) Phylogenetic tree for a select TP53-mutant case—the detailed evolution of this case is shown in the affiliated Mendeley Dataset. Pathogenic mutations are annotated with the TP53 mutation highlighted in bold. (C) The fraction of genetic alterations affecting genes involved in the PI3K pathway (y axis) at each phase of melanoma progression. (D) Phylogenetic trees for select PTEN-mutant cases—the detailed evolutions of these cases are shown in the affiliated Mendeley Dataset. Pathogenic mutations are annotated with the PTEN mutations highlighted in bold. See also Figure S7.

Figure 6.. Distinct Mutational Signatures Are Apparent at Specific Evolutionary Time Points

(A–C) The point mutation (A) and copy-number (B) burden at each phase of melanoma progression. Line, median; box, interquartile range (25%–75%); whiskers, 2 SDs above and below the median of the data; circles, outlier data points (C). The copy-number landscape at each phase of progression. Copy-number alterations reaching and remaining above a frequency of 10% are highlighted: red, gain; blue, loss. (D) Canonical phylogenetic trees corresponding to the four main progression trajectories constructed from the median trunk and branch lengths of all the individual cases (upper). The fraction of UV radiation-induced mutations within the trunks and branches of each progression trajectory (lower): D, descendant;P, precursor, mean ± 95% confidence intervals are shown.

Figure 7.. Somatic Alterations in Key Signaling Pathways that Drive Melanoma Appear at Specific Points in the Melanoma Progression Cascade

Each heatmap reflects the frequency that a given pathway is activated (red) or inactivated (blue) at a specific point in the melanoma progression cascade.