Identification of diagnostic and prognostic genetic alterations in uveal melanoma using RNA sequencing

Abstract

Uveal melanoma is a lethal intraocular tumour, in which the presence of various genetic alterations correlates with the risk of metastatic dissemination and survival. Here, we tested the detectability of all key mutations and chromosomal changes from RNA sequencing data in 80 primary uveal melanomas studied by The Cancer Genome Atlas (TCGA) initiative, and in five prospective cases. Whereas unsupervised gene expression profiling strongly indicated the presence of chromosome 3 alterations, it was not reliable in identifying other alterations. Though, the presence of both chromosome 3 and 8q copy number alterations could be successfully inferred from expressed allelic imbalances of heterozygous common single nucleotide polymorphisms. Most mutations were adequately recognised in the RNA by their nucleotide changes (all genes), alternative splicing around the mutation (BAP1) and transcriptome-wide aberrant splicing (SF3B1). Notably, in the TCGA cohort we detected previously unreported mutations in BAP1 (n = 3) and EIF1AX (n = 5), that were missed by the original DNA sequencing. In our prospective cohort, all genetic alterations were successfully identified by combining the described approaches. In conclusion, a transcriptional analysis presents insights into the expressed tumour genotype and its phenotypic consequences and may augment or even substitute DNA-based approaches, with potential applicability in research and clinical practice.

Keywords: Genetic profiling; RNA sequencing; Uveal melanoma.

Fig. 1

Cohort-wide GEP to estimate the presence of genetic alterations in our training cohort of uveal melanomas studied by TCGA (n = 80). (A) Two-dimensional UMAP analysis to identify two clusters of tumours: GEP class I and II. (B) Chromosomal location of signature genes characterising the two clusters. (C) Relative gene expression levels of signature genes characterising the two clusters in relation to the genetic alterations present per individual tumour.

Fig. 2

RNA-inferred allelic imbalances in relation to DNA-confirmed copy number alterations. The figures for all TCGA tumours can be found in Supplementary Data 1. (A) Example of a tumour showing balanced expression of chromosome 3, but imbalanced expression of chromosome 8q, in line with a disomy 3 and copy number alteration of chromosome 8q. Additionally, alterations affecting chromosome 6p, 6q and 17q (partial) are correctly identified. (B) Example of a tumour showing imbalanced expression of both chromosome 3 and 8q, in line with copy number alterations of the two chromosomes. Additionally, an alteration affecting chromosome 6p (partial) is correctly identified. (C) Overview of our RNA-derived observations in comparison with the DNA-determined copy number values in the 80 TCGA tumours. Balanced expression correctly identified all tumours with disomy 3 (n = 40/40), and imbalanced expression nearly always meant the tumour had lost (part) of chromosome 3 (n = 39/40). Concerning chromosome 8q, most tumours with balanced expression had a disomy 8q (n = 21/27), and all tumours with imbalanced expression had a gain or amplification of this chromosome (n = 53/53). In aggregate, the presence or absence of a copy number loss of chromosome 3 or increase of 8q was correctly identified from RNA sequencing data in 79/80 (99%) and 74/80 (93%) of the uveal melanomas.

Fig. 3

Detectability of mutations via RNA sequencing data in our training cohort of uveal melanomas studied by TCGA (n = 80). All details are described in Supplementary Table 2. (A) Summary of detected mutations in comparison to findings in exome-captured DNA sequencing. At the RNA level, Gα_q mutations were correctly identified in 74/78 (95%), BAP1 mutations in 38/40 (95%), SF3B1 mutations in 15/15 (100%) and EIF1AX mutations in 14/15 (93%) mutant uveal melanomas. Note that various mutations were missed by the initial DNA-based analyses and detected by RNA sequencing only. (B) Examples of detected Gα_q signalling mutations in RNA sequencing data. (C) Examples of detected BAP1 mutations in RNA sequencing data: although some were identified directly via nucleotide variation, others showed alternative splicing around the mutant position. (D) Example of detected SF3B1 mutation in RNA sequencing data. (E) SF3B1-mutant tumours are known to result in a recurrent profile of aberrant splice junction usage ^,. This signature of alternative splicing was effectively used to identify SF3B1-mutant tumours in the TCGA cohort. (F) Examples of detected EIF1AX mutations in RNA sequencing data.

Fig. 4

Analysis of our prospective cohort of uveal melanomas (n = 5). (A) Two-dimensional UMAP analysis of cohort-wide GEP and application of TCGA-derived GEP-classifier in relation to the RNA- and DNA-based detectability of genetic alterations. (B) Representative examples of RNA-inferred allelic (im)balances to identify chromosomal alterations. In UM-4 and UM-5, imbalanced expression of genes of chromosome 3 correctly indicates the presence of a copy number alteration affecting this chromosome. In contrast, in UM-4, no imbalance is observed on chromosome 8q, in line with a disomy 8q, but an alteration and consequent imbalance is present in UM-5. The figures for all tumours from our cohort can be found in Supplementary Data 2. (C) BAP1 alterations (mutations and mutation-associated alternative splicing) as observed in the RNA sequencing data. (D) Analysis of the previously described signature of alternative splicing to identify SF3B1-mutant tumours.