The Integrated Genomic Landscape of Thymic Epithelial Tumors

Abstract

Thymic epithelial tumors (TETs) are one of the rarest adult malignancies. Among TETs, thymoma is the most predominant, characterized by a unique association with autoimmune diseases, followed by thymic carcinoma, which is less common but more clinically aggressive. Using multi-platform omics analyses on 117 TETs, we define four subtypes of these tumors defined by genomic hallmarks and an association with survival and World Health Organization histological subtype. We further demonstrate a marked prevalence of a thymoma-specific mutated oncogene, GTF2I, and explore its biological effects on multi-platform analysis. We further observe enrichment of mutations in HRAS, NRAS, and TP53. Last, we identify a molecular link between thymoma and the autoimmune disease myasthenia gravis, characterized by tumoral overexpression of muscle autoantigens, and increased aneuploidy.

Keywords: TCGA; autoimmunity; genomics; myasthenia gravis; proteomics; thymic carcinoma; thymic epithelial tumors; thymoma; transcriptomics.

Copyright © 2018 Elsevier Inc. All rights reserved.

Figures

Figure 1

The landscape of DNA mutation in TETs. The matrix in the center of the figure represents individual mutations in TET patients, color-coded by type of mutation, for the significantly mutated genes which include: GTF2I, HRAS,NRAS, and TP53. The rate of synonymous and non-synonymous mutations for each sample is displayed at the top of the matrix. The barplot on the left of the center matrix shows the number of mutations in each gene. The barplot to the right of the matrix displays the q-values for the most significantly mutated genes. The bottom half of the figure depicts the arm level sCNAs for each sample color-coded by the type of CNA. The barplot on the left depicts the number of total sCNAs for each sample. See also Figure S2 and S3.

Figure 2

Tumor mutation burden (TMB) in TETs (THYM) compared to 21 other cancers profiled by the TCGA. The proportional presence of mutational transitions and transversions by cancer lineage are depicted in the histograms at the bottom of the figure.

Figure 3

Integrative unsupervised clustering of subtypes from five data platforms. (A) Consensus clustering separated TET samples into four molecular subtypes (n=117). The blue and white heatmap at the top shows sample consensus. The blue and yellow heatmap in the center shows the correlation to each individual data type cluster membership centroid. The bottom displays the presence (black) or absence (gray) of a mutation in one of the four significantly mutated genes. (B) Summary of samples in each cluster by WHO histological subtype. (C) Survival differences across molecular subtypes. Survival data was missing for a sample in cluster 3. (D) Map of samples generated from TumorMap colored by pathology status. Samples are placed according to similarities in their genomic profiles integrating all the platforms. For each cluster, single-platform hallmarks are listed above the thin line, whereas PARADIGM results are listed below the thin line. See also Figures S3, S4, and S5.

Figure 4

Multi-platform analysis of the thymoma specific oncogene GTF2I. (A) Lollipop plot ofGTF2I demonstrating all the mutations observed in GTF2I. Green boxes mark the GTF2I-like repeat regions. (B) The frequency of somatic mutations in GTF2I in other cancer lineages as compared to TETs. (C)GTF2I mutational gene expression signature in thymoma type A and AB. (D)GTF2I methylation signature in type A and AB. Ten misclassified samples are also reported, based on the RNAseq data. (E) Box plots of pathway scores of GTF2I mutant (red) and wild type (blue) tumors. Box plot display the median value, upper and lower quartiles, the whiskers represent the interquartile range, and outliers are marked with dots. P values are based on the ANOVA test.

Figure 5

Patterns of sCNA and gene expression associated with autoimmunity. (A-B) The prevalence of altered chromosome arms were compared between the positive (MG+) and negative (MG−) status of myasthenia gravis for all samples across histology types (A) and only the subset of samples in B1, B2, B3 histology types (B). For the boxplots: line in the box indicates the median; lower and upper hinges correspond to the first and third quartiles; upper and lower whiskers extend to 1.5 × interquartile range; outlier data are shown as points. (C) Gene-level sCNA frequency landscape for samples with B1, B2, B3 histology, comparing between the history of myasthenia gravis status. The χ2 test of independence was applied to arm-level sCNA for each chromosome arm to determine significantly enriched events between MG+ and MG- status. False discovery rate (q) less than 0.05 (−log q value = 3) are shown for gains (red) and deletions (blue). (D) Log2 normalized gene expression of selected differentially expressed genes. See also Figure S6.

Figure 6

Genomic analysis of thymic carcinomas. (A) The landscape of DNA mutation in Type TC tumors. The matrix at the top of the panel depicts clinical information. The center of the panel depicts individual mutations in Type TC tumors, color-coded by type of mutation, for the previously identified significantly mutated genes and focal copy number changes. The bottom half of the panel depicts the arm level sCNAs for each sample color-coded by the type of CNA. The barplot on the left depicts the number of total sCNAs for each sample. (B) A box plot demonstrating the TMB of samples by histology. Lines in the boxes indicate the median; lower and upper hinges correspond to the first and third quartiles; upper and lower whiskers extend to 1.5 × interquartile range; outlier data are shown as points. In order to avoid skewing the results, one hypermutated TC sample and one TC sample with a TMB=0 were excluded. (C) The normalized profile of COSMIC signature 6 (microsatellite unstable tumors) and the SNV mutation spectra of the hyper-mutated TC sample (TCGA-ZB-A966) along 96 base substitution types in tri-nucleotide sequence motifs (top) and the heatmap of cosine similarity between the mutation spectra in TCGA-ZB-A966 and thirty curated COSMIC signatures.