Genomic, transcriptomic, and immunogenomic landscape of over 1300 sarcomas of diverse histology subtypes

Abstract

Given their rarity and diversity, a fundamental understanding of the genomic underpinnings for many sarcoma subtypes is still lacking. To better define the molecular landscape of this group of diseases, we perform matched whole exome sequencing and RNA sequencing on a cohort of 1340 sarcoma tumor specimens. We identify recurrent somatic mutations and observe an increased mutational burden in metastatic vs. primary samples (p < 0.001). We observe frequent copy number alterations including whole genome doubling, with this feature being more common in metastatic tumors (p = 0.026). Estimation of immune cell abundances followed by hierarchical clustering identifies five immune subtypes ranging from low to high and we observe inferior overall survival in immune deplete clusters compared to immune enriched (p < 0.01). Interestingly, GIST predominantly form a distinct "immune intermediate" cluster that is marked by a specific enrichment for NK cells (FDR < 0.01).

Fig. 1. COSMIC Tier 1 genes significantly mutated in sarcomas.

Top 10 most significantly mutated genes from robust regression in each histology subtype intersected with the COSMIC Tier 1 gene list (A) shows the high mutation frequency of TP53 across all histology subtypes and histology-specific mutations like KIT. Tumor mutation burden of all non-silent mutations in sarcomas compared to TCGA PanCancer Study (B) show similar tumor mutation burden in ORIEN compared to TCGA SARC.

Fig. 2. Copy number variation landscape of sarcomas.

Arm-level copy number alterations in primary (A) and metastatic (B) disease. Green bar indicates samples that had >50% of segments lengths classified as having GAIN or AMP copy number. Samples were clustered within their respective histology subtype, and the histology subtypes of were clustered for Primary disease.

Fig. 3. The gene expression landscape of sarcomas.

Sarcomas were grouped according to their gene expression similarities and displayed in an UMAP projection (A). Four histologies (Leiomyosarcomas, GISTs, Myxoid liposarcomas, and dedifferentiated/well-differentiated liposarcomas) showed cohesive clusters (B). Differential expression analysis indicated that each of the four histologies expressed a unique set of genes (C). Gene set enrichment analysis (GSEA) supported the high-level differences observed in differential expression analysis (D).

Fig. 4. Sarcomas samples classified according to the level of immune infiltration.

Cell type scores (MCPcounter) showed five distinct immune groups of sarcomas, each with a distinct transcriptional profile. Tumor mutation burden (TMB), estimated as log(mut/MB) was slightly higher in the immune group E (A). Each histology in the data set was composed by different proportions of the five immune groups (B). GSEA also highlighted the differences in gene expression among the immune groups (C). Survival analysis of patients with tumors in the different immune groups showed higher overall survival for those with immune group C-E tumors (D). Asterisks (*) in (B) indicate p-value < 0.05 from two-sided Fisher’s Exact tests for association between number of samples in each immune group and number primary/metastatic samples. An “x” next to the histology indicated that the Fisher’s test could not be completed.