Predicting transposable element-derived neoantigens in ATRX-mutated sarcoma.

Abstract

11529 Background: Epigenetic dysregulation is a key mechanism in cancer biology, driving both immune evasion and the generation of potentially druggable neoantigen targets. Transposable elements (TEs) are proviral sequences mostly found in epigenetically repressed genomic regions. In cancer, TEs are one potential source of neoantigens. The tumor suppressor gene ATRX is an epigenetic regulator that is lost in up to 30% of soft tissue sarcomas. Loss of function of this heterochromatin modulator leads to widespread alterations to transcription patterns. Our prior work in sarcoma, along with studies in other tumors, showed that ATRX loss leads to the de-repression of TEs. We hypothesize that this mechanism leads to TE neoantigen presentation on MHC class I (MHC-I) molecules. Methods: Leveraging the gEVE database of proviral sequences with open reading frames, we examined the differential expression of TEs using RNA sequencing of multiple models comparing ATRX lost and ATRX wild-type tumors. Analysis included our previously published data from human undifferentiated pleomorphic sarcoma cell lines (2 clones, 3 replicates per group; GEO: GSE240030) and a publicly available dataset from a mouse model of soft tissue sarcoma (n = 3 per group; GEO: GSE167537). Additionally, we analyzed analogous data from a mouse model of glioblastoma multiforme (GBM), another malignancy with frequent ATRX loss (n = 3 per group; GEO: GSE178113). RNA-seq data was analyzed using Salmon for transcript-to-genome pseudo-alignment, DESeq2 for differential TE expression analysis, and arcasHLA for sample-level MHC-I subtype identification. Differentially overexpressed TEs in ATRX mutant tumors were analyzed for predicted neoantigens using the NetMHCpan4.1 package. Results: Across multiple models, ATRX loss of function alterations were associated with both activation and repression of TEs. Within two sarcoma contexts, upregulated TEs were found to have protein-coding sequences with the potential to lead to MHC-I neoantigen presentation. Within a human undifferentiated pleomorphic sarcoma cell line with ATRX loss, there were 7 upregulated TEs shared between the clones, each predicted to contain 12 – 50 MHC-I binding epitopes. Within a mouse soft tissue sarcoma model, 7 TEs were upregulated with ATRX loss, each with 5 – 28 MHC-I binding epitopes. Since ATRX loss alters TE expression in other malignancies, parallel data from a mouse GBM model was additionally analyzed. In this model, 308 TEs were found to be upregulated with ATRX loss, and all were predicted to generate neoantigens. ERVs and LINE1 elements were recurrently identified, making up 97 – 100% of the activated TEs. Conclusions: Transposable element transcription altered by ATRX loss in sarcoma and GBM creates a predicted source of MHC-I-binding neoantigens. These neoantigens come from TEs of the ERV and LINE1 families.