In-depth analysis of endogenous retrovirus expression in glioblastoma.

Abstract

Background: Human endogenous retroviruses (HERVs) are remnants of ancient viral infections and comprise 6-8% of the human genome. Their biological functions in cancer remain poorly understood, especially in glioblastoma, the most common and deadly primary brain cancer in adults. Prior studies on HERV expression in glioblastoma have yielded conflicting results. Here, we employed orthogonal computational pipelines to address these limitations.

Results: Locus-specific analysis revealed marked heterogeneity of transcription among HERVs within the same clade. We found that individual HERV proviruses are more than twice as likely to be underexpressed in glioblastoma than overexpressed and that most differentially expressed HERVs are exonized within transcripts. These HERVs are enriched in the 3'-terminal exons of transcripts, associated with alternative polyadenylation and contribute conserved polyadenylation signals. We identified HERV expression patterns associated with glioblastoma subtypes and anatomic features. We also identified three proviruses or proviral fragments of particular interest including one associated with survival and one exonized within a currently unannotated cancer-specific transcript. Among the most recently integrated clade of HERVs, excluding solo-LTRs, only the HML-2 provirus at 1q22 is overexpressed in glioblastoma. We report previously undescribed transcripts incorporating this provirus that may encode several proteins.

Conclusions: This study represents the first systematic exploration of the heterogeneity of HERV expression between anatomic features of any cancer. It shows that exaptation of HERV polyadenylation signals and HERV-associated APA are widespread in the human transcriptome and identifies critical structural information regarding the HML-2 1q22 provirus transcript, which has been the focus of several recent analyses. Our findings underscore the importance of locus-specific and anatomic feature-specific HERV analysis and suggest structural and functional roles for HERVs in glioblastoma-associated transcripts.