An expression signature of 19 human endogenous retroviruses identifies immunogenic luminal breast cancers likely to respond to immunotherapy.

Background: Human endogenous retroviruses (HERVs) are single-stranded RNA viruses that have integrated into the human germline, and whose sequences represent approximately 8% of the human genome. In cancer cells, certain HERV sequences can be re-expressed, thereby promoting an innate and adaptive immune response, particularly through the induction of an interferon response in tumors, known as "viral mimicry". We sought to elucidate the links between HERV expression and tumor immunogenicity in different biological subtypes of early breast cancer (eBC). Patients and methods: We used publicly available RNA-seq gene expression data of breast cancer samples and corresponding matched normal data from The Cancer Genome Atlas (TCGA). HERV sequences were detected and quantified using Telescope software. For each patient, we computed an overall HERV expression and a signature of HERVs whose expression is associated with biological signals of intra-tumor viral mimicry. Correlations between these expression signatures, signs of tumor immunogenicity, and patient prognosis were examined. Results: Overall HERV tumor load varies little among the different entities of eBC and does not appear to impact tumor immunogenicity, regardless of the biological subtype. In contrast, restricting the analyses to HERVs whose expression is associated with viral mimicry hallmarks, it was possible to identify a signature of 19 HERVs whose expression is associated with signs of tumor immunogenicity. Moreover, this "Breast HERV19" expression signature was associated with better prognosis in HR Conclusion: In eBC, our "Breast HERV19" expression signature makes it possible to isolate a subgroup of HR