Strong association of endogenous retroviral element expression and genetically-defined acute myeloid leukemia subtypes
Endogenous retroviral elements (ERVs) are remnants of retroviruses that have become permanently integrated into host genomes. These repetitive genomic elements collectively comprise approximately 8% of the human genome. ERVs are normally silenced, and their activation has been linked to multiple types of cancer, including acute myeloid leukemia (AML). Additionally, treatment with hypomethylating agents decitabine or azacitidine has been shown to activate ERV expression, which induces cell death. Increasing evidence suggests that ERVs can act as gene promotors and enhancers, bind hematopoietic transcription factors such as GATA1/2 and TAL1, and modulate immune responses. However, their overall contribution to acute leukemia remains poorly understood. ERV detection is complicated due to their repetitive nature but several bioinformatic tools have been developed to address that challenge.
Here, we used the bioinformatic algorithm Telescope to quantify ERV expression from RNA sequencing data in more than 800 AML and healthy control samples. Differences in ERV expression between AML and healthy samples were analyzed, as well as between different subtypes of AML. Finally, the subtype-specific overexpression of one ERV was validated in primary patient material using qPCR.
In the 831 samples studied here, a total of 3459 ERVs were expressed, with large heterogeneity between samples. Dimension reduction revealed clear grouping of samples according to ERV expression, indicating that programs of ERVs are expressed in specific subsets of samples. Inspection of this grouping revealed a clear separation between healthy and AML samples. Additionally, AML samples separated by genetic subtype. NPM1‐mutated cases formed at least three distinct groups. Samples harboring RUNX1::RUNX1T1, CBFB::MYH11, PML::RARA fusion genes or CEBPA mutations each formed distinct groups that contained few other AML samples. Comparative analysis of RUNX1::RUNX1T1, CBFB::MYH11, and PML::RARA samples identified over 900 differentially expressed ERVs. In a qPCR validation experiment using independent primary AML samples, the most significantly upregulated ERV in RUNX1::RUNX1T1 (HERVH_6q12f) was confirmed to be more highly expressed in RUNX1::RUNX1T1 samples compared to samples with other translocations.
In conclusion, we showed that AML and healthy samples exhibit distinct ERV expression programs, and found sets of ERVs that are expressed in specific AML subtypes. Thus, ERVs are revealed as a separate class of RNA molecules that further characterize the complex AML transcriptome.