Disrupting MLV integrase:BET protein interaction biases integration into quiescent chromatin and delays but does not eliminate tumor activation in a MYC/Runx2 mouse model.

Author summary Many different retroviruses, including murine leukemia virus (MLV), are used as vectors for human gene therapy and cancer immunotherapies (CAR-T) because of their stable and efficient delivery of genetic material into the host DNA. However, this process can result in activation and/or disruption of cellular genes, which has resulted in the outgrowth of tumors in previous clinical trials. Of critical importance is the preferred location within the host genome at which the retrovirus integrates. Our study presents a modified MLV virus that has lost the ability to bind to the host BET proteins, the drivers of insertion at promoter/enhancer regions of highly activated genes. This is the first direct study within an animal model to examine whether infection with this type of modified MLV virus affects tumor progression. We found that the modified virus improved the survival of the well-characterized MYC/Runx2 mouse compared to infections with the wild-type MLV. Most modified viral integrants mapped into less active regions of the genome, but some integration events still occurred near cancer-related genes. Thus, while the modified MLV virus can be a safer vector than the wildtype virus, it still maintains the potential to activate oncogenes. This study provides new insights on how to improve the safety of MLV retroviral vectors. Murine leukemia virus (MLV) integrase (IN) lacking the C-terminal tail peptide (TP) loses its interaction with the host bromodomain and extraterminal (BET) proteins and displays decreased integration at promoter/enhancers and transcriptional start sites/CpG islands. MLV lacking the IN TP via an altered open reading frame was used to infect tumorigenesis mouse model (MYC/Runx2) animals to observe integration patterns and phenotypic effects, but viral passage resulted in the restoration of the IN TP through small deletions. Mice subsequently infected with an MLV IN lacking the TP coding sequence (TP-) showed an improved median survival by 15 days compared to wild type (WT) MLV infection. Recombination with polytropic endogenous retrovirus (ERV), Pmv20, was identified in seven mice displaying both fast and slow tumorigenesis, highlighting the strong selection within the mouse to maintain the full-length IN protein. Mapping the genomic locations of MLV in tumors from an infected mouse with no observed recombination with ERVs, TP(-)16, showed fewer integrations at TSS and CpG islands, compared to integrations observed in WT tumors. However, this mouse succumbed to the tumor in relatively rapid fashion (34 days). Analysis of the top copy number integrants in the TP(-)16 tumor revealed their proximity to known MLV common insertion sites genes while maintaining the MLV IN TP- genotype. Furthermore, integration mapping in K562 cells revealed an insertion preference of MLV IN TP- within chromatin profile states associated with weakly transcribed heterochromatin with fewer integrations at histone marks associated with BET proteins (H3K4me1/2/3, and H3K27Ac). While MLV IN TP- showed a decreased overall rate of tumorigenesis compared to WT virus in the MYC/Runx2 model, MLV integration still occurred at regions associated with oncogenic driver genes independently from the influence of BET proteins, either stochastically or through trans-complementation by functional endogenous Gag-Pol protein.