Defining the lentiviral integrome in human hematopoietic cells

Retroviral integration is a non-random process, whereby pre-integration complexes of different viruses recognize components or features of the host cell chromatin in a specific fashion. By using deep sequencing technology, we mapped >60,000 MLV and HIV integration sites in the genome of human CD34+ hematopoietic stem/progenitor cells and >16,000 sites in peripheral blood T-lymphocytes, and defined genome-wide integration maps in both cell types. MLV integrations cluster around regulatory elements (promoters, enhancers, and evolutionarily conserved non-coding regions) of genes involved in hematopoietic functions, and to chromatin regions bearing epigenetic marks of active or poised transcription. On the contrary, HIV integrations are clustered in regions marked by histone modifications associated to the body of transcribed genes (H3K36me3 and H2BK5me1), and are under-represented in regulatory regions. Although >90% of the genes targeted by HIV integration are expressed by Affymetrix analysis, expressed genes are not equally targeted. By a rigorous statistical analysis, we define a set of <400 genes that are targeted by HIV at significantly higher frequency than matched random controls after normalization for gene length, and a smaller set of genes that are targeted at significantly lower frequency. Functional clustering analysis shows that highly targeted genes are involved in chromatin remodeling and transcription, and are enriched in housekeeping functions. This analysis identifies a set of “high-risk” genes in hematopoietic cells, the function of which is more likely to be influenced by lentiviral vector integration in clinical gene therapy. Many of these genes are over-represented in collections of lentiviral vector integrations from patients treated by gene therapy, indicating that lentiviral “common integration sites” are determined by the HIV target site selection rather than clonal dominance in vivo.