Cell-Specific Transcriptional Regulatory Domains Attract Gamma Retroviral Integration in the Human Genome

Gene transfer vectors derived from γ-retroviruses target at high frequency genes involved in the control of growth and differentiation of the target cell, and may induce insertional tumors or pre-neoplastic clonal expansions in patients treated by gene therapy. The gene expression program of the host cell appears instrumental in directing γ-retroviral integration, but the molecular basis of this phenomenon is poorly understood. We recently reported a bioinformatics analysis of the distribution of transcription factor binding sites (TFBSs) flanking ∼4,000 proviruses in human hematopoietic and non-hematopoietic cells. We showed that γ-retroviral, but not lentiviral, vectors integrate in genomic regions enriched in cell-type specific subsets of TFBSs. Analysis of sequences flanking the integration sites of Moloney leukemia virus (MLV)-derived vectors with modified long terminal repeats (LTRs), and of a human immunodeficiency virus (HIV)-based vector packaged with the MLV integrase, showed that the MLV integrase and LTR enhancer are the viral determinants of the selection of TFBS-rich regions. The study indicated that γ-retroviruses have evolved a peculiar strategy to interact with the host cell chromatin, based on a cooperation between the integrase and TFs bound to the LTR enhancers before integration to tether pre-integration complexes (PICs) to transcriptionally active regulatory regions. To further explore the link between cell-specific transcription and MLV integration, we performed deep sequencing of MLV insertions in human CD34+ hematopoietic cells (n=33,162), T lymphocytes (n=7,996), and skin keratinocytes (n=3,020). Compared to matched random distributions, MLV integrations are highly clustered along the chromosomes, with peaks of frequency around transcriptionally active loci. Genes within or near the insertion clusters are shared in part among the three cell types, but are mostly involved in tissue-specific development, differentiation and functions, and differentially expressed, as evaluated by gene expression arrays. Integration peaks overlap with non-coding elements highly conserved among mammals, again meaning that genomic regions presumably involved in transcriptional regulation are attractive for MLV PICs. The status of chromatin around retroviral and random sites was checked by ChIP-on-chip technology, revealing marks of active transcription (H3K9ac, H3K4me2, H3K4me3) on regions flanking the sole MLV insertions. Taken together, our observations state an overt preference of γ-retroviral vectors for genomic regions actively engaged in transcriptional regulation. We speculate that MLV PICs are directed to these regions by interaction with general components of the enhancer-binding complexes, rather than with specific TFs or TF families.