Increasing use of hematopoietic stem cells for retroviral vector-mediated gene therapy and recent reports on insertional mutagenesis in mice and humans have created intense interest to characterize vector integrations on a genomic level. We studied retrovirally transduced human peripheral blood progenitor cells with bone marrow-repopulating ability in immune-deficient mice. By using a highly sensitive and specific ligation-mediated polymerase chain reaction (PCR) followed by sequencing of vector integration sites, we found a multitude of simultaneously active human stem cell clones 8 weeks after transplantation. Vector integrations occurred with significantly increased frequency into chromosomes 17 and 19 and into specific regions of chromosomes 6, 13, and 16, although most of the chromosomes were targeted. Preferred genomic target sites have previously only been reported for wild-type retroviruses. Our findings reveal for the first time that retroviral vector integration into human marrow-repopulating cells can be nonrandom (P ‫؍‬ .000 37). (Blood. 2003;

Reports on insertional "genotoxicity" in patients have created intense interest in characterizing retroviral vector integrations on the genomic level. The retroviral vector SF91m3 was used for transduction of human peripheral blood progenitor cells (PBPC). These PBPC were transplanted into nonobese diabetic/severe combined immunodeficient mice. A total of 186 retroviral vector integration sites were isolated by ligation-mediated PCR from chimeric mouse bone marrow of five PBPC donors, sequenced, and blasted against the human genome. Preferred integration near the transcription start regions, within CpG islands, and within Alu regions was observed. Detailed analysis of targeted RefSeq genes showed a favored integration within the first intron. Integrations were most common in genes coding for signaling proteins, transcription factors, and kinases. In all genes targeted independently multiple times the respective orientation of the provirus within the gene was identical, indicating integration hot spot regions and similar steric determinants for integration sites. Possible explanations for these findings could be nonrandom vector integration, clonal selection due to gene expression interference, or engraftment issues related to gene insertion in signaling and cell cycle genes. The low frequency of integrations in exons may be reassuring as to the safety of retroviral gene therapy with normal human PBPC.

These investigations indicate that integration of lentiviral vectors in human repopulating cells capable of engrafting NOD/SCID mice preferentially occur in coding regions of the human genome. Nevertheless, the clustering of integrations at the transcriptional start is not as high as that observed for gammaretroviral vectors.