Targeting retroviral integration?

C an retroviral integration be targeted to preselected locations in the human genome? If so, targeting might improve the safety of retroviral vectors for use in gene therapy. Directing integration of vectors to benign locations in the human genome, for example, might reduce the risk of transformation by insertional activation of oncogenes. This issue has become a lot less theoretical with the recent report from Li and coworkers describing transformation by a gene therapy vector [1]. Interpreting this experiment is not straightforward, however, in part because the marker gene transduced by the vector probably contributed to oncogenesis. However, this result does increase interest in possible means for directing retroviral integration to preselected locations in the human genome. The first efforts to target retroviral integration involved fusing retroviral integrase enzymes to sequencespecific DNA-binding domains [2–5]. The integrase enzyme carries out the initial DNA breaking and joining reactions that connect the reverse-transcribed retroviral DNA to the host chromosome [6]. Normally integration takes place with little target sequence specificity [6]. Tests with in vitro integration reactions using preintegration complexes isolated from virus-infected cells showed relatively slight target sequence preferences [7,8]. More recently some biases have been detected in integration in vivo, but these are of a more global nature and are not dependent on local primary sequence [9]. Fusions of integrase to DNA-binding domains were made by several groups and the modified integrase enzymes were tested for targeting in vitro. The first report described a fusion of HIV-1 integrase to the sequence-specific DNA binding domain of phage lambda repressor [2,3]. Repressor normally needs to form a dimer to bind DNA, and initial binding studies showed that the repressor–integrase fusion was indeed capable of proper dimerization and binding to lambda operators (repressor binding sites). In optimized reactions in vitro, integration was found to take place preferentially near lambda operators in target DNA. Favored sites were found to lie on the same face of the DNA helix as the operator sites, suggesting that the bound repressor–integrase fusion captured target DNA by looping out the intervening sequences. Later studies demonstrated that the LexA DNA-binding domain could also be used to direct integration targeting in vitro [4,5], and other integrases could be used as fusion partners [5].