Intentionally integrating in oncogenes

Gene therapy is currently plagued by the problem of vector integration near oncogenes. Although different retroviral (including lentiviral) vectors have distinct preferences for integrating near genes and/or their promoters, these typically vary only 2- to 3-fold in magnitude. Given the large numbers of transduced cells required for stem-cell gene therapy, one should assume that vectors can and will integrate near every potential oncogene at clinically relevant frequencies. The real question is whether a specific vector provirus will activate nearby proto-oncogenes and cause a malignancy.

This was dramatically demonstrated in the treatment of X-linked severe combined immunodeficiency (X-SCID), when multiple patients developed leukemias linked to gammaretroviral vector integration near the LMO2 gene. To overcome this type of genotoxicity, gene therapists have scrambled to develop safer vectors, and now need to prove that they are safer before beginning new clinical trials. In this issue of Blood, Ryu and colleagues tackle this problem with an assay that specifically measures LMO2 gene activation by vector constructs integrated at the LMO2 locus in a human lymphoid cell line. The article is a technical tour de force that combines gene targeting with adeno-associated virus (AAV) vectors, Cre-mediated transgene cassette exchanges, and self-complimentary AAV vectors expressing Cre recombinase.

The authors show that major increases in LMO2 transcript and protein levels occur when a gammaretroviral long terminal repeat (LTR)–driven GFP gene is present in the first intron of LMO2, and they address how this can be avoided by new vector designs. One improvement is to limit transgene expression to a particular cell type, as shown by the lack of LMO2 activation when an erythroid-specific transgene construct was used. Of course, a different result might have been obtained in erythroid cells, but this still limits gene activation to a subset of potentially transformable cell types. Another promising modification is the inclusion of insulator elements. Originally incorporated in vectors to limit chromosomal position effects on vector transgenes, the authors prove the inverse can also occur, since transgenes flanked with these ele-ments activated LMO2 at lower levels. Although the effect was not absolute, this demonstration of insulator function at a relevant human chromosomal locus goes a long way toward justifying their inclusion in the next generation of vectors. The assay is also ideally suited for testing different types of retroviral vector backbones. A lentiviral vector with insulator elements and an internal cellular EF1α promoter did not activate LMO2.

Because the assay is performed in one cell type at one locus, it is not the broadest way to screen for vector genotoxicity. Instead, the power of this approach lies in its specificity, since different vector designs can be tested at the desired proto-oncogene locus in an appropriate human cell type. In the case of X-SCID, we know that LMO2 gene activation is a consistent problem, and as we learn more about which proto-oncogenes are activated in other clinical situations, analogous assays can be developed for other diseases.