More Than One Mutation Is Required for Leukemia Development Following Retroviral Vector-Mediated Gene Transfer.

As recent case reports have suggested that proto-oncogene activation by replication-deficient retroviral vectors may trigger leukemia in mice and men, there is an urgent need for models allowing predictive investigations into the pathomechanisms of this complication. Previously, we have shown that dose-escalated replication-defective retrovirus vectors (such as HaMDR1 or SF91dsRedpre) may trigger combinatorial leukemia-initiating events in single cells, with a probability that allows prospective studies in simple mouse models (C57Bl/6). Meanwhile, we have performed molecular analyses for 5 different leukemias, harboring 6 to 12 vector insertions according to Southern blot. 34 insertions recovered from these clones by LM-PCR and LAM-PCR showed that 4 leukemias had combinatorial hits in at least one established proto-oncogene (class 1 genes: Hivep1, Mllt3, Fli1, HoxA7, Brd2, Evi1, Csfr3, Bcl11a, Ski) with at least one hit in other signaling genes (class 2 genes). One leukemia for which 2 insertions remain to be identified showed at least 3 class 2 hits, so far without a class 1 event. All leukemias presented with hits in genes involved in cytoplasmic signal cascades along with genes contributing to nuclear regulatory networks. Some of these leukemias had a normal karyotype. A comparison with 62 insertion sites from healthy control animals (determined in peripheral blood DNA harvested 7 months after transplantation) revealed a significant enrichment of class 1 hits in leukemias (14-fold over expected and 3-fold over control). Healthy animals still showed 6 hits (~10%) in class 1 genes, obviously not being sufficient for leukemia induction. Three of these occurred upstream of Evi1, one upstream of Nfic, one upstream of Pdclg2, and one in Bcl11a. Interestingly, upstream vector insertion was preferred in class 1 gene hits. Overall, insertions recovered from both normal and leukemic samples revealed an overrepresentation of hits within (or upstream of) known or predicted genes (89 and 90%, respectively). In line with the data of a related murine marking study (Kustikova, Fehse et al., ASH 2004), our study strongly suggests that single hits in class 1 or class 2 genes are not sufficient to induce leukemia in mice. However, certain combinations of such events in single cells may well be leukemogenic with a given latency. Our findings establish a powerful approach for both oncogene discovery and safety validation in gene therapy.