Foamy Virus Backbone Has Insulator Properties Which Remarkably Reduce Its Genotoxicity Potential

Michael Goodman, Paritha Arumugam, Devin Pillis, Danielle Lynn, Johannes C.M. Van der Loo, Mehdi Keddache, Thomas R. Bauer Jr., Dennis D. Hickstein, David W Russell and Punam Malik


Strong viral enhancers in γ-retrovirus vectors (GV) have caused cellular proto-oncogene activation and leukemia in gene therapy trials, necessitating the use of cellular promoters in enhancer-less integrating vectors. However, data is now emerging that inadequate transgene expression from cellular promoters may limit successful correction. Vectors derived from foamy virus (FV), a nonpathogenic retrovirus, have a higher preference for non-genic integrations than GV/lentiviral vectors (LV), and may be less genotoxic. We constructed GV, LV and FV driven either by the spleen focus forming virus (SFFV) or the murine stem cell virus (MSCV) enhancer/promoters, all driving eGFP expression, and compared their relative genotoxicity using an in vitro immortalization assay on primary hematopoietic stem/progenitor cells (HSPC). In this assay, integration near a protooncogene/gene promoting cell proliferation results in quantifiable HSPC immortalization. Strong viral enhancer/promoters from SFFV or MSCV in FV caused a remarkably low immortalization of HSPC compared to analogous LV or GV: compared to the immortalization frequency of HSPC with the SFFV-GV in this assay, SFFV-LV and MSCV-LV had 12- and 14-fold lower immortalization frequency, while the SFFV-FV and MSCV-FV showed a 155- and 414-fold lower immortalization frequency, respectively. Immortalized clones had multiple (3-10) integrated copies. Integration site analysis of FV immortalized clones revealed a majority of integrants in non-gene regions; those in genic regions targeted cell proliferation or proto-oncogenes, as previously reported. FV has been previously reported to have 2-fold higher insertions in non-genic regions and higher, but nearly half the propensity to target promoters compared to GV. However, this remarkably reduced genotoxicity with FV was not explained by the integration pattern. We therefore hypothesized that FV backbone may contain sequences that have an enhancer blocking/insulator effect. Studies on chromatin insulators have shown that enhancer-blocking property of insulators is mediated via binding of CTCF to its consensus sequences within the insulator. Indeed, an in silico analysis for CCCTC-binding factor (CTCF) binding sites in the vector backbone sequences showed more predicted CTCF binding sites in the FV than in GV or LV (26, 8, and 6, respectively). To functionally validate the enhancer-blocking effect of the FV backbone and ensure that only effects of the vector backbone would be measured, without the confounding influence of integration site or the enhancer/promoter/transgene, we inserted SFFV-GV, SFFV-LV and SFFV-FV into a clinically relevant proto-oncogene, LMO2, using CRISPR/Cas9, and assessed LMO2 expression. LMO2 upregulation has previously resulted in leukemias in the X-linked severe combined immune-deficiency and Wiscott-Aldrich syndrome (WAS) GV-mediated gene transfer trials; notably SFFV-GV was used in the WAS trial and caused leukemias in 8 of 10 patients from insertional oncogenesis. HeLa cells were transfected with the proviral donor plasmids and the guide-RNA/spCas9 plasmids and GFP+ cells sorted and cloned. Nearly all clones derived had one intact LMO2 allele, while the other alleles had GV/LV/FV proviral insertions. We next assessed LMO2 mRNA and protein expression in these clones. We found a minimal increase in LMO2 mRNA expression with SFFV-FV, in sharp contrast to significantly increased LMO2 expression with SFFV-GV and SFFV-LV by qRT-PCR (Figure 1A). Overall, the SFFV enhancer in GV demonstrated the greatest fold-increase in LMO2 expression (median increase of 280+/-23-fold over unmodified HeLa cells), followed by the SFFV enhancer in LV (median 200+/-27-fold increase). However, the same SFFV enhancer in FV only showed a 45+/-7-fold median increase in expression. Western blot analysis for LMO2 protein expression in three clones for each vector showed no detectable LMO2 expression in SFFV-FV clones, which was similar to baseline in mock (non-targeted) HeLa cells (Figure 1B). However, significantly higher LMO2 protein was detectable in GV and LV clones. Hence, the remarkably low genotoxic potential of FV, even those carrying strong viral enhancers appears to be, in large part, from an insulator property of FV sequences, making FV ideal for situations where high transgene expression, necessitating strong enhancers is required for a therapeutic effect.

Disclosures No relevant conflicts of interest to declare.

  • * Asterisk with author names denotes non-ASH members.