817. Nonrandom Insertion Site Preferences for the SB Transposon In Vitro and In Vivo

Abstract

The recent demonstration that HIV-1, MLV, and AAV all strongly favor integration into active genes has raised significant concerns about the future of gene insertion-based therapies. The Sleeping Beauty (SB) transposon has shown tremendous success in preclinical gene therapy models and is a promising alternative to viral-based in vivo delivery methods, and herein, we investigated its target site preferences. Using a plasmid rescue strategy, we isolated 524 unique, unselected transposition events from adult mouse liver and compared the chromosomal features associated with these coordinates with 10,000 computer-simulated random integrations. SB insertions were distributed evenly on each chromosome, but were highly favored in 2-6 bp simple sequence repeats (p < 0.0001) and were strongly disfavored in genomic LINEs (p < 0.0001). Also, SB showed a marked preference for integration into localized AT-rich palindromes, especially the consensus sequence GNATATATANC (p < 0.01). These attributes are consistent with a bendable target DNA structure and suggest a two-fold dyad symmetry at the insertion site. We also mapped SB insertions with respect to the RefSeq genes defined by the UCSC mouse genome database. Results showed that ~20% (103/524) of transposon insertions mapped within at least one RefSeq gene, which was not statistically different from the control set (p=0.3). These intragenic insertions were distributed evenly along the transcripts without orientational preference, but were almost exclusively within introns. Interestingly, all five of the exon insertions we isolated were in the reverse orientation of transcription and all mapped near the 3′ end of genes, very close to the translational stop site. Moreover, even though SB does not appear to actively target intragenic regions, it showed a weak but significant tendency to insert upstream of genes (p=0.009), especially ± 5 kb of the transcriptional start site (p < 0.005) and within known CpG islands (p=0.0001). We therefore analyzed an Affymetrix chip dataset available on the web to test whether preferences for promoter regions might be due in part to transcriptional activity. However, results showed no statistical difference between the 44 SB-targeted and 70 randomly-targeted RefSeq genes represented on the array (p=0.25), suggesting that chromatin accessibility, rather than active tethering, may be involved in SB's regional insertion site preferences. Currently, we are analyzing a large number of insertions from mouse and human cell lines in order to test the generality of these preferences in different tissues and/or species, and are determining whether or not genetic selection introduces any bias during SB insertion site selection. Collectively, these data demonstrate a nonrandom distribution for SB insertions in mammals, and suggest that SB integration may be significantly safer for stable in vivo gene delivery than many widely used viral-based vectors.