Genomic landscape of human, bat, and ex vivo DNA transposon integrations

Abstract

The integration and fixation preferences of DNA transposons, one of the major classes of eukaryotic transposable elements, have never been evaluated comprehensively on a genome-wide scale. Here, we present a detailed study of the distribution of DNA transposons in the human and bat genomes. We studied three groups of DNA transposons that integrated at different evolutionary times: 1) ancient (>40 My) and currently inactive human elements, 2) younger (<40 My) bat elements, and 3) ex vivo integrations of piggyBat and Sleeping Beauty elements in HeLa cells. Although the distribution of ex vivo elements reflected integration preferences, the distribution of human and (to a lesser extent) bat elements was also affected by selection. We used regression techniques (linear, negative binomial, and logistic regression models with multiple predictors) applied to 20-kb and 1-Mb windows to investigate how the genomic landscape in the vicinity of DNA transposons contributes to their integration and fixation. Our models indicate that genomic landscape explains 16-79% of variability in DNA transposon genome-wide distribution. Importantly, we not only confirmed previously identified predictors (e.g., DNA conformation and recombination hotspots) but also identified several novel predictors (e.g., signatures of double-strand breaks and telomere hexamer). Ex vivo integrations showed a bias toward actively transcribed regions. Older DNA transposons were located in genomic regions scarce in most conserved elements-likely reflecting purifying selection. Our study highlights how DNA transposons are integral to the evolution of bat and human genomes, and has implications for the development of DNA transposon assays for gene therapy and mutagenesis applications.

Keywords: DNA transposons; Myotis lucifugus genome; human genome; integration preferences; logistic regression; multiple linear regression; negative binomial regression.

Fig. 1.

Effects of genomic features in various human DNA tranposon models. The intensity of the color is proportional to the RCVE of each predictor in each model, and the color encodes the sign of the effect - positive in red or negative in blue. White are not significant predictors. LR, logistic regression; RS, R-squared; DE, deviance explained.

Fig. 2.

Effects of genomic features in various bat DNA transposon models. The intensity of the color is proportional to the RCVE of each predictor in each model, and the color encodes the sign of the effect—positive in red or negative in blue. White are not significant predictors. LR, logistic regression; RS, R-squared; DE, deviance explained.