Hypotheses for testing deviations from random integration: evidence for nonrandom retroviral integration

The use of a recently developed in vitro model for retroviral integration provides a means of statistically testing hypotheses concerning the distribution of integration sites and hypotheses about the sequence of proviral orientations. In this study, three null hypotheses are formulated and applied to previously published data. Statistical analyses of these data suggest that the distribution of integration sites may not be uniform, and the sequence of proviral orientations is not random. On the basis of these results and the observed clustering of orientations, it was postulated that if a DNA sequence was involved in nonrandom proviral integration, that sequence would be found in the regions where the orientations change direction with respect to the target DNA ("I" regions). Computer analyses for homologous and complementary DNA sequences were performed on all possible pairs of identifiable "I" regions. A common sequence (at least 8 bp in size) was found in three out of four regions and that sequence was absent elsewhere in the target DNA. A model, with features of recombination reminiscent of chi sequences in bacteria, is proposed that may account for these results.