Abstract
The sequences required for integration of retroviral DNA have been analyzed in vitro. However, the in vitro experiments do not agree on which sequences are required for integration: for example, whether or not the conserved CA dinucleotide in the 3' end of the viral DNA is required for normal integration. At least a portion of the problem is due to differences in the experimental conditions used in the in vitro assays. To avoid the issue of what experimental conditions to use, we took an in vivo approach. We made mutations in the 5' end of the U3 sequence of the Rous sarcoma virus (RSV)-derived vector RSVP(A)Z. We present evidence that, in RSV, the CA dinucleotide in the 5' end of U3 is not essential for appropriate integration. This result differs from the results seen with mutations in the U5 end, where the CA appears to be essential for proper integration in vivo. In addition, based on the structure of circular viral DNAs smaller than the full-length viral genome, our results suggest that there is little, if any, integrase-mediated autointegration of RSV linear DNA in vivo.