Abstract
The genome of all retroviruses consists of two identical RNAs noncovalently linked near their 5‘ end. Dimerization of genomic RNA is thought to modulate several steps in the retroviral life cycle, such as recombination, translation, and encapsidation. The kissing-loop model of HIV-1 genome dimerization [Laughrea, M., & Jetté, L. (1994) Biochemistry 33, 13464−13474; Skripkin et al. (1994) Proc. Natl. Acad. Sci. U.S.A. 91, 4945−4949] posits that the 248−270 region of the HIV-1 genome, by forming a hairpin and initiating dimerization through a loop−loop interaction, is the full or at least the core dimerization domain of HIV-1 RNA. Here, we show by nested deletion analysis that the 3‘ boundary of the HIV-1 dimerization domain is immediately downstream of hairpin 248−270 and that the isolated region 248−271 dimerizes at least as readily as longer RNAs. Among various HIV-1Lai RNA transcripts containing hairpin 248−270, all form two types of dimer, as is implicit in the kissing-loop model. The high-stability dimer resists semidenaturing conditions and the low-stability dimer cannot, which is consistent with the model. At physiological temperatures, low-stability dimers are usually formed, as if dimerization without nucleocapsid proteins corresponded to loop−loop interaction without switching from intra- to interstrand hydrogen bonding. Our results show that the 3‘ DLS (a sequence immediately 3‘ from the 5‘ splice junction and originally thought to be the dimerization domain of the HIV-1 genome) and adjacent nucleotides are not necessary for efficient dimerization of HIV-1Lai RNA at low and high ionic strength. Upstream of hairpin 248−270 exists another “DLS-like” sequence that we name 5‘ DLS: like the isolated 3‘ DLS, the isolated 5‘ DLS forms an apparently nonphysiological structure that can become substantially dimeric at high ionic strength.