Loop–loop interactions among nucleic acids
constitute an important form of molecular recognition in
a variety of biological systems. In HIV-1, genomic dimerization
involves an intermolecular RNA loop–loop interaction
at the dimerization initiation site (DIS), a hairpin located
in the 5′ noncoding region that contains an autocomplementary
sequence in the loop. Only two major DIS loop sequence
variants are observed among natural viral isolates. To
investigate sequence and structural constraints on genomic
RNA dimerization as well as loop–loop interactions
in general, we randomized several or all of the nucleotides
in the DIS loop and selected in vitro for dimerization-competent
sequences. Surprisingly, increasing interloop complementarity
above a threshold of 6 bp did not enhance dimerization,
although the combinations of nucleotides forming the theoretically
most stable hexanucleotide duplexes were selected. Noncanonical
interactions contributed significantly to the stability
and/or specificity of the dimeric complexes as demonstrated
by the overwhelming bias for noncanonical base pairs closing
the loop and covariations between flanking and central
loop nucleotides. Degeneration of the entire loop yielded
a complex population of dimerization-competent sequences
whose consensus sequence resembles that of wild-type HIV-1.
We conclude from these findings that the DIS has evolved
to satisfy simultaneous constraints for optimal dimerization
affinity and the capacity for homodimerization. Furthermore,
the most constrained features of the DIS identified by
our experiments could be the basis for the rational design
of DIS-targeted antiviral compounds.