Structural basis for engineering of retinoic acid receptor isotype-selective agonists and antagonists.

BACKGROUND: Many synthetic retinoids have been generated that exhibit a distinct pattern of agonist/antagonist activities with the three retinoic acid receptors (RARalpha, RARbeta and RARgamma). Because these retinoids are selective tools with which to dissect the pleiotropic functions of the natural pan-agonist, retinoic acid, and might constitute new therapeutic drugs, we have determined the structural basis of their receptor specificity and compared their activities in animal and yeast cells. RESULTS: There are only three divergent amino acid residues in the ligand binding pockets (LBPs) of RARalpha, RARbeta and RARgamma. We demonstrate here that the ability of monospecific (class I) retinoid agonists and antagonists to bind to and induce or inhibit transactivation by a given isotype is directly linked to the nature of these residues. The agonist/antagonist potential of class II retinoids, which bind to all three RARs but depending on the RAR isotype have the potential to act as agonists or antagonists, was also largely determined by the three divergent LBP residues. These mutational studies were complemented by modelling, on the basis of the three-dimensional structures of the RAR ligand-binding domains, and a comparison of the retinoid agonist/antagonist activities in animal and yeast cells. CONCLUSIONS: Our results reveal the rational basis of RAR isotype selectivity, explain the existence of class I and II retinoids, and provide a structural concept of ligand-mediated antagonism. Interestingly, the agonist/antagonist characteristics of retinoids are not conserved in yeast cells, suggesting that yeast co-regulators interact with RARs in a different way than the animal cell homologues do.