We have used the yeast two-hybrid system to localize the ligand-dependent dimerization domain of the estrogen receptor-alpha (ER) to region E in vivo. In this system, the cDNAs corresponding to the A-D, E, E/F, A-E (deltaF), and full-length (wtER) domains of the human ER were each cloned into the yeast two-hybrid vectors GAL4 DB and GAL4 TA and expressed in different combinations in yeast harboring a GAL1-lacZ reporter. The reporter was used as a relative measure of the interaction between the ER domains, through reconstitution of GAL4 activity. We found that the interaction of E or E/F domains of the ER with full-length ER is estradiol dependent and estrogen responsive element independent, as measured by the reconstitution of GAL4 activity from GAL4-E domain-containing fusion protein interactions. In the presence of F domain, this activity is reduced 10-fold. The results suggest that sequences in the F domain are inhibitory to the dimerization signal that is present in the E region. We propose that the full-length ER contains intrinsic dimerization restraints contributed by regions outside domain E that are released upon binding hormone agonist. In addition, we have demonstrated that coactivator RIP140 is able to interact with the ER in vivo at the E domain of the receptor in the presence of estrogen. Yeast two-hybrid analysis shows that RIP140 does not homodimerize in the presence or absence of estrogens. We present evidence showing that the ER has the inherent ability to interact with RIP140 in the presence of antiestrogens, but sequences inherent in the ER itself that are present outside of the E domain compromise this ability.