DNA shuffling method for generating estrogen receptor alpha and beta chimeras in yeast

To facilitate our study of the molecular basis for the estrogen receptor (ER) subtype selectivity of novel ligands, we used DNA shuffling to construct chimeric ERs having ligand binding domains derived from both ER alpha and ER beta. The efficiency of chimera generation was low with traditional DNA shuffling protocols. Furthermore, ER ligand binding domain sequences lack convenient restriction sites for introducing chimeric ligand binding domain sequences into expression vectors. To overcome these problems, we developed a modified strategy whereby chimeric sequences were exclusively amplified from among the reassembled products from DNA shuffling using a special pair of PCR primers whose 3' ends specifically match the alpha and beta sequences, respectively, and whose 5' ends match sequences outside the ER beta ligand binding domain. When chimeric ligand binding domain DNA sequences, amplified with these primers, were co-transformed into a yeast strain with a linearized expression vector for ER beta, an active expression vector was produced by homologous recombination. Twenty-two different crossover sites were found; most occurred when there was a stretch of eight or more identical base pairs in both sequences, and many were concentrated in the regions important for studying ligand binding and transactivation. This method should prove to be useful for generating chimeric gene products from parent templates that share relatively low sequence identity.     

Stability of the ligand-estrogen receptor interaction depends on estrogen response element flanking sequences and cellular factors

To determine whether accessory proteins mediate the ligand- and DNA sequence-dependent specificity of estrogen receptor (ER) interaction with DNA, the binding of partly purified vs highly purified bovine ER to various estrogen response elements (EREs) was measured in the presence of different ER ligands. Partly purified estradiol-liganded ER (E₂-ER) binds cooperatively to stereoaligned tandem EREs flanked by naturally occurring AT-rich sequences, with a stoichiometry of one E₂-ER dimer per ERE. In contrast, highly purified E₂-ER binds with a 10-fold lower affinity and non-cooperatively to EREs flanked by the AT-rich region. Moreover, the binding stoichiometry of highly purified E₂-ER was 0.5 E₂-ER dimer, or one monomer per ERE, independent of the ERE flanking sequence. Interestingly, the binding of ER liganded with the antiestrogen 4-hydroxytamoxifen (4-OHT-ER) was non-cooperative with an apparent stoichiometry of 0.5 4-OHT-ER dimer per ERE, regardless of ER purity or ERE flanking sequence. We recently showed that when 4-OHT-ER binds DNA, one molecule of 4-OHT dissociates from the dimeric 4-OHT-ER-ERE complex, accounting for the reduced apparent binding stoichiometry. In contrast, ER covalently bound by tamoxifen aziridine (TAz) gave an ERE binding stoichiometry of one TAz-ER dimer per ERE, and TAz-ER binds cooperatively to multiple AT-rich EREs, regardless of the purity of the receptor. We have obtained evidence that purification of ER removes an accessory protein(s) that interacts with ER in a sequence- and/or DNA conformational-dependent manner, resulting in stabilization of E₂, but not 4-OHT, in the ligand binding domain when the receptor binds to DNA. We postulate that retention of ligand by ER maintains the receptor in a conformation necessary to achieve high-affinity, cooperative ERE binding.