Estrogen receptors (ERs) alpha, beta and gamma are ligand-dependent transcription factors that regulate vertebrate reproduction, cell survival and other physiological processes. Here, we report an integrative analysis of mammalian and teleost receptors, including the first structure of an ERgamma ligand binding domain (LBD), showing that structural divergence acquired during evolution is accommodated by different strategies within functional regions to preserve intrinsic regulatory mechanisms. Supervised-learning and comparative Markov modelling uncover highly constrained network structures essential for allostery and protein folding, revealing a fundamental regulatory architecture and source of selective pressure in ER+ breast cancer and reproductive disorders. Finally, cross-referencing molecular constraints to genome sequencing reveal evolutionary origins underlying natural genetic variation in humans and widespread disruption of constraints. This work provides structural insights into the conservation of gene regulation by essential transcription factors and has implications for precision medicine.