NEWS AND VIEWS NUCLEAR RECEPTORS------------------------------------------------------------ Dymer, dymer binding tight Ben Luisi and Leonard Freedman EARLY in the evolution of multicellular organisms, a fortuitous accident enabled tissues to communicate over long dis- tances. The event, a fusion of two genes encoding small proteins binding DNA and lipophilic molecules, respectively, permit- ted diffusible signals to have a direct effect on gene expression. So successful was the modular molecular design that it has been maintained throughout the animal king- dom, and is seen in the conserved domains of a large family of ligand-activated trans- cription factors known as the hormone or nuclear receptors. The DNA-binding domains of several family members have been characterized structurally, revealing a common, metal-bearing fold 1 . With the report of Bourguet et al. on page 377 of this issue 2 , the ligand-binding domain (hereafter LBD) is now also set on a stereochemical foundation. The au- thors present the crystal structure of the apo-LBD from the human receptor for 9-cis retinoic acid (a-isoform, or RXR-a), and it seems likely that its fold, like that of the DNA-binding domain, is common throughout the receptor family. This study also provides the first structural glimpse at the mysterious transcriptional activation domain which resides in the LBD carboxy terminus. Many of the nuclear receptors dimerize through the LBD. The RXR-a receptor appears to associate avidly not only with itself, but also with other receptors, in- cluding those for vitamin D, thyroid hor- mone and all-trans retinoic acid. This association confers an enhanced affinity for a distinct DNA target on each heterodimer 3 , permitting the 'cross-talk' of different ligands and potentially ba- roque physiological effects. Bourguet and colleagues' X-ray struc- ture of the RXR-a LBD shows that the dimer interface coincides with a crys- tallographic symmetry element and is dyad-symmetrical (see figure). It had been noted earlier that the LBD bears a conserved heptad repeat of leucine and other hydrophobic residues, and in 1990 Forman and Samuels 4 put forward the inspired hypothesis that this motif forms a coiled-coil dimerization surface in analogy with the interface found in the leucine zipper and helix-loop-helix transcription factors. It turns out that in the LBD crystal structure, the dimer interface is composed of packed helices and the heptad residues do not directly contribute to the interface. Instead, these residues are interspersed throughout the structure to provide architectural support. As predicted 5 , two helices from each monomer participate in the dimer interface, resulting in a packed helical cluster. NATURE · VOL 375 · 1 JUNE 1995 Judging from the conservation of the hydrophobic scaffold, it seems that the LBD architecture is conserved, but details of the dimer interface are likely to vary. This is expected because certain recep- tors, such as those binding steroids, seem only to form homodimers and do not cross-talk. Variability at the dimer inter- face may therefore define partner choice and, consequently, result in either tight compartmentalization or complex cross- talk in the effects of various ligands. This Symmetry of ligand-binding domain and re- sponse elements for nuclear receptors. The dyad symmetry of the LBD dimer interface stands in contrast to the translational sym- metry of the nuclear receptors' DNA targets, which are either direct or inverted repeats 1 · 3 . Steroid receptors form homodimers and bind inverted repeats whereas nuclear receptors, which form heterodimers, recognize direct repeats. The DNA-binding domain can medi- ate recognition of these repeats through protein-protein interactions 1 · 14 . The match of dimeric receptor to directly repeating element can only be explained if the peptide connec- tion between the DNA- and ligand-binding domains is flexible, like a telephone cord. The LBD dimer therefore serves as a physical tether. DNA binding benefits from this design, because the tether decreases the entropic cost of binding and contributes to an implicit cooperativity. type of functional discrimination is analo- gous to the assembly equilibrium of Jun, Fos and other leucine-zipper transcription factors. Ligand binding is required by most nuclear receptors for modulating gene expression, but the RXR-a LBD crystal structure has been determined in the absence of 9-cis retinoic acid, so it is not clear how the ligand acts structurally. Ligand binding affects dimer/monomer equilibrium in certain receptors 6 ; con- versely, dimerization can affect ligand affinity 7 . These observations indicate that the ligand-binding cavity and the dimer interface are communicating - which, conceivably, could result in quaternary structural change in the dimer. There is spectroscopic evidence for minor secon- dary structural changes with ligand bind- ing in some receptors 8 , and the ligand can also decrease protease sensitivitl. The ligand-binding cavity can be readily mod- elled to fit 9-cis retinoic acid with little conformational adjustment 2 , so the structural responses to ligand may be triggered by very subtle changes in the pocket. The LBD also encompasses the trans- criptional activation domain, called AF-2, which interacts with auxiliary proteins 10 to communicate with the RNA polymerase transcription complex. AF-2 is part of an a-helix and is not disordered as is some- times expected of activation domains. The helix is partly exposed on the surface of the receptor, but some of its buried res- idues have been genetically defined 11 • 12 as mediating interactions with co-adaptors 2 . Ligand binding, which triggers transcrip- tional activation, presumably affects the presentation of AF-2. Two mechanisms can be envisaged: first, that tertiary and secondary structural alterations unveil the hidden AF-2 residues; second, that subunit rearrangements occur through a quaternary transition. If substantiated, the conformational change of AF-2 may explain the stereochemical puzzle of how some reg- ulatory cofactors interact promiscuously, yet with presumed specificity, with many dissimilar proteins. Perhaps the mechan- ism involves refolding of secondary structure to yield a variety of stable com- plexes (there is analogous structural plasticity in other regulatory systems 13 ). Furthermore, if co-adaptors recognize the spacing and orientation of the dyad- related activation helices, then a simple rotation of the two LBDs brought about by ligand-induced quaternary change could also act as a regulatory switch. Subtle differences in the rotation or fold could effect the accommodation of distinct cell-specific or developmental factors. Studies of the DNA-binding domain have revealed how a few amino acids can generate the rich diversity of DNA recog- 359