Nuclear receptors are ligand-activated transcription factors that modulate gene regulatory networks from embryonic development to adult physiology and thus represent major targets for clinical interventions in many diseases. Most nuclear receptors function either as homodimers or as heterodimers. The dimerization is crucial for gene regulation by nuclear receptors, by extending the repertoire of binding sites in the promoters or the enhancers of target genes via combinatorial interactions. Here, we focused our attention on an unusual structural variation of the α-helix, called π-turn that is present in helix H7 of the ligand-binding domain of RXR and HNF4. By tracing back the complex evolutionary history of the π-turn, we demonstrate that it was present ancestrally and then independently lost in several nuclear receptor lineages. Importantly, the evolutionary history of the π-turn motif is parallel to the evolutionary diversification of the nuclear receptor dimerization ability from ancestral homodimers to derived heterodimers. We then carried out structural and biophysical analyses, in particular through point mutation studies of key RXR signature residues and showed that this motif plays a critical role in the network of interactions stabilizing homodimers. We further showed that the π-turn was instrumental in allowing a flexible heterodimeric interface of RXR in order to accommodate multiple interfaces with numerous partners and critical for the emergence of high affinity receptors. Altogether, our work allows to identify a functional role for the π-turn in oligomerization of nuclear receptors and reveals how this motif is linked to the emergence of a critical biological function. We conclude that the π-turn can be viewed as a structural exaptation that has contributed to enlarging the functional repertoire of nuclear receptors.

Ancestral character reconstruction and stochastic mapping were performed under R version 3.2.2 using the make.simmap function as implemented in the phytools package version 0.5.0. Character evolution was inferred using a model of symmetrical transition rates between the character states (SYM). 10 000 character histories were sampled to allow the incorporation of the uncertainty associated with the transition between different states. Inferred state frequencies for ancestral nodes were plotted using the describe.simmap function.

This dataset contains all the files necessary to reproduce Figure 3 of the associated paper. Those file are:

- script_Fig3.R : the R script necessary to load the data and process them as indicated in the methods section. The outputs of character mappings are also indicated in the script file, in order that the user can compare them with the results he would get on his/her own computer.

- tree_Fig3.nex: the backbone tree used for character mapping

- Fig3A_RxxxE.csv: the data table containing the presence-absence data regarding the RxxxE motif for each nuclear receptor. Those characters were plotted on panel A of Figure 3.

- Fig3B_dimerization.csv: the data table containing the data on dimerization modes accross nuclear receptors. Those character were plotted on panel B of Figure 3.