Nuclear receptors (NRs) are transcription factors that regulate genes involved in the development, homeostasis and metabolism of the organisms. Their expression is dynamic and, in the liver, follows a circadian pattern. Among the 48 members of this protein super-family, the retinoid receptor X (RXR) has a special status as an obligatory heterodimer partner for several other NRs. Unlike most of its partners, RXR is suitable for chromatin immunoprecipitation (ChIP) experiments, and is therefore a candidate of choice for genome-wide characterization of NR-dependent regulations. The aim of this thesis was to map the repertoire of RXR chromatin binding sites and explore the ability of RXR to regulate the rhythmic functions of the liver during the day. For that purpose, we used a pan- RXR antibody to perform a ChIP-seq experiment on mouse liver at 7 different time-points of the day over 24 hours.
Faced with the complexity of multiple comparisons between the 7 sets of ChIP-seq peaks obtained at the genome scale, we have developed an in-house method for analyzing the 7 samples together. Just over 30,000 binding sites have been identified, divided into 4 categories: solitary peaks, clustered peaks, high plateaus (wide regions with continuous high tag density), and low plateaus. RXR peaks were preferentially located in promoter regions, and the expression of three-quarters of the corresponding putative RXR target genes, representing almost half of mouse genes, was detectable by microarray. The abundance of RXR binding sites for each gene, but not the intensity of the signal observed at each binding site, correlated with gene expression levels. Intriguingly, the high plateaus most often cover the entire sequence of small and highly expressed genes, from promoter to polyadenylation site. The plateaus signals were remarkably similar to the Pol-II signals, also obtained in ChIP-seq. Taken together with the fact that nearly 40% of the peaks located in promoters exactly overlap with the transcription initiation site (TSS), it suggests a possible direct contribution of RXR to Pol - II's dependent transcription machinery.
Analysis of the density variation of the RXR peaks during the circadian or ultradian rhythm reveals that only 10% of RXR binding sites are considered rhythmic, and only 10% of their putative target genes (a total of 219 genes) presented a rhythmic expression. This disproportion implies that the transcriptional response requires more than the binding rate of RXR.
After gating out the RXR peaks presenting high Pol-II co-occupancy, motif analyses of the sequences under each remaining peaks were performed to identify the associated heterodimeric partners. The binding sites of RXR coupled to a partner are made of two repeats of a canonical hexamer, separated by 1 to 5 base-pairs, and were identified in more than 90% of the remaining peaks. The direct repeat with one base pair spacing (DR1) was the dominant predicted binding mode, followed by the other direct repeat DR2-5, and finally by the IR1-2 inverted repeats. Half of the genes used in this analysis are adornedwith multiple binding modes, and are therefore likely regulated by multiple NRs. Metabolic functions and circadian rhythm maintenance are particularly well represented in this subset of genes.
In summary, these results demonstrate that RXR and its partners interact directly with a significant number of liver genes, although RXR cannot be considered as the driving force behind the internal regulation of the liver clock. A key discovery is the close association of RXR with Pol-II, found on more than 8000 genes. Their co-occupancy, which could be observed on whole genes as well as on TSS, suggests a close involvement of RXR in the transcription machinery, particularly for genes highly expressed in the liver.