The superfamily of nuclear receptors (NRs) are important transcription factors that regulate functions as diverse as reproduction, differentiation, development, metabolism, metamorphosis and homeostasis. The human genome contains 48 NR genes. Classical endocrine NRs and adopted orphan NRs are ligand-inducible, whereas for more than 20 orphan NRs still a ligand has yet to be identified. Most NRs, such as vitamin D receptor and the adopted orphan NRs pregnane X receptor (PXR) and constitutive androstane receptor (CAR), preferentially form heterodimeric complexes with the retinoid X receptor (RXR) and bind to specific REs in the promoter regions of hormone/xenobiotic responsive genes. REs are formed by two hexameric core binding motifs of the consensus sequence RGKTCA (R = A or G, K = G or T). In most of the presently known REs the core binding motifs where found in a directly repeated (DR) arrangement. In the remaining REs the motifs are either in an inverted (IR) or everted (ER) repeat arrangement. In all cases the number of spacing nucleotides is critical.
It is thought that around 100-200 genes out of the 35000 human genes are regulated by one nuclear hormone. Statistically there are around 6000 binding sites of one type of REs (for example DR4) in the non-coding region of the human genome. This discrepancy between apparent and truly responsive genes is further compounded by the fact that several NRs share the same type of REs.
Results
Our experimental studies show that the complex of the RXR with either the receptors PXR or CAR have comparable functionality on a RE of the rat pit-1 gene that is formed by a DR4-type RE. For both NRs the sequence of the 5´-flanking dinucleotides have a modulatory effect of up to 3- and 20-fold for the upstream and downstream hexamers, respectively. This suggests that more likely octameric motifs instead of hexameric sequences should be considered as specific NR binding sites.
For the in silico analysis we used the data-mining tool NUBIScan (Podvinec et.al., Molecular Endocrinology 16:1269-1279, 2002) which uses the standard hexameric motif to predict NR binding sites. For our specific purposes we generated a hexameric weight matrix with the RGKTCA motif. As a final step, for a more precise prediction of hormone responsiveness, we re-analysed the found REs with our octamer concept.
To test our hypothesis and for a better understanding of the network of regulation amongst NRs, we first analysed all members of the human NR superfamily. In total of 1710936 nucleotides of the maximal regulatory region were taken into account. We found the statistical expected number of perfect consensus ER6- and ER9-type REs (92%). Interestingly we also found that DR5-type REs are under-represented (31%). In contrast, DR3- and DR4-type REs are over-represented (276%).
In closer examination of the sequences we found that many of the DR4-type REs are associated with an additional hexamer site which creates a DR4-DR2 motif. Flanking sequence analysis revealed that this DR4-DR2 element is in fact inside an ALU repetive sequence.
We used several experimental assay systems (gelshift, reporter-gene-assays, RT-PCR and ChIP) to confirm that these sequences have the ability to bind NRs and activate gene expression.
Conclusions
The finding that the 5´-flanking dinucleotides have a drastic and receptor-specific effect on heterodimer complex formation shows that octameric instead of hexameric sequences should be considered as specific NR binding sites. This strategy gives a more definitve list of putative REs in promoter sequences and a more precise prediction of hormone responsiveness of the respective genes. This concept is supported by the results of our analysis of the promoter-region of the 48 members of the human NRs superfamily and verified by several experimental assay systems.
