ERR and ERR share high sequence homology at their DNA
ERRα and ERRγ share high sequence homology at their DNA binding domain (98% identity) and bind response elements in a similar fashion (Fig. 3 and Liu et al., 2005), yet ERRγ shows high variability in its capacity to stimulate the different response elements, whereas ERRα does not (Fig. 6A). For example, ERRγ binds the consensus steroidogenic factor 1 response MCB-613 (SF-1RE) and thyroid hormone response element (TRE), but it only activates the SF-1RE moderately and is inactive on the TRE, whereas ERRα activates both elements (Heard et al., 2000, Sanyal et al., 2002). Additionally, the osteopontin (OPN) gene promoter is very sensitive to ERRα stimulation but not to ERRγ (Bonnelye and Aubin, 2005, Huppunen et al., 2004). These studies demonstrate that the ERRα and ERRγ have very different molecular and biochemical characteristics. Indeed, our molecular study showed that the conformation of ERRγ is influenced by binding to the MHRE (Fig. 6B) and the changes of conformation could lead to the recruitment of coactivators or release of corepressors, thus affecting its transactivation function. The biological implication of ERRγ conformational change upon binding to different response elements is not clear but may be important to its functional roles, especially since it has a broad range of binding specificity (Razzaque et al., 2004). Another major difference between ERRα and ERRγ is that ERRγ interacts strongly with SRC1, a coactivator not only required for the transactivation function of nuclear receptors but also needed to increase the potency of PGC-1α (Puigserver et al., 1999). Specifically, the AF2 domain of the ERRγ is required for the protein–protein interaction with coactivators. On the other hand, ERRα interacts with the coactivator weakly whether AF2 domain is present suggesting, that the region of receptors and coactivators interaction is different between ERRα and ERRγ. Interestingly, both ERRα and ERRγ bind PGC-1α identically. The differences in coactivator interaction are also reflected in their coactivation function, in which PGC-1α is the most potent coactivator for both ERRα and ERRγ whereas SRC-1 is a potent coactivator for ERRγ. Since SRC-1 is ubiquitously expressed, it may explain why ERRγ actively stimulates the ERRα gene expression in the absence of PGC-1α in several cell lines tested (Fig. 6D and Liu et al., 2005). How ERRα represses the ERRγ-stimulated MHRE activity is unclear, nonetheless, we can rule out the competition for binding to the DNA since ERRα mutants deficient in DNA binding (p-box mut) retain the repression function, whereas the N-terminus containing the DNA binding domain lost the repression function. These study suggest that the repression function of the ERRα lies at the C-terminal region. In EMSA study, both ERRα and ERRγ were present on the MHRE although it was not clear whether they form heterodimers. Moreover, the ERRα and ERRγ interact with each other in vitro and in vivo. It has been shown that heterodimerization between ERRα and ERRγ inhibits the transactivation function of each other, in contrast to the homodimerization which is needed for their transcriptional activity (Horard et al., 2004a, Horard et al., 2004b, Huppunen and Aarnisalo, 2004). It is also known that one of the dimerization interfaces of ERRα resides in the C-terminal region (Horard et al., 2004a, Horard et al., 2004b), implicating that the heterodimerization between the ERRs is one of the mechanism for ERRα repression. We detected both ERRα and ERRγ on the chromatin of MHRE by ChIP assay and showed that overexpression of ERRγ significantly increases its occupancy on the chromatin of MHRE without affecting the ERRα presence, and vice versa (Fig. 7A). MHRE is composed of multiple ERE half sites with various spacing and orientations (Liu et al., 2003) and the two reversed ERE half sites (TGACCTTCA) separated with 14bp are important for ERRs binding and transcriptional activity (Fig. 3 and Liu et al., 2003). It is unusual for a nuclear receptor to bind two ERE half site with such wide space in between. Nonetheless, ERRγ has been reported to bind a broad range of sequences, including the monovalent binding sites (Razzaque et al., 2004), whereas ERRα also binds variety of response sequences (Barry et al., 2006, Johnston et al., 1997) including the NRRE-1 of the medium chain acyl-coenzyme A dehydrogenase (MCAD) gene at sites with 14bp in the middle (Maehara et al., 2003). Recently, Giguere\'s laboratory demonstrated that a single nucleotide change in ERRα binding site can affect the mode of PGC-1α activation of its target promoters (Barry et al., 2006). Thus, the binding characteristics of ERRs may influence the recruitment of coactivators and the transactivation activity.