High Affinity Heme Binding to a Heme Regulatory Motif on the Nuclear Receptor Rev-erbβ Leads to Its Degradation and Indirectly Regulates Its Interaction with Nuclear Receptor Corepressor*

Rev-erbα and Rev-erbβ are heme-binding nuclear receptors (NR) that repress the transcription of genes involved in regulating metabolism, inflammation, and the circadian clock. Previous gene expression and co-immunoprecipitation studies led to a model in which heme binding to Rev-erbα recruits nuclear receptor corepressor 1 (NCoR1) into an active repressor complex. However, in contradiction, biochemical and crystallographic studies have shown that heme decreases the affinity of the ligand-binding domain of Rev-erb NRs for NCoR1 peptides. One explanation for this discrepancy is that the ligand-binding domain and NCoR1 peptides used for in vitro studies cannot replicate the key features of the full-length proteins used in cellular studies. However, the combined in vitro and cellular results described here demonstrate that heme does not directly promote interactions between full-length Rev-erbβ (FLRev-erbβ) and an NCoR1 construct encompassing all three NR interaction domains. NCoR1 tightly binds both apo- and heme-replete FLRev-erbβ·DNA complexes; furthermore, heme, at high concentrations, destabilizes the FLRev-erbβ·NCoR1 complex. The interaction between FLRev-erbβ and NCoR1 as well as Rev-erbβ repression at the Bmal1 promoter appear to be modulated by another cellular factor(s), at least one of which is related to the ubiquitin-proteasome pathway. Our studies suggest that heme is involved in regulating the degradation of Rev-erbβ in a manner consistent with its role in circadian rhythm maintenance. Finally, the very slow rate constant (10−6 s−1) of heme dissociation from Rev-erbβ rules out a prior proposal that Rev-erbβ acts as an intracellular heme sensor.

Nuclear receptors (NRs) 2 are eukaryotic transcription factors that activate or repress gene expression in response to binding small signaling molecules such as steroid hormones, lipophilic vitamins, and fatty acids (1). Genes regulated by NRs are involved in a myriad of cellular processes ranging from metabolic homeostasis to growth and development. The subjects of this article, Rev-erb␣ and Rev-erb␤, are NRs that have overlapping functions and tissue expression patterns (2)(3)(4)(5)(6)(7)(8)(9) and are under circadian control (10 -12). Rev-erb NRs repress the transcription of key genes, e.g. Bmal1, CLOCK, and Rev-erb␣ itself, involved in regulating the molecular clock (13)(14)(15). This is accomplished in part through the formation of a heterodimeric Bmal1-CLOCK complex that activates the transcription of clock-controlled genes including Rev-erb␣/␤, completing a critical feedback loop required for circadian rhythm maintenance (6,7). Rev-erb␣ and Rev-erb␤ also regulate the expression of genes involved in gluconeogenesis, lipid homeostasis, and immune responses, ultimately synchronizing these processes to the diurnal cycle (16 -21).
The multiple functions of NRs (DNA, ligand, and coregulator binding) are accomplished through their modular structure (22,23). The N-terminal A/B domain is hypervariable, is involved in ligand-independent transcriptional regulation, and is subject to phosphorylation. This is followed by a conserved DNA-binding domain (DBD, the C domain) containing two zinc-finger motifs that poise the receptor to interact with specific response elements in the promoter of target genes. A flexible hinge region (the D domain) links the DBD to the C-terminal globular ligand-binding domain (LBD or the E domain). The LBD of NRs is generally composed of 12 ␣-helices arranged in a three-layered bundle that forms a hydrophobic ligand pocket. In the absence of ligand, some NRs, such as the thyroid hormone receptor (THR) and retinoic acid receptor, repress gene expression by interacting with transcription corepressors including nuclear receptor corepressor 1 (NCoR1) and silencing mediator of retinoid and thyroid receptors (SMRT) (24); binding of NR antagonists can also lead to corepressor recruitment (25). Both NCoR1 and SMRT are large (Ͼ250 kDa) pro-teins that bind histone deacetylase complexes and contain multiple NR interaction domains at their C termini (26). Interaction domain motifs are composed of short amphipathic helices with an I/LXXI/VIXXXL/F/Y consensus that associate with NRs by binding to a hydrophobic cleft on the LBD (27). Binding of the specific ligand, i.e. thyroid hormone for the THR, to the LBD causes ␣-helix 12 of the NR to adopt a compact conformation against the LBD that partially occludes the corepressor cleft and exposes a region that recognizes transcriptional coactivator proteins like the p160 steroid receptor coactivator. These coactivators contain LXXLL NR box motifs that also form helices and bind to the shorter cleft once occupied by corepressor interaction domains (28). Rev-erb␣ and Rev-erb␤ isoforms are unique to the NR superfamily in that they lack ␣-helix 12, rendering them unable to associate with a coactivator. Thus, these NRs are thought to function exclusively as repressors through binding of NCoR1, although Rev-erb␤ has been shown to activate Srebp-1c in skeletal muscle (29,30).
Identifying the ligand and how binding of the ligand affects various properties of the NR, such as its subcellular localization, stability, oligomeric state, repressor/activator activity, and interactions with DNA and/or a coregulator, are essential for understanding the role of the NR. Many NRs are classed as orphan receptors, meaning that their ligands are unknown. Rev-erb␣ and Rev-erb␤ were first identified as orphan members of the NR family (31,32). When the Drosophila homolog of Rev-erb, E75, was found to bind heme (33), various studies were performed that ultimately demonstrated that Rev-erb␣ and Rev-erb␤ also bind heme and that this interaction appears to be responsible for recruitment of NCoR1, leading to repression of Rev-erb target genes (16,34), which include important regulators of glucose and lipid metabolism, immune responses, and the molecular clock (16 -21).
Intriguingly, the LBDs of E75, Rev-erb␣ and Rev-erb␤, contain a heme regulatory motif (HRM), which consists of a Cys-Pro dipeptide generally followed by a hydrophobic residue (35). For Rev-erb␤, the relevant sequence is C 374 NTGGRMHLVC 384 PMSK, where Cys-384 (Cys-418 for Rev-erb␣) and His-568 serve as heme axial ligands (36). In most HRMs, the Cys residue is followed by a Pro, where the conformational inflexibility of the Pro directs residues C-terminal to the Cys-Pro dipeptide away from the heme face, poising the Cys thiolate for an interaction with heme (37). Cys-384 is involved in a redox-regulated ligand switch in which it forms a disulfide bond with Cys-374, which lowers heme affinity (38). HRMs have also been proposed to control the activity or stability of several regulatory proteins including an iron-responsive regulator in bacteria (39), a kinase (eIF2␣ kinase or heme-regulated inhibitor (HRI)) (40), a mammalian transcription repressor (Bach1) (41), 5-aminolevulinate synthase (42), and the yeast transcription factor Hap1, which mediates the effects of oxygen on transcription (43). In 5-aminolevulinate synthase, three HRMs facilitate the heme-dependent inhibition of mitochondrial import of the synthase, while in the mammalian iron-responsive protein IRP2, heme binding to the HRM appears to regulate protein degradation (44). In the case of heme oxygenase-2, it was concluded that one heme can bind to each of the two HRMs at the C terminus as well as to the active site, where it undergoes conversion to biliverdin, CO, and iron (45,46).
The Drosophila Rev-erb homolog, E75, acts as a repressor by heterodimerizing with DHR3, inhibiting its ability to activate the transcription of genes involved in fly development (47). The Fe 2ϩ -heme form of E75 preferentially interacts with DHR3, whereas NO binding to form the Fe 2ϩ -heme⅐NO complex destabilizes the heterodimer, thus implicating heme binding, its redox state, and interaction with diatomic gases as regulators of E75 activity (33,47). Early studies suggested that the heme redox state and NO do not play a role in Rev-erb function, indicating that the regulatory role of heme may differ among the various Rev-erb homologs (34). On the other hand, the results of more recent experiments indicate that NO destabilizes the Rev-erb␣/␤⅐NCoR1 interaction akin to the regulatory mechanism of E75 (36), and thus, additional data are required to assess the role of NO in regulating Rev-erb activity.
The results of co-immunoprecipitation (co-IP), chromatin immunoprecipitation (ChIP), and gene expression experiments indicate that heme promotes the interaction of Rev-erb␣ with NCoR1; thus, Rev-erb NRs were proposed to act as heme sensors (16,34). These results implied that heme acts as a classical nuclear receptor ligand in which heme binding leads to a conformational change that increases the affinity of the Rev-erb NR for NCoR1, ultimately amplifying gene repression. However, in contrast to the heme-dependent NCoR1 recruitment model, several biochemical studies indicate that heme binding actually destabilizes the Rev-erb⅐NCoR1 complex. For example, titration of purified Rev-erb␣/␤ LBDs with Fe 3ϩ -and Fe 2ϩ -heme decreases the affinity of Rev-erb for small peptides encompassing the NCoR1 interaction domains (16,34,36). In addition, when the crystal structures of apoRev-erb␤ LBD (29), Fe 3ϩheme-bound Rev-erb␤ LBD (36), and Rev-erb␣ LBD in complex with the NCoR1 interaction domain 1 (ID1) peptide (27) are compared, it is apparent that heme binding induces structural rearrangements that would cause significant steric clash between ␣-helix 3 of the LBD and a unique anti-parallel ␤-sheet of ID1. Furthermore, Trp402 and the heme axial ligand His-568 (Rev-erb␤ amino acid numbering), which provide hydrophobic interactions in the heme-binding pocket, is unfavorably positioned to accommodate heme in the ID1⅐Rev-erb␣ structure. In sum, although studies with full-length proteins in cells provided a model in which heme has a robust stimulatory effect on the interaction between Rev-erb␣ and NCoR1 (16), biochemical and structural studies (16,27,29,34,36) indicate that heme binding may destabilize the Rev-erb␣/␤⅐NCoR1 complex. However, it should be re-emphasized that these in vitro studies were performed with the isolated LBD and NCoR1 peptides. Because of their generally intractable nature (poor solubility, instability, etc.), there are few studies with full-length NRs and their coregulators. However, concerns have been expressed about the use of domains in studying the biochemistry of NRs (see for example, references 23 and 48). One hypothesis to explain why the in vitro studies cannot reproduce the role of heme in promoting complex formation between Rev-erb␣/␤ and NCoR1 full-length proteins in vivo is that the truncated constructs omit features that are important for heme-dependent regulation.
To test the first hypothesis, we generated an NCoR1 construct encompassing all three NR interaction domains and fulllength Rev-erb␤ (FLRev-erb␤). Biochemical studies indicate that heme dissociates from Rev-erb␤ too slowly to equilibrate with the heme pool and thus would be unable to act, as suggested previously (16,49), as an intracellular heme sensor. We also demonstrate that heme has only a minimal effect on the binding of NCoR1 to FLRev-erb␤, refuting our original hypothesis that full-length proteins could reproduce the effects of heme observed in the cell. Instead, our co-IP results demonstrate that endogenous NCoR1 in cell extracts interacts with wild-type Rev-erb␤ but not with a variant defective in heme binding, indicating that another factor present in extracts mediates the heme-dependent interaction of Rev-erb␤ with NCoR1. Potential candidates are proteasome-related proteins or a post-translational modification such as ubiquitination, because we found through mass spectroscopic analyses of the co-IP experiment that an ubiquitin-ligase associates in a hemedependent manner with Rev-erb␤. In addition, apoRev-erb␤ is more stable than the heme-replete protein in cells. Our results also suggest that there are differences in the mechanisms by which heme affects the repressor functions of Rev-erb␣ and Rev-erb␤. Although heme binding is dispensable for Rev-erb␤ repression of the Bmal1 promoter, it modestly enhances Rev-erb␣ repressor activity.

Experimental Procedures
Plasmids-For the heterologous expression and purification of human FLRev-erb␤ in Escherichia coli, two constructs were generated: one contained the sequence from the mammalian gene collection (MGC), pGB1-FLRev-erb␤-MGC, and the other was from the human genome sequence (pGB1-FLRev-erb␤-HGS). The MGC clone differs from the human genome sequence (GenBank TM accession number NM_005126.4) by three missense mutations that result in amino acid substitutions P21H, Q282K, and P288R. The sequences of all plasmids used in this study were verified at the University of Michigan Sequencing Core (Ann Arbor, MI).
To prepare the pGB1-FLRev-erb␤-MGC construct, the Rev-erb␤ open reading frame (ORF) lacking the ATG start codon was amplified with PCR using PfuUltra II HS polymerase (Agilent Technologies, Santa Clara, CA), mammalian gene collection clone 5274113 (GenBank TM accession number BC045613.1, coding for human NR1D2) as template, and FLRev-erb␤ ligation-independent cloning (LIC) forward and reverse primers (refer to Table 1 for complete primer sequences and plasmid synopses). LIC techniques were utilized to clone the isolated PCR fragment into a modified pMCSG7 vector containing an N-terminal hexahistidine-tagged protein G, B1 domain (GB1) with a C-terminal tobacco etch virus (TEV) proteolytic cleavage site (50,51). The resulting plasmid encodes a translational fusion between GB1 and FLRev-erb␤ with an internal TEV protease cleavage site for downstream removal of GB1 and the His tag. TEV cleavage resulted in three residues from the TEV recognition sequence remaining on the N terminus of FLRev-erb␤ with the following sequence, NH 2 -SNAE-, where Glu (E) is the first residue of Rev-erb␤.
Site-directed mutagenesis of pGB1-FLRev-erb␤-MGC was carried out to generate an additional clone identical to the human genome sequence, designated pGB1-FLRev-erb␤-HGS, using the QuikChange protocol (Agilent Technologies). PCR was conducted using overlapping primers containing H21P, K282Q, and R288P mutations. Mutant PCR products were digested with DpnI (New England Biolabs, Ipswich, MA) overnight at 37°C prior to transformation into E. coli Top10 cells (Life Technologies). pGB1-FLRev-erb␤-HGS served as the template for generation of heme axial ligand mutants H568F and H568F/C384A.
To generate a clone of FLRev-erb␤ for expression in mammalian cells, the ORF from pGB1-FLRev-erb␤-HGS was amplified with PCR using a forward primer containing a 5Ј-KpnI restriction site and a consensus Kozak sequence and a reverse primer with a 5Ј-XhoI restriction site. The PCR fragment was digested with the indicated endonucleases and then ligated into similarly digested pcDNA3.1(ϩ) resulting in pcDNA3.1(ϩ)-FLRev-erb␤. The H568F, C384A, and H568F/C384A mutations were introduced into pcDNA3.1(ϩ)-FLRev-erb␤ as described above.
To express Rev-erb␣ and the corresponding H602F variant, the Rev-erb␣ ORF from mammalian gene collection clone 6190140 (GenBank TM accession number BC047875.1, coding for human NR1D1) was PCR-amplified using a forward primer consisting of a consensus Kozak sequence and a 5Ј-KpnI restriction site and a reverse primer with a 5Ј-NotI restriction site. The PCR fragment was digested with the indicated endonucleases and ligated into similarly digested pcDNA3.1(ϩ) yielding pcDNA3.1(ϩ)-Rev-erb␣. The H602F mutation was introduced into this vector as described above.
Expression and purification of soluble forms of NCoR1 containing all three interaction domains and of full-length THR␤1 were accomplished by generating maltose-binding protein (MBP) fusion constructs pMBP-NCoR1-540 and pMBP-THR␤1, respectively. DNA encoding the C-terminal 540 amino acids of NCoR1 was amplified by PCR using ORFeome clone HOC23341 (Genecopoeia, Rockville, MD; GenBank TM accession number BC172437) as template, and NCoR540 LIC forward and reverse primers. The isolated PCR fragment was cloned using LIC methodology into pMCSG9 (52), resulting in a translational fusion between MBP and the C-terminal 540 residues of NCoR1. The same approach was taken when generating an expression clone for THR␤1. The THR␤1 ORF was amplified by PCR using ORFeome clone GC-OG01669 (Genecopoeia; GenBank TM accession number BC106929) as template, THR␤1 LIC as forward primer, and THR␤1 LIC as reverse primer. The resultant product was cloned into pMCSG9 generating pMBP-THR␤1, a translational fusion between MBP and THR␤1 with an internal TEV protease cleavage site for downstream removal of MBP.
Protein Purification-Unless otherwise specified, all materials were reagent grade and were purchased from either Sigma-Aldrich or Thermo Fisher Scientific. FLRev-erb␤ MGC, HGS, and heme axial ligand mutants were purified from E. coli BL21(DE3) (Life Technologies) transformed with the appropriate plasmid as described above. The cells were grown with shaking in 2ϫ YT medium (16 g of tryptone, 10 g of yeast extract, and 5 g of NaCl/liter (Fisher Bioreagents)) supplemented with 100 g ml Ϫ1 ampicillin at 37°C. Once the culture reached an optical density at 600 nm (OD 600 ) of 1.0, the temperature was reduced to 20°C, and 30 -60 min later protein expression was induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside (IPTG, Gold Biotechnology, St. Louis, MO) and 100 M zinc acetate. The cells were grown under the same conditions for 16 h prior to harvesting by centrifugation. Cell pellets were suspended in TNG buffer (50 mM Tris, pH 8.0, 300 mM NaCl, and 10% glycerol) with 1 mM 2-mercaptoethanol, 1 mM EDTA, 0.5 mM PMSF, 0.1% Tween 20, 6 mM benzamidine, 2ϫ Complete EDTA-free protease inhibitors (Hoffmann-La Roche), 5 g ml Ϫ1 DNaseI, and 1 mg ml Ϫ1 lysozyme. The cell slurry was stirred for 20 min at 4°C prior to sonication (unless otherwise indicated, all purification steps were at 4°C). Cell lysates were cleared by centrifugation at 38,000 ϫ g for 1 h, and the soluble extract applied to a 2.5 ϫ 5-cm nickel-nitrilotriacetic acid-agarose (Qiagen, Valencia, CA) column equilibrated in TNG buffer with 1 mM 2-mercaptoethanol. The column was washed with the same buffer for 1 column volume prior to applying a linear gradient from 10 to 250 mM imidazole. Fractions containing partially pure protein verified by SDS-PAGE were pooled and the protein content determined. TEV protease (prepared as described previously (53)) was added to the pool at a mass:mass ratio of 1.5:20 TEV:GB1-FLRev-erb␤ and dialyzed against 50 mM Tris, pH 8.0, 150 mM NaCl, 10% glycerol, 5 mM sodium citrate, 5 mM DTT, and 10 M ZnCl 2 at 4°C for 16 h. The TEV digest reaction was then dialyzed against 50 mM Tris, pH 8.0, 10 mM NaCl, 10% glycerol, and 5 mM DTT at room temperature for 4 h prior to application of the pool onto a 2.5 ϫ 5-cm Q-Sepharose Fast Flow (Sigma) column equilibrated in the dialysate buffer (the remaining purification steps were performed at 4°C). The column was washed thoroughly (4 -5 column volumes) prior to the application of a linear gradient from 10 to 500 mM NaCl. Fractions containing partially pure FLRev-erb␤ free of undigested GB1-FLRev-erb␤ were identified with SDS-PAGE and pooled, and the NaCl concentration was adjusted to 0.5 M by the slow addition of solid NaCl crystals. The pool was concentrated with 10-kDa molecular weight cutoff (MWCO) Amicon centrifugal filter devices (Millipore, Billerica, MA) and then loaded onto a 1.5 ϫ 90-cm Sephacryl S100 high resolution (Sigma) gel filtration column equilibrated in TNGD buffer (50 mM Tris, pH 8.0, 300 mM NaCl, 10% glycerol, and 5 mM DTT). Fractions containing pure FLRev-erb␤ were identified with SDS-PAGE, pooled, concentrated to ϳ10 mg ml Ϫ1 with 10-kDa MWCO Amicon devices, and snap-frozen in 20 -100-l aliquots in liquid nitrogen prior to storage at Ϫ80°C. Densitometric analysis of SDS-polyacrylamide gels with ImageJ (54) indicated wild-type and variant FLRev-erb␤ preparations were Ͼ90% pure.
MBP-NCoR1-540 was purified from E. coli BL21(DE3) transformed with pMBP-NCoR1-540. Cells were grown with shaking in Terrific Broth (Fisher Bioreagents) supplemented with 0.4% glycerol and ampicillin (100 g ml Ϫ1 ) at 37°C until the culture reached an A 600 of 3.0. The temperature was lowered to 18°C for 1.5 h prior to induction by 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside, and the cells continued to grow under the same conditions for 3 h before harvesting by centrif-ugation. Cell pellets were suspended in TNGD buffer and further supplemented with 10 mM EDTA, 1ϫ protease inhibitors, 0.5 mM PMSF, 0.2% Tween 20, 6 mM benzamidine, 20 g ml Ϫ1 DNaseI, and 0.5 mg ml Ϫ1 lysozyme. The cell slurry was sonicated and a soluble extract prepared by centrifugation at 100,000 ϫ g for 45 min at 4°C. The extract was loaded onto a 2.5 ϫ 6-cm amylose column (New England Biolabs) equilibrated in TNGD, and the column was washed with 5 volumes of the same buffer before applying 5 volumes of TNGD containing 20 mM MgCl 2 and 10 mM ATP. The ATP wash was necessary to eliminate a contaminant with an approximate molecular mass of 70 kDa (likely DnaK, a common contaminant in preparations of heterologously expressed protein) (55). The column was reequilibrated in TNGD prior to elution of bound protein in the same buffer containing 20 mM maltose. The presence of partially purified MBP-NCoR1540 was confirmed with SDS-PAGE, and the elution pool was concentrated with 10-kDa MWCO Amicon devices prior to loading onto a 1.5 ϫ 90-cm Sephacryl S300 high-resolution (Sigma) column equilibrated in TNGD. Fractions containing partially pure protein were identified with SDS-PAGE, concentrated as described to ϳ20 mg ml Ϫ1 , and then aliquoted and snap-frozen in liquid nitrogen prior to storage at Ϫ80°C. Densitometric analysis of SDS-polyacrylamide gels indicated that MBP-NCoR1-540 preparations were ϳ80% pure. For control experiments utilizing MBP alone, E. coli BL21(DE3) cells transformed with pMCSG9 (encoding an N-terminally hexahistidine-tagged MBP) were grown with shaking in Lennox Broth (Fisher Bioreagents) supplemented with 200 g ml Ϫ1 ampicillin at 37°C until the culture reached an A 600 of 0.7. Expression was induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside, and the cells were incubated under the same conditions for an additional 16 h prior to harvesting by centrifugation. The pellets were suspended in 25 mM Tris, pH 8.0, 150 mM NaCl, 5% glycerol, and 5 mM DTT (0.5ϫ TNGD buffer) supplemented with 1 mM EDTA and 1ϫ protease inhibitors, sonicated, and a soluble extract prepared as described for MBP-NCoR1-540. The extracts were loaded onto a 2.5 ϫ 6-cm amylose column that had been equilibrated in 0.5ϫ TNGD, and the column was washed with 5 volumes of the same buffer before eluting bound protein with 20 mM maltose in 0.5ϫ TNGD. The presence of highly pure MBP was confirmed by SDS-PAGE, and aliquots were stored at Ϫ80°C. THR␤1 was purified from E. coli BL21(DE3) transformed with pMBP-THR␤1. The cells were grown, harvested, and lysed as described for GB1-FLRev-erb␤ expression, except that the lysis buffer for the THR␤1 cells contained 5 mM DTT instead of 2-mercaptoethanol and 1ϫ protease inhibitors, but it included all other supplements at the aforementioned concentrations. The soluble extract was loaded onto a 2.5 ϫ 10-cm amylose column that had been equilibrated in TNGD and then thoroughly washed prior to eluting the protein with 20 mM maltose. TEV protease was added to the MBP-THR␤1 pool at a ratio of 0.5:20 (w/w), and the digest was dialyzed against 50 mM Tris, pH 8.0, 15 mM NaCl, 10% glycerol, and 5 mM DTT at 4°C for 16 h. The pool was applied to a 2.5 ϫ 5-cm Q-Sepharose Fast Flow column equilibrated in the dialysate and the column washed with 2 volumes of the same buffer prior to applying a linear gradient from 15 mM to 500 mM NaCl. Fractions containing THR␤1 were confirmed by SDS-PAGE, pooled, and dialyzed against 20 mM Tris, pH 8.0, 5% glycerol, 20 mM MgCl 2 , and 5 mM DTT for 16 h at 4°C. The pool was applied to a 1.5 ϫ 1.5-cm heparin-agarose column (Sigma) equilibrated in dialysate buffer and the column washed with 10 volumes of the same buffer before applying 25 volumes of buffer containing 5 mM ATP, which effectively eliminated a ϳ70-kDa contaminant, likely DnaK. The protein was eluted from the column in the same buffer (lacking ATP) but with 1.5 M NaCl. The presence of pure THR␤1 (Ͼ85%) was confirmed with SDS-PAGE prior to buffer exchanging into 0.5ϫ TNGD by repeated concentration and dilution in 10-kDa MWCO Microcon centrifugal filter devices (Millipore) and storage of the protein at Ϫ80°C after snap freezing the aliquots in liquid nitrogen.
Analytical Gel Filtration-50-l samples of FLRev-erb␤ MGC and HGS (0.5-1 mg ml Ϫ1 ) were injected onto a Shodex KW-803 high pressure liquid chromatography (HPLC) column (8 ϫ 300 mm) attached to a Shimadzu HPLC system fitted with a diode array detector. Samples were run at 1 ml min Ϫ1 in 50 mM Bis-Tris, pH 7.0, 150 mM NaCl, 5% glycerol, and 0.5 mM TCEP. Prior to injection of FLRev-erb␤, the column was standardized with blue dextran to determine the void volume, Bio-Rad gel filtration standards (thyroglobulin, 670 kDa; ␥-globulin, 158 kDa; ovalbumin, 44 kDa; myoglobin, 17 kDa; and vitamin B 12 , 1.35 kDa), bovine serum albumin (66.4 kDa), apoferritin (443 kDa), and cytochrome c (12.4 kDa). A standard curve was generated by plotting the logarithm of the molecular masses of the standards versus elution volume/void volume (V e /V 0 ) and fitting the data with a linear line, which yielded an R 2 of 0.9905 (data not shown). The molecular masses of FLRev-erb␤ MGC and HGS were determined using the equation derived from the standards plot and V e /V 0 of the major eluting species.
Heme and CO Binding Assays-The affinity of Fe 3ϩ -heme and Fe 2ϩ -heme for FLRev-erb␤ MGC, HGS, H568F, and H568F/C384A was assessed with difference UV-visible spectrophotometry. All spectra in this study were acquired using either a Shimadzu model UV-2600 or UV-2501PC double beam spectrophotometer equipped with a TCC temperature controller set to 20°C. Prior to heme titration, proteins (100 -200 M) were thiol-reduced with 25 mM TCEP at ambient temperature for 1 h in an anaerobic chamber (Ͻ2 ppm O 2 ) prior to buffer exchanging to TNG lacking reductant using Micro Bio-Spin 6 gel filtration columns (Bio-Rad). Wild-type FLRev-erb␤ (MGC or HGS) was diluted to 0.8 M in 1 ml of 0.5ϫ TNG (Ͻ1% dilution factor), and the sample was sealed with a rubber stopper in a quartz cuvette to maintain oxygen-free conditions during the experiment carried out at the bench. A parallel cuvette containing 1 ml of buffer was used as the reference. Fe 3ϩ -and Fe 2ϩ -heme titrants were prepared by dissolving solid hemin (5 mg ml Ϫ1 ; Frontier Scientific, Logan, UT) in 0.1 M NaOH with 10% dimethyl sulfoxide (DMSO), filtering the solution, and diluting it to ϳ100 M in 0.5ϫ TNG containing 5% DMSO for Fe 3ϩ -heme. Fe 2ϩ -heme was prepared by dilution in the same buffer but with 20% DMSO (which is necessary to avoid precipitation of Fe 2ϩ -heme) followed by the addition of 10 mM dithionite. Titrant concentrations were determined by dilution in 0.1 M NaOH using ⑀ ϭ 58.44 mM Ϫ1 cm Ϫ1 at 385 nm (56) (prior to the addition of dithionite for Fe 2ϩ -heme) and were maintained under anaerobic conditions in stoppered serum vials during the titration. One-l increments of the stocks were added to both the reference and experimental cuvettes, mixed by inversion, and the spectrum monitored until equilibrium was reached; then the final spectrum was recorded. The absorbance of the Soret band maxima for each titration was plotted as a function of heme concentration, and the data were fit to Equation 1 (all fitting and statistical analysis in this study was performed with SigmaPlot software (Systat Software, Inc., San Jose, CA)), which accounts for the concentration of free ligand in equilibrium titrations and is ideal for tight-binding interactions where K d Ͻ Ͻ protein concentration; thus the [free ligand] is negligible compared with the [total ligand] (57).
In this equation, A is the observed absorbance from the experiment (but can represent any observed signal, i.e. aniosotropy), A B is the difference absorbance of fully bound protein, [P] is the protein concentration, [L] is ligand concentration (heme), and K d is the dissociation constant.
For titrations of H568F and H568F/C384A, 4 M protein was used. In addition, a high concentration of DMSO was required to stabilize concentrated (1 mM) Fe 2ϩ -heme stocks used in these titrations; thus, titrants were prepared by dissolving hemin at 5 mg ml Ϫ1 in 100% DMSO, filtering the solution, and determining the concentration of the stock by serial dilution in DMSO and ⑀ ϭ 170 mM Ϫ1 cm Ϫ1 at 406 nm (58). The stock was diluted to 1 mM in 50% DMSO for Fe 3ϩ -heme or further reduced with 10 -20 mM dithionite for Fe 2ϩ -heme. Two-l increments of the titrants were added to both the reference and experimental cuvettes, and the spectra were recorded and data analyzed as described for MGC and HGS.
To determine the affinity of CO for FLRev-erb␤ MGC/HGS , thiol-reduced proteins were diluted under anaerobic conditions to 3 M in 3 ml total volume 0.5ϫ TNG buffer containing 2 M Fe 3ϩ -heme in a 4-ml capacity quartz cuvette. The cuvette was sealed with a rubber stopper, and dithionite was added to generate the Fe 2ϩ -heme complex. One-l increments of COsaturated water were titrated into the cuvette using a gas-tight syringe, and the ensuing spectral changes were monitored until equilibrium was achieved. The concentration of CO in the titrant was determined by a second titration against the Fe 2ϩ -heme⅐myoglobin complex using the extinction coefficient for the CO complex of 207 mM Ϫ1 cm Ϫ1 at 424 nm (59).
Determination of Fe 3ϩ -Heme Off-rates with Excess Apomyoglobin-Apomyoglobin was prepared by extracting heme from equine skeletal muscle myoglobin (Sigma) using methyl ethyl ketone as described (60). Briefly, solid myoglobin was dissolved in 10 mM potassium P i , pH 7.0, and the solution was acidified to ϳpH 3 by the addition of HCl. Heme was removed with three 1:1 methyl ethyl ketone extractions on ice, thereby separating the protein-containing aqueous phase between extractions. Denatured apomyoglobin was refolded by dialysis against 10 mM potassium P i , pH 7.0, at 4°C overnight and then concentrated to ϳ2 mM (⑀ ϭ 14 mM Ϫ1 cm Ϫ1 at 280 nm based on the calculated extinction coefficient from the ExPASy server (61) and protein database accession P68082 for equine myoglobin) prior to snap freezing samples in liquid nitrogen and storage at Ϫ80°C. Before an experiment, apomyoglobin was bufferexchanged to 0.5ϫ TNG by concentration and dilution in 10-kDa MWCO Microcon devices. Heme transfer experiments were set up and performed under anaerobic conditions in cuvettes sealed with rubber septa. 0.7-ml assay volumes were used, which is sufficient to encompass the light beam of the spectrophotometer when a 1.7-ml quartz cuvette is raised by placing a flat Teflon cap in the bottom of the sample cell. Assays were prepared in 0.5ϫ TNG buffer containing 1.5 M Fe 3ϩheme and 3 M thiol-reduced wild-type FLRev-erb␤ HGS, H568F, or H568F/C384A (the maximum dilution factor of protein is Ͻ5%); an identical experiment with heme alone was performed as a control. The cuvettes were sealed and then brought out of the anaerobic chamber, and the spectrum was recorded at 20°C. To initiate the heme transfer reaction, apomyoglobin was injected (30 M final concentration, Ͻ2% dilution factor), the sample mixed by inversion (dead time from mixing was ϳ20 s), and absorbance changes at 408 nm (the position of the ␥-band maximum for holomyoglobin) monitored over time. At the conclusion of the time course, a spectrum was recorded to confirm heme transfer to apomyoglobin. The kinetic data fit best to a double exponential rate equation. All reported rate constants and ⌬absorbance % values for these experiments are averages Ϯ S.D. of three independent experiments.
Electron Paramagnetic Resonance (EPR)-As-purified FLRev-erb␤ MGC was thiol-reduced and buffer exchanged under anaerobic conditions as described above. The sample was concentrated to 97 M using 10-kDa MWCO Microcon devices, and heme was added from a concentrated stock in TNG buffer (prepared by dissolving heme at 20 mg ml Ϫ1 in 0.1 M NaOH, filtering, and then diluting it in TNG buffer, confirming the concentration as described above); this resulted in a final [heme] of 86 M and [protein] of 93 M. The sample was transferred to an EPR tube and frozen in liquid nitrogen prior to collecting spectra on an X-band Bruker EMX spectrometer (Bruker Biospin Corp., Billerica, MA) containing an Oxford ITC4 temperature controller, a Hewlett-Packard model 5340 automatic frequency counter, and a Bruker gaussmeter with the following instrument parameters: 10 K, 9.3829 GHz microwave frequency, 0.32 milliwatt power, 6.3 ϫ 10 5 receiver gain, 8.31 G modulation amplitude, 100 kHz modulation frequency, 163.84 ms time constant, and 8 total scans with 2048 points each.
Fluorescence Anisotropy-To establish a K d for the interaction of fluorescein-labeled NCoR1 ID1 (F-ID1, Life Technologies; fluorescein-RTHRLITLADHICQIITQDFARN) with FLRev-erb␤ MGC , a 10 -12.5 M stock of protein in 0.5ϫ TNG buffer with 0.5 mM TCEP was titrated into 1.5 ml of the same buffer containing 10 nM F-ID1 at 25°C. After the addition of protein, the anisotropy of F-ID1 was measured with a PTI Quantamaster fluorometer (controlled by FeliX software) by vertically polarized excitation at 495 nm (6-nm slit width) and detection of parallel and perpendicular emission intensities at 525 nm (6-nm slit width). Intensity was recorded as the average of a 60-s integration time with 1 reading s Ϫ1 . Care was taken to assure that no change in intensity occurred during the reading, indicating that equilibrium had been reached. Total fluorescence and anisotropy were calculated using Equation 2, where A is the anisotropy of F-ID1, I v and I h are the vertical and horizontal emission intensities, respectively, and G is an instrument correction factor determined by excitation of the F-ID1 solution with horizontally polarized light and calculating the ratio of the resulting emission intensities I v /I h (62). Protein titration led to a concentration-dependent quenching of F-ID1, where the total fluorescence of the system is equal to the denominator of Equation 2. Thus, the raw anisotropy data were corrected for quenching using Equation 3, as described previously (63).
Here, A corr is the corrected anisotropy, A is the measured anisotropy, A f and A b are the anisotropy of F-ID1 in the free and bound state, respectively, and Q f and Q b are the total fluorescence of free and bound F-ID1, respectively. A b is estimated from the hyperbolic fit of the raw anisotropy data, whereas Q b is determined from the hyperbolic decay fit of the total fluorescence. The corrected anisotropy data are plotted versus [FLRev-erb␤ MGC ] and fit with Equation 1, holding the [F-ID1] constant at 10 nM. The effect of Fe 3ϩ -heme on the F-ID1⅐FLRev-erb␤ MGC complex was assessed by titration of heme in 0.5ϫ TNG with 0.5 mM TCEP into 1.5 ml of the same buffer containing 10 nM F-ID1 and 25 nM FLRev-erb␤ MGC . After the addition of heme the anisotropy of F-ID1 was determined as described above and the data fit with a hyperbolic decay function to estimate IC 50 .
The DNA binding activities of FLRev-erb␤ MGC, HGS, H568F, and H568F/C384A were measured by preparing 0.15-ml assays in 0.5ϫ TNGD containing 1 ng l Ϫ1 poly(deoxyinosinic-deoxycytidylic) acid competitor DNA (poly(dI-dC), Sigma), 10 nM fluorescein amidite (FAM)-labeled Rev-DR2 or FAM-Bmal1 promoter DNA duplexes (see Table 1 for sequence; duplexes were prepared by Integrated DNA Technologies (Coralville, IA) from HPLC-purified single-stranded oligonucleotides and used without further purification), and varying concentrations of the aforementioned proteins. The assay mixtures were incubated at 4°C for 2 h prior to transferring 0.1 ml of each assay to a black polystyrene nonbinding surface 96-well plate (Corning) and reading the anisotropy of the FAM duplexes with a Tecan Safire microplate reader with the appropriate settings for fluorescein detection. Protein titration had a negligible effect on FAM duplex total fluorescence. Raw anisotropy data were plotted versus [protein] and the data fit with the Hill equation.
To test the effect of Fe 2ϩ -heme on the interaction of NCoR1 and FLRev-erb␤, we obtained a Texas Red (TR)-labeled Rev-DR2 duplex, as TR is resistant to dithionite quenching (64) (obtained from Integrated DNA Technologies and prepared as described for FAM duplexes). The K d for the interaction between MBP-NCoR1-540 and FLRev-erb␤ HGS ⅐TR-Rev-DR2 was determined by preparing 0.3-ml assay mixtures containing 1 ng l Ϫ1 poly(dI-dC), 20 nM TR-Rev-DR2, 100 nM FLRev-erb␤ HGS , and increasing concentrations of MBP-NCoR1-540 in 0.5ϫ TNGD. A parallel control experiment was performed with MBP instead of MBP-NCoR1-540. The assays were incubated for 2 h at 4°C prior to quantifying the anisotropy of TR-Rev-DR2 using the microplate method, 0.2 ml of each sample, and excitation and emission wavelengths of 590 and 630 nm, respectively. The effect of Fe 3ϩ -heme and Fe 2ϩ -heme on the interaction between MBP-NCoR1-540 and the FLRev-erb␤ HGS ⅐Rev-DR2 complex was tested using two approaches. Either Fe 3ϩ -heme was added at a stoichiometry of 1.5:1 heme: FLRev-erb␤ to the titration described above or assays containing a fixed [MBP-NCoR1-540] of 30 nM were mixed with increasing concentrations of Fe 3ϩ -heme. In both experiments, Fe 3ϩ -heme was converted to Fe 2ϩ -heme by the addition of 5 mM dithionite diluted from a freshly prepared stock of 100 mM in 100 mM Tris, pH 8.0, and the anisotropy was recorded within 10 min to avoid heme reoxidation.
Electrophoretic Mobility Shift Assays (EMSA)-20-l assays were prepared in 0.5ϫ TNGD buffer containing 25 ng l Ϫ1 poly(dI-dC), 1 M FAM-Rev-DR2 or FAM-Bmal1 promoter, and varying concentrations of wild-type and variant FLRev-erb␤, Fe 3ϩ -heme and MBP or MBP-NCoR1-540. Mixtures were incubated on ice for 2 h prior to loading 10 l of each sample in the wells of a NativePAGE Novex 4 -16% gradient polyacrylamide gel (Life Technologies) that had been prerun at 100 V for 1 h at ambient temperature in 50 mM bis-Tris and 50 mM Tricine, pH 6.8, running buffer. Samples were separated under the same conditions prior to imaging the wet gel with an Amersham Biosciences Typhoon Trioϩ using the appropriate settings for fluorescein detection.
For experiments with THR␤1, samples were prepared as described above, except they contained 0.25 mg ml Ϫ1 BSA, 1 M FAM-labeled THR␤1 response element from the human placental lactogen gene promoter (65, 66) (FAM-hPL promoter, Table 1; the duplex was prepared as described but required additional purification by gel filtration chromatography to eliminate a high molecular mass contaminant, as described in the analytical gel filtration section above), and varying concentrations of heme or T 3 (Sigma) (BSA was omitted from those assays containing heme). T 3 was prepared by dissolving the solid to 20 mg ml Ϫ1 in 0.1 M NaOH and determining the concentration with ⑀ ϭ 4.66 mM Ϫ1 cm Ϫ1 at 320 nm in 0.1 M NaOH (56). The stock was diluted in 0.5ϫ TNGD containing 1 mg ml Ϫ1 BSA prior to dilution in EMSA samples.
Cell Culture and Transfection-HEK293 cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies) with 4.5 g liter Ϫ1 glucose, 0.11 g liter Ϫ1 pyruvate, 0.584 g liter Ϫ1 glutamine, 10% fetal bovine serum, 100 units ml Ϫ1 penicillin, and 100 g ml Ϫ1 streptomycin at 37°C in a humidified incubator maintained at 5.0% CO 2 density. Prior to an experiment, confluent cells on 10-cm-diameter plates (60.1 cm 2 growth area) were harvested using 1-2 ml of 0.25% trypsin/ EDTA dissociation reagent (Life Technologies). Trypsinized cells were brought to a total volume of 10 ml with medium, the cells were pelleted at Ͻ1000 ϫ g prior to aspirating the supernatant, and the pellet was suspended in the same volume of Opti-MEM (Life Technologies) with 2.5% FBS and 4.5 g liter Ϫ1 glucose (low serum medium). Unless otherwise stated, cells were diluted 5-fold into the same medium (ϳ5 ϫ 10 6 cells/10cm-diameter plate, 10 ml total medium), which resulted in 60 -70% confluency 24 h after dilution; this ratio was used to determine the number of cells to seed into plates with different growth areas. Transfections were carried out 24 h after seeding cells using Lipofectamine LTX and Plus reagents (Life Technologies) according to manufacturer's recommendations with a DNA (g):LTX (l) ratio of 1:1.9 -2.3.
Determination of Cellular Heme Content-The concentration of heme in soluble cell extracts was determined by iron extraction with supersaturated oxalic acid and quantitation of the resulting protoporphyrins with fluorescence spectroscopy as described previously (67) with modifications. Briefly, cell extracts (50 -200 g) prepared with either reporter lysis buffer (Promega, Madison, WI) or Triton X-100 lysis buffer (20 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 1.5 mM MgCl 2 , 2 mM TCEP, and 1% Triton X-100) were diluted in supersaturated oxalic acid 1:20 in sealed glass HPLC vials, mixed, and immediately heated at 121°C for 30 min in an autoclave. A parallel set of samples was left unheated. The samples were cooled and transferred to a black microplate, and the fluorescence emission spectrum of protoporphyrin was recorded from 550 -700 nm with excitation at 400 nm in a Tecan Safire microplate reader. Heme concentration (nmol mg protein Ϫ1 ) was determined using a standard curve of hemin prepared in DMSO (the standards were treated the same as those samples containing cell extract) and quantified after subtraction of background emission (662 nm) from protoporphyrin in the unheated samples from the intensity of the heated sample at the same wavelength.
Cellular Fractionation-HEK293 cells grown in 10-cm-diameter dishes were transfected with 12.6 g of pcDNA3.1(ϩ)-FLRev-erb␤ wild type, H568F, or H568F/C384A expression vectors. After an additional 24 h incubation, the cells were washed thoroughly with PBS (1.1 mM KH 2 PO 4 , 3 mM Na 2 HPO 4 -7H 2 O, and 155 mM NaCl, pH 7.4) prior to fractionation into cytosolic and nuclear compartments as described (70). Briefly, the pelleted cells were suspended in 3-fold the pellet volume of 10 mM HEPES, pH 7.4, 0.33 M sucrose, 1 mM MgCl 2 , 0.1% Triton X-100, and 1ϫ protease inhibitors and incubated on ice for 15 min; the insoluble debris/nuclei were pelleted by centrifugation at 17,000 ϫ g for 10 min at 4°C. The soluble fraction comprising the cytosolic components was removed, and the pellet was washed thoroughly with the same buffer prior to suspension in 10 mM HEPES, pH 7.4, 0.45 M NaCl, 1 mM MgCl 2 , and 1ϫ protease inhibitors. Soluble nuclear components were extracted with frequent agitation and incubation on ice for 30 min prior to centrifuging and pelleting the remaining insoluble material at 17,000 ϫ g for 5 min at 4°C, with the soluble nuclear proteins remaining in the supernatant.
For experiments determining the effect of intracellular heme on the distribution of endogenous Rev-erb␤, rapidly growing HEK293 cells (ϳ50% confluent) in low serum medium were incubated with 1 mM succinylacetone (SA) for 42 h prior to harvesting and fractionating as described above, except HEPES was replaced with Tris buffer in the cytoplasmic extraction buffer, and the nuclear pellet was lysed in Triton X-100 lysis ORF from the Mammalian Gene Collection cloned into a modified pMCSG7 vector resulting in a translational fusion between the His 6 -GB1 solubility tag and FLRev-erb␤ with an internal TEV protease cleavage site for protein production in E. coli (this study). pGB1-FLRev-erb␤-HGS pGB1-FLRev-erb␤-MGC carrying H21P, K282Q, and R288P mutations reflecting the Human Genome Sequence; parent vector for H568F and H568F/C384A expression clones (this study). pcDNA3.1(ϩ)-FLRev-erb␤ The Rev-erb␤ ORF from pGB1-FLRev-erb␤-HGS was subcloned into pcDNA3.1(ϩ) using a forward primer containing a consensus Kozak sequence for expression in mammalian cells; parent vector for H568F, C384A, and H568F/C384A expression clones (this study). pcDNA3.1(ϩ)-Rev-erb␣ The Rev-erb␣ ORF from mammalian gene collection vector 6190140 was subcloned into pcDNA3.1(ϩ) for overexpression of Rev-erb␣ in mammalian cells; parent vector for the H602F expression clone (this study). pMBP-NCoR1-540 DNA coding for the C-terminal 540 amino acids of NCoR1 was cloned into pMCSG9 resulting in a translational fusion between MBP and NCoR1-540 with an internal TEV protease cleavage site (this study). pMBP-THR␤1 The THR␤1 ORF was cloned into pMCSG9 resulting in a translational fusion between MBP and THR␤1 with an internal TEV protease cleavage site (this study). pCMV-␤Gal Mammalian expression vector producing ␤-galactosidase for measuring transfection efficiency and normalizing luciferase activity (provided by Dr. Daniel Bochar, University of Michigan). pGL3Basic-BMal1 Luciferase reporter vector driven by the core Bmal1 promoter (encompassing Ϫ422 to ϩ108 bp with respect to the transcription start site) from M. musculus (71).
Co-immunoprecipitation and Mass Spectrometric Identification of Proteins in SDS Gels-One approach to determining the extent of endogenous NCoR1 co-immunoprecipitation with Rev-erb␤ is to make use of recombinant proteins. In those assays, HEK293 cells (Ͼ80% confluent) were washed thoroughly with PBS prior to lysing the cells with Triton X-100 lysis buffer containing 1.5ϫ protease inhibitors. For some experiments, the cells were depleted of heme by treatment with 2 mM SA for 24 h prior to harvesting. In 0.5 ml total volume of lysis buffer, 1 mg of soluble extract was mixed with 5 g of recombinant FLRev-erb␤ HGS, H568F, or H568F/C384A variants (a parallel set of control experiments contained no recombinant protein) and 0, 5, 10, or 20 M Fe 3ϩ -heme, and the mixtures were incubated on ice for 1 h prior to adding Rev-erb␤ LBD antibody to each sample (9.3 g). A second control experiment contained the same mass of nonspecific IgY antibodies added to assays with recombinant protein. In both cases, control IPs led to no detectable FLRev-erb␤ HGS or NCoR1 precipitation (data not shown). After an additional hour of incubation on ice, 40 l of PrecipHen-agarose (50% agarose slurry, Aves Labs) was added, and the samples were incubated overnight at 4°C with rocking and protected from light. The beads were washed thoroughly with Triton X-100 lysis buffer (except that the Triton concentration was 0.1%, instead of 1.0%) prior to eluting the protein complexes by boiling for 10 min in 2ϫ SDS loading buffer. Complexes were separated by SDS-PAGE and electroblotted as described, and the membranes were probed with Rev-erb␤ QK-6, NCoR1 C-20, or MycBP2 antibodies.
A second approach for assessing the role of heme in NCoR1 binding to Rev-erb␤ was to overexpress the wild-type receptor or the H568F variant via transient transfection of HEK293FT cells. 40 -50% confluent cultures in 10-cm-diameter dishes grown in low serum medium were transfected with 20 g of pcDNA3.1(ϩ) parent vector as a control, pcDNA3.1(ϩ)-FLRev-erb␤ for overexpression of wild-type Rev-erb␤, or the analogous expression vector for the H568F variant. 24 h after transfection, 20 M MG-132 (Sigma; stock solution is 10 mM in DMSO) was added to the medium, and the cells were incubated for an additional 4 h prior to washing with PBS, lysing in Triton X-100 lysis buffer containing 1.5ϫ protease inhibitors, and preparing IPs as described above for recombinant proteins.
Mass spectrometry was used to identify the proteins that interacted with FLRev-erb␤. The co-IP utilizing recombinant protein was scaled up 2-fold (in terms of recombinant protein and cell extract mass), and the eluted complexes were separated by SDS-PAGE, staining the gel with Coomassie Brilliant Blue. Protein bands of interest were excised from the gel and subjected to in-gel reduction and alkylation (DTT/iodoacetamide) followed by tryptic digestion (at 37°C for 16 h) and extraction with 5% acetonitrile in 0.1% formic acid. The samples were analyzed by capillary HPLC-electrospray ionization tandem mass spectrometry (HPLC-ESI-MS/MS) on a Thermo Scientific LTQ Orbitrap Velos Pro mass spectrometer. On-line HPLC separation was accomplished with an Eksigent/AB Sciex NanoLC-Ultra 2D HPLC system using a PicoFrit TM (New Objective) 75 m i.d. column packed to 15 cm with C18 adsorbent reverse phase medium (Vydac, 218MS 5 m, 300 Å). Precursor ions were acquired in the Orbitrap in centroid mode at 60,000 resolution (m/z 400); data-dependent collision-induced dissociation (CID) spectra of the six most intense ions in the precursor scan were acquired at the same time in the linear trap (30% normalized collision energy). Mascot (Matrix Science) was used to search the spectra against the human subset of the NCBInr database (NCBInr 20140522, Homo sapiens (human) (279,814 sequences and 14,178,194,136 residues)) concatenated with a database of common protein contaminants (contaminants 20120713 (247 sequences, 128,130 residues)). Cysteine carbamidomethylation was set as a fixed modification, and methionine oxidation and deamidation of glutamine and asparagine were considered as variable modifications; trypsin was specified as the proteolytic enzyme, with one missed cleavage allowed. A subset search of the identified proteins by X! Tandem, cross-correlation with the Mascot results, and determination of protein and peptide identity probabilities were accomplished by Scaffold (Proteome Software). The thresholds for acceptance of peptide and protein assignments in Scaffold were 95 and 99%, respectively.
Cycloheximide Degradation Assays-For those experiments measuring the half-life of overexpressed Rev-erb␤, HEK293 cells grown in 3.5-cm-diameter dishes (2 ml total volume medium) were transfected with 3 g of pcDNA3.1(ϩ)-FLRev-erb␤ or the expression vectors for H568F/C384A and H568F variants. After 24 h, the medium was replaced, and the cells were incubated for another 24 h prior to adding 20 g ml Ϫ1 cycloheximide (CHX, Santa Cruz Biotechnology; 20 mg ml Ϫ1 stocks were prepared in DMSO and used immediately). In some experiments, 20 M MG-132 was added at the same time as CHX. At the appropriate time points after CHX addition, cells were washed with PBS, lysed in radioimmune precipitation assay buffer (Life Technologies) containing 1ϫ protease inhibitors, and soluble extracts were electroblotted after SDS-PAGE separation as described, probing membranes with either Rev-erb␤ QK-6 or actin antibodies.
Repression Assays-HEK293 cells grown in 6-well plates were transfected with 0.1 g of pCMV-␤Gal control vector (provided by Dr. Daniel Bochar, University of Michigan), 0.2 g of pGL3Basic-Bmal1 luciferase reporter (71), varying masses of pcDNA3.1(ϩ)-FLRev-erb␤ or pcDNA3.1(ϩ)-Rev-erb␣ (or the corresponding vectors for expression of heme axial ligand mutants), and parent pcDNA3.1(ϩ) vector to a total mass of 3.5 g/well. 24 h after transfection, cells were washed with PBS and lysed in reporter lysis buffer with one freeze-thaw cycle, and the soluble extract was assayed for luciferase activity by mixing 20 l of extract with 100 l of luciferase substrate (Promega) and detecting luminescence with a Tecan Safire2 microplate reader over a 10-s integration time. The relative luciferase units were normalized to ␤-galactosidase activity, which was measured by the colorimetric detection of o-nitrophenol from the hydrolysis of o-nitrophenyl-␤-D-galactopyranoside (␤-galactosidase enzyme assay system, Promega).

Results
Purification of FLRev-erb␤ and Its Ability to Bind CO and Heme-Two Rev-erb␤ (NR1D2) sequences exist in GenBank TM , one corresponding to the mammalian gene collection (72) and the other to the human genome reference sequence. MGC differs from HGS by three missense mutations that result in amino acid substitutions P21H, Q282K, and P288R and may represent single nucleotide polymorphisms of NR1D2. To test whether these amino acid substitutions affect Rev-erb␤ ligand binding, both clones were heterologously expressed in E. coli as translational fusions to a His 6 -GB1 solubility tag with a TEV protease recognition sequence between the fusion partners. GB1-FLRev-erb␤ MGC and HGS were purified by nickel-nitrilotriacetic acid-agarose and digested with TEV protease to cleave the GB1 tag, and FLRev-erb␤ was isolated from the protease and undigested fusion by anion exchange chromatography. A final preparative gel filtration step resulted in preparations that were Ͼ90% pure and free of GB1 (Fig. 1, lanes 1 and 2). After these purification steps, the MGC and HGS forms of FLRev-erb␤ contained 2.2 and 2.9% bound heme, respectively, as determined using an oxalic acid fluorescence assay (average of four replicates; see "Experimental Procedures" for details). The native molecular masses of HGS and MGC (with monomeric molecular masses of 64.8 and 64.9 kDa, respectively) were estimated by analytical gel filtration chromatography to be 142 Ϯ 5 and 123 Ϯ 10 kDa, respectively (average of duplicate HPLC runs Ϯ range), indicating that FLRev-erb␤ associates into a dimeric quaternary structure (Fig. 2).
The ability of FLRev-erb␤ MGC and HGS to bind heme and CO was assessed by UV-visible and EPR spectroscopy. Because thiol oxidation has been shown to compromise the heme binding activity of the isolated Rev-erb␤ LBD (38), we chose to study the thiol-reduced form of FLRev-erb␤. Thus, prior to heme binding analyses, proteins were reduced with an excess of TCEP in an anaerobic chamber, buffer-exchanged to eliminate reductant, and maintained under anaerobic conditions throughout the assay. The UV-visible spectra of Fe 3ϩ -heme with a 2-fold excess of FLRev-erb␤ MGC or HGS (to ensure the majority of the heme was bound) were identical and exhibited features of a six-coordinate low-spin heme with thiolate ligation (Fig. 3A) (73). The spectra are characterized by a weak shoulder at 650 nm, broad ␣and ␤-bands at 577 and 543 nm, respectively, a Soret (␥ band) at 423 nm, and a ␦-feature at 358 nm. The EPR spectrum of the Fe 3ϩ -heme⅐FLRev-erb␤ MGC complex (Fig. 4) exhibits a rhombic feature with g values of 1.88, 2.28, and 2.47, similar to those reported for cystathionine-␤synthase and HRI (Table 2), which bind heme as six-coordinate complexes with His/Cys ligation (74 -77). Other minor features in the EPR spectrum are likely due to the heterogeneity of the axial cysteine sulfhydryl protonation state (73). The addition of 1 mM dithionite generates Fe 2ϩ -heme complexes (Fig. 3B) with sharper ␣and ␤-bands at 559 and 530 nm, respectively, and a 3 nm red-shifted Soret band at 426 nm, whereas Fe 2ϩ -CO adducts (Fig. 3C) have broader ␣/␤-bands (569 and 539 nm) and a 5 nm blue-shifted Soret peak at 421 nm. Further, titration of CO-saturated water into a solution of Fe 2ϩ -heme⅐FLRev-erb␤ HGS indicates that CO binds with high affinity in a 1:1 complex as described previously for the isolated LBD (Fig. 3D). Similar results were obtained with FLRev-erb MGC (data not shown). Thus, both the UV-visible and EPR spectra of heme/CO-bound MGC/HGS closely resemble those of the equivalent complexes prepared with thiol-reduced Rev-erb␤ LBD (38) ( Table 2), indicating that the LBD and FLRev-erb␤ are competent to bind both redox forms of heme and CO and that additional domains provided by the full-length constructs do not impart unique heme binding characteristics.
The crystal structure of the Rev-erb␤ LBD in complex with Fe 3ϩ -heme indicates that His-568 and Cys-384 serve as axial ligands coordinating the heme iron atom (36). In an attempt to generate a form of FLRev-erb␤ that is deficient in heme binding, the H568F and H568F/C384A variants were purified (Fig.  1, lanes 3 and 4) and their heme binding properties characterized. The UV-visible spectrum of Fe 3ϩ -heme in the presence of a 2-fold excess of H568F is distinct from those of the wild-type proteins, most notably by the low relative ratio of the ␥-band absorbance (422 nm) to that of the pronounced ␦-band at 370 nm (␥ 422 /␦ 370 ϭ 0.92; wild-type HGS, ␥ 423 /␦ 358 ϭ 2.43 (Fig.  3A)). This spectrum closely resembles that of an HRI truncation mutant that lacks the His axial ligand and binds Fe 3ϩ -heme with Cys/H 2 O ligation (75). The Fe 3ϩ -heme⅐H568F/C384A spectrum is similar to that of H568F, with a Soret band at 423 nm and slightly blue-shifted ␦-band at 366 nm, with the ␥ 423 / ␦ 366 ratio equal to 1.36. Fe 2ϩ -heme complexes of H568F and H568F/C384A resemble those of HGS and MGC with overlapping ␣/␤-bands and slightly blue-shifted Soret bands at 424 nm; a unique feature of the variants is a weak shoulder at 455 nm (Fig. 3B). Similarly, the spectra of the Fe 2ϩ -heme⅐CO complexes with the mutants are nearly identical to those of the wild-type proteins (Fig. 3C).
Having established that wild-type and variant FLRev-erb␤ are competent to bind Fe 3ϩ /Fe 2ϩ -heme, it was imperative to measure the K d values for the heme⅐FLRev-erb␤ complex to determine whether the full-length proteins could bind heme at the low intracellular concentrations of free/exchangeable heme, which has been estimated to be Ͻ0.2 M in mature eryth-   rocytes (78,79). Difference UV-visible spectrophotometric titrations of Fe 3ϩ -heme into anaerobic solutions of thiol-reduced MGC (Fig. 5A) and HGS establish K d values of 7 Ϯ 2 and 6 Ϯ 2 nM, respectively, and titration of Fe 2ϩ -heme (Fig. 5C for MGC) yields K d values of 24 Ϯ 4 and 23 Ϯ 5 nM, respectively. All of these values are well below 0.2 M, suggesting that FLRev-erb␤ tightly binds both redox forms of heme in the cell. Furthermore, the identical K d values shared between HGS and MGC indicate that the minor differences in primary amino sequence between isoforms have no influence on heme binding capacity. The high concentrations of heme required for titration of the H568F and H568F/C384A variants lead to precipitation; thus, only data collected prior to the observation of protein precipitation were included in the fits. Nonetheless, in stark contrast to the wild-type proteins, Fe 3ϩ -heme titrations of H568F and H568F/C384A (Fig. 5B) yield K d values of 23 Ϯ 3 and 14 Ϯ 1 M, respectively, with similar affinities observed for Fe 2ϩ -heme titrations (see Fig. 5D for the double mutant) of 18 Ϯ 2 and 9 Ϯ 1 M. Considering that the affinity of the mutants for heme is ϳ1000-fold lower than wild-type and K d values are ϳ50-fold higher than the estimated concentration of the free heme pool, they will exist predominantly as apoprotein in the cell.
The K d for wild-type proteins were determined by using a quadratic equation (Equation 1), instead of a simple Michaelis-Menten-type binding equation, to fit the data. This is because the K d values are so far below the protein concentration that the concentration of free heme (the x axis in the simple ligandbinding isotherm) is negligible compared with that of total heme. In some cases, the affinity is so high that even the quadratic treatment does not provide correct K d values. For example, a recent report from our laboratory showed that the K d for Fe 3ϩ -heme binding to heme oxygenase-2 determined with stoichiometric equilibrium titrations was 70-fold higher (14 nM) than the value (0.2 nM) calculated from heme on-and off-rates (45). Thus, although Equation 1 is useful for fitting data where the K d Ͻ Ͻ the protein concentration, it may still overestimate the true K d , suggesting that those low nanomolar values reported herein for FLRev-erb␤ may represent an upper limit of K d .
For the reasons just discussed, we measured the Fe 3ϩ -heme off-rates, the controlling factor in distinguishing heme affinity among a variety of heme proteins (49,75). We mixed FLRev-erb␤⅐Fe 3ϩ -heme complexes with 10-fold excess apomyoglobin and monitored the conversion of apo-to holomyoglobin at 408 nm ( Fig. 6 and Table 3). In a control experiment, when we reacted excess apomyoglobin (30 M) with 1.5 M free Fe 3ϩheme (Fig. 6D), complete conversion to holomyoglobin occurred within the mixing dead time of ϳ20 s. Mixing excess apomyoglobin with a solution of 1.5 M Fe 3ϩ -heme and 3 M FLRev-erb␤ HGS leads to very slow increases in absorbance at 408 nm (Fig. 6A) that fit best to a double exponential equation, yielding off-rate constants of k off-fast ϭ (1.5 Ϯ 0.4) ϫ 10 Ϫ4 s Ϫ1 and k off-slow ϭ (3.9 Ϯ 0.8) ϫ 10 Ϫ6 s Ϫ1 . The fast phase makes up 18 Ϯ 3% of the total ⌬absorbance, indicating that the second slower phase is representative of the majority of the FLRev-erb␤ HGS ⅐heme population. Upon completion of the kinetics experiment at 4000 min, the ␥-band is centered at 415 nm, approximately halfway between the positions of holomyoglobin (408 nm) and holoHGS (423 nm) (Fig. 6A), revealing incomplete heme transfer. In addition, the total ⌬absorbance for both fast and slow phases (ϳ0.05) is only ϳ50% of the expected total (0.10 -0.11) based on assays with heme axial ligand variants (discussed below). To ensure that apomyoglobin could still bind heme after the long incubation of the assay, 6 M Fe 3ϩheme was added to the cuvette, yielding a spectrum characteristic of holomyoglobin (Fig. 6A, blue spectrum; note the ␥-band is sharply centered at 408 nm). However, FLRev-erb␤ did not bind any added heme. An identical experiment with the Rev-erb␤ LBD alone also exhibits nearly identical biphasic off-rates but with heme transfer proceeding to near completion, suggest-  JANUARY 29, 2016 • VOLUME 291 • NUMBER 5

JOURNAL OF BIOLOGICAL CHEMISTRY 2207
ing that the FLRev-erb␤ undergoes denaturation during the long incubation times of the assay.
An identical experiment performed with H568F demonstrates that heme transfer to apomyoglobin reaches completion within 10 min (Fig. 6B). These data also fit best to a double exponential equation yielding off-rate constants of k off-fast ϭ (2.1 Ϯ 0.4) ϫ 10 Ϫ2 s Ϫ1 and k off-slow ϭ (5 Ϯ 1) ϫ 10 Ϫ3 s Ϫ1 , with the first phase representing the majority of the total ⌬absorbance at 64 Ϯ 9%. The spectrum acquired at the end of the kinetics experiment indicates complete heme transfer by the sharp ␥-band centered at 408 nm (Fig. 6B), essentially identical to that spectrum acquired after mixing free Fe 3ϩ -heme with apomyoglobin (Fig. 6D). Complete heme transfer from the H568F/C384A variant also occurred within 10 min, yielding rate constants of k off-fast ϭ (4.3 Ϯ 0.3) ϫ 10 Ϫ2 s Ϫ1 and k off-slow ϭ (2.4 Ϯ 1.0) ϫ 10 Ϫ3 s Ϫ1 , with the fast phase comprising the majority (88 Ϯ 1%) of the total ⌬absorbance (Fig. 6C). These data demonstrate that k off-slow for HGS is ϳ600 times slower than the slowest phase of the heme axial ligand mutants (k off-slow for H568F/C384A), indicating that off-rates are responsible for the vast differences in K d values observed in equilibrium titrations for wild-type FLRev-erb␤ versus H568F and H568F/C384A.
FLRev-erb␤ and Its Interactions with DNA and NCoR1-To ensure that FLRev-erb␤ MGC, HGS, H568F, and H568F/ C384A variants are competent to bind DNA, fluorescence anisotropy was used to measure the affinity of these proteins for two fluorescein-labeled DNA duplexes, which serve as Rev-erb response elements, termed Rev-REs (Table 1 and Fig. 7). FAM-Rev-DR2, is a synthetic construct comprised of two tandem Rev-REs separated by 2 bp, whereas the FAM-Bmal1 promoter encompasses the two Rev-REs found in the core Bmal1 promoter separated by 30 bp (from the 3Ј-end of the distal Rev-RE to the 5Ј-start of the proximal hexameric Rev-RE) (13).
Increasing concentrations of recombinant HGS, MGC, H568F, and H568F/C384A correspondingly lead to increased anisotropy of both the FAM-Rev-DR2 and FAM-Bmal1 promoter elements (Fig. 7, A and B). The data fit best to the Hill  H568F (B), and H568F/C384A (C) under anaerobic conditions, and the spectrum was recorded (black spectrum). 30 M apomyoglobin was added, and absorbance changes at 408 nm were monitored over time (insets: 2-s intervals for axial ligand mutants and 60 s for the wild type; raw data are represented as black circles). The kinetic data fit best to a double exponential (red fit, inset) yielding rate constants for the fast and slow phases (see Table 3). At the conclusion of the time course, a spectrum was recorded to confirm the extent of heme transfer to apomyoglobin (red spectrum). Note the incomplete transfer from wild-type FLRev-erb␤ to apomyoglobin by the position of the Soret band maximum at 415 nm in the spectrum taken at 4000 min after the addition of apomyoglobin (A). Further, to ensure apomyoglobin was still competent to bind heme after the long time frame of the experiment in A, 6 M Fe 3ϩ -heme was added yielding the blue spectrum characteristic of holomyogobin with a Soret band sharply centered at 408 nm. An identical control experiment performed with 1.5 M free Fe 3ϩ -heme (D) demonstrates heme binding to apomyoglobin goes to completion within the mixing dead time of ϳ20 s. A spectrum of 30 M apomyoglobin is included as a reference (D, blue spectrum).

TABLE 3 Fe 3؉ -heme dissociation rate constants for FLRev-erb␤ HGS , H568F, and H568F/C384A
Rate constants are derived from double exponential fits of the raw data (see Fig. 6).  Fig. 7C). In agreement with prior studies, Rev-erb NRs bind cooperatively as a homodimer to the Rev-DR2 construct (80 -82). However, our results show that cooperative interactions occur even when the spacing between the Rev-REs is much longer. The affinities of H568F and H568F/C384A for FAM-Rev-DR2 and FAM-Bmal1 promoter were ϳ2and 1.5-fold lower, respectively, than those of the wild-type proteins, suggesting that mutations in the LBD may affect DNA binding. However, these minor changes in affinity could be due to vari-able levels of Zn 2ϩ loading of the two 4-Cys zinc-fingers in the DBDs of the recombinant proteins. In any case, wild-type and heme axial ligand mutants bind Rev-REs with high affinity, a result supported by an EMSA demonstrating that both H568F and H568F/C384A variants bind FAM-Rev-DR2 and FAM-Bmal1 promoter elements with similar capacity to wild-type FLRev-erb␤ MGC (Fig. 8, compare odd numbered lanes).

Protein
The interaction of FLRev-erb␤ MGC with a fluorescein-labeled peptide corresponding to NCoR1 ID1 (F-ID1) was assessed with fluorescence anisotropy. Titration of heme-free FLRev-erb␤ MGC into a solution containing 10 nM F-ID1 led to a concentration-dependent increase in anisotropy that was accompanied by quenching of peptide fluorescence (Fig. 9A). The anisotropy data were corrected for the quenching effect and fit well to Equation 1 yielding a K d of 15 Ϯ 11 nM (average Ϯ S.D. of three independent titrations), thus indicating that recombinant FLRev-erb␤ is competent to bind NCoR1 with high affinity in the absence of heme. In fact, FLRev-erb␤ exhibits a higher affinity for this NCoR1 peptide than that of the isolated allophycocyanin-labeled apoRev-erb␣ LBD for a europium chelate-labeled NCoR1 ID1 peptide (K d of 135 nM) determined with fluorescence energy transfer (27). Thus, the above experiments clearly demonstrate that FLRev-erb␤ binds heme, CO, DNA, and an NCoR1 peptide with a high affinity similar to the isolated domains.

Binding Studies Using Purified Components Showing That Heme Does Not Promote the Interaction of FLRev-erb␤ with
NCoR1-Because heme titration has been shown to lead to dissociation of the complex between NCoR1 peptides and Rev-erb␣/␤ LBDs, we performed a similar experiment with the FLRev-erb␤ MGC ⅐F-ID1 complex. As observed with the LBD, the addition of Fe 3ϩ -heme destabilizes the complex between ID1 and the full-length NR with an IC 50 of 124 nM (average IC 50 of two independent titrations) (Fig. 9B). We were unable to test the effect of Fe 2ϩ -heme in this system, as F-ID1 fluorescence was paradoxically quenched under anaerobic conditions. Nonetheless, we can unambiguously conclude that additional domains provided by FLRev-erb␤ do not impart Fe 3ϩ -hemedependent stimulation of NCoR1 peptide binding. Thus, we developed a new hypothesis that, because the two N-terminal NCoR1 interaction domains are critical for an interaction with Rev-erb␣ in the cell (83), the disconnect between in vivo and in vitro results originates from the use of a peptide encompassing only one of three NR interaction domains found in NCoR1.
To mimic the interactions of Rev-erb␤ with full-length NCoR1, we generated a translational fusion between MBP and the C-terminal 540 residues of NCoR1 encompassing all three interaction domains (MBP-NCoR1-540). The construct was heterologously expressed in E. coli and purified to ϳ80% homogeneity by successive chromatography on amylose affinity resin and Sephacryl S300 gel filtration medium (Fig. 1, lane 5). Using EMSAs with native gradient polyacrylamide gels, we resolved the ternary complexes among the FAM-Bmal1 promoter (or the FAM-Rev-DR2 elements), FLRev-erb␤, and MBP-NCoR1-540, ultimately imaging the protein-DNA complexes using the fluorescein tags. The interaction of FLRev-erb␤ MGC with FAM-Bmal1 promoter reveals at least three protein-DNA complexes (Fig. 10A, lanes 1-4), likely representing a mixed population of MGC monomers and dimers binding to the two distant Rev-REs or a cooperative complex where Rev-erb␤ binding bends the DNA to allow an interaction between the two sites. The addition of MBP-NCoR1-540 causes a supershift in the MGC⅐FAM-Bmal1 promoter complex, resolving at least four distinct corepressor complexes (Fig. 10A, lanes 5-8). MBP alone does not affect migration of the MGC-promoter complex (Fig. 10A, lane 9), indicating that the supershift is the result of a specific interaction between the C-terminal 540 residues of NCoR1 and FLRev-erb␤. Because the purified receptors are predominantly heme-free, these results indicate that heme is not required for the interaction between NCoR1 and FLRev-erb␤.
To further address the effect of heme on the interaction between Rev-erb␤ and NCoR1, we added Fe 3ϩ -heme at a 1:1 stoichiometry (with respect to the concentration of FLRev-erb␤) to the assay mixture and found no effect on this interaction at different concentrations of corepressor (Fig. 10A, compare lanes 6 and 8 (heme-free samples) with lanes 10 and 12 (same samples with 1:1 heme)). An identical experiment with FLRev-erb␤ HGS yielded the same results (Fig. 10B). High levels of Fe 3ϩ -heme (10:1, heme:FLRev-erb␤ ratio) decreased the overall intensity of the protein-DNA complexes, suggesting that heme at high concentrations may inhibit DNA binding (Fig. 10A, lanes 11 and 13). An alternative interpretation is that heme induces DNA degradation, perhaps by catalyzing the generation of reactive oxygen species by Fenton chemistry. Similar experiments show that the H568F and H568F/C384A variants bind MBP-NCoR1-540 to the extent of the wild type with identical supershifting patterns on both the FAM-Bmal1 promoter and FAM-Rev-DR2 elements (Fig. 8, compare even numbered  lanes).
FLRev-erb␤ MGC and HGS binding to FAM-Rev-DR2 reveals an intense band likely representing the homodimer (Fig  10, C and D, lanes 1-4). MBP-NCoR1-540 supershifts the protein-DNA complex (Fig 10, C and D, lanes 5-8) and a 1:1 heme: FLRev-erb␤ ratio has no effect on the Rev-erb␤-NCoR1 interaction at two concentrations of corepressor (Fig 10, C and D,  lanes 10 and 12), although we do observe a low intensity supershifted band at elevated heme concentrations (10:1 heme:FL-Rev-erb␤ ratio) and an intermediate concentration of MBP-NCoR1-540 (Fig 10, C and D, lane 11). However, it is unclear if this supershift is due to heme promoting the interaction between Rev-erb␤ and NCoR1 or simply to an effect of heme on the electrophoretic mobility of the heme-bound complex. If heme stimulated the interaction we would expect this band to be at least the same intensity as the complex in Fig 10, C  and D, lane 10, with 1:1 heme, supporting the notion that  heme impacts electrophoretic mobility. This phenomenon is restricted to the Rev-DR2 element, as we did not see pronounced heme-dependent supershifting of bands in the experiments with the FAM-Bmal1 promoter.
Because of these apparent minor effects of Fe 3ϩ -heme on Rev-erb␤ and MBP-NCoR1-540 binding, it was important to quantitatively assess how much heme is bound to FLRev-erb␤ under the EMSA conditions and to what extent MBP-NCoR1-540 affects heme binding. The concentrations of the proteins and DNA described for the assay mixtures in Fig. 10D, lanes 4 (FLRev-erb␤ HGS ⅐Rev-DR2 complex) and 8 (FLRev-erb␤ HGS ⅐ Rev-DR2 complex bound by MBP-NCoR1-540), were scaled up 3-fold and included 3 M Fe 3ϩ -heme, i.e. retaining a 1:1 heme:FLRev-erb␤ ratio. Parallel assays were performed without heme, and the UV-visible spectra were recorded for all of the samples. The spectra of heme-free samples were subtracted from those of the heme-containing samples to yield the difference spectra shown in Fig. 11A. Then, 5 mM dithionite was added to the samples to generate the Fe 2ϩ -heme complexes, and the UV-visible spectra were recorded after reaching equilibrium (Fig. 11B). For both the Fe 3ϩ -heme and Fe 2ϩ -heme complexes, the Soret absorbance of samples containing the corepressor is slightly (ϳ10%) lower than that of the sample lacking MBP-NCoR1-540 (also observed in an independent experiment). If heme displaces the ID1 corepressor peptide, excess corepressor should reciprocally displace heme. To assess the extent of heme loading in each sample, extinction coefficients were calculated for Fe 3ϩ -heme and Fe 2ϩ -heme complexes with FLRev-erb␤ HGS/MGC based on those spectra acquired in Fig. 3, A and B, with excess protein-heme. For Fe 3ϩheme, we took into account the minor absorbance at 423 nm from residual heme in the as purified protein (Fig. 3A, dashed  black trace), which yielded ⑀ ϭ 90 mM Ϫ1 cm Ϫ1 at 423 nm. This value is in good agreement with the estimated extinction coef-ficient for the Fe 3ϩ -heme complex with thiol-reduced Rev-erb␤ LBD of ⑀ ϭ 79 mM Ϫ1 cm Ϫ1 at 424 nm (based on the spectrum described in Ref. 38). For the Fe 2ϩ -heme complexes, ⑀ ϭ 127 mM Ϫ1 cm Ϫ1 and 130 mM Ϫ1 cm Ϫ1 at 426 nm for FLRev-erb␤ MGC and FLRev-erb␤ HGS , respectively. Based on these extinction coefficients and those spectra shown in Fig. 11A, the concentration of Fe 3ϩ -heme bound to FLRev-erb␤ HGS in complex with Rev-DR2 and MBP-NCoR1-540 is 2.6 M, indicating 87% heme loading based on the total [FLRev-erb␤ HGS ] of 3 M in the assay; the sample lacking corepressor contains 2.8 M bound heme or 93% holoprotein. Reduction to Fe 2ϩ -heme uniformly decreased the levels of heme loading (Fig. 11B); the sample containing MBP-NCoR1-540 is loaded with 2.2 M heme or 73% holoprotein, whereas the sample without corepressor is loaded with 2.4 M heme, i.e. 80% holoprotein. In either case, these data demonstrate that FLRev-erb␤ binds both redox forms of heme under the EMSA conditions and that corepressor binding has only a minor effect on the level of heme loading.
Thus, our results support a model in which heme, Rev-erb␤, and NCoR1 form a ternary complex in which stoichiometric heme does not affect corepressor binding. As a positive control and to ensure that the EMSA technique is sensitive enough to detect ligand-dependent changes in NCoR1 binding to a NR, we tested the effect of T 3 on the interaction of MBP-NCoR1-540 with purified THR␤1. Previous EMSA experiments have shown that T 3 binding to THR␤1 decreases the affinity of the receptor for NCoR1 and DNA (84). Full-length THR␤1 was heterologously expressed in E. coli as a fusion with MBP, and the protein was purified with amylose affinity resin. After TEV digestion and successive chromatography on anion exchange and heparin-agarose, tagless THR␤1 preparations were Ͼ85% pure (Fig. 1, lane 7). For DNA we use a FAM-labeled duplex corresponding to a region of the human placental lactogen gene promoter that is composed of two tandem THR␤1 response elements separated by 4 base pairs (FAM-hPL promoter; see Table 1) (65,66). THR␤1 interacts with the FAM-hPL promoter yielding two distinct complexes likely representing the receptor binding as a monomer or homodimer (Fig. 12A, lanes  1-4). MBP-NCoR1-540, but not MBP alone, supershifts the protein-DNA complexes (Fig. 12A, lanes 5-9), and inclusion of 1:1 T 3 :THR␤1 decreases the intensity of the corepressor complex by 65% (determined with densitometry) with a concomitant increase in the intensity of the apparent monomer-DNA complex (Fig. 12A, lane 10). Higher concentrations of T 3 did not lead to increased dissociation of MBP-NCoR1-540, indicating that the system had reached equilibrium at stoichiometric concentrations of T 3 . These data demonstrate that our EMSA with purified components could distinguish ligand-dependent changes in corepressor affinity.
To further assess whether heme has a specific effect on the interactions between FLRev-erb␤ and NCoR1, we performed heme titrations with the THR␤1 system, which is responsive to T 3 but not to heme. Intriguingly, as seen with FLRev-erb␤, high levels of Fe 3ϩ -heme cause dissociation of THR␤1 from the FAM-hPL promoter without effecting corepressor binding at two concentrations of MBP-NCoR1-540 (Fig. 12B, compare lanes 2 and 5 (heme-free samples) with lanes 4 and 7 (the same samples with a 10:1 ratio of heme:THR␤1). On the other hand, there was little effect of a 1:1 ratio of heme:THR␤1 on band intensities (Fig. 12B, lanes 3 and 6). This result suggests that, at high concentrations, free heme may inhibit the binding of NRs to DNA, perhaps because of some Fenton chemistry-related reactions.
To obtain a more quantitative assessment of the interaction between MBP-NCoR1-540 and DNA-bound FLRev-erb␤ HGS and the effect of Fe 3ϩ /Fe 2ϩ -heme on the complex, we developed a fluorescence anisotropy assay. We obtained a Texas Red (TR)-labeled Rev-DR2 element because, unlike other fluorescence probes, TR is resistant to dithionite quenching (64), which allowed us to determine the effect of Fe 2ϩ -heme on the system. When increasing concentrations of MBP-NCoR1-540 were mixed with 20 nM TR-Rev-DR2 and 100 nM HGS, we observed increases in anisotropy that could be ascribed to MBP-NCoR1-540 binding to the binary complex between FLRev-erb␤ HGS and TR-Rev-DR2. The effect was saturable with corepressor concentration, and the addition of MBP alone yielded no change in anisotropy (Fig. 13A). A titration of 20 nM TR-Rev-DR2 with FLRev-erb␤ HGS yielded a K 0.5 of 83 Ϯ 1 nM and a Hill coefficient of 1.6 Ϯ 0.4, demonstrating that a mixture of 20 nM TR-Rev-DR2 and 100 nM HGS would contain 11.6 nM protein-DNA complex. Thus, we fit the MBP-NCoR1-540 titration data with Equation 1 holding the [P] term (in this case, [P] ϭ binary complex of TR-Rev-DR2 and HGS) constant at 11.6 nM, yielding a K d of 35 Ϯ 4 nM. We caution that MBP-NCoR1-540 preparations were only ϳ80% pure, thus the concentrations displayed in Fig. 13 are best estimates based on a determination with the Bradford assay. Nonetheless, the measured K d is similar to that for the F-ID1 peptide (15 Ϯ 11 nM), verifying this value and suggesting that studies using the ID1 peptide accurately reflect interactions with the full-length protein. Inclusion of Fe 3ϩ /Fe 2ϩ -heme at a 1.5:1 heme:HGS ratio did not affect the K d (Fig. 13A). Another approach to assay the effect of heme is to hold the concentration of MBP-NCoR1-540 constant at 30 nM (ϳK d ) and increase the [Fe 3ϩ /Fe 2ϩheme] (Fig. 13B). Both redox forms of heme cause a decrease in anisotropy at high concentrations, with no effect at concentrations close to that of FLRev-erb␤ (100 nM). Conversion to Fe 2ϩheme by the addition of dithionite leads to Ͻ10% quenching of TR-Rev-DR2 fluorescence, likely accounting for the uniform decrease in anisotropy of these samples (Fig. 13B). Fitting the data with a four-parameter logistics curve generates IC 50 values for Fe 3ϩ /Fe 2ϩ -heme of 1.0 Ϯ 0.2 and 1.0 Ϯ 0.1 M, respectively. Comparing these data with the anisotropy values of free 20 nM TR-Rev-DR2, 20 nM TR-Rev-DR2 plus 100 nM FLRev-erb␤ HGS , and the same mixture but with saturating levels of MBP-NCoR1-540 (400 nM) shows that heme titration decreases the anisotropy to the level of free DNA, suggesting that HGS dissociates from TR-Rev-DR2 at high heme concentrations (Fig.  13C). These data agree with our EMSA results in which high levels of Fe 3ϩ -heme decrease the intensity of bands correlating to protein-DNA complexes. FIGURE 11. Rev-DR2⅐FLRev-erb␤ HGS binds Fe 3؉ /Fe 2؉ -heme in a ternary complex with MBP-NCoR1-540. To test whether heme and NCoR1 can bind simultaneously to the FLRev-erb␤ HGS ⅐Rev-DR2 complex, the assay mixtures described in Fig. 10D, lane 4 (FLRev-erb␤ HGS ⅐Rev-DR2 complex, black spectra) and lane 8 (FLRev-erb␤ HGS ⅐Rev-DR2 complex fully bound by MBP-NCoR1-540, red spectra), were scaled-up 3-fold and included 3 M Fe 3ϩ -heme, i.e. a 1:1 ratio of heme to FLRev-erb␤ HGS . Unlabeled Rev-DR2 duplex was used to avoid overlap of the fluorescein spectrum with that of heme. A parallel set of assays was prepared without heme. Samples were incubated on ice for at least 2 h prior to recording the UV-visible spectrum. The spectra of heme-free samples were subtracted from heme-containing samples, generating the difference spectra shown in A. 5 mM dithionite was added to the samples in A under anaerobic conditions, the cuvettes were sealed, and the UV-visible spectrum was recorded until no change was observed. Difference spectra of Fe 2ϩ -heme complexes are depicted in B.
In summary, the data presented in this section clearly demonstrate that heme and NCoR1 can bind simultaneously to Rev-erb␤. Furthermore, at stoichiometric levels, the binding of heme has no impact on the affinity of the receptor for NCoR1; only high [heme] causes the breakdown of the NCoR1⅐Rev-erb␤⅐DNA complex.

An Unknown Factor Present in Cell Extracts Mediates the Heme-dependent Interaction of Full-length NCoR1 with
FLRev-erb␤-To extend our results to a system with full-length NCoR1 and to assess whether there is a missing cellular component that might induce the heme-dependent regulation of this interaction, we tested the effect of heme binding on the ability of FLRev-erb␤ to interact with endogenous NCoR1 from HEK293 cell extracts in a co-IP experiment. Initially we were unable to detect co-precipitation of NCoR1 with endogenous Rev-erb␤. However, the levels of the endogenous receptor in HEK293 cells are low, i.e. Ͻ5 ng/100 g of cell extract or Ͻ0.005% (based on comparing the relative intensities of bands for endogenous Rev-erb␤ and known masses of recombinant full-length protein; data not shown). The addition of recombinant FLRev-erb␤ HGS to heme-depleted HEK293 extracts at a 100-fold excess to endogenous protein co-precipitated appreciable levels of NCoR1; however, this was unaffected by heme (Fig. 14A), suggesting that the interaction with NCoR1 in this assay is dependent on the concentration of Rev-erb␤, requiring high levels of the receptor to perhaps outcompete other NCoR1-binding proteins. Although FLRev-erb␤ HGS interacts with NCoR1 in heme-depleted extracts, the H568F and H568F/ C384A variants paradoxically failed to interact with the corepressor (Fig. 14B). This result is especially puzzling considering that both recombinant H568F and H568F/C384A interacted with MBP-NCoR1-540 to the same extent as wild-type protein in our EMSA (Fig. 8). Additionally, it suggests that the low levels of heme present in SA-depleted extracts were sufficient to bind excess apoFLRev-erb␤ HGS and impart the heme-dependent NCoR1 interaction.
To ensure that the use of recombinant proteins in the aforementioned IPs did not lead to artifactual data, we carried out a similar IP but overexpressed wild-type and H568F Rev-erb␤ in HEK293FT cells. In addition, the proteasome inhibitor MG-132 was added to cells prior to lysis to increase Rev-erb␤ levels (later we describe that MG-132 rescues Rev-erb␤ from degradation, implicating the proteasome in Rev-erb␤ turnover). In accord with the results using recombinant H568F, wild-type Rev-erb␤ immunoprecipitated higher levels of NCoR1 than did the H568F variant (Fig. 14C). Of note, when comparing the relative intensities of the bands for overexpressed wild-type and H568F Rev-erb␤ in extracts versus 50 ng of recombinant protein, we can estimate that the levels of Rev-erb␤ in these cells is also Ͼ100-fold higher than the endogenous protein. In any case, these data disagree with our binding studies carried out with purified components, suggesting that another cellular factor in HEK293 extracts mediates the apparent heme-dependent NCoR1 interaction with Rev-erb␤ under these conditions, including high receptor concentrations and high detergent (from the cell lysis buffer) levels.
In an attempt to identify the component that mediates the heme-dependent recruitment of NCoR1 to Rev-erb␤, the co-IP experiment described in Fig. 14B was scaled up 2-fold, and the precipitated complexes were separated by SDS-PAGE (Fig.  14D). At least one high molecular mass band of Ͼ250 kDa appeared to bind uniquely to the wild-type protein versus H568F/C384A. The portion of the gel encompassing this region was excised from both the wild-type and H568F/C384A lanes, and the proteins in both gel slabs were identified by mass spectrometry. As expected, NCoR1 was enriched in the gel slab from the wild-type co-IP over that of H568F/C384A, validating the assay (Fig. 14E). Interestingly, we also identified MycBP2 associated with both wild-type and H568F/C384A proteins (Fig. 14, D and E). A Western blot confirmed that 2.6 Ϯ 1.0-fold (average of three replicate experiments Ϯ S.D.) more MycBP2 co-precipitated with wild-type Rev-erb␤ versus H568F/C384A ( Fig. 14F), indicating that the interaction is modestly heme-dependent but not to the same degree as NCoR1. MycBP2 is an E3 ubiquitin-ligase that has been identified previously as interacting with and facilitating the degradation of Rev-erb␣; however, in that study (94) it was shown that heme had no impact on the apparent rate of Rev-erb␣ degradation. Nonetheless, we chose to investigate whether heme had any influence on regulating Rev-erb␤ degradation.
Heme Binding Inversely Controls Rev-erb␤ Protein Levels, Likely through a Proteasome-dependent Pathway-Rev-erb␣ is a critical regulator of adipogenesis where its degradation is important for potentiating differentiation. Given that heme leads to a decrease in the steady-state level of Rev-erb␣ (85, 86), we tested whether heme binding affects the stability of Rev-erb␤. Indeed, steady-state levels of Rev-erb␤ H568F are 5.7-fold higher than wild-type protein in transiently transfected HEK293 cells (Fig. 15, A and C). Likewise, the analogous H602F variant of Rev-erb␣ is at a 1.9-fold higher level than the wild type, indicating that heme promotes degradation of both Reverb isoforms; however this effect appears to be more pronounced for Rev-erb␤ (Fig. 15, B and C). Similarly, the half-life of the H568F and H568F/C384A Rev-erb␤ variants in CHXtreated cells increased ϳ2-fold versus the wild type (Fig. 15, D  and E). The addition of MG-132, a proteasome inhibitor, completely reversed the effects of CHX on the wild-type protein (Fig. 15, F and G), demonstrating that the proteasome is responsible for degradation of Rev-erb␤, as suggested for Rev-erb␣ (85). Thus, heme binding inversely controls the concentration of Rev-erb␤ in the cell, likely by targeting the protein for proteasomal degradation. It remains unclear what aspect of this process is regulated by heme (perhaps the rate of ubiquitination, likely through MycBP2, or some other facet of the degradation pathway). In any case, it is tempting to speculate that whatever the degradation process regulated by heme, it may also play a role in mediating the interaction between Rev-erb␤ and NCoR1 as observed in our co-IP.
Heme Has a Minor Effect on Rev-erb␤ Subcellular Localization-Given that heme binding regulates Rev-erb␤ protein levels in the cell, we were curious as to whether the accumulation of apoRev-erb␤ was specific to the cytosol or nucleus. Further, although Rev-erb␣/␤ have been shown to be predominantly nuclear proteins (87), the effect of heme on Rev-erb␤ subcellular localization has not been tested. We transiently transfected HEK293 cells with expression vectors for the wildtype, H568F, and H568F/C384A forms of Rev-erb␤, fractionated the cells into cytosolic and nuclear compartments, and determined the relative amounts of Rev-erb␤ in each fraction (Fig. 16A). Extracts from cells that had not been transfected with a Rev-erb␤ expression vector yielded undetectable levels of endogenous protein under similar Western blotting conditions and x-ray film exposure times (data not shown). The purity of the fractions was determined by blotting for compartment-specific proteins including Hsp90 for the cytosol and lamin B1 and NCoR1 for the nucleus (Fig. 16B). Wild-type and heme axial ligand mutants were similarly distributed throughout the cytosol and nucleus, suggesting that heme binding does not regulate subcellular localization. We also examined the effect of heme on the steady-state levels and cellular localization of endogenous Rev-erb␤ in HEK293 cells. SA treatment for 42 h, which lowers heme levels from 0.095 Ϯ 0.008 nmol mg protein Ϫ1 for control cells to 0.044 Ϯ 0.001 nmol mg protein Ϫ1 (the average of three experiments Ϯ S.D.) (Fig. 16C), leads to a 2.3-fold increase in cytosolic Rev-erb␤ with little effect on protein levels in the nucleus (based on densitometry analysis of the bands in Fig 16C). These results are consistent with the increase in steady-state levels observed with the overexpressed H568F variant (Fig 15A). Further, lowering the intracellular heme did not have a pronounced effect on the overall distribution of Rev-erb␤ between cellular compartments, in agreement with those studies in transiently transfected cells overexpressing wild-type Rev-erb␤ and heme axial ligand variants (Fig. 16A).
Heme Does Not Affect Repressor Activity of Rev-erb␤, but It Enhances That of Rev-erb␣-The conflicting data on the role of heme binding in recruitment of NCoR1 to Rev-erb␤ in our co-IP and binding studies prompted us to test whether heme regulates the repressor activity of Rev-erb␤ in the cell. Further, gene expression studies indicate that heme markedly stimulates the interaction of Rev-erb␣ with NCoR1 (16,34). Thus, we tested the ability of wild-type Rev-erb␤ and heme axial ligand mutants H568F, C384A, and H568F/C384A to repress the transcription of a luciferase reporter construct driven by the core Bmal1 promoter from Mus musculus encompassing both Rev-REs and representing Ϫ422 to ϩ108 bp with respect to the transcription start site (pGL3Basic-Bmal1 (71)). It should be noted that the region of the Bmal1 promoter encompassing the proximal and distal Rev-REs is identical in mouse, rat, and human (13) and that both Rev-erb␣/␤ isoforms are known to bind to the Bmal1 promoter and repress transcription in a complex with NCoR1 (6). After transient transfection with pGL3Basic-Bmal1, pCMV-␤Gal (as a transfection efficiency control) and wild-type or mutant Rev-erb␤ expression vector, HEK293 cells were incubated for 24 h, lysed, and assayed for repressor activity expressed as fold change compared with the reporter control (Fig. 17A). Transfection of 0.5 g of wild-type Rev-erb␤ vector repressed luciferase expression to 0.33 Ϯ 0.04fold of the control, and increasing the amount of Rev-erb␤ vector to 2.5 g enhanced repression to 0.072 Ϯ 0.007-fold relative to the control, consistent with higher levels of Rev-erb␤ in those cells (Fig. 17B, compare lanes 1-3). The observed repression is strictly dependent on overexpressed Rev-erb␤ because undetectable levels of endogenous protein are visible in the control lane of the immunoblot. Transfection of 0.5 g of H568F, C384A, and H568F/C384A vectors repressed luciferase expression to 0.41 Ϯ 0.05-, 0.36 Ϯ 0.01-, and 0.42 Ϯ 0.06-fold, respectively, indicating that the mutations have a consistent but minor inhibitory effect on the repressor activity of Rev-erb␤. B, a similar experiment as described for A, except the IPs did not contain heme, and recombinant FLRev-erb␤ HGS , H568F, and H568F/C384A were added to extracts. A control IP containing no recombinant protein was added to demonstrate the specificity of the NCoR1 interaction with the recombinant proteins (Cont.). C, HEK293FT cells were transiently transfected with pcDNA3.1(ϩ) parent vector as a control, pcDNA3.1(ϩ)-FLRev-erb␤ for overexpression of wild-type Rev-erb␤ (wt), or the analogous expression vector for the H568F variant. Rev-erb␤ LBD antibody was added to the cell extracts, and the level of NCoR1 co-purifying with Rev-erb␤ was assessed with immunoblotting. Input is 50 g of extract (1/20th of the mass used in the IP) and Lane R ϭ 50 ng of recombinant FLRev-erb␤ HGS . D, Coomassie-stained SDS-PAG demonstrating high molecular weight bands associated with wild-type FLRev-erb␤ MGC versus the H568F/ C384A mutant. The asterisk indicates the location of the gel that was excised from the wild-type and H568F/C384A IP lanes, and those proteins within the gel slabs were identified with mass spectrometry. 5 g each of pure FLRev-erb␤ MGC (wt) and H568F/C384A are shown for reference (denoted by black bars above the appropriate lanes). E, proteins with high total spectrum counts from the MS analysis are listed for both the wild-type and H568F/C384A samples. F, a similar IP as described in B, except the membrane was probed with a MycBP2 antibody showing a modest heme dependence to the MycBP2/Rev-erb␤ interaction. The absence of MycBP2 in the control lane (Cont.) lacking recombinant protein indicates that the interaction of recombinant Rev-erb␤ with the E3 ligase is specific.
Further, protein levels of axial ligand variants are similar to the wild type (Fig. 17B, compare lane 2 with lanes 4 -6), indicating that repression of the Bmal1 promoter depends on the concentration of Rev-erb␤ and not on the ability of the receptor to bind heme.
In the repression assays (Fig. 17B), we did not observe the higher steady-state protein levels of the heme axial ligand variants relative to the wild type described in Fig. 15A. Perhaps this is because of a much higher expression level of Rev-erb␤ due to the increased amount of Rev-erb␤ expression vector used in the assay described in Fig. 15A. Furthermore, it is likely that there is a subtle balance between the rate of Rev-erb␤ degradation and the amount of the NR such that the stabilization is observed only in cells expressing high levels of Rev-erb␤ or depleted of heme for long periods of time. Differences in cell lysis reagents between studies may also affect the amount of Rev-erb␤ extrac- FIGURE 15. Western blotting analysis indicates that heme induces Rev-erb␤ protein degradation through a proteasome-dependent pathway. A, a Western blot demonstrating steady-state levels of Rev-erb␤ in HEK293 cells transiently transfected with pcDNA3.1(ϩ) parent vector (Cont., control), or the corresponding expression vectors for wild-type Rev-erb␤ (wt) or the H568F variant (H568F). Cells (60 -70% confluent in 35-mm-diameter plates) were transfected with 4 g of vector and incubated for 24 h, and a soluble extract was prepared in radioimmune precipitation assay buffer. 30 g of each extract was electroblotted and probed with Rev-erb␤ QK-6 antibody (top panel) or actin as a loading control (bottom panel). B, same experiment as described for A, except cells were transfected with expression vectors for wild-type Rev-erb␣ (wt) or the H602F variant (H602F). Membranes were probed with Rev-erb␣ RS-14 antibody (top panel) or actin as a loading control (bottom panel). C, the intensities of the bands in A and B were quantified by densitometry, and the mean intensities Ϯ S.D. of three experiments were plotted. For the Rev-erb␣ analysis, the low level intensity of the band in the control lane was subtracted from the intensities of the bands for wild-type and H602F Rev-erb␣ from the same experiment. D, HEK293 cells transiently transfected with pcDNA3.1(ϩ)-FLRev-erb␤ (wild-type) or the analogous expression vector for H568F (top experiment) or the H568F/C384A variant (bottom experiment) were treated with 20 g ml Ϫ1 CHX. Cells were lysed and soluble extracts prepared at the indicated time points post-CHX addition. 50 g of each extract was electroblotted and probed with Rev-erb␤ QK-6 or an actin antibody as a loading control. Lane R, 25 ng of FLRev-erb␤ HGS . E, the intensities of the bands in D were quantified by densitometry, and the mean intensities Ϯ S.D. of three experiments were plotted. F, a replicate CHX half-life experiment as described in D for cells overexpressing wild-type Rev-erb␤. Lane R, 5 ng of FLRev-erb␤ HGS . G, the same cells as described in F were treated with either 20 g ml Ϫ1 CHX or 20 g ml Ϫ1 CHX plus 20 M MG-132 (denoted by black bars above the appropriate lanes; three biological replicates of each set are depicted). Cells were incubated for an additional 4 h prior to preparing extracts and blotting with Rev-erb␤ QK-6, indicating that the proteasome degrades Rev-erb␤ (i.e. a direct comparison can be made between time points from F and G, as these experiments were carried out on the same day). . The cytosolic and nuclear fractions are indicated by black bars above the corresponding lanes, and duplicate biological replicates (Rep.) are shown for each fraction. B, the purity of cytosolic and nuclear fractions assessed by probing the membranes described in A with antibodies against lamin B1 (nuclear envelope), NCoR1 (nucleus), and Hsp90 (cytosol). A representative blot is shown for each protein with duplicate biological replicates. C, the effect of altering intracellular heme on the distribution and steady-state level of endogenous Rev-erb␤. HEK293 cells were incubated with 1 mM SA for 42 h or left untreated as a control (indicated by black bars above the appropriate lanes). The blot is arranged as described in A, with duplicate biological replicates shown for each fraction; 5 ng of FLRev-erb␤ MGC as a positive control (lane R). Blots assessing the purity of the fractions are also depicted.
tion. In any case, the nearly equivalent levels of wild-type Rev-erb␤ and heme-binding variants in the repression assays underscore the conclusion that it is the concentration of Rev-erb␤ in cells that dictates repression of the reporter and not its ability to bind heme.
A similar experiment was performed with cells co-transfected with 0.5 g of expression vectors for wild-type or H602F Rev-erb␣ (Fig. 17C). The wild-type Rev-erb␣ vector represses luciferase expression to 0.53 Ϯ 0.07-fold of the control versus the H602F mutant with a value of 0.82 Ϯ 0.15-fold. Similarly, transfection of 2.5 g wild-type Rev-erb␣ vector enhances repression to 0.19 Ϯ 0.02-fold of the control, whereas the same mass of H602F vector leads to 0.40 Ϯ 0.03-fold repression. These data agree with previously published results for similar reporter-based assays with Rev-erb␣ and the H602F variant (16,34) and suggest that heme enhances Rev-erb␣ repressor activity to a greater degree than Rev-erb␤. The results also indicate that heme binding is only partially responsible for Rev-erb␣ repressor activity. It may be that Rev-erb␣/␤ competition with the transcriptional activator, human retinoic acid receptor-related orphan receptor, at Rev-REs in the Bmal1 promoter is predominantly responsible for the observed repression. Perhaps promoter elements with closely spaced Rev-REs, e.g. those resembling the canonical Rev-DR2, are subject to higher degrees of NCoR1 contribution to repression. Another interpretation is that the reporter system is insensitive to regulation by endogenous NCoR1 due to overexpression of the promoter element and Rev-erb NRs. In any case, although heme binding is dispensable for Rev-erb␤ repressor activity, it modestly stimulates that of Rev-erb␣.

Discussion
Determination of the Properties of FLRev-erb␤ and Demonstration That All Its Domains Are Intact-To test our hypothesis related to the discrepancy between cellular and in vitro studies, we developed a method to obtain the full-length receptor and showed that it binds all of its ligands, e.g. heme, CO, DNA, and NCoR1 (both a peptide and a construct containing all three NR-binding domains), with high affinity. In doing so, our work represents the first report of a homogeneous preparation of an active full-length Rev-erb NR and legitimizes the use of the construct to test the effect of heme on its functional properties.
FLRev-erb␤ purifies as a homodimer, in accord with studies of the LBD (apoRev-erb␤ and the Rev-erb␣ LBD⅐ID1 complex) (27,29) and DBD (Rev-erb␣ DBD bound to the Rev-DR2 promoter element) (80,81). FLRev-erb␤ binds with high affinity (K 0.5 values Ͻ100 nM) and cooperatively to DNA Rev-REs, yielding Hill coefficients between 1.6 and 2.0 for interactions with the TR/FAM-Rev-DR2 and FAM-Bmal1 promoter. These results are similar to those of THR␤ DBD binding as a homodimer to an inverted palindrome (with a K 0.5 of 130 nM) (88) and of the isolated Rev-erb␤ DBD binding to Rev-DR2 with an apparent K 0.5 of 270 nM (82).
FLRev-erb␤ also tightly binds heme, with K d values below 30 nM for the complex between FLRev-erb␤ MGC/HGS and either ferric or ferrous heme, values that are similar to the low nM K d value for the complex between heme and thiol-reduced Rev-erb␤ LBD (38). Furthermore, the UV-visible spectra of the heme⅐FLRev-erb␤ complexes are nearly identical to those of heme⅐LBD (38) (see Table 2). Thus, the LBD in the context of the full-length protein retains its heme coordination environment and its high affinity for heme. Furthermore, because these K d values are below the 0.1-0.2 M concentration of the intracellular exchangeable heme pool (78,79), the receptors would FIGURE 17. Heme binding enhances the repressor activity of Rev-erb␣ but has a negligible effect on Rev-erb␤ repression of the Bmal1 promoter. A, HEK293 cells were transfected with 0.1 g of ␤-galactosidase control vector, 0.2 g of pGL3Basic-Bmal1 luciferase reporter, and varying masses of pcDNA3.1(ϩ)-FLRev-erb␤ wild-type (wt) or variant expression vectors (indicated below the graph). 24 h after transfection, luciferase activity was measured as described under "Experimental Procedures." Data are normalized to ␤-galactosidase activity and are represented as -fold change in luciferase activity based on the reporter control. Three biological replicates were performed with mean Ϯ S.D. shown. B, Western blotting analysis of 50 g of soluble extracts from a representative replicate experiment described in A, probed with Rev-erb␤ QK-6 antibody (top panel), or actin (bottom panel) as a loading control. 5 ng of recombinant FLRev-erb␤ HGS as a positive control (lane R). A nonspecific reactive band (NS) and a potential Rev-erb␤ degradation product (Deg) are indicated with arrows. C, a similar experiment as described in A, except cells were co-transfected with the ␤-galactosidase control vector, Bmal1 luciferase reporter, and varying masses of pcDNA3.1(ϩ)-Rev-erb␣ (wt) or the corresponding expression vector for the H602F variant (indicated below graph). Bars represent the average Ϯ S.D. of three biological replicates. be expected to exist predominantly within the cell in the hemebound state. On the other hand, the H568F and H568F/C384A variants, with 1000-fold lower heme affinity than the wild-type protein, would exist as apoproteins within the cell.
In accord with the results of equilibrium titrations, heme dissociates at least 600-fold faster from the H568F and H568F/ C384A variants relative to wild-type protein. The heme dissociation rates correlate with the role of heme in a particular protein, from relatively fast (10 Ϫ2 -10 Ϫ3 s Ϫ1 ) off-rate constants for heme sensors and transporters such as HRI and BSA, which must bind and release heme quickly as metabolic conditions (such as heme availability) change to 10 Ϫ4 -10 Ϫ7 s Ϫ1 for hemoproteins that interact with diatomic gases (75,89). Thus, the slow Fe 3ϩ -heme dissociation rate constant for FLRev-erb␤ (0.3 day Ϫ1 at 20°C) is more akin to those for globins than those for heme sensors, indicating that Rev-erb␤ would be ineffective in its proposed (16) role as a heme sensor. The instability of wildtype FLRev-erb␤ during the long off-rate measurements prevented us from performing them at physiological temperatures; nonetheless, increasing the temperature from 20 to 37°C increased the rate of heme dissociation from myoglobin by ϳ3-fold (60). Assuming a similar Arrhenius relationship for Rev-erb␤, the rate constant (10 Ϫ5 s Ϫ1 ) for heme dissociation remains well within the range of proteins like globins that cannot readily exchange heme.
Related to the potential role of Rev-erb␤ in diatomic gas binding, both Rev-erb isoforms bind NO and CO, and our laboratory has shown avid CO binding (K d ϭ 6 -60 nM) (Ref. 38 and herein). Although CO is reported to have little effect on Rev-erb␤ repressor activity (36), one study reports that NO negligibly affects Rev-erb␣ repression (34), whereas another indicates that NO relieves Rev-erb␣/␤ repression of target genes by 2-4 fold (36). This effect of repression could occur by disrupting interactions with corepressors (like NCoR1) or proteins that regulate steady-state levels of Rev-erb␤. Further studies are needed to characterize the interactions of NO and CO with Rev-erb and their potential effects on activity.
The FLRev-erb␤ construct tightly binds an NCoR1 ID1 peptide (K d ϭ 15 nM) with similar affinity to that of the isolated Rev-erb␣ LBD (27). Thus, in sum, our experiments validate the idea that the functions of the LBD and the DBD are all intact in our FLRev-erb␤ construct. This is important, because few fulllength NRs have been expressed in an active state. Furthermore, although the two subunits of the dimer (either full-length or DBDs) exhibit cooperative DNA binding, each of the domains of FLRev-erb␤ apparently act as independent modular units.
An Unknown Factor Present in Cell Extracts Mediates the Effects of Heme on the Interaction of Rev-erb␤ and NCoR1-Although both the ID1 and ID3 motifs of NCoR1 are important for interacting with Rev-erb␣ (83), the peptide encompassing only ID1 appears to accurately mimic interactions between FLRev-erb␤ and the MBP-NCoR1-540 construct, which comprises all three NR interaction domains. Furthermore, gel shift and fluorescence anisotropy assays show that the apo form of FLRev-erb␤, as well as variants that poorly bind heme, interacts with MBP-NCoR1-540 and promoter DNA (Rev-DR2 or the core Bmal1 promoter). Also, the addition of Fe 2ϩ /Fe 3ϩ -heme only marginally affects the affinity of FLRev-erb␤ for NCoR1.
On the other hand, unlike wild-type Rev-erb␤, the H568F and H568F/C384A variants (whether added as purified recombinant protein or expressed in the cell) failed to interact with NCoR1 in extracts, reminiscent of a previous co-IP study with the Rev-erb␣ H602F variant (16). Furthermore, we could only co-immunoprecipitate NCoR1 when Rev-erb␤ was in Ͼ100fold excess of endogenous levels. It is likely that this dose effect results from the competition of Rev-erb␤ with other NCoR1binding proteins under the conditions of this assay.
These combined results strongly suggest that an additional cellular factor present in HEK293 extracts is required in order to observe the effect of heme on the interaction between NCoR1 and Rev-erb␤, at least in the context of the co-IP assay. Our MS/bioinformatics studies suggest that one factor is the proteasomal E3 ligase MycBP2. For example, binding of heme to Rev-erb␤ may enhance its ubiquitination by MycBP2. Such interactions may also be directly or indirectly affected by a posttranslational modification. For example, glycogen synthase kinase 3␤ phosphorylates two conserved serine residues in the A/B region of Rev-erb␣, stabilizing the protein; however the effect of heme on the interaction of these proteins has not been explored (90). Acetylation is another potential post-translational modification regulated by heme binding. The histone acetyltransferase Tip60 acetylates a conserved Rev-erb␤ lysine within a consensus RXKK sequence, abrogating repression of the apoC-III gene. Repressor activity is restored by HDAC1, which interacts with Rev-erb␤ and deacetylates the regulatory lysine (91). In light of our results, it is imperative to determine whether heme affects the interaction of Rev-erb␤ with Tip60 or HDAC1.
Another possibility is that heme regulates the interaction of Rev-erb␤ with a protein in HEK293 extracts that regulates NCoR1 recruitment. One possibility is ZNHIT-1, a zinc-finger HIT domain protein known to interact with the Rev-erb␤ LBD through a FXXLL motif similar to the LXXLL motif found in NR coactivators. ZNHIT-1 binds Rev-erb␤ on the apoC-III promoter and inhibits its repressor activity; whether ZNHIT-1 accomplishes this through competitive binding with NCoR1 has not been tested (92). Further, because heme, ZNHIT-1, and NCoR1 bind to the LBD and have differential effects on repression, it will be critical to ascertain whether heme binding affects the interaction of Rev-erb␤ or NCoR1 with ZNHIT-1.
Cellular Studies of the Role of Heme in Regulating Rev-erb␤ Repressor Activity and Turnover-The Rev-erb␣ H602F variant, which is deficient in heme binding, is unable to repress transcription or to interact with NCoR1 at the Bmal1 promoter (16,34). In part, these results laid the groundwork for the hemedependent NCoR1 recruitment model described in the introduction to this work. We replicated those results, demonstrating that heme binding enhances Rev-erb␣ repressor activity by ϳ2-fold over basal repression by the apoprotein. However, surprisingly, although the Rev-erb␤ H568F and H568F/C384A variants are highly deficient in binding heme, they still interact with NCoR1 and DNA (Fig. 8) and retain repressor activity in promoter fusion studies (Fig. 17), providing consistency between in vitro and cell-based experiments in indicating that heme has different effects on the ␣versus ␤-isoforms of Reverb. Again, the co-IP study demonstrating that H568F failed to bind NCoR1 provides the only conflicting data that we observed against a model in which heme binding is dispensable for Rev-erb␤ repressor activity and an interaction with NCoR1.
Heme-dependent regulation of the proteolysis/stability of Rev-erb NRs would be important in at least two physiological processes: adipogenesis and the circadian cycle. Although peroxisome proliferator-activated receptor ␥ induces the high levels of Rev-erb␣ necessary for the early stages of preadipocyte differentiation (93), repression of peroxisome proliferator-activated receptor ␥2 and proteasomal degradation of Rev-erb␣ are required for later differentiation stages (85). Heme levels increase throughout adipogenesis at a rate closely matching that of Rev-erb␣ degradation. Similarly, decreasing cellular heme levels with 3-amino-1,2,4-triazole increases the steadystate levels of Rev-erb␣ in adipocytes (86). However, heme did not affect degradation of Rev-erb␣ in cells treated with the translation inhibitor cycloheximide (94).
The role of heme in controlling the proteolysis of Rev-erb␤ is consistent with prior results on other proteins that function in the negative limb of the circadian clock (94 -96), where Rev-erb␣ and Period (Per) have been shown to undergo ubiquitination and proteasomal degradation. In this process, Bmal1-CLOCK heterodimers activate the transcription of clock-controlled genes, including Rev-erb␣/␤, Per, and Cry. The binding of heme to the HRM of Period 2 (Per2) enhances its proteasomal degradation in immortalized cells (97). Heme also has been shown to dampen Per2 rhythm in mouse suprachiasmatic nucleus explants (98). Per forms a heterodimeric complex with Cry that inhibits the transcriptional activator function of Bmal1-CLOCK, whereas Rev-erb NRs feed back, repressing the transcription of Bmal1 and CLOCK, which also has been shown to bind heme (99), although this binding is claimed to be nonspecific (100). Significantly, the CLOCK homolog in the forebrain, NPAS2, is also a hemoprotein that binds CO, which ultimately disrupts NPAS2 DNA binding (101), although heme binds to the CO-sensing PAS-A domain with bis-His ligation and not via an HRM (102). Thus, heme binding to clock proteins appears to represent an important mechanism in circadian rhythm maintenance.
The circadian clock interconnects various aspects of heme and cellular metabolism. The expression of key heme proteins involved in electron transport and xenobiotic metabolism (P450s) is entrained with the circadian rhythm (103). Furthermore, heme biosynthesis itself is controlled by the molecular clock through the regulation of expression of the rate-determining enzyme aminolevulinic acid synthase 1 (ALAS1). The Bmal1-NPAS heterodimer and PGC-1␣ (a coactivator of nuclear receptors and regulator of metabolism), which are both under circadian control, activate expression of ALAS1 and Rev-erb␣, with Rev-erb␣ completing the feedback loop by binding heme as a ligand, recruiting NCoR1, and repressing expression of Bmal1 and PGC-1␣ (34, 104 -107). The interconnecting regulatory loops of heme biosynthesis, the molecular clock, and metabolism are exemplified in a recent study demonstrating that a diet high in iron increases the amplitude of circadian PGC-1␣ expression, which raises intracellular heme levels. In turn, elevated heme leads to NCoR1 recruitment to Rev-erb␣ and repression of key enzymes involved in gluconeogenesis, ultimately lowering serum glucose concentrations. Thus, dietary signals like iron control metabolism through the regulation of core clock protein expression and heme biosynthesis (108).
Conclusion-Many of the key studies supporting a role for heme in recruiting NCoR1 to Rev-erb NRs are based on experiments with the ␣-isoform (16,21,34,86,109), which is more extensively studied, likely because of its discovery prior to Rev-erb␤ and its unique context in the genome being transcribed on the opposite strand of the gene coding for THR␣ on chromosome 17 (2,9), whereas Rev-erb␤ is encoded on chromosome 3. Because the DBD and LBD of the two receptors share a high level of homology and nearly identical heme binding properties (9,36), it has been assumed that the mechanism of heme regulation would be the same for both isoforms. Further, several reports demonstrate that Rev-erb␤ and Rev-erb␣ share a significant overlap in the genes they repress, especially those involved in lipid homeostasis and the molecular clock, namely Bmal1 (6,7,110). This redundancy likely provides protection against clock and metabolic dysfunction, as knockdown or deletion of one Rev-erb isoform has only modest effects on these processes, whereas loss of both leads to marked dyslipidemia and loss of rhythmic clock gene expression (6,7). In this study we have shown that Rev-erb␤ behaves like Rev-erb␣ in terms of heme-dependent NCoR1 binding in co-IPs, Nonetheless, heme appears to have a more pronounced effect on stabilizing Rev-erb␤ H568F versus Rev-erb␣ H602F, but it also enhances Rev-erb␣ repressor activity to a greater degree than the ␤-isoform (although this effect is modest). Thus, there seems to be subtle differences in the regulatory role of heme in terms of Rev-erb turnover and repressor activity. Two recent reports demonstrate that Rev-erb␣ and -␤ bind at or near the same Rev-REs in the promoters of target genes (6, 7), provoking the question, do the isoforms form heterodimers that cooperatively repress transcription? Perhaps, the effects of heme and NCoR1 act on a Rev-erb␣/␤ heterodimer. Intriguingly, both Rev-erb␣ and Rev-erb␤ repress the transcription of Rev-erb␣; however, neither receptor binds to the Rev-erb␤ promoter (6). Thus, heme-induced degradation may act as a post-translational mechanism to control the protein levels of Rev-erb␤, maintaining its ability to rhythmically repress the transcription of genes in metabolic pathways, thereby synchronizing metabolism with the molecular clock.
Author Contributions-E. L. C., N. G., and S. W. R. designed the study, and E. L. C. and N. G. performed the experiments. E. L. C. and S. W. R. composed the manuscript, and all authors reviewed and approved its content.