Transcription Factor NF-Y Is a Functional Regulator of the Transcription of Core Clock Gene Bmal1 ♦

Background: The regulation of the core clock gene Bmal1 is not fully understood despite many studies. Results: NF-Y cooperates with Sp1 to activate the transcription of Bmal1 by binding to the CCAAT-boxes in the Bmal1 promoter. Conclusion: NF-Y/Sp1 and Rev-Erbα serve as positive and negative regulators of Bmal1 expression, respectively. Significance: Elucidation of Bmal1 regulation provides a better understanding of its biological functions and the circadian clock overall. The circadian clock enables organisms to adjust to daily environmental changes and synchronize multiple molecular, biochemical, physiological, and behavioral processes accordingly. In mammalian clock work, Bmal1 is the most important core clock gene, which works with another core clock gene Clock to drive the expression of other clock genes and clock-controlled genes. However, the regulation of Bmal1 has not been fully understood. This work was aimed at identifying the positive regulator(s) of Bmal1 transcription. A series of 5′ deletion reporter constructs was generated, and binding site mutations of mouse Bmal1 promoter fragments were cloned into pGL3-basic and pGL3(R2.1)-basic plasmids and transfected into NIH 3T3 cells. Luciferase activity was either measured 48 h after transfection or recorded for 4 days after serum shock. DNA affinity precipitation assay was used to detect the transcription factors binding to Bmal1 promoter. Small interfering RNA against nuclear factor Y, subunit A (NF-YA) and dominant negative NF-YA were employed to study the role of NF-Y in Bmal1 transcription regulation. Deletion and mutation analyses identified two clusters of CCAAT/GC-boxes at the proximal region of Bmal1 promoter as the activating cis-elements. Bmal1 promoter activity was up-regulated by NF-Y and/or Sp1 and repressed by dominant negative NF-YA or siRNA against NF-YA. The activation of Bmal1 promoter activity by NF-Y and Sp1 was inhibited by Rev-Erbα. DNA affinity precipitation assay showed that NF-Y and Sp1 bound to the two CCAAT/GC clusters of Bmal1 promoter. These results indicate that NF-Y is a functional activator of Bmal1 transcription and it cooperates with Sp1 and Rev-Erbα to generate the daily cycle of Bmal1 expression.

mutating ROR␣ did not show either drastic decrease of Bmal1 expression or disruption of Per2 expression rhythm (5). Using small interfering RNA knockdown, ROR␣ also had only minimal effects on Bmal1 expression in mouse embryonic fibroblasts (6). It is of great interest to identify the dominant activator of Bmal1 transcription because high level Bmal1 expression is required for generating oscillation of both the positive and the negative loops.
The mammalian CCAAT-box-binding factor nuclear factor Y (NF-Y) is an evolutionarily conserved transcription factor presented from yeast to human (7). A functional NF-Y factor consists of three different subunits, NF-YA, NF-YB, and NF-YC. Both the NF-YB and the NF-YC subunits contain a histone-fold motif and interact with each other to form a NF-YB/NF-YC heterodimer. The B/C complex then interacts with the NF-YA subunit to form the NF-Y heterotrimeric transcription factor. All three subunits are required for the binding of NF-Y complex to DNA and NF-Y-directed transcription (7).
The CCAAT motif (Y-box) is present in promoters of many mammalian genes spanning from developmentally controlled and tissue-specific, housekeeping and inducible, to cell cycleregulated. The position of the CCAAT-box in promoters is strongly biased, and it is typically found as a single copy element in either forward or reverse direction between Ϫ30 and Ϫ150 bp from major transcription start sites (8). Bmal1 promoters from mouse to human are highly conserved at the proximal region where multiple copies of CCAAT-boxes are located upstream of the transcription start site. Although NF-Y was implicated as a regulator of the circadian clock (9), there has been no experimental evidence to confirm its role in the transcriptional regulation of clock genes. Here we report the identification of NF-Y as the major activator of Bmal1 transcription.

MATERIALS AND METHODS
Cell Culture, Transient Transfection, and Luciferase Assays-Mouse embryonic fibroblast NIH 3T3 cells (ATCC, Manassas, VA) were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% bovine serum, 100 IU/ml penicillin, and 100 g/ml streptomycin (all from Invitrogen, Shanghai, China) at 37°C in a humidified atmosphere (5% CO 2 ). On the day before transfection, NIH 3T3 cells were seeded into 12-well plates to have a confluence about 30 -50% on the day of transfection. Reporter constructs and expression vectors were transfected using Lipofectamine Plus (Invitrogen) according to manufacturer's instructions. 50 g of pCMV-lacZ (Promega, Madison, WI) was transfected as the internal control. The total DNA amount was kept at 1.0 mg/well. Corresponding empty vectors were used to make up the differences among treatments. Luciferase activity was measured 48 h later using a microtiter plate luminometer (Dynex, Chantilly, VA) and luciferase assay kit (Promega). All experiments were done by triplicates and performed at least three times independently. Results were shown as relative luciferase activity and expressed as mean Ϯ S.E.
Plasmids-The mouse Bmal1 promoter luciferase reporter constructs were generated by PCR (see Table 1 for primer sequences) and digested with MluI and XhoI and then ligated into pGL3-basic (Promega) predigested with the same enzymes. For bioluminescence recording, the same MluI/XhoI promoter fragments were transferred into pGL3-Basic(R2.1) (Promega). The mutation of the NF-Y-, Sp1-, or ROR␣-binding site was done by site-directed mutagenesis with the primers listed in Table 1. Heterologous promoter reporter vectors were generated by cloning custom-made oligonucleotides corresponding to either the proximal CCAAT/GC cluster or the ROR␣-binding element of murine Bmal1 promoter (refer to Table 1 for primer sequences) into BamHI/XhoI sites of enhancerless luciferase reporter vector pT81. All the constructs were verified by DNA sequencing. NF-Y expression constructs (⌬4NF-YA, ⌬4NF-YB, and ⌬4NF-YC) were gifts from Dr. Mantovani (Universita di Milano, Italy). Dominant negative NF-YA (YA m29) was generated by site-directed mutagenesis as described previously (10). pCMV-Bmal1 and mPer1 promoter reporter constructs have been described previously (11).
Real-time Recording of the Circadian Expression of Bmal1-NIH 3T3 cells were seeded in 35-mm plates with a density of greater than 95% confluence and transfected the next day with the indicated DNA constructs by Lipofectamine Plus. 24 h after transfection, cells were treated with 50% horse serum for 2 h and replaced with culture medium without serum. Luciferin was added at a final concentration of 1 M. Light emission was measured and integrated for 1 min at intervals of 15 min by an LM-2400 photon detection unit (Hamamatsu, Japan).
Preparation of Nuclear Extract and DNA Affinity Precipitation Assays (DAPA)-Nuclear extracts from 24 h post-serumshocked NIH 3T3 cells were prepared as described previously (19). DAPA have been described previously (19). The sequences for the oligonucleotides used in DAPA were as follows (with mutated nucleotides in lowercase): 1-WT, 5Ј-CTC CAT TGG TGG GCG GGG AAG G; 1-NF-Y, 5Ј-CTC CAA ACG TGG GCG GGG AGG G; 1-Sp1, 5Ј-CTC CAT TGG TGG TAT GGG and 2-Mut, 5Ј-GCG AAA CGT GGT ATG GAG TAT GGG CCT. 5Ј-Biotinylated double-stranded oligonucleotides corresponding to wild type or mutant (CCAAT-and GC-box mutated individually or combined) sequences of either distal or proximal clusters of Y-box/GC-box were incubated with NIH 3T3 nuclear extract. The protein-DNA complexes were pulled down by Tetralink TM avidin resin and resolved on PAGE gel followed by immunoblotting with antibodies against NF-YA (G-2) and Sp1 (PEP2) (Santa Cruz Biotechnology, Inc. Santa Cruz, CA). Small Interfering RNA (siRNA) against NF-YA-siRNAs were designed using the Invitrogen RNAi designing tool. Two pairs of RNA oligonucleotides targeting different regions of NF-YA and a control pair (enhanced green fluorescence protein) were custom-synthesized by Sangon (Shanghai, China). Doublestranded siRNA was transfected into NIH 3T3 cells using Lipofectamine 2000 (Invitrogen, Shanghai, China) according to the manufacturer's instructions. The protein or mRNA level was measured 48 h after transfection by immunoblotting with specified antibodies or quantitative real-time PCR using the ABI Prism 7000 detection system (Applied Biosystems, Foster City, CA).

Identification of Promoter Region Conferring Bmal1
Activation-It has been shown that Bmal1 expression is negatively regulated by Rev-Erb␣, which drives the circadian oscillation of Bmal1 expression and forms the stabilizing loop in the positive limb of the circadian clock. However, the identity of the activator of Bmal1 transcription has been elusive. We employed in vitro real-time expression monitoring systems to identify the cis-element in the mouse Bmal1 promoter driving its transcription. A series of 5Ј deletion reporter constructs was generated using a destabilized luciferase reporter vector to enhance the oscillation resolution. The reporter constructs were transfected into NIH 3T3 cells individually, and their bioluminescence was recorded. The expression level and circadian rhythm of Bmal1 were gradually reduced from Ϫ275 to Ϫ107 deletion constructs (Fig. 1, A and B). Further truncating Bmal1 promoter to Ϫ77, the promoter activity was further diminished (Fig. 1A) and circadian oscillation was almost flattened (Fig.  1B), indicating that the cis-element(s) critical for activation of Bmal1 transcription were localized within Bmal1 promoter regions between Ϫ275 to Ϫ77. As all promoter reporter constructs had intact transcription start site and both ROR-binding sites, the loss of transcriptional activity and rhythmicity indicated the requirement of (an) additional cis-element(s) within the promoter for the up-regulated expression of the Bmal1 gene.
NF-Y Activates Bmal1 Transcription through the Inverted CCAAT-boxes-The proximal regions of human and mouse Bmal1 promoter are highly conserved, and they include several inverted CCAAT motifs (Y-box), two clusters of Y-box and GC-boxes, as well as two ROR-binding sites ( Fig. 2A). To determine whether NF-Y is truly an activator of Bmal1 gene, we first co-transfected mouse Bmal1 luciferase reporter gene with NF-Y expression constructs in transient transfection assay. Cotransfection of NF-Y yielded 4 -10-fold increase of mouse Bmal1 promoter activity (Fig. 2B). Co-transfection of the dominant negative form of NF-YA expression vector (⌬4NF-YAm29) with Bmal1-luc strongly inhibited Bmal1 promoter activity ( Fig. 2C) but maintained detectable oscillation (Fig.  2D), implicating the involvement of NF-Y in activating Bmal1 FIGURE 1. Identification of promoter element(s) responsible for mouse Bmal1 gene activation and rhythmic expression. A, 5Ј deletion Bmal1 reporter constructs were transfected into NIH 3T3 cells. Cells were treated with 50% horse serum for 2 h on the next day followed by culturing in regular DMEM without serum and 1 M luciferin (Luc) and placed in LM-2400 for bioluminescence recording. The highest activity recording of each construct was set as 1, respectively. B, data were normalized to the average luciferase activity produced over the duration of the recording and are plotted as the difference from the centered 24-h moving average. expression. Two lines of experiments were carried out to further dissect the sequence requirements of mouse Bmal1 promoter for NF-Y activation. First, a series of 5Ј-truncated promoter reporter genes were co-transfected with or without NF-Y constructs to locate the region conferring significant activation of mouse Bmal1 promoter. As deleting from Ϫ255 to Ϫ165 then to Ϫ107 caused gradual decreases of activation of Bmal1 expression, further deletion from Ϫ107 to Ϫ77 resulted in significant loss of activation (Fig. 2E). Next, we characterized the possible roles of two clusters of Y-box/GC-box in mediating NF-Y activation of Bmal1 expression. The Y-boxes and GCboxes were mutated individually or in combination within each cluster. Mutating the GC-box of distal cluster (C1) did not have significant effect on either promoter activity or activation by NF-Y. Mutation of the Y-box of C1 or GC-boxes of proximal cluster (C2) had a moderate effect on NF-Y activation of Bmal1 expression. However, mutation of proximal Y-box caused more than 50% decrease of such activation (Fig. 2F). Furthermore, deletion of a small fragment from Ϫ118 to Ϫ77, which contained the proximal Y-box/GC-box cluster, from the Ϫ318 construct almost totally abolished the activation of Bmal1 expression by NF-Y (last lane of Fig. 2E). Alternatively, when the proximal cluster of Y-box/GC-boxes (C2) was cloned into an enhancerless luciferase reporter vector pT81, it was specifically activated by NF-Y, whereas pT81-RORE did not show any activation by NF-Y when compared with empty pT81 vector (Fig.  2G).

NF-Y and Sp1 Bind to Two Y-box/GC-box Clusters-To test
whether NF-Y actually binds to those two CCAAT-boxes, DAPA were performed. Wild type oligonucleotides of both distal (C1) and proximal (C2) elements pulled down NF-Y and Sp1, whereas mutating Y-box and GC-box simultaneously abolished their ability to bind either protein (Fig. 3, A and B). Individual mutation of Y-boxes reduced both NF-Y and Sp1 binding, whereas mutation of GC-boxes only reduced Sp1 binding capability but did not have any effect on NF-Y binding (Fig. 3, A and B). To confirm the specificity of such binding, we then performed competitive DAPA. Each wild type biotinylated oligonucleotide was incubated with NIH 3T3 cell nuclear extracts alone or with 10 times the amount of nonbiotinylated oligonucleotide (wild type or mutants) added. Wild type oligonucleotide competed away NF-Y and Sp1 binding to its own sequence as well as the other cluster of Y-box/GC-box (Fig. 3, C  and D). The oligomers with mutated Y-box competed away Sp1 binding but not NF-Y binding, whereas mutation at the GC-box abrogated its ability to compete for Sp1 but not NF-Y binding (Fig. 3, C and D).
NF-Y Activation of Bmal1 Expression Is Inhibited by Rev-Erb␣-As Rev-Erb␣ has been shown to be a negative regulator of Bmal1 transcription to generate the oscillation of Bmal1 expression, we next tested whether NF-Y activation of Bmal1 was repressed by Rev-Erb␣. In a transient transfection luciferase assay, we co-transfected Bmal1 promoter reporter with NF-Y and an increasing amount of Rev-Erb␣ FIGURE 2. NF-Y is the functional transcription activator of Bmal1 gene. A, alignment of human (hBmal) and mouse (mBmal) Bmal1 promoter proximal region. The important and highly conserved transcription factor-binding sites are shown in boldface and underlined. B, Bmal1 luciferase (Bm-luc) reporter gene (d318) was co-transfected with NF-Y or empty pSG5 vectors into NIH 3T3 cells. Luciferase activity was measured 48 h later. C, Bmal1 luciferase reporter gene (d318) was co-transfected with dominant negative NF-YA or empty pSG5 vectors into NIH 3T3 cells. Luciferase activity was measured 48 h later. The remaining luciferase activity was shown. D, NIH 3T3 cells were transfected with Bmal1 promoter construct in combination with control vector or pSG5-YAm29 and treated the same way as in Fig. 1. E, 5Ј-truncated Bmal1 promoter constructs were co-transfected with NF-Y or empty pSG5 vector. The ratio of luciferase activity of each reporter with and without NF-Y was shown. F, Bmal1 promoter (d318) reporter constructs with mutated different CCAAT-box or/and Sp1 site were cotransfected with either pSG5 or NF-Y expression constructs. The relative luciferase activity (left panel) and -fold of activation by NF-Y (right panel) were shown. G, pT81-carrying the proximal CCAAT/GC-box/GC-box (C2) and ROR␣-binding sites of mouse Bmal1 promoter were co-transfected with NF-Y, Sp1, and/or Rev-Erb␣. The luciferase activity relative to ␤-galactosidase activity was shown. Error bars in B, C, E, F, and G indicate mean Ϯ S.E. expression construct. The activation of Bmal1 promoter activity by NF-Y was dose-dependently repressed by Rev-Erb␣ (Fig. 4A). Furthermore, Bmal1 promoter was activated by NF-Y and Sp1 synergistically, which was repressed by Rev-Erb␣ (Fig. 4B).
The Elements in Bmal1 Promoter Binding Positive and Negative Regulators Are Separable-It has been shown that two ROR-binding sites are required for the repression and rhythmic expression of Bmal1 gene. On the other hand, although it was suggested that ROR␣ bound to RORE to activate Bmal1 transcription, we identified the sequence between Ϫ165 and Ϫ77 (especially from Ϫ107 to Ϫ77) of Bmal1 promoter as the more important cis-elements for positive regulation of Bmal1 expression. To determine the role of RORE and the cis-element located within Ϫ107 to Ϫ77 in the transcriptional activation of Bmal1, we made two reporter constructs to specifically address this issue. The first construct had both ROR␣-binding sites mutated in the Ϫ275 construct (mRORE), and the second one had wild type ROR␣-binding sites but a deletion of the sequence from Ϫ118 to Ϫ77 (d118 -77). These two reporter constructs showed completely opposite behavior. Although mRORE had high basal Bmal1 expression, d188 -77 showed minimal promoter activity (Fig. 5A). Consequently, in serumshocked NIH 3T3 cells, mRORE showed constantly high level expression, whereas d118 -77 had extremely low level expression and without oscillation (Fig. 5B). Furthermore, mutating the Y-box within the region between Ϫ118 and Ϫ77 caused significant decrease of Bmal1 promoter activity in the presence of NF-Y but had no effect on the repression by Rev-Erb␣ in the presence or absence of NF-Y. On the contrary, disrupting the ROR␣-binding sites abolished the ability of Rev-Erb␣ to inhibit Bmal1 expression at basal or NF-Y-activated state (Fig. 5C).
Endogenous Expression of Bmal1 and Its Target Gene Is Reduced by Knocking Down NF-Y Activity-The ability of NF-Y to regulate Bmal1 expression in vivo was tested using siRNA against NF-YA. Two siRNAs targeting different regions of mouse NF-YA were transfected into NIH 3T3 cells, and protein levels were measured 48 h after transfection. One siRNA (siYA-2) drastically knocked down the expression level of NF-YA, whereas the other one (siYA-1) did not have any effect on NF-YA level (Fig. 6A). Consequently, the Bmal1 level in siYA-2 treated cells was also decreased significantly, but that in siYA-1-treated cells did not change (Fig. 6, A and B). When siRNAs targeting NF-YA were co-transfected with Bmal1 or Per1 promoter luciferase reporter, siYA-2 significantly knocked down both Bmal1 and Per1 promoter activity, whereas siYA-1 did not have any effect (Fig. 6, C and D). Overexpression of Bmal1 restored Per1 promoter activity in siYA-2-treated cells to control level (Fig. 6D).  It has been shown that Bmal1 expression is inhibited by another clock-controlled gene, Rev-Erb␣, which drives the circadian oscillation of Bmal1 expression (4). Although it is speculated that ROR␣ acts as the activator of Bmal1 transcription, mice with mutated ROR␣ (and embryonic fibroblasts originating from such mice) or cells with decreased ROR␣ level by RNA FIGURE 5. The cis-elements responsible for Bmal1 transcription activation and repression are separated. A, The d275 Bmal1-luc construct (WT), with a deletion from Ϫ77 to Ϫ118 (D118 -77), or with both ROR␣-binding sites mutated (mRORE) was transfected into NIH 3T3 cells. Luciferase activity was measured 48 h later. B, WT, D118 -77, or mRORE was transfected into NIH 3T3 cells and treated as in Fig. 1. C, WT, D118 -77, or mRORE was co-transfected with NF-Y and/or Rev-Erb␣ into NIH 3T3 cells. Luciferase (Luc) activity was measured 48 h later. Error bars in A and C indicate mean Ϯ S.E.  interference had Bmal1 expression level and rhythm comparable with that of wild type. Moreover, other results (12) and our results showed that ROR␣-binding sites were not sufficient to drive rhythmic Bmal1 expression as the promoter reporter constructs containing intact transcription start site and both ROR␣-binding sites had extremely low promoter activity and flattened oscillation, which was different from the conclusion drawn by Ueda et al. (13) when they fused a DNA fragment containing ROR␣-binding sites of Bmal1 promoter with SV40 promoter. This approach had two major flaws. First, it completely changed the configuration around ROR␣-binding sites in the promoter and the distance of ROR␣-binding sites from the transcription start site. Second, the SV40 promoter masked the need for the activating elements of the Bmal1 promoter. Moreover, mutating the ROR␣-binding sites resulted in significantly high promoter activity and dampened rhythm at peak level of the Bmal1 gene, indicating that the ROR␣-binding sites of Bmal1 promoter served primarily as negative-regulating elements for Rev-Erb␣ binding and that the role of ROR␣ activation in Bmal1 expression was marginal.
Deleting the proximal cluster or mutating the Y-box and two GC-boxes of the cluster greatly reduced Bmal1 level and flattened the oscillation of Bmal1 expression at the trough level, indicating that the activation of Bmal1 transcription was mainly mediated through the proximal Y-box/GC-box cluster. As NF-Y and Sp1 cooperatively regulate many genes (14 -18), it was not surprising to see that Bmal1 transcription was up-regulated by NF-Y and Sp1 through a tandem of one Y-box and two GC-boxes. NF-Y bound to CCAAT-box of various promoters and recruited RNA polymerase II and general transcription factors onto those promoters (7). NF-Y has been shown to regulate the transcription of genes involved in development (14), cancer (14,15), cell growth (16), and cell cycle control (17). A previous study (9) implicated a possible role for NF-Y in the regulation of the circadian clock, but we for the first time provided direct evidences that NF-Y activated Bmal1 transcription by binding to the CCAAT-boxes in Bmal1 promoter.
Based on previously published data and our results (4 -6, 12), we propose a new model for the clock work and Bmal1 regulation (Fig. 7). Unlike other clock genes, the oscillation of Bmal1 expression is solely achieved by the oscillation of negative regulator Rev-Erb␣. The high basal level of Bmal1 expression is driven by NF-Y/Sp1. Bmal1 and Clock activate the transcription of per1-3, cry1, cry2, Rev-Erb␣, and other clock-controlled genes, which is repressed by Per/Cry, whereas the transcription of Bmal1 is repressed by Rev-Erb␣.