CLIF, a Novel Cycle-like Factor, Regulates the Circadian Oscillation of Plasminogen Activator Inhibitor-1 Gene Expression*

The onset of myocardial infarction occurs frequently in the early morning, and it may partly result from circadian variation of fibrinolytic activity. Plasminogen activator inhibitor-1 activity shows a circadian oscillation and may account for the morning onset of myocardial infarction. However, the molecular mechanisms regulating this circadian oscillation remain unknown. Recent evidence indicates that basic helix-loop-helix (bHLH)/PAS domain transcription factors play a crucial role in controlling the biological clock that controls circadian rhythm. We isolated a novel bHLH/PAS protein, cycle-like factor (CLIF) from human umbilical vein endothelial cells. CLIF shares high homology with Drosophila CYCLE, one of the essential transcriptional regulators of circadian rhythm. CLIF is expressed in endothelial cells and neurons in the brain, including the suprachiasmatic nucleus, the center of the circadian clock. In endothelial cells, CLIF forms a heterodimer with CLOCK and up-regulates the PAI-1 gene through E-box sites. Furthermore, Period2 and Cryptochrome1, whose expression show a circadian oscillation in peripheral tissues, inhibit the PAI-1 promoter activation by the CLOCK:CLIF heterodimer. These results suggest that CLIF regulates the circadian oscillation of PAI-1 gene expression in endothelial cells. In addition, the results potentially provide a molecular basis for the morning onset of myocardial infarction.

The onset of myocardial infarction occurs frequently in the early morning. It may result in part from the circadian variation of fibrinolytic activity. Plasminogen activator inhibitor-1 (PAI-1) 1 is the major component of inhibitors of fibrinolysis activity. PAI-1 activity shows a clear circadian oscillation peaking in the early morning, and this may account for the morning onset of myocardial infarctions (1,2). However, the molecular mechanisms regulating this circadian oscillation remain unknown.
Every organism from bacteria to humans has an internal biological clock that adapts its activity to a circadian rhythm. Recently, the molecular mechanisms underlying these circadian processes are beginning to be elucidated (3)(4)(5). The biological clock is composed of transcriptional-translational feedback loops (3). Many of these components, including CLOCK, CYCLE, and PERIOD (PER), belong to the basic helix-loophelix (bHLH)/PAS domain family of transcription factors. Thus, the PAS domain plays a crucial role in regulating the biological clock (4). In mammals, CLOCK and BMAL1, the mammalian counterpart of Drosophila CYCLE, induce Per and Cryptochrome (Cry) gene expression (6). The PER and CRY proteins in turn act as negative components of the feedback loop by suppressing CLOCK:BMAL1-mediated transcription through CACGTG E-box enhancer elements (7,8).
The central circadian pacemaker in mammals is localized in the hypothalamic suprachiasmatic nucleus (SCN) (4). Recent studies have shown circadian oscillations of various transcripts in both peripheral tissues and cultured cells. Their underlying rhythm is inferred to be due to the same mechanisms that are present in the SCN (9 -11). These peripheral clocks are synchronized with the central clock by yet unidentified humoral factors (5,10,(12)(13)(14). The functional significance of the mammalian peripheral clocks is unknown.
We report here the identification of a novel bHLH/PAS domain transcription factor, CLIF (cycle like factor) that shares high homology with Drosophila CYCLE. Since CLIF is expressed in vascular endothelial cells, we hypothesize that CLIF may regulate the circadian oscillation of PAI-1 gene expression in endothelial cells. We show that CLIF forms a heterodimer with CLOCK and in endothelial cells up-regulates the PAI-1 gene through its E-box sites. Furthermore, PER2 and CRY1, whose expression show a circadian oscillation in peripheral tissues, inhibit the PAI-1 promoter activation by the CLOCK: CLIF heterodimer. This may account for the circadian oscillation of PAI-1 gene expression.

EXPERIMENTAL PROCEDURES
Construction of Plasmids-The endothelial PAS domain protein 1 (EPAS1) bait plasmid pBDGAL4-EPAS1 was constructed by amplification of cDNA encoding amino acids (aa) 30 -380 of the human EPAS1 protein by the polymerase chain reaction (PCR). The amplified fragment was cloned into the vector pBDGAL4 Cam (Stratagene). The plasmids for GAL4 DNA binding domain fusion proteins were constructed by amplification of cDNAs encoding bHLH/PAS domain (ARNT, aa 70 -508; BMAL1, aa 64 -445; CLIF, aa 74 -445) by PCR. The . The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF256215.
Yeast Two-hybrid Screening-A human umbilical vein endothelial cell (HUVEC) cDNA library was constructed by cloning HUVEC cDNAs into the vector HybriZAP-2.1 (Stratagene). The yeast strain YRG-2 (Stratagene) was made competent by the lithium acetate method (17). Approximately 10 10 yeast cells were transformed with 100 g of the EPAS1 bait plasmid pBDGAL4-EPAS1 and 100 g of library plasmid. The screening and sequencing of positive clones were performed as described previously (18).
Construction of Recombinant Adenoviruses-Recombinant adenovirus expressing CLIF (AdCMV.CLIF) was prepared by standard homologous recombination techniques using the replication-defective E1/E3deleted serotype 5 human adenovirus (19). Adenovirus expressing CLOCK (AdCMV.CLOCK) was generated using the AdEasy system as described previously (20). In brief, the full-length mouse CLOCK cDNA was cloned into the shuttle vector pAdTrack-CMV. After recombination with the vector pAdEasy-1 in bacteria, high titer virus was generated in 293 cells and purified by cesium chloride ultracentrifugation. The titer of the purified adenovirus was determined in 293 cells by plaque assay techniques.
RNA Isolation and Northern Analysis-Total RNA was prepared from mouse tissues and cultured cells with the RNeasy kit (QIAGEN). Northern analysis was performed as described previously (15). To obtain a probe for mouse CLIF, a fragment of the mouse CLIF cDNA was amplified by reverse transcriptase-PCR using primers based on the PAS domain sequences of the hCLIF cDNA (5Ј-AAGATGTTGCCAAAG-TAAAGGA-3Ј, 5Ј-ATTCCAGTTCTTTTGTCCAAGG-3Ј). The mouse CLIF cDNA shares 80.8% homology to hCLIF cDNA (data not shown).
In Situ Hybridization-In situ hybridization studies were performed with paraffin sections as detailed previously (21,22). Briefly, denatured, acetylated, and prehybridized tissue sections were hybridized for 12 h at 50°C with 250 ng/ml of 14-biotin-UTP-labeled hCLIF antisense and sense cRNA probes. The hCLIF riboprobes were transcribed from a 447-bp (nucleotides 1600 -2046) hCLIF cDNA fragment. After stringent washing and digestion of nonspecifically bound probe, the hybridized cRNAs were detected with VectorBrown substrate (Vector Laboratories). The sections were lightly counterstained with hematoxylin.
Gel Mobility Shift Assay-Gel mobility shift assays were performed using in vitro translated proteins as described previously (15). The 32-bp double-stranded oligonucleotide probe (5Ј-CTGGACACGTGGG-GAGACAATCACGTGGCTGG-3Ј) contains the two E-boxes derived from the sequence of the PAI-1 promoter.
Transient Transfection Assays-Bovine aortic endothelial cells (BAEC) were plated in six-well dishes and transfected with 15 ng of reporter construct and 1 g of expression constructs using Lipo-fectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions. To correct for variation in transfection efficiency, we cotransfected 15 ng of pCMV␤ (CLONTECH) in all experiments. Cell extracts were prepared 48 h after transfection by a detergent lysis method (Promega). Luciferase activity and ␤-galactosidase assays were performed as described previously (23). The ratio of luciferase activity to ␤-galactosidase activity in each sample served as a measure of normalized luciferase activity. Each construct was transfected at least four times, and each transfection was done in triplicate. Data for each construct are presented as the mean Ϯ S.E.

RESULTS AND DISCUSSION
Yeast Two-hybrid Screening-EPAS1 is a bHLH/PAS domain transcription factor that is expressed preferentially in vascular endothelial cells (24). To identify novel bHLH/PAS domain transcription factors in the cardiovascular system, we screened a HUVEC cDNA library by a yeast two-hybrid system using the HLH/PAS domain of EPAS1 as a bait. Eleven clones out of approximately 2 ϫ 10 6 independent transformants were positive for both HIS3 and lacZ. Of the 11 positive clones, two represented overlapping cDNAs encoding the same novel bHLH/PAS domain protein. We screened a HUVEC cDNA phage library using this cDNA as a probe and obtained 10 additional clones for further analysis. The open reading frame encoded a novel polypeptide of 602 amino acids with a calculated molecular mass of 67 kDa. The putative initiation codon (GCGATGG) is in agreement with the Kozak consensus sequence.
Homology and Phylogenetic Analysis of CLIF-Computer data base searches revealed that the amino acid sequence of this bHLH/PAS protein is homologous to Drosophila CYCLE and its mammalian homolog BMAL1 (44.1 and 44.9% identity, respectively) (25,26). Thus we designated this novel bHLH/ PAS protein as CLIF (CYCLE-like factor). The protein sequence alignment of hCLIF, hBMAL1, and Drosophila CYCLE revealed a particularly high degree of homology in the bHLH and PAS domains (Fig. 1A). Furthermore, a phylogenetic analysis within the ARNT superfamily revealed that CLIF and BMAL1 are most closely related to Drosophila CYCLE and diverge early from other members of the ARNT family (Fig.  1B). Since Drosophila CYCLE shows the highest homology to BMAL1 and CLIF with comparable identities, we conclude that CLIF is the second mammalian homolog of Drosophila CYCLE.
CLIF Specifically Interacts with EPAS1 and CLOCK-BMAL1 is known to heterodimerize with both EPAS1 and CLOCK (27). To determine whether CLIF also interacts with EPAS1 and CLOCK, quantitative yeast ␤-galactosidase assays were performed. Similar to BMAL1, CLIF interacted with EPAS1 and CLOCK, whereas ARNT interacted only with EPAS1 (Fig. 1C). This interaction of CLIF is specific because the combination of CLIF and other bHLH/PAS domain proteins, ARNT or BMAL1 was not sufficient to activate ␤-galactosidase activity in yeast (data not shown).
Expression Pattern of CLIF in Adult Tissues-To examine the expression pattern of CLIF, we performed Northern blot analysis on RNA isolated from human tissues. The human CLIF probe hybridized to four mRNAs of 8, 6, 2.4 and 2.2 kilobases. CLIF is expressed most abundantly in the brain and placenta, with lower expression in the heart, thymus, kidney, liver, and lung ( Fig. 2A). In contrast to the restricted tissue distribution of CLIF, ARNT and BMAL1 were expressed in many tissues, including the brain, heart, and skeletal muscles as reported previously ( Fig. 2A) (26,27).
We further performed in situ hybridization on human tissues to identify the cellular expression of CLIF. CLIF was expressed in endothelial cells and neurons in the brain, including the SCN, i.e. the center of the circadian clock (Fig. 2B). We also detected CLIF mRNA transcripts in endothelial cells in the heart, lung, and kidney (data not shown). Additional analysis by gel mobility shift assays demonstrated that CLIF formed a heterodimer with CLOCK and bound to the E-box within the Per1 promoter (data not shown). Furthermore, transient transfection of CLOCK and CLIF together transactivated the Per-1 promoter to the same extent as CLOCK and BMAL1 (data not shown) (6). Taken together, CLIF as a heterodimerization partner of CLOCK may contribute to the control of circadian rhythm in both the central and peripheral clocks.
The endothelial cell regulation of vascular tone and fibrinolytic activity shows circadian variation (28,29). For instance, the frequent onset of myocardial infarction during the early morning may partly result from the circadian variation of fibrinolytic activity (2,30). In the basal state, PAI-1 is produced mainly in endothelial cells, with the highest activity in the early morning. This elevated PAI-1 activity corresponds with the peak occurrence of myocardial infarctions (31). Until now, the molecular mechanisms regulating this type of circadian rhythm have not been elucidated. We hypothesize that peripheral endothelial cells also contain a biological pacemaker and that CLOCK and CLIF are the critical components that activate genes with circadian variations such as PAI-1.
Circadian Expression of PAI-1 mRNA in Peripheral Organs-To address this hypothesis, we first determined whether PAI-1 mRNA levels exhibited circadian variation. Indeed, Northern blot analysis of the mouse heart (Fig. 3A), kidney (Fig. 3B), brain, and lung (data not shown) showed circadian variation of PAI-1 mRNA levels with peak expression in the evening. This circadian oscillation pattern of mouse PAI-1 mRNA is antiphase to that of human PAI-1 activity, potentially due to the fact that rodents are nocturnal whereas humans are diurnal. To our knowledge, this is the first demonstration that the circadian oscillation of PAI-1 activity is regulated at mRNA level. The blots were then reprobed to analyze CLIF and BMAL1 expression in relation to PAI-1. These studies revealed that CLIF mRNA was constitutively expressed, whereas BMAL1 mRNA levels oscillated as reported previously (9) (Fig.  3C). The constant circadian expression pattern of CLIF is similar to that of Drosophila CYCLE (25). Per2 and Cry1 are important negative regulators of the biological clocks (32). Per2 and Cry1 expression followed a clear circadian oscillation (Fig.  3C). In these experiments, Clock was constitutively expressed as reported previously (data not shown) (33).
Overexpression of Clock and CLIF Induce PAI-1 mRNA Expression-We next examined whether the CLOCK:CLIF heterodimer is able to increase the endogenous PAI-1 mRNA in HUVEC. Using adenovirus-mediated gene transfer, we demonstrated that overexpression of CLOCK resulted in a dose-dependent increase in PAI-1 mRNA levels relative to the control cells infected with a GFP-expressing adenovirus (Fig. 3D). Coinfection of adenovirus expressing CLIF with CLOCK further increased the PAI-1 mRNA levels (Fig. 3D).
CLOCK and CLIF Transactivates the PAI-1 Promoter through the E-box Sites-To further elucidate the mechanisms by which CLOCK and CLIF increase PAI-1 mRNA levels, we Binding between SV40 T antigen (T) and p53 was used as a control. Yeasts were harvested in mid-log phase, and quantitative ␤-galactosidase assays were performed as described previously (18).

FIG. 2. Expression of CLIF mRNA in human tissues.
A, Northern blot analysis of human tissues. A, a commercially prepared multipletissue Northern blot (CLONTECH) was hybridized with the indicated probes. To correct for differences in RNA loading, the filters were rehybridized with a radiolabeled actin probe. Abbreviations: B, brain; H, heart; Sk, skeletal muscle; C, colon; T, thymus; Sp, spleen; K, kidney; Li, liver; I, intestine; P, placenta; Lu, lung; Le, leukocyte. B, in situ hybridization analysis of CLIF expression in the human brain. Human brain sections were hybridized with hCLIF antisense and sense cRNA probes labeled with biotin. The in situ hybridization signals were detected with VectorBrown substrate. The sections were lightly counterstained with hematoxylin. Arrow, endothelial cells; arrowheads, neuronal cells in the SCN. performed transient transfection assays using human PAI-1 promoter/luciferase reporter plasmids (16). Cotransfection of CLOCK and CLIF increased PAI-1 promoter activity by 5-fold (Fig. 4A). Deleting bp Ϫ800 to Ϫ550 of the PAI-1 promoter markedly diminished this induction by CLOCK and CLIF (Fig.  4B). Two E-box elements (CACGTG), which are the consensus binding sites of CLOCK and BMAL1 (6), are located at bp Ϫ677 to Ϫ672 and at bp Ϫ562 to Ϫ557. We generated point mutations in these E-boxes to determine whether they are important for the transactivation of PAI-1 promoter by CLOCK and CLIF. Indeed, mutation in either E-box decreased the transactivation of PAI-1 promoter (Fig. 4B). Mutation of both E-boxes nearly abolished the CLOCK and CLIF mediated transactivation (Fig.  4B). Thus, both E-box elements are important for the CLOCK: CLIF transactivation of the PAI-1 promoter.
To determine whether the CLOCK:CLIF heterodimer transactivates the PAI-1 promoter by directly binding to the Eboxes, we performed gel mobility shift assays with in vitro translated CLOCK and CLIF proteins and a labeled probe containing the PAI-1 E-boxes. In the presence of CLOCK, a weak shifted band was observed. In the presence of both CLOCK and CLIF proteins, however, robust DNA binding activity was detected (Fig. 4C).
PER2 or CRY1 Inhibits Transactivation of the PAI-1 Promoter by CLOCK:CLIF-In the central circadian pacemaker, PER and CRY function by inhibiting CLOCK:BMAL1-mediated transcription of their own promoter through CACGTG E-box enhancer elements (7,8). We found that Per2 and Cry1 mRNAs showed a clear circadian oscillation in peripheral tissues (Fig. 3B). Therefore, we asked whether PER2 or CRY1 could affect the CLOCK:CLIF activation of PAI-1. Cotransfection of PER2 partially inhibited CLOCK:CLIF-dependent transcription, while cotransfection of CRY1 abolished the induction of the PAI-1 promoter (Fig. 4D). These data suggest that PER2 and/or CRY1 suppress the CLOCK:CLIF-mediated transcription of PAI-1, resulting in the circadian oscillation of PAI-1 gene expression.
We identified that both E-boxes in PAI-1 promoter are responsible for its activation by CLOCK:CLIF. It is noteworthy that the E-box at bp Ϫ677 to Ϫ672 overlaps with the sequence of the 4G/5G polymorphism of the PAI-1 promoter, which correlates with serum levels of PAI-1 and the frequency of ischemic diseases (34,35). Thus, the 4G/5G polymorphism of PAI-1 promoter may affect the binding of CLOCK:CLIF to this E-box or interaction with adjacent transcription factors, resulting in an altered circadian expression of the PAI-1 gene. Therefore, it may be interesting to see the relation between the circadian oscillation pattern of PAI-1 activity and the 4G/5G polymorphism. . Total RNA was isolated 48 h after infection. Northern blot analysis was performed using a human PAI-1 probe. The blot was hybridized with human CLOCK and CLIF probes to confirm their expression and with 18 S rRNA to normalize for loading.

FIG. 4. CLOCK and CLIF transactivate the PAI-1 promoter through two E-box sites.
A, CLOCK and CLIF transactivate the PAI-1 promoter. The indicated expression plasmids (0.5 g each) were cotransfected into BAEC with the PAI-1 promoter (Ϫ800 bp)/luciferase reporter plasmid. B, mutation of the E-boxes abolishes the transactivation of the PAI-1 promoter. The E-box at bp Ϫ677 to Ϫ672 was mutated from CACGTG to GCTAGT (m1), and the E-box at bp Ϫ562 to Ϫ557 was mutated from CACGTG to TCGCTC (m2). Deletion series of PAI-1 promoter (Ϫ549 and Ϫ187) or mutated PAI-1 promoter (m1, m2, and m1ϩ2) reporter plasmids were transfected with the indicated expression plasmids. C, binding of CLOCK:CLIF heterodimer to the E-box of the PAI-1 gene. Gel mobility shift assays were performed as described under "Experimental Procedures" using the indicated in vitro translated proteins and a 32-bp double-stranded oligonucleotide (5Ј-CTGGA-CACGTGGGGAGACAATCACGTGGCTGG-3Ј) probe containing the two E-boxes derived from the sequence of the PAI-1 promoter. Binding experiments were also performed in the presence of unlabeled E-box consensus oligonucleotide (E-Box) or unrelated nonspecific oligonucleotide (NS) at 100-fold molar excess. D, PER2 or CRY1 inhibits transactivation of the PAI-1 promoter by CLOCK:CLIF. The indicated expression plasmids (0.3 g each) were cotransfected into BAEC with the PAI-1/luciferase reporter plasmid. A, B, and D, -fold induction represents the ratio (mean Ϯ S.E.) of luciferase activity in cells transfected with expression plasmid to that in cells transfected with empty vector (pcDNA3).
Recent data indicate that the mammalian circadian system is hierarchically organized, with self-sustained oscillators in the SCN entraining dampened oscillators in the periphery (12). However, the physiological significance of these peripheral oscillators remains unknown. Our results suggest that a peripheral tissue pacemaker regulates the circadian expression of PAI-1 gene directly. There are two heterodimers containing CLOCK; BMAL1:CLOCK dimer and CLIF:CLOCK dimer. The oscillation pattern and distribution of these two dimers are different in peripheral tissues. Therefore it may be possible that the combination of CLOCK:CLIF or CLOCK:BMAL1 together with PER and CRY function to regulate the peripheral pacemakers differently. Thus circadian rhythm in mammals may be regulated by both the central and peripheral tissue pacemakers.
In addition, since PAI-1 plays a crucial role in controlling fibrinolytic activity, the results potentially provide a molecular basis for the circadian variation of myocardial infarction.