Discovery of peroxisome proliferator–activated receptor α (PPARα) activators with a ligand-screening system using a human PPARα-expressing cell line

Peroxisome proliferator–activated receptor α (PPARα) is a ligand-activated transcription factor that belongs to the superfamily of nuclear hormone receptors. PPARα is mainly expressed in the liver, where it activates fatty acid oxidation and lipoprotein metabolism and improves plasma lipid profiles. Therefore, PPARα activators are often used to treat patients with dyslipidemia. To discover additional PPARα activators as potential compounds for use in hypolipidemic drugs, here we established human hepatoblastoma cell lines with luciferase reporter expression from the promoters containing peroxisome proliferator–responsive elements (PPREs) and tetracycline-regulated expression of full-length human PPARα to quantify the effects of chemical ligands on PPARα activity. Using the established cell-based PPARα-activator screening system to screen a library of >12,000 chemical compounds, we identified several hit compounds with basic chemical skeletons different from those of known PPARα agonists. One of the hit compounds, a 1H-pyrazolo[3,4-b]pyridine-4-carboxylic acid derivative we termed compound 3, selectively up-regulated PPARα transcriptional activity, leading to PPARα target gene expression both in vitro and in vivo. Of note, the half-maximal effective concentrations of the hit compounds were lower than that of the known PPARα ligand fenofibrate. Finally, fenofibrate or compound 3 treatment of high fructose–fed rats having elevated plasma triglyceride levels for 14 days indicated that compound 3 reduces plasma triglyceride levels with similar efficiency as fenofibrate. These observations raise the possibility that 1H-pyrazolo[3,4-b]pyridine-4-carboxylic acid derivatives might be effective drug candidates for selective targeting of PPARα to manage dyslipidemia.

Peroxisome proliferator-activated receptor ␣ (PPAR␣) is a ligand-activated transcription factor that belongs to the superfamily of nuclear hormone receptors. PPAR␣ is mainly expressed in the liver, where it activates fatty acid oxidation and lipoprotein metabolism and improves plasma lipid profiles. Therefore, PPAR␣ activators are often used to treat patients with dyslipidemia. To discover additional PPAR␣ activators as potential compounds for use in hypolipidemic drugs, here we established human hepatoblastoma cell lines with luciferase reporter expression from the promoters containing peroxisome proliferator-responsive elements (PPREs) and tetracycline-regulated expression of full-length human PPAR␣ to quantify the effects of chemical ligands on PPAR␣ activity. Using the established cell-based PPAR␣-activator screening system to screen a library of >12,000 chemical compounds, we identified several hit compounds with basic chemical skeletons different from those of known PPAR␣ agonists. One of the hit compounds, a 1H-pyrazolo[3,4-b]pyridine-4-carboxylic acid derivative we termed compound 3, selectively up-regulated PPAR␣ transcriptional activity, leading to PPAR␣ target gene expression both in vitro and in vivo. Of note, the half-maximal effective concentrations of the hit compounds were lower than that of the known PPAR␣ ligand fenofibrate. Finally, fenofibrate or compound 3 treatment of high fructose-fed rats having elevated plasma triglyceride levels for 14 days indicated that compound 3 reduces plasma triglyceride levels with similar efficiency as fenofibrate. These observations raise the possibility that 1H-pyrazolo [

3,4-b]pyridine-4-carboxylic acid derivatives might be effective drug candidates for selective targeting of PPAR␣ to manage dyslipidemia.
Peroxisome proliferator-activated receptor ␣ (PPAR␣) 4 is a ligand-activated transcription factor that belongs to the nuclear hormone receptor superfamily. PPAR␣ binds to a direct repeat of two hexanucleotides, spaced by one nucleotide (DR1 peroxisome proliferator-responsive element (PPRE) motif), as heterodimers with the other nuclear receptor, retinoid X receptor (RXR), and subsequently induces target gene expression (1,2). PPAR␣ is mainly expressed in the liver, where it activates fatty acid oxidation and lipoprotein metabolism and improves plasma lipid profiles (lowering triglycerides and raising highdensity lipoprotein cholesterol in humans); PPAR␣ activators have been used to treat dyslipidemia (3)(4)(5). In addition, because residual risk factors of cardiovascular events, such as high triglycerides and low high-density lipoprotein cholesterol, need to be considered for the prevention of coronary events, PPAR␣ activators are thought to be candidates for reducing residual risk (6). Thus, PPAR␣ represents a drug target in the treatment of diseases involved in metabolic syndrome, among others, and PPAR␣ activators are effective against these diseases.
Screening for small molecules that enhance PPAR␣ transactivation activity is one approach for developing hypolipidemic drugs. PPAR␣ possesses four functional domains, including an N-terminal A/B domain containing a ligand-independent activation function 1, DNA-binding domain (DBD), hinge region, and C-terminal ligand-binding domain (LBD). The PPAR␣-RXR heterodimer binds to the PPRE located in the promoter of target genes via DBD, and ligand-bound LBD then associates with the coactivator complex, leading to expression of the target genes (2,7). Although the full-length crystal structure of the PPAR␣-RXR heterodimer has not been resolved, the crystal structure of the intact PPAR␥-RXR␣ heterodimer bound to PPRE with ligands and coactivator peptides has been solved (8,9). In the structure, the PPAR␥-LBD interacts directly with DBDs of both PPAR␥ and RXR␣; these interactions may affect their DNA-binding properties. The main interaction site domains located in the amino acid sequences between PPAR subtypes are well conserved; therefore, PPAR␣ and RXR␣ may form a complex on PPRE similar to the PPAR␥-RXR␣ heterodimer (8). From this viewpoint, a reporter gene assay using full-length PPAR␣ and the reporter plasmid containing the PPRE would provide a detection system for PPAR␣ activators under conditions resembling those in vivo.
In previous work, we established a human hepatoblastoma cell line tightly regulated by tetracycline (Tet) that can be induced to express full-length human PPAR␣ by removing Tet from the culture medium (HepG2-tet-off-hPPAR␣). Further, we identified numerous human PPAR␣ target genes (10 -13). In this study, we established a stable reporter cell line in which a reporter plasmid containing a putative PPRE was incorporated into the genomic DNA of HepG2-tet-off-hPPAR␣ cells to find PPAR␣ activators. We also screened a chemical library with the established human PPAR␣ reporter cell line and successfully identified several hit compounds with a basic skeleton different from those of known PPAR␣ agonists. Furthermore, one of the hit compounds, a 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivative, up-regulated PPAR␣ target genes in vitro and in vivo. When fructose-drinking rats with hypertriglyceridemia were treated with this compound, plasma triglyceride levels were reduced. These results suggest that the established reporter cell lines are useful cell-based screening systems for finding PPAR␣ activators and ameliorating metabolic syndrome.

Establishment of reporter cell lines that can be induced to express full-length human PPAR␣
To detect PPAR␣ activators as potential compounds for hypolipidemic drugs, we established reporter cell lines that could be used to quantify the effects of test compound-induced PPAR␣ activity. PPAR␣ binds to PPREs as a heterodimer with RXR and activates target gene transcription (2). Thus, to construct the PPAR␣-responsive reporter plasmid, the PPRE fragment was cloned into the upstream region of the SV40 promoter driving the luciferase reporter gene. To evaluate the specificity of ligands for PPAR␣, it is important to compare luciferase activity between reporter cell lines with or without PPAR␣ expression. Previously, we established a Tet-regulated human hepatoblastoma cell line that can be induced to express full-length human PPAR␣ via removal of Tet from the culture medium (HepG2-tet-off-hPPAR␣); the expression levels of PPAR␣ in this established cell line can be tightly controlled by modifying the concentration of Tet (10). Thus, to establish reporter cell lines, we transfected the reporter plasmids into HepG2-tet-off-hPPAR␣ cells, and cells were selected with blasticidin. Finally, we obtained a stable reporter cell line (HepG2tet-off-hPPAR␣-Luc) in which the reporter plasmid was incorporated into the genomic DNA of HepG2-tet-off-hPPAR␣ cells.
To determine whether a PPAR␣ ligand induces luciferase activity of this reporter gene, we cultured HepG2-tet-off-hPPAR␣-Luc cells in medium with or without Tet to regulate PPAR␣ expression. These cells were subsequently incubated with various concentrations of PPAR ligands and used for reporter gene assays (Fig. 1A). PPAR␣ ligands (fenofibric acid or Wy-14643) induced reporter gene luciferase activity in the absence of Tet (TetϪ), which allowed PPAR␣ expression (Fig.  S1A), but not in the presence of Tet (Tetϩ) (Fig. S1B). To evaluate PPAR␣ ligand specificity, luciferase activity in TetϪ cells was divided by the activity observed in Tetϩ cells. Induction of luciferase activity from the reporter gene via the liganded PPAR␣ was observed in a dose-dependent manner (Fig. 1B). In contrast, PPAR␦ and PPAR␥ ligands (GW501516 for PPAR␦ or ciglitizone for PPAR␥) did not affect luciferase activity of the reporter cells (Fig. 1B). Based on these results, we determined that the established reporter cell line is a useful detection system for PPAR␣ activators.

Identification of novel PPAR␣ activators using the established PPAR␣ reporter cell line
To identify potential compounds that stimulate PPAR␣ transactivation activity, a two-step screening process was performed using HepG2-tet-off-hPPAR␣-Luc cells. In the first step, HepG2-tet-off-hPPAR␣-Luc cells were incubated in TetϪ medium to allow PPAR␣ expression, and compounds that activated the reporter gene activity were identified. These cells were subsequently incubated with 12,467 compounds listed in the Osaka University compound library at a final concentration of 10 M for 24 h and used for the reporter gene assays. DMSO (a solvent) was used as a negative control, and GW7647 (PPAR␣ ligand) served as a positive control ( Fig. 2A). The top 300 compounds with activities greater than 50% were selected for further investigations. To evaluate the specificity of compounds for PPAR␣, HepG2-tet-off-hPPAR␣-Luc cells were then incubated in TetϪ medium to allow PPAR␣ expression and in Tetϩ medium to suppress PPAR␣ expression as a counterscreening. The cells were treated with DMSO, GW7647, or 300 individual compounds (10 M). After 24 h, reporter gene assays were performed (Fig. 2B). Interestingly, six of the top 10 compounds were 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivatives having a common chemical structure (Fig. 2C); these compounds have not been reported as PPAR␣ activators (Table 1).
Because PPAR␣ ligands up-regulate PPAR␣ activity through the LBD, we used a GAL4-hPPAR␣-LBD chimera reporter assay system to confirm the effects of nine of the 10 compounds (which we were able to repurchase) on PPAR␣ transactivation. We co-transfected both the GAL4-hPPAR␣-LBD expression

Discovery of novel selective PPAR␣ ligands
vector and the reporter plasmid into HepG2 cells, and these cells were subsequently incubated with each compound (10 M), fenofibrate (10 M), or DMSO for 24 h and used in the reporter gene assays. These nine compounds up-regulated PPAR␣ transcriptional activities; compounds with a common chemical structure (compounds 1 and 3-6) showed high activity (Fig. 3A). The activities of the compounds were completely inhibited in the presence of GW6471, a PPAR␣-specific antagonist (14), indicating that these compounds induced the transcriptional activity of PPAR␣ via the LBD (Fig. 3, B-J). For further analysis, we selected compounds having the highest activity (compounds 1 and 3) from those with a common chemical structure (compounds 1 and 3-6).

Compounds 1 and 3 are novel PPAR␣ ligands
Whereas three PPAR subtypes (␣, ␦, and ␥) have a closely conserved amino acid sequence in the LBD, the functions of each subtype are distinct (2,15). Thus, we examined the effect of subtype specificity and dose dependence of the compounds on the transcriptional activities of each PPAR subtype using the GAL4-hPPAR-LBD chimera reporter assay system. Com-pounds 1 and 3, which have a common chemical structure, induced PPAR␣ activity in a dose-dependent manner; the EC 50 for these compounds (2.06 and 1.78 M for compounds 1 and 3, respectively) was lower than that of fenofibrate (Ͼ21.84 M) (Fig. 4, A, B, and G). These compounds slightly enhanced PPAR␥-dependent transactivation and had no effect on PPAR␦ activity, indicating that these compounds are selective PPAR␣ activators. Similarly, other compounds (compounds 7-10) selectively activated PPAR␣; however, they were less effective activators than compounds 1 and 3 ( Fig. 4, C-F).
Because the activities of compounds 1 and 3 were higher than those of the other compounds, we also tested whether compounds 1 and 3, as well as fenofibric acid, an active form of fenofibrate, bind to the hPPAR␣-LBD. Results of a cellfree time-resolved FRET (TR-FRET) PPAR␣ competitive binding assay demonstrated that fenofibric acid binds to hPPAR␣-LBD by competitively and dose-dependently displacing Fluormone TM pan-PPAR green, as indicated by a reduction in the 518 nm/488 nm ratio; an IC 50 of 45.1 M was determined and agrees with a previous report (Fig. 5) (16). Compounds 1 and 3 also bound to hPPAR␣-LBD in a dose-dependent man-

Discovery of novel selective PPAR␣ ligands
ner, with IC 50 values of 12.9 and 11.2 M, respectively, indicating that the binding affinity of these compounds is about 4-fold stronger than that of fenofibric acid (Fig. 5). Furthermore, the TR-FRET PPAR␣ coactivator assay revealed that both compounds 1 and 3 recruited the coactivator peptide (fluorescein-PPAR␥ coactivator 1␣ (PGC1␣) coactivator peptide) to hPPAR␣-LBD in a dose-dependent manner (Fig. S2A). We have also performed reporter assays using the human solute carrier family 25, member 20 (SLC25A20) promoter driving the luciferase reporter gene (p4205 construct), which contains the PPAR-responsive element (12). We co-transfected either the PGC1␣ expression plasmid or the vector plasmid as a control together with the reporter plasmid and the hPPAR␣ expression plasmid into HepG2 cells. These cells were subsequently incubated with compounds 1 and 3, fenofibrate, or DMSO for 24 h and used for reporter gene assays. Induction of the luciferase activity of the human SLC25A20 reporter plasmid was observed upon treatment with the compounds as well as fenofibrate. The expression of PGC1␣ resulted in a further increase in promoter activity (Fig. S2B). These results indicate that compounds 1 and 3 (1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivatives) are novel PPAR␣ ligands and recruited at least PGC1␣ to PPAR␣-LBD and up-regulated PPAR␣ transcriptional activity.

Compound 3 induces PPAR␣ target gene expression in vitro and in vivo
We next performed a series of experiments confirming that 1H-pyrazolo[3,4-b]pyridine-4-carboxylic acid derivatives are efficient ligands for endogenous PPAR␣ target genes. As such, functional liver cell 4 (FLC4) cells were treated with compound 1 or 3 and then analyzed via real-time RT-PCR. Pyruvate dehydrogenase kinase 4 (PDK4) is a known PPAR␣ target gene (17); treatment with compounds 1 and 3 up-regulated the mRNA expression of PDK4 in a dose-dependent manner. In contrast, cyclophilin A, used as an internal control, was unaffected ( Fig. 6 and Fig. S3). Further, the effects of compound 3 on PDK4 mRNA expression in FLC4 cells were comparable with those of fenofibrate; however, fenofibrate was used at a maximum dose of 10 M, because of the cytotoxic effects in FLC4 cells (Fig. S4).
Next, we examined whether treatment with 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivatives regulates PPAR␣ target gene expression in vivo. Based on the above results, we chose compound 3 for use in further in vivo experiments. The liver is the main site of PPAR␣ expression; therefore, liver samples were collected from mice orally administered either compound 3 (10, 30, or 100 mg/kg/day), fenofibrate (100 mg/kg/ day), or vehicle alone for 7 days for RNA isolation. Compound 3 dose-dependently induced expression of PPAR␣ target genes in mouse livers, including acyl-CoA oxidase 1 (ACOX1),   SLC25A20, and peroxisomal biogenesis factor 11␣ (PEX11A) (17), in a manner similar to that observed for fenofibrate; ␤-actin, used as an internal control, was ineffective at inducing the expression of PPAR␣ target genes (Fig. 7). In addition, we evaluated the safety of compound 3. When compound 3 was administered to mice, body weight was not affected (Fig. S5A). Both compound 3 and fenofibrate increased liver weight (Fig. S5B). This increase may be the result of PPAR␣ activation (18). No change in aspartate aminotransferase (AST) was observed upon administration of compound 3. Although alanine aminotransferase (ALT) was not affected up to a dose of 30 mg/kg, ALT was elevated at high-dose administration (100 mg/kg/day) (but only around 50 IU/liter). These data indicate that compound 3 induces PPAR␣ target genes in vitro in human FLC4 cells and in vivo in mouse livers.

Compound 3 has potent hypolipidemic effects in fructose-fed rats
Finally, we examined whether 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivatives are potential compounds for use as hypolipidemic drugs by treating a hypertriglyceridemia animal model. To elucidate the effects of compound 3 on serum triglyceride levels, we used a well-established fructose-fed rat model (19). After fructose feeding (10% fructose in drinking water) for 2 weeks, serum triglyceride levels (191 Ϯ 49 mg/dl) of rats fed fructose (10% fructose in drinking water) for 2 weeks were higher than those from rats consuming normal drinking water (112 Ϯ 56 mg/dl). The fructose-fed rats were then divided into five groups and administered compound 3 or fenofibrate

Discovery of novel selective PPAR␣ ligands
for an additional 2 weeks along with the fructose (Fig. 8A, Administration period). As shown in Fig. 8A, administration of neither compound 3 nor fenofibrate affected body weight. However, fructose feeding affected liver weights (Fig. 8B, Nor-mal versus Vehicle), whereupon administration of compound 3 or fenofibrate resulted in an increase in liver weight (Fig. 8B). Nonetheless, serum liver transaminases (AST and ALT) were measured as indicators of hepatotoxicity, and serum levels of AST and ALT after treatment with compound 3 were within the normal range (Fig. S6, A and B). Other general biochemical markers, such as creatine kinase (CK; muscle damage marker) and blood urea nitrogen (UN; renal damage marker) were also within the normal range (Fig. S6, C and D). Moreover, serum triglyceride levels were lower in rats treated with compound 3 and fenofibrate under fructose-feeding conditions than those in the vehicle controls after 2 weeks (Fig. 8C). In contrast, no change in the serum levels of low-density lipoprotein cholesterol and high-density lipoprotein cholesterol was observed at this dosage (Fig. S6, E and F). After the 2-week treatment period, hepatic expression of PPAR␣ target genes, such as ACOX1, SLC25A20, and PEX11A, was enhanced in rats administered compound 3 and fenofibrate, whereas ␤-actin was not affected (Fig. 8, D-G). These data suggest that compound 3 attenuates hypertriglyceridemia in vivo.

Discussion
PPAR␣ binds to a PPRE as a heterodimer with RXR␣. In the presence of a ligand, helix 12 of the ligand-bound PPAR␣-LBD is then stabilized in the active conformation and associates with the coactivator complex (20). Although PPAR␣ can interact with various cofactors, some ligands (selective PPAR␣ modulators) induce ligand-specific PPAR␣-LBD structures, and these different conformations recruit specific cofactors (7,21). Moreover, studies of the full-length PPAR␥-RXR␣-PPRE complex structure show that PPAR␥-LBD interacts directly with both the DBD and RXR␣ and modulates the DNA-binding ability (8). From these viewpoints, it is important to examine the transcriptional activity of the full-length PPAR␣-RXR heterodimer when screening for drug discovery (8,9). In this study, we

Discovery of novel selective PPAR␣ ligands
applied this concept to develop a PPAR␣ ligand screening system using a previously established HepG2-tet-off-hPPAR␣ cell line that can be induced to express full-length human PPAR␣ (10). We then established a reporter cell line using full-length PPAR␣ and a reporter plasmid containing the PPRE; these cells then provided a detection system for PPAR␣ activators under conditions resembling those observed in vivo. Although further investigations are required, this screening enabled us to obtain compounds that bind to areas other than the LBD pocket and affect PPAR␣-RXR␣ interactions and transactivation properties (9).
Cell-based luciferase reporter gene assays are useful for drug discovery. However, unexpected factors such as transcriptional activation of the luciferase gene by target-independent pathways or factors influencing luciferase stability and turnover might affect luciferase activity during screening (22). Moreover, other members of the nuclear receptor family, such as PPAR␦, PPAR␥, hepatocyte nuclear factor-4 (HNF4), chicken ovalbumin upstream promoter transcriptional factor I (COUP-TFI), COUP-TFII, or thyroid hormone receptor (TR) might bind to the PPREs and potentially affect PPAR␣ transcriptional activity (23)(24)(25)(26). PPAR␣ expression was not induced by a Tet-controlled transactivator with Tet in the Tet-off system; therefore, we used our established reporter cells cultured in Tet-containing medium as a counterscreening assay (10,27). PPAR␣ expression was induced in the reporter cells after removal of Tet from the culture medium, enabling us to evaluate the specificity of test compounds for activating PPAR␣ by comparing luciferase activity with or without Tet in the culture medium. Indeed, we evaluated several known ligands using this reporter cell line and confirmed that the cells were useful for detecting PPAR␣ ligands (Fig. 1). Based on these results, the established reporter cell line is a powerful tool for screening PPAR␣ activators. In addition, this technology can be applied for screening activators targeting other nuclear receptors.
LBDs of the PPARs containing 13 ␣-helices and small fourstranded ␤-sheets are very large (about 1400 Å) and exhibit Y-shaped pockets (20,28). The large ligand-binding pockets bind with various types of ligands, such as fatty acids and synthetic fibrates. Usually, PPAR ligands have three regions, including an acidic head part, aromatic linker part, and hydrophobic tail part (29). The acidic head part of the ligand interacts with the arm I region, including a residue of helix 12 of the LBD, and the hydrophobic tail part interacts with the arm II or III region (30). In this study, we screened a chemical library using established reporter cells and identified PPAR␣ ligand hit compounds having a common 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivative skeleton ( Fig. 2C and Table 1). For these compounds, both the carboxylic acid (the acidic head part) and hydrophobic aromatic tail directly bound to the pyrazolopyridine ring (Fig. S7); the structure of these compounds differs from those of typical PPAR ligands. However, 1H-pyrazolo [3,4b]pyridine-4-carboxylic acid derivatives might bind to the PPAR␣ ligand-binding pocket, as determined via TR-FRET PPAR␣ competitive binding assays and luciferase-based antagonist assays. In addition, TR-FRET PPAR␣ coactivator assays revealed that 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivatives recruited PGC1␣ to PPAR␣ and activated it. However, it is still unclear whether other coactivators, such as steroid receptor coactivator-1 (SRC-1), will be recruited to PPAR␣ by these compounds. Because the PPAR␣ structure complexed with the ligand and recruited specific cofactors to selectively modulate the target genes (5, 7), it will be important to elucidate the structure of PPAR␣-LBD and 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivatives in the future.
Compound 3, a 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivative, up-regulated PPAR␣ transcriptional activity in a dose-dependent manner both in vivo and in vitro. Although ALT increased at a high dose (100 mg/kg/day) of compound 3 in the mouse study (levels were around 50 IU/liter), it was not affected up to a dose of 30 mg/kg, and the PPAR␣ activity was Male rats received either a 10% fructose challenge in the drinking water or normal drinking water (Normal). Two weeks later, fructose-fed rats were orally treated with compound 3 (1, 3, or 10 mg/kg/day), fenofibrate (30 mg/kg/day), or 0.5% methylcellulose (Vehicle) for 2 subsequent weeks (Administration period). A, rat body weights. There were no significant differences in body weight between the vehicle and treatment groups. B, relative liver weights. C, the mean percentage change from baseline in serum triglyceride (TG) levels after 2 weeks of each treatment. D-G, effects of compound 3 on hepatic mRNA levels of PPAR␣ target genes. ACOX1 (D), SLC25A20 (E), PEX11A (F), and ␤-actin (G) mRNA levels in rat liver were measured using real-time RT-PCR and normalized to ␤ 2 -microglobulin mRNA relative to the vehicle set as 1. Values are expressed as means Ϯ S.D. (error bars) (n ϭ 5). The small black dots are data points. Significant differences between the values compared with the vehicle were determined using a two-sample t test (Normal versus Vehicle; † †, p Ͻ 0.01; † † †, p Ͻ 0.001) or Dunnett's test (Vehicle versus treatment; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001).

Discovery of novel selective PPAR␣ ligands
observed (the expression levels of the target genes were increased) at this dose. When high fructose-supplemented rats with elevated plasma triglyceride levels were treated with compound 3 or fenofibrate for 14 days, an increase in liver weights was observed (Fig. 8B), probably attributable to PPAR␣ activation (18). However, the levels of AST and ALT were normal (Fig. S6). Furthermore, treatment with compound 3 reduced plasma triglyceride levels (Fig. 8C). These results suggest that 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivatives might be effective drugs to treat hyperlipidemia and reduce residual risk. In addition, recent studies have suggested that PPAR␣ is a good therapeutic target for nonalcoholic steatohepatitis (31)(32)(33). Although further investigations are needed, this compound might be a drug candidate in the treatment of metabolic syndrome and nonalcoholic steatohepatitis.
In conclusion, we engineered reporter cell lines that can be used to quantify ligand-induced PPAR␣ activity using a Tetregulatable human hepatoblastoma cell line that can be induced to express full-length human PPAR␣. By screening a chemical library with this cell-based PPAR␣-activator screening system, we successfully identified several hit compounds with a basic skeleton different from those of known PPAR␣ agonists. One of the hit compounds, a 1H-pyrazolo [3,4-b]pyridine-4-carboxylic acid derivative, up-regulated PPAR␣ transcriptional activity and induced PPAR␣ target genes in vitro and in vivo. Further, it reduced serum triglyceride levels in fructose-fed rats. These data suggest that these reporter cell lines provide a powerful detection system for PPAR␣ activators as potential compounds for hypolipidemic drugs.

Plasmid construction
A luciferase reporter plasmid containing PPRE was generated using standard cloning techniques. Two copies of the ACOX1 PPRE-coding oligonucleotide, which is a PPAR␣-binding element (35), were constructed by annealing the forward oligonucleotide 5Ј-CCGGACCAGGACAAAGGTCACGTTC-GGACCAGGACAAAGGTCACGTTCGGAGCT-3Ј (including a PPRE, underlined) and reverse oligonucleotide 5Ј-CCGAAC-GTGACCTTTGTCCTGGTCCGAACGTGACCTTTGTCC-TGGTCCGGGTAC-3Ј. The annealed oligonucleotide was cloned into the KpnI-SacI sites of a PGV-P2 vector (Toyo Ink, Tokyo, Japan). This plasmid was digested with KpnI and bluntended with T4 DNA polymerase. The luciferase reporter gene containing the PPRE in the promoter region was then released from this plasmid via digestion with BamHI; this fragment was then inserted into the SmaI-BamHI sites of a pEF-Bsd vector (Invitrogen) to generate a luciferase reporter plasmid (termed pEF-Bsd-PPREx2-Luc). All constructs were verified by sequencing.

Generation of a stably transfected human PPAR␣ reporter cell line
HepG2-tet-off-hPPAR␣ cells were transfected with pEF-Bsd-PPREx2-Luc using TransIT-LT1 transfection reagents (Mirus). Stable cell lines were isolated using 4 g/ml blasticidin S (Kaken Pharmaceutical, Tokyo, Japan). These clones were further screened by checking the luciferase activity; one clonal cell line demonstrated a high-level expression of luciferase via a PPAR␣ ligand and was designated as the HepG2-tet-off-hPPAR␣-Luc cell line and used for subsequent experiments.

Luciferase assays using a human PPAR␣ reporter cell line
HepG2-tet-off-hPPAR␣-Luc cells (1.3 ϫ 10 4 cells/well) were seeded in 96-well plates and incubated in DMEM supplemented with 10% charcoal dextran-treated FBS with or without 2 g/ml Tet. The cells were treated with various concentrations of compounds. Firefly luciferase activity was quantified using a luciferase assay system (Promega) and a luminometer (Berthold Technologies, Bad Wildbad, Germany). To evaluate the specificity of compounds for PPAR␣, luciferase activity levels were determined in cells cultured in TetϪ medium and divided by those observed in cells cultured in Tetϩ medium (see below).

Two-step screening using a human PPAR␣ reporter cell line
To identify PPAR␣ activators, a two-step screening process was performed. In the first step, HepG2-tet-off-hPPAR␣-Luc cells (4 ϫ 10 4 cells/well) were seeded in 384-well plates and incubated in DMEM supplemented with 10% charcoal dextrantreated FBS without Tet for 24 h. The cells were treated with 0.1% DMSO (negative control), 1 M GW7647 (positive control), or the screening compounds (10 M each). After 24 h, firefly luciferase activity was measured using a Steady-Glo luciferase assay system (Promega) and the functional drug screening system 7000 (FDSS7000, Hamamatsu Photonics, Hamamatsu, Japan). Activity of the test compounds was calculated using the following formula.

Discovery of novel selective PPAR␣ ligands
In the second step, HepG2-tet-off-hPPAR␣-Luc cells (4 ϫ 10 4 cells/well) were seeded in 384-well plates and incubated in DMEM supplemented with 10% charcoal dextran-treated FBS with or without 2 g/ml Tet for 24 h. The cells were treated with 0.1% DMSO (negative control), 1 M GW7647 (positive control), or the screening compounds (10 M each) in duplicate. After 24 h, firefly luciferase activity was measured as described above. The average luciferase activity for each compound was calculated from duplicate runs. To evaluate the specificity of compounds for PPAR␣, luciferase activity of cells cultured in TetϪ medium was divided by that from cells cultured in Tetϩ medium. Activity of the test compounds was calculated using the following formula.

Animal treatments
Mice and rats were used in this study. Animal experiments using mice were performed at TransGenic Inc. (Kobe, Japan). Male C57BL/6J mice were maintained in an air-conditioned room (20 -26°C) with a 12-h light/dark cycle and were given free access to food (standard chow) and water. Mice (8 -10 weeks old) were administered compound 3 (10, 30, or 100 mg/kg body weight/day), fenofibrate (100 mg/kg body weight/ day), or 0.5% methylcellulose (400 centipoise, Wako Pure Chemicals) via gavage for 7 days. All mice were killed under anesthesia 24 h after the final administration, and the livers were isolated and stored at Ϫ80°C until further analysis.
Animal experiments using rats were performed at Shin Nippon Biomedical Laboratories, Ltd. (Kagoshima, Japan). Male Sprague-Dawley rats were maintained in an air-conditioned room (22°C) with a 12-h light/dark cycle and were given free access to food (standard chow) and water. Rats were randomly divided into two groups, including a normal group receiving untreated drinking water and a fructose-challenge group receiving 10% fructose in the drinking water for 2 weeks to establish a hyper-triglyceridemic animal model (19). After 2 weeks, blood was collected from the external jugular vein to evaluate serum triglyceride levels. The fructose-supplemented rats were then divided into five groups according to serum triglyceride levels. These groups were administered compound 3 (1, 3, or 10 mg/kg body weight/day), fenofibrate (30 mg/kg body weight/day), or 0.5% methylcellulose (Metolose SM-400, Shin-Etsu Chemical Co., Tokyo, Japan) via oral gavage, as well as a 10% fructose challenge through the drinking water for 2 subsequent weeks. A normal group was also administered 0.5% methylcellulose via daily oral gavage for 2 weeks. Body weights were measured weekly in the fed state. At the end of the experimental period (8 weeks old), blood was collected from the abdominal aorta, and the livers were collected and weighed under anesthesia. A portion of the left lateral lobe was excised and stored at Ϫ80°C until further analysis. Serum triglyceride levels were enzymatically measured using the glycerol phosphate oxidase-N-(3-sulfopropyl)-3-methoxy-5-methylaniline glycerol elimination method. The percentage change in serum triglyceride levels from baseline was obtained from the differences between pre-and post-treatment levels divided by the pretreatment levels and multiplied by 100.
All animal experiments were approved by the Experimental Animal Care and Use Committee at Osaka University, Trans-Genic Inc., and Shin Nippon Biomedical Laboratories, Ltd.

RNA extraction and quantitative real-time RT-PCR
Real-time RT-PCR was performed as described previously (12). Briefly, total RNA was isolated from cultured cells using the QuickGene RNA cultured cell HC kit S (KURABO, Osaka, Japan) or from frozen livers using ISOGEN with a spin column (Nippon Gene, Tokyo, Japan) according to the manufacturer's instructions. First-strand cDNA was synthesized from total RNA from each sample using the SuperScript TM first-strand synthesis system for RT-PCR (Invitrogen). The cDNAs were used as templates for individual PCRs using specific primer sets (Table S1). PCRs were carried out using a QuantiTect TM SYBR Green PCR kit (Qiagen). ␤ 2 -Microglobulin was used to normalize each expression data set.

LanthaScreen TM TR-FRET PPAR␣ competitive binding assays
The LanthaScreen TM TR-FRET PPAR␣ competitive binding assay was performed according to the manufacturer's instructions (Invitrogen). Test compounds were incubated for 3 h at room temperature with GST-hPPAR␣-LBD (5 nM), LanthaScreen TM terbium-labeled anti-GST antibody (5 nM), and Fluormone TM pan-PPAR green (20 nM). The TR-FRET emission was measured with a SpectraMax M5e microplate reader (Molecular Devices); results are expressed as the ratio of fluorescence intensity at 518 nm (fluorescein emission excited by terbium emissions) and 488 nm (terbium emissions).

Statistical analysis
Data from dose-response experiments were analyzed with the add-on package for dose-response curves (drc) for R, and a four-parameter log-logistic model was fitted to the data as selected by the drc modelFit function (37). Statistical analyses

Discovery of novel selective PPAR␣ ligands
were performed using a two-sample t test or Dunnett's multiple-comparison test using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan) (38), which is a graphical user interface for R (R Foundation for Statistical Computing, Vienna, Austria). More precisely, it is a modified version of R commander designed to add statistical functions that are frequently used in biostatistics.