A small-molecule inhibitor of Tcf/β-catenin signaling down-regulates PPARγ and PPARδ activities

The Wnt signaling pathway is important in normal development and in cancer (reviewed in refs. 1, 2). This signaling pathway is regulated by Wnt ligands, the APC-Axin complex and β-catenin. In development, signaling that stabilizes β-catenin is mainly mediated by the Wnt ligands. However, in human cancers such as hepatocellular and colorectal tumors, β-catenin stabilization is often the result of mutation in the tumor suppressor genes Axin and APC, or in the proto-oncogene β-catenin. β-Catenin stabilization leads to its accumulation and subsequent translocation to the nucleus, where it forms complexes with transcription factors of the Tcf/Lef family. The transcriptionally active β-catenin/Tcf complex exerts its cell proliferation and tumorigenic effects by promoting the transcription of growth controlling genes like c-Myc and cyclin D1. Wnt/β-catenin signaling pathway plays a central role in regulating the balance between stem cell growth and differentiation. Thus, the degree of Wnt signaling activation is an important modulator of the stem cells cancerous potential. In colorectal cancer, the necessary initiating APC or β-catenin mutations are not sufficient for maximum Wnt activation. Other intrinsic somatic mutations such as the oncogene Ras, as well as extrinsic factors like prostaglandins, are likely to play a rate-limiting role in Wnt signaling activation and contribute to the cancer development from the initial stem cell transformation to the metastasis stage (3). Increasing evidence shows that suppression of Wnt signaling can be achieved by targeting pathways that cross-talk with the Wnt signaling pathway. For example, both cyclooxygenase-2 and integrin-linked kinase small-molecule inhibitors attenuate β-catenin/Tcf–dependent transcription in colorectal cancer cells harboring mutated Wnt signaling (reviewed in ref. 4).

Peroxisome proliferator–activated receptors (PPAR) are members of the nuclear hormone receptor superfamily. The ligand-activated transcription by these receptors requires heterodimerization with retinoid X receptor, interaction with different coactivators as well as binding to PPAR-response elements (PPRE). There are three PPAR isotypes (α, γ, and δ), differing in their tissue distribution, physiologic functions, and ligand specificity. Fatty acids and their derivatives are natural PPAR agonists. Synthetic agonists have been reported for all PPAR isoforms, whereas antagonists have been identified only for PPARγ.

The physical interaction between β-catenin and PPARγ suggests a possible mechanism of cross-talk between the Wnt and the PPAR signaling pathways. On the transcriptional level, β-catenin enhances PPARγ activity whereas PPARδ is a target for β-catenin/Tcf regulation (5, 6). A clear interaction between these pathways is observed during adipogenesis. Adipogenic differentiation is regulated by reciprocal inhibitory signals between PPARγ and Wnt ligands; Wnt1 and Wnt10 promote the growth of preadipocytes whereas PPARγ agonists repress the Wnt/β-catenin signaling and advance their differentiation (7).

The contribution of PPARγ and PPARδ to Wnt/β-catenin–induced carcinogenesis remains unclear, as there are genetic and pharmacologic data suggesting that PPARs either promote or inhibit colon cancer (reviewed in refs. 8, 9). For example, treatment of Apc^min mice, which are predisposed to intestinal polyposis with the PPARγ agonist troglitazone, enhanced colon polyp development (10, 11), whereas treatment with pioglitazone, another PPARγ agonist, suppressed polyp formation (12). Similarly, conflicting results were observed using the PPARδ agonists GW501516 and GW0742 (13, 14).

Another class of compounds that modulate both Wnt/β-catenin and PPAR activity is the nonsteroidal anti-inflammatory drugs (NSAID). The pharmacologic action of NSAIDs has been attributed to their inhibition of cyclooxygenase activity. However, a large number of studies have suggested that the anticarcinogenic efficacy of NSAIDs is independent of their cyclooxygenase inhibition. Several NSAIDs, including indomethacin, weakly interact with PPARγ and stimulate its activity (15, 16). Indomethacin and other NSAIDs were also reported to suppress PPARδ activity (6, 17). The role of PPAR regulation to the overall ability of NSAIDs to repress the Wnt/β-catenin pathway is not fully understood.

Here, we report the identification of a low molecular weight compound (FH535) that suppresses β-catenin/Tcf–mediated transcription. FH535 behaves as a dual PPARγ and PPARδ antagonist that is able to inhibit GRIP1 and β-catenin recruitment. Comparisons between FH535 and the PPARγ antagonist GW9662, which allows β-catenin recruitment, suggest that inhibition of the Wnt/β-catenin pathway may require modulation of the interaction between PPARs and β-catenin.

The chemical library used for the screen is part of the DIVERSet collection from ChemBridge. ChemBridge was also the source for FH535 and its analogues. L165041 was purchased from Calbiochem-EMD Biosciences. All other PPAR ligands and nitric oxide–donating aspirin were obtained from Cayman Chemical.

Three copies of the optimized or mutated Tcf-binding element from TOPFLASH or FOPFLASH (18) driving a secreted alkaline phosphatase reporter gene were cloned into pCEP4 plasmid (Invitrogen), replacing the cytomegalovirus promoter. The plasmids were transfected into HepG2 cells, and hygromycin-resistant clones were pooled. Library screening was done at 20 μmol/L concentration in HepG2 serum-free media (19). Hits were tested in the HCT116 cell line for inhibition of TOPFLASH luciferase activity but not for inhibition of a reporter activity controlled from β-actin promoter (details of the reporter construct used in the screen and the stepwise characterization scheme for prioritization of hits is available in the Supplemental Materials).¹

TransIT-LT1 transfection reagent (Mirus) was used according to the manufacturer's instructions. After transfection of reporter constructs (18–24 h), cells were trypsinized and replated in drug-containing assay plates for another 24 h. Alkaline phosphatase and luciferase activities were measured using CSPD Emerald-II (Tropix-Applied Biosystems) or steadylite-HTS (Perkin-Elmer), respectively. When an internal control for transfection efficiency was required, the pCMV/β-galactosidase vector was cotransfected with the luciferase reporter constructs and β-galactosidase activity was assayed with Galacto-Light Plus (Tropix-Applied Biosystems). All assays were done in triplicate using a 96-well plate format and a TopCount-NXT luminescence counter (Packard).

The optimized and mutated Tcf-binding element–driven luciferase reporters (TOPFLASH, FOPFLASH) were provided by Dr. B. Vogelstein (Johns Hopkins Oncology Center, Baltimore, MD). The PPRE-driven luciferase reporter (ptk-PPRx3-luc; ref. 20) was a gift from Dr. R.M. Evans (Salk Institute, La Jolla, CA). The PPARγ protein fused to VP16 activation domain or the gal4 DNA-binding domain (pVP16-PPARγ2, pECE72gal4-PPARγLBD) and gal4-NCoR (pECE72gal4-NcoR) plasmids were made available by Dr. R.N. Cohen (University of Chicago, Chicago, IL). The gal4-GRIP plasmid (pSG424-mGRIP1; ref. 21) was a gift from Dr. F. Saatcioglu (University of Oslo, Oslo, Norway). Full-length mouse PPARδ coding sequence was amplified and fused in-frame downstream of VP16 activation domain in the pVP16 vector (Clontech) or downstream of the gal4 DNA-binding domain in the pSG424 vector (22). The PPARγC285A point mutation was introduced into the human PPARγ gene by replacing the MscI-BsaAI fragment with a PCR product amplified with the primers 5′-CCATCCGCATCTTTCAGGGCGCGCAGTTTCG-3′ and 5′-GTGCTCTGTGACGATCTGCCTGAGG-3′. The PPARδC248A point mutation was introduced into the mouse PPARδ gene by replacing the BstXI-BsaAI fragment with a fragment composed of 5′-GTGTTCTACCGGGCCCAGTCCACCACA-3′ and 5′-GTGGACTGGGCCCGGTAGAACAC-3′. The gal4/β-catenin fusion protein was constructed by cloning the gal4 DNA-binding domain upstream of the full-length human β−catenin gene in pCS2+ vector. The UAS–thymidine kinase reporter [p(UAS)5 tk-LUC] harbors five gal4-binding sites upstream of a minimal thymidine kinase promoter followed by the luciferase gene (a gift from S.J. Collins, Fred Hutchinson Cancer Research Center, Seattle, WA).

Cell lines were obtained from the American Type Culture Collection, except for the following cell lines: Huh7 was obtained from Dr. M. Katze (University of Washington, Seattle, WA). HCC15 and NCI-H1703 were a gift of Dr. A. Gazdar (UT Southwestern Medical Center, Dallas, TX). HCT116 PPARδ^+/+ WT and HCT116 PPARδ^−/− KO1 (23) were generously provided by Dr. B. Vogelstein (Johns Hopkins Oncology Center, Baltimore, MD). All cell lines were cultured at 37°C and 5% CO₂. The appropriate medium was supplemented with 10% fetal bovine serum (FBS), 2 mmol/L of l-glutamine, 100 units/mL of penicillin, and 100 μg/mL of streptomycin. In HCT116 serum-free medium, the FBS was substituted with 10 μg/mL of insulin, 5.5 μg/mL of transferrin, 35 nmol/L of sodium selenite, and 10 ng/mL of human epidermal growth factor (Life Technologies-Invitrogen). Cell viability was determined by the modified ³H-thymidine incorporation assay (24). Briefly, cells were plated in 96-well microplates for 24 h and treated in triplicate with various concentrations of the test compound. After 48 h of compound exposure, the cells were incubated for an additional 48 h in compound-free medium. The cells were then incubated in medium containing ³H-thymidine for 24 h, washed and mixed with the scintillant in the 96-well plate. Individual wells were counted with a 96-well scintillation counter (TopCount, Packard Instruments) and the LC₅₀ was calculated.

Total RNAs were extracted from HCT116 cells by the RNeasy kit and cDNA was synthesized with Moloney murine leukemia virus reverse transcriptase (Invitrogen). The primer sequences for amplification were: GAPDH-F 5′-ATGATCTTGAGGCTGTTG-3′, GAPDH-R 5′-CTCAGACACCATGGGGAA-3′; TCF-4F 5′-TTCAAAGACGACGGCGAACAG-3′, TCF-4R 5′-TTGCTGTACGTGATAAGAGGCG-3′. PCR amplification was done using Taq Polymerase (Fisher) for 20 cycles at 50°C (GAPDH) and 57°C (TCF-4) annealing temperatures.

To identify inhibitors of the Wnt/β-catenin pathway, we modified the existing TOPFLASH reporter system (18) to suit a high-throughput screen. Episomal reporter constructs containing three copies of either an optimized or mutated Tcf-binding element were stably transfected into HepG₂ hepatocellular carcinoma cells expressing high levels of nuclear β-catenin. Robust reporter activity was detected in clones containing the optimized element, which was 10-fold higher than activity in clones expressing the mutated element (data not shown).

Using this cell-based reporter system, we screened a diverse library of 11,600 low molecular weight compounds for the inhibition of β-catenin/Tcf–mediated transcription. Initially, two structurally related compounds (FH525 and FH614; Fig. 1A) were found to be potent inhibitors. Thirteen additional analogues were tested and another two active compounds were identified (FH535 and FH610; Fig. 1A). The nitro group located at the para or meta position in all four compounds is critical for function. Substitution of this nitro group with any of six other functional groups (OH, ethyl, ethylcarbonate, benzamide, O-benzyl, para-tolylsulfonamide) abolishes activity. In four cell lines harboring deregulation of the Wnt/β-catenin pathway, FH535 was found to be the best β-catenin/Tcf inhibitor (Fig. 1C). FH535 and two additional compounds (FH610 and FH614) were more active than nitric oxide–donating aspirin, a NSAID that has been shown to disrupt the formation of the β-catenin/Tcf complex at the drug concentration used here (ref. 25; Fig. 1C).

The newly identified β-catenin/Tcf inhibitors also share structural similarity to the known PPARγ antagonists GW9662 and T0070907 (refs. 26, 27; Fig. 1B). All of the compounds contain the nitro group, differing mainly in the central amide or sulfonamide groups as well as their orientation. Because PPARγ and PPARδ have been implicated in Wnt/β-catenin pathway regulation (reviewed in ref. 8), we tested the ability of our most active compound, FH535, to antagonize these PPARs.

First, we investigated the transactivation of a reporter construct containing three copies of the acyl-CoA oxidase PPAR-response element (20). In HCT116, this reporter is active and is sensitive to the addition of PPARγ and PPARδ agonists (Fig. 2). FH535 inhibits the reporter-dependent activity driven from natural PPAR ligands found in the cellular environment (Fig. 2A) as well as from the added ligands (Fig. 2B and C). A more direct assay for PPAR activity uses PPAR fused to the gal4 DNA-binding domain. Treatment with PPARγ or PPARδ agonists causes the activation of the UAS–thymidine kinase reporter (see Materials and Methods for details) in cells cotransfected with the relevant PPAR-gal4 chimera (Fig. 3A and B). This PPAR agonist–dependent transactivation is inhibited when FH535 is present.

We next evaluated the role of PPARδ in the inhibition of β-catenin/Tcf–mediated transcription by FH535. PPARδ-null HCT116 cells are more resistant to FH535 treatment than their matched paired cells expressing the wild-type PPARδ protein (Fig. 3C). However, deletion of PPARδ is not sufficient to fully counteract FH535 inhibition, suggesting that PPARδ is not the only target by which FH535 inhibits the Wnt/β-catenin pathway. Parallel to these findings, GW9662 is unable to inhibit β-catenin/Tcf–signaling regardless of PPARδ cellular levels (Fig. 3C).

The PPARγ antagonists GW9662 and T0070907 irreversibly modify Cys²⁸⁵ in the PPARγ ligand–binding site via a nucleophilic aromatic substitution of chlorine (26, 27). The same cysteine residue is essential for the activity and covalent binding of some PPARγ agonists such as 15d-PGJ2 but not for rosiglitazone (28). Among our four active compounds, only FH614 has a chemical structure capable of cysteine residue arylation (Fig. 1C). This strongly suggests that the antagonistic activity of our compounds does not require the modification of a cysteine. However, because PPARγ Cys²⁸⁵ and its equivalent Cys²⁴⁸ in PPARδ may be important for a noncovalent interaction with the compounds, we mutated this cysteine to alanine and found that these changes had no significant bearing on FH535 antagonism (Fig. 3A and B). Thus, FH535 activity does not require the same PPARγ binding residues that GW9662 uses. The use of different PPAR residues for ligand binding was also observed for the PPARγ agonists, rosiglitazone and lysophosphatidic acid (29).

We noticed that inhibition of PPRE-dependent activity in defined medium requires lower FH535 concentrations than in serum-containing medium (Fig. 2B and C). Because a variety of fatty acids and their metabolites are secreted into plasma and are naturally present in the serum, it is likely that serum contains PPAR agonists capable of suppressing the effects of FH535. The opposing relationship between FH535 and PPAR agonists with regard to PPAR transactivation (Fig. 3A and B), raises the possibility that PPAR agonists can also counteract FH535 inhibition of the Wnt/β-catenin pathway. First, we tested the effect of serum on this FH535 activity. FH535 is five times more active in defined medium than medium containing serum (Fig. 4A). Adding albumin to the defined medium reduced FH535 activity only 2-fold (data not shown). Next, 10 natural and synthetic PPAR ligands representing a broad spectrum of binding affinities were tested. Of those examined, lysophosphatidic acid was the only ligand found to reduce the inhibition activity of FH535 on the β-catenin/Tcf reporter (Fig. 4B). Lysophosphatidic acid is a pleiotropic growth factor–like lipid that mediates its effects through the activation of G protein–coupled receptors LPA1-4 and PPARγ (30). Thus, lysophosphatidic acid could either directly interfere with FH535 inhibition of PPARγ or the activation of lysophosphatidic acid receptors could indirectly suppress the activity of FH535.

To investigate the mechanism in which FH535 suppresses β-catenin/Tcf and PPAR-dependent transactivation, we focused on β-catenin and the coactivator GRIP1 because they are activators of both pathways (5, 31, 32). PPARs recruitment of these factors and the corepressors NCoR and SMRT was studied using a mammalian two-hybrid assay, in which activation of a gal4-dependent reporter is regulated by the interaction between VP16-PPARs with the gal4-transcription factor chimera. FH535 inhibits GRIP1 but not the corepressors recruitment to PPARδ (Fig. 5A). A similar pattern of repressing coactivator recruitment, whereas allowing corepressor binding, was observed for GW9662 (26). However, the recruitment of β-catenin to PPARγ is inhibited by FH535 but not by GW9662 (Fig. 5B). In this regard, GW9662 behaves like the vitamin D receptor antagonist, ZK159222, that is unable to prevent the recruitment of β-catenin to vitamin D receptors, whereas maintaining the ability to inhibit the recruitment of other coactivators (33).

Because PPAR ligand regulates the transcription of only a subset of genes, we looked for genes targeted by FH535. We focused on genes most important for Wnt/β-catenin signaling. β-Catenin levels were unaffected by FH535 (data not shown), whereas TCF4 transcription was suppressed in FH535-treated HCT116 cells (Fig. 5C).

Inhibition of Wnt/β-catenin signaling in some cancer cell lines that overexpress this pathway was shown to be toxic. For example, deletion of the activated mutant β-catenin in SW48 colon carcinoma cells is lethal but causes only a weak growth inhibition of HCT116 cells (34, 35). Because FH535 is an effective β-catenin/Tcf inhibitor (Fig. 1C), we tested its toxicity in 12 carcinoma cell lines expressing this signaling pathway (Table 1). The LC₅₀ of most carcinomas tested against FH535 was 5 to 15 μmol/L. At this range, β-catenin/Tcf–dependent transactivation is inhibited by >50% (Figs. 1C and 4A). Only the two colon carcinomas, SW48 and HCT116, were resistant to the compound's toxicity at concentrations of up to 30 μmol/L. This suggests that FH535 inhibition of β-catenin/Tcf–dependent transactivation does not extend to the other oncogenic functions of β-catenin essential for the survival of SW48 cells. Interestingly, FH535 is toxic to A549 and RKO cell lines that express low levels of β-catenin/Tcf–dependent activity but respond to treatment with Wnt ligands. Cells that do not express the Wnt/β-catenin pathway, the primary fibroblasts and the immortal intestine cell line IEC6, were unaffected by FH535 at concentrations of up to 30 μmol/L.

Several groups studying the therapeutic potential of PPARγ and PPARδ ligands to suppress Wnt/β-catenin pathway both in cell culture and in in vivo conditions reported conflicting results (reviewed in ref. 8). This suggests that only a specific type of PPAR modulation in a defined cellular environment could lead to the inhibition of Wnt/β-catenin pathway. Comparison of FH535 activities with other PPAR ligands provides an insight to the requirement for Wnt/β-catenin pathway inhibition.

FH535 and GW9662 share similar structure and both are antagonistic to PPARγ, but FH535 is unique in its ability to inhibit the Wnt/β-catenin pathway (Fig. 3C). One explanation for this difference is the target specificity of the compounds. FH535 antagonizes both PPARγ and PPARδ whereas GW9662 is specific for PPARγ. GW9662 is unable to reduce Tcf/β-catenin transactivation in cells expressing or lacking the PPARδ gene (Fig. 3C). In contrast, FH535 maintains partial antagonistic activity even in PPARδ-deficient cells (Fig. 3C). Thus, the inability of GW9662 to antagonize PPARδ cannot be the only reason for the difference between the two compounds, and suggests that they might also differ in the ability to inhibit the Wnt/β-catenin pathway via PPARγ.

GW9662 requires Cys²⁸⁵ to covalently bind PPARγ (26), whereas FH535's antagonistic activity does not depend on this cysteine residue (Fig. 3A). Similarly, the PPARγ agonist 15ΔPGJ2, but not rosiglitazone, use Cys²⁸⁵ for binding and transactivation (28). The nature of the interactions between the PPARγ ligands and specific residues in the PPARγ-binding domain leads to selectivity in coactivator recruitment. For example, recruitment of GRIP1 and SRC1 coactivators by 15ΔPGJ2 is superior compared with troglitazone, an analogue of rosiglitazone (31). Thus, lysophosphatidic acid, in contrast to the other PPAR agonists we tested (Fig. 3B), may direct the recruitment of specific coactivators that contribute to Wnt/β-catenin signaling regulation and are the targets of FH535 inhibition. Coactivator recruitment is believed to be a crucial step in PPAR-targeted gene activation because it is important for chromatin remodeling and for interaction with the basic transcription machinery (36). This suggests that the differences between FH535 and GW9662, with regard to binding PPARγ and β-catenin recruitment, may form the basis for the compounds' different abilities to regulate β-catenin/Tcf–dependent genes.

The ability of PPARγ to bind β-catenin and regulate β-catenin/Tcf activity is shared with other nuclear receptors like the orphan nuclear receptor LRH1 and the androgen receptor. LRH1 was suggested to serve as a direct coactivator for Tcf/β-catenin transactivation from the cyclin D1 promoter (37). Androgen receptor can sequester β-catenin from Tcf4 or promote β-catenin/Tcf4 interaction depending on the biological system analyzed (38, 39). The direct binding of androgen receptor to β-catenin and Tcf4 is thought to mediate the regulation of Wnt signaling. Like the androgen receptor, PPARγ interacts with β-catenin and is found in a complex with Tcf4 (5, 40). Thus, it is conceivable that FH535 inhibits a complex containing PPARγ β-catenin and Tcf/Lef proteins, depriving the complex bound to Tcf/Lef DNA sites from trans-activating. FH535 may achieve this by preventing the recruitment of GRIP1 to this complex because GRIP1 transactivates both PPARγ and Tcf/Lef proteins (31, 32). Alternatively, FH535 could suppress the transcription of selected PPARγ and PPARδ-targeted genes necessary for Wnt/β-catenin pathway activation. Tcf4 transcription is sensitive to FH535 inhibition and thus it is a good candidate for the transcriptional regulation of PPAR (Fig. 5C). The proposed mechanisms for FH535 activity are not mutually exclusive because the LRH1/β-catenin complex coactivates Tcf4 at the cyclin D1 promoter and also regulates cyclin E1 transcription by direct binding to the LRH1-RE site in the cyclin E1 promoter (37).

FH535 shows some functional similarity to R-etodolac, a stereoisomer of the NSAID etodolac. Unlike FH535, R-etodolac is a weak activator of the PPARγ-dependent reporter (16). However, in the presence of a strong PPARγ agonist, R-etodolac blocks the recruitment of the coactivator PBP by PPARγ. This antagonistic function of R-etodolac was suggested to explain its inhibition of the Wnt/β-catenin pathway (40). Diclofenac and indomethacin are also NSAIDs capable of both Wnt/β-catenin pathway inhibition and antagonizing strong activation of PPARγ (41, 42). All three compounds have been shown to bind and activate PPARγ. Thus, R-etodolac, diclofenac, and indomethacin are PPARγ ligands and may inhibit the Wnt/β-catenin pathway via their function as antagonists of PPARγ. Natural PPARγ agonists in the cancer cell environment could provide the necessary switch from a weakly agonistic to antagonistic function of these three compounds. The same natural PPARγ agonists can be expected to have the opposite effect on a true PPARγ antagonist's ability to suppress the Wnt/β-catenin pathway. This was observed for FH535 with the addition of either serum or lysophosphatidic acid (Fig. 4A and B).

FH535 inhibits Wnt/β-catenin signaling in cell lines that are resistant or sensitive to the toxic effect of FH535 (Fig. 1C; Table 1). FH535′s effect on cell viability is complex and depends on several factors. First, inhibition of Wnt/β-catenin signaling is not toxic to all cell lines with elevated Wnt signaling (34, 35). Additionally, PPAR antagonists have general toxicity to some cells. This toxicity is independent of the ability to inhibit the Wnt/β-catenin signaling pathway because GW9662 and T0070907 are toxic to some cancer cell lines (43). Also, the effect of FH535 depends on the production and cellular concentration of PPAR agonists that antagonize its activity (Fig. 4B).

In this article, we show that FH535 is a tool to study the cross-interaction between the Wnt/β-catenin and the PPAR signaling pathways. FH535 is a more potent Wnt/β-catenin inhibitor than NO-aspirin (Fig. 1C), a leading experimental chemopreventive compound against colon cancer that inhibits intestinal tumors in APC^min mice (17). We expect that therapeutic concentration of FH535 can be achieved in mice because T0070907 was shown to reduce the metastasis of injected cancerous cells in mice (44) and GW9662 suppressed mice obesity induced by a high-fat diet (45).

We thank Sondra Goehle and Kevin Schutz for technical assistance.