EP1203101A1 - Ppar delta links apc to chemopreventive drugs - Google Patents

Abstract

PPARδ was identified as a target of APC suppression through the analysis of global gene expression profiles in human colorectal cancer cells. PPARδ expression is elevated in primary colorectal cancers and significantly repressed by APC in colorectal cancer cells. This repression is mediated by two Tcf-4-responsive elements in the PPARδ promotor. Reporters containing PPARδ-responsive elements are repressed by sulindac, a non-steroidal anti-inflammatory (NSAID) agent which can reduce the size and number of colon tumors in humans and animals with APC mutations. Furthermore, sulindac is able to specifically disrupt the ability of PPARδ to bind its cognate recognition sequences. These findings suggest a model wherein NSAIDs inhibit tumorigenesis through post-transcriptional modification of a gene that is normally regulated by APC. This novel molecular target for NSAIDs can be used to develop more effective chemopreventive agents for colorectal tumors.

Description

PPARδ LINKS APC TO DRUGS

This invention was made using grant funds from the U.S. National Institutes of Health (CA57345 and CA62924). Therefore the government retains some rights in the present invention.

TECHNICAL FIELD OF THE INVENTION

The invention relates to the area of cancer and gastroenterological therapeutic agents. More particularly, the invention relates to the area of screening assays for therapeutic agents.

BACKGROUND OF THE INVENTION With an estimated 150,000 new cases and 55,000 deaths per year, colorectal cancer is the second leading cause of cancer deaths in the United States. Over half of the U.S. population will develop a colorectal tumor during their lifetime, and these tumors will progress to malignancy in approximately 10% of the cases. The high prevalence of this disease and aging nature of the population make effective prevention an important public health and economic concern.

The gradual progression of a normal colon epithelial cell to a malignant tumor occurs over several decades. During this time, the neoplasm progresses through a series of well-recognized histopathological stages, from microscopic lesions to grossly visible benign tumors (adenomas) to widely disseminated malignant disease (carcinomas). Molecular genetic studies have identified a series of genetic alterations that commonly underlie this progression (reviewed in Kinzler and Vogelstein, 1996).

In terms of prevention, the alterations that occur early in this process are of most interest. Alterations of the APC tumor suppressor pathway are the earliest genetic alterations known to occur in these tumors, and are likely to represent the initiating event. Most colorectal carcinomas and adenomas carry inactivating mutations of the APC gene. Likewise, inherited mutations of APC cause FAP, characterized by the development of hundreds to thousands of colorectal adenomas (reviewed in Kinzler and Vogelstein, 1996).

Recent studies have started to provide insights into how APC might exert its tumor suppressive effects. Although APC has been shown to associate with at least a dozen proteins, its association with β-catenin seems to be of special importance for its tumor suppressor function. The β-catenin protein was originally identified through its association with E-cadherin and role in cellular adhesion (reviewed in Kemler, 1993).

More recently it has been recognized that β-catenin has a separate cellular role, serving as a signal transducer in the Wg/WNT pathway. In the colon, β-catenin binds to the

Tcf-4 transcription factor, providing a domain which activates genes containing Tcf-4 binding sites in their regulatory regions (Behrens et al, 1996; Molenaar et al, 1996). The product of the wild type APC gene inhibits this β-catenin/Tcf-4 mediated transcription, while disease-associated APC mutants are deficient in this ability

(Korinek et al, 1997; Morin et al, 1997). APC inhibition of β-catenin Tcf-4 mediated transcription is accomplished through the binding of β-catenin to APC (Rubinfeld et al,

1993; Su et al, 1993). This binding facilitates the phosphorylation of β-catenin by the serine/threonine kinase GSK3β, leading to β-catenin degradation by ubiquitin-dependent proteolysis (Aberle et al, 1997; Munemitsu et al, 1995; Orford et al, 1997; Rubinfeld et al, 1996).

The importance of this pathway in colorectal tumorigenesis has been substantiated by the identification of oncogenic mutations of β-catenin in a significant fraction of those colon cancers that lack APC mutation (Morin et al, 1997; Ilyas et al,

1997; Iwao et al, 1998; Kitaeva et al, 1997; Sparks et al, 1998). These oncogenic forms of β-catenin are mutated within the GSK3β phosphorylation domain and are thereby able to cause constitutive activation of Tcf-4/β-catenin transcriptional activity in the presence of intact APC (Morin et al, 1997; Rubinfeld et al, 1997). In the great majority of colorectal cancers, mutation of APC or β-catenin leads to loss of this inhibition, resulting in increased β-catenin/Tcf-mediated transcription of downstream target genes that are likely to be essential for cell proliferation. Identification of these downstream target genes is therefore crucial for understanding the mechanisms through which APC regulates cell proliferation and functions as a tumor suppressor and for developing effective therapeutics for cancer chemoprevention.

SUMMARY OF THE INVENTION It is an object of the present invention to provide tools and methods of screening for potential cancer therapeutics. This and other objects of the invention are provided by one or more of the embodiments described below.

One embodiment of the invention is an isolated subgenomic polynucleotide comprising a PPARδ binding element comprising a nucleotide sequence selected from the group consisting of SEQ ID NOS:l-21 and nucleotides 3-9 of SEQ ID NO:21

(Figure 3B) and an RXR binding element comprising a nucleotide sequence selected from the group consisting of SEQ ID NOS:22-42 and nucleotides 3-7 of SEQ ID NO:50 (Figure 3A).

Another embodiment of the invention is an isolated subgenomic polynucleotide comprising at least 2 copies of a PPARδ binding element comprising a nucleotide sequence selected from the group consisting of SEQ ID NOS:l-21 and nucleotides 3-9 of SEQ ID NO:21 (Figure 3B).

Even another embodiment of the invention is a nucleic acid construct comprising at least one PPARδ binding element comprising a nucleotide sequence selected from the group consisting of SEQ ID NOS: 1-21 and nucleotides 3-9 of SEQ

ID NO:21, a minimal promoter, and a reporter gene. The PPARδ binding element is upstream from the minimal promoter, and the minimal promoter is upstream from the reporter gene. The minimal promoter regulates transcription of the reporter gene.

The invention provides a method of pre-screening agents for therapeutic use. Binding of a PPARδ protein to a DNA molecule comprising a PPARδ binding element is measured in the presence and in the absence of a test substance. The amount of binding of the PPARδ protein in the presence of the test substance is compared to amount of binding of the PPARδ protein in the absence of the test substance. A test substance which decreases the amount of binding is a candidate agent for use in cancer therapy. A test substance which increases the amount of binding is a candidate agent for ameliorating negative side effects of NSAIDs. The invention provides another method of pre-screening agents for therapeutic use. A transfected cell is contacted with a test substance. The transfected cell contains a PPARδ protein and a reporter construct comprising a reporter gene. The reporter gene encodes an assayable product, a minimal promoter upstream from and regulating transcription of the reporter gene, and at least one copy of a PPARδ binding element upstream of the minimal promoter. Whether expression of the reporter gene is decreased or increased by the test substance is determined. A test substance which decreases the amount of expression of the reporter gene is a candidate agent for use in cancer therapy. A test substance which increases the amount of expression of the reporter gene is a candidate agent for ameliorating negative side effects of NSAIDs.

The invention provides yet another method of pre-screening agents for therapeutic use. RNA polymerase, ribonucleotides, and PPARδ protein are added to a reporter construct. The reporter construct comprises a reporter gene which encodes an assayable product and at least one copy of a PPARδ binding element upstream from a minimal promoter. The minimal promoter is upstream from and controls transcription of the reporter gene. The step of adding is effected in the presence and absence of a test substance. Whether transcription of the reporter gene is decreased or increased in the presence of the test substance is determined. A test substance which decreases the amount of transcription of the reporter gene is a candidate agent for use in cancer therapy. A test substance which increases the amount of transcription of the reporter gene is a candidate agent for ameliorating negative side-effects of NSAIDs.

Even another embodiment of the invention is a method of identifying candidate drugs for use in FAP patients, patients with APC or β-catenin mutations, or patients with increased risk of developing cancer. A cell having no wild-type APC or a mutant β-catenin is contacted with a test compound. Transcription in the cell of a Tcf- responsive reporter gene is measured. The Tcf-responsive reporter gene comprises a Tcf-4 binding element selected from the group consisting of CTTTGAT (TREl) and CTTTCAT (TRE2). A test compound which decreases transcription of the reporter gene is a candidate drug for cancer therapy. Still another embodiment of the invention is a method of identifying candidate drugs for use in for use in FAP patients, patients with APC or β-catenin mutations, or patients with decreased risk of developing cancer. A Tcf-responsive reporter gene is contacted with a test compound under conditions in which the reporter gene is transcribed in the absence of the test compound. The Tcf-responsive reporter gene comprises a Tcf-4 binding element selected from the group consisting of CTTTGAT (TRE1) and CTTTCAT (TRE2). Transcription of the Tcf-responsive reporter gene is measured. A test compound which decreases transcription of the Tcf-responsive reporter gene is a candidate drug for cancer therapy.

The invention thus provides tools and methods for identifying potential therapeutic agents for cancer treatment and for ameliorating negative side effects of NSAIDs.

RRTEF DESCRIPTION OF THE DRAWINGS

Figures lA-C. Expression of PPARδ inhuman colorectal cancer cells. Figure

1A. Decreased expression of PPARδ following induction of APC in human colorectal cancer cells. Expression of APC (HT29-APC) or β-galactosidase (HT29-GAL) was induced with 110 μM ZnCl₂ for the indicated times in HT29 colorectal cancer cells containing the respective genes under the control of a modified metallothionein promotor. Total RNA (10 μg) was isolated and analyzed by Northern blot analysis with probes specific for PPARδ and PPARγ. Figure IB. Increased expression of PPARδ in primary human colorectal cancers. Northern blot analyses with probes specific to PPARδ and PPARγ were performed on total RNA (10 μg) isolated from matched primary colorectal cancers (C) and normal colon epithelium (N) removed from four different patients. Figure IC. Expression of PPARδ in human colorectal cancers is dependent on Tcf-4-mediated transcription. Colorectal cancer cells with increased β-catenin/Tcf-4 mediated transcription due to either APC (SW480, DLD1) or β-catenin (HCT116) mutations were either mock infected (Con) or infected with adeno virus expressing GFP (GFP) or a dominant negative mutant of Tcf-4 (dnTcf). Total RNA (10 μg) was isolated and analyzed by Northern blot analysis with probes specific for PPARδ and PPARγ.

Figures 2A-E. APC Regulates PPARδ expression through β-catenin/Tcf-4-mediated transcription. Figure 2A. PPARδ promotor. A restriction map of the 3.1 kb region upstream of the first exon of PPARδ is shown. Restriction fragments BE, NE, HE, DE, BN, NH, HD and NP were used to construct reporters for measuring APC and β-catenin responsiveness. Filled boxes represent potential Tcf-4 binding sites, and open sites represent the same sites engineered to contain mutations that abolish Tcf-4 binding. mNP represents fragment NP with both potential Tcf-4 binding sites mutated. TRE1 and TRE2 contain four repeats of the two Tcf-4 binding sites, respectively. mTREl and mTRE2 are mutant forms of TRE1 and TRE2. Figure

2B. The PPARδ promotor is repressed by APC and dominant negative Tcf-4. SW480 colorectal cancer cells were transfected with the indicated PPARδ promotor luciferase reporters (0.4 μg), with a β-galactosidase expression vector (0.2 μg pCMVβ), and with 1.0 μg of either a vector control (Vector), an expression vector for APC (APC) or for a dominant negative form of Tcf-4 (dnTcf). Luciferase activity is reported relative to the vector control after normalizing for transfection efficiency through β-galactosidase activity. Bars represent the means of three independent replicates, with error bars being the unbiased standard deviations. Figure 2C. APC and dnTcf responsiveness is mediated by two putative Tcf-4 binding sites. PPARδ promotor fragments with intact and mutated Tcf-4 binding sites were tested for APC and dnTcf-4 responsiveness as described in Figure 2B. Bars represent the means of three independent replicates, with error bars representing the unbiased standard deviations. Figure 2D. β-catenin transactivation maps to the same promotor regions mediating APC and dnTcf responsiveness. The 293 human cell line was transfected with the indicated PPARδ promotor luciferase reporters (0.4 μg), with a β-galactosidase expression vector (0.2 μg pCMVβ) and with 1.0 μg of either a no insert control (Vector) or an oncogenic β-catenin (β-catenin) expression vector. Luciferase activity was reported as described for Figure 2B. Bars represent the means of three independent replicates, with error bars representing the unbiased standard deviations. Figure 2E. Putative Tcf-4 binding sites in the PPARδ promotor bind Tcf-4. GEMS A was performed using ³²P-labeled probes containing either putative Tcf-4 binding sites TREl or TRE2. GEMS A was performed in the presence of a GST fusion protein containing the Tcf-4 DNA binding domain as indicated. Wild type (wt) or mutant (mut) competitors corresponding to the Tcf-4 binding sites were used as indicated. Figures 3 A-G. Development of a PPARδ-Specific Reporter. Figure 3 A. RXR consensus binding site. PCR products of a randomized ohgonucleotide template that bound a GST fusion protein containing the DNA binding domain of RXR were selected, cloned, and sequenced. The sequences of twenty-eight clones are shown, manually aligned to derive the consensus binding sequence indicated at the bottom.

Figure 3B. PPARδ consensus binding site. PCR products of a randomized ohgonucleotide template that bound a GST fusion protein containing the DNA binding domain of PPARδ were selected, cloned and sequenced. The sequences of twenty clones are shown, manually aligned to derive the consensus binding sequence indicated at the bottom. Figure 3C. Binding Specificity of PPARα, PPARδ and PPARγ.

Oligonucleotides containing the indicated binding elements (DRE or ACO) were ³²P-labeled and incubated with GST fusion proteins containing either the PPARα, PPARδ, PPARγ, RXR, or no DNA binding domain (-). DNA binding was assessed by GEMS A, where "Probe" indicates the unbound probe and "Shifted" indicates bound probe. Figure 3D. DRE confers PPARδ responsiveness. The 293 human cell line was transfected with the indicated (DRE or ACO) luciferase reporters (0.3 μg), with a β-galactosidase expression vector (0.2 μg pCMVβ), and with 1.0 μg of either a vector control (Vector), PPARδ, or PPARγ expression vectors. Luciferase activity was calculated as described in Figure 2B. Bars represent the means of three independent replicates with the error bars representing the unbiased standard deviations. Figure 3E .

Binding specificity of PPARδ RXRα and PPARγ/RXR heterodimers. Oligonucleotides containing the indicated binding elements (DRE or ACO) were ³²P- labeled and incubated with in vitro translated PPARδ, PPARγ, and RXRα as indicated. The binding was supplemented with PPARδ ligand cPGI (10 μM) and PPARγ ligand BRL 49653 (10 μM) were added as indicated. DNA binding was assessed by GEMS A where "Probe" indicates the unbound probe and "Shifted" indicates bound probe. Figure 3F. DRE confers PPARδ but not PPARγ responsiveness. The 293 human cell line was transfected with DRE luciferase reporter (0.3 μg), a β-galactosidase expression vector (0.2 μg pCMVβ), and with 1.0 μg of either empty vector (Control), PPARδ, or PPARγ expression vectors. Where indicated, cells were treated with the PPARδ ligand cPGI (20 μM) or the PPARγ ligand BRL 49653 (20 μM). Luciferase activity was reported as relative luciferase activity after correction for transfection efficiency using β-galactosidase activity. Bars represent the means of three independent replicates, with the error bars representing the unbiased standard deviations. Figure 3G. ACO confers PPARγ but not PPARδ responsiveness. The 293 human cell line was transfected with ACO luciferase reporter (0.3 μg), a β-galactosidase expression vector (0.2 μg pCMVβ), and with 1.0 μg of either empty vector (Control), PPARδ, or PPARγ expression vectors. Where indicated, cells were treated with the PPARδ ligand cPGI (20 μM) or the PPARγ ligand BRL 49653 (20 μM). Luciferase activity was reported as relative luciferase activity after correction for transfection efficiency using β-galactosidase activity. Bars represent the means of three independent replicates, with the error bars representing the unbiased standard deviations.

Figures 4A-C. PPARδ activity is regulated by APC, β-catenin and sulindac. Figure 4 A. APC and dnTcf specifically repress PPARδ activity. PPARδ and PPARγ activity was assessed with the DRE and ACO luciferase reporters, respectively. SW480 colorectal cancer cells were transfected with the indicated luciferase reporters (0.4 μg of DRE or ACO), with a β-galactosidase expression vector (0.2 μg pCMVβ), and with

1.0 μg of either a vector control (Vector), APC, or a dominant negative Tcf-4 (dnTcf) expression vector. Luciferase activity was calculated as described in Figure 2B. Bars represent the means of three independent replicates, with the error bars being the unbiased standard deviations. Figure 4B. β-catenin expression increases PPARδ activity. The 293 human cell line was transfected with the indicated luciferase reporters

(0.4 μg of DRE or ACO), with a β-galactosidase expression vector (0.2 μg pCMVβ), and with 0.8 μg of either a no insert control (Vector) or an oncogenic β-catenin expression vector. Figure 4C. Sulindac specifically represses PPARδ activity. PPARδ and PPARγ activity was assessed as transcriptional activity of the DRE and ACO luciferase reporters, respectively. HCT116 and SW480 colorectal cancer cells were transfected with the indicated luciferase reporters (1.0 μg of DRE or ACO) and with a β-galactosidase expression vector (0.2 μg pCMVβ). Cells were allowed to recover for 20 hours after transfection and were then treated for 10 hours with the indicated concentrations (μM) of sulindac sulfide. Luciferase activity was reported relative to the control (0) after normalizing for transfection efficiency.

Figures 5A-E. Fluorescence microscopy of uninfected (Figure 5A), AdGFP (Figure 5B) or AdPPARδ (Figure 5C) infected HCT116 cells treated with 125 uM of sulindac sulfide, showing that PPARδ can partially protect colon cancer cells from sulindac-induced apoptosis. HCT116 and SW480 cells were either mock infected (Uninfected) or infected with adenovirus expressing GFP (AdGFP) or PPARδ (AdPPARδ). Twenty hours after infection, cells were treated for 42 hours with sulindac sulfide. Apoptosis was assessed by the presence of apoptotic nuclei (condensation and fragmentation) after Hoechst 33258 staining. Figure 5D. Bars represent the fraction of apoptotic nuclei after treatment with the indicated adenoviruses and concentration of sulindac sulfide (μM). Figure 5E. PPARδ rescues sulindac sulfide inhibition of clonal growth. Cells were infected with the indicated adenovirus, treated with the indicated concentrations of sulindac sulfide, and plated. Clonal growth was scored as colony formation after six days. Colonies were visualized by staining with Crystal Violet (upper panel) and enumerated (lower panel).

Figures 6A-E. Mechanism of suppression of PPARδ by NSAIDs. Figure 6A.

NSAIDs do not affect PPARδ expression. HCT116 and SW480 cells were treated with the indicated concentration (μM) of sulindac sulfide for 36 hours, and RNA was isolated. Northern blot analysis was performed on 10 μg of total RNA with a probe specific for PPARδ. Figure 6B. NSAIDs suppress PPARδ DNA binding. The DRE binding element was ³²P-labeled and incubated with no lysate (Probe only), a non-programmed in vitro translation lysate (Blank Lysate) or in vitro translated PPARδ (δ), RXR (RXR), or both (δ + RXR). PPARδ + RXR was included in all lysates treated with the indicated NSAIDs. DNA binding was assessed by GEMSA, where "Probe" indicates the unbound probe and "Shifted" indicates bound probe. Figure 6C. NSAIDs do not suppress PPARγ DNA binding. DNA binding activity was assessed as in Figure

6B except the ACO DNA binding element was used as probe. Figure 6D. NSAIDs suppress PPARδ DNA binding. The DRE binding element was ³²P-labeled and incubated with no lysate (Probe only) or in vitro translated PPARδ (δ), RXRα (RXRα), or both (δ + RXRα). PPARδ + RXRα + cPGI (10 μM) was included in all lysates treated with the indicated NSAIDs. DNA binding was assessed by GEMSA, where

"Probe" indicates the unbound probe and "Shifted" indicates bound probe. Figure 6E. NSAIDs do not suppress PPARγ DNA binding. The ACO binding element was ³²P- labeled and incubated with no lysate (Probe only) or in vitro translated PPARγ (γ),

RXRα (RXRα), or both (γ + RXRα). PPARγ + RXRα + BRL 49653 (10 μM) was included in all lysates treated with the indicated NSAIDs. DNA binding was assessed by GEMSA, where "Probe" indicates the unbound probe and "Shifted" indicates bound probe.

Figure 7. Unified model for APC- and NSATD-mediated suppression of colorectal cancer. Elements indicated in blue have been shown to have a tumor suppressive effect, whereas elements in red have been shown to promote tumor formation. The effects of items in boxes have been demonstrated by genetic alterations. LOX = 5'-lipoxygenase, sPLA2 = secretory phospholipase 2, and COX = cyclooxygenase.

DETAILED DESCRIPTION

Using SAGE technology to analyze global gene expression profiles in human colorectal cancer cells, we have identified peroxisome proliferator-activated receptor delta (PPARδ, a.k.aPPARβ, NUC1, and FAAR; Amri et al, 1995; Jow and Mukherjee, 1995; Schmidt et al, 1992) as another target of the APC pathway. PPARδ belongs to the nuclear receptor superfamily, which includes the steroid hormone, thyroid hormone, retinoid, and PPAR subfamilies as well as a growing number of orphan receptors (Kastner et al, 1995; Lemberger et al, 1996; Mangelsdorf et al, 1995). The PPAR subfamily comprises at least three distinct subtypes found in vertebrate species: PPARα (Dreyer et al, 1992), PPARδ (Amri et al, 1995; Jow and Mukherjee, 1995; Schmidt et al, 1992), and PPARγ (Tontonoz et al, 1994). The nuclear receptor family members function as ligand-dependent sequence-specific activators of transcription (Lemberger et al, 1996; Mangelsdorf et al, 1995). The PPARs were initially shown to be activated by peroxisome proliferators and hypolipidemic drugs of the fibrate class, and later by natural fatty acids and prostaglandins (Forman et al, 1997; Forman et al, 1995; Keller et al, 1993; Kliewer et al., 1995; Kliewer et al, 1997; Xu et al, 1999; Yu et al, 1995).

To evaluate the functional role of PPARδ, we used in vitro DNA-binding selection to develop a reporter molecule specific for PPARδ-activated transcription.

Using this reporter, we found that the NSAIDs sulindac and indomethacin could mimic the effects of APC by down-regulating the transcriptional activity of PPARδ but not that of PPARγ. The basis of this inhibition appears to be a direct disruption of the

DNA-binding ability of PPARδ/RXR heterodimers. Consistent with the functional significance of the above interaction, the sulindac-induced apoptosis and inhibition of cell growth of colon cancer cells could be partially rescued by overexpression of PPARδ. These observations demonstrate that the APC tumor suppressor and NSAIDs inhibit a mutual target, PPARδ, thereby providing an unexpected link between the genetic alterations underlying tumor development and a clinically proven effective cancer chemopreventive agent.

In addition to providing insights into the mechanism of APC tumor suppression, this information can be used to pre-screen agents for use in cancer therapy or the treatment of other conditions in which decreased cellular proliferation is desired, such as hyperplastic or dysplastic conditions. In particular, the development of agents that specifically target PPARδ can lead to more efficacious and less toxic means for colorectal cancer chemoprevention.

Subgenomic polynucleotides and nucleic acid constructs can be used to identify test substances which down-regulate the transcriptional activity of PPARδ. Subgenomic polynucleotides of the invention contain less than a whole chromosome and can be single- or double-stranded genomic or cDNA. Preferably the polynucleotides are isolated free of other cellular components, such as membrane components, proteins, and lipids. They can be made by a cell and isolated, or synthesized in the laboratory using an amplification method such as PCR or using an automatic synthesizer. Methods for purifying and isolating DNA are routine and are known in the art.

The isolated subgenomic polynucleotides contain a PPARδ binding element and an RXR binding element. The nucleotide sequence of the PPARδ binding element can be selected, for example, from any of the nucleotide sequences shown in Figure 3B (SEQ ID NOS: 1-21), including the consensus nucleotide sequence CGCTCAC (nucleotides 3-9 of SEQ ID NO:21). PPARδ binding elements with other nucleotide sequences which bind PPARδ protein can also be used in subgenomic polynucleotides of the invention. Such binding elements can be identified, for example, by carrying out assays which can detect PPARδ protein-DNA binding, such as DNA footprinting, electrophoretic mobility shift assays, or immunoprecipitation of PPARδ-DNA complexes using antibodies specific for PPARδ. Such methods are well known in the art.

The nucleotide sequence of the RXR binding element can be selected from any of the nucleotide sequences shown in Figure 3A (SEQ ID NOS:22-50), including the consensus sequence GGTCA (nucleotides 3-7 of SEQ ID NO:50). Other RXR binding elements which bind RXR can be identified as described for PPARδ binding elements, above. The PPARδ and RXR binding elements can be located directly adjacent to each other in the subgenomic polynucleotide, as shown in SEQ ID NO:78, or can be separated by any number of nucleotides which still permits functional binding of a PPARδ/RXR heterodimer, such as 1, 2, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100 nucleotides. If desired, the isolated subgenomic polynucleotide can comprise 1, 2, 3, 4, or more copies of the PPARδ binding element. Multiple copies of the RXR binding element can also be included.

Isolated subgenomic polynucleotides comprising a PPARδ binding element can be attached to a solid support and used to selectively bind PPARδ and remove it from other cellular components. Suitable solid supports include, but are not limited to, insoluble polymers, such as a column chromatography matrix, glass or plastic slides, tissue culture plates, microtiter wells, tubes, or particles such as beads, including but not limited to latex, polystyrene, or glass beads. Any method known in the art can be used to attach a subgenomic polynucleotide to the solid support, including use of covalent and non-covalent linkages, passive absoφtion, or pairs of binding moieties attached respectively to the subgenomic polynucleotide and the solid support.

PPARδ binding elements of the invention can be present in a nucleotide construct, which can be prepared using standard recombinant DNA techniques. Nucleic acid constructs can be linear or circular molecules, with or without replication sequences. Nucleic acid constructs of the invention contain at least 1, 2, 3, or 4 or more copies of the PPARδ binding element.

If desired, a nucleic acid construct can comprise a reporter gene which encodes an assayable product, such as β-galactosidase, luciferase, β-glucuronidase, green fluorescent protein (GFP), autofluorescent proteins, including blue fluorescent protein (BFP), glutathione-S-transferase (GST), luciferase, horseradish peroxidase (HRP), or chloramphenicol acetyltransferase (CAT). Many such reporter genes are known in the art. The reporter gene can be under the control of a minimal promoter, such that in the absence of PPARδ the reporter gene is not expressed or is expressed only at low levels. Such reporter gene constructs can be used, for example, in methods for pre- screening agents for use in cancer therapy, which are described below. In these reporter gene constructs, the minimal promoter is upstream from the reporter gene, and at least one copy of the PPARδ binding element is upstream from the minimal promoter.

Optionally, 2, 3, 4, or more copies of the PPARδ binding element can be present. Suitable minimal promoters include, for example, the minimal CMV promoter (Boshart et al, 1985) and the promoters for TK (Nordeen, 1988), IL-2, and MMTV. If desired, the reporter construct can include one or more RXR binding elements upstream of the minimal promoter.

In other reporter constructs, a reporter gene is under the control of a Tcf-4 binding element. The Tcf-4 binding element can be CTTTGAT (TRE1) or CTTTCAT

(TRE2). Tcf-4-responsive reporter constructs can comprise at least 1 , 2, 3, or 4 or more of either or both Tcf-4 binding elements, or can comprise nucleotides - 1543 to -759 of PPARδ.

The invention provides various methods of pre-screening agents for use in cancer therapy. These methods measure either PPARδ protein binding to its binding element or PPARδ-dependent transcription in response to a test substance. It is also possible to screen agents for use in cancer therapy by measuring transcription of PPARδ itself in response to a test substance. Test substances which can be screened can be pharmacologic agents already known in the art or can be compounds previously unknown to have any pharmacological activity. Test substances can be naturally occurring or designed in the laboratory. They can be isolated from microorganisms, animals, or plants, or can be produced recombinantly or synthesized by chemical methods known in the art.

In one embodiment of the invention, binding of a PPARδ protein to a DNA molecule comprising a PPARδ binding element is measured in the presence and absence of a test substance. Binding can be measured either m a crude nuclear extract of a mammalian tissue, including human tissue, or a human or other mammalian cell line. Preferably the extract either lacks wild-type APC or contains a mutant β-catenin which permits transcription of PPARδ even in the presence of wild-type APC. Thus, suitable extracts can be prepared from colorectal cancer tissue obtained from mammals, including humans, or from colorectal cancer cell lines, such as HT29, SW480, HCTl 16, and DLDl cells. Methods of preparing nuclear extracts are well known in the art, and any such method can be used. Alternatively, binding can be measured in a reconstituted in vitro system. DNA molecules comprising PPARδ binding elements are described above. PPARδ protein can be purified from tissues or cell lines, chemically synthesized, or produced recombinantly, for example using the primer pairs shown in SEQ ID NOS:70 and 71 and in SEQ ID NOS:76 and 77 to amplify the human PPARδ coding sequence in an in vitro transcription-coupled translation system (see Example

1). Measurement of the binding of the PPARδ protein to the PPARδ binding element can be carried out using any method known in the art for detecting DNA- protein binding, such as gel electrophoretic mobility shift assays (GEMSA), DNA footprinting, or immunoprecipitation of bound and unbound PPARδ protein using PPARδ-specific antibodies. PPARδ-specific probes for use in GEMSA or footprinting assays preferably comprise a detectable label. Either radiolabels or nonisotopic labels, such as chemiluminescent, fluorescent, or enzymatic labels, can be used. Optionally, binding can be measured in the presence of known agonists or antagonists of PPARδ regulated transcription. Suitable antagonists include NSAIDs, such as sulindac, indomethacin, and other COX inhibitors (for a complete list, see Goodman & Gilman's THE PHARMACOLOGICAL BASIS OF THERAPEUTICS, 9th ed, McGraw Hill, and Cada et al, FACTS AND COMPARISONS, J.B. Lippincott, 1999, including the July 1999 update).

The amount of binding of the PPARδ protein to the PPARδ binding element in the presence of the test substance is compared to the amount of binding of the PPARδ protein to the PPARδ binding element in the absence of the test substance. The comparison can be quantitative, for example by reference to a standard curve, or qualitative. A test substance which decreases the amount of binding of PPARδ protein to the PPARδ binding element is a candidate drug for use in cancer therapy. Preferably, binding is decreased by at least 25, 50, 75, 85, 90, 95, 97, or 98 percent.

In another method of pre-screening agents for use in cancer therapy, a transfected cell containing a Tcf-responsive reporter construct and a PPARδ protein is contacted with a test substance. The cell can be either stably or transiently transfected. Introduction of reporter constructs can be carried out in culture or in vivo. Preferably, the transfected cell either lacks wild-type APC or contains a mutant β-catenin. Appropriate cells are, for example, colorectal cancer cells, present either in situ in a mammalian body or in vitro in a tissue culture preparation. Colorectal cancer cells can be isolated from patients and placed in tissue culture or established colorectal cancer cell lines, such as HT29, SW480, HCTl 16, and DLDl, can be used. Methods of transfecting nucleic acid constructs into cells are well known and include, but are not limited to, transfection with naked or encapsulated nucleic acids, cellular fusion, protoplast fusion, viral infection, and electroporation. The PPARδ protein can be PPARδ protein which is either endogenous to the cell or which is added to the cell, for example by transfecting the cell with a nucleic acid construct encoding PPARδ protein, or both.

Expression of the reporter gene can be determined by any method suitable for detecting the assayable product of the particular reporter gene used, including biochemical, immunological, or visual detection methods. Expression of the reporter gene can also be determined by detecting its mRNA, for example using Northern or dot blots or in situ hybridization. A test substance which decreases the amount of expression of the reporter gene is a candidate drug for use in cancer therapy. The decrease in expression of the reporter gene can be determined qualitatively or quantitatively, for example by reference to a standard curve. Preferably, the test substance decreases expression of the reporter gene by at least 25, 50, 75, 85, 90, 95,

97, or 98 percent. Optionally, expression of the reporter gene can be measured in the presence of an agonist or antagonist of PPARδ regulated transcription.

In another method, agents are pre-screened for use in cancer therapy by measuring transcription of the reporter gene in the presence of RNA polymerase, ribonucleotides, and PPARδ protein. As in the methods described above, the PPARδ protein can be purified, synthesized chemically, produced recombinantly, or synthesized by an in vitro translation reaction. RNA polymerases and ribonucleotides are readily available commercially. The addition of the RNA polymerase, ribonucleotides, and PPARδ protein to the reporter construct is effected in the presence and absence of the test substance, and transcription of the reporter gene is determined. As in the methods described above, transcription can be determined, for example, using Northern or dot blots, or by measuring the assayable product of the reporter gene. A test substance which decreases the amount of transcription of the reporter gene, preferably by at least 25, 50, 75, 85, 90, 95, 97, or 98 percent, is a candidate for use in cancer therapy. If desired, transcription of the reporter gene can be measured in the presence of a known agonist or antagonist of PPARδ regulated transcription. The invention also provides methods for identifying candidate drugs for use in

FAP patients, patients with APC or β-catenin mutations, or patients with increased risk of developing cancer. In one embodiment, a cell having no wild-type APC or which has a mutant β-catenin is contacted with a test compound, and transcription in the cell of a Tcf-responsive reporter gene is measured. Constructs comprising Tcf-response reporter genes can be introduced into the cells as described above, and the cell can be contacted with the test compound. Alternatively, the Tcf-responsive reporter gene can be contacted with the test compound in a reconstituted in vitro system under conditions in which the reporter gene is transcribed in the absence of the test compound. Conditions which permit in vitro transcription are well known in the art (see Example 1).

A cell which has no wild-type APC either produces an APC protein defective in β-catenin binding or regulation or produces no detectable APC protein at all. Cells which have no wild-type APC include primary colorectal cells isolated from FAP patients or other patients whose colorectal cells bear APC mutations, as well as cell lines such as HT29, SW480, or DLDl. A cell which has mutant β-catenin produces a β-catenin protein which is super-active or which is defective in APC binding or which is resistant to APC regulation. Cells which have mutant β-catenin include primary colorectal cells isolated from FAP patients or other patients whose colorectal cells produce mutant β-catenin. Other cells which have no wild-type APC or which have mutant β-catenin can be identified by assaying candidate cells for production of wild- type APC or β-catenin protein or mRNA, by detecting mutations in APC or β-catenin coding sequences, or by assaying Tcf-4/β-catenin-dependent transcription, using standard molecular biological or immunological techniques.

Transcription of the Tcf-responsive reporter gene is measured in the presence of the test compound and compared with transcription of the Tcf-responsive reporter gene in the absence of the test compound. As with the methods described above, either reporter gene mRNA or the encoded assayable product can be measured. A test compound which decreases transcription of the reporter gene is a candidate drug for treating FAP patients, patients with APC or β-catenin mutations, or patients with increased risk of developing cancer. Preferably, reporter gene expression is decreased by at least 25, 50, 75, 85, 90, 95, 97, or 98 percent. Because PPARδ-dependent gene transcription is believed to result in cell proliferation, the invention also provides methods for identifying test compounds which can be used to encourage cell proliferation or to prevent apoptosis of cells which are dying prematurely in a disease state such as Alzheimer's Disease, AIDS, muscular dystrophy, amyotrophic lateral sclerosis, or other muscle wasting diseases, autoimmune diseases, heart attack, stroke, ischemic heart disease, kidney failure, septic shock, or a disease in which the cell is infected with a pathogen, such as a virus, bacterium, fungus, mycoplasm, or protozoan, to promote healing of the stomach or intestines, or to ameliorate negative side effects of NSAIDs, such as gastric and intestinal ulceration. PPARδ agonists can also be used to block harmful effects of NSAJDS. PPARδ DNA binding activity and PPARδ-dependent transcription are measured as described above for the methods for screening test compounds as cancer therapeutics. In this embodiment of the invention, however, test compounds which increase transcription of PPARδ protein, PPARδ protein binding to a PPARδ binding element, or expression of a reporter gene which is under the control of a PPARδ binding element are identified as candidates for use in encouraging cell proliferation.

The above disclosure generally describes the present invention. A more complete understanding can be obtained by reference to the following specific examples which are provided herein for purposes of illustration only, and are not intended to limit the scope of the invention. All references cited in this disclosure are expressly incoφorated herein.

EXAMPLE 1

Methods used in the examples below

Chemicals, Cell Culture and Medium. Human colorectal cancer cells HT29,

HCTl 16, SW480, and DLDl were maintained in McCoy's 5A medium (Life Technologies, MD) supplemented with 10% fetal bovine serum (HyClone, UT), 100 units / ml penicillin, and 100 μg/ml of streptomycin. Human embryonic kidney cells 293 were maintained in DMEM (Life Technologies) supplemented with 10% fetal bovine serum, 100 units / ml penicillin, and 100 μg/ml of streptomycin. Sulindac derivatives and indomethacin were purchased from BIOMOL. BRL49653 and cPGI were purchased from American Radiolabeled Chemicals and Cayman Chemical Company, respectively. Unless otherwise indicated, all chemicals were purchased from

Sigma (St. Louis, MO).

Serial Analysis of Gene Expression (SAGE). As previously described (He etal, 1998), SAGE was performed on mRNA harvested from exponentially growing HT29-APC and HT29-β-Gal cells 9 hours after zinc induction. A total of 55,233 and 59,752 tags were obtained from APC-expressing and control cells. Analysis of internal linker controls revealed a sequencing error rate of 0.065 per tag, corresponding to a sequencing error rate of 0.0067 per base. This was in good agreement with instrument specifications and previous estimates of SAGE tag errors based on the analysis of the complete yeast genome (Velculescu et al, 1997). These tags represented 14,346 unique transcripts, of which 7,811 transcripts appeared at least twice. Expression differences were considered significant if they had a _felse < 0.1 as determined by Monte Carlo simulations and they were at least four-old in magnitude.

In Vitro DNA-Binding Site-Selection for PPAR δand RXR. GST fusion proteins containing the N-terminal DNA-binding domains of human PPARδ and human RXR were constructed by PCR amplifying the cDNA coding sequences of residues 1-249 of

PPARδ and residues 1-224 of RXR and cloning them into pGEX-2TK vector. As controls, GST fusion proteins containing the DNA-binding domains of human PPARα (amino acids 1-249) and PPARγ (amino acids 1-248) were also constructed. The fusion proteins were produced and purified according to the manufacturer's protocol. To identify the potential consensus DNA sequence motifs recognized by PPARδ and RXR, a previously described in vitro site selection procedure was utilized. Briefly, for binding to the PPARδ and RXR proteins, the following oligonucleotide was synthesized: 5 '-TAGTAAACACTCTATCAATTGG(N)₂₀TCTAG-

AAAGCTTGTCGACGC-3' (SEQ ID NO:51), where "N" represents an equimolar mixture of each nucleotide. Using this oligonucleotide as template, a random duplex pool was generated by PCR amplification with primers that hybridized to the flanking sequences. The fusion proteins were mixed with the random duplex pool and subjected to GEMSA (see below).

A broad region of the gel predicted to contain DNA-protein complexes (from control binding experiment) was excised. Gel slices were homogenized, incubated at 65 °C for 30 min, and passed through Spin-X column (Costar). Eluted DNA was extracted with phenol-chloroform, precipitated with ethanol, re-amplified by PCR, and subjected to the next round of binding. Following completion of the third round of selection-amplification, PCR products were cloned into pZero 2.1 (Invitrogen). The 60-bp probes corresponding to single clones were generated for GEMSA by direct colony PCR using the following ³²P- labeled primers: 5'-TAGTAAACACTCTATCAATTGG-3' (SEQ ID NO:52) and

5'-GTCCAGTATCGTTTACAGCC-3' (SEQ ID NO:53).

To determine the DNA sequences contained within single clones, inserts were PCR amplified by using M13 forward and reverse primers and the PCR products were sequenced with Amersham's Thermosequenase kit and an SP6 primer. Gel Electrophoresis Mobility Shift Assays (GEMSA). DNA-binding assays were performed essentially as described (Zawel et al, 1998). For binding to PCR products derived from in vitro site selections, 1.0-1.5 μg of protein and 50 ng of DNA were used. For binding to oligonucleotides, 0.3-0.5 μg of protein and 0.5 ng of DNA were used. For competitions, a 100-fold excess of unlabeled probe was used. For GEMSA with GST fusion proteins, 0.3-0.5 μg of fusion protein and 0.5 ng of ³²P kinase labeled (~ 10⁶ dpm) DNA were used. The probes for Tcf-4 binding were as previously reported (Korinek et al, 1997). For GEMSA with in vitro translated proteins, 0.1 to 0.2 μl of programmed lysate and ³²P-labeled probe (~ 10⁶ dpm) was used. The DRE probe was formed by annealing 5'-GCGTGAGCGCTCACAGGTCAATTCG-3' (SEQ ID NO:78) and 5'-CCGAATTGACCTGTGAGCGCTCACG-3' (SEQ ID NO:79). The ACO probe was formed by annealing S'-GCGGACCAGGACAAAGGTCACGTTC-S¹ (SEQ ID NO:80) and 5'-CGAACGTGACCTπGTCCTGGTCCG-3' (SEQ ID NO:81).

Construction of a PPAR δResponsive Reporter. The following oligonucleotides containing PPARδ and RXR recognition motifs that were identified from in vitro site-selection approach were synthesized: 5'-CTAGCGTG-

AGCGCTCACAGGTCAATTCGGTGAGCGCTCACAGGTCAATTCG-3' (SEQ ID N O : 5 4 ) a n d 5 ' - C T A GCGAATTGACCTGTGAGCGCTCACCGAATTGACCTGTGAGC-GCTCACG-3' (SEQ ID NO:55). As a control, the following oligonucleotides containing a PPARα and PPARγ responsive element from the acyl-CoA oxidase promotor were also synthesized: 5 '-CTAGCGGACCAGGACAAAGGTCACGTTCGGA- CCAGGACAAAGGTCACGTTCG-3' (SEQ ID NO:56) and

5'-CTAC GAACGTGACCITrGTCCTC^TCCGAACGTGACCTTTGTCCTGGTC- CG-3' (SEQ ID NO:57). The oligonucleotide cassettes were dimerized and cloned into pB V-Luc, a luciferase reporter plasmid with very low basal activity. All constructs were verified by DNA sequencing. Constructions of PPARδ Promotor Reporters. To identify the genomic sequence of the human PPARδ promotor, the following PCR primers were used to screen a BAC library (Research Genetics): 5'-CCTGTAGAGGTCCATCTGCGTTC-3' (SEQ ID NO:58) and 5'-CATGCTGTGGTCCCCCATTGAGC-3' (SEQ ID NO:59). Three independent BAC clones containing the PPARδ promotor sequence were obtained. Upon subcloning and sequencing, the genomic sequence of 3.1-kb immediately upstream of the first exon was determined (GenBank Accession # ).

For the construction of PPARδ promotor reporters, corresponding restriction fragments (illustrated in Figure 2A) were subcloned into pB V-Luc. The following primer pair was used to PCR amplify the mutant NP fragment: 5'-CTAC TAGCGAGGGTGCATCGTCAATGTTTTGTGTGGGAAG-3' (SEQ ID

N O : 6 0 ) a n d

5'-CCGGAATTCTAGGGACGATGACGATGAACAAAGCTTGACTC-3' (SEQ ID NO:61). The following oligonucleotide pairs were used for dimerization to construct corresponding reporters in pB V-Luc: 5'-CTAGCATGTCTTTGTAC- TCGATGTCTTTGTACTCG-3' (SEQ ID NO:62) and

5'-CTAGCGAGTACAAAGAC-ATCGAGTACAAAGACATG-3' (SEQ ID NO:63) for p4XTREl-Luc; 5'-CΓAGCATGTCTTTGGCCTCGATGTCTTTGGCCTCG-3' ( S E Q I D N O : 6 4 ) a n d

5'-CTAGCGAGGCCAAAGACATCGAGGCCAAAGACATG-3' (SEQ ID NO:65) forp4XmTREl-Luc;5'-CTAC<TTGGCTTTCATCTGATTGGCTTTCATCTGAG-3'

( S E Q I D N O : 6 6 ) a n d

5'-CTAGCTCAGATGAAAGCCAATCAGATGAAAGCCAAG-3' (SEQ ID NO :67)for p4XTRE2-Luc; and 5 '-CTAGCTTGGCTTTCG-

CCTGATTGGCTTTCGCCTGAG-3' (SEQ ID NO:68) and

5'-CTAGCTCAGGCGAAA-GCCAATCAGGCGAAAGCCAAG-3' (SEQIDNO:69) for p4XmTRE2-Luc. Transfections and Reporter Assays. Exponentially growing cells were seeded to

12-well tissue culture plates and each assay was carried out in triplicate. Reporter plasmid, effector plasmid and β-gal control plasmid were transfected into cells using LipofectAmine (Life Technologies). Twenty-four hours after transfection, cells were lysed and collected for assays of luciferase activity using Promega' s Luciferase Assay System.

In vitro Transcription and Translation Assays. The full-length proteins of PPARδ, PPARγ, and RXRα were generated by in vitro transcription-coupled translation using the Single Tube Protein System 3 kit (Novagen). Briefly, the following primer pairs were used to amplify the coding sequences of PPARδ, PPARγ, a n d R X R :

5'-CτGATCCTAATACGACTCACTATAGGGAGACCACCATGGAGCAGCCACAG- GAGGAAGCC-3' (SEQ ID NO:70) and 5'-TTTTTTTTAGTAC- ATGTCCTTGTAGATCTC-3' (SEQ ID NO:71) for PPARδ; 5'-GGATCCTAATACGACTCACTATAGGGAGACCACCATGGCAACCATGG- TT GA C A C AGAGAT C - 3 ' ( S E Q ID N O : 72) and

5'-TTTTTTTTAGTACAAGTCCTTGTAGATCTCC-3' (SEQ ID NO:73) for PPARγ; a n d 5'-GGATCCTAATACGACTCACTATAGGGAGACCACCATGGACACCAAACAT- TTCCTGCCGC-3' (SEQ ID NO:74) and 5'-TTTTTTTTAAGTCA- TTTGGTGCGGCGCCTCC-3 ' (SEQ ID NO:75) for RXRα. The full-length proteins were produced according to the manufacturer's protocol.

Generation of Recombinant Adenovirus Expressing PPARδ. The following PCR primer pair was used to amplify human PPARδ coding sequence: 5'-AGAATGCC JCCGCTGCTCGAGGAATGGAGCAGCCACAGGAGGAAGCC-3' ( S E Q I D N O : 7 6 ) a n d

5'-CGCGGATCCTCTAGATTAGTACATGTCCTTGTAGATCT-C-3' (SEQ ID NO:77). The PCR product was cloned into pCMV-HAHA, which contained a double HA-tag driven by a CMV promotor. Authentic PPARδ coding sequence was verified by DNA sequencing and its expression was confirmed by Western blot using an anti-HA antibody. The expression cassette of HA-tagged PPARδ was further subcloned into pAdTrack vector, which also expressed green fluorescent protein. The AdPPARδ recombinant virus was subsequently generated and purified by using the AdEasy system as previously described (He et al, 1998). AdMYC was generated in a similar fashion.

EXAMPLE 2

APC Represses PPARδ Expression

The effects of APC on gene expression were explored using SAGE analysis as previously reported (He et al, 1998). Briefly, gene expression was examined in a human colorectal cancer cell line with inducible wild type APC (HT29-APC) and a control cell line with an inducible lacZ gene (HT29-β-Gal) nine hours after induction.

SAGE analysis of 55,233 and 59,752 tags from APC-expressing and control cells, respectively, led to the identification of 14,346 different transcripts, the great majority of which were not differentially expressed. Because biochemical studies have indicated that APC directly represses Tcf-4/β-catenin-mediated transcription, we focused on the repressed transcripts. One of the most highly repressed tags corresponded to PPARδ (24 tags in HT29-β-Gal vs. 5 tags in HT29-APC).

To confirm the SAGE data, we performed northern blot analysis of RNA from HT29-APC and HT29β-Gal cells using PPAR probes (Figure 1A). Repression of

PPARδ expression was evident as early as 3 hours after APC induction whereas no change was detectable in HT29 β-Gal cells even 9 hours after induction. In contrast, expression of PPARδ was not affected by expression of APC, and the other known PPAR subfamily member, PPARα, was not expressed at detectable levels in the presence or absence of wild type APC (Figure 1A and data not shown).

The ability of APC to repress PPARδ expression in vitro suggested that expression of PPARδ should be elevated in primary colorectal cancers, where the APC pathway is inactivated by mutations in either APC or β-catenin. We tested this hypothesis through evaluation of PPAR expression in paired samples of primary colorectal cancers and normal colorectal mucosa from the same patients. Northern blot analysis revealed a marked increase in PPARβ expression in each of four cancers studied (Figure IB). In contrast, there was no increase in PPARγ expression in the cancers of these patients (Figure IB).

EXAMPLE 3

APC inhibits Tcf-41 β-Catenin Mediated Transcription of the PPARδ Gene To address the basis for APC repression of PPΛRδ expression, we isolated and sequenced a 3.1-kb genomic fragment containing the region upstream of the PPARδ transcription start site (GenBank Accession # ) and used it to analyze APC responsiveness (Figure 2). A luciferase reporter construct containing this fragment (BE) upstream of a minimal promotor was markedly repressed by APC expression (Figures 2A and 2B). Similar analysis of a series of nested deletions and promotor fragments identified two APC-responsive fragments (Fragment NH and HD, Figures 2A and 2B). Examination of the sequence of these fragments revealed two putative Tcf-4 binding sites, one (TRE1) located 1,543 bp upstream of the PPARδ transcription start site in Fragment NH and the other (TRE2) located 759 bp upstream in Fragment HD.

To determine whether these sites were responsible for APC responsiveness, we tested a fragment spanning these two sites with either intact Tcf-4 binding sites (fragment NP) or with alterations predicted to destroy the putative Tcf-4 binding sites (fragment mNP). Fragment NP demonstrated marked APC repression which was completely abrogated by disruption of the putative Tcf-4 binding sites (Figure 2C).

Moreover, either of the putative Tcf-4 binding sites in isolation could confer APC responsiveness in a sequence specific manner (compare TRE1 vs. mTREl and TRE2 vs. mTRE2 in Figure 2C).

As noted above, the most obvious basis for the APC responsiveness would be inhibition of β-catenin/Tcf-4 mediated transcription. Consistent with this, there was a perfect concordance between the ability of APC to repress expression from the PPARδ promotor reporters in colorectal cancer cell lines containing inactivated APC genes (Figures 2B and 2C) and the ability of oncogenic β-catenin to induce transcriptional activity of these reporters in 293 cells with wild type APC function (Figure 2D). Likewise, there was a perfect concordance between APC responsiveness and the ability of a dominant negative Tcf-4 (dnTcf) expression vector to inhibit transcriptional activity in colorectal cancer cell lines (Figures 2B and 2C). As with APC responsiveness, the β-catenin transactivation and the dnTcf repression were abrogated by mutation of the putative Tcf-4 binding sites (Figures 2C and 2D). The ability of Tcf-4 to directly bind to the PPARδ TRE sites was demonstrated by gel electrophoresis mobility shift assays (GEMSA). Both putative binding sites demonstrated significant

Tcf-4 binding which was inhibited by their cognate wild type binding sequences but not by their mutant counteφarts (Figure 2E).

The above results suggested that APC repressed the expression of PPARδ by interfering with β-catenin/Tcf-4 mediated transcription and that alterations in this pathway could lead to increased expression of PPARδ in colorectal cancers. To further evaluate the generality of this pathway, we examined the ability of dnTcf to interfere with PPARδ expression in other human colorectal cancer cell lines with defined APC pathway alterations. Like the HT29 cells in which PPARδ expression was first identified (Figure 1 A), SW480 and DLDl cells contain inactivating mutations of APC. HCTl 16 cells have an activating mutation of b-catenin. As expected from the study of primary tumors (Figure IB), PPARδ expression was readily detected in all the lines (Figure IC). Moreover, PPARδ expression was inhibited in each line by infection with an adenovirus containing a dnTcf expression cassette but not by a control adenovirus containing a GFP expression cassette (Figure IC). In contrast, PPARγ expression was barely detectable in SW480 cells and dnTcf had no effect on PPARγ expression in any of the lines tested.

EXAMPLE 4

Definition of PPARδResponsive Elements (DRE)

To further explore the functional significance of PPARδ repression, we sought to develop reporters for PPARδ function. Although the biological functions as well as downstream targets of PPARδ were virtually unknown, studies of other PPAR family members have defined a prototypic response element. Maximum DNA binding and activation is achieved through heterodimerization between a PPAR protein and RXR, though each protein alone can bind to its cognate recognition element, probably as a homodimer (Gearing et al, 1993; Iseemann et al, 1993). Accordingly, the prototypic

PPAR response element ACO from the acyl-CoA oxidase gene promotor contains two copies of the core binding sequence AGGTCA separated by one base pair (Juge-Aubry et al, 1997; Mangelsdorf, 1995; Lemberger et al, 1996; Tugwood et al, 1992). PPARα and PPARγ bind this consensus efficiently whereas PPARδ does not (see below). To define a PPARδ responsive element, we performed in vitro binding site selection for both PPARδ and RXR. Analysis of 28 binding sites selected with a GST fusion protein containing the DNA binding domain of RXR identified (A/G)GGTCA as the core consensus for RXR (Figure 3 A). This sequence matches previously described RXR binding sites. A similar selection was performed with a GST fusion protein containing the putative DNA binding domain of PPARδ. Analysis of 20 sites identified through this selection revealed a novel binding consensus (CGCTCAC) which was distinct from the previously defined PPARα/γ consensus (Figure 3B).

We predicted that the combination of the PPARδ and RXR consensus sequences should form an efficient responsive element for PPARδ/RXR heterodimers in vivo and also create a PPARδ-binding element in vitro. To test these predictions, we first generated an ohgonucleotide containing a putative PPARδ responsive element (DRE,

5'-CGCTCACAGGTCA-3') (SEQ ID NO:78) by joining the PPARδ and RXR consensus binding sites. GEMSA analysis of DRE revealed strong binding to PPARδ but not to PPARα or PPARγ (Figure 3C). In contrast, the prototypic PPAR responsive element ACO (5'-AGGACAAAGGTCA-3') (SEQ ID NO:79) bound PPARα and PPARγ but not PPARδ (Figure 3C). RXR demonstrated weaker binding to both responsive elements.

To test the specificity of these response elements in cells, we constructed luciferase reporters containing either the DRE or ACO elements. As expected, transfection of 293 cells with PPARδ resulted in strong activation of the DRE reporter but not the ACO reporter (Figure 3D). In contrast, expression of PPARγ in 293 cells resulted in activation of the ACO reporter but not the DRE reporter (Figure 3D). The above results indicate that DRE represents an effective and specific reporter of PPARδ function. EXAMPLE 5

PPARδ Function is Specifically Regulated by the APC/β-Catenin/Tcf-4 Pathway

The above findings suggested that PPARδ activity was regulated by APC/β-cateninTcf-4 pathway at the transcriptional level. To address the functional consequences of this transcriptional regulation in colorectal cancer cells, we used the

PPARδ-specific reporters described above. Transfection of wild type APC into a human colorectal cancer cell line containing endogenous mutant APC resulted in down regulation of the PPARδ reporter DRE but had no effect on the PPARα/γ responsive reporter ACO (Figure 4A). The lack of any effect on PPARα/γ demonstrated the specificity of this inhibition and made it unlikely that the suppressive effects were due to non-specific toxicity from expression of a tumor suppressor gene. Moreover, transfection of a dnTcf-4 expression vector specifically repressed the PPARδ reporter but not the PPARα/γ reporter.

To further eliminate the possibility of non-specific toxic effects, we determined the ability of β-catenin to positively regulate PPARδ activity. Expression of oncogenic β-catenin mutants in human fibroblast cells activated the PPARδ reporter but did not activate the PPARα/β reporter (Figure 4B).

EXAMPLE 6

NSAIDs Suppress PPARδ Activity The effectiveness of NSAIDs at suppressing colorectal tumorigenesis has raised the suspicion that these compounds may somehow be linked to the genetic alterations that drive tumorigenesis in this organ. The identification of PPARδ as a target of the APC tumor suppressive pathway suggested a specific link. Both precursors and products involved in eicosanoid metabolism have recently been shown to be ligands for PPARs (Forman et al, 1997; Forman et al, 1995; Keller et al, 1993; Kliewer et al,

1995; Kliewer et al, 1997; Xu et al, 1999; Yu et al, 1995). The ability of NSAIDs to perturb eicosanoid metabohsm suggested that PPARs may be an ultimate target of NSAIDs in suppressing tumorigenesis (Prescott and White, 1996) and the above findings suggest that PPARδ could be a specific target. To explore this possibility, we tested the effects of the NSAID sulindac on

PPARδ function. Sulindac has been shown to effectively suppress intestinal tumorigenesis in both humans (Giardiello et al, 1993; Labayle et al, 1991; Nugent et al, 1993; Rigau et al, 1991; Thorson et al, 1994; Waddell et al, 1989; Winde et al, 1993; Winde et al, 1995) and mice (Beazer-Barclay et al, 1996; Chiu et al, 1997; Jacoby et al, 1996; Mahmoud et al, 1998), and this inhibition is associated with the induction of apoptosis (Mahmoud et al, 1998; Pasricha et al, 1995). Analogously, sulindac sulfide, the active metabolite of sulindac, has been shown to induce apoptosis in human colorectal cancer cells (Chan et al, 1998; Hanif et al, 1996; Piazza et al, 1995; Shiff et al, 1995). Sulindac sulfide treatment resulted in a dose-dependent repression of PPARδ activity in colorectal cancer cells, as assessed with the DRE reporter (Figure 4C). A similar dose dependent suppression of PPARδ was observed with indomethacin, another NSATD (data not shown). A greater than two-fold repression was observed at low concentrations of sulindac sulfide and a greater than ten-fold reduction was noted at levels of sulindac sulfide that induced substantial degrees of apoptosis in these cells (Figure 4C and 5D). In contrast, sulindac sulfide had only a modest effect (less than 25% repression) on PPARα/γ activity, assessed with the

ACO reporter (Figure 4C).

EXAMPLE 7

Expression of PPAR δ Partially Rescues Sulindac Sulfide-Induced Apoptosis

As noted above, sulindac sulfide has been shown to induce apoptosis of tumor cells in vitro and in vivo. If suppression of PPARδ activity were contributing to this apoptotic activity, overexpression of PPARδ might be expected to protect against sulindac sulfide-induced apoptosis. To test this possibility, we constructed an adenovirus (AdPPARδ) to express PPARδ as well as a green fluorescent protein (GFP) marker, using AdEasy technology (He et al, 1998). The ability of AdPPARδ to suppress sulindac sulfide-induced apoptosis was compared to that of AdGFP, which contained only the GFP marker gene.

Expression of AdPPARδ resulted in nearly a five-fold decrease in apoptosis in

HCTl 16 cells treated with 100 or 125 μM sulindac sulfide (Figures 5A - 5D). Similar results were obtained with the SW480 cell line (Figure 5D). However, the suppression of apoptosis could be overridden at higher concentrations of sulindac sulfide (150 μM,

Figure 5D). The results were further confirmed and extended by the abilityof AdPPARδ to rescue inhibition of clonal cell growth by sulindac sulfide. Treatment of cells with 100 or 125 μM sulindac sulfide resulted in an aproximately 5-fold drop in number of colonies (Figure 5E). This drop could be completely rescued by infection with AdPPARδ, which actually resulted in a slightly increased number (~ 15%) of colonies.

In contrast, the APC target and prototypic oncogene c-MYC could not rescue the inhibition of clonal growth. As with the apoptosis assay the protective effects of AdPPARδ could be abrogated at higher concentrations of sulindac sulfide.

EXAMPLE S Sulindac Sulfide Directly Disrupts the DNA-Binding Ability of PPARδ/RXR

Heterodimers

The inhibition of PPARδ activity by sulindac sulfide could have been direct or indirect. In the case of APC, the suppression of APC is indirect, resulting from APC's inhibition of Tcf-4/β-catenin-mediated transcriptional activation of the PPARδ promotor. To determine whether sulindac also acted at the transcriptional level, we first examined expression of PPARδ following sulindac sulfide treatment. Concentrations of sulindac sulfide that resulted in marked suppression of PPARδ activity and apoptosis had no effect on the level of PPARδ transcripts, as assessed by Northern blot analysis (Figure 6A). This result contrasted markedly with the effect of APC on such transcripts (Figure

IC). Given the ability of PPARs to bind precursors and products of the eicosanoid pathway (Forman et al, 1997; Forman et al, 1995; Keller et al, 1993; Kliewer et al, 1995; Kliewer et al, 1997; Xu et al, 1999; Yu et al, 1995), the effects of NSAIDs on PPARδ could be due to their ability to perturb eicosanoid metabolism. However, several studies have suggested that the chemopreventive effects of NSAIDs are not simply related to their ability to suppress prostaglandin synthesis. We therefore considered an alternative possibility, namely that NSAIDs might act by directly inhibiting PPARδ activity.

To address this possibility, we tested the ability of sulindac to inhibit PPARδ/RXR heterodimer DNA binding activity in vitro. Sulindac sulfide was able to inhibit binding of PPARδ/RXR heterodimers to the DRE element (Figure 6B). Binding to DRE was also inhibited by the NSAIDs indomethacin and the sulindac sulfide-related compound sulindac sulfone (Figure 6B). The relative concentrations of sulindac sulfide, indomethacin, and sulindac sulfone required to inhibit binding to DRE were roughly concordant with the concentrations required to induce apoptosis in colorectal cancer cells, with sulindac sulfide being the most potent and sulindac sulfone the least (Figures 5D, 6B, and data not shown). Neither sulindac sulfide, indomethacin, nor sulindac sulfone had any effect on binding of PPARγ/RXR heterodimers in an analogous assay performed with the ACO element (Figure 6C). The effect of sulindac on PPARδ/RXRα heterodimer DNA binding activity was not simply due to competition with cPGI, because similar inhibitory response could be demonstrated in the absence of cPGI using GEMSA conditions that allow detection of DNA binding in the absence of ligand stimulation.

EXAMPLE 9

Model of APC- and NSAID-suppression of intestinal tumorigenesis These observations allow the formation of a model ofhow APC and NSAIDs may operate to suppress intestinal tumorigenesis (Figure 7). Inmost cancers, inactivating mutations of the APC tumor suppressive pathway lead to elevated levels of β-catenin/Tcf-4 mediated transcription (Korinek et al, 1997; Morin et al., 1997). In rare colorectal cancers without APC mutations, β-catenin mutations mat render it resistant to APC mediated degradation result in elevated β-catenin Tcf-4 mediated transcription (Morin et al, 1997). In either case, this increased β-catenin/Tcf-4 activity leads to increased transcription of growth-promoting genes. Accordingly, restoration of APC function to colorectal cancer cells with defective APC function results in growth suppression and apoptosis (Morin et al, 1996). The genes which have been postulated to mediate the growth-promoting effects of β-catenin/Tcf-4 activity include those encoded by the c-MYC oncogene (He et al, 1998) and the cyclin DI gene (Tetsu and McCormick, 1999), among others (WISP, c-jun and fra-1) (Mann et al, 1999; Pennica et al, 1998). The present findings suggest that PPARδ represents a β-catenin/Tcf-4 target with particular importance for chemoprevention. Whereas APC or β-catenin mutations can result in increased PPARδ activity, NSAIDs can compensate for this defect by suppressing PPARδ activity and promoting apoptosis. This suppression of PPARδ is mediated in part by the ability of some NSAIDs to directly inhibit the DNA binding activity of PPAR. In addition, because fatty acids and eicosanoids can act as ligands and modifiers of PPAR activity (Forman et al, 1997; Forman et al, 1995; Keller et al, 1993; Kliewer et al, 1995; Kliewer et al, 1997; Prescott and White, 1996; Xu et al, 1999; Yu et al, 1995; and unpublished observations of the inventors), PPARδ activity might be repressed by the NS AJD-mediated changes in eicosanoid metabolism. This model can help explain several features of NSATD mediated chemoprevention. First, the remarkable effectiveness of some NSAIDs in the prevention of colorectal adenomas can now be linked to specific genetic defects that underlie the initiation of these tumors and to the ability of NSAIDs to counter-balance the functional consequences of these genetic defects.

Second, although NSATD functions have classically been linked to their inhibition of COX activity and the resulting inhibition of prostaglandin synthesis, several studies have suggested that the chemopreventive and apoptosis-inducing activities of NSAIDs are not entirely related to the inhibition of COX or to the decreased levels of prostaglandins. These results may be explained by the ability of some NSAIDs to directly inhibit PPARδ. Indeed, the sulindac derivative sulindac sulfone, which is devoid of COX-inhibitory activity, has apoptotic activity in vitro and chemopreventive activity in vivo when used at high concentrations, and has been proposed as a chemopreventive agent that lacks the toxicity associated with traditional NSAIDs (Mahmoud et al, 1998; Piazza et al, 1997; Piazza et al, 1995). Sulindac sulfone inhibited PPARδ activity, albeit at higher concentrations that required for sulindac sulfide, consistent with its reduced chemopreventive and apoptosis-promoting activity. Third, recent studies have demonstrated that PPARγ agonists promote intestinal tumorigenesis in the Min mouse while the same agonists inhibit the growth of human colorectal cancer cells (Brockman et al, 1998; Lefebvre et al, 1998; Saez et al, 1998; Sarraf et al, 1998). Although the conclusions of these studies were contradictory, they clearly demonstrated the ability of PPAR ligands to modify intestinal tumor growth. Whether the differences in the responses to PPARγ agonists are due to differences between humans and mice, and whether the observed effects are due to effects on PPARγ or on both PPARγ and PPARδ, will require further investigation. An important role for PPARs in intestinal tumorigenesis is further suggested by the recent identification of loss of function mutations in one allele of PPARγ in 4 of 55 sporadic colorectal cancers (Sarraf et al, 1999).

Fourth, the ability of COX2 expression to modulate apoptosis (Tsujii and Dubois, 1995) and intestinal tumorigenesis (Oshima et al, 1996) may be partially related to its ability to alter the spectrum of Ugands for PPARδ and other PPARs. In this regard, it is interesting to note that the PPARδ ligand cPGI can partially rescue infertility resulting from COX-2 deficiency (Lim et al, 1999).

Finally, the ability of dietary fatty acids and secreted phospholipases to modify the spectrum of PPARδ ligands and thus alter PPARδ activity could account for their ability to affect colorectal cancer risk (Dietrich et al, 1993; MacPhee et al, 1995; Vandεn Heuvel, 1999; Willett t α/., 1990).

REFERENCES

Aberle et al, 1997, EMBOJ. 16, 3797-804 Amri et al, 1995, J. Biol Chem. 270, 2367-71

Beazer-Barclay et al, 1996, Carcinogenesis 17, 1757-60

Behrens et al, 1996, Nature 382, 638-42

Boshart et α/.(1985), Cell 41, 521-30

Brockman etal, 1998, Gastroenterology 115, 1049-55 Chan et al, 1998, Proc. Natl. Acad. Sci. U.S.A. 95, 681-86

Chiu et al, 1997, Cancer Research 57, 4267-73

Dietrich et al, 1993, Cell 75, 631-39

Dreyer et al, 1992, Cell 68, 879-87

Forman et al, 1995, Cell 83, 803-12 Forman et al, 1997, Proc. Natl. Acad. Sci. U.S.A. 94, 4312-17

Gearing et al, 1993, Proc. Natl. Acad. Sci. U.S.A. 90, 1440-44

Giardiello et al, 1993, N. Engl. J. Med. 328, 1313-16

Hanif et al, 1996, Biochem. Pharmacol. 52, 237-45

He et al, 1998, Science 281, 1509-12 Ilyas et al, 1997, Proc. Natl. Acad. Sci. U.S.A. 94, 10330-34

Iseemann et al, 1993, J. Mol. Endocrinol 11, 37-47 Iwao et al, 1998, Cancer Research 58, 1021-26

Jacoby et al, 1996, Cancer Research 56, 710-14

Jow and Mukherjee, 1995, J Biol Chem. 270, 3836-40

Juge-Aubry et al, 1997, J. Biol Chem. 272, 25252-59 Kastner et al, 1995, Cell 83, 859-69

Keller et al, 1993, Proc. Natl Acad. Sci. U.S.A. 90, 2160-64

Kemler, 1993, Trends Genet. 9, 317-21

Kinzler and Vogelstein, 1996, Cell 87, 159-70

Kitaeva et al, 1997, Cancer Research 57, 4478-81 Kliewer et al, 1995, Cell 83, 813-19

Kliewer et al, 1997, Proc. Natl. Acad Sci. U.S.A. 94, 4318-23

Korinek et al, 1997, Science 275, 1784-87

Labayle et al, 1991, Gastroenterology 101, 635-39 Lefebvre et al, 1998, Nature Med. 4, 1053-57 Lemberger et al, 1996, Ann. Rev. Cell. Devel. Biol. 12, 335-63

Urn et al, 1999, Genes Dev. 13, 1561-74.

Lockhart et al, 1996, Nature Biotechnology 14, 1675

MacPhee et al, 1995, Cell 81,951-66

Mahmoud et al, 1998, Carcinogenesis 19, 87-91 Mangelsdorf et al, 1995, Cell 83, 835-39

Mann et al, 1999, Proc. Natl. Acad. Sci. U.S.A. 96, 1603-08

Molenaar et al, 1996, Cell 86, 391-99

Morin et al, 1996, Proc. Natl Acad. Sci. U.S.A. 93, 7950-54

Morin et al, 1997, Science 275, 1787-90 Munemitsu et al, 1995, Proc. Natl. Acad. Sci. U.S.A. 92, 3046-50

Nordeen, 1988, BioTechniques 6, 454-48

Nugent et al, 1993, Br. J. Surg. 80, 1618-19

Orford et al, 1997, J. Biol. Chem. 272, 24735-38

Oshima et al, 1996, Cell 87, 803-09. Pasricha et al, 1995, Gastroenterology 109, 994-98

Pennica et al, 1998, Proc. Natl. Acad. Sci. U.S.A. 95, 14717-14722

Piazza et al, 1995, Cancer Research 57, 2909-15 Prescott and White, 1996, Cell 87, 783-86 Rigau et α/., 1991, Ann Intern. Med 115, 952-54 Rubinfeld et al, 1993, Science 262, 1731-34 Rubinfeld et al, 1996, Science 272, 1023-25 Rubinfeld et al, 1997, Science 275, 1790-92

Saez et al, 1998, Nature Med. 4, 1058-61 Sanaf et al , 1998, Nature Med. 4, 1046-52 Sarraf et α/., 1999, Mol Cell 3, 700-804. Schmidt et al , 1992, Mol Endocrinol 6, 1634-41 Shiff et al, 1995, J. Clin. Invest. 96, 491-503

Sparks et al, 1998, Cancer Research 58, 1130-34 Su et al, 1993, Science 262, 1734-37 Tetsu and McCormick, 1999, Nature 398, 422-26 Thorson et al, 1994, Lancet 343, 180 Tontonoz et al, 1994, Cell 79, 1147-56

Tsujii and Dubois, 1995, Cell 83, 493-501. Tugwood et al, 1992, EMBOJ. 11, 433-39 Vanden Heuvel, 1999, J. Nutri. 129, 575-80 Velculescu et al, 1997, Cell 88, 243-51 Waddell et al, 1989, Am J. Surg. 157, 175-79

Willett et al, 1990, N. Engl. J. Med 323, 1664-72 Winde et al, 1993, Int J Colorectal Dis 8, 13-17 Winde et al , 1995, Dis Colon Rectum 38, 813-30 Xu et al, 1999, Molecular Cell 3, 397-403 Yu et al, 1995, J. flio/. Chem. 270, 23975-83

Zawel et al, 1998, Molecular Cell 1, 611-17

Claims

1. An isolated subgenomic polynucleotide comprising a PPARδ binding element comprising a nucleotide sequence selected from the group consisting of SEQ ID NOS :1-21 and nucleotides 3-9 of SEQ ID NO:21 (Figure 3B) and an RXR binding element comprising a nucleotide sequence selected from the group consisting of SEQ

ID NOS:22-50 and nucleotides 3-7 of SEQ ID NO:50 (Figure 3 A).

2. The polynucleotide of claim 1 wherein the PPARδ binding element is within 100 nucleotides of the RXR binding element.

3. The polynucleotide of claim 1 which is attached to a solid support.

4. The polynucleotide of claim 1 which is attached to an insoluble polymer.

5. An isolated subgenomic polynucleotide comprising at least 2 copies of a PPARδ binding element comprising a nucleotide sequence selected from the group consisting of SEQ ID NOS: 1-21 and nucleotides 3-9 of SEQ ID NO:21 (Figure 3B).

6. The polynucleotide of claim 5 which comprises at least 4 copies of the PPARδ binding element.

7. The polynucleotide of claim 5 which is attached to a solid support.

8. The polynucleotide of claim 5 which is attached to an insoluble polymer.

9. A nucleic acid construct comprising at least one PPARδ binding element comprising a nucleotide sequence selected from the group consisting of SEQ ED NOS:l-21 and nucleotides 3-9 of SEQ ID NO:21, a minimal promoter, and a reporter gene, wherein the PPARδ binding element is upstream from the minimal promoter and the minimal promoter is upstream from the reporter gene, wherein the minimal promoter regulates transcription of the reporter gene.

10. The nucleic acid construct of claim 9 which comprises at least 2 copies of the PPARδ binding element.

11. The nucleic acid construct of claim 9 which comprises at least 3 copies of the PPARδ binding element.

12. The nucleic acid construct of claim 9 which comprises at least 4 copies of the PPARδ binding element.

13. The nucleic acid construct of claim 9 which further comprises an RXR binding element comprising a nucleotide sequence selected from the group consisting of SEQ ID NOS:22-50 and nucleotides 3-7 of SEQ ID NO:50 upstream from the reporter gene.

14. The nucleic acid construct of claim 13 wherein the PPARδ binding element is within 100 nucleotides of the RXR binding element.

15. A method of pre-screening agents for therapeutic use, comprising: measuring binding of a PPARδ protein to a DNA molecule comprising a PPARδ binding element in the presence and in the absence of a test substance; and comparing amount of binding of the PPARδ protein in the presence of the test substance to amount of binding of the PPARδ protein in the absence of the test substance, a test substance which decreases the amount of binding being a candidate agent for use in cancer therapy and a test substance which increases the amount of binding being a candidate agent for ameliorating negative side effects of NSAIDs.

16. The method of claim 15 wherein the step of measuring employs observation of a shift in mobility of the DNA molecule on gel electrophoresis.

17. The method of claim 15 wherein the DNA molecule is radiolabeled.

18. The method of claim 15 wherein binding is measured in the presence of an NSATD.

19. A method of pre-screening agents for therapeutic use, comprising: contacting a transfected cell with a test substance, wherein the transfected cell contains a PPARδ protein and a reporter construct comprising a reporter gene which encodes an assayable product, a minimal promoter upstream from and regulating transcription of the reporter gene, and at least one copy of a PPARδ binding element upstream of the minimal promoter; and determining whether expression of the reporter gene is decreased or increased by the test substance, a test substance which decreases the amount of expression of the reporter gene being a candidate agent for use in cancer therapy and a test substance which increases the amount of expression of the reporter gene being a candidate agent for ameliorating negative side effects of NSAIDs.

20. The method of claim 19 wherein the reporter gene construct further comprises an RXR binding element upstream of the minimal promoter.

21. The method of claim 19 wherein the transfected cell is in culture.

22. The method of claim 19 wherein the transfected cell is in a mammalian body.

23. The method of claim 19 wherein expression of the reporter gene is determined in the presence of an NSAID.

24. A method of pre-screening agents for therapeutic use, comprising: adding RNA polymerase, ribonucleotides, and PPARδ protein to a reporter construct which comprises a reporter gene which encodes an assayable product, at least one copy of a PPARδ binding element upstream from a minimal promoter, and the minimal promoter upstream from and controlling transcription of the reporter gene, the step of adding being effected in the presence and absence of a test substance; and determining whether transcription of the reporter gene is decreased or increased in the presence of the test substance, a test substance which decreases the amount of transcription of the reporter gene being a candidate agent for use in cancer therapy and a test substance which increases the amount of transcription of the reporter gene being a candidate agent for ameliorating negative side effects of NSAIDs.

25. The method of claim 24 wherein transcription of the reporter gene is determined in the presence of an NSATD.

26. A method of identifying candidate drugs for use in FAP patients, patients with APC or β-catenin mutations, or patients with increased risk of developing cancer, comprising the steps of: contacting a cell having no wild-type APC or a mutant β-catenin with a test compound; measuring transcription in the cell of a Tcf-responsive reporter gene, wherein the Tcf-responsive reporter gene comprises a Tcf-4 binding element selected from the group consisting of CTTTGAT (TRE1) and CTTTCAT (TRE2), wherein a test compound which decreases transcription of the reporter gene is a candidate drug for cancer therapy.

27. The method of claim 26 wherein the Tcf-responsive reporter gene comprises both TRE1 and TRE2.

28. The method of claim 26 wherein the Tcf-responsive reporter gene comprises nucleotides -1543 to -759 of PPARδ.

29. The method of claim 26 wherein the Tcf-responsive reporter gene comprises at least two copies of TRE1 or TRE2.

30. The method of claim 26 wherein the Tcf-responsive reporter gene comprises at least four copies of TREl or TRE2.

31. The method of claim 26 wherein the cell produces an APC protein defective in β-catenin binding or regulation.

32. The method of claim 26 wherein the cell produces a β-catenin protein which is super-active, or which is defective in APC binding or resistant to APC regulation.

33. The method of claim 26 wherein the cell produces no detectable APC protein.

34. A method of identifying candidate drugs for use in FAP patients, patients with APC or β-catenin mutations, or patients with decreased risk of developing cancer, comprising the steps of: contacting a Tcf-responsive reporter gene with a test compound under conditions in which the reporter gene is transcribed in the absence of the test compound, wherein the Tcf-responsive reporter gene comprises a Tcf-4 binding element selected from the group consisting of CTTTGAT (TREl) and CTTTCAT (TRE2); and measuring transcription of the Tcf-responsive reporter gene; wherein a test compound which decreases said transcription is a candidate drug for cancer therapy.

35. The method of claim 34 wherein the Tcf-responsive reporter gene comprises both TREl and TRE2.

36. The method of claim 34 wherein the Tcf-responsive reporter gene comprises nucleotides -1543 to -759 of PPARb.

37. The method of claim 34 wherein the Tcf-responsive reporter gene comprises at least two copies of TREl or TRE2.

38. The method of claim 34 wherein the Tcf-responsive reporter gene comprises at least four copies of TREl or TRE2.