Peroxisome proliferator-activated receptor gamma-mediated differentiation: a mutation in colon cancer cells reveals divergent and cell type-specific mechanisms.

Activation of the nuclear hormone receptor peroxisome proliferator-activated receptor gamma (PPARgamma) inhibits cell growth and induces differentiation in both adipocyte and epithelial cell lineages, although it is unclear whether this occurs through common or cell-type specific mechanisms. We have identified four human colon cancer cell lines that do no undergo growth inhibition or induce markers of differentiation after exposure to PPARgamma agonists. Sequence analysis of the PPARgamma gene revealed that all four cell lines contain a previously unidentified point mutation in the ninth alpha-helix of the ligand binding domain at codon 422 (K422Q). The mutant receptor did not exhibit any defects in DNA binding or retinoid X receptor heterodimerization and was transcriptionally active in an artificial reporter assay. However, only retroviral transduction of the wild-type (WT), but not mutant, receptor could restore PPARgamma ligand-induced growth inhibition and differentiation in resistant colon cancer cell lines. In contrast, there was no difference in the ability of fibroblast cells expressing WT or K422Q mutant receptor to undergo growth inhibition, express adipocyte differentiation markers, or uptake lipid after treatment with a PPARgamma agonist. Finally, analysis of direct PPARgamma target genes in colon cancer cells expressing the WT or K422Q mutant allele suggests that the mutation may disrupt the ability of PPARgamma to repress the basal expression of a subset of genes in the absence of exogenous ligand. Collectively, these data argue that codon 422 may be a part of a co-factor(s) interaction domain necessary for PPARgamma to induce terminal differentiation in epithelial, but not adipocyte, cell lineages and argues that the receptor induces growth inhibition and differentiation via cell lineage-specific mechanisms.

The induction of terminal cellular differentiation is a complex process requiring the initiation of a gene expression pattern that ultimately results in both cell cycle withdrawal and the expression of a set of proteins that are necessary to carry out the specialized function of the differentiated cell. Transcription factors often serve as driving forces for cellular differentiation. For example, muscle cell differentiation is critically dependent on the MyoD family of basic helix-loop-helix transcription factors (1). Expression of MyoD family members initiates a cascade of temporal-specific gene expression patterns ultimately leading to growth arrest and differentiation. Cancer is essentially a state of de-differentiation in which the malignant cells do not undergo the normal maturation process that leads to cessation of cell growth. The ability of transcription factors to initiate terminal differentiation pathways has been exploited as a form of differentiation therapy for cancer that can serve as an alternative to more toxic chemotherapeutic regimens. For example, ligand activation of the retinoic acid receptor ␣-promyelocytic leukemia fusion protein using alltrans retinoic acid has been successfully used in the treatment of acute promyelocytic leukemia (2).
Peroxisome proliferator-activated receptor ␥ (PPAR␥) 1 is a ligand-activated transcription factor that is capable of initiating terminal differentiation pathways. PPAR␥ and related subtypes PPAR␣ and PPAR␦ are members of the nuclear hormone receptor gene superfamily (3) that form functional heterodimers with members of the retinoid X receptor (RXR) family of nuclear receptors (4). PPARs play fundamental roles in metabolic homeostasis, primarily as regulators of fatty acid storage and catabolism (5). Putative endogenous ligands for PPAR␥ include both polyunsaturated fatty acids and the eicosanoids 15-deoxy⌬ 12,14 -PGJ 2 (6, 7), 13-hydroxyoctadecadienoic acid, and 15-hydroxyeicosatetraenoic acid (8), but their respective roles in PPAR␥ signaling in vivo remains unclear. High affinity synthetic ligands that selectively activate PPAR␥ include the thiazolidinediones, a class of insulin sensitizing drugs currently in use for the treatment of insulin-resistant diabetes mellitus (9).
PPAR␥ appears to play a dominant role in the differentiation of adipocytes. Early experiments established that ectopic expression of PPAR␥ in fibroblasts resulted in conversion of the cells to adipocytes (10). More recent studies using mice null for the PPAR␥ gene have confirmed this essential role in adipogenesis (11,12). The cellular response induced by PPAR␥ during adipogenesis involves both cell cycle withdrawal and the expression of lipogenic-related genes such as the fatty acidbinding protein aP2 (13). The growth arrest pathway is characterized by a G 1 cell cycle arrest and the induction of cyclindependent kinase inhibitors p18 and p21 (14). PPAR␥ has also been shown to restrict S phase entry by inhibiting the DNA binding activity of E2F/DP (15).
Current evidence suggests that PPAR␥ can induce differentiation pathways beyond adipocytes. For example, activating ligands of PPAR␥ inhibit the proliferation rates of epithelial cells derived from breast, prostate, stomach, and lung (16 -19). In the colon, levels of PPAR␥ mRNA are nearly equivalent to that found in adipocytes (20) with the highest levels of receptor expression observed in the post-mitotic, differentiated epithelial cells facing the lumen (21). Consistent with this expression pattern, exposure of cultured human colon cancer cells to PPAR␥ agonists induces growth inhibition associated with a delay in the G 1 phase of the cell cycle and an increase in several markers of cellular differentiation (22)(23)(24).
Whether the anti-neoplastic, pro-differentiation effects of PPAR␥ ligands in the colon operate in vivo is not clear. Agonists of the receptor will reduce pre-malignant intestinal lesions in rats treated with the carcinogen azoxymethane (25) but slightly increase colon polyps in Adenomatous polyposis coli mutant mice that are predisposed to intestinal adenomas (26,27). However, Sarraf et al. (28) have reported that 8% of primary colorectal tumors contained a loss of function point mutation in one allele of the PPAR␥ gene, emphasizing that the receptor is likely to have a tumor suppressive function in the colon. Four unique mutations in PPAR␥ were identified in the study; one resulted in a truncated protein that lacked the entire ligand binding domain whereas the other three mutations caused defects in the binding of either synthetic or natural ligands.
Although activation of PPAR␥ will initiate pathways leading to growth arrest in both colon epithelial and adipocyte lineages, it is unknown whether this occurs through similar or distinct mechanisms. For example, in both cell types activation of the receptor eventually leads to a G 1 arrest and an increase in cell-specific differentiation markers. Does this occur through the initial regulation of identical target genes in both cell types, and does PPAR␥ require common co-regulator interactions in both instances?
Here we report the detection of an identical exonic mutation (K422Q) in the PPAR␥ gene in four distinct colon cancer cell lines that are refractory to the decrease in cell growth or increase in differentiation markers normally induced by activators of PPAR␥. Only introduction of the WT, but not mutant, receptor was able to restore PPAR␥ ligand sensitivity in the resistant colon cancer cell lines. In contrast, there was no difference in the ability of WT or K422Q receptor to induce adipocyte differentiation. Analysis of direct PPAR␥ target genes in colon cancer cells expressing the WT or K422Q mutant allele suggests that the mutation may be non-functional because of an inability of the apo-receptor to basally repress certain target genes. These results argue that codon 422 may be a part of a co-factor(s) interaction domain necessary for PPAR␥ to induce terminal differentiation in epithelial, but not adipocyte, cell lineages.

EXPERIMENTAL PROCEDURES
Receptor Ligands-All synthetic PPAR ligands were from Glaxo-SmithKline and dissolved in Me 2 SO. 15-Deoxy⌬ 12,14 -PGJ 2 was purchased from Cayman Chemical.
Cell Culture-The HCT 15, COLO 205, HCT 116, HT-29, and NIH 3T3 cell lines were purchased from ATCC. 293-EBNA cells were purchased from Invitrogen. The MOSER S cell line was a gift from M. Brattain (University of Texas Health Sciences, San Antonio, TX). The MIP 101 and Clone A cell lines were a gift from L. B. Chen (Dana Farber Cancer Institute, Boston, MA). The HCA-7 cell line was obtained from S. Kirkland (University of London, London, United Kingdom). Cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (FBS; Hyclone), L-glutamine (2 mmol/liter), penicillin (100 units/ml), and streptomycin (100 g/ml) in a 5% CO 2 atmosphere with constant humidity. For all experiments in which a receptor ligand was added, cells were grown in the above media except regular 10% FBS was replaced with 10% charcoalstripped FBS (Hyclone).
Plasmids-Full-length WT PPAR␥ was cloned into pBLUESCRIPT KSϩ. This was used as a template to generate PPAR␥ K422Q using oligonucleotide-directed in vitro mutagenesis (Muta-Gene; Bio-Rad). Both WT and K422Q PPAR␥ were cloned into pCDNA3.0 (Invitrogen) for use in transient transfection and EMSA experiments. HA-tagged WT and K422Q PPAR␥ were generated by PCR using Pfu Turbo Taq polymerase (Stratagene) and the proper non-tagged cDNA as a template. The 5Ј primer contained a XhoI site, the full-length HA epitope, and a partial region of PPAR␥ starting at codon 2. The 3Ј primer contained a HpaI site and a partial region of PPAR␥ starting at codon 479 (stop codon). Each amplicon was digested and cloned into the XhoI/HpaI site of the retroviral expression vector pMSCVpuro (Clontech). All plasmids were sequenced to avoid unwanted mutations.
Western Blot Analysis-Cells were harvested in ice-cold 1ϫ phosphate-buffered saline, and cell pellets were lysed in radioimmune precipitation assay buffer. Centrifuged lysates (50 g) from each cell line were fractionated on a 4 -20% gradient SDS-polyacrylamide gel and electrophoretically transferred to a polyvinlylidene difluoride membrane (PerkinElmer Life Sciences). Membranes were blocked for 1 h at room temperature in Tris-buffered saline containing 0.1% Tween 20 and 5% powdered milk. The following primary antibodies were used: monoclonal anti-HA antibody clone HA.11 (1:1000; Babco), monoclonal anti-PPAR␥ (1:500; Santa Cruz Biotechnology, Inc.), and monoclonal anti-keratin 18 and 20 antibodies (1:1000; NeoMarkers). This was followed by incubation with donkey anti-mouse horseradish peroxidaseconjugated secondary antibody (Jackson ImmunoResearch Laboratories) at a dilution of 1:50,000 for 1 h. Detection of immunoreactive polypeptides was accomplished using an enhanced chemiluminescence system (Amersham Biosciences).
Mutation Detection-Mutations in the PPAR␥ gene in the COLO 205,  MIP 101, and Clone A cell lines were detected using a combination of denaturing gradient gel electrophoresis and direct sequencing as described previously (28,29). PPAR␥ mutations in the HCT 15, MOSER S, HT-29, HCT 116, and HCA-7 cell lines were detected by automated dideoxy sequence analysis of PCR products that span the coding region of PPAR␥1 using primers sets described previously (30).
Cell Growth Measurements-The day after initial seeding, cells were exposed to Dulbecco's modified Eagle's medium containing 10% char-FIG. 2. DNA binding and transcriptional activity of K422Q PPAR␥. A, the PPAR␥ ligand binding domain (residues 207-477) is depicted as a ribbon diagram in blue (Protein Data Bank code 2PRG). Helix 9 (residues 402-425) is highlighted in red with Lys-422 rendered in space-filling mode. The molecular modeling and graphical representations were performed using Insight II 2000 software with an R12000 Silicon Graphics Octane work station. B, amino acid sequence alignment of codons 412-432 of human PPAR␥ compared with human PPAR subtypes ␣ and ␦ and to PPAR␥ from a range of species. Lys-422 (arrow) is conserved in the PPAR␥ from all species reported in GenBank TM , including the range of species shown here. C, EMSA of WT and K422Q PPAR␥. In vitro translated WT or K422Q PPAR␥ was combined with increasing amounts of RXR␣. The receptor complexes were incubated with a 32 P-labeled oligonucleotide containing the PPRE from the acyl-coA oxidase promoter for 20 min followed by resolution on a 5% non-denaturing polyacrylamide gel and detection by autoradiography. The first lane on the left is a sample with probe only. D and E, CV-1 cells were transiently transfected with PPRE3-tk-luc, pRL-SV40, and WT or K422Q PPAR␥/pcDNA3.0 and treated with increasing doses of either rosiglitazone (D) or 15-deoxy⌬ 12,14 -PGJ 2 (E) for 24 -36 h. Cells were harvested, and the dual luciferase assay was performed. Data are represented as -fold activation over vehicle-treated cells and represent the mean from three independent experiments each done in triplicate. Error bars, S.E. coal stripped FBS and either 0.1% Me 2 SO or the indicated ligand. Cells were exposed to fresh medium and compound every 48 h. Cells were counted after 6 days of treatment using a Coulter counter. Each experiment was done in triplicate.
EMSA-EMSAs were done based on methods reported by Schulman et al. (31). PPAR and RXR receptors were synthesized using a T7 Quick TNT in vitro transcription/translation kit (Promega). 1 determined by staining nuclear DNA with propidium iodide (50 g/ml) followed by measuring the relative DNA content of nuclei using a Facsort fluorescence-activated sorter (BD Biosciences). The proportion of nuclei in each phase of the cell cycle was determined using MODFIT DNA analysis software (BD Biosciences).
Tumor Growth in Athymic Mice-Athymic mice (Harlan Sprague-Dawley, Inc.) were injected subcutaneously in the dorsal flanks with 5 ϫ 10 6 cells of the HCT 15 cells expressing WT or K422Q PPAR␥ in a volume of 0.10 ml of 1ϫ phosphate-buffered saline. Dosing was started 10 -15 days post-injection for each cell line when the mean tumor volumes were ϳ75 mm 3 . Mice were then orally gavaged five times/week with either vehicle (0.5% methylcellulose in 0.05 N HCl) or 10 mg/kg of rosiglitazone (in a total volume of 0.10 ml per mouse). Rosiglitazone was formulated daily by first dissolving the compound in 0.1 N HCL that had been pre-warmed to 40°C followed by the addition of an equal volume of 1% methylcellulose. The size of each tumor was determined by direct measurement of tumor dimensions. The volume was calculated according to the equation (V ϭ [L ϫ W2] ϫ 0.5), where V ϭ volume, L ϭ length, and W ϭ width.
Adipogenesis Assay-Virally infected NIH 3T3 cells were exposed at confluence to dexamethasone (1 M) for 24 h followed by treatment with vehicle or rosiglitazone for 7 days with media changed every 48 h. Cells were then fixed and stained with Oil Red O (Sigma).
Northern Hybridization Analysis-Northern blot analysis was performed as described previously (32). The indicated cell lines were treated with either 0.1% Me 2 SO or 2.0 M rosiglitazone. Total RNA (20 g) from each sample was fractionated on a 1.2% agarose-formaldehyde gel and transferred to a Hybond-NX nylon membrane (Amersham Biosciences). Filters were pre-hybridized for 4 h at 42°C in Ultrahyb (Ambion). Hybridization was conducted in the same buffer in the presence of a 32 P-radiolabeled cDNA fragment of the indicated gene. Blots were washed 4 ϫ 15 min at 50°C in 2ϫ SSC, 0.1% SDS and once for 30 min in 1ϫ SSC, 0.1% SDS. Membranes were then exposed to a phosphorimager, screen and images were analyzed using a Cyclone Storage Phosphor System and Optiquant software (Hewlitt-Packard).
Synthesis of cDNA Probes for Northern Blots-A partial cDNA fragment for adipophilin was generated by PCR with M13 forward and reverse primers using a sequence-validated human IMAGE cDNA clone (Research Genetics) as a template. Partial cDNA fragments for Gob-4 and TSC-22 were generated using reverse transcriptase PCR and genespecific primers corresponding to base pairs (each from the translational start site) 283-550 for Gob-4 and 1-425 for TSC-22. The template for these PCR reactions was a random primed cDNA library of MOSER S colon carcinoma cells treated with either 0.1% Me 2 SO or 1 M rosiglitazone. The Gob-4 PCR products were cloned into the pCR2.1-TOPO vector (Invitrogen), and the TSC-22 product was cloned into pPCR-Script (Stratagene). All plasmids were sequenced to confirm gene identity. The aP2 cDNA fragment was obtained from Youfei Guan (Vanderbilt University, Nashville, TN).

PPAR␥ Ligand Sensitivity and PPAR␥ Gene Mutations in a Panel of Human Colorectal Cancer Cell
Lines-In our initial survey of the biological response of human colorectal carcinoma cells to PPAR␥ agonists, we noticed that some cell lines were resistant to the growth inhibitory effects of PPAR␥ ligands. A panel of eight cell lines (four sensitive and the four that were resistant) was chosen for further study. All eight cell lines expressed relatively equivalent levels of PPAR␥ protein (Fig.  1). The ability of the high affinity, PPAR␥ subtype-selective agonist rosiglitazone to induce growth inhibition in each of the eight lines was tested. Four of these cell lines (MOSER S, HCT 116, HCA-7, and HT-29) were growth-inhibited in the presence of a PPAR␥ agonist, whereas the other four (HCT 15, MIP 101, Clone A, and COLO 205) were not affected (Table I).
There are a number of explanations for why a particular cell line could be resistant to activators of PPAR␥ despite expressing robust levels of the receptor. Because somatic loss of function mutations have been identified in a subset of colorectal tumors, we sought to determine whether PPAR␥ ligand resistance in the four cell lines could be because of a loss of function mutation in the PPAR␥ gene. All four resistant lines contained a monoallelic point mutation in the PPAR␥ gene at codon 422 resulting in a change from lysine (Lys) to glutamine (Gln) (K422Q); this mutation was not found in the four sensitive cell lines (Table I). The correlation between the K422Q allele and lack of sensitivity to PPAR␥ ligands provided suggestive, but not definitive, evidence that this mutation caused the HCT 15, MIP 101, Clone A, and COLO 205 cell lines to be resistant to the growth inhibitory effects of PPAR␥ ligands.
Characterization of K422Q Mutant Allele-No previous studies documenting the sequence of the PPAR␥ gene in various malignancies or from individuals at risk for diabetes or obesity have reported mutations at codon 422 of the receptor. Lys-422 lies within the ninth ␣-helix (H9) of the ligand binding domain of the receptor. Crystallographic studies of PPAR␥/RXR␣ heterodimers suggest a role for H9 in receptor dimerization (33). However, these studies found no direct role for Lys-422 in any polar interactions found at the dimer interface. X-ray crystal-lography of PPAR homodimers revealed Lys-422 to be located at the receptor surface and exposed to solvent, suggesting the possibility of involvement in co-factor interactions ( Fig. 2A) (34). Lys-422 is conserved in the PPAR␥ cDNAs from all species reported in the NCBI Entrez nucleotide data base, including the six different species shown in Fig. 2B. However, Lys-422 is not conserved in either PPAR␣ or PPAR␦, both of which encode a Gln at the homologous amino acid (Gln-413 and Gln-386, respectively) (Fig. 2B). Because codon 422 of PPAR␥ in the resistant cell lines is mutated to an amino acid (Gln) that is normally present in the homologous positions of WT PPAR␣ and PPAR␦, it is unlikely that the K422Q mutation disrupts an important structural interaction common to all three PPARs. In fact, it may be that the Lys at position 422 present in WT PPAR␥ is responsible for an interaction unique to the ␥ subtype. As no obvious function has been ascribed to Lys-422, we first characterized what effects the K422Q mutation might have on WT receptor activity. There was no difference in the DNA binding activity of WT PPAR␥/RXR␣ or K422Q PPAR␥/RXR␣ on a PPRE from the acyl-coA oxidase promoter (Fig. 2C). Identical results were observed using RXR␤ and RXR␥ (data not shown). Transcriptional activity was assayed in cells transfected with either receptor cDNA and the PPRE3-tk-luc reporter vector that contains a luciferase cDNA downstream of three tandem repeats of the PPRE from the acyl-coA oxidase gene (35). There were no significant differences between WT and K422Q PPAR␥ in the ability of either a synthetic (rosiglitazone) or natural (15-deoxy⌬ 12,14 -PGJ 2 ) ligand to induce transcriptional activation (Fig. 2, D and E).
Wild-type, but Not K422Q, PPAR␥ Can Rescue PPAR␥ Ligand Unresponsiveness in Resistant Colon Cancer Cells-Because WT and K422Q PPAR␥ showed equivalent activity in DNA binding and transactivation assays, it was not clear whether the presence of the K422Q mutant allele was the reason the resistant cells were refractory to the growth inhibitory effects of PPAR␥ ligands. To directly test this hypothesis, one of the resistant cell lines, the HCT 15 cells, was retrovirally transduced with HA-tagged WT or K422Q PPAR␥ and assayed for PPAR␥ ligand-induced growth inhibition and differentiation. Three different pooled stable cell lines, HCT 15-pMSCV (vector), HCT 15-PPAR␥ WT, and HCT 15-PPAR␥ K422Q, were generated. Both the HCT 15-PPAR␥ WT and HCT 15-PPAR␥ K422Q cell lines expressed equivalent levels of the WT or mutant receptor protein (Fig. 3A). As observed with the transiently transfected receptors, there were no differences in ligand-induced transactivation between WT and K422Q PPAR␥ in the stable cell lines (Fig. 3B).
Exposure of HCT 15-PPAR␥ WT, but not HCT 15-pMSCV or HCT 15-PPAR␥ K422Q, cells to a synthetic (rosiglitazone) or natural (15-deoxy⌬ 12,14 -PGJ 2 ) PPAR␥ agonist induced a dosedependent decrease in cell number (Fig. 4, A and B). Similarly, only HCT 15 cells expressing the WT (but not mutant) receptor could undergo a partial arrest in the G 1 phase of the cell cycle after extended exposure to a PPAR␥ agonist (Fig. 4C). Identical results were obtained in vivo using a nude mouse xenograft model of tumor growth. Mice bearing tumors consisting of HCT 15 cell expressing either WT or K422Q PPAR␥ were treated by oral gavage with vehicle or 10 mg/kg body weight of rosiglitazone. A significant reduction in tumor volume was seen only in HCT 15 cells transduced with the WT PPAR␥ allele (Fig. 5, A  and B).
Some members of the keratin and carcinoembryonic antigen (CEA) family of proteins represent markers of intestinal epithelial cell differentiation and have been shown to be induced by PPAR␥ ligands in colon cancer cells (22,24). Rosiglitazone was able to increase protein levels of keratin 18 or 20 or the RNA levels of the CEA cell adhesion molecule CEACAM6 (also known as nonspecific cross-reacting antigen or NCA) only in HCT 15 cells expressing WT, but not K422Q, PPAR␥ (Fig. 6).
In summary, despite the fact that both WT and K422Q PPAR␥ have comparable DNA binding and trans-activation activities, only the WT receptor could rescue the functional resistance of the parental HCT 15 cells. Cumulatively, these data suggest that the four PPAR␥ ligand-resistant cell lines identified in this study are resistant because of the presence of the K422Q allele.
Both WT and K422Q PPAR␥ Can Induce Adipocyte Differentiation-The above data indicate that codon 422 may be important in the ability of ligand occupied PPAR␥ to initiate terminal differentiation pathways in colon epithelial cells. To determine whether this was also true in the case of adipocyte differenti-ation, NIH 3T3 cells were retrovirally infected with vector (NIH 3T3-pMSCV), HA-tagged WT PPAR␥ (NIH 3T3-PPAR␥ WT), or HA-tagged K422Q PPAR␥ (NIH 3T3-PPAR␥ K422Q). Both WT and K422Q receptors were expressed at equivalent levels (Fig. 7A). Unlike the case with colon cancer cells, exposure of cells expressing either WT or K422Q PPAR␥ resulted in growth inhibition (Fig. 7B). Similarly, both WT and K422Q PPAR␥ could induce fibroblasts to differentiate into adipocytes, as indicated by ligand-induced increases in the adipocyte differentiation marker aP2 (Fig. 7C) and lipid accumulation (Fig.  7D).
The K422Q Apo-receptor Cannot Repress the Basal Expression of Target Genes in Colon Cancer Cells-The K422Q allele could not induce colon epithelial differentiation markers. However, these are proteins that are only elevated after 5-7 days of treatment with PPAR␥ agonists and are unlikely to be direct PPAR␥ target genes. To better understand why the presence of the K422Q mutation causes functional resistance to PPAR␥ agonists in colon cancer cells, we determined the effects the mutation had on the regulation of direct PPAR␥ target genes. Three PPAR␥ target genes, adipophillin, Gob-4, and TSC-22, which are all induced within 12 h of PPAR␥ ligand exposure in colon cancer cells, were selected for further study. We have previously identified adipophilin and Gob-4 as PPAR␥ target genes in a different colorectal cancer cell line (24). Adipophilin (also known as adipose differentiation-related factor) is a protein involved in fatty acid storage (36) whereas Gob-4 is a secreted protein associated with mature intestinal goblet cells (37). TSC-22 is a putative leucine zipper containing transcription factor originally identified as a TGF-␤ inducible gene (38). We have, in a previous microarray screen, identified and characterized TSC-22 as a direct PPAR␥ target gene in colorectal cancer cells. 2 The expression levels of these three PPAR␥ target genes was determined by Northern blot hybridization in HCT 15-pMSCV, HCT 15-PPAR␥ WT, and HCT 15-PPAR␥ K422Q cell lines treated with vehicle or rosiglitazone (Fig. 8). There was a 3-fold difference between WT and K422Q PPAR␥ in ligand-dependent 2 R. A. Gupta and R. N. DuBois, unpublished results. induction of a target gene (i.e. adipophilin). The mutant receptor showed no defect in ligand-dependent repression of a target gene (i.e. Gob-4). However, as compared with WT apo-PPAR␥, the most striking defect of the K422Q mutation was an inability of the apo-receptor to repress the basal expression of the PPAR␥ target gene TSC-22.
The ability of some members of the nuclear hormone receptor superfamily, notably the RARs and thyroid hormone receptors, to actively repress gene expression in the absence of ligand has been well documented (reviewed in Refs. 39 and 40). This activity is dependent on their ability to bind to transcriptional co-regulators termed co-repressors. Two of the best characterized co-repressors are N-CoR (nuclear receptor co-repressor) (41) and SMRT (silencing mediator for RAR and thyroid hormone receptors) (42). Although prior studies have failed to demonstrate the ability of either N-CoR or SMRT to bind DNAbound apo-PPAR␥, both of these proteins can bind to PPAR␥ in solution, suggesting their possible involvement in apo-PPAR␥mediated transcriptional repression (43). Thus, we tested whether apo-K422Q PPAR␥ is unable to repress basal TSC-22 expression because of a defect in co-repressor binding. However, in a mammalian two-hybrid assay, there was no difference between WT and K422Q PPAR␥ in their ability to bind to, or exhibit ligand-dependent release from, N-CoR or SMRT (data not shown).

DISCUSSION
Previous studies have established that activation of PPAR␥ is capable of inducing terminal differentiation in both adipocyte and epithelial cell lineages. Here we report the identification of four colorectal cell lines that are resistant to PPAR␥ ligand-induced growth inhibition. Each cell line contained a point mutation in one allele of the PPAR␥ gene that resulted in a change from Lys to Gln at codon 422 (K422Q). We also provide functional evidence to indicate that the ligand-activated mutant receptor is unable to induce colorectal cancer cell differentiation but is able to induce adipocyte differentiation.
The K422Q mutation appears to be defective in the basal repression of certain target genes in colon cancer cells. These results are relevant in understanding two broader issues, 1) loss of normal PPAR␥ signaling during the development of colorectal cancer and 2) mechanisms of PPAR␥-mediated differentiation in distinct cell types.
Loss of Normal PPAR␥ Signaling during the Development of Colorectal Cancer-Sarraf et al. (28) identified four of 55 primary colorectal tumors that harbored loss of function somatic mutations in either exon 3 or 5 of the PPAR␥ gene. Because PPAR␥ induces growth arrest in cultured colon cancer cells, these results suggested that some colorectal tumors undergo genetic selection for loss of PPAR␥ signaling during the development of colorectal cancer. Our results provide further support for this hypothesis by documenting four established colon cancer cell lines that are all functionally resistant to PPAR␥ ligands and that all contain a mutant PPAR␥ allele.
Information on the four patients from whom the resistant HCT 15, MIP 101, Clone A, and COLO 205 cell lines were derived is limited. Thus, it is unknown whether the K422Q mutation is a germline mutation, occurred in the primary colorectal tumors of these individuals, or was established only after these cells were cultured in vitro. The K422Q mutation was not found in an earlier study screening for PPAR␥ gene mutations in 55 primary colorectal tumors. A more recent study failed to detect any PPAR␥ gene mutations in a large number of clinical samples including cancers derived from the breast, colon, prostate, and lung (44). The conclusion of the study was that loss of function mutations of the receptor is extremely rare in cancer. However, this study limited its screen to exons 3 and 5 of PPAR␥. The K422Q mutation lies in exon 6, and our results emphasize that studies designed to screen for PPAR␥ gene mutations in cancers should be extended to include exons spanning the entire coding region. We are currently screening clinical samples of colorectal cancer to determine the incidence of the K422Q mutation in primary colorectal tumors. It will also be of interest to know whether the K422Q mutation is associated with metabolic syndromes such as insulin resistance or obesity for which germline PPAR␥ mutations have been identified previously (45,46).
The K422Q mutation in the four resistant cell lines was only identified in one allele of the gene whereas the other allele encoded for wild-type receptor. Similarly, in the earlier report by Sarraf and colleagues (28), the four PPAR␥ mutations found in primary colorectal cancers were all monoallelic with no evidence for loss of heterozygosity. However, it is not clear in these instances whether the remaining WT receptor is being expressed. In the HCT 15 cells, 10 independent PPAR␥ cDNA fragments that span exon 6 were cloned and sequenced using reverse transcriptase PCR; all 10 clones contained the K422Q mutation (data not shown). This would suggest that, at least in this cell line, the WT receptor is not expressed or is present at very low levels. In fact, in tumors that contain one mutated allele of PPAR␥, the other allele might be silenced through alterative mechanisms (e.g. promoter methylation).
Nevertheless, PPAR␥ does not appear to fit the classic Knudtson "two hit" hypothesis in which both alleles of a tumor suppressor are genetically inactivated. As opposed to studies that are limited only to analysis of DNA from primary colorectal cancers, our experiments with established cell lines has allowed us to conduct functional experiments that demonstrate that colon cancer cells with one mutant and one wild-type PPAR␥ allele are resistant to PPAR␥ ligand-induced growth inhibition. In theory, this loss of normal PPAR␥ signaling could occur through a dominant-negative or haploinsufficiency mechanism. It is possible that in a situation where both WT and K422Q PPAR␥ are co-expressed in the same cell line, the K422Q receptor could out-compete the WT receptor for a limiting number of binding sites (e.g. to RXR or to specific DNA elements) and thus inhibit the function of the WT receptor. Alternatively, it could simply be a gene dosage effect. PPAR␥Ϯ embryonic stem cells have a reduced capacity to differentiate into adipocytes as compared with WT cells (11), and mice heterozygous for PPAR␥ have a greater incidence of colorectal cancer as compared with control animals (47); both of these experiments establish that haploinsufficiency can occur for the PPAR␥ locus.
Mechanisms of PPAR␥-mediated Differentiation in Distinct Cell Types-The molecular mechanisms by which PPAR␥ initiates terminal differentiation pathways remain largely unknown. Here we have identified a PPAR␥ mutation that highlights a region of the receptor that appears to be essential for an interaction (and activity) necessary for receptor-induced colon epithelial, but not adipocyte, differentiation. What is the nature of this interaction? Analysis of direct genes regulated by either WT or K422Q PPAR␥ in the HCT 15 cells suggests that the mutation may cause a defect in the ability of the aporeceptor to repress gene expression. Thus, the mutation may disrupt interactions with an as yet unidentified co-repressor. These findings also imply that this type of target gene repression is only important in the ability of PPAR␥ to induce terminal differentiation in specific cell lineages (i.e. colon epithelial) and not others. It is also possible that the K422Q is defective in another unidentified receptor activity independent of repression (and that the defect in gene repression is a secondary event).
Finally, our evidence that PPAR␥ induces colon epithelial and adipocyte differentiation through distinct mechanisms implies that selective PPAR␥ modulators could be developed that specially target differentiation in particular cell lineages. For example, a PPAR␥ ligand capable of inducing epithelial, but not adipocyte, differentiation might limit the undesirable side effects because of modulation of adipocytes in treatments primarily aimed at inducing terminal differentiation of an epithelial cancer. As a first step in pursuing this line of research, our current focus is to identify biochemical target(s) that differentially bind to WT or K422Q PPAR␥.