Peroxisome Proliferator-activated Receptor γ and Transforming Growth Factor-β Pathways Inhibit Intestinal Epithelial Cell Growth by Regulating Levels of TSC-22*

Peroxisome proliferator-activated receptor γ (PPARγ) and transforming growth factor-β (TGF-β) are key regulators of epithelial cell biology. However, the molecular mechanisms by which either pathway induces growth inhibition and differentiation are incompletely understood. We have identified transforming growth factor-simulated clone-22 (TSC-22) as a target gene of both pathways in intestinal epithelial cells. TSC-22 is member of a family of leucine zipper containing transcription factors with repressor activity. Although little is known regarding its function in mammals, theDrosophila homolog of TSC-22,bunched, plays an essential role in fly development. The ability of PPARγ to induce TSC-22 was not dependent on an intact TGF-β1 signaling pathway and was specific for the γ isoform. Localization studies revealed that TSC-22 mRNA is enriched in the postmitotic epithelial compartment of the normal human colon. Cells transfected with wild-type TSC-22 exhibited reduced growth rates and increased levels of p21 compared with vector-transfected cells. Furthermore, transfection with a dominant negative TSC-22 in which both repressor domains were deleted was able to reverse the p21 induction and growth inhibition caused by activation of either the PPARγ or TGF-β pathways. These results place TSC-22 as an important downstream component of PPARγ and TGF-β signaling during intestinal epithelial cell differentiation.

The pathways that induce intestinal epithelial differentiation are complex and multigenic (1,2). Two biologically important regulators of intestinal epithelial cell growth are the peroxisomal proliferator-activated receptor-␥ (PPAR␥) 1 and transforming growth factor-␤ (TGF-␤) pathways. PPAR isoforms ␣, ␦, and ␥ constitute a family of orphan nuclear hormone receptors that serve to integrate dietary fat intake with changes in the expression of genes involved in both fatty acid oxidation and storage (3). PPAR␥ was originally defined as an essential component of the adipocyte differentiation program (4). However, the range of biological functions that PPAR␥ regulates is much wider than originally thought and the receptor has been shown to modulate the growth of a wide variety of cell types (5)(6)(7). This is perhaps most evident in the colon, where activation of the receptor inhibits the growth of a broad range of colon cancer cell lines and induces markers of differentiation (8,9). The effects of PPAR␥ on colorectal carcinogenesis may be dependent on the model system used because activators of the receptor slightly increase the number of colonic adenomas in the min mouse (a murine model of familial adenomatous polyposis) (10,11). However, ϳ8% of primary colorectal tumors were found to harbor a loss of function mutation in the PPAR␥ gene emphasizing the putative role of this receptor as a tumor suppressor in humans (12).
The TGF-␤ family of growth factors regulate a plethora of biological processes including embryonic development, wound healing, angiogenesis, proliferation, and differentiation of cells (reviewed in Ref. 13). This latter function has been well defined in the colon, where TGF-␤ is a potent inhibitor of colonic epithelial cell growth (14,15). Loss of normal TGF-␤ responsiveness occurs commonly during the development of colorectal cancers associated with microsatellite instability and genetic lesions that disrupt the TGF-␤ pathway have been identified, including loss of function mutations in the TGF-␤ type II receptor (16,17), SMAD4 (18), and SMAD2 (19). Under these circumstances continued expression of TGF-␤ paradoxically leads to enhanced tumor growth through stimulation of angiogenesis, extracellular matrix production, and immuno suppression (20).
Thus, although both PPAR␥ and TGF-␤ are key regulators of epithelial cell biology, the molecular mechanisms by which either pathway induces growth inhibition and differentiation are incompletely understood. Here we have identified transforming growth factor simulated clone-22 (TSC-22) as a target gene of both pathways in intestinal epithelial cells. Functional studies with wild-type and dominant negative forms of TSC-22 suggest that TSC-22 is an important downstream component of PPAR␥ and TGF-␤ signaling during intestinal epithelial cell differentiation.

Cell Culture and Materials
The MOSER S (M-S) (21) and MOSER R (M-R) (22) colon carcinoma lines were a gift of M. Brattain (University of Texas Health Sciences, San Antonio, TX). The FET cell line was a gift of W. Grady (Vanderbilt University, Nashville, TN) and the CBS cell line was a gift of H. Moses (Vanderbilt University). COS7 cells were purchased from ATCC (Rockville, MD). Cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone), L-glutamine (2 mmol/liter), penicillin (100 units/ml), and streptomycin (100 g/ml) in a 5% CO 2 atmosphere with constant humidity. TGF-␤1, cycloheximide, and 5,6-dichloro-␤-D-ribofuranosyl-benzimidazole (DRB) were purchased from Sigma. Synthetic PPAR and RXR ligands were synthesized at GlaxoSmithKline (Research Triangle Park, NC) and dissolved in Me 2 SO. RNA was isolated using the TRI reagent (Molecular Research Center, Cincinnati, OH). For all experiments in which TGF-␤1 or a PPAR ligand were added, the cells were grown in the above media except that regular fetal bovine serum was replaced by charcoal-stripped fetal bovine serum (Hyclone).

High Density cDNA Microarray Screening
Human named genes GENEFILTERS Release I blots (Research Genetics Inc., Huntsville, AL) were probed with [ 33 P]dCTP-labeled reverse transcriptase-derived cDNA. The cDNA was synthesized from RNA isolated from exponentially growing M-S cells that had been exposed for 12 or 24 h to one of the following treatments: 0.1% 4 mM HCl containing 0.1% bovine serum albumin, 0.1% Me 2 SO, 1 M rosiglitazone, or 2 ng/ml TGF-␤1. Membranes were hybridized and washed according to the manufacturer's instructions. The blots were imaged using the Cyclone Storage Phosphor System (Hewlett-Packard) and imported into the Pathways Software (Research Genetics Inc., Huntsville, AL) to identify differences in the intensity of cDNA spots between different treatments.

Northern Hybridization Analysis
Northern blot analysis was performed as described previously (23). Although representative Northern blot images are shown, all Northern blot experiments were repeated a minimum of two times. Briefly, total RNA (20 g) was fractionated on a 1.2% agarose-formaldehyde gel and transferred to a Hybond-NX nylon membrane (Amersham Biosciences). Filters were prehybridized for 4 h at 42°C in Ultrahyb (Ambion, Austin, TX). Hybridization was conducted in the same buffer in the presence of 32 P radiolabeled cDNA of a partial fragment of human TSC-22. Blots were washed 4ϫ for 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. Quantification of mRNA intensities for Fig. 4 (A-D) was done using Optiquant Software (Hewlett-Packard).

DNA Constructs
For all PCR reactions, Pfu Turbo Taq Polymerase (Stratagene, La Jolla, CA) was used. Every construct generated was sequenced to ensure the absence of any unwanted mutations. Full-length TSC-22 was cloned using PCR from a cDNA library of M-S cells treated with 1 M rosiglitazone for 24 h. The product was cloned into the pPCR-Script cloning vector (Stratagene, La Jolla, CA) and this plasmid was utilized as a template for all subsequent PCR reactions. For all primers, the 5Ј primer was designed to include a NotI site and nucleotides encoding the full-length HA or FLAG epitope. Each 3Ј primer was designed to include a BglII site. For HA TSC-22 wild-type (wt) or FLAG TSC-22 wt, the 5Ј primer included a partial region of TSC-22 starting at the second codon and the 3Ј primer included a partial region starting at codon 145 (stop codon). For HA TSC-22dn, in which both repressor domains (RD) were deleted, the 5Ј primer included a partial region of TSC-22 beginning at codon 38. The 3Ј primer included a stop codon and a partial region starting at codon 102. For HA TSC-22dncon, both RDs were deleted and two highly conserved leucine residues in the leucine zipper domain were mutated to alanine (L91A and L97A). The 5Ј primer was identical to that used for HA TSC-22dn. The 3Ј primer was also the same except for a mismatch at codons 91 and 97 such that the amplified product would contain the L91A and L97A mutations. See Fig. 5A for a schematic of all constructs generated. Each PCR product was digested with NotI/BglII and subcloned into the pcDNA3.1Zeo expression vector (Invitrogen, La Jolla, CA).

Transient Transfections
Exponentially growing COS7 cells were transiently transfected with the appropriate plasmid using FuGENE 6 (Roche Molecular Biochemicals, Indianapolis, IN) at a lipid:DNA ratio of 3.5:1 and with a DNA concentration of 1.5 g/ml in Opti-MEM (Invitrogen). 6 h post-transfection, regular media was added back and the cells were allowed to grow for another 48 h prior to harvest.

Immunoprecipitations
Anti-HA affinity matrix (100 l; Roche Molecular Biosciences) was added to 500 g of COS7 cell lysate and the cells were incubated in NET-N buffer (150 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl, pH 8.0, 0.5% Nonidet P-40, and 10% glycerol) at 4°C for 4 h. Beads were then washed 3ϫ with NET-N buffer. Protein was eluted by the addition of 1ϫ SDS loading buffer followed by a 5-min incubation at 100°C. Transfection-M-S cells were transfected with the appropriate vector using FuGENE 6 at a lipid:DNA ratio of 3:1 and a DNA concentration of 5 g/ml in Opti-MEM. After 6 h of transfection, normal media was added back and the cells were allowed to grow for another 72 h.

Stable Cell Lines
Selection-Cells were then split into media containing 200 g/ml zeocin (Invitrogen, La Jolla, CA) and selected for 2 weeks. Resistant clones were combined into two unique pools and expanded. Zeocin was maintained in the media prior to all experiments at a dose of 100 g/ml.

Cell Growth Measurements
Cells were plated at a density of 5 ϫ 10 4 and replaced the next day with Dulbecco's modified Eagle's medium containing 10% charcoalstripped fetal bovine serum and rosiglitazone or TGF-␤1. Cells were counted at the indicated times using a Coulter counter. Each experiment was done in triplicate.

In Situ Hybridization
In situ hybridization was performed as described previously (24). Sense or antisense 35 S-labeled cRNA probes were generated from human TSC-22. The probes had specific activities at 2 ϫ 10 9 disintegrations per minute (dpm)/g. Sections hybridized with the sense probes did not exhibit any positive autoradiographic signals and served as negative controls.

Cellular Response of Colorectal Cancer Cells to PPAR␥ and
TGF-␤1-We noticed a striking similarity in the cellular response to either TGF-␤1 or PPAR␥ ligands in colorectal cancer cells with intact PPAR␥ and TGF-␤ signaling pathways. For example, in the M-S, CBS, and FET colon carcinoma lines, exposure to the high affinity PPAR␥ ligand rosiglitazone (25) or TGF-␤1 results in accumulation of cells in the G 1 phase of the cell cycle (data not shown) and a decrease in cell growth (Fig.  1A). Moreover, in all three cell lines, addition of either agent results in an increase in the p21 (Fig. 1B) and keratin 20 (data not shown). Elevated levels of the cyclin-dependent kinase inhibitor p21 (26,27) and keratin 20 (28) are associated with intestinal epithelial differentiation in various model systems and p21 is a well characterized TGF-␤-regulated gene (29). Because both PPAR␥ and TGF-␤1 regulate such a wide spectrum of physiological processes, the genomic response to either pathway is complex. This makes it difficult to identify those target genes that play functionally important roles in cell growth and differentiation. These data led us to hypothesize that one way to identify relevant target genes of PPAR␥ and TGF-␤1 in the setting of colon epithelial cell growth and differentiation is to focus on the common subset of downstream target genes regulated by both pathways.
Identification of TSC-22 as a PPAR␥ and TGF-␤ Target Gene in Colon Epithelial Cells-Microarray analysis was utilized to determine the genomic response of one of these lines, the M-S cells, after exposure to rosiglitazone or TGF-␤1 for 12 or 24 h. A subset of genes was commonly regulated by either pathway, including several members of the keratin and carcinogenic embryonic antigen superfamilies. However, time course experiments suggested that these genes were not likely to be direct targets of PPAR␥ or TGF-␤ and may represent end points (rather than effectors) of the differentiation process. One promising candidate rapidly induced by either treatment was TSC-22, a gene originally identified as a TGF-␤1-stimulated gene in osteoblast cells (30). Northern blot analysis confirmed that either PPAR␥ or TGF-␤ could induce TSC-22 in the M-S, CBS, and FET cell lines (Fig. 2). We further became interested in studying this gene when in situ hybridization of TSC-22 in the normal colon demonstrated that its expression was enriched in the postmitotic epithelial compartment of the normal human colon, where the most differentiated enterocytes reside (Fig. 3).
Transcriptional Regulation of TSC-22 by PPAR␥ and TGF-␤1-The M-S cells were chosen as a system to study TSC-22 regulation by PPAR␥ or TGF-␤1 and to test the hypothesis that TSC-22 is a regulator of intestinal epithelial differentiation. The induction of TSC-22 by either PPAR␥ or TGF-␤1 was both time-and dose-dependent (Fig. 4, A-D). Following treatment with either rosiglitazone or TGF-␤1, anti-TSC-22 antiserum detected an appropriately sized polypeptide, which was absent in untreated cells (Fig. 4E). We conclude that TSC-22 protein levels increase as well. TSC-22 has been shown to be a direct target gene of TGF-␤1 (30). We were interested in determining whether it is also a target for PPAR␥ as well. Cells were pretreated with the protein synthesis inhibitor cycloheximide or the RNA polymerase II inhibitor DRB followed by rosiglitazone. DRB co-treatment blocked the ability of PPAR␥ to induce TSC-22 suggesting that the ability of PPAR␥ to increase steady-state RNA levels of the gene was dependent on de novo transcription rather than through an increase in mRNA stability (Fig. 5). Furthermore, whereas cycloheximide slightly induced TSC-22, co-treatment of rosiglitazone and cycloheximide led to a superinduction (Fig. 5), suggesting that de novo protein synthesis was not required for PPAR␥ to induce TSC-22. This result suggested that PPAR␥ was not inducing TSC-22 by first increasing the levels of TGF-␤ and in fact rosiglitazone was not able to induce TGF-␤1 in the parental M-S cells (data not shown).
As TSC-22 has not previously been shown to be a target of PPARs, we wanted to determine the specificity and selectivity of this induction. Cells were treated with two structurally distinct PPAR␥ ligands (the thiazolidinedione-based agonist rosiglitazone or the tyrosine-based agonist GW7845 (31)), an irreversible PPAR␥ antagonist (GW9662 (32)), or rosiglitazone plus GW9662. TSC-22 was induced by both PPAR␥ agonists and the induction by rosiglitazone could be blocked by cotreatment with the PPAR␥ antagonist (Fig. 6). Cells were also treated with a PPAR␣ selective ligand (GW7647 (33)), a dual PPAR␣/␦ ligand (GW2433 (34)), an RXR specific ligand (LG100268), or a combination of rosiglitazone and LG100268. Despite the fact that M-S cells express both PPAR␣ and PPAR␦ (data not shown), neither of these other two PPAR isoforms was able to regulate TSC-22 expression (Fig. 6).
The Ability of PPAR␥ to Induce TSC-22 Is Independent of TGF-␤1-To further clarify the issue of whether PPAR␥ signaling in the colon was dependent on the TGF-␤ pathway, the M-R cell line was utilized. The M-R line is a subclone of the parental M-S cells that are relatively refractory to the growth inhibitory effects of TGF-␤ (22). PPAR␥, but not TGF-␤1, could induce expression of TSC-22 in the M-R cells (Fig. 7A). Finally, activators of PPAR␥ inhibited the growth of either cell line by equivalent amounts, whereas the M-S-R cells (as has been previously reported) were relatively resistant to TGF-␤ induced growth inhibition (Fig. 7B).

Overexpression of Wild-type TSC-22 Inhibits Colon Epithelial Cell Growth and Induces Elevated Levels of p21-
The functional role of TSC-22 in mediating any of the biological effects induced by either PPAR␥ or TGF-␤ is unknown. To address this issue we focused on two experimental strategies: 1) to determine whether expression of TSC-22 in the M-S cells could recapitulate any of the phenotypic changes induced by PPAR␥ or TGF-␤1, and 2) to determine whether inhibition of normal TSC-22 function blocks the phenotypic changes induced by either treatment. Kester et al. (35) have defined 4 functional domains within TSC-22 (Fig. 8A). Using both glutathione Stransferase pull-down and mammalian two-hybrid assays, they demonstrated that TSC-22 could form homodimers via the leucine zipper domain. They also showed that TSC-22 has transcriptional repressor activity when fused to a heterologous DNA-binding domain and identified two domains within TSC-22, RD 1 and 2, which were in large part responsible for this effect. Finally, they were able to demonstrate that titration of a mutant TSC-22 in which both RD 1 and 2 were deleted could inhibit the repressor activity of the wild-type protein and thus act as a dominant-negative (dn) inhibitor. Based on these results, we constructed three different HA-tagged TSC-22 constructs for stable introduction of this gene into the M-S cells: TSC-22 wt, TSC-22dn (in which both RD 1 and 2 were deleted), and TSC-22dncon (which, in addition to both RD domains being deleted, contains mutations in two highly conserved leucine residues within the leucine zipper domain) (Fig. 8A). Co-immunoprecipitation experiments confirmed that TSC-22dn, but not TSC-22dncon, could dimerize with TSC-22 wt (Fig. 8B). This last construct was made to help properly interpret any potential artifacts because of expression of TSC-22dn that were independent of its ability to dimerize and inhibit the function of the wild-type protein.
Each of these three constructs (plus empty vector) was used to generate eight unique pools of stably transfected cells. Each of the pooled cell lines was found to express relatively equivalent protein levels of the integrated cDNA (Fig. 8, C and D). The two different pools of M-S cells expressing TSC-22 wt were found to have significantly reduced growth rates compared with vector-transfected cells (Fig. 9A). They also displayed higher levels of p21 protein but showed no difference in the levels of keratin 20 protein (Fig. 9B).
Overexpression of Dominant Negative TSC-22 Partially Inhibits PPAR␥ Ligand and TGF-␤1-induced Growth Inhibition and p21 Induction-We next tested the ability of TSC-22dn and TSC-22dncon to block the ability of PPAR␥ or TGF-␤ to inhibit the growth of M-S cells. In the two different pools of cells expressing TSC-22dn, the ability of PPAR␥ or TGF-␤1 to inhibit growth was reduced by ϳ60% (Fig. 10A). Importantly, no differences in the inhibitory activity of either PPAR␥ or TGF-␤1 were seen when comparing vector and TSC-22dncontransfected cells (Fig. 10A). Northern blot analysis confirmed that PPAR␥ and TGF-␤1 could still induce TSC-22 in these cell lines (data not shown). Finally, the ability of rosiglitazone or TGF-␤1 to induce p21, but not keratin 20, was greatly dimin-ished in M-S cells expressing TSC-22dn but not TSC-22dncon (Fig. 10B). DISCUSSION The complex mechanisms by which the undifferentiated stem cells of the intestine give rise to differentiated cells with specialized functions remain incompletely understood. Some of the key pathways that govern this process in the intestine have been identified from understanding genetic lesions found in colorectal cancer (36). For example, a majority of human colorectal cancers contain a loss of function mutation in the tumor suppressor gene adenomatous polyposis coli (37). Subsequent studies have suggested a critical role for this gene in exerting strict growth controls such that the renewing population of intestinal epithelial cells is kept at a constant number. Likewise, loss of function mutations have been identified in genes involved in the TGF-␤ signaling pathway and PPAR␥. Further studies have shown that activation of either of these pathways in cell culture models can lead to growth inhibition and the induction of markers of differentiation. These experiments emphasize the biological relevance of these two pathways in intestinal epithelial cell biology. In this study, we used microarray technology to identify TSC-22 as a target gene of both PPAR␥ and TGF-␤ in intestinal epithelial cells. We further show using wild-type and dominant negative forms of TSC-22 that this gene plays an important role in the ability of either PPAR␥ or TGF-␤ to inhibit cell growth.
There are several reasons why the results presented here are likely to be biologically relevant. First of all, TSC-22 is localized to the most differentiated epithelia in the normal human colon. PPAR␥ and TGF-␤1 have been previously shown to be most strongly expressed in vivo in the most differentiated epithelial cells of the colonic mucosa (38,39). Thus, all three of these genes appear to be co-localized in the normal human colon to the most differentiated enterocytes. Second, PPAR␥ and TGF-␤1 were shown to regulate TSC-22 in multiple colon cancer cell lines, emphasizing the fact that the present findings are not simply limited to one unique colon cancer cell clone. In addition, the functional experiments in which a dominant negative TSC-22 could reverse the growth inhibition caused by PPAR␥ ligands or TGF-␤1 are likely because of its ability to inhibit normal TSC-22 function. This is because no such reversal was seen with a control dominant negative construct that cannot dimerize with and inhibit wild-type TSC-22. We also provide evidence that there is specificity to this putative signaling system. For example, TSC-22 was not induced by PPAR␣ or -␦ (neither of which induces epithelial cell differentiation). In addition, the induction of TSC-22 does not appear to be a general phenomenon of agents that inhibit cell growth because the RXR ligand LG100268 did not induce TSC-22 (Fig.  6) but does inhibit the growth of multiple cell types, including the M-S cells (data not shown). Finally, TSC-22 did not have any effect on the expression of keratin 20 suggesting that TSC-22 is involved in one facet of the differentiation program (control of cell growth) but is not involved in pathways that lead to changes in the expression of certain structural proteins. TSC-22 is highly conserved during evolution, with the human protein sequence being 98.5% identical to the mouse and rat proteins (40). TSC-22 has been shown to a have a solution structure similar to members of the bZIP family of transcription factors (41). Similar to bZIP proteins, TSC-22 contains a leucine zipper domain and, analogous to the basic domain in bZIP family members, has a highly conserved sequence known as the TSC domain or box. In addition, TSC-22 is also homologous to the Drosophila melanogaster gene shortsighted (shs) or bunched, which is required for normal development of oocytes, larval peripheral nervous system, and adult eye and wing (42)(43)(44). Recent evidence also implicates an important role for TSC-22 in the developing feather tract (45). There is also evidence suggesting that TSC-22 can negatively regulate the growth of cancer cells. For example, antisense expression of TSC-22 in a salivary gland carcinoma line inhibits the growth of these cells in vivo (46) and overexpression of TSC-22 appears to sensitize carcinoma cells to certain anti-cancer chemotherapeutic drugs (47).
Evidence to date suggests that TSC-22 is part of a family of proteins that act as transcriptional repressors. As mentioned earlier, TSC-22 has been shown to homodimerize and exhibit transcriptional repressor activity when fused to a heterologous DNA-binding domain (35). In fact, our results with the TSC-22dn protein in which both repressor domains were deleted provide functional evidence in a biological system for the importance of these domains. This notion is further supported by the finding that the TSC-22 Drosophila homolog, bunched, is a powerful repressor of the enhancer trap reporter A359 (44). It is also possible that TSC-22 may modify gene expression through protein-protein interactions. Recently, a TSC-22 homologue (termed THG-1) was cloned and found capable of forming heterodimers with TSC-22 (35). (THG-1 was not found to be expressed in the M-S cells.) Exactly how either a TSC-22 homodimer complex or a TSC-22⅐THG-1 heterodimer complex can regulate the expression of a particular target gene is unknown.
One possible target gene of TSC-22 may be p21. However, at this point it is unclear whether TSC-22 increases p21 levels directly or indirectly. Although our work provides evidence that TSC-22 is involved in the signaling pathway by which both PPAR␥ and TGF-␤ induce p21, there are other mechanisms by which activation of either pathway may modulate p21 expression. For example, TGF-␤ has been shown to activate p21 proximal promoter activity via two consensus Sp1-binding sites (48,49). How (or if) TSC-22 may act to enhance basal Sp1 transcriptional activity is unknown. Clearly, a major goal of future work will be to identify genes that are directly regulated by TSC-22 and determine the mechanism(s) by which TSC-22 modulates the expression of these genes. Determining TSC-22 target genes in the intestine will lead to much better understanding of the exact role of this protein during colon epithelial differentiation.
In summary, the data reported here point to a previously unidentified role for the putative transcriptional repressor TSC-22 as a regulator of intestinal epithelial cell differentiation. A large percentage of advanced colorectal tumors lose their responsiveness to growth inhibition induced by TGF-␤. It will be of interest to determine what percentage of these colorectal tumors also contain mutations in PPAR␥. If this is not found to be a common occurrence, then our present studies predict that treatment with activators of PPAR␥ may provide a way to bypass the loss of normal TGF-␤ signaling that occurs during the progression of these tumors.
Acknowledgments-We thank Howard Crawford for help throughout this project, Chris Ferris for reviewing the manuscript, and Jian Tan B, each indicated cell line was treated with vehicle, 1 M rosiglitazone, or 2 ng/ml TGF-␤1, and protein lysates were collected after 4 days. 50 g of total protein was loaded and membranes were probed with anti-p21 or anti-keratin 20 antibodies. and Sudhansu K. Dey for their generous help with the in situ hybridization studies.