Molecular Alterations Associated with Sulindac-Resistant Colon Tumors in Apc^Min/+ Mice

Sulindac and other nonsteroidal anti-inflammatory drugs (NSAID) have been shown to be effective chemopreventive agents for colorectal cancer (1, 2). In human familial adenomatous polyposis (FAP) patients, sulindac treatment causes an inhibition of aberrant crypt foci formation as well as a reduction in the number and size of adenomas (3–7). Unfortunately, long-term sulindac treatment has several drawbacks. Side effects include gastrointestinal bleeding and ulceration, which limit its clinical use (8). In addition, numerous studies have found that sulindac often does not cause a complete regression of adenomas in FAP patients, and in some cases, colorectal cancer develops in patients with FAP after receiving sulindac treatment (4, 9–13). Also, in the Apc^Min/+ mouse, the murine model of FAP, sulindac was found to inhibit small intestinal tumors, but actually increased colon tumor incidence, multiplicity, and volume (14).

The mechanism by which sulindac can inhibit intestinal tumorigenesis is strongly dependent on its ability to inhibit cyclooxygenase (COX)-1 and COX-2 enzymes and, therefore, the production of proliferative and inflammatory prostaglandins (PG), including, most notably, PGE₂ (15, 16). However, there is evidence that sulindac can also act via COX-1/COX-2–independent mechanisms (17, 18). For example, the deregulation of Wnt/β-catenin signaling and subsequent accumulation of nuclear β-catenin is very common in colon adenomas and is mainly due to mutations in the Apc and β-catenin genes (19). It has been shown, however, that sulindac treatment can induce the degradation of β-catenin protein in colon cancer cells, thus inhibiting its nuclear translocation (20). Furthermore, in normal differentiated cells, β-catenin is maintained as part of a protein complex at the plasma membrane, binding E-cadherin to the actin cytoskeleton (21). E-cadherin regulates cell adhesion in epithelial cells and is attached to the actin cytoskeleton through interactions with α-catenin and β-catenin (22). It is believed that downregulation of E-cadherin causes the initiation of an abnormal epithelial-mesenchymal transition that occurs in invasive cancer cells (23). Interestingly, sulindac has been shown to increase the production of E-cadherin protein in cancer cells (24). This may in fact be a result of the increased pool of β-catenin that is available to bind to E-cadherin (22).

Sulindac has also been shown to target other signaling pathways. For example, microarray experiments have shown that the cyclin-dependent kinase inhibitor p21^WAF1/cip1 is significantly upregulated in response to sulindac treatment in colon cancer cell lines and rectal biopsies, resulting in inhibition of cell proliferation (25). Subsequent studies using a variety of different mouse models, including Apc^Min/+ or Apc1638^+/−, p21^+/+, p21^+/−, or p21^−/−, have shown that p21 may be critical for the tumor-suppressive properties of sulindac, and that disruption of even one p21 allele may be sufficient to abrogate intestinal tumor inhibition by sulindac (26, 27). Other cellular targets may also mediate the effects of sulindac. For example, sulindac can bind to and activate peroxisome proliferator–activated receptors (PPARγ or PPARδ), which have been shown to act either as a tumor suppressor or as an oncogene in colon cancer, respectively (28, 29). Also, in prostate cells, it has been shown that PPARγ is required for both growth inhibition and p21 upregulation by sulindac sulfide (30).

In the following study, we have evaluated the effects of sulindac on colon cancer using Apc^Min/+ mice. As anticipated, sulindac treatment resulted in a profound suppression of the growth of tumors in the small intestine. However, in the colon, the effects of sulindac treatment were less straightforward. Although 9 weeks of sulindac exposure inhibited the growth of tumors, tumor multiplicity was actually significantly increased. Importantly, PGE₂ production was suppressed by sulindac to comparable extents in the two organs, indicating that regulation of non-COX targets of sulindac during tumorigenesis might be responsible for the observed increase in colon tumor multiplicity. The survival of tumors that have lost p21 and PPARγ expression following sulindac treatment may underlie the relative lack of efficacy of the drug in the colons of Apc^Min/+ mice and may provide new insights into the incomplete suppression of adenomas in a subset of human patients.

Beginning at 5 weeks of age, Apc^Min/+ mice were given sulindac (Sigma-Aldrich) at 180 ppm in drinking water buffered with 40 mmol/L sodium phosphate ad libitum. Control mice were given water without drug. All mice were maintained in a temperature-controlled, light-cycled room and allowed free access to drinking water (with or without sulindac) and standard diet (LM-485, Harlan Laboratories). Animals and food were weighed once weekly and mice were checked weekly for signs of weight loss or lethargy indicating intestinal obstruction or anemia associated with tumors. Animal experiments were conducted with approval from the Center for Laboratory Animal Care Committee, University of Connecticut Health Center.

Both control and sulindac-treated Apc^Min/+ mice were sacrificed at 14 weeks of age (9 weeks of sulindac treatment) for polyp scoring, histologic analyses, and RNA/protein/lipid extraction from tissue. The entire small intestine and colon were harvested and flushed with ice-cold PBS and excised longitudinally. Specimens were fixed flat in 10% neutral buffered formalin solution for 4 hours and stored in 70% ethanol. Tissues were stained with 0.2% methylene blue and the number and size of tumors were scored under a dissecting microscope. Small intestines and colons were then Swiss-rolled and paraffin-embedded and sections were mounted onto glass slides and stained with H&E for histologic analysis. Tissues (normal and tumor) from a portion of control and sulindac-treated mice were snap-frozen in liquid nitrogen at the time of sacrifice for RNA/protein extraction and eicosanoid analysis.

The wild-type (WT) HCT116 cells and a p21^−/− variant of HCT116 cells were generously provided by Dr. Bert Vogelstein (Johns Hopkins University, Baltimore, MD) and maintained in McCoy's 5A medium supplemented with 10% (v/v) fetal bovine serum and 1% penicillin/streptomycin. Cells were treated with sulindac sulfide (Sigma-Aldrich) at varying concentrations (as indicated) for 24 hours.

Apoptosis induced by sulindac sulfide in cell culture was assessed by (a) quantification of cytoplasmic histone-associated DNA fragmentation and (b) flow cytometric analysis of cells with sub-G₀/G₁ DNA. The effect of sulindac on cytoplasmic histone-associated DNA fragmentation was determined using the Cell Death Detection ELISA^PLUS kit (Roche Diagnostics) according to the manufacturer's instructions. The sub-G₀/G₁ fraction in control and sulindac-treated cells was analyzed by flow cytometry. Briefly, cells were harvested with trypsin-EDTA, washed with 1× PBS, fixed in 70% ethanol, and stored at −20°C overnight. Cells were then resuspended in a staining solution containing 500 μg/mL RNase A and 500 μg/mL propidium iodide and incubated at 37°C for 1 hour. Analysis of DNA content was done using the BD FACSCalibur system and FlowJo software (Becton-Dickinson).

PGD₂, PGE₂, PGF_2α, 6-keto-PGF_1α, and thromboxane B₂ were quantified in tissue samples by a modification of the method of Luderer et al. (31) and Nichols et al. (32). Tissue (50-300 mg) from the distal small intestine or colon was snap-frozen in liquid N₂. Each homogenized tissue sample was added to 2 mL of 100% methanol, 10 μmol/L indomethacin, and 25 ng of PGD₂, PGE₂, PGF_2α, 6-keto-PGF_1α, and thromboxane B₂-D4 standard. After centrifugation, the supernatant was acidified with 1× PBS (pH 3) and PGs were extracted twice with 2 mL of chloroform. Samples were dried down under N₂ gas. PGs were derivatized using the method of Waddell et al. (33). PGs were first treated with 2% methoxylamine hydrochloride in pyridine (30 μL). After incubating overnight at room temperature, the samples were dried under N₂, dissolved in acetonitrile (30 μL), and treated with pentafluorobenzyl bromide [35% (v/v) in acetonitrile; 10 μL] and diisopropylethylamine (10 μL). The samples were vortexed, incubated for 20 minutes at 40°C, and evaporated under N₂. The residue was treated with bistrimethylsilyl-trifluoroacetamide (50 μL) and incubated at room temperature for 4 to 5 days. Gas chromatography-mass spectrometry (GC/MS) was carried out on a HP 5890 gas chromatograph interfaced with a 5988A mass spectrometer (Hewlett-Packard). PG samples were applied to a SPB-1 column (12 m × 0.2 mm, 0.33 μm film thickness; Supelco, Inc.) held at 100°C. PG samples were analyzed using a temperature program of 2°C/min from 100°C to 240°C. The injector block was held at 260°C and the transfer tube was maintained at 280°C. PG derivatives were detected using electron capture-negative chemical ionization (34). PG levels were quantified using selected ion monitoring of the characteristic base peak ions of the deuterated and authentic PGs.

Polyps and normal-appearing mucosa were harvested from the colons of 14-week-old Apc^Min/+ mice. Tissues were homogenized and total RNA was extracted with TRizol reagent (Invitrogen Corp.). cDNA was synthesized using SuperScript III according to the manufacturer's protocol (Invitrogen). mRNA expression levels were examined with TaqMan gene expression assays (Applied Biosystems, Inc.) using primer-probe combinations for Bax (Mm00432050_m1), Bcl-2 (Mm00477631_m1), or CDKN1a (p21; Mm00432448_m1). PCR amplification on a 7500 real-time PCR instrument (Applied Biosystems) was carried out by denaturing cDNA at 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. mRNA expression levels were normalized to hypoxanthine phosphoriboxyltransferase 1 (HPRT1).

For protein extraction, tissue was incubated in SDS buffer [0.125 mol/L Tris (pH 6.8), 2% SDS, 10% glycerol] at 70°C for 10 minutes, followed by sonication and centrifugation at 14,000 rpm for 15 minutes at 4°C. The supernatant was removed and quantified for total protein. Cells were lysed in a buffer containing 20 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, and 1% Triton X-100 supplemented with protease and phosphatase inhibitors (50 mmol/L NaF, 10 mmol/L Na β-glycerol phosphate, 5 mmol/L Na pyrophosphate, 1 mmol/L sodium vanadate, 1 μg/mL leupeptin, and 1 mmol/L phenylmethylsulfonyl fluoride). Following centrifugation at 14,000 rpm for 15 minutes at 4°C, the supernatant was removed and quantified for total protein. Thirty micrograms of protein were incubated at 95°C for 5 minutes with 4× sample loading buffer containing 8% β-mercaptoethanol and loaded onto a 10% to 15% polyacrylamide gel. Proteins were transferred onto an Immobilon-P membrane (Millipore Corp.) and blocked in 5% nonfat dry milk in TBS-T (1× TBS, 0.1% Tween 20) for 1 hour. Blots were incubated at 4°C overnight with primary antibody [i.e., anti-p21 (1:4,000; BD Biosciences), anti-PPARγ (1:1,000; Cell Signaling), anti-PARP (1:1,000; Cell Signaling), or anti–β-actin (1:6,000; Santa Cruz Biotechnology)]. Blots were washed multiple times and incubated with goat α-mouse IgG horseradish peroxidase (1:10,000; Upstate Biotechnology) or donkey α-rabbit IgG horseradish peroxidase (1:4,000; Santa Cruz Biotechnology) for 45 minutes at room temperature. Horseradish peroxidase was visualized with enhanced luminol reagent (Millipore).

Staining was done as described in our previous study (35). Briefly, small intestine and colon tissues were Swiss-rolled, paraffin-embedded and sectioned at 7-μm thickness. Tissue sections were deparaffinized and incubated with 3% hydrogen peroxide for 20 minutes at room temperature. Sections were subjected to antigen retrieval and blocked with 10% normal goat serum in PBS. Sections were then incubated overnight at 4°C with anti-p21 (1:100; Neomarkers), anti–proliferating cell nuclear antigen (PCNA; 1:150; Novacastra Laboratories), anti–cleaved caspase-3 (1:200; Cell Signaling), anti–E-cadherin (1:1,000; Abcam), anti–β-catenin (1:2,000; Sigma-Aldrich), or anti-p53 (1:1,000; Sigma-Aldrich) antibodies. Sections were washed and incubated with biotinylated antimouse secondary antibody for 30 minutes at room temperature. Sections were washed and then incubated with avidin-biotin complex reagent (Vector Laboratories) for 30 minutes at room temperature, followed by signal detection with 3,3′-diaminobenzidine solution (Vector Laboratories). Tissues were counterstained with hematoxylin.

The area and density of E-cadherin and caspase-3 immunostaining were determined using Image-Pro Plus 7.0 software (Media Cybernetics) using the area and density (sum) measurements.

For the comparison of size and multiplicity of polyps as well as the analysis of PG levels, Bax/Bcl-2 expression, and immunostaining pixel density, statistical analyses were done using Student's t test. P < 0.05 was considered statistically significant.

To determine the effect of sulindac treatment on colon tumorigenesis in Apc^Min/+ mice, we compared the multiplicity and size of colon polyps in sulindac-treated and control mice. Consistent with a previous study (14), analysis of methylene blue–stained colons revealed a 71% increase in colon tumor multiplicity following sulindac treatment (1.6 ± 0.4 versus 5.5 ± 0.6 in control and sulindac treated, respectively; P < 0.0001; Fig. 1A). However, morphometric analysis revealed a significant reduction in the size of the adenomas. There were no colon tumors >5 mm in diameter present in the sulindac-treated mice (0 ± 0 versus 0.9 ± 0.3 in the control group; P < 0.05). The number of small (1-2 mm) polyps increased by 97% (3.3 ± 0.3 versus 0.1 ± 0.1, respectively; P < 0.0001) and the number of medium (2-5 mm) polyps increased by 73% (2.2 ± 0.3 versus 0.6 ± 0.3, respectively; P < 0.01) in sulindac-treated compared with control mice (Fig. 1A). All tumors in both groups were in the distal region of the colon. These results suggest that although sulindac cannot prevent tumor initiation in the colons of Apc^Min/+ mice, tumor progression is markedly suppressed.

As shown in Fig. 1B, further histologic examination of the colons revealed the presence of large, polypoid, noninvasive tumors in the untreated controls. However, polyps from sulindac-treated mice were generally smaller in size and largely sessile in appearance. Histologically, the polyps shared similar features, and in both groups there was no evidence of tumor invasion into the submucosa (Fig. 1B).

As anticipated, sulindac treatment significantly reduced the total number of small intestinal polyps by 95% relative to untreated controls (2.3 ± 0.8 versus 42.0 ± 3.3, respectively; P < 0.0001; Fig. 1C). Sulindac treatment caused a similar reduction in the number of small (94%; 2.0 ± 0.7 versus 33.4 ± 3.1, respectively; P < 0.0001), medium (96%; 0.3 ± 0.2 versus 8.3 ± 1.3, respectively; P < 0.001), and large (100%; 0 ± 0 versus 0.3 ± 0.2, respectively; not significant) small intestinal adenomas (Fig. 1C). This indicates that sulindac is able to inhibit both the initiation and the progression of tumors in the small intestine of Apc^Min/+ mice.

To determine whether sulindac was effectively targeting COX-1/COX-2 activity in the small intestine and colon, we measured the tissue levels of a panel of prostanoids by GC/MS, including PGE₂, PGD₂, PGF_2α, thromboxane B₂, and 6-keto-PGF_1α. As shown in Fig. 2A and B, PGE₂ was the major PG produced in both organs. Sulindac treatment, however, reduced the levels of PGE₂ to comparable extents in the small intestine (27%) and colon (38%; Fig. 2A and B) relative to control tissue. Sulindac treatment also decreased the levels of PGD₂ (43% in the small intestine and 64% in the colon), 6-keto-PGF_1α (stable breakdown product of PGI₂; 72% in the small intestine and 72% in the colon), and PGF_2α (56% in the small intestine and 69% in the colon) to comparable extents in the two organs (Fig. 2A and B). The equivalent suppression of prostanoids in the small intestine and colon, despite very different effects on tumorigenesis, suggests that the increased number of small adenomas in the colon may be a result of regulation of other events.

To determine whether sulindac treatment affected cell turnover in the colon, immunohistochemical analyses of PCNA and caspase-3 were done on sulindac-treated and untreated control tissues. As shown in representative images in Fig. 3A (a-d), sulindac treatment did not affect the extent of PCNA staining in either colon polyps or normal colon crypts relative to untreated controls (56.8% versus 65.6%, respectively; P > 0.05). Also shown in Fig. 3A (e and f) is caspase-3 immunostaining in colon polyps from both control and sulindac-treated mice. Sulindac treatment, however, had no effect on the levels of caspase-3 staining (P > 0.05).

Sulindac-induced apoptosis has been shown to be dependent on mitochondrial apoptotic proteins (36). As shown in Fig. 3B, sulindac treatment moderately decreased the Bax (proapoptotic protein)/Bcl-2 (antiapoptotic protein) mRNA expression ratio in colon polyps compared with normal mucosa by 52% (1.19 versus 2.48, respectively; P = 0.1). In comparison, in untreated control tissues, the Bax/Bcl-2 expression ratio was slightly increased (33%) in colon polyps relative to normal mucosa (1.08 versus 0.72, respectively; P > 0.05), indicating a modest decrease in apoptosis within polyps from sulindac-treated mice relative to normal mucosa.

Suppression of PGE₂ production has been associated with reduced accumulation of nuclear β-catenin in colon cancer cells as well as in polyps (35, 37). Thus, to gain additional insight into the increased colon tumor multiplicity found in the colon after sulindac exposure, the cellular localization of β-catenin and E-cadherin was evaluated. As shown in Fig. 4A (a and c) and B, immunohistochemical analysis of E-cadherin showed a dramatic reduction in membrane-associated E-cadherin staining in untreated Apc^Min/+ polyps compared with normal adjacent mucosa (4.0 × 10⁶ versus 6.3 × 10⁶ pixels/frame, respectively; P < 0.01). However, sulindac treatment resulted in a profound protection against this tumor-associated loss of E-cadherin (Fig. 4A, b and d). In fact, as shown in Fig. 4B, the loss of membrane staining of E-cadherin in polyp tissue was completely inhibited by sulindac treatment.

Because E-cadherin and β-catenin exist in a protein complex at the plasma membrane (22), we next wanted to determine whether altered E-cadherin levels might affect β-catenin localization. Although the accumulation of nuclear β-catenin is clearly associated with increased cell proliferation (19), there is evidence that elevated nuclear levels might actually become proapoptotic (38). As expected (Fig. 4A, e and f), intense nuclear staining of β-catenin was found consistently within the polyp tissue of untreated controls. However, sulindac treatment resulted in a complete abolition of nuclear β-catenin within the colon polyps (P < 0.0001, Fig. 4C). In addition, there was no evidence of nuclear β-catenin staining in the normal colon mucosa from either group (data not shown). In the small intestine, sulindac treatment had no effect on the expression of nuclear β-catenin. Strong nuclear staining was observed in both sulindac-treated and control polyp tissues whereas normal small intestinal mucosa did not exhibit nuclear β-catenin expression (data not shown). Taken together, these data suggest that the ability of sulindac to inhibit loss of E-cadherin in colon polyps may contribute to the significant reduction in nuclear β-catenin.

The increase in colon polyp multiplicity occurred despite continuous exposure to sulindac, suggesting that a subset of lesions have developed sulindac resistance. It is possible that the failure to respond to sulindac may be a result of the acquisition of an altered molecular signature within the polyps. p21 has been shown to be upregulated by sulindac treatment, and its expression has been shown to be required for tumor suppression in response to sulindac (25, 26). As shown in Fig. 5A and B, we performed immunoblot and immunohistochemical analyses of p21 in control and sulindac-treated colons. In the absence of sulindac, p21 expression was upregulated in colon polyps relative to normal mucosa (4 of 4 cases). Following sulindac treatment, however, p21 levels were not detectable in the polyps (Fig. 5A and B). However, p21 mRNA analysis did not reveal expression differences between the control and sulindac-treated groups (data not shown). p21 expression in the small intestine, assessed by immunohistochemistry, was found to be highly variable in tumors from both groups (data not shown). In addition, we did not detect any differences in p53 expression between the control and the sulindac-treated groups (data not shown).

It has recently been reported that upregulation of p21, with associated growth suppression, following sulindac treatment may be mediated in part through activation of the nuclear receptor protein PPARγ (30). As shown in Fig. 5A, PPARγ expression was maintained in polyp tissue in the absence of sulindac treatment. However, exposure to sulindac resulted in the growth of polyps that failed to express PPARγ (Fig. 5A), suggesting that sulindac-resistant polyps may also be able to survive because they have lost PPARγ. These data suggest that during tumorigenesis, a subset of cells will lose expression of p21 and PPARγ, possibly facilitating their clonal expansion and contributing to their resistance to sulindac.

Based on our observation that colon polyps that lose p21 can survive in the presence of sulindac, we wanted to further investigate the mechanism for this resistance in vitro using HCT116 WT and p21^−/− cells. As shown in Fig. 6A, sulindac sulfide increased PARP cleavage in HCT116 WT cells in a dose-related manner, consistent with previous studies (39). This response was abrogated in the absence of p21. Apoptosis as assessed by enrichment of histone/DNA cytoplasmic complexes was also significantly inhibited in p21^−/− cells compared with WT cells in response to 90 and 120 μmol/L sulindac sulfide (Fig. 6B). In addition, the percentages of sub-G₀/G₁ phase cells (apoptotic) were reduced in p21^−/− cells compared with WT cells in response to 90 and 120 μmol/L sulindac sulfide (Fig. 6C). These data indicate that the cells that lose p21 do not respond to sulindac sulfide due to inhibition of apoptosis and cell cycle arrest.

Human clinical trails have shown that NSAIDs, including sulindac, which inhibit COX-1 and COX-2, are effective in reducing the recurrence of colorectal adenomas (4, 5). Recent reports, however, suggest that sulindac therapy may be associated with the development of resistance among a subset of patients. This outcome includes the failure of sulindac to promote the regression of adenomas and/or the development of a “breakthrough” carcinoma in patients while on sulindac treatment (4, 9–13). Thus, it becomes of paramount importance to begin to elucidate the mechanisms of sulindac resistance to predict those individuals who will benefit most from NSAID treatment. In addition, gaining a better understanding of underlying molecular mechanisms of this effect may facilitate the development of new chemoprevention strategies. For these reasons, we have tested the efficacy of sulindac in inhibiting tumorigenesis in a widely used mouse model of intestinal cancer, Apc^Min/+, with a focus on colonic rather than small intestinal tumors. We report that sulindac is effective at reducing the size of colon adenomas, indicating inhibition of progression. However, the number of colon adenomas is significantly increased by sulindac exposure, consistent with an earlier study (14). Together, these results suggest that sulindac fails to block early events associated with tumor initiation.

To begin to evaluate potential mechanisms by which sulindac can inhibit colon tumor progression, we measured the concentrations of a panel of prostanoids in the small intestine and colon. Sulindac is a potent inhibitor of both COX-1 and COX-2 and has been shown to effectively suppress PGE₂ production within the intestine (15, 16). As expected, PGE₂ levels were markedly reduced in the small intestine by sulindac, consistent with its tumor-suppressive properties (Fig. 1C). However, in the colon, where tumor numbers were actually increased by sulindac treatment, PGE₂ production was inhibited to a comparable extent as in the small intestine, suggesting that resistance to drug treatment in this organ is not directly related to the degree of PGE₂ suppression. Thus, we began to consider the possibility that regulation of other events might account for the drug-associated increase in colon tumor multiplicity and decrease in size.

Sulindac is known to affect a number of non-COX cellular pathways that may directly affect tumor cell growth (40). For example, it has been shown to disrupt Wnt signaling by inhibiting the nuclear translocation of β-catenin, thus impairing the transcription of T-cell factor/lymphoid enhancer factor target genes (20). We found that sulindac abolished nuclear β-catenin within the colon polyps while increasing its membrane localization. To better define a mechanism for this effect, we examined the levels of E-cadherin expression in the presence and absence of sulindac because E-cadherin can anchor β-catenin at cell-cell junctions (22). Repression of E-cadherin, often occurring during tumor cell invasion, results in nuclear accumulation of β-catenin and activation of target genes (41, 42). In the present study, we found a dramatic reduction in E-cadherin levels in untreated Apc^Min/+ colon polyps, suggesting a high potential for growth and invasion. Interestingly, sulindac completely protected against this loss of E-cadherin. Thus, it seems likely that the sulindac-induced maintenance of E-cadherin in colon polyps is causing an increase in membrane-bound β-catenin, inhibiting its nuclear accumulation and suppressing the growth of the tumors.

Recently, a “just right signaling model” for colorectal cancer has been proposed by several research groups (38, 43). This model suggests that during colon carcinogenesis, APC must acquire specific mutations that will cause sufficient accumulation of nuclear β-catenin to promote transcription of proliferative target genes. However, excessive nuclear β-catenin accumulation has also been shown to promote apoptosis and is therefore unlikely to provide a selective advantage during tumorigenesis (44). It is thought that these elevated levels of nuclear β-catenin will result in a broader change in gene expression, which increases the likelihood of conflicting downstream signals, thus inducing an apoptotic response (43). This model holds true in FAP and sporadic tumors as well as in APC-mutant mouse models, although there is tissue and species specificity (45–47). Therefore, it is possible that the observed sulindac-mediated suppression of nuclear β-catenin and increased membrane localization in Apc^Min/+ colon polyps might have dual effects of inhibiting growth while simultaneously creating a resistance to apoptosis, thus increasing tumor multiplicity (Figs. 1A and 4A). In contrast, nuclear β-catenin expression in the small intestinal tumors was not affected by sulindac treatment (data not shown). In the small intestinal tissue, the level of β-catenin/TCF signaling may be enough to maintain a balance of proliferation and apoptosis. Recently, there has been interest in developing chemoprevention strategies to target Wnt/β-catenin–mediated signaling (48). Our study points to a potential caveat to developing therapies that completely inhibit the nuclear translocation or enhance the proteosomal degradation of β-catenin as a means for tumor prevention. The possibility exists that complete suppression of nuclear β-catenin may ultimately weaken the apoptotic response to sulindac, paradoxically facilitating tumor initiation. The increased formation of colon tumors in sulindac-treated Apc^Min/+ mice may arise from excessive β-catenin inhibition.

Sulindac is known to induce transcription of p21^WAF1/cip1, a key inhibitor of the cell cycle, which has been shown to be required for drug-dependent suppression of tumor growth (25, 26). In accordance with these earlier findings, we also observed that HCT116 p21^−/− cells have reduced levels of apoptosis and cell cycle arrest in comparison with their WT counterparts in response to treatment with sulindac sulfide. In addition, the nuclear receptor PPARγ can be activated by direct sulindac binding and has been implicated in sulindac-induced growth arrest and transcriptional activation of p21 (30). In fact, PPARγ activation by other ligands (e.g., pioglitazone, bezafibrate) has also been shown to cause a significant reduction in adenoma multiplicity in Apc^Min/+ mice (49). Our study found that the majority of remaining colon polyps sustained a loss of p21 and PPARγ expression, suggesting the evolution of an altered molecular signature within nonresponding tumors that have been chronically exposed to sulindac. While the underlying mechanisms that may account for this loss of protein expression remain to be determined, these data suggest that increasing p21 and/or PPARγ levels in tumor cells may provide a pathway for these cells to overcome resistance to sulindac. One limiting factor to these observations was the inability to compare small-size versus large-size tumors in the sulindac-treated group because there were no large-size tumors (>5 mm) present. These data would have been more complete if it had been possible to perform a size-matched comparison of colon tumors between the two groups.

Taken together, our data suggest that colon tumors that evade sulindac suppression possess an altered molecular signature that potentially confers drug resistance. In particular, the loss of p21 and PPARγ expression may provide a fundamental mechanism for incomplete sulindac chemoprevention. In fact, such a mechanism may extend to other chemopreventive agents that induce cell cycle arrest. For example, celecoxib, like sulindac, has been shown to induce transcription of p21 (50). Therefore, loss of p21 and/or PPARγ in tumors may impart chemoresistance to a large number of therapeutic agents. These findings may provide new insights into designing novel clinical therapies targeting multiple molecular pathways to overcome sulindac resistance.

No potential conflicts of interest were disclosed.

Grant Support: NIH grant 5R01CA114635 (D.W. Rosenberg).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.