Suppression of Tumor Formation by a Cyclooxygenase-2 Inhibitor and a Peroxisome Proliferator-Activated Receptor γ Agonist in an In vivo Mouse Model of Spontaneous Breast Cancer

Cyclooxygenase-2 (COX-2) represents one potential chemoprevention target for breast cancer. COX-2 is up-regulated in human breast cancer and is a key enzyme in the production of prostaglandins, which stimulate cell proliferation, inhibit apoptosis, promote metastasis, and promote angiogenesis in mammary tumor cells (3). Selective COX-2 inhibitors, such as celecoxib, inhibit breast cancer cell growth in vitro and in mouse models (4–6). Epidemiologic studies suggest that both nonsteroidal anti-inflammatory drugs and selective COX-2 inhibitors reduce the risk of breast cancer in both primary and secondary prevention settings (7, 8).

Like COX-2, peroxisome proliferator-activated receptor γ (PPARγ) is up-regulated in breast cancer cells (9). PPARγ is a transcription factor that heterodimerizes with retinoid X receptor and binds to specific DNA response elements in gene promoters (10). Targets of PPARγ include genes involved in metabolism and lipid transport, resulting in adipocyte differentiation. To activate transcription, PPARγ must bind an activating ligand (11). Synthetic PPARγ agonists known as thiozolidinediones are currently used to treat type 2 diabetes. N-(9-fluorenyl-methyloxycarbonyl)-l-leucine (F-L-Leu) as a novel PPARγ agonist that lacks the thiozolidinedione ring has potent insulin-sensitizing activity but weaker adipogenic activity (12). Exposure of human breast cancer cell lines to thiozolidinedione has been shown to inhibit cell proliferation in cell culture as well as in nude mice (13). 15-Deoxy-Δ^12,14-prostaglandin J2 (15PGJ₂) is an endogenous ligand of PPARγthat is a downstream product of COX-2 and thus PPARγ activity may be down-regulated in the presence of COX-2 inhibitors.

As described above, there is good evidence that both COX-2 inhibitors and PPARγ agonists have antitumor effects individually. However, because the key endogenous ligand of PPARγ is 15PGJ₂, and this ligand is a downstream product of COX-2, it is unlikely that the antitumorigenic activity of COX-2 inhibitors is being mediated by up-regulation of PPARγ function. In fact, inhibition of COX-2 may actually inhibit the antitumor effects of PPARγ because of reduced 15PGJ₂ production. If this is the case, then addition of a PPARγ agonist to a COX-2 inhibitor may have a cooperative or synergistic effect because both antitumorigenesis pathways would be functional. The central hypothesis for this study is that a combination of both COX-2 inhibitors and PPARγ agonists may be more effective in inhibiting the growth of breast cancer cells than either agent alone.

To examine the effects of celecoxib and F-L-Leu, we have chosen to use the C3 (1)-SV40 tumor antigen (Tag) mouse, originally developed in the laboratory of Dr. Jeffrey Green at the National Cancer Institute (14). This mouse expresses SV40 large Tag under control of the rat prostatic steroid binding protein promoter. Somewhat unexpectedly, it was found that female C3 (1)-SV40 Tag mice consistently develop ductal adenocarcinoma by 6 months of age due to expression of SV40 Tag mRNA in breast tissue (15). Time course studies show that there is a progression observed in the breast tissue of these animals over time, starting with hyperplasia (2 months), progressing to nodular lesions (3 months), and then to adenocarcinomas (4-6 months of age; ref. 16). SV40 large Tag is known to inhibit p53 and RB function, inactivation of which is common in human breast cancer (17, 18). This model has also been used to test several other chemopreventive agents for breast cancer, including dehydroepiandrosterone, the ornithine decarboxylase inhibitor DFMO, and the antiangiogenic compound endostatin (19, 20).

Here, we have examined the effects of the selective COX-2 inhibitor celecoxib and the PPARγ agonist F-L-Leu on tumor formation in C3 (1)-SV40 Tag mice. Our results show that the combination of celecoxib and PPARγ is particularly effective at delaying breast tumor formation in these animals.

MMAC-1 cell line. A mouse mammary adenocarcinoma from a C3 (1)-SV40 Tag mouse was aseptically removed, rinsed with cold PBS, minced with a scalpel, and transferred into 10 mL of 0.02% collagenase and rotated in 37°C water bath for 2 h. The cellular material was then washed four times with serum-free DMEM and the cells were recovered by centrifugation and transferred to 1.6 mL of low calcium medium in a swine skin gelatin–coated T-25 flask to select for epithelial cell growth. The next day, the supernatant (enriched for tumor cells) was transferred to a new flask. Once the cells adhered, 3 mL of additional medium were added.

In vitro drug studies. MMAC-1 cells in logarithmic growth phase were washed once with PBS, detached with trypsin, and suspended to a final concentration of 5 ×10⁴/mL in fresh low calcium medium. Cells (1 ×10⁴) were placed in a 96-well flat-bottomed tissue culture plate. After 24 h, supernatant was discarded and new medium containing drug was added. After 48 h of incubation at 37°C, viable cell growth was determined using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (Promega). Celecoxib was purchased from Toronto Research Chemical, Inc. F-L-Leu was purchased from Sigma-Aldrich.

Transgenic mice and diets. Female C3 (1)-SV40 Tag transgenic mice were weaned at 21 to 30 d and then put on the experimental diet as indicated. Virgin animals were used because of the acceleration of the development of mammary tumors due to pregnancy. Animals were monitored weekly for the appearance of tumor by visual inspection and palpation.

Animals were fed AIN-76A rodent diet as their control chow (Bio-Serv). The drug(s) was added to this chow during the manufacturing process at the indicated concentration (Bio-Serv custom diets S5282, F5152, F5153, S5345, S5346, S5347, and S5348). The chow was sterilized by γ-radiation.

Western blot analysis. MMAC-1 cells were collected, washed with PBS, and resuspended in lysis buffer (Cell Signaling Technology). Samples were then incubated at 4°C for 20 min followed by centrifugation at 13,000 rpm for 10 min at 4°C. The supernatant was then assessed for protein concentration using the Bio-Rad protein assay solution and a Bio-Rad SmartSpec3000 machine (Bio-Rad). Protein (30 μg) was loaded onto a NuPage 10% Tris-acetate gel (Invitrogen) and, after electrophoresis, transferred to polyvinylidene difluoride membrane. Membranes were blocked with 5% skimmed milk-PBS-Tween for 1 h and then probed with the primary antibodies, either anti-vimentin (1: 1,000; Sigma-Aldrich) or anti-keratin (1: 100; Abcam). After washing in PBS-Tween, the membranes were incubated with a donkey anti-rabbit horseradish peroxidase–linked secondary antibody (1: 1,000; Amersham Biosciences) for 1 h. The signal was visualized by chemiluminescence (SuperSignal West Dura Extended Duration Substrate, Pierce Biotechnology).

Reverse transcription-PCR. MMAC-1 cells were grown to 75% confluency in six-well plates, either with no treatment or treatment with celecoxib (50 μmol/L) or F-L-Lou (80 μmol/L). Cells were harvested and RNA was extracted using the Qiagen Total RNA Isolation kit. Reverse transcription-PCR was done using the SuperScript One-Step Reverse Transcription-PCR kit (Invitrogen) according to the manufacturer's instructions. The following primers were used: COX-2, 5′-CTTTGCCCAGCACTTCACCCATC-3′ (forward) and 5′-TCCAAAGGTGCTCGGCTTCCAG-3′ (reverse); PPARγ, 5′-TCTCCGTGATGGAAGACCACTCG-3′ (forward) and 5′-TTCTGGAGCACCTTGGCGAACAGC-3′ (reverse).

Quantitative real-time PCR. Total RNA was extracted using the PicoPure RNA kit (Molecular Devices) according to the manufacturer's instructions. First-strand cDNA was generated from total RNA from each sample using the High Capacity cDNA kit (Applied Biosystems) according to the manufacturer's instructions. COX-2, PPARγ, 18S rRNA, and β-actin was measured using Taqman technology on an ABI Prism 7900 Sequence Detection System. Reactions were prepared in triplicate for each gene using Taqman Gene Expression Master Mix and the following Taqman Gene Expression Assays (Applied Biosystems): COX-2 (Mm00478374_m1), PPARγ (Mm00440945_m1), β-actin (Mm00607939_s1), and 18S rRNA (Hs99999901_s1). Plates were loaded and reactions were cycled using Taqman universal cycling conditions according to the manufacturer's instructions. During thermal cycling, the threshold cycle (C_t) is defined as the cycle number when amplification of a specific PCR product can be detected. An average COX-2 or PPARγ C_t value and an average 18S rRNA C_t or β-actin C_t value were calculated from the three replicate reactions. The average C_t value for the housekeeping gene β-actin or 18S rRNA was subtracted from the average C_t value for each target gene (COX-2 and PPARγ) to normalize the amount of sample RNA added to the reaction. Relative quantification describes the fold change in expression of a gene of interest in a test sample relative to a calibrator sample. With the comparative C_t (ΔΔC_t) method, the following formula was used to determine the level of the target gene mRNA in the drug-treated samples relative to the level found in the untreated sample (Applied Biosystems User Bulletin #2, October 2001): relative quantification = 2^−ΔΔCt, where ΔΔC_t = average ΔC_t (drug treated) − average ΔC_t (untreated).

Immunohistochemistry. Mammary tumor material was fixed in buffered formalin and embedded in paraffin. The paraffin blocks were cut into 5-μm-thick sections that were placed on positively charged slides. The sections were dewaxed in xylene and hydrated through graded ethanol. Heat-induced antigen retrieval was then done in 10 mmol/L sodium citrate (pH 6.0) in a microwave for 10 min. The endogenous peroxidase activity was blocked by immersing the slides in 3% H₂O₂ in PBS for 30 min. After 30 min of incubation with goat blocking serum, slides were incubated with either a 1:500 dilution of COX-2 rabbit polyclonal antibodies (Cayman Chemical) or a 1:50 dilution of PPARγ mouse monoclonal antibodies (Santa Cruz Biotechnology) at 4°C overnight followed by incubation with goat anti-rabbit biotinylated secondary antibody for COX-2 or goat anti-mouse biotinylated secondary antibody for 30 min followed by streptavidin peroxidase for 30 min at room temperature. 3,3′-Diaminobenzidine (Sigma-Aldrich) substrate chromogen was applied for 4 min until the brown color developed. The slides were counterstained with hematoxylin and mounted (Permount). The COX-2 blocking peptide (Cayman Chemical) was used in combination with COX-2 polyclonal antibody in a 2:1 (v/v) ratio in a final concentration of 1:500 in PBS solution. Each slide was treated with 100 μL of the cocktail solution and then developed as indicated with the antibody alone. PPARγ blocking peptide (Santa Cruz Biotechnology) was used in conjunction with PPARγ mouse monoclonal antibodies in a 5:1 (v/v) ratio in a final concentration of 1:50 in PBS solution, and then each slide was treated with 100 μL of the cocktail solution and developed as indicated with the antibody alone.

Survival analysis. Animals were monitored for tumor formation by visual inspection and palpation. When the tumor reached 10% of the animal's body mass, the animal was euthanized and recorded as having died of tumor formation. Animals that died for other reasons were censored. Survival analysis was done using GraphPad Prism 5.0. Statistical testing was done using a log-rank test.

Characterization of C3 (1)-SV40 Tag mice. We obtained a mating pair of C3 (1)-SV40 Tag mice from Dr. Denise Connolly at Fox Chase who had recently imported them from Dr. Green's laboratory for another study. As expected, on aging, we found that ∼100% of the female C3 (1)-SV40 Tag mice developed histologically confirmed mammary carcinomas between 131 and 261 days of age with a median age of onset of 232 days (Fig. 1A). It should be noted that this was somewhat longer than the 140 days initially described in Maroulakou et al. (14) and may reflect differences in the feeding and housing of the mice or perhaps a reduction in the number of integrated transgenes. The tumors that developed were quite aggressive, and usually within 3 weeks after initial detection the animals had to be sacrificed because tumor volume had exceeded the maximum size allowed by the Fox Chase Cancer Center Laboratory Animal Facility (Fig. 1B). Consistent with the findings of Dr. Green and colleagues, we were able to distinguish between early lesions confined to the inner lining of the breast epithelium and more advanced lesions resembling invasive adenocarcinomas (Fig. 1C). We also examined the tumors from C3 (1)-SV40 Tag females for expression of COX-2 and PPARγ using immunohistochemistry (Fig. 2) and found high levels of both COX-2 and PPARγ protein. Based on these findings, we concluded that the C3 (1)-SV40 Tag mouse model is suitable to test the chemopreventive effects of celecoxib and F-L-Leu.

Creation and characterization of MMAC-1 cell line as an epithelial tumor cell line. We created a mouse breast adenocarcinoma cell line by culturing a tumor derived from a C3 (1)-SV40 Tag mouse (Fig. 3A). To confirm that the cells grown out from the tumor were actually tumor cells and not contaminating stromal fibroblasts, we examined the abundance of keratin and vimentin by Western blot (Fig. 3B). Epithelial cells generally express keratin and do not express vimentin. Consistent with this expectation, we found that MMAC-1 cells exhibited abundant keratin expression and undetectable levels of vimentin. This expression pattern was identical to that observed for a human breast cancer cell line (MCF-7) and was opposite of that observed for a mouse fibroblast line and for tissue from the breast region of a healthy female mouse. We also examined COX-2 and PPARγ mRNA expression in the presence and absence of both celecoxib and the PPARγ agonist F-L-Leu (Fig. 3C). From these results, we conclude that the MMAC-1 cell line is epithelial in origin and expresses both COX-2 and PPARγ.

Celecoxib and F-L-Leu are cytotoxic and exhibit cooperativity on MMAC-1 cells. The antiproliferative effect of celecoxib and F-L-Leu was evaluated on MMAC-1 cells in vitro (Fig. 4A). Significant dose-dependent growth inhibition was observed with both drugs. However, celecoxib was significantly more potent than F-L-Leu at inhibiting cell growth. The IC₅₀ for celecoxib was 79 μmol/L compared with 242 μmol/L for F-L-Leu.

We next examined how cells responded to a combination of celecoxib and F-L-Leu (Fig. 4B). In this experiment, we exposed cells to 50 μmol/L celecoxib alone, which resulted in growth inhibition of 9.9%. We also examined F-L-Leu alone at concentrations of 80, 160, and 400 μmol/L and found that MMAC-1 cells are inhibited 16.1%, 36.3%, and 78.6%, respectively. Treating the cells with combinations of 50 μmol/L celecoxib and either 80, 160, or 400 μmol/L of F-L-Leu resulted in 68%, 56.5%, and 90% growth inhibition, respectively. The level of growth inhibition observed, especially at lower concentrations of F-L-Leu, was greater than what would be expected if the drugs were simply working additively. This indicates that celecoxib and F-L-Leu exhibit cooperativity with respect to growth inhibition of MMAC-1 cells.

F-L-Leu and celecoxib inhibit breast tumor formation in C3 (1)-SV40 Tag mice. We next examined the effects of F-L-Leu and celecoxib on breast tumor formation in C3 (1)-SV40 Tag mice. Mice at the time of weaning were given either a control diet or a diet containing celecoxib at a concentration of 1,500, 750, or 375 mg/kg of chow or F-L-Leu at a concentration of 250, 125, or 62.5 mg/kg of chow. Animals were then monitored for up to 400 days for tumors.

We found that all three concentrations of celecoxib tested significantly delayed tumor onset (Fig. 5A; Table 1). For animals on 1,500 mg/kg celecoxib, the median age of tumor appearance was 292 days (versus control, P < 0.0006). Animals on 750 mg/kg celecoxib had a median age of tumor appearance of 307 days (versus control, P < 0.0007). Animals on 375 mg/kg celecoxib had a median age of tumor appearance of 293 days (versus control, P < 0.0002). There was no significant difference in the age of tumor appearance for the different celecoxib concentrations used.

With regard to the PPARγ agonist F-L-Leu, we observed high levels of toxicity with 250 mg/kg doses and this arm was discontinued. We did not observe any statistical difference in the median age of tumor formation with either 125 or 62.5 mg/kg of F-L-Leu compared with controls (Fig. 5B; Table 1). These results suggest that F-L-Leu alone does not inhibit breast tumor formation in C3 (1)-SV40 Tag mice. However, addition of F-L-Leu to celecoxib did significantly increase tumor latency (Fig. 5C; Table 1). A combination of 750 mg/kg celecoxib and 62.5 mg/kg F-L-Leu resulted in an increase in median tumor latency from 307 days (750 mg/kg celecoxib alone) to 393 days (P < 0.007). These results show that a combination of celecoxib and F-L-Leu is more effective than celecoxib alone at inhibiting breast adenocarcinoma development in C3 (1)-SV40 Tag mice.

In conducting these experiments, we noticed that although tumor formation was delayed in celecoxib-treated animals, once tumors became detectable they seemed to grow as rapidly as in the untreated animals. One possible explanation for this phenomenon would be that the tumors in the celecoxib-treated animals might overexpress COX-2, thus making them resistant to celecoxib. Therefore, we examined the relative amount of COX-2 mRNA present in the breast tumor of an untreated, celecoxib-treated, and F-L-Leu–treated mouse. In addition, we examined PPARγ mRNA levels. As shown in Fig. 6, we found that COX-2 mRNA levels were 22-fold elevated in the tumor from a celecoxib-treated mouse compared with the tumor from an untreated mouse. Interestingly, PPARγ levels were 4-fold elevated in the celecoxib-treated animal. We also observed significant elevations in both the COX-2 and PPARγ mRNA in the F-L-Leu–treated animal. These findings are consistent with the notion that resistance to celecoxib may be related to overexpression of the COX-2 protein. In addition, these results suggest that there may be cross-talk between the COX-2 and PPARγ pathways.

In this study, we have examined the effects of the selective COX-2 inhibitor celecoxib and the PPARγ agonist F-L-Leu both singly and in combination for their effects on tumor formation in a mouse model of breast cancer. We found that celecoxib, but not F-L-Leu, significantly delayed tumor development. However, the strongest inhibition occurred when both agents were combined. Our results support the idea that targeting COX-2 and PPARγ in combination may be a potential chemopreventive strategy for human breast cancer.

Our results are consistent with other published studies that show that the use of selective COX-2 inhibitors can slow breast tumor formation in mice. Lanza-Jacoby et al. (6) showed that celecoxib in mouse diet (900 mg/kg) could inhibit breast tumor formation in HER2/neu mice, increasing the mean age of incidence from 266 to 291 days. We observed a somewhat larger effect in our C3 (1)-SV40 Tag mouse, with the mean age of incidence increasing from 232 to 307 days in mice fed a diet with 750 mg/kg celecoxib. We did not see a statistically significant difference in the survival curves for the different doses of celecoxib, suggesting that lower doses may be equally effective.

We did not observe inhibition of tumor formation in mice given F-L-Leu alone. We found that high doses of F-L-Leu (250 mg/kg) were associated with significant toxicity as indicated by premature death. Lower doses of F-L-Leu (125 and 62.5 mg/kg) were well tolerated but failed to result in increased tumor latency. To the best of our knowledge, this is the first study to examine the effect of a PPARγ agonist on a mouse model of breast cancer. However, a recent study did find that the PPARγ agonist ciglitizone given i.p. prolonged survival of nude mice injected with human OVCAR-3 ovarian carcinoma cells (21).

The most effective tumor inhibition was observed in mice fed a diet with both F-L-Leu and celecoxib. Mice given the combination diet lived on average 22% longer than mice on celecoxib alone. Our in vivo findings are consistent with our in vitro results in which we found that F-L-Leu and celecoxib acted cooperatively to inhibit tumor cell growth. A possible mechanism by which PPARγ agonists affect breast tumor growth may involve PTEN. PTEN is a tumor suppressor gene that is inactivated in a variety of tumors, including breast tumors. Overexpression of PTEN leads to both cell cycle arrest and increased apoptosis (22). It has previously been shown that the PTEN promoter contains a PPARγ binding site and that exposure of MCF-7 cells to the PPARγ agonists lovastatin and rosiglitazone can induce expression of the tumor suppressor gene PTEN (22). It is interesting to note that we did not observe tumor growth effects of F-L-Leu alone in vivo. This may be because it is not possible to achieve high enough levels of F-L-Leu in the blood. Alternatively, it may be that cells grown in culture are more sensitive to PTEN activation than cells inside a tissue.

Clinically, several prospective epidemiologic studies have shown that use of nonsteroidal anti-inflammatory drugs, inhibitors of COX-1 and COX-2, is associated with a lower risk of breast cancer, but other studies have not (8). With regard to selective COX-2 inhibitors, a recent case-control study indicated that women taking these compounds daily for 2 years or more had a 3-fold reduced risk of breast cancer (7). Less is known about the clinical effects of PPARγ agonists in humans, although a pilot study of rosiglitazone therapy in women with breast cancer showed that it had no effect on Ki67 expression in tumor tissue (23). However, the same study did find that the drug was well tolerated without serious adverse events. The work presented here supports the idea that a combination of COX-2 inhibitors and PPARγ agonists together in subclinical doses may be useful for chemoprevention in women at high risk of developing breast cancer in either primary or secondary settings. Future clinical studies will determine if this concept is feasible.

No potential conflicts of interest were disclosed.

We thank Drs. Denise Connolly and Jeffrey Green for the C3 (1)-SV40 Tag mice used in the study, the facilities at Fox Chase Cancer Center (Laboratory Animal Facility, Experimental Histopathology Facility, Tissue Culture Facility, and Genomics Facility), and Drs. Margie Clapper and Mary Daly for critical reading of this manuscript.