ANTICANCER RESEARCH 25: 225-234 (2005)

Growth Inhibitory Effect of Celecoxib and Rofecoxib on Human Colorectal Carcinoma Cell Lines

BRUNO BUECHER1*, DELPHINE BOUANCHEAU1*, ALEXIS BROQUET1, STÉPHANE BEZIEAU3, MARC G. DENIS1, CHRISTIAN BONNET2, MARIE-FRANÇOISE HEYMANN1, ANNE JARRY1, JEAN-PAUL GALMICHE1 and HERVÉ M. BLOTTIÈRE1,2

Centre de Recherche en Nutrition Humaine de Nantes, 1INSERM U539 and 2INRA-UFDNH, CHU Hôtel-Dieu, 44035 Nantes; 3LEPA, Laboratoire d’Etude du Polymorphisme de l’ADN, Faculté de Médecine, 44035 Nantes, France

Abstract. Background: Epidemiological studies have revealed shown to play a central role in several physiological processes a protective effect of NSAIDs, which principally target including platelet aggregation and parturition (3). COX-2 is a cyclooxygenase (COX)-1 and COX-2, on the development of highly inducible gene, that is activated by various stimuli colorectal cancer. Increased expression of COX-2 was shown including pro-inflammatory cytokines and growth factors (4). in colorectal adenocarcinoma. However, some effects were In 1994, Eberhart et al. were the first to describe a shown to be COX-independent. Here, we compared two significant elevation of COX-2 expression in most human selective COX-2 inhibitors for their effect on the growth of colorectal carcinomas, but also in approximately 50% of colorectal tumour cells in vitro. Materials and Methods: Fifteen adenomas (5). Such findings were later documented by tumour cell lines were characterized for COX-1 and COX-2 others (6, 7). Moreover, a similar elevated expression has expression by Western blot and RT-PCR. The effect of been reported in azoxymethane-induced colonic tumours in celecoxib and rofecoxib on their growth was assessed by rats (8) and in APC mutant mice (9). It is associated with staining of DNA with crystal violet. Results: COX-2 expression increased PG levels in the tissues, especially PGE2 (10). varied among cell lines, whereas COX-1 was always expressed. The fact that such elevated expression was found in human Rofecoxib displayed a limited dose-related effect on cell premalignant adenomas and also in polyps in APC mutant proliferation, whereas celecoxib strongly inhibited cell growth mice argues for an implification as an early event during the at high concentrations. Both effects appeared COX-2- course of carcinogenesis. Indeed, Tsujii and DuBois independent. Conclusion: Rofecoxib, which is devoid of demonstrated that transfecting rat intestinal epithelial cells apoptotic effect at high concentration but efficient at with a COX-2 expression vector, leading to its constitutive pharmacological concentrations, revealed a potential new expression, provokes phenotypic alterations such as mechanism of action of NSAIDs towards colorectal cancer. modulation of their adhesion properties and inhibition of induced apoptosis (11). These changes enhanced their Cyclooxygenases (COX) are the key enzymes in the tumorigenic potential and were reversed by sulindac conversion of arachidonic acid to prostaglandins (PG) and sulphide. Moreover, Liu et al., using transgenic mice in other eicosanoids (1). Two genes have been identified, which COX-2 expression was under a promoter targeting although at the protein level several isoforms have been selectively the mammary glands, showed that COX-2 described resulting from alternative splicing (2). COX-1 has overexpression alone was sufficient to induce mammary been shown to be constitutively expressed in a variety of carcinomas (12). tissues including gastro-intestinal epithelium and has been Therefore, the targeting of COX-2 with more or less selective inhibitors was proposed as a new therapeutic approach for colorectal cancer. These inhibitors were shown to display chemopreventive effects in different models of *Both authors contributed equally to this work. colorectal carcinogenesis (for review see 13). Moreover, epidemiological studies revealed that regular consumption Correspondence to: H.M. Blottière, Ph.D., UNSA-INRA, Domaine de Vilvert, 78352 Jouy-en Josas cedex, France. Tel: 33 (0)1 34 65 23 of aspirin or other non-steroid anti-inflammatory drugs 19, Fax: 33 (0) 1 34 65 23 11, e-mail: [email protected] (NSAIDs) reduces the risk of colorectal cancer (14-16). Finally, it was demonstrated that sulindac administration, Key Words: COX-2, COX-1, colorectal cancer, NSAIDs, proliferation. and more recently celecoxib, to patients with familial

0250-7005/2005 $2.00+.40 225 ANTICANCER RESEARCH 25: 225-234 (2005) adenomatous polyposis resulted in a reduction in the horseradish peroxidase-linked polyclonal antibody to rabbit Ig number and size of colorectal polyps (17, 18). Although (Jackson Labs, West-Grove, PA, USA). Specific bands were sulindac was inefficient as a primary chemoprevention agent revealed using the ECLì detection system, according to the manufacturers’ instruction (Amersham Pharmacia biotech, Les (19), it was found efficient in the retained rectal segment of Ullis, France). FAP patients after prophylactic colectomy (20). All these findings suggest that COX-2 plays an important Cell proliferation assay. Cells were seeded in 24-well plates at 1x104 part in colorectal carcinogenesis and that NSAIDs, especially cells/well, allowed to grow for a day, then exposed in quadruplicate COX-2 selective inhibitors, have potential chemopreventive wells to increasing concentrations of celecoxib or rofecoxib. After 6 properties in human colorectal cancer. However, the exact days in culture with the drugs, the cell number was estimated using mechanisms by which NSAIDs inhibit colon carcinogenesis the colorimetric crystal violet assay, as described previously (28). are still elusive, and some may be independent of COX-2 2 expression (21). In prostate cancer cells, Song et al. RT-PCR analysis. Total RNA from cells plated onto 25-cm flasks was isolated using TRIzol Reagentì (Invitrogen). cDNA was questioned whether COX-2 was a "player" or a "spectator" in generated on 1 Ìg of total RNA in a reaction volume of 20 Ìl, using COX-2 inhibitor-induced apoptosis (22). Indeed, COX-2 Superscript II M-MLV reverse transcriptase (RT; Invitrogen). PCR inhibitors have been shown to inhibit cell proliferation, was done in the linear range of amplification (determined for each induce apoptosis, inhibit angiogenesis and stimulate the primer pair-cDNA combination). Standard PCR reactions were immune system (23-26). Although reduction in performed with 1 Ìl of the cDNA solution, 1 ÌM of each primer prostaglandins synthesis may account for these activities, solution, 200 ÌM of each dNTP, 1.5 mM MgCl2, Goldstar DNA other mechanisms unrelated to COX-2 may also be involved. polymerase reaction buffer and 0.5 units of Goldstar DNA polymerase (Eurogentec, Seraing, Belgium). Each PCR cycle The aim of our study was to compare the effect of two consisted of 1 min at 92ÆC, 1 min at 58ÆC and 1 min at 72ÆC. selective COX-2 inhibitors, namely celecoxib (Celebrexì, Primers were: Pharmacia) and rofecoxib (Vioxxì, Merck), on colorectal ‚-actin Forward 5'-GGCATCGTGATGGACTCCG-3' adenocarcinoma cell growth. These cells were also Reverse 5'GCTGGAAGGTGGACAGCGA-3'. characterized for COX-1 and COX-2 expression and some COX-1 Forward 5'-GTTCAACACCTCCATGTTGGTGGAC-3' other features of colonic tumours. Reverse 5'-TGGTGTTGAGGCAGACCAGCTTC -3'; COX-2 Forward 5’- TTCAAATGAGATTGTGGGAAAAT -3’ Materials and Methods Reverse 5’-AGATCATCTCTGCCTGAGTATCTT-3’.

Cell lines and reagents. Caco-2, Colo-205, HCT-116, HCT-EB, HCT- Real-time quantitative RT-PCR. The amplification conditions of GEO, HRT-18, HT-29, LS174T, LS180, SW48, SW480, SW620, COX-2, ‚-actin and glyceraldehyde-3 phosphate dehydrogenase SW948 and SW1116 human colorectal adenocarcinoma cell lines (GAPDH) templates were optimised for the RotorGene 2000 were characterised previously (27). Cells were cultured in RPMI instrument (Ozyme, Saint Quentin en Yvelines, France). PCR 1640 supplemented with 10% foetal calf serum (FCS), 2 mM L- amplifications were performed using Titanium Taq DNA glutamine, 50 U/ml penicillin and 50 mg/ml streptomycin, except for polymerase (Clontech, Palo Alto, CA, USA). The reaction HCT-EB and HCT-GEO, which were cultured in DMEM containing mixture contained 2 Ìl of the supplied 10X Titanium Taq PCR 25 mM glucose instead of RPMI. Caco-2 cells were also cultured in buffer (containing magnesium chloride), 2 Ìl of a 1/1,000 dilution DMEM supplemented with 20% FCS. All tissue culture reagents of SYBR Green I (Roche Molecular Biochemicals, Meylan, were from Invitrogen (Cergy Pontoise, France), except FCS which France), 1 Ìl of each primer (0.4 ÌM each), 0.4 Ìl of Titanium was provided by Sigma (L’isle Dabeau, France). Celecoxib was from Taq DNA polymerase, 0.5 Ìl of dNTPs (10 mM each) and PCR- Pharmacia (St Quentin en Yvelines, France) and rofecoxib from grade water to a volume of 15 Ìl. Microtubes (0.2 ml) were Merck (Rahway, NJ, USA). The drugs were dissolved in loaded with 15 Ìl of this master mix and 5 Ìl of the template dimethylsulfoxide (DMSO) to obtain a stock solution of 20 mM. (cDNA diluted 1/100) and the run was initiated. The cycling conditions were as follows: denaturation for 5 min at 95ÆC, Western blot analysis. Cells were plated onto 25-cm2 flasks at a amplification for 45 cycles, with denaturation for 5 sec at 95ÆC, density of 2x106 cells per flask and allowed to grow for 48 h. After annealing for 15 sec at 66ÆC and extension for 20 sec at 72ÆC. To washes with phosphate-buffered saline (PBS), proteins were exclude primer-dimer artefacts, fluorescence was not measured at extracted at 4ÆC for 1h with 500 Ìl of lysis buffer containing 10 mM the end of the extension step, but at a temperature (83ÆC) above Tris-HCl, (pH 7.4), 20 mM NaCl, 5 mM MgCl2, 0.5% Nonidet P- the melting point of primer-dimers and below the melting point 40, 1 mM phenylmethanesulphonyl fluoride, 2.1 ÌM leupeptin, 1.4 of the specific PCR product (90ÆC). ÌM pepstatin (Sigma). Homogenates were centrifuged at 10,000 x Primers were chosen on separate exons to amplify cDNA but g to remove cellular debris, and the protein concentration in the not genomic DNA. The following primers were used: supernatant was determined using the DC protein assay (Biorad, COX-2 Forward 5’-GCCCTTCCTCCTGTGCC -3’ Marne-la-coquette, France). Samples (30 Ìg total protein) were Reverse 5’-AATCAGGAAGCTGCTTTTTAC-3’ separated in a denaturing polyacrylamide gel (7.5%) and ‚-actin Forward 5’-GGAGCAATGATCTTGATCTTC-3’ transferred to a nitrocellulose membrane. COX-1 and COX-2 Reverse 5’-CCTTCCTGGGCATGGAGTCCTG-3’ proteins were detected using rabbit polyclonal antibodies (Santa GAPDH Forward 5’-TGAACGGGAAGCTCACTGG-3’ Cruz Laboratories, Santa Crux, CA, USA), followed by a Reverse 5’-TCCACCACCCTGTTGCTGTA-3’

226 Buecher et al: Rofecoxib, Celecoxib and Colorectal Tumour Cell Proliferation

Figure 1. Expression of COX-1 in colorectal adenocarcinoma cell lines. A. Western blot analysis. B. RT-PCR analysis. ‚-actin was used as a normalisation gene. Results are representative of at least three experiments. Lane 1, SW1116; 2, HT-29; 3, SW707; 4, LS180; 5, LS174T; 6, HCT116; 7, SW48; 8, SW480; 9, SW620; 10, SW948; 11, HRT18; 12, HCT-EB; 13, HCT-GEO; 14, Colo205; 15, Caco-2; 16, negative control for RT-PCR.

Figure 2. Expression of COX-2 in colorectal adenocarcinoma cell lines. A. Western blot analysis. B. RT-PCR analysis. ‚-actin was used as a normalisation gene. Results are representative of at least three experiments. Lane 1, SW1116; 2, HT-29; 3, SW707; 4, LS180; 5, LS174T; 6, HCT116; 7, SW48; 8, SW480; 9, SW620; 10, SW948; 11, HRT18; 12, HCT-EB; 13, HCT-GEO; 14, Colo205; 15, Caco-2; 16, negative control for RT-PCR.

An external standard curve was generated with serial 5-fold After completion of the cycling process, samples were subjected dilutions (1/40, 1/200 and 1/1000) of a control sample (cDNA from to a temperature ramp from 60ÆC to 99ÆC, with continuous SW1116 cells). The reference curve was constructed by plotting the fluorescence monitoring for melting curve analysis. For each relative amounts of these dilutions vs the corresponding Ct amplification, apart from primer-dimers, a single narrow peak was (threshold cycle) values. The correlation coefficient of these curves obtained at the expected melting temperature (86ÆC), indicating was always greater than 0.99. The amount of COX-2, ‚-actin or specific amplification without significant by-products. GAPDH transcripts was calculated from these standard curves using the RotorGene software. Samples were tested in triplicate Mononucleotide repeat microsatellites analysis. Colorectal cancer cell and the average values were used for quantification. For each lines were suspended in 20 mM Tris-HCl pH 7.4, 5 mM EDTA, sample, the ratio between the relative amount of COX-2 and the 0.4% SDS, 1 mg/ml proteinase K, at 37ÆC overnight for DNA mean between ‚-actin and GAPDH was calculated to compensate extraction. Genomic DNA was precipitated by ammonium acetate for variations in quantity or quality of starting mRNA as well as for 3M final and ethanol. Finally, DNA was rinsed in 70% ethanol and differences in reverse transcriptase efficiency. resuspended in TE buffer (20 mM Tris-HCl pH 7.4, 1mM EDTA).

227 ANTICANCER RESEARCH 25: 225-234 (2005)

Table I. Characterisation of the different human colorectal Table II. Effect of celecoxib and rofecoxib on the proliferation of human adenocarcinoma cell lines used. Expression of COX-2 mRNA was colorectal adenocarcinoma cell lines. Results are expressed as the percentage assessed by real time RT-PCR. Results are expressed as the ratio between of cell remaining in the presence of 10 ÌM and 50 ÌM as compared to the relative amount of COX-2 mRNA and the average of the relative control. The amount of cells from triplicate wells was determined using amounts of two normalisation genes ‚-actin and GAPDH. Three to five crystal violet staining. SEM did not exceed 10%. determinations performed in triplicate were done. Replication error (RER) status was determined by analysis of BAT-26 and BAT-25 microsatellites, Celecoxib Rofecoxib as described in the Materials and Methods section. Differentiation 10 ÌM 50 ÌM 10 ÌM 50 ÌM capacities of the cell lines were described in detail by Blottiere et al. (27). It was based on cell polarity, CEA and brush border hydrolase expression, Caco-2 69.3 3.9 38.1 11.8 and histopathological features of tumours after graft into Nude mice. Colo-205 60.7 0.0 99.9 13.5 (+++): well-differentiated; (++): moderately-differentiated; (+): poorly- HCT-116 72.7 6.3 74.4 57.4 differentiated. n.c.: not characterised. HCT-EB 11.1 1.7 23.9 21.4 HCT-GEO 9.3 3.3 26.0 24.4 COX-2 mRNA RER status Differentiation HRT-18 58.7 5.8 76.6 49.0 Caco-2 125.8 - +++ HT-29 60.6 6.0 70.0 43.7 Colo-205 0 - ++ LS174T 76.3 3.4 51.5 34.9 HCT-116 30.3 + n.c. LS180 57.7 1.9 58.7 27.9 HCT-EB 12.7 - +++ SW48 52.4 5.6 84.4 44.3 HCT-GEO 0 - +++ SW480 67.0 9.0 79.5 37.9 HRT-18 0 - ++ SW620 79.2 16.0 79.0 70.6 HT-29 129.9 - ++ SW707 16.3 0.4 14.0 13.6 LS174T 24.7 + +/++ SW948 40.1 5.5 53.9 35.8 LS180 24.7 + +/++ SW1116 60.4 14.3 39.9 52.3 SW48 154.8 - + SW480 0 - + SW620 0 - + SW707 7.6 + + SW948 0 - ++ SW1116 1022.6 - +++ according to the manufacturer’s instructions. Cell DNA content was then analysed by flow cytometry using an EPICS XL (Beckman-Coulter). Raw data for the distribution of DNA content retrieved from the EPICS XL were treated using Multicycle AV software (Phoenix Flow Software, San Diego, CA, USA) to Primers used to amplify BAT-25 and BAT-26 were as described generate DNA content histograms, and curve fitting analysis was (29). One of each pair of PCR amplification primers is performed to obtain the percentage of G0/G1 through G2/M fluorescently-labelled with either FAM (BAT25, blue), or TET populations. Sub-diploid cells were considered as apoptotic cells. (BAT26, green). The amplification for the two mononucleotide repeats was performed in the same PCR reaction mixture Results containing 10 mM Tris-HCl, 1.5 mM MgCl2, 0.01% (w/v) gelatin, 50 mM KCl, 200 mM of each deoxynucleoside triphosphate (Pharmacia, Uppsala, Sweden), 10 pmol of each primer, 1.25U of Expression of cyclooxygenase-1 protein and mRNA. A great AmpliTaq Gold (Applied Biosystems, Foster City, USA) and 10 ng discrepancy exists in the literature considering the of DNA in a total volume of 20 Ìl. PCR was performed for 35 expression of COX for each particular cell line, thus we first cycles of 0.5 min at 94ÆC, 0.5 min at 55 and 3 min at 70ÆC. A step characterised these cells for both COX isoforms. of 15 min at 95ÆC was performed before the cycles. The fluorescent Considering COX-1 expression, all cell lines exhibited PCR products were electrophoresed on an ABI PRISMì 310 COX-1 mRNA expression as assessed by semi-quantitative Genetic Analyser (Applied Biosystems), and results analysis was RT-PCR (Figure1). At the protein level, we also detected performed with Genotyper 2.5.2 software (Applied Biosystems). All replication error negative (RER-) cell lines showed only minor COX-1 in all cell lines, albeit that SW480 presented lower size variation for BAT 26 and BAT 25 alleles not exceeding 2 bp protein level than the others. from one cell line to another, as described previously (30). RER+ cell lines showed BAT26 and/or BAT25 alleles shortening from -4 Expression of cyclooxygenase-2 protein and mRNA. As bp to -15 bp compared to RER- cell lines. opposed to COX-1 expression, COX-2 expression varied among cell lines (Figure 2A). At the protein level, a very Cell cycle analysis. Cells in exponential growth phase were exposed strong expression was observed in SW1116 and in Caco-2 for 24 h to rofecoxib and celecoxib at increasing concentrations or to DMSO as control. The cells were then harvested using a cells; a lower expression was found in the HT-29 and in trypsin/EDTA solution and stained with propidium iodide using the SW48 cell lines. A barely detectable COX-2 signal was also DNA-prep Coulter kit (Beckman-Coulter, Villepinte, France), obtained in the LS174T and LS180 lines. We failed to

228 Buecher et al: Rofecoxib, Celecoxib and Colorectal Tumour Cell Proliferation

Figure 3. Effect of celecoxib and rofecoxib on growth of HT-29 (triangle), SW1116 (lozenge), HCT-116 (square) and SW480 (cross) cells. Cells in 24- well plates were exposed to the drugs for six days. Cell number was determined by DNA staining using crystal violet. The results are given as % ± SEM of values obtained with controls.

detect COX-2 protein in the other cell lines tested. The Table III. Effect of celecoxib and rofecoxib on cell cycle distribution on results of semi-quantitative RT-PCR analysis (Figure 2B) HCT116 cells after 24 hours. Results representative of one experiment of corroborate the Western blot analysis. In order to the three performed in duplicate are expressed as mean % of the cells in the different phases of the cell cycle. 20,000 cells were analyzed. accurately quantify COX-2 transcripts, real-time RT-PCR were also performed using another pair of primers. The Sub-diploid G1 S G2/M level of COX-2 mRNA was reported to two normalisation genes, ‚-actin and GAPDH. The high expression in Control 2.0 31.1 52.4 16.5 SW1116 was confirmed with almost 10 times more mRNA Celecoxib 10 ÌM 2.2 35.9 50.2 13.9 than in HT-29 or Caco-2 (Table I). Moreover, we were also 20 ÌM 1.4 37.6 47.8 14.6 able to detect COX-2 mRNA in three other cell lines, 50 ÌM 7.1 55.4 34.8 9.8 HCT-116, HCT-EB and SW707. 100 ÌM 12.6 34.2 25.4 40.4 Rofecoxib Relationship between COX-2 expression and the phenotype of 10 ÌM 2.6 31.6 50.2 18.2 20 ÌM 2.0 31.1 49.2 19.7 cell lines. We attempted to test whether COX-2 expression 50 ÌM 2.2 34.6 47.4 18.0 was linked to one of the two pathways of genomic 100 ÌM 2.3 35.0 44.9 20.1 instability. Thus, we characterised the genetic alteration of these cell lines by studying BAT-25 and BAT-26 microsatellite loci (Table I). Microsatellite stability (MSS, also termed RER-) corresponded to cell lines in which both alleles did not exceed 2bp from one cell line to another. was detected, COX-2 mRNA were detected in the 4 RER+ Microsatellite instability (MSI, also termed RER+) cell lines. Therefore, no correlation was noticed between displayed alleles shortening from -4 to -15 bp as compared microsatellite status and COX-2 expression (p>0.05). to MSS cell lines. Most of the cell lines tested were RER- We then addressed the question of whether there was a (also termed MSS, microsatellite stability), especially the link between the differentiation capacities of the cells and highly COX-2-expressing cells SW1116 and Caco-2. Four COX-2 expression. The differentiation status of the cell lines cell lines were found unstable; HCT-116, LS-174T, LS180 was described in detail previously (27) and summarised in and SW707. Surprisingly, and as opposed to what has been Table I. It was based on several criteria, including polarity, published elsewhere (30), SW48, which expressed COX-2, CEA and brush border hydrolase expression, and was observed stable. Although, low or no COX-2 protein histopathological features of tumors after graft into Nude

229 ANTICANCER RESEARCH 25: 225-234 (2005) mice. Among the 4 well-differentiated cell lines, 2 expressed cell cycle. At 100 ÌM, the growth inhibition was maximal, highly COX-2 protein, and 2 were negative. No correlation apoptosis became very important and G1 arrest was not was found between the differentiation capacities of a cell line observed. With rofecoxib, we did not observed any and the expression of COX-2 (p>0.05). Moreover, we studied induction of apoptosis and a mild G1 accumulation was the expression of COX-2 in SW1116 cells at different noticed. Similar results were obtained with SW-1116 and differentiation stage, i.e. below confluence (exponentially HT-29 cells (not shown). growing undifferentiated cells), just confluent, and 1 and 2 weeks after confluence (the later being well-differentiated, Discussion presenting domes and highly expressing alkaline phosphatase and dipeptidyl aminopeptidase IV). Using SW1116, we In the present study, we compared the effect of two observed that the differentiation stage did not influence selective COX-2 inhibitors, celecoxib and rofecoxib, on the COX-2 expression (data not shown). growth of a panel of human colorectal cancer cell lines. We characterised the expression of both COX isoforms in all Sensitivity to celecoxib- and rofecoxib-induced inhibition of cell these lines by Western blot and RT-PCR. We clearly found proliferation. We examined the effect of two selective COX- that, at high concentration, celecoxib was the most efficient 2 inhibitors on the proliferation of the different colorectal drug in inhibiting cell proliferation, whereas at lower tumour cell lines in vitro. Cell number was evaluated by concentration, both molecules were almost equally efficient. crystal violet staining of DNA. The effect of increasing Thus, the two compounds present different mechanisms of concentrations of the drugs was measured after 6 days. As action. We also observed that the inhibitory effect appears shown in Figure 3, the two compounds inhibited cell growth. unrelated to COX-2 status. The effect of celecoxib and rofecoxib appeared independent First, we performed a full characterisation of COX of COX-2 expression, as illustrated for the 4 cell lines: isoforms expression in all the cell lines. This is an important SW1116, which highly expresses COX-2, HT-29 which also issue since, depending on the studies, COX status is still expresses COX-2 but at a lower level, HCT-116 which controversial. For instance, HCT-116 was described to expresses COX-2 mRNA but not the protein, and SW480 express none of the COX (31) or to be COX-2-negative and which was completely negative. The percentage of inhibition express low COX-1 (32). However, as observed in our study, found for two concentrations of celecoxib and rofecoxib, 10 others found COX-1 protein and mRNA in HCT-116 (4). and 50 ÌM, are reported in Table II. Clearly, there was no Considering COX-2, HCT-116 was found COX-2-negative correlation between this parameter and COX-2 expression. both at the protein and mRNA level (4). In our study, we Interestingly, the anti-proliferative response differed found a low COX-2 mRNA expression by real-time RT-PCR, between the two compounds. Indeed, we observed that, but not by semi-quantitative RT-PCR. A similar low COX-2 when increasing the concentration of celecoxib from 10 ÌM mRNA expression was also documented in HCT-116 (33). to 50 or 100 ÌM, the percentage of remaining cells fell Differences in technical approaches may explain these almost completely, whereas for rofecoxib a plateau was discrepancies, since the antibodies used and the Western observed. Due to this phenomenon, we were unable to conditions varied from one laboratory to another. Moreover, calculate the IC50 for rofecoxib and therefore presented the at the mRNA level, the analysis conditions, Northern blot or values obtained with two concentrations of both inhibitors. RT-PCR and, in the latter, the number of cycles performed For example, on SW1116 cells, at 10 ÌM, rofecoxib inhibited are important. In that matter, the use of real-time RT-PCR cell growth by almost 60%, however at 50 ÌM, the inhibition allowed a gain in quantitative analysis. The finding that COX- was less than 50%. With celecoxib, at 10 ÌM a 40% 1 was expressed in all the cell lines analysed is important to inhibition was noticed, but it reached 85% at 50 ÌM. consider in order to analyse the mechanism of action of At 50 ÌM or above, celecoxib was more effective than NSAIDs. Moreover, no study reported any reduced or rofecoxib on all cell lines tested. However, at 10 ÌM suppressed COX-1 expression in colorectal tumours in situ. rofecoxib was more efficient than celecoxib on several cell We then addressed the question of the relationship lines e.g. Caco-2, LS174T and SW1116. Others cell lines, between COX-2 expression and the phenotype of the cells. including LS180, HCT-116, SW707 and SW620, were First, we found no association between COX-2 expression identically sensitive to both drugs. Again, this was unrelated and the differentiation process. As quoted by Shao et al. (8), to COX-2 expression (p>0.05). the well-differentiated cells including Caco-2, HCA-7 and The effect of celecoxib and rofecoxib on cell cycle LS174T expressed COX-2, whereas poorly-differentiated distribution and on apoptosis are presented in Table III. cells such as DLD-1 were negative. In our study, the well- We observed that celecoxib dose-dependently induced differentiated cells Caco-2 and SW1116 strongly expressed apoptosis as revealed by the presence of sub-diploid cells. COX-2, however other well differentiated cells like HCT- This was associated with an arrest in the G1-phase of the EB and HCT-GEO were negative (very low COX-2 mRNA

230 Buecher et al: Rofecoxib, Celecoxib and Colorectal Tumour Cell Proliferation was detected by real-time RT-PCR). On the contrary, the At high concentrations (50 ÌM and above), celecoxib was poorly-differentiated cells SW48 were found COX-2- shown to induce apoptosis, whereas rofecoxib did not (39, positive. In human, it has been shown that moderately- to 40). On HCT-116, SW1116 and HT-29 cells, we also well-differentiated carcinomas showed significantly higher observed that celecoxib used at high concentration rapidly COX-2 immunoreactivity than poorly-differentiated induced apoptosis, while rofecoxib was without apoptotic tumours (34). However, it was also reported that the effect. A recent work, performed in prostate cancer cells, transfection of COX-2 RNA antisense in rat intestinal demonstrated that celecoxib-induced apoptosis was epithelial (RIE) cells resulted in loss of COX-2 expression independent of COX-2 (22), as suggested in our study. This and increased differentiation, as measured through alkaline pro-apoptotic effect is now well characterised and involves phosphatase activity. On the contrary, RIE cells transfected the AKT/PKB pathway through PDK1 inhibition (41). with the sense COX-2 expression vector showed high COX- At low concentrations, the two drugs inhibited cell 2 expression and low alkaline phosphatase activity (11). proliferation, again by a mechanism independent of COX-2 Finally, inhibition of COX-2 by sulindac also resulted in inhibition, the effect being observed whatever the COX-2 increased alkaline phosphatase activity. In our hands, high status was. It is questionable whether COX-1 is involved in expression of COX-2 in SW1116 cells did not interfere with such an inhibitory effect. First, COX-1 was expressed in all its differentiation, and two weeks confluent well- the cell lines tested. Moreover, a COX-1 selective inhibitor differentiated SW1116 cells expressed COX-2 to a similar showed growth inhibition capacities, and was almost devoid level as non-confluent undifferentiated SW1116 cells. of apoptotic effect on colorectal cell lines (42). The potential We further investigated the microsatellite status of the role of COX-1 in neoplasia was also revealed in a murine cell lines by studying BAT-25 and BAT-26 alleles since, in model of intestinal tumorigenesis, where genetic disruption human, reduced COX-2 protein was described in colorectal of COX-1 led to a reduction of intestinal tumorigenesis carcinoma with defective mismatch repair system (7). Alone, similarly to what was found for COX-2 (43). However, BAT-26 was described as a good indicator of replication rofecoxib and celecoxib, used at low concentrations, error phenotype in colorectal cancer cells (30), therefore, displayed a similar effect on cell proliferation, but were the association of the two alleles allowed a reliable shown to very differently inhibit COX-1 activity. Indeed, the phenotypic characterisation, although for clinical studies the IC50 towards COX-1 was 1.2 ÌM for celecoxib and 63 ÌM for actual guidelines recommend the analysis of 5 markers (35). rofecoxib (36). Thus, it seems unlikely that COX-1 is Only 4 cell lines were found to be RER+. Among these 4 involved in such an effect and the precise mechanism of lines, none expressed a high level of COX-2 protein, action remains to be characterised. however, the small number analysed did not allow a highly Understanding of this mechanism of action is of reliable correlation analysis although, based on COX-2 importance because, in vivo, both in human and in murine mRNA level, no correlation was observed (p>0.05). It is models of carcinogenesis, the apoptotic effect, which was noteworthy that SW48 cells, which expressed COX-2 observed at high concentration (above 50 ÌM), may not protein, were found to present microsatellite stability in our play a major role in celecoxib and other NSAIDs inhibition hands, whereas it was noted unstable previously (30). of tumour development. Indeed, in human plasma, We then compared the anti-proliferative properties of two celecoxib concentrations were 10 to 100 times lower, i.e. COX-2 inhibitors, celecoxib and rofecoxib. These two between 0.5 and 5 ÌM. Similarly, in Min mice the serum inhibitors were selected because they represent the last concentration of celecoxib found to inhibit tumour growth generation of NSAIDs which selectively target COX-2. The was 2-5 ÌM. Moreover, rofecoxib, which has almost no IC50 towards COX-2 were very close, and varied from 0.34 pro-apoptotic effect, also showed a potent anti-tumour to 0.96 ÌM for celecoxib and from 0.31 to 0.84 ÌM for effect in vivo (44). rofecoxib, depending on studies (36, 37). In our hands, In summary, our results show that rofecoxib and celecoxib was the most potent COX-2 inhibitor for growth celecoxib differently affected the growth of colorectal inhibition of colorectal cancer cells whatever their COX-2 cancer cells, both in a COX-2-independent manner. status was. These results support previous findings obtained Celecoxib induced apoptosis whereas rofecoxib did not. in leukaemia cell lines and epithelial cancer cell lines other The implication of COX-1, although unlikely because of than colorectal (38). At the highest celecoxib concentration the low efficacy of rofecoxib towards COX-1, deserves (100 ÌM), an almost complete inhibition of proliferation was supplementary studies. Furthermore, our findings suggest seen, whereas at this concentration rofecoxib was much less that a potentially new mechanism may be involved in the efficient. On the contrary, at lower concentration (10 ÌM) rofecoxib effect on cell growth. This mechanism, which both drugs were effective, some cells being more efficiently remains to be characterized, seems independent of COX-2 growth-inhibited with rofecoxib than with celecoxib. Thus, it expression, and may also be independent of the appears that a dual mechanisms occurred. PDK1/AKT pathways revealed by celecoxib.

231 ANTICANCER RESEARCH 25: 225-234 (2005)

Acknowledgements 15 DuBois RN, Giardiello FM and Smalley WE: Nonsteroidal anti-inflammatory drugs, eicosanoids, and colorectal cancer This work was supported by a grant from La Ligue contre le prevention. Gastroenterol Clin North Am 25: 773-791, 1996. Cancer, Comity of Loire-Atlantique, France. The authors 16 Sandler RS, Galanko JC, Murray SC, Helm JF and Woosley gratefully acknowledge Dr F. Blachier for helpful discussion. JT: Aspirin and nonsteroidal anti-inflammatory agents and risk for colorectal adenomas. Gastroenterology 114: 441- References 447, 1998. 17 Debinski HS, Trojan J, Nugent KP, Spigelman AD and Phillips 1 Dubois RN, Abramson SB, Crofford L, Gupta RA, Simon LS, RK: Effect of sulindac on small polyps in familial adenomatous Van De Putte LB and Lipsky PE: Cyclooxygenase in biology polyposis. Lancet 345: 855-856, 1995. and disease. FASEB J 12: 1063-1073, 1998. 18 Steinbach G, Lynch PM, Phillips RK, Wallace MH, Hawk E, 2 Chandrasekharan NV, Dai H, Roos KL, Evanson NK, Tomsik Gordon GB, Wakabayashi N, Saunders B, Shen Y, Fujimura T, J, Elton TS and Simmons DL: COX-3, a cyclooxygenase-1 Su LK and Levin B: The effect of celecoxib, a cyclooxygenase-2 variant inhibited by acetaminophen and other analgesic/ inhibitor, in familial adenomatous polyposis. N Engl J Med 342: antipyretic drugs: cloning, structure, and expression. Proc Natl 1946-1952, 2000. Acad Sci USA 99: 13926-13931, 2002. 19 Giardiello FM, Yang VW, Hylind LM, Krush AJ, Petersen GM, 3 Smith WL and Langenbach R: Why there are two Trimbath JD, Piantadosi S, Garrett E, Geiman DE, Hubbard W, cyclooxygenase isozymes. J Clin Invest 107: 1491-1495, 2001. Offerhaus GJA and Hamilton SR: Primary chemoprevention of 4 Smith WL, DeWitt DL and Garavito RM. Cyclooxygenases: familial adenomatous polyposis with sulindac. N Engl J Med 346: structural, cellular, and molecular biology. Annu Rev Biochem 1054-1059, 2002. 69: 145-182, 2000. 20 Cruz-Correa M, Hylind LM, Romans KE, Booker SV and 5 Eberhart CE, Coffey RJ, Radhika A, Giardiello FM, Giardiello FM: Long term treatment with sulindac in familial Ferrenbach S and DuBois RN: Up-regulation of cyclooxygenase adenomatous polyposis: a prospective cohort study. 2 gene expression in human colorectal adenomas and Gastroenterology 122: 641-645, 2002. adenocarcinomas. Gastroenterology 107: 1183-1188, 1994. 21 Zhang X, Morham SG, Langenbach R and Young DA: Malignant 6 Sano H, Kawahito Y, Wilder RL, Hashiramoto A, Mukai S, transformation and antineoplastic actions of nonsteroidal Asai K, Kimura S, Kato H, Kondo M and Hla T: Expression of antiinflammatory drugs (NSAIDs) on cyclooxygenase-null embryo cyclooxygenase-1 and -2 in human colorectal cancer. Cancer fibroblasts. J Exp Med 190: 451-459, 1999. Res 55: 3785-3789, 1995. 22 Song X, Lin HP, Johnson AJ, Tseng PH, Yang YT, Kulp SK and 7 Karnes WE Jr, Shattuck-Brandt R, Burgart LJ, DuBois RN, Chen CS: Cyclooxygenase-2, player or spectator in Tester DJ, Cunningham JM, Kim CY, McDonnell SK, Schaid DJ cyclooxygenase-2 inhibitor-induced apoptosis in prostate cancer and Thibodeau SN: Reduced COX-2 protein in colorectal cancer cells. J Natl Cancer Inst 94: 585-5891, 2002. with defective mismatch repair. Cancer Res 58: 5473-5477, 1998. 23 Shiff SJ, Qiao L, Tsai LL and Rigas B: Sulindac sulfide, an 8 Shao J, Sheng H, Aramandla R, Pereira MA, Lubet RA, Hawk aspirin-like compound, inhibits proliferation, causes cell cycle E, Grogan L, Kirsch IR, Washington MK, Beauchamp RD and quiescence, and induces apoptosis in HT-29 colon DuBois RN: Coordinate regulation of cyclooxygenase-2 and adenocarcinoma cells. J Clin Invest 96: 491-503, 1995. TGF-beta1 in replication error-positive colon cancer and 24 Elder DJ, Halton DE, Crew TE and Paraskeva C: Apoptosis azoxymethane-induced rat colonic tumors. Carcinogenesis 20: induction and cyclooxygenase-2 regulation in human colorectal 185-191, 1999. adenoma and carcinoma cell lines by the cyclooxygenase-2- 9 Boolbol SK, Dannenberg AJ, Chadburn A, Martucci C, Guo selective non-steroidal anti-inflammatory drug NS-398. Int J XJ, Ramonetti JT, Abreu-Goris M, Newmark HL, Lipkin ML, Cancer 86: 553-560, 2000. DeCosse JJ and Bertagnolli MM: Cyclooxygenase-2 25 Zhang Z and DuBois RN: Par-4, a proapoptotic gene, is overexpression and tumor formation are blocked by sulindac in regulated by NSAIDs in human colon carcinoma cells. a murine model of familial adenomatous polyposis. Cancer Res Gastroenterology 118: 1012-1017, 2000. 56: 2556-2560, 1996. 26 Gupta RA and DuBois RN: Colorectal cancer prevention and 10 Rigas B, Goldman IS and Levine L: Altered eicosanoid levels treatment by inhibition of cyclooxygenase-2. Nature Rev Cancer in human colon cancer. J Lab Clin Med 122: 518-523, 1993. 1: 11-21, 2001. 11 Tsujii M and DuBois RN: Alterations in cellular adhesion and 27 Blottiere HM, Zennadi R, Gregoire M, Aillet G, Denis MG, apoptosis in epithelial cells overexpressing prostaglandin Meflah K and Le Pendu J: Analysis of the relationship between endoperoxide synthase 2. Cell 83: 493-501, 1995. stage of differentiation and NK/LAK susceptibility of colon 12 Liu CH, Chang SH, Narko K, Trifan OC, Wu MT, Smith E, carcinoma cells. Int J Cancer 53: 409-417, 1993. Haudenschild C, Lane TF and Hla T: Overexpression of 28 Siavoshian S, Segain JP, Kornprobst M, Bonnet C, Cherbut C, cyclooxygenase-2 is sufficient to induce tumorigenesis in Galmiche JP and Blottiere HM: Butyrate and trichostatin A transgenic mice. J Biol Chem 276: 18563-18569, 2001. effects on the proliferation/differentiation of human intestinal 13 Buecher B, Heymann MF and Blottiere HM: Rationale and epithelial cells: induction of cyclin D3 and p21 expression. Gut prospects for the use of cyclooxygenase-2 (COX-2) inhibitors in 46: 507-514, 2000. colorectal cancer. Gastroenterol. Clin Biol 25: 967-978, 2001. 29 Parsons R, Myeroff LL, Liu B, Willson JK, Markowitz SD, 14 Paganini-Hill A, Hsu G, Ross RK and Henderson BE: Aspirin Kinzler KW and Vogelstein B: Microsatellite instability and use and reduced risk of fatal colon cancer. N Engl J Med 326: mutations of the transforming growth factor beta type II receptor 1290-1291, 1992. gene in colorectal cancer. Cancer Res 55: 5548-5550, 1995.

232 Buecher et al: Rofecoxib, Celecoxib and Colorectal Tumour Cell Proliferation

30 Hoang JM, Cottu PH, Thuille B, Salmon RJ, Thomas G and 38 Waskewich C, Blumenthal RD, Li H, Stein R, Goldenberg DM Hamelin R: BAT-26, an indicator of the replication error and Burton J: Celecoxib exhibits the greatest potency amongst phenotype in colorectal cancers and cell lines. Cancer Res 57: cyclooxygenase (COX) inhibitors for growth inhibition of COX- 300-303, 1997. 2-negative hematopoietic and epithelial cell lines. Cancer Res 31 Boudreau MD, Sohn KH, Rhee SH, Lee SW, Hunt JD and 62: 2029-2033, 2002. Hwang DH: Suppression of tumor cell growth both in nude 39 Yamazaki R, Kusunoki N, Matsuzaki T, Hashimoto S and mice and in culture by n-3 polyunsaturated fatty acids: Kawai S: Selective cyclooxygenase-2 inhibitors show a mediation through cyclooxygenase-independent pathways. differential ability to inhibit proliferation and induce apoptosis Cancer Res 61: 1386-1391, 2001. of colon adenocarcinoma cells. FEBS Lett 531: 278-284, 2002. 32 Sheng H, Shao J, Kirkland SC, Isakson P, Coffey RJ, Morrow J, 40 Zhu J, Song X, Lin HP, Young DC, Yan S, Marquez VE and Beauchamp RD and DuBois RN: Inhibition of human colon Chen CS: Using cyclooxygenase-2 inhibitors as molecular cancer cell growth by selective inhibition of cyclooxygenase-2. J platforms to develop a new class of apoptosis-inducing agents. J Clin Invest 99: 2254-2259, 1997. Natl Cancer Inst 94: 1745-1757, 2002. 33 Hsi LC, Baek SJ and Eling TE: Lack of cyclooxygenase-2 41 Arico S, Pattingre S, Bauvy C, Gane P, Barbat A, Codogno P activity in HT-29 human colorectal carcinoma cells. Exp Cell and Ogier-Denis E: Celecoxib induces apoptosis by inhibiting Res 256: 563-5670, 2000. 3-phosphoinositide-dependent protein kinase-1 activity in the 34 Masferrer JL, Leahy KM, Koki AT, Zweifel BS, Settle SL, human colon cancer HT-29 cell line. J Biol Chem 277: 27613- Woerner BM, Edwards DA, Flickinger AG, Moore RJ and 27621, 2002. Seibert K: Antiangiogenic and antitumor activities of 42 Grösch S, Tegeder I, Niederberger E, Bräutigam L and cyclooxygenase-2 inhibitors. Cancer Res 60: 1306-1311, 2000. Geisslinger G: COX-2 independent induction of cell cycle arrest 35 Boland CR, Thibodeau SN, Hamilton SR, Sidransky D, and apoptosis in colon cancer cells by the selective COX-2 Eshleman JR, Burt RW, Meltzer SJ, Rodriguez-Bigas MA, inhibitor celecoxib. FASEB J 15: 2742-2754, 2001. Fodde R, Ranzani GN and Srivastava S: A National Cancer 43 Chulada PC, Thompson MB, Mahler JF, Doyle CM, Gaul BW, Institute Workshop on microsatellite instability for cancer Lee C et al: Genetic disruption of ptgs-1, as well as ptgs-2, detection and familial predisposition: development of reduces intestinal tumorigenesis in Min mice. Cancer Res 60: international criteria for the determination of microsatellite 4705-4708, 2000. instability in colorectal cancer. Cancer Res 58: 5248-5257, 1998. 44 Yao M, Kargman S, Lam EC, Kelly CR, Zheng Y, Luk P, 36 Warner TD, Giuliano F, Vojnovic I, Bukasa A, Mitchell JA and Kwong E, Evans JF and Wolfe MM: Inhibition of Vane JR: Nonsteroid drug selectivities for cyclo-oxygenase-1 cyclooxygenase-2 by rofecoxib attenuates the growth and rather than cyclo-oxygenase-2 are associated with human metastatic potential of colorectal carcinoma in mice. Cancer gastrointestinal toxicity: a full in vitro analysis. Proc Natl Acad Res 63: 586-592, 2003. Sci USA 96: 7563-7568, 1999. 37 Chan CC, Boyce S, Brideau C, Charleson S, Cromlish W, Ethier D, Evans J, Ford-Hutchinson AW, Forrest MJ, Gauthier JY, Gordon R, Gresser M et al: Rofecoxib [Vioxx, MK-0966; 4-(4'- methylsulfonylphenyl)-3-phenyl-2-(5H)-furanone]: a potent and orally active cyclooxygenase-2 inhibitor. Pharmacological and Received August 6, 2004 biochemical profiles. J Pharmacol Exp Ther 290: 551-560, 1999. Accepted November 19, 2004

233