P-Glycoprotein Mediates Celecoxib-Induced Apoptosis in Multiple Drug-Resistant Cell Lines Free

One type of drug resistance to anticancer compounds is mediated by the overexpression of the P-glycoprotein, the product of multidrug resistant (MDR1) gene (1). P-glycoprotein is a membrane-associated glycoprotein and a member of the ATP-binding cassette superfamily of membrane transport proteins (2). P-glycoprotein works like an energy-dependent efflux pump to extrude anticancer drugs and other compounds from tumor cells (3).

Previous studies have shown increased levels of the inducible enzyme cyclooxygenase-2 (COX-2) in various types of carcinoma, including hepatocellular carcinoma (4). Cyclooxygenases (COX-1 and COX-2) are enzymes that catalyze the metabolism of arachidonic acid to prostaglandins and thromboxanes (5). COX-1 is constitutively expressed in many tissues and is responsible for various physiologic functions (6–8). COX-2 is an immediate-early inducible gene that responds to various stimuli, such as mitogens, cytokines, and growth factors; and is involved in inflammation, regulation of cell growth, differentiation, prevention of apoptosis, tumorigenesis, and angiogenesis (7, 9–14).

The expression of P-glycoprotein and COX-2 in cancerous tissues varies according to several factors, decreasing with the increase of anaplasia (4, 15). This observation suggests that P-glycoprotein and COX-2 can play a role in early stage of carcinogenesis and in malignity in many organs, liver included, although very few links have been shown between P-glycoprotein and COX-2 pathways. We showed that the development of the MDR phenotype is associated with a constitutive expression of COX-2 and inducible nitric oxide synthase in a human hepatocellular carcinoma cell line and these enzymes may enhance the angiogenic activity of the MDR-positive hepatocellular carcinoma cell line (16, 17). Moreover, a causal link between expression and activity of COX-2 and P-glycoprotein was shown (18–21). Cells expressing the MDR phenotype are more resistant to several agents that trigger cell apoptosis compared with parental drug-sensitive cells (22). This phenomenon can be linked to the fact that MDR-positive cells exhibit a block in the release of mitochondrial cytochrome c into the cytosol and, because of this, they seem protected from undergoing apoptosis (23). This feature of MDR-positive cells seems to be dependent on overexpression of Bcl-x_L, an antiapoptotic protein that belongs to the Bcl-2 family (23). Bcl-x_L and Bcl-2 act on the release of cytochrome c from mitochondria by modulating the activity of the voltage-dependent anion channel (24). Their overexpression blocks cytochrome c release and activation of caspase-3 (25). Moreover, the transfection of human hepatocellular carcinoma cell line PLC/PRF/5 with a plasmid vector containing recombinant bcl-2 makes these cells resistant to doxorubicin-induced cytotoxicity, suggesting a possible link between P-glycoprotein and Bcl-2 family members expression (26).

COX-2 is thought to play a role in cell apoptosis, but the signaling pathways affected by COX-2 expression and/or activity are still not fully understood (27). The COX inhibitors, known as coxibs, inhibit the proliferation and induce apoptosis in cultured hepatocellular carcinoma cells, although these inhibitors are known to mediate effects through both COX-dependent and COX-independent mechanisms (28). Celecoxib induces cytochrome c release, activation of caspase-9 and caspase-3 and eventually apoptosis in hepatocellular carcinoma cells (29, 30). However, its mechanism of action should be further explored. It is possible to prevent celecoxib-induced cell death by z-Val Ala Asp (Ome)-fluoromethylketone (z-VAD), a wide-spectrum inhibitor of caspase activity (29). Celecoxib, a well-known inhibitor of COX-2 activity, induces apoptosis also in cells that lack the COX-2 enzyme (31).

Because COX-2 and P-glycoprotein proteins are expressed in hepatocellular carcinoma (16, 32), they could affect the apoptotic pathways and cooperate in the development of the MDR phenotype, at least in human hepatocellular carcinoma. In recent years, inhibitors of COX-2, celecoxib included, have been proposed in the treatment of hepatocellular carcinoma, and in fact, the inhibition of COX-2 was proved to possess antiproliferative action (33).

Based on these observations, this work was done to examine the effect of celecoxib on different cell lines expressing the MDR phenotype. To address this issue, the expressions of P-glycoprotein, COX-2, Bcl-2, and Bcl-x_L; the release of mitochondrial cytochrome c; the translocation of Bax to the mitochondria; the percentage of fragmented Hoechst-positive nuclei, of Annexin V–positive cells, and of terminal deoxynucleotidyl transferase (TdT)–mediated nick end labeling (TUNEL)–stained cells; and the cell death were assessed at basal conditions and after exposure to celecoxib in different cell lines.

Cell lines. Experiments were done mainly on a human hepatocellular carcinoma cell line PLC/PRF/5 (34), by using the drug-sensitive clone (P5) and the highly MDR subclone [P1(0.5)], and confirmative experiments were done on human colon cancer drug-sensitive (HT-29) and drug-resistant (HT-29-dx) cells, on human ovarian cancer cells (IGROV-1), and on murine NIH3T3 drug-sensitive (PSI-2) and drug-resistant (PN1A) cells.

The P1(0.5) subclone was developed by serial prolonged exposures to increasing concentrations of doxorubicin (Pharmacia & Upjohn) starting from parental P5 cells, and subsequently cultured in DMEM containing 0.5 μg/mL doxorubicin and 10% fetal bovine serum (FBS), as previously reported (16). The human colon carcinoma cells HT-29 and the subpopulation of doxorubicin-resistant cells (HT29-dx) were kindly donated by Dr. A. Bosia (Department of Genetics, Biology and Biochemistry, University of Turin, Turin, Italy). The HT-29 cells were cultured in RPMI 1640 supplemented with 10% FBS; the HT-29-dx were maintained in RPMI 1640 supplemented with 10% FBS and 34 nmol/L doxorubicin, as described by C. Riganti et al. (35). The human ovarian adenocarcinoma cell line (IGROV-1), kindly provided by Dr. J.H. Schellens (the Netherlands Cancer Institute, Amsterdam, the Netherlands), was maintained in RPMI 1640 supplemented with 10% FBS (36). Two subclones of NIH3T3 cells, one of which is drug-resistant (PN1A) and the other drug-sensitive (PSI-2), were originally provided by J.M. Croop (Pediatric Hematology/Oncology, James Whitcomb Riley Hospital for Children, Indianapolis, IN) and were cultured in DMEM supplemented with 10% FBS; 0.1 μg/mL doxorubicin was added to the PN1A cell medium, as previously described (37). The PN1A cell line was generated by transfection of PSI-2 cells with expression vector pBA-mdr, which contained mouse mdr1 cDNA. All cell lines were cultured at 37°C in an atmosphere containing 5% CO₂.

Cell treatment. Because only a part of celecoxib added to medium containing 10% FBS is protein-unbound and hence available for interacting on cells, doxorubicin-free, serum-free medium containing celecoxib was used in all experiments, as described in the details for each experimental condition (38).

Western blot analysis. Cells were seeded in complete medium for 24 h and then exposed to doxorubicin-free, serum-free medium containing 10 or 50 μmol/L celecoxib (a kind gift of Prof. E. Masini, Department of Preclinical and Clinical Pharmacology, University of Florence, Florence, Italy). After 3, 6, and 24 h, total proteins were extracted and evaluated by Western blot analysis, as previously described (16), using each of the following primary antibodies: anti–COX-2 polyclonal antibody (Cayman Chemical); anti–P-glycoprotein (C219) monoclonal antibody (Calbiochem); anti-Bax polyclonal antibody, anti–Bcl-2 monoclonal antibody, anti–Bcl-x_L monoclonal antibody, and anti–cytochrome c polyclonal antibody (Oncogene Research Products); anti–caspase-3 (E8) monoclonal antibody (Santa Cruz Biotechnology); anti–β-actin monoclonal antibody (Sigma Chemical Co.).

Small interfering RNA transfection. Positive control GAPDH small interfering RNA (siRNA), siCONTROL nontargeting siRNA and a pool of four different double-stranded RNA oligonucleotides directed against MDR1 (siGENOME SMARTpool, human ABCB1) were purchased from Dharmacon. P1(0.5) cells were seeded in 60-mm dishes to reach 30% to 50% confluency after 24 h of incubation, and transfected with a total of 100 nmol/L siRNA using LipofectAMINE 2000 (Invitrogen Life Technologies, Inc.) in antibiotic-free medium, according to the manufacturer's instructions. Forty-eight hours after transfection, total RNA was extracted for reverse transcription-PCR (RT-PCR) analysis, and 72 h after transfection, total proteins were extracted for Western blot analysis.

Reverse transcription-PCR. Total RNA was extracted from cells with TRIzol reagent (Invitrogen Life Technologies) and quantified by absorbance spectroscopy. The reverse transcription reaction was done by heating for 5 min at 65°C 5 ng of total RNA, 250 ng of random primer, and 1 mmol/L deoxyribonucleoside triphosphate (dNTP). After cooling at 4°C, the mixture was incubated for 50 min at 37°C in the reverse transcription reaction mixture (10 mmol/L DTT, 20 units RNAsin; Invitrogen), and 200 units of Moloney murine leukemia virus reverse transcriptase (Invitrogen Life Technologies). Samples were then heated at 70°C for 15 min. Five microliters of the synthesized cDNA was then amplified in a 55 μL reaction mixture containing Taq polymerase buffer, 1 mmol/L dNTP, 2.5 mmol/L MgCl₂, 0.5 μmol/L of each primer, and 0.5 units Taq polymerase (Invitrogen Life Technologies). The primers used were as follows: for MDR1, 5′-ATATCAGCAGCCCACATCAT-3′ and 5′-GAAGCACTGGGATGTCCGGT-3′; for GAPDH, 5′-GCCAAAAGGGTCATCATCTC-3′ and 5′-GTAGAGGCAGGGATGATGTTC-3′ (39). Amplification cycles were 94°C for 3 min, then 33 cycles at 94°C for 1 min, 58°C for 1 min, 72°C for 1.5 min, followed by 72°C for 15 min. Aliquots of the PCR product were electrophoresed on 1.5% agarose gels, and PCR fragments were visualized by ethidium bromide staining (39).

Preparation of mitochondrial fractions. Cells were washed twice with ice-cold PBS, trypsinized, and collected by centrifugation; the pellet was suspended in 500 μL of ice-cold buffer A [20 mmol/L HEPES (pH 7.5), 1.5 mmol/L MgCl₂, 10 mmol/L KCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 1 mmol/L DTT, 0.1 mmol/L phenylmethylsulfonyl fluoride, and 10 μg/mL each leupeptin, aprotinin, and pepstatin A] containing 250 mmol/L sucrose.

Unbroken cells, large plasma membrane pieces, and nuclei were removed by centrifuging the homogenates at 1,000 × g at 4°C for 10 min. The resulting supernatant was subjected to 10,000 × g centrifugation at 4°C for 20 min. The pellet fraction (i.e., mitochondria) was first washed with the above buffer A containing sucrose and then solubilized in 50 μL of TNC buffer [10 mmol/L Tris acetate (pH 8.0), 0.5% Igepal CA-630, 5 mmol/L CaCl₂]. The supernatant was recentrifuged at 100,000 × g (4°C, 1 h) to generate cytosol.

Immunofluorescence analysis of cytochrome c release. Cells were plated onto a coverslip and were grown in culture medium containing 10% FBS. After 24 h, cells were treated with 10 μmol/L celecoxib in doxorubicin-free, serum-free medium for 3 h and then were fixed with paraformaldehyde (3.7% w/v), permeabilized with 0.05% (v/v) Triton X-100 for 10 min, blocked with 10% (w/v) normal goat serum, then incubated overnight with primary antibody at an appropriate dilution (1:500). The purified anti–cytochrome c monoclonal antibody was obtained from BD Biosciences PharMingen (BD Biosciences). After washing, samples were incubated with FITC-conjugated secondary antibodies (Molecular Probes) at room temperature in the dark. FITC-labeled cells were analyzed by fluorescence Nikon microscope using standard fluorescein filters (excitation 488 nm).

Hoechst assay. Nuclear morphology was assessed using Hoechst staining. Cells were plated onto coverslip and were grown in complete medium. After 24 h at 37°C, the cells were rinsed to remove the detached cells, and then exposed to doxorubicin-free, serum-free medium containing 10 μmol/L celecoxib or 10 μmol/L verapamil or 100 or 200 μmol/L taurolithocholic acid for 24 h. Afterward, the supernatant was removed and cells were washed twice with PBS. Then, the cells were fixed for 20 min at room temperature in freshly prepared 3.7% paraformaldehyde in PBS, then rinsed thrice with PBS. After that, 0.5% Triton X-100 was added to the sample and left for 5 min at room temperature. Cells were washed twice with PBS and stained for 30 min at 37°C in Hoechst 33258 1 μg/mL (Sigma Chemical); then, they were rinsed with PBS and examined with a fluorescence microscope (UV light). Fragmented nuclei were expressed as percentage of fragmented Hoechst-positive nuclei versus the total Hoechst-positive nuclei.

Annexin V assay. Drug-sensitive and drug-resistant cells were plated in 60-mm Petri dishes and grown in DMEM supplemented with 10% FBS. The next day, cells were treated with 10 μmol/L celecoxib in doxorubicin-free, serum-free medium for 24 h. Afterward, both floating and attached cells were collected by brief trypsinization. Cells were stained with an Annexin V–Fluos staining kit (Roche Applied Science), and apoptosis was evaluated by using the FACScan flow cytometer (Becton Dickinson).

TUNEL assay. DNA fragmentation during apoptosis was detected by the TUNEL technique using an apoptosis detection kit. Briefly, drug-sensitive and drug-resistant cells were plated in 60-mm Petri dishes and grown in DMEM supplemented with 10% FBS. The next day, cells were treated with 10 μmol/L celecoxib in doxorubicin-free, serum-free medium for 24 h. Cells were fixed in 10% (v/v) formalin and, subsequently, a 250 μL TUNEL reaction mixture containing biotinylated nucleotide and TdT was added to cells for 60 min at 37°C in the dark. For negative control, TdT was omitted from the reaction mixture. The reaction was analyzed by Nikon fluorescent microscope.

Cell death assay. P5 and P1(0.5) cells were plated in 60-mm Petri dishes and grown in DMEM supplemented with 10% FBS. On the next day, cells were treated with 10 μmol/L celecoxib in doxorubicin-free, serum-free medium; moreover, 1 h before the addition of celecoxib, cells were preincubated with a cell permeable pan-caspase-inhibitor z-VAD at 20 μmol/L (Sigma). After 6, 12, 18, and 24 h, the dead cells were identified using the trypan blue staining test. Adherent cells were detached, resuspended in DMEM, and both adherent and suspended cells were mixed with an equal volume of 0.04% (w/v) trypan blue. The percentage of dead cells stained was determined by counting.

Statistical analysis. The significance of the differences between groups was determined by ANOVA followed by the Bonferroni post hoc test, using GraphPad Prism version 4.0 for Windows (GraphPad Software). P < 0.05 was considered significant.

Celecoxib modifies P-glycoprotein expression in MDR-positive cells. First, we confirmed that P1(0.5) cells have markedly higher expression of P-glycoprotein and COX-2 compared with parental cells P5 (Fig. 1A,, i). To evaluate whether COX-2 activity induced the expression of P-glycoprotein in our model, we determined whether the selective inhibition of COX-2 activity by celecoxib modified P-glycoprotein and/or COX-2 expression in P1(0.5) and in parental P5 cells. The levels of P-glycoprotein and COX-2 expression in P1(0.5) cells cultured in a doxorubicin-free, serum-free medium for 3, 6, and 24 h were not modified (Fig. 1A,, ii). Western blot analysis clearly showed that the P-glycoprotein expression was reduced in P1(0.5) cells treated for 3 h with 10 μmol/L celecoxib and the reduced expression persisted at least up to 24 h, whereas COX-2 expression was not altered (Fig. 1B,, i and ii). On the contrary, cell exposure to a 5-fold higher concentration of celecoxib (50 μmol/L) weakly increased COX-2 expression at 6 h and P-glycoprotein expression at 24 h in P1(0.5) cells (Fig. 1C). P-glycoprotein and COX-2 proteins were not expressed in P5 parental cells at basal conditions and their exposure to celecoxib did not change the expression of these proteins (Fig. 1A,, i, B, i, and C). Because it is known that COX-2 induces P-glycoprotein expression but it is unknown whether P-glycoprotein can modify COX-2 expression, we evaluated the effect of inhibiting the expression of MDR1 by siRNA on COX-2 expression. The MDR1 siRNA reduced the levels of MDR1 mRNA (Fig. 1D,, i) and abolished the expression of P-glycoprotein (Fig. 1D,, ii). siRNA targeting GAPDH strongly reduced GAPDH transcript levels (data not shown). The inhibition of P-glycoprotein expression was not observed using the nonspecific control siRNA or the mock transfection control (data not shown). MDR1 siRNA did not affect the expression of COX-2 in P1(0.5) cells (Fig. 1D , ii). The inhibitory effect of celecoxib on P-glycoprotein expression was also evaluated in different cell lines (Supplementary Fig. S1). The levels of P-glycoprotein in HT-29-dx, IGROV-1, and PN1A cells cultured in a doxorubicin-free, serum-free medium for 3, 6, and 24 h were not modified (data not shown). The expression of P-glycoprotein was reduced in HT-29-dx after 3, 6, and 24 h of treatment with 10 μmol/L celecoxib (Supplementary Fig. S1A, i and iii). On the contrary, under the same experimental conditions, the P-glycoprotein levels were not affected in parental HT-29 cells (Supplementary Fig. S1A, i and ii). As shown in Supplementary Fig. S1A, iv, celecoxib reduced P-glycoprotein expression only after 6 h in IGROV-1 cells. Moreover, P-glycoprotein expression was reduced in mdr1 cDNA transfected PN1A cells after treatment with 10 μmol/L celecoxib for 3, 6, and 24 h (Supplementary Fig. S1C, i and iii). No changes in the expression of P-glycoprotein were observed in parental PSI-2 cells treated with celecoxib (Supplementary Fig. S1C, i and ii).

Celecoxib modifies Bcl-x_L and Bcl-2 expression in MDR-positive cells. To examine a possible association between P-glycoprotein expression and resistance to mitochondrial-related apoptosis, the expression of Bcl-x_L, Bcl-2, and Bax proteins was determined in parental P5 and in MDR- and COX-2–positive P1(0.5) cells. Western blot analysis showed that antiapoptotic proteins Bcl-x_L and Bcl-2 were more expressed in P1(0.5) cells than in parental P5 cells at basal conditions. On the contrary, the expression of Bax, the proapoptotic member of Bcl-2 family, was higher in parental P5 than in P1(0.5) cells (Fig. 2A,, i). The levels of Bcl-x_L, Bcl-2, and Bax expression in P1(0.5) cells cultured in doxorubicin-free, serum-free medium up to 24 h were unmodified (Fig. 2A,, ii). Interestingly, 10 μmol/L celecoxib strongly reduced Bcl-x_L expression after 3 and 6 h and weakly after 24 h (Fig. 2B,, i and ii); celecoxib also reduced Bcl-2 expression at 6 h in P1(0.5) cells (Fig. 2B,, i and ii). On the contrary, 50 μmol/L celecoxib did not alter the expression of these proteins in P1(0.5) cells (data not shown). No significant changes in Bcl-x_L or Bcl-2 expression were observed in parental P5 cells (Fig. 2B , i).

To test the role of P-glycoprotein in the process of mitochondrial apoptosis induction, we did siRNA experiments in P1(0.5) cells. The MDR1 siRNA reduced the expression of antiapoptotic proteins Bcl-x_L and Bcl-2, compared with basal conditions, whereas 10 μmol/L celecoxib did not modify the Bcl-x_L and Bcl-2 expression in P1(0.5) cells transfected with MDR1 siRNA (Fig. 2C).

Because P-glycoprotein activity may affect mitochondrial apoptosis induction, we assessed the effect of verapamil and taurolithocholic acid, two inhibitors of P-glycoprotein activity (40), on P-glycoprotein, procaspase-3, Bcl-xL, Bcl-2, and Bax expression in MDR-positive cells. As shown in Supplementary Fig. S2A and B, the treatment with 10 μmol/L verapamil or 100 or 200 μmol/L taurolithocholic acid for 3, 6, or 24 h did not modify the expression of these proteins in P1(0.5) cell line. Moreover, the two inhibitors of P-glycoprotein activity did not alter the percentage of fragmented Hoechst-positive nuclei (Supplementary Fig. S2C).

Celecoxib induces Bax translocation and mitochondrial cytochrome c release in MDR-positive cells. The expression of Bax did not change in P5 and P1(0.5) cells after exposure to 10 μmol/L celecoxib (Fig. 3A). However, a different subcellular localization of Bax between cytosolic and mitochondrial fractions was observed in P1(0.5) cells after their exposure to 10 μmol/L celecoxib for 3, 6, or 24 h. As shown in Fig. 3B, a partial translocation of Bax to mitochondria was observed after 3 and 6 h of treatment with 10 μmol/L celecoxib. The subcellular localization of Bax did not change in parental P5 cells after exposure to celecoxib (Fig. 3B).

To test the hypothesis that 10 μmol/L celecoxib could restore mitochondrial cytochrome c release into the cytosol, MDR- and COX-2–positive cells were treated with 10 or 50 μmol/L celecoxib for 3, 6, and 24 h, then cells were fractioned and cytosolic and mitochondrial cytochrome c were analyzed by Western blot analysis. As shown in Fig. 4A, cytochrome c was released into the cytosol following treatment of P1(0.5) cells with 10 μmol/L celecoxib for 3, 6, and, especially, 24 h. On the contrary, cells treated with a higher concentration of celecoxib (50 μmol/L) did not release cytochrome c from mitochondria into cytosol (Fig. 4A). Immunofluorescence analysis on P5 and P1(0.5) cells treated for 3 h with 10 μmol/L celecoxib confirmed results of Western blot analysis, showing the effect of celecoxib on cytochrome c release into cytosol only in P1(0.5) cells (Fig. 4B). On the contrary, no changes in mitochondrial cytochrome c release were observed when parental P5 cells were treated with celecoxib (Fig. 4A and B).

Effect of celecoxib on apoptosis and cell death of MDR-positive and MDR-negative cells. Because Bax translocation from cytosol to the mitochondria and the mitochondrial cytochrome c release into the cytosol play a crucial role in the induction of cellular apoptosis, the expression of the precursor form of caspase-3 (procaspase-3) was evaluated by Western blot analysis. Moreover, the percentage of fragmented nuclei, estimated by histochemical nuclear staining with Hoechst 33258, the percentage of Annexin V– and TUNEL-positive cells were evaluated in P5 and P1(0.5) cell lines treated with celecoxib. Evident activation of caspase-3 was shown by degradation of its precursor form after treatment with 10 μmol/L celecoxib at 3 and 6 h in P1(0.5) cells (Fig. 5A), at 3, 6, and 24 h in HT-29-dx cells and at 3 and 6 h in IGROV-1 cells (Supplementary Fig. S1A, iii and iv). On the contrary, the procaspase-3 levels were not affected in parental P5 and HT-29 cells (Fig. 5A and Supplementary Fig. S1A ii). In addition, treatment with 50 μmol/L celecoxib for 3, 6, or 24 h did not modify the procaspase-3 levels and the percentage of cells with fragmented Hoechst-positive nuclei in P1(0.5) cells (Fig. 5A and C).

The inhibition of P-glycoprotein expression by MDR1 siRNA transfection in P1(0.5) cells prevented the activating effect of celecoxib on caspase-3 (Fig. 5B).

Treatment of P1(0.5) cells with 10 μmol/L celecoxib for 24 h significantly increased the percentage of cells with fragmented Hoechst-positive nuclei and the percentage of Annexin V– and TUNEL-positive cells, whereas the same concentration of celecoxib had no effect on P5 cells (Fig. 5C ; and D, i and ii).

Furthermore, treatment with 10 μmol/L celecoxib increased the number of cells in apoptosis in all P-glycoprotein overexpressing cell lines (Supplementary Fig. S1B and S1D, i and ii). The effect of 10 μmol/L celecoxib for 6, 12, 18, and 24 h on the death of P5 and P1(0.5) cells by trypan blue dye exclusion assay was also evaluated. As shown in Fig. 6B, celecoxib significantly increased cell death in P1(0.5) cells compared with doxorubicin-free, serum-free controls, whereas the same concentration of celecoxib had no effect on viability of P5 cells (Fig. 6A). Moreover, preincubation with 20 μmol/L Z-VAD, a universal inhibitor of caspases, abrogated the enhanced cell death induced by celecoxib treatment in MDR-positive cells (Fig. 6B).

The presence of intrinsic or acquired MDR is the major factor responsible for chemotherapy failure in cancer patients who are undergoing chemotherapy. The MDR phenotype is characterized by the overexpression of P-glycoprotein in plasma membrane that works as a pump to extrude anticancer drugs from cells. However, in addition to the overexpression of P-glycoprotein, the MDR phenotype is associated with other modifications of cell biology that make cancer cells more resistant to many other mechanisms of cell damage (2, 41, 42). We have previously shown that the development of the MDR phenotype is associated with the constitutive expression of COX-2 in hepatocellular carcinoma cell lines (16). P-glycoprotein can regulate the mitochondrion-dependent apoptotic pathway, conferring resistance to a caspase-dependent mechanism of apoptosis induced by antineoplastic agents (23). The antitumor effect of COX-2 inhibitors has been described in recent years although the molecular mechanisms involved remain elusive (43, 44). The relationship between COX-2 and P-glycoprotein suggests that celecoxib, a specific COX-2 activity inhibitor, may improve results of chemotherapy by increasing the sensitivity of tumor cells to anticancer drugs.

In the present study, we investigated the effect of celecoxib on COX-2 and P-glycoprotein expression and on factors involved in the mitochondrial apoptosis pathway, in MDR-positive cell lines, and in their drug-sensitive parental cell clones. To determine whether the possible antitumor effects of celecoxib could be dose dependent, two concentrations of celecoxib were used, the lowest (10 μmol/L) that can be reached in the clinic and another that is 5-fold higher (50 μmol/L; ref. 45). To our surprise, the lower concentration of celecoxib, in addition to the inhibitory effect on COX-2 activity (16), markedly reduced the expression of P-glycoprotein without affecting COX-2 expression, whereas 50 μmol/L celecoxib, although inhibited COX-2 activity, weakly enhanced the expression of P-glycoprotein and COX-2. It is known that the higher concentration of celecoxib inhibits both COX-1 and COX-2 activities (46); thus, it is likely that total inhibition of arachidonic acid metabolism due to celecoxib could be sensed by cells as being damaging and could trigger a cell response with overexpression of COX-2 and P-glycoprotein proteins. These results are in agreement with other studies that have shown a COX-2–mediated regulation of P-glycoprotein expression (18). The observation that only 10 μmol/L celecoxib reduce P-glycoprotein expression suggests a direct inhibitory effect of celecoxib on P-glycoprotein, which has no effect on COX-2 expression in MDR-positive cells transfected with MDR1 siRNA. Moreover, although 10 and 50 μmol/L celecoxib inhibit COX-2 activity, only the lower concentration sharply reduces P-glycoprotein expression without altering COX-2 expression.

The double-face activity of celecoxib, depending on its concentration, on the expression of COX-2 and P-glycoprotein proteins, was also shown for nitric oxide on apoptosis. In fact, NO can either be proapoptotic or antiapoptotic according to its concentration (47). However, this is the first report to document that celecoxib can inhibit the expression of one protein (i.e., P-glycoprotein) at low concentration and, on the contrary, can increase its expression when a higher concentration is involved. It is unknown whether this aspect of the biology of celecoxib can explain the same controversial results observed in clinical trials. Clearly, it is a point that should be kept in mind when planning studies using this drug.

Based on these preliminary data, we decided to investigate the effect of celecoxib at a low concentration, which can be achieved in the serum of patients after standard anti-inflammatory and antineoplastic therapy, rather than using a higher concentration, which is not clinically achievable (45). First, we found that the MDR-positive clone P1(0.5) has a higher expression of antiapoptotic proteins Bcl-x_L and Bcl-2 and lower expression of proapoptotic Bax compared with P5 cells. Expressions of Bcl-x_L and Bcl-2 seemed to be correlated with that of P-glycoprotein, as evaluated by MDR1 siRNA transfection experiments. These data confirmed what is reported in the literature on the association between P-glycoprotein expression and the abrogation of cytochrome c release in mitochondrial apoptotic pathway (23).

Although it is known that celecoxib induces apoptosis in cultured hepatocellular carcinoma cells through COX-dependent and COX-independent mechanisms (28, 48), the precise function of celecoxib on the apoptotic signaling pathways remains uncertain. Zhang et al. (49) have shown that celecoxib at 50 μmol/L induced apoptosis through a mechanism involving Akt inactivation without altering Bcl-2 and Bcl-x_L protein levels in rat cholangiocarcinoma cells. In agreement with them, we found that 50 μmol/L celecoxib did not alter Bcl-x_L or Bcl-2 expression (data not shown). Only the 10 μmol/L celecoxib reduced both Bcl-x_L or Bcl-2 protein expression in P-glycoprotein–overexpressing cells and remained without effect in the same cells after transfection with MDR1 siRNA. Thus, the inhibitory effect of 10 μmol/L celecoxib on Bcl-x_L or Bcl-2 seems to be dependent on the expression of P-glycoprotein because its suppression by the MDR1 siRNA made 10 μmol/L celecoxib completely ineffective.

To assess the role of the activity of P-glycoprotein in this mechanism, experiments were done to test the effect of known P-glycoprotein activity inhibitors, such as verapamil and taurolithocholic acid, on the behavior of Bcl-xL, Bcl-2, and Bax proteins. These two compounds did not modify the expression of P-glycoprotein, procaspase-3, Bcl-xL, Bcl-2, and Bax proteins, ruling out the possibility that P-glycoprotein activity can modulate the expression of these proteins. The proapoptotic effect of celecoxib was further confirmed by the observation that 10 μmol/L celecoxib, although not affecting the expression of the proapoptotic protein Bax, caused its translocation from cytosol to mitochondria in P1(0.5) cells.

These data show that celecoxib restores the release of cytochrome c from the mitochondria when MDR-positive cells are exposed to 10 μmol/L celecoxib. This effect was shown only in the P-glycoprotein–overexpressing cell clones, whereas no effects were observed in no P-glycoprotein–overexpressing cells. Release of cytochrome c into cytosol initiates the activation of caspase-3 and increases the number of apoptotic cells. Moreover, the celecoxib-induced activation of caspase-3 is mediated by P-glycoprotein expression because the procaspase-3 cleavage was abolished when cells were transfected with the MDR1 siRNA. Recently, Mantovani et al. (50) have shown that P-glycoprotein is cleaved in a caspase-3–dependent manner. Considering the findings of the present work together with those by Mantovani et al., it can be hypothesized that the expression of P-glycoprotein blocks cytochrome c release from mitochondria to cytosol and, consequently, the activation of caspase-3. Therefore, the celecoxib-induced decrease in P-glycoprotein expression causes the release of cytochrome c and the activation of caspase-3. This enzyme cleaves P-glycoprotein, priming a caspase-3–mediated amplification of the apoptotic signaling pathways. If this is true, for the first time a self-enhancing mechanism (celecoxib-mediated reduction of P-glycoprotein expression leads to cytochrome c release and to the activation of caspase-3, which, in turn, can reduce P-glycoprotein levels) has been identified and could be considered a step to a no-return pathway for apoptosis in tumor cells. Celecoxib significantly increased the percentage of apoptotic cells, estimated by Hoechst 33258 staining, Annexin V assay, and TUNEL assay; and cell death, estimated by trypan blue exclusion assay, in P1(0.5) cells, without affecting apoptosis and viability of P5 cells. Accordingly, the preincubation of cells with z-VAD, a universal inhibitor of caspases, abrogated the enhanced cell death induced by celecoxib treatment. The recent evidence that P-glycoprotein is expressed and active in mitochondria membrane further contributes to the understanding of molecular mechanisms of these phenomena (37).

In conclusion, this work confirms that celecoxib exerts more than one pharmacologic action on cells. In the model of human hepatocellular carcinoma cells overexpressing the MDR phenotype as well as in other MDR-positive cells, celecoxib exerts a direct effect on the expression of P-glycoprotein and therefore on apoptotic pathways. MDR phenotype–expressing cells are intrinsically resistant to apoptosis; however, for the first time, we show that apoptosis can be restored by a P-glycoprotein–dependent mechanism in MDR-positive cell lines. The use of low doses of COX-2 inhibitors seems to be a promising approach to improving cancer chemotherapy outcome because celecoxib, which directly reduces P-glycoprotein expression, might enhance the susceptibility of tumor cells to apoptosis induced by anticancer drugs.

Grant support: Italian Ministry of University, Scientific and Technological Research (Ministero dell'Istruzione, dell'Università e della Ricerca, Progetto Nazionale cofinanziato COFIN N. 2002067115), the University of Florence (R. Mazzanti), and Faculty of Pharmacy of Novara (L. Tessitore).

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.

We thank Prof. I.M. Arias (Cell Biology and Metabolism Branch, National Institute of Child Health and Human Development, NIH, Bethesda, MD) for his critical and stimulating discussion on the pathophysiology of liver disease, and Prof. A.M. Vannucchi and Dr. C. Bogani (Department of Hematology, Azienda Ospedaliero-Universitaria Careggi, University of Florence, Florence, Italy) for their technical expertise in flow cytometry.