Induction but not inhibition of COX-2 confers human lung cancer cell apoptosis by celecoxib.

The antitumorigenic mechanism of the selective cyclooxygenase-2 (COX-2) inhibitor celecoxib is still a matter of debate. Among different structurally related COX-2 inhibitors, only celecoxib was found to cause apoptosis and cell death of human lung cancer cells (IC₅₀ values of 19.96 µM [A549], 12.48 µM [H460], and 41.39 µM [H358]) that was paralleled by a time- and concentration-dependent upregulation of COX-2 and peroxisome proliferator-activated receptor γ (PPARγ) at mRNA and protein levels. Apoptotic death of celecoxib-treated cancer cells was suppressed by the PPARγ antagonist GW9662 and by siRNA targeting PPARγ and, surprisingly, also by the selective COX-2 inhibitor NS-398 and siRNA targeting COX-2. NS-398 (1 µM) was shown to suppress celecoxib-induced COX-2 activity. Among the COX-2-dependent prostaglandins (PG) induced upon celecoxib treatment, PGD₂ and 15-deoxy-Δ¹²,¹⁴-PGJ₂ were found to induce a cytosol-to-nucleus translocation of PPARγ as well as a PPARγ-dependent apoptosis. Celecoxib-elicited PPARγ translocation was inhibited by NS-398. Finally, a COX-2- and PPARγ-dependent cytotoxic action of celecoxib was proven for primary human lung tumor cells. Together, our data demonstrate a proapoptotic mechanism of celecoxib involving initial upregulation of COX-2 and PPARγ and a subsequent nuclear translocation of PPARγ by COX-2-dependent PGs.

adenomatous polyposis (FAP). In this context, a six-month, twice-daily treatment with 400 mg of celecoxib was shown to lead to a signifi cant reduction in the number of colorectal polyps in patients with FAP ( 2 ). Another study suggested celecoxib for prevention of colorectal adenomas ( 3 ). Recent reports indicate celecoxib as treatment and preventive option for lung cancer (4)(5)(6)(7) and to enhance the response to classical chemotherapeutics in early stage non-small cell lung cancer (NSCLC) ( 8 ). These studies have attracted particular interest given that lung cancer is worldwide the most common cancer in terms of both incidence and mortality and that the response and remission rates in NSCLC patients still remains relatively low ( 9 ).

Analyses of nuclear proteins
Following incubation, cells were adjusted to a consistent cell number and lysed in 12.5 mM NaF, 25 mM ␤ -glycerophosphate, 25 mM para-nitrophenyl phosphate, and 2.5 mM NaVO 3 . After a centrifugation step, pellets were resuspended in 1 ml of a hypotonic buffer containing 20 mM HEPES (pH 7.5), 5 mM NaF, 10 µM Na 2 MoO 4 , and 0.1 mM EDTA. Afterwards, cells were allowed to swell on ice for 15 min, and then 50 µl of a 10% (w/v) Nonidet ® P-40 solution was added to each well and the solution was gently shaken. Following centrifugation of the homogenate, supernatants were carefully rinsed, and nuclear pellets were resuspended in 40 µl of complete lysis buffer containing 50 mM HEPES (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1% (v/v) Triton ® X-100, 10% (v/v) glycerol, 1 mM PMSF, 1 µg/ml leupeptin, 10 µg/ml aprotinin, and 0.1% (w/v) SDS. Thereafter, tubes were shaken on ice for 30 min, and a debris spin out was performed by centrifugation at 14,000 g for 10 min. Supernatants were used for determination of nuclear protein by Western blot as described above.

Analysis of nuclear PPAR ␥
For confocal imaging of PPAR ␥ and nuclear regions, fi xed cells were incubated with a PPAR ␥ antibody (Biomol GmbH) and a lamin A/C antibody (New England Biolabs GmbH). Secondary antibodies were a goat anti-rabbit Alexa Fluor ® 555-labeled IgG for detection of PPAR ␥ and a goat anti-mouse Alexa Fluor ® 488labeled IgG for detection of lamin A/C (Life Technologies Corporation, Darmstadt, Germany). All antibodies were diluted in PBS containing 0.3% (v/v) Triton ® X-100 and 1% (v/v) BSA.

Determination of COX-2 activity
To assess the effect of celecoxib on COX-2 activity and PG production by A549, H460, and H358 cells, two different experimental protocols were used.
In the fi rst approach , cells seeded in 24-well plates at a density of 2 × 10 5 cells per well and grown to confl uence were treated with aspirin (250 µM) for 2.5 h to inactivate endogenous COX activity. Thereafter, cells were extensively washed and incubated with celecoxib (30 µM for A549 and H460 cells, 50 µM for H358 cells) or vehicle for 24 h (A549), 48 h (H460), or 18 h (H358) to induce COX-2. Following extensive washing and medium change, NS-398 (1 µM) or celecoxib (1 µM) were added to the cultures, and the incubation was continued for another 30 min. Arachidonic acid (30 µM) was added subsequently, and cells were incubated in a fi nal volume of 300 µl for a further 15 min. Afterwards, the medium was removed and analyzed for PGE 2 , PGD 2 , and 15d-PGJ 2 .
In the second approach, experimental conditions comparable to those used for analyses of cytotoxicity and apoptosis were used. To this end, cells seeded in 24-well plates at a density of 2 × 10 5 cells per well and grown to confl uence were preincubated with NS-398 (1 µM) or its vehicle for 1 h, followed by 24 h (A549), 48 h (H460), or 18 h (H358) combined incubation with celecoxib at 30 µM (A549, H460) or 50 µM (H358). The fi nal volume of the supernatant was 300 µl. Afterwards, the medium was removed and analyzed for PGE 2 , PGD 2 , and 15d-PGJ 2 . that celecoxib may likewise induce COX-2 expression in cancer cells ( 37,38 ) and may elicit increased expression ( 14,(39)(40)(41) or activation ( 42 ) of PPAR ␥ . However, the functional consequence or crosstalks between these regulations have not been addressed. This study investigates a potential contribution and coordinated action of COX-2 and PPAR ␥ within the celecoxib-induced apoptosis of human NSCLC cell lines and primary lung cancer cells.
Primary lung tumor cells were obtained from resections of a brain metastasis of a 67-year-old male Caucasian (patient #1) and a 46year-old female Caucasian (patient #2) with NSCLC. Patients had been informed about the establishment of cellular models from their tumors and had given informed consent in written form. The procedure was approved by the institutional ethical committee. Samples from metastasis were excised, stored at 4°C in PBS, and immediately transferred to the laboratory. Samples were minced and single-cell suspensions were generated. Cells from patient #1 were passaged fi ve times in DMEM containing 20% FCS, 100 U/ml penicillin, and 100 µg/ml streptomycin for six weeks with medium change twice per week. Passages 5-8 were used for experiments. Experiments with cells of patient #2 were performed using passage 1.
All incubations were performed in serum-free medium. PBS was used as vehicle for test substances with a fi nal concentration of 0.1% (v/v) DMSO (for COX-2 inhibitors, GW9662, and PGs).

Statistics
Comparisons between groups were performed with Student two-tailed t -test or with one-way ANOVA plus post hoc Bonferroni test using GraphPad Prism 5.00 (GraphPad Software, San Diego, CA). IC 50 values were calculated by nonlinear regression of log(inhibitor) versus response.

Impact of different selective COX-2 inhibitors on apoptotic cell death
Analysis of the effects of different selective COX-2 inhibitors on viability and apoptosis of human lung cancer cells Cell culture media were used to evaluate PG levels using enzyme immunoassay kits (PGE 2 , PGD 2 kits from Cayman Chemical; 15d-PGJ 2 kit from Enzo Life Sciences). In both assays, PG levels were normalized to whole cell protein for the decreases in cell number elicited by celecoxib´s cytotoxic action, and subsequently expressed as percentage of vehicle control (100%).

Cell viability and apoptosis
Cells seeded at a density of 5 × 10 3 cells per well in 96-well, fl atbottom microplates (viability) or at 1 × 10 5 cells per well in 24-well plates (apoptosis) and grown to confl uence were used for incubations. Cell viability and apoptosis were analyzed using WST-1 test and Cell Death Detection ELISA PLUS kit, respectively, (both from Roche Diagnostics, Mannheim, Germany) ( 25 ).  the limited availability of primary tumor cells, analyses were restricted to key experiments, confi rming an upregulation of COX-2 and PPAR ␥ mRNA and protein at one time point ( Fig. 2B ).
In the second approach ( Fig. 4B ), the levels of PG were evaluated without exogenously added arachidonic acid. In this experimental setting, which was also used for analyses of cytotoxicity and apoptosis, NS-398 at 1 µM was added to cells 1 h prior to celecoxib (30-50 µM) followed by longterm (at least 18 h) coincubation with celecoxib. Collectively, these experiments ( Fig. 4B ) revealed results similar to the COX-2 activity assays shown in Fig. 4A in that the celecoxib-driven increase of the three PGs was also inhibited by NS-398 ( Fig. 4B ).

Impact of COX-2 and PPAR ␥ on celecoxib-induced apoptotic cell death
To investigate a potential involvement of COX-2 and PPAR ␥ in celecoxib-induced apoptotic cell death, experiments using NS-398 and the PPAR ␥ antagonist GW9662 were performed. As shown in Table 1 , NS-398 and GW9662 inhibited both apoptosis and cell death caused by celecoxib in each cell line as well as in primary cells.
To further substantiate the role of COX-2 and PPAR ␥ in celecoxib-induced apoptotic cell death and to exclude possible off-target effects of NS-398 and GW9662, transfection experiments were performed using siRNA targeting COX-2 cell lines tested, a greater than 1.5-fold induction of COX-2 and PPAR ␥ protein expression was unique for celecoxib (i.e., not shared by etoricoxib, rofecoxib, and valdecoxib).

Time course of celecoxib-induced COX-2, PPAR ␥ , and L-PGDS expression and PG production
In each cell line, celecoxib elicited a time-dependent increased expression of COX-2 and PPAR ␥ both at the mRNA ( Fig. 3A -C , left) and protein levels ( Fig. 3A-C ,  right). In addition, celecoxib treatment was associated with increased protein levels of lipocalin-type PGD synthase (L-PGDS), which catalyses the isomerization of PGH 2 to PGD 2 ( Fig. 3A-C , right).
Further time-course experiments revealed signifi cant increased concentrations of PGD 2 and 15d-PGJ 2 in supernatants of all cell lines within 12-24 h incubation with celecoxib ( Fig. 3D , middle and right). PGE 2 levels became signifi cantly decreased after 2 h treatment with celecoxib in all cell lines and increased above vehicle control levels thereafter (i.e., after 6 h in H460, 12 h in H358 cells, and 24 h in A549 cells; Fig. 3D , left). Noteworthy, experiments monitoring PGD 2 levels in the cell culture media of vehicle-and celecoxib-treated cells revealed a similar timecourse as compared with PGE 2 , with an initial drop of PGD 2 after 2 h (A549, H358) or 6 h (H460) and a subsequent increase of PGD 2 in all cell lines ( Fig. 3D , middle).

Impact of celecoxib on COX-2 activity and PG production
To evaluate whether the upregulation of PG production by celecoxib is causally linked to increased COX-2 expression and thus sensitive toward inhibition of COX-2 activity, concentrations of PG were measured in cell culture media using two different experimental settings.  As further shown in Fig. 6A (PPAR ␥ in nuclei), the nuclear accumulation of PPAR ␥ by celecoxib was completely abrogated by NS-398 in all three cell lines, suggesting that NS-398 confers an inhibition of the celecoxib-induced cytosol-to-nuclear translocation of PPAR ␥ . This fi nding was substantiated by confocal imaging of intracellular PPAR ␥ , which appeared more restricted to nuclear regions in cells treated with celecoxib in the absence of NS-398 ( Fig. 6C ).
On the basis of the time-course experiments ( Fig. 3A-C ) demonstrating COX-2 mRNA to be induced by celecoxib prior to PPAR ␥ mRNA, a possible involvement of COX-2 in the celecoxib-induced PPAR ␥ expression was tested further. However, NS-398 did not reverse the celecoxibinduced increase of total PPAR ␥ levels in A549 and only slightly decreased this response in H460 and H358 ( Fig. 6A , PPAR ␥ in total lysates, middle blots). Likewise, NS-398 did not suppress but rather slightly increased celecoxib-induced total COX-2 protein levels in all cell lines ( Fig. 6A ). and PPAR ␥ . Celecoxib-induced DNA fragmentation and loss of viability were signifi cantly inhibited by knockdown of COX-2 ( Fig. 5A , Table 2 ) and PPAR ␥ ( Fig. 5B and Table 2 ) using siRNA approaches.

Infl uence of NS-398 and GW9662 on the celecoxibmodulated expression and intracellular distribution of COX-2 and PPAR ␥
To elucidate a potential coordinated action of COX-2 and PPAR ␥ , the impact of NS-398 and GW9662 on celecoxib-induced expression of COX-2 and PPAR ␥ in total cell lysates and fractions of nuclei from A549, H460, and H358 cells was investigated.
According to Fig. 6A , B (PPAR ␥ in nuclei), a profound translocation of PPAR ␥ to nuclear regions became evident when cells were treated with celecoxib for the same time periods in which substantial PG accumulation ( Figs. 3D and 4B ), apoptosis ( Fig. 5

DISCUSSION
The present study demonstrates induction of COX-2 expression followed by activation of PPAR ␥ as key events within the proapoptotic action of the selective COX-2 inhibitor celecoxib on human lung cancer cells. The mechanism elicited by celecoxib was shown to include an initial upregulation of COX-2 and PPAR ␥ and a subsequent PPAR ␥ activation by de novo-synthesized, COX-2-derived PGs, eventually leading to apoptosis.
There are several lines of evidence supporting this hitherto unknown antitumorigenic pathway of celecoxib. First, celecoxib at high concentrations (30-50 µM) caused a profound upregulation of COX-2 and PPAR ␥ mRNA and protein expression in three lung cancer cell lines as well as in primary lung cancer cells. Second, long-term treatment of cell lines with celecoxib was shown to result in increases of PGE 2 , PGD 2 , and 15d-PGJ 2 that were sensitive to NS-398, thus indicating a functionally active COX-2 enzyme. Noteworthy, celecoxib also elicited an increased expression of L-PGDS, which may further contribute to the increases of PGD 2 and its dehydration product 15d-PGJ 2 . Third, inhibition of COX-2 and PPAR ␥ by knockdown and smallmolecule approaches was demonstrated to suppress celecoxib-induced apoptotic cell death. Fourth, celecoxibinduced translocation of PPAR ␥ from cytosol to nucleus, an established marker of PPAR ␥ activation (43)(44)(45), was inhibited by NS-398, suggesting that COX-2-dependent PGs generated upon celecoxib treatment confer the observed activation of PPAR ␥ . In line with this notion, exogenously added PGD 2 and 15d-PGJ 2 elicited PPAR ␥ translocation and PPAR ␥ -dependent apoptosis, which is in agreement with other studies demonstrating that anticancerogenic effects of PGD 2 and 15d-PGJ 2 occur via PPAR ␥ ( 27,28,(31)(32)(33). Noteworthy, NS-398 left 15d-PGJ 2 -induced PPAR ␥ translocation observed within a 4 h treatment period virtually unaltered, thus excluding a direct inhibitory effect of the COX-2 inhibitor on PPAR ␥ activation.
Although PPAR ␥ was demonstrated to be involved in COX-2 expression in some reports (46)(47)(48), the data Similar to the effect of NS-398, the PPAR ␥ antagonist GW9662 inhibited celecoxib-induced accumulation of PPAR ␥ in nuclear regions ( Fig. 6B ), confi rming PPAR ␥ activation to be involved in this response. By contrast, GW9662 did not suppress COX-2 levels in nuclear fractions and total cell lysates ( Fig. 6B ).
In control experiments, a potential, not PG-driven activation of PPAR ␥ by celecoxib was addressed following 2 h incubation with celecoxib. As shown in Fig. 3D , none of the analyzed PGs became elevated by celecoxib treatment within this early time frame. According to Western blot analyses of nuclear fractions ( Fig. 6D ), 2 h treatment of cells with celecoxib was not accompanied by a translocation of PPAR ␥ into the nuclei, indicating that celecoxib does not confer direct activation of PPAR ␥ .

Impact of exogenous PGs on nuclear and total levels of PPAR ␥
To further confi rm a link between the celecoxib-induced elevation of COX-2-dependent PGs and subsequent PPAR ␥ activation eventually leading to cancer cell death, the impact of exogenously added PGs on nuclear accumulation of PPAR ␥ and apoptosis was investigated.
According to Fig. 7A , PGD 2 and 15d-PGJ 2 (but not PGE 2 ) induced a nuclear accumulation of PPAR ␥ within 4 h incubation without affecting total PPAR ␥ expression. Upregulation of both nuclear and total PPAR ␥ levels by PGD 2 and 15d-PGJ 2 was observed following longer incubation periods, whereas again PGE 2 was inactive in this respect ( Fig. 7B ).
To exclude the possibility that off-target effects rather than COX-2 inhibition confer the inhibitory action of NS-398 on PPAR ␥ translocation, closing experiments addressed the impact of NS-398 on PPAR ␥ activation elicited by another stimulus. On the basis of the data presented in Fig. 7A-C , 15d-PGJ 2 was chosen as PPAR ␥ activator for this purpose. According to Fig. 7D , the PPAR ␥ accumulation by 15d-PGJ 2 was not impaired by NS-398, thus excluding an unspecifi c action of the COX-2 inhibitor ( Fig. 7D ).  presented for celecoxib do not confi rm this notion. NS-398 did virtually not alter total PPAR ␥ expression in A549 and only faintly suppressed celecoxib-induced PPAR ␥ expression in H460 and H358 cells. Vice versa, inhibition of PPAR ␥ by GW9662 did not infl uence celecoxib-induced COX-2 expression in any cell line. Furthermore, there are reports demonstrating several NSAIDs to elicit direct activation of PPAR ␥ ( 49,50 ). In the case of celecoxib, direct activation of PPAR ␥ was observed in rat mesangial cells ( 42 ). On the other hand, celecoxib failed to elicit such effect in rheumatoid synovial fi broblasts ( 51 ), which is in line with our data from lung cancer cells that did not reveal a direct, COX-2-independent PPAR ␥ activation by celecoxib within a 2 h time frame. In addition, others reported a stimulatory action of celecoxib ( 14,(39)(40)(41) as well as PGD 2 and 15d-PGJ 2 ( 52 ) on PPAR ␥ expression. In agreement with the latter fi nding, we observed an upregulation of PPAR ␥ expression following prolonged exposure of cells to PGD 2 and 15d-PGJ 2 .
The most remarkable content of our study is an induction of COX-2 by a COX-2 inhibitor, leading to the apparently contradictory fi nding that celecoxib under certain conditions is able to antagonize its pharmacologically intended COX-2 inhibitory action. However, a thorough literature search revealed several studies reporting an induction of COX-2 expression by COX inhibitors ( 48,(53)(54)(55)(56)(57)(58)(59), including celecoxib ( 37,38,60 ). Furthermore, celecoxib was shown to enhance PGE 2 release from hematopoietic cancer cells at a concentration of 40 µM, but to exert inhibitory effects at 10 µM ( 61 ). Other studies yielded profound increases of cervical ( 62 ) and fetal plasma PGE 2 levels ( 63 ) in pregnant rabbits following administration of celecoxib.
In the present study, modulation of PG formation by celecoxib was addressed by use of two different experimental protocols. In the fi rst approach, activity assays, including a washout of cells preinduced with celecoxib and a subsequent exogenous addition of arachidonic acid, were performed. The data obtained with this protocol indicate an enzymatically active, celecoxib-induced COX-2 protein whose activity was fully abolished in the presence of 1 µM concentrations of both NS-398 and celecoxib.
In the second approach with incubation protocols comparable to those used for analyses of cytotoxicity, apoptosis, and PPAR ␥ translocation, PG levels were measured in culture media of cells treated with celecoxib in the presence or absence of NS-398. The outcome of this assay may be infl uenced by several aspects (i.e., transcription, translation, and activation) of the COX-2 pathway ( 64 ). In consequence, the induction or inhibition of PGE 2 release observed under these conditions is mainly dependent on the incubation time with celecoxib. Thus, high concentrations of celecoxib conferred inhibition of PGE 2 release following a 2 h incubation period that became obviously counteracted in the continuing time period by a celecoxib-induced expression of COX-2 and L-PGDS. The resulting increase of PGE 2 as well as PGD 2 and 15d-PGJ 2 by celecoxib was sensitive to NS-398, thereby corroborating the data on viability, apoptosis, and PPAR ␥ activation obtained with the same incubation protocol. In view of the fact that NS-398 did not suppress the celecoxib-induced COX-2 expression, it is assumed that celecoxib loses parts of its COX-2-inhibitory function in the presence of prooxidant factors in the media of apoptotic cells, whereas NS-398 maintains its COX-2-inhibitory potency under the same conditions. In fact, a previous investigation encouraged a reclassifi cation of COX inhibitors with regard to their ability to interfere with oxidation state of the enzyme and/or essential radicals in the reaction. The study showed that the COX-2 inhibitory effects of some inhibitors (naproxen, ibuprofen, rofecoxib) were signifi cantly blunted by increasing intracellular hydroperoxide levels, whereas the inhibitory effects of others (diclofenac, indomethacin) were essentially unaffected ( 65 ). Ongoing studies have to evaluate whether celecoxib may also infl uence other parameters involved in PG synthesis, such as cytosolic phospholipase A 2 , which has been demonstrated to be induced by high concentrations of celecoxib (50-200 µM) in Lewis lung carcinoma cells ( 41 ) but which appeared to be downregulated in a murine hepatoma cell line exposed to 200 and 400 µM celecoxib ( 66 ).
In our hands, the proapoptotic effect of celecoxib was not a group effect shared by other COX-2 inhibitors with a diaryl heterocyclic structure. Such unique celecoxib effects are in line with earlier reports that demonstrated that celecoxib but not other COX-2 inhibitors induced apoptosis in rheumatoid synovial fi broblasts ( 51 ) and in human colon cancer cells ( 67 ). The lack of group effect of COX-2 inhibitors on lung tumor cell apoptosis presented here may be due to an intracellular accumulation of celecoxib. In fact, a recent investigation of different tumor cell types incubated with diverse COX-2 inhibitors yielded 5-to 10-fold higher intracellular levels of celecoxib as compared with etoricoxib, valdecoxib, lumiracoxib, and rofecoxib ( 68 ). In a further analysis, evidence was provided for an integration of celecoxib into cellular phospholipid membranes resulting in a disturbance of membrane integrity ( 68 ). Consequently, an accumulation of celecoxib in humans has been suggested as a basis of its diverse actions independent of COX-2 inhibition, despite comparatively low plasma concentrations, which have been reported to reach a maximum of 7.67 µM following single-dose administration of celecoxib at 800 mg to human volunteers ( 1 ).
PPAR ␥ by de novo-synthesized, COX-2-derived PGs. Further studies addressing the impact of celecoxib on these parameters in vivo are suggested to better understand the antitumorigenic action of celecoxib.
Collectively, this is the fi rst study to provide insights into the functional consequence of celecoxib-induced COX-2 expression, and it presents a hitherto unknown proapoptotic mechanism of celecoxib comprising the activation of ## P < 0.01, ### P < 0.001 versus corresponding PGD 2 group; ++ P < 0.01, +++ P < 0.001 versus corresponding 15d-PGJ 2 group; one-way ANOVA plus Bonferroni test.