Deoxycholate induces COX-2 expression via Erk1/2-, p38-MAPK and AP-1-dependent mechanisms in esophageal cancer cells

Phorbol 12-myristrate 13-acetate (PMA), acetylsalicidic acid, sodium deoxycholate (DCA) and ursodeoxycholate (UDCA) were from Sigma Chemical Co. (St. Louis, MO). PD 98059 (2'-amino-3'-methoxyflavone), SB 203580 (4-[4'-fluorophenyl]-2 [4'-methylsulfinylphenyl]-5-[4'-pyridyl] imidazole), Z-VAD-FMK (Z-Val-Ala-Asp-CH₂F), Z-DEVD-FMK (Z-Asp(OCH₃)-Glu(OCH₃)-Val-Asp(OCH₃)-FMK), U0126 (1,4-diamino-2,3-dicyano-1,4-bis [2-aminophenylthio] butadiene), phorbol 12,13-dibutyrate (PdBu) and anisomycin were from Calbiochem (LA Jolla, CA). Poly(dI-dC) and T4 polynucleotide kinase were from Amersham Biosciences (Buckinghamshire, UK).

The SKGT4 cell line, derived from a well-differentiated adenocarcinoma arising in Barrett's epithelium of the distal esophagus [34] was generously provided by Dr. David Schrump (Bethesda, MA). The gastric adenocarcinoma cell line AGS was from ECACC (Salisbury, UK). Both cell lines were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, 4 mM L-Glutamine, 50 units/ml penicillin and 50 μg/ml streptomycin (GIBCO BRL, Life Technologies, Paisley, Scotland) at 37°C in a humidified atmosphere containing 5% CO₂.

Control and treated cells were harvested in ice cold phosphate buffered saline (PBS) and nuclear extracts were prepared as described previously [35]. EMSA was performed on nuclear extracts with a double stranded 19-mer oligonucleotides containing the AP-1 binding motif, TGACTCA (12-O-tetradecanoylphorbol-13-acetate response element) as previously described [36]. For supershift analysis, 450 ng of rabbit polyclonal antibodies against c-Jun, Fra-1, and c-Fos or unlabelled oligonucleotides, as a control, were mixed with 4 μg of nuclear extract 30 minutes prior to the binding reaction. Samples were subjected to 4% native polyacrylamide gels. Gels were dried and resulting AP-1 DNA binding complexes visualised by autoradiography.

Affinity precipitation of DNA binding proteins was performed with the optimal binding sequence for AP-1 (5'-CGC TTG ATG AGT CAG CCG GAA-3') (Sigma genosys UK) as previously described [37] with the following modifications: total protein content was standardized to 300 – 400 μg/sample using a protein assay (Bio-Rad), according to the manufacturer's instructions. Equal protein content during affinity precipitation was assessed on acetone-precipitated supernatants.

SKGT4 cells were stimulated and total cell lysates obtained using 25 mM Tris-HCl, pH 7.9, 0.2% NP-40, 15 mM NaCl, 1 mM sodium fluoride, 5% glycerol (v/v), 0.05 mM EDTA, 1 mM Na₃VO₄ and 1 mM PMSF and 10 μg leupeptin (Sigma) and incubating on ice for 20 minutes. Cell nuclei and debris were eliminated by centrifugation at 10,000 × g for 10 min. The total protein content per sample was standardized to 50–100 μg as above.

Equal amounts of proteins were separated on a 10% SDS-polyacrylamide gel and transferred onto a PVDF membrane (Millipore Corp., Bedford, MA). Membranes were incubated with specific antibodies against Erk1/2, p38 and JNK (at a dilution of 1:1000) (Santa Cruz, CA) or their corresponding phosphorylated forms Erk1/2 (p-Tyr204-Erk1/2), JNK (p-Thr-183/Tyr-185-JNK) (Santa Cruz, CA) and p38 (p-Thr180/Tyr182-p38) (at a dilution of 1:50) (New England Biolabs, Hertfordshire, UK) overnight at 4°C. Antibodies against COX-2 (Cayman chemicals, Alexis Corp, UK) and PARP (Biosource International UK) (at a dilution of 1:100) were also used and incubated overnight at 4°C. Membranes were then incubated with corresponding secondary horseradish peroxidase-conjugated antibodies (Dako, Bucks, UK) (at a dilution of 1:2000) for 1 h at room temperature. Specific immunocomplexes were visualised using the ECL detection system (Amersham Biosciences, Buckinghamshire, UK). For sequential detection, membranes were stripped in 100 mM 2-Mercaptoethanol, 2% SDS, and 62.5 mM Tris pH 6.8 for 30–45 min at 50°C.

SKGT4 cells (2.5 × 10⁴) were plated in a flat-bottomed micro-titre plate and incubated for 24 hr at 37°C and 5% CO₂. Cells were incubated either with increasing concentrations of DCA (0 – 500 μM) or over a period of 1 – 24 hr with 300 μM DCA. Following stimulation, MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide) (Promega Inc., Madison, WI.) was added and cells were further incubated for 1 to 3 hr. Absorbance was measured at 490 nm. Viability is expressed as the percentage of cells remaining in cultures treated with bile acids relative to untreated controls.

DCA-induced toxicity was quantified using the cell death detection kit (Roche diagnostics, Penzberg, Germany) according to the manufacturer's standard protocols. Absorbance was measured at 405 nm using an ELISA plate reader.

DCA regulates gene transcription through AP-1 activation in colonic cells [28]. We examined the possible link between DCA, AP-1 in esophageal adenocarcinoma SKGT4 cells, a cell line derived from a well-differentiated adenocarcinoma arising in Barrett's epithelium of the distal esophagus [34]. DCA is present at micromolar concentrations (0–300 μM) in esophageal aspirates [18], doses which have been previously shown to be optimal for DCA signaling. SKGT4 cells were exposed to 300 μM DCA from 1 – 24 hr and then analyzed for AP-1 DNA binding activity by EMSA. PMA treated AGS cells were used as a positive control. DCA induces increased AP-1 DNA binding activity as compared to unstimulated cells, in a time dependent manner (Figure 1A). DCA-induced AP-1 activation is biphasic, being markedly induced after 1 hr of stimulation, peaking again at 6 hr and returning to basal levels at the later time points, 12 hr and 24 hr (Figure 1A). We have also demonstrated a similar profile of AP-1 activation in another esophageal adenocarcinoma line OE-33 (data not shown).

AP-1 dimer composition is crucial in determining the induction of specific target genes and consequent cellular responses. EMSA was used to determine which Jun and Fos proteins form part of the DCA-induced AP-1 complex. Nuclear extracts from SKGT4 cells stimulated with 300 μM DCA, were incubated with antibodies against c-Fos, Fra-1 and c-Jun prior to incubation with the radiolabelled AP-1 probe. Blocking antibodies that prevent the binding of the corresponding transcription factor rather than causing a supershift were used [38]. The specificity of the DCA-induced AP-1 complex was further verified by the addition of unlabeled AP-1 oligonucleotide. DCA induces clear AP-1 DNA binding activity, which is reduced by addition of antibodies against Fra-1 and c-Jun but not by the anti-c-Fos antibody (Figure 1B). Addition of a pool of all the antibodies completely abrogated the formation of the DCA-induced AP-1 complex. These data suggest that Fra-1 and c-Jun, but not c-Fos, are members of the DCA-induced AP-1 complex.

To further characterize the composition of the DCA-induced AP-1 complex, total cell lysates were prepared from SKGT4 cells treated with 300 μM DCA for 1 – 6 hr and used for DNA affinity precipitation assays with the AP-1 consensus sequence (5'-CGC TTG ATG AGT CAG CCG GAA-3'). Only active forms of the Jun and Fos proteins are able to bind to this oligonucleotide and can be therefore affinity purified and detected by Western blot analysis with specific antibodies. For these assays, we concentrated on c-Jun and Fra-1 as suggested by EMSA and also included JunB and JunD, as they have been shown to respectively counteract or enhance c-Jun activity [21]. These experiments show that DCA stimulates a time dependent increase in Fra-1, JunB and c-Jun DNA binding activity (Figure 1C). No activation of JunD was observed at any stimulation time (data not shown). DCA induces strong Fra-1 binding activity after 1 hr, which is sustained for at least 6 hr of stimulation. Fra-1 is detected as two bands with distinct electrophoretic mobility: a slower migrating, more prominent band and a fainter faster migrating band (Figure 1C). After prolonged stimulation, the slower species is stabilized while the faster migrating species is no longer detected (Figure 1C). These two electrophoretic mobility forms of Fra-1, which most likely correspond to different phosphorylation states, have been previously reported in other cell types [39]. Similarly, DCA-induced JunB activity is clear at 1 hr and remains elevated for up to 6 hr of stimulation (Figure 1C). On the other hand, DCA induces a weak and more transient activation of c-Jun, which is maximal at 4 hr and is consistently weak and even negligible at 6 hr (Figure 1C). These data indicate that DCA induces AP-1 complexes composed of Fra-1, JunB and c-Jun at early stages of stimulation, but only of Fra-1 and JunB at 6 hrs. Fos and Jun proteins can form heterodimers while the only the members of the Jun family can homodimerise [23]. Therefore, the possible types of early induced complexes are c-Jun/c-Jun, c-Jun/JunB, c-Jun/Fra-1, JunB/Fra-1 or JunB/JunB, while only the latter two would be present at later stages.

Induction of AP-1 DNA binding activity can be achieved by activation of pre-existing Fos/Jun proteins or through induction of de novo protein expression [40]. To differentiate between these two possibilities, the protein expression levels of these molecules was assessed by Western blotting in SKGT4 cells following DCA treatment (300 μM) for 1 – 6 hr. SKGT4 cells express basal levels of Fra-1, JunB and c-Jun (Figure 1D). The expression levels of all three proteins are further enhanced by DCA treatment. An increase in Fra-1 and JunB protein levels is observed within 1 hour of stimulation and remains constant for up to 6 hours (Figure 1D). DCA induces a lesser increase in c-Jun protein expression as compared to Fra-1 and JunB, which decreases by 6 hours (Figure 1D). JunB is detected as three distinct bands while c-Jun is generally found as a doublet. Multiple electrophoretic mobility forms of JunB and c-Jun attributed to different phosphorylation status have previously been reported [40]. The presence of basal expression levels together with the matching kinetics of enhanced protein expression and those of DNA binding activity for Fra-1, JunB and c-Jun, suggest that DCA induces AP-1 DNA binding activity through activation of pre-existing molecules as well as either induction of de novo protein synthesis or increased protein stability. Sustained activation of AP-1 components has been associated with oncogenic transformation [41]. As c-Jun is only transiently activated by DCA, we concentrated on Fra-1 and JunB in subsequent experiments.

Increased concentrations of bile acids (> 200 μM) associated with higher severity of disease, have been observed in esophageal aspirates in patients with erosive esophagitis and Barrett's esophagus [16, 18]. The contribution of various doses of DCA (0 – 500 μM) following prolonged stimulation (6 hours) was examined on Fra-1 and JunB DNA binding activity in SKGT4 cells using the affinity precipitation assay. DCA induces a dose dependent increase in the DNA binding activity of Fra-1 and JunB at 6 hours of stimulation (Figure 2). Low concentrations of DCA stimulate a modest increase while stronger activation of Fra-1 and JunB is detected at and above 300 μM DCA (Figure 2). These data show that strong activation of AP-1 is achieved by DCA at the concentrations observed in vivo in patients with Barrett's esophagus [18].

AP-1 activation is mainly regulated by MAPKs. We therefore examined the ability of DCA to activate Erk1/2, p38 and JNK in SKGT4 cells using Western blot analysis with specific antibodies that recognize the active phosphorylated forms of these proteins: Erk1/2 (p-Tyr204-Erk1/2), p38 (p-Thr180/Tyr182-p38) and JNK (p-Thr-183/Tyr-185-JNK). The well known Erk1/2, p38 and JNK activators phorbol 12, 13-dibutyrate (PdBu) and anisomycin (ANIS) were respectively used as positive controls (Figure 3A–E). Time course analyses show that 300 μM DCA induces sustained activation of Erk1/2 and p38, which are detected as early as 15 minutes and persist for at least 6 hr of stimulation (Figures 3A–C), while JNK is not activated at any time tested (Figure 3E). DCA does not influence the protein expression levels of any of these MAPKs (Figure 3A–E lower panels).

Mek1/2 and MKK3/6 are the respective upstream activators of Erk1/2 and p38 [42, 43]. The PD98059 and SB203580 compounds, known specific inhibitors of Mek1/2 and MKK3/6 were used to verify activation of Erk1/2 and p38 in response to DCA. SKGT4 cells were incubated with 50 μM PD98059 or 2 μM of the SB203580 for 30 minutes prior to the addition of 300 μM DCA. PD98059 completely abolishes basal, PdBu and DCA-induced Erk1/2 activity at all the time points tested (Figure 3B). Similarly, the SB203580 compound inhibits the activation of p38 in response to DCA and to anisomycin at all tested time points (Figure 3D). These data show that DCA activates the MAPKs Erk1/2 and p38 without affecting their protein expression levels, but it is unable to regulate JNK activation or protein expression.

The pharmacological inhibitors PD98059 and SB203580 were respectively used to corroborate the contribution of the Raf-Mek1/2-Erk1/2 and the MKK3/6-p38 pathways in DCA-induced DNA binding of Fra-1 and JunB. SKGT4 cells were pre-treated with 10 μM PD98059 or 2 μM SB203580 for 30 min prior to stimulation with 300 μM DCA for 6 hr and DNA affinity precipitation assays were performed. Pre-treatment of SKGT4 cells with 10 μM PD98059 impairs and diminishes DCA-induced activation of Fra-1 and JunB, respectively (Figure 3F). The SB203580 compound completely abolishes DCA-induced Fra-1 DNA binding while having no effect on DCA-induced JunB DNA binding (Figure 3F). These data indicate that both Raf-Mek1/2-Erk1/2 and MKK3/6-p38 are involved in DCA-induced Fra-1 activation, while only Raf-Mek1/2-Erk1/2 is upstream of JunB activation.

Bile acids, in particular DCA, inhibit proliferation and are potent inducers of apoptosis in several cell types including, hepatocytes and colonic cells [10, 13, 28]. Activation of AP-1 can have both anti-apoptotic and pro-apoptotic functions depending on the cellular context [20]. Since DCA induces sustained (6 hour) activation of AP-1 in SKGT4 cells (Figure 1A), its possible contribution to deregulated cell survival and apoptosis was examined. SKGT4 cells were stimulated with 300 μM DCA or 300 μM ursodeoxycholic acid (UDCA) for 0 – 6 hr. We have previously shown that UDCA, in contrast to DCA, does not induce AP-1 transcription factor activation in colon cancer cells. In fact, it inhibits interleukin-1 beta and deoxycholic acid-induced activation of NF-kappaB and AP-1 in these cells [44]. Cell proliferation was assessed using the MTT assay. DCA induces a dose and time dependent decrease in cellular proliferation, which is initially observed within the first hour of treatment, remains at similar levels up to 8 hr and is more pronounced at 12 and 24 hr (Figure 4A). This decrease is clear at 300 μM DCA and higher concentrations, being statistically significant (p < 0.05) at 400 – 500 μM (Figure 4B). Dramatic morphological changes indicative of apoptosis are also observed at 6 hr of DCA treatment at concentrations in excess of 300 μM (data not shown). In comparison, cells stimulated with UDCA show identical proliferation patterns and morphology as compared to untreated cells at all times and concentrations tested.

DNA fragmentation and PARP cleavage, two of the hallmarks of apoptosis, were respectively assessed by quantifying cytoplasmic histone-associated DNA fragments (mono and oligonucleosomes) by ELISA and Western blotting using a specific antibody that recognizes the 85 kDa cleaved PARP fragment. DCA dose response and kinetic studies showed that a low level of PARP cleavage was detected at 6 hr post DCA treatment and persisted for up to 24 hr (Figure 4C). This effect is dose-dependent, being observed at 300 – 500 μM DCA but not at lower concentrations (Figure 4D). Similarly, DCA induced a 1.5-fold increase in DNA fragmentation after 6 hr, which increased after 24 hr (3-fold) (Figure 4E). DNA fragmentation with low concentrations of DCA (100 – 200 μM) was similar to resting cells, while increasing concentrations (300 – 500 μM) resulted in a steady rise in DNA damage (Figure 4F). Taken together, these data show that DCA induces a reduction in cell proliferation, which is accompanied by low levels of apoptosis. These effects are sustained and dose dependent, being observed at high DCA concentrations similar to those found in patients with erosive esophagitis and Barrett's esophagus [16, 18].

PARP cleavage can occur via caspase-dependent and independent mechanisms [45]. The broad-spectrum caspase inhibitor, Z-Val-Ala-Asp-CH₂F (Z-VAD-FMK), and the specific caspase-3 inhibitor, Z-Asp(OCH₃)-Glu(OCH₃)-Val-Asp(OCH₃)-FMK (Z-DEVD-FMK), were employed to assess the role of caspases in DCA-induced PARP cleavage. SKGT4 cells were pretreated for 1 hr with 50 μM of either Z-VAD-FMK or Z-DEVD-FMK and stimulated with 400 μM DCA, a concentration which induced significant levels of PARP cleavage (Figure 4) for 6 hr. Unstimulated SKGT4 cells showed negligible levels of PARP cleavage and DNA fragmentation. Both Z-VAD-FMK and Z-DEVD-FMK completely abolished DCA-induced PARP cleavage while partially inhibiting DNA fragmentation (Figures 5A and 5B). These data indicate that DCA-induced PARP cleavage is caspase-3 dependent, while DNA fragmentation is only partially dependent on this pathway.

Interestingly, the levels of DCA-induced PARP-cleavage plateau and do not increase progressively. This suggests that a compensatory survival mechanism might be concomitantly regulated by DCA. Enhanced protein expression of COX-2 has been correlated with cellular proliferation and resistance to apoptosis in various cell types [13, 19]. Therefore the induction of COX-2 protein expression by DCA in SKGT4 cells was examined using Western blot analysis. COX-2 is not expressed in unstimulated cells, but it is readily induced after 4 hr of DCA stimulation (Figure 6A). Maximal induction is achieved at 6 hours with 300 μM DCA (Figure 6A). In agreement with previous reports, COX-1 protein is not constitutively expressed in this cell line.

COX-2 protein expression can be regulated at transcriptional and posttranscriptional levels by MAPKs and by AP-1 through binding to the CREB site in the COX-2 gene promoter in various cell types [29, 32]. Since DCA induces AP-1 activity through the activation of Erk1/2 and p38 (Figure 6B), we explored the involvement of these pathways in the regulation of COX-2 protein expression in our system. SKGT4 cells were pre-treated with 10 μM PD98059, 2 μM SB203580 or 1 μM Go6976 for 30 minutes prior to addition of 300 μM DCA for 6 hr. (GO6976 was used as a positive control in this experiment as it has previously been shown to inhibit COX-2 expression [46]. The induction of COX-2 protein expression in response to DCA was strongly inhibited by all three compounds (Figure 6B). These results demonstrate that the Erk1/2 and p38 pathways are not only responsible for DCA-induced Fra-1 and JunB activation but also for induction of COX-2 protein expression.

The kinetics of COX-2 protein induction in response to DCA correlates with those of PARP cleavage and DNA fragmentation (Figures 4 and 6A). To examine the possible anti-apoptotic role of COX-2 in our system, the induction of COX-2 expression was inhibited using acetylsalicylic acid (ASA) and the levels of apoptosis assessed. Treatment of SKGT4 cells with 5 mM ASA for 30 min prior to the addition of 400 μM DCA resulted in a dramatic reduction in COX-2 expression in response to DCA (Figure 6C). DCA induced clear PARP cleavage in comparison to untreated cells, an effect that was enhanced by pre-treatment with ASA (Figure 6D). Levels of DNA fragmentation were similar in unstimulated cells and cells treated with ASA alone. DCA induced a four-fold increase in DNA fragmentation that was further increased by pre-treatment with ASA (Figure 6E). These data show that inhibition of COX-2 expression readily enhances the apoptotic markers induced upon DCA exposure, confirming a potential anti-apoptotic role of COX-2 in this system.