Cell culture material was obtained from Gibco (Invitrogen, Cergy Pontoise, France). Unless mentioned, chemical products and reagents were obtained from Sigma (France).

Antibodies were purchased from indicated companies: COX-2 (Cayman, #160107), mPGES-1 (Cayman, #160140), actin (Chemicon, #MAB1501R).

15-PGDH inhibitor (CAY10397) was purchased from Cayman (#70130) (Interchim, France).

³H-PGE2 (0.1 μCi/μl) was purchased from Amersham Biosciences.

Immunoblots were quantified using the ImageJ software (NIH, USA).

Every experiment was repeated at least 3 independent times unless otherwise stated.

Statistical analyses were performed using the GraphPad software (San Diego, CA 92130 USA) (Student unpaired t-test, *: p < 0.05, **: p < 0.01).

Patient materials as well as records (diagnosis, age, sex, date of death) were used with confidentiality according to French laws and recommendations of the French National Committee of Ethic. Tumor samples were collected from adult patients after surgical resection at the Department of anatomo-pathology of the Hospital of Nantes over the years 2002-2003. The clinical information of the patients is summarized in additional file 1 table SI, and additional file 2 table SII. Control tissue was obtained from normal colon tissue found at the periphery of the resected tumor.

Cells were washed twice in PBS, then total RNA was isolated using the RNAwiz (Ambion, Austin, TX, USA) according to the manufacturer's instructions with DNAse I treatment. After RNA quantification using the Nano Drop (Nano Drop ND-1000, Thermo Fisher Scientific, Waltham, MA, USA), the quality of the RNA was determined in an Agilent 2100 Bioanalyzer (Agilent, Palto Alto, CA, USA) using the Labchip RNA 6000 kit. A minimum RNA Integrity Number (RIN) value of 8 was required 11 . Total RNA (1 μg) was reverse transcribed in a final volume of 20 μl using the Superscript II kit (Invitrogen, France). Subsequently the cDNA was diluted to a final concentration of 20 ng/μl, for use in Q-PCR.

The PCR reaction contained 40 ng cDNA in a reaction volume of 25 μl, 1× Brilliant II SYBR Green Q-PCR master mix, 200 nM reverse and forward primers and 30 nM Sybr Green. Thermo-cycling conditions were 95°C for 10 min followed by 40 cycles at 95°C for 1 min, 60°C for 45 s and 72°C for 30 s. Gene expression values were normalized to housekeeping gene (GAPDH) and relative expression values were calculated based on the comparative ΔΔCT-method with adherent cells used as a reference for each cell type 12 .

GAPDH: sense primer: 5'-GAAGGTGAAGGTCGGAGTC-3'

antisense primer: 5'-GAAGATGGTGATGGGATTTC-3'

COX-2: sense primer: 5'-CAGCCATACAGCAAATCC-3'

antisense primer: 5'-ATCCTGTCCGGGTACAAT-3'

mPGES-1: sense primer: 5'-AGGAAGACCAGGAAGTGC-3'

antisense primer: 5'-ACGACATGGAGACCATCTAC-3'

MRP4: sense primer: 5'-AAGTGAACAACCTCCAGTTCCA-3'

antisense primer: 5'-CCGGAGCTTTCAGAATTGAC-3'

15-PGDH: sense primer: 5'-AAGCAAAATGGAGGTGAAGGC-3'

antisense primer: 5'-TGGCATTCAGTCTCACACCAC-3'

HCT-116 and HCT-116^Bax-/- cells (described in 13 ) were grown in McCOY's 5A medium containing 10% fetal calf serum, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. SW1116 cells were grown in RPMI medium containing 10% fetal calf serum, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. The cells were transfected by a plasmid encoding for the sequence of mPGES-1 cDNA subcloned into pDEST12.2 vector (Invitrogen), or by the mock plasmid containing no coding sequence 10 . Plasmid DNA (5 μg) was introduced into 10⁶ cells by electroporation (GenePulser, Bio-Rad) using 200 V/cm and 250 μF. Transfected cells were selected and further cultured in a medium containing 1 mg/ml G418. The mock-transfected cells were used as a control for the mPGES-1 transfected cells in the expression and viability experiments.

Microinjection was performed as described by Cartron et al. 14 . PGE₂ was co-injected with a dextran coupled to a fluorochrome (Oregon Green, Molecular Probes). The instantaneous intracellular concentration of compounds achieved by the microinjection is about one tenth of the initial concentration in the injected solution. The percentage of fluorescent cells exhibiting morphological apoptotic features was evaluated every hour following PGE₂ microinjection using an inverted fluorescent microscope (DMIRE2, Leica France).

HCT-116^Bax-/- cells were seeded in a 96-well culture plate the day before experiment. 5 μl ³H-PGE₂ was added to the cells. Inhibitors of 15-PGDH (CAY10397, 15 μM) and MRP4 (ketoprofen, 1 μM 15 ) were added in every other well. After the indicated incubation time (0 min, 30 min and 1 h), cells were rinsed and harvested. The amount of radiolabeled PGE₂ present in the cells was quantified by beta-emission measurement (LS 6500 liquid scintillation counter, Beckman Coulter).

Total cell lysates were carried out with RIPA buffer (PBS, 1% NP-40, 0.5% Na-deoxycholate, 0.1% SDS, proteases inhibitor cocktail (Roche, Meylan, France)) and protein concentrations were measured by the Bradford technique. DEVDase activity was measured using the fluorometric CaspACE Assay System (Promega) and normalized to the sample protein concentration as described previously 10 .

Total cell lysates were obtained with RIPA buffer and separated by SDS-PAGE. Proteins were transferred onto PVDF membranes by semi-dry transfer. Membranes were successively probed with the indicated antibodies and revealed by ECL with peroxidase-coupled secondary antibodies.

SW1116 cells were plated the day before treatment. PGE₂ (10 μM) was added to the culture medium. After 10 min, CAY10397 (15 μM) and ketoprofen (1 μM) were added, and cells were treated for 30 h. Cell death was then assessed by trypan blue staining.

The role played by PGE₂ in cell survival/proliferation or in cell death is still highly debated. Given the observation we made in glioblastoma that mPGES-1 overexpression was correlated to an increased survival of patients, we studied the expression of COX-2 and mPGES-1 in nine human colon cancer samples (Figure 1A). mPGES-1 expression was very inconstant in the tumours, some expressing very high levels of the protein whereas mPGES-1 was hardly detectable in others. COX-2 expression was weakly detected in all the samples. We then measured mPGES-1 and COX-2 transcripts in seven additional human colon cancer samples and in the non cancerous corresponding tissue by qPCR (Figure 1B). Similarly, mPGES-1 expression appeared very inconstant whereas COX-2 was regularly overexpressed in cancer tissue compared to control (p < 0.003 in 5 out of 7 patients), even if the mRNA relative expression level was low (see mPGES-1, left graph, for comparison). Of note, in two out of three tumours overexpressing mPGES-1, COX-2 was largely overexpressed. In our hands, it thus seems that mPGES-1 was not always up-regulated in colon cancers.

We have previously shown in glioblastoma primary cultures that mPGES-1 exogenous overexpression sensitized the cells to apoptosis 10 . We thus considered if this observation could be reproduced in human colon cancer cells in vitro, when isolated from the stroma. We first analysed the expression of mPGES-1 and COX-2 in two model human colon cancer cell lines (HCT-116 and SW1116). As shown in Figure 2A, mPGES-1 expression was hardly detectable in both cell lines (Figure 2A right, lanes 1 and 3). In contrast, COX-2 was abundantly expressed in SW1116 cells, but not in HCT-116 cells. HCT-116 cells have often been presented as COX-2 deficient cells, whereas they rather seem to constitutively express COX-2, unlike non-cancer cells, even at lower level than other cell lines. We induced mPGES-1 overexpression in both cell lines by plasmid transfection. Of note, two rounds of transfection were necessary to induce a steady expression of mPGES-1 in HCT-116 cells. As we had observed earlier in glioblastoma cells, the modulation of mPGES-1 expression had an effect on the expression of COX-2. Surprisingly, COX-2 expression was increased in HCT-116 cells overexpressing mPGES-1 whereas it was not significantly modified in SW1116 transfected cells (Figure 2A left, lanes 2 and 4). We verified that the transfection of mPGES-1 in these two cell lines and in two additional colon cell lines (HCT-8 and HT29) was accompanied by an increase in intracellular PGE₂ concentration (additional file 3 figure S1A). Since the expression of COX-2 was not modified by the transfection in SW1116 cells, we used these cells for the rest of the experiment. SW1116 cells were subjected to UV-B irradiation to induce apoptosis (25 J/cm², 10 min). The specific caspase-3 activity was measured in the cell lysate after 30 hours. As shown in Figure 2B, mPGES-1 overexpression led to a higher caspase-3 activity, attesting to a higher sensitivity of the cells to apoptosis induction. Of note, we measured a constitutive caspase-3 activity in all the four mPGES-1-transfected cells, even if this activity was very low and was not associated to a significant cell death (additional file 3 figure S1B).

The conflicting observations that extracellular PGE₂ promotes colon cancer cells survival and proliferation 3 4 8 while mPGES-1 overexpression increases apoptosis in colon cancer cells (Figure 2B andadditional file 3 figure S1B) prompted us to study the effect of PGE₂ microinjection directly into the cytoplasm of colon cancer cells. The treatment of the four colon cancer cell lines with PGE₂ extracellular concentrations ranging from 0.1 μM to 100 μM induced no more than 10% cell death compared to control cells. Yet HCT-116 cells exhibited the highest variation coefficient in cell death following to PGE₂ extracellular treatment (additional file 4 figure S2). As shown in Figure 3A, PGE₂ microinjection induces a significant cell death within 5 hours in both SW1116 and HCT-116 cell lines, whereas SW1116 cells are only sensitive to the highest concentration injected (left). Of note, the microinjection of a 20 μM PGE₂ solution results in a transient intracellular increase of about 2 μM, which is consistent with the intracellular levels we have previously measured during apoptosis induction (unpublished results). Consistent with the direct effect of PGE₂ on Bax activation we have previously demonstrated, we observed that the cell death induced by PGE₂ in colon cancer cells was dependent on Bax expression. Indeed, Bax-deficient HCT-116^Bax-/- cells appeared resistant to PGE₂ microinjection (Figure 3B).

Since PGE₂ microinjection was able to induce colon cancer cell death, whereas extracellular PGE₂ could not, we tried to force PGE₂ cytoplasmic accumulation by pharmacological means. PGE₂ intracellular concentration is the result of several regulated processes, among which are its synthesis, its enzymatic degradation through the activity of 15-PGDH (15-prostaglandin E₂ dehydrogenase), and its membrane transport, mainly through the export protein MRP4. We thus analysed the level of the transcripts of the synthesis enzymes COX-2 and mPGES-1 and also those of 15-PGDH and MRP4 in both cell lines HCT-116 and SW1116 (Figure 4A). It was noticeable that SW1116 cells expressed a high level of 15-PGDH compared to HCT-116, possibly explaining the significant difference of sensitivity to PGE₂ microinjection (Figure 3A). Ketoprofen was used, at a non-COX-2 inhibiting dose, to block the MRP4-mediated exportation of PGE₂ 15 , while CAY10397 was used to inhibit PGE₂ enzymatic degradation by 15-PGDH. Treating the cells with both drugs caused a significant cytoplasmic accumulation of ³H-PGE₂, as measured in the resistant HCT-116^Bax-/- cells (Figure 4B, black squares), whereas ³H-PGE₂ remained extracellular in the absence of MRP4 inhibition (Figure 4B, open circles). Given this result, we treated SW1116 cells with extracellular PGE₂ and with various combinations of the inhibitors described and measured cell death after 30 hours by Trypan blue exclusion. As shown in Figure 4C, the combination of both inhibitors induced a significant cell death even in the absence of exogenous PGE₂. However, cell death increased from 21 to 38% when PGE₂ was added. A significant, yet slightly lower cell death induction was observed when either CAY10397 or ketoprofen was omitted, indicating that both PGE₂ enzymatic degradation and exportation participate in the survival of these. We verified in SW1116 cells that the treatment applied effectively induced an increase in PGE₂ intracellular concentration and that caspase 3 was activated, testifying apoptotic cell death (additional file 5 figure S3A). Of note, the treatment described above also enabled PGE₂ intracellular accumulation and cell death in the three additional colon cell lines used (additional file 5 figure S3B); besides, we confirmed the singularity of HCT-116 cells compared to the three other cell lines since PGE₂ internalisation was measured even in the absence of the inhibitors, making their sensitivity to extracellular PGE₂ clearer. We also verified, using the HCT-116 cell line and its Bax-deficient counterpart (HCT-116^Bax-/-), that the cell death induced by the combination of extracellular PGE₂ and the inhibitors described above was Bax-dependent (Figure 4D).