The human colorectal adeno-carcinoma-derived cell line HCA7 was a kind gift from Dr Susan Kirkland, (Imperial College, London); cells were mycoplasma tested on receipt (Mycoalert Plus mycoplasma detection kit; Lonza Group Ltd.). The human colorectal adenocarcinoma-derived cell line HT-29 and human rectal adenocarcinoma-derived cell line SW837 were obtained from the American Type Culture Collection (ATCC; cat. nos. HTB-38 and CCL-235, respectively) and mycoplasma tested on receipt. All experiments were performed within 6 passages and cells were routinely characterised (as described below). The human colorectal adenocarcinoma-derived HCT116, HCT15, SW480, SW620, LOVO and LS174T cells, and the specifically rectal cell line SW1463, were all obtained from ATCC (cat. nos. CCL-247, CCL-225, CCL-228, CCL-227, CCL-229, CL-118 and CCL-234, respectively), cells were mycoplasma tested on receipt and experiments were performed within 6 passages of receipt. All cell lines were maintained as previously described (32). Cell lines were grown in DMEM (Gibco; Thermo Fisher Scientific, Inc.), supplemented with 10% FBS, 2 mM glutamine (Gibco; Thermo Fisher Scientific, Inc.), 100 U/ml penicillin and 100 µg/ml streptomycin (Gibco; Thermo Fisher Scientific, Inc.). The human colorectal adenoma-PC/AA/C1, S/AN/C1, S/RG/C2 and transformed adenoma-PC/AA/C1/SB10 cell lines were derived in this laboratory and grown as described previously (33,34). Growth medium was DMEM supplemented with 20% FBS, 1 µg/ml hydrocortisone sodium succinate, 0.2 units/ml insulin, 2 mM glutamine, 100 U/ml penicillin and 100 µg/ml streptomycin. All cell lines were routinely assessed for microbial contamination (including mycoplasma), and molecularly characterised using an inhouse panel of cellular and molecular markers to check that cell lines have not been cross contaminated (every 3-6 months; data not shown). Stocks were securely catalogued and stored, and passage numbers were strictly adhered to prevent phenotypic drift.

Conditioned medium (CM; 24 h) was harvested from BCL-3 siRNA, COX-2 siRNA or negative control transfected HCA7 cells, filtered and used to culture S/RG/C2 adenoma for 72 h (CM was refreshed every 24 h). Add back experiments were carried out using 1 µM dimethyl PGE2 (Sigma-Aldrich; Merck KGaA) to the CM prior to culture.

HCA7 cells were treated with 75 µM NS-398 (Sigma-Aldrich; Merck KGaA), for 24 h, using a dose selective for COX-2 inhibition (35). Both HCA7 and HT-29 cells were treated with 100 ng/ml TNF-α (Insight Biotechnology Ltd.), and SW837 cells were treated with 10 ng/ml IL-1β (Insight Biotechnology Ltd.), both for up to 72 h. The concentration of TNF-α and IL-1β used had been previously optimized (36).

Growth medium from individual flasks was collected and the number of floating cells counted. Attached cell yield was determined by trypsinising and counting the number of viable cells from the same flask. The floating cells are represented as a proportion of the total cell population (floating and attached), and used as a measure of cell death, as described previously (37).

Cells were transfected using Lipofectamine RNAiMAX (Invitrogen, Thermo Fisher Scientific, Inc.), according to the manufacturer's protocol, with small interfering RNAs (siRNAs final concentration 50 nM; GE Healthcare Dharmacon, Inc.) targeting BCL-3 (AGA CAC GCC UCU CCA UAU U) or COX-2, for which four different siRNA sequences were pooled (GGA CUU AUG GGU AAU GU UA, CAU CAA CAC UGC CUC AAU U, GAA AUU UGA CCC AGA ACU A and GAA UUA CCC AGU UUG UUG A) (23). Cells were incubated overnight at 37˚C, before changing the medium. Samples were prepared up to 72 h after transfection.

Total RNA was extracted from cells using TRI-reagent (Sigma-Aldrich; Merck KGaA), an RNeasy mini kit (Qiagen, Inc.) was utilised according to manufacturer's protocol with an additional on-column DNase digestion step (RNase-Free DNase Set; Qiagen, Inc.), prior to complementary (c)DNA synthesis. The RNA concentration of samples was measured using a NanoDrop (Thermo Fisher Scientific, Inc.).

cDNA was synthesised from 2 µg RNA by RT, using the RNA-dependent DNA polymerase, Moloney murine leukaemia virus reverse transcriptase, (Promega Corporation), at 40˚C for 1 h. A second tube, without reverse transcriptase was used as the negative control (no RT). The samples were diluted further giving a final concentration of 10 ng/µl.

Following optimisation of primers and ensuring the annealing temperature provided ~100% amplification efficiency per cycle (data not shown), qPCR was performed as previously described (24), using SYBR Green PCR mix (Qiagen, Inc.) and the following Qiagen QuantiTect primers, at a dilution of 1:10: BCL-3, cat. no. QT00040040; and COX-2, cat. no. QT00040586; with gene expression normalised interchangeably with both housekeeping genes TATA-binding protein (TBP; cat. no. QT00000721) or Hypoxanthine Phosphoribosyl transferase (HPRT; cat. no. QT00059066). The QuantiTect primer assays sets are designed to have annealing temperatures of 55˚C. For PCR, a 40-cycle program was performed (denaturing, 15 sec at 94˚C; annealing, 30 sec at 55˚C; and extension, 30 sec at 72˚C) using a MxPro 3005P Real-time Thermal Cycler (Agilent Technologies, Inc.); samples were amplified in triplicate with one no RT well per condition. Amplification data was analysed using MxPro software version 4.10 (Agilent Technologies, Inc.), using the 2^-ΔΔCq method (38).

Whole-cell lysates were prepared in situ, on ice, using 100 µl 1x lysis buffer (Cell Signaling Technology, Inc.) with the addition of a Protease Inhibitor Cocktail Tablet (Roche Diagnostics) per 10 ml lysis buffer. The cell debris were removed by centrifugation at 1˚C for 10 min, at 18,500 x g. Protein concentration of the cell lysate was determined using a Bio-Rad DC Protein assay kit (Bio-Rad Laboratories, Inc.), according to the manufacturer's protocol. Absorption was measured in duplicate at 750 nm using an iMark Microplate Absorbance Reader (Bio-Rad Laboratories, Inc.). Samples of 100 µg total protein were prepared in a volume of 20 µl (adjusted with distilled water), and 5 µl 5x Laemmli buffer was added to each lysate sample and boiled for 5 min.

Mini-Protean 3 Electrophoresis Cells (Bio-Rad Laboratories, Inc.) were used to cast 9% acrylamide resolving gels, using 1.5 mm spacers. Gels were run at 100 V for ~15 min, allowing the samples to move through the stacking gel, before voltage was increased to 180 V for ~1 h, or until the blue dye front had migrated through the gel. Following separation of the protein samples using SDS-PAGE, the proteins were transferred onto Immobilon-P, a PVDF membrane (EMD Millipore). The gel and membrane were then assembled into a Transblot Cell (Bio-Rad Laboratories, Inc.), and a voltage of 100 V was applied for 1.5 h. The membrane was blocked in 5% (w/v) milk blocking buffer for a minimum of 1 h at room temperature, prior to immunoblotting, as previously described (39), using the following antibodies diluted in 0.5% (w/v) milk dilution buffer: Rabbit polyclonal anti-BCL-3 (1:2,000; cat. no. 23959-1-AP; ProteinTech Group, Inc.), goat polyclonal anti-COX-1 (1:500; cat. no. sc-1752; Santa Cruz Biotechnology, Inc.), goat polyclonal anti-COX-2 (1:500; cat. no. sc-1745; Santa Cruz Biotechnology, Inc.) or rabbit polyclonal anti-15-PGDH (1:2,000; cat. no. ab37148; Abcam). The membranes were incubated overnight at 4˚C, after which they were washed three times 10 min each in Tris-Buffered Saline-0.1% (v/v) Tween-20 (TBS-T). Subsequently, the membranes were incubated with an appropriate horseradish peroxidise-conjugated secondary antibody (1:1,000, anti-mouse IgG, cat. no. A4416; anti-rabbit IgG, cat. no. A6154, 1:30,000; anti-goat IgG, cat. no. A9452; all from Sigma-Aldrich; Merck KGaA) in 0.5% (w/v) milk dilution buffer. After an 1 h incubation at room temperature on a rocker, the membranes were washed three times, 10 min each, in TBS-T, and rinsed once with distilled water prior to visualisation of the signals using LumiGLO Peroxidase Chemiluminescence Substrate (Kirkegaard & Perry Laboratories, Inc.). The signal was detected using x-ray films developed in a Compact X4 Film Processor (XOgraph Imaging Systems Ltd.). The length of time films were exposed was dependent on the strength of the signal (between 1-20 min). Equal loading was confirmed using a monoclonal mouse anti-α-tubulin incubated for 2 h at room temperature (1:10,000; cat. no. T9026; Sigma-Aldrich; Merck KGaA).

Blots were quantified using ImageJ version 1.52p (National Institutes of Health). Densitometry analysis was used to measure the change in intensity of individual western blot bands.

PGE₂ released by cells into culture media was quantified using a PGE₂ enzyme immunoassay solid well kit (Cayman Chemical Company), that used a high-affinity PGE₂ monoclonal antibody for quantification of PGE₂. Medium collected from 70% confluent HCA7 cells was snap frozen in liquid nitrogen (-196˚C) and stored at -70˚C. PGE₂ levels were determined using an immunoassay kit (Cayman Chemical Company), as described previously (40). Cells were pre-treated with TNF-α (100 ng/ml, Insight Biotechnology Ltd.) for 16 h or NS-398 (75 µM, Sigma-Aldrich; Merck KGaA) for 24 h as required. Samples were performed in triplicate.

For statistical analysis, a one sample t-test, a Student's t-test, or a one-way ANOVA with Tukey's post-hoc test was used to compare groups in GraphPad Prism version 7 (GraphPad Software, Inc.). Results are expressed as the mean ± standard error of the mean, where a minimum of three independent experiments were performed. Where a single experiment was performed with multiple technical repeats, the data are presented as the mean of the technical repeats ± standard deviation.

BCL-3 consistently repressed PTGS2 expression in a number of human colorectal cancer cell lines in focused mini-arrays (data not shown). Therefore, to determine whether BCL-3 regulated COX-2/PGE₂ signalling in colorectal cancer, HCA7 human colorectal cancer cells, which endogenously express high levels of COX-2 and high BCL-3 expression (Fig. 1A), were transfected with BCL-3 siRNA and the expression of COX-2 was determined (Fig. 1B). Cells were transfected with BCL-3 siRNA or non-targeting control. To confirm the specificity of the siRNA and to control for off-target effects, BCL-3 expression was silenced using four separate siRNA sequences (A, B, C and D). Transfection with all four siRNA sequences separately resulted in a reduction in the levels of COX-2 protein (Fig. 1B). Sequence A was used in all subsequent experiments. The protein expression levels of both BCL-3 and COX-2 in HCA7 cells were assessed 24, 48 and 72 h post-transfection (Fig. 1C). Western blot analysis showed that the levels of COX-2 protein were reduced following knockdown of BCL-3 at all-time points (Fig. 1C). BCL-3 and COX-2 mRNA levels were evaluated using RT-qPCR. Results are presented as the fold change of the negative siRNA transfected levels (Fig. 1C). Knockdown of BCL-3 expression resulted in >60% reduction in COX-2 mRNA levels, confirming that knockdown of BCL-3 in colorectal cancer cells decreased both the COX-2 mRNA and protein expression levels.

To determine whether BCL-3 regulated COX-2 expression in other colorectal cancer cell lines, HT-29 (which has an intermediate level of endogenous COX-2 protein expression; Fig. 1D) and the SW837 rectal cancer cell lines (low endogenous COX-2 protein expression; Fig. 1E) were assessed. Transfection with BCL-3 siRNA also resulted in downregulation of COX-2 mRNA and protein expression levels in both cell lines.

To confirm that the decrease in COX-2 expression following transfection with BCL-3 siRNA resulted in reduced COX-2 activity, PGE₂ levels were measured in the culture medium. HCA7 cells were used for this experiment as the high endogenous COX-2 expression allowed PGE₂ to be measured directly by ELISA (Fig. 2A). COX-1 and 15-PGDH protein expression were also assessed by western blotting 72 h after BCL-3 siRNA transfection (Fig. 2B) to determine whether expression of any other enzymes which regulate PGE₂ levels were altered. Furthermore, cells were treated with NS-398 to validate that changes in PGE₂ detected in the cell culture medium were due to COX-2 and not COX-1 activity (Fig. 2A). The results showed that PGE₂ production was significantly reduced when BCL-3 expression was knocked down (P=0.048); there was a ~30% decrease in basal PGE₂ production in BCL-3 siRNA transfected cells compared with the negative siRNA control. The levels of PGE₂ were reduced by the COX-2 selective inhibitor NS-398 by >90% (Fig. 2A). Finally, COX-1 and 15-PGDH levels were not altered by transfection with the BCL-3 siRNA (Fig. 2B). Taken together, these findings show that the observed decrease in PGE₂ production in the BCL-3 siRNA transfected cells was due to a loss of COX-2 activity.

Subsequently, whether the regulation of PGE₂ observed by suppression of BCL-3 affected survival of other tumour cells was determined. CM exchange experiments were performed using a PGE₂ sensitive adenoma derived cell line (S/RG/C2). It was previously shown that PGE₂ promoted the growth of S/RG/C2 (41). CM from the BCL-3 siRNA, COX-2 siRNA or negative control transfected HCA7 cells were harvested and used to culture the S/RG/C2 adenoma derived cell lines. The number of HCA7 cells were not affected by BCL-3 knockdown or COX-2 downregulation during the preparation of CM over the 72 h period, (Fig. S1A). Treating S/RG/C2 cells with CM from the BCL-3 siRNA transfected HCA7 cells resulted in a significant increase in floating cells (P=0.015). Cell death was rescued by addition of dimethyl PGE₂ to the CM prior to culture. In addition, the degree of cell death was similar to that observed in cells grown in CM from HCA7 cells transfected with the COX-2 siRNA (Fig. 2C). These findings suggest that the changes in PGE₂ secretion caused by knockdown of BCL-3 affected the survival of PGE₂ sensitive tumour cells.

Together these results showed that knockdown of BCL-3 expression downregulated expression of COX-2 and significantly reduced the quantity of potentially pro-tumorigenic PGE₂ produced by the cancer cells.

COX-2 protein has been identified as an immediate-early response gene that although normally absent from most cells, may be induced at sites of inflammation in response to cytokines such as TNF-α and IL-1β (31). These cytokines have also been shown to increase COX-2 expression in colorectal cancer cells (4,42,43). To determine whether targeting BCL-3 reduced the induction of COX-2 in response to inflammatory cytokines, HCA7 and HT-29 cells were treated with 100 ng/ml TNF-α for up to 72 h (Fig. 3), and SW837 were treated with 10 ng/ml IL-1β (Fig. 4) as SW837 cells are more responsive to IL-1β compared with TNF-α. BCL-3 and COX-2 protein expression were measured over 72 h. BCL-3 protein levels increased after 2 h of treatment with TNF-α (Fig. 3A), reaching peak levels after 4-8 h of treatment and returning to basal levels by 72 h. COX-2 expression followed a similar pattern to BCL-3 induction; COX-2 was induced after 4-8 h, and remained high for 24-48 h when treated with TNF-α, consistent with its regulation with BCL-3 (Fig. 3A). Notably, the increased expression of COX-2 in response to TNF-α treatment (as shown in Fig. 3B at 4 h) was found to be partially blocked following the knockdown of BCL-3 in both HCA7 and HT-29 cell lines. Suppression of BCL-3 also reduced PGE₂ production by the HCA7 cells in response to TNF-α treatment (Fig. 3C). Similar results were observed in the SW837 cells treated with 10 ng/ml IL-1β (Fig. 4). Due to the low endogenous levels of COX-2 protein expression in the SW837 cells, expression of both BCL-3 and COX-2 were also measured using RT-qPCR (Fig. 4B and D). There was a rapid induction of BCL-3 (within 2 h) following treatment with IL-1β, followed by a more prolonged induction of COX-2 (Fig. 4A and B). In addition, the increase in COX-2 expression 2-4 h after IL-1β treatment was partially blocked following knockdown of BCL-3 (Fig. 4C and D). These data show that inhibition of BCL-3 expression not only reduced basal COX-2 expression levels but also signifi-cantly supressed cytokine induction mediated by COX-2. These results suggest that expression of BCL-3 may be an important determinant in the COX-2-mediated response to inflammatory cytokines.