Background
Colorectal cancer (CRC) remains a leading cause of cancer death, with worldwide 1 million new cases each year and as many as half a million cancer deaths annually [
1]. Cyclooxygenase-2 (COX-2) expression is increased in the majority of colorectal tumours [
2] and this induction is associated with advanced tumour stage and correlates with poor clinical outcomes [
3]. Non-steroidal anti-inflammatory drugs (NSAIDs), which inhibit COX activity, show anti-neoplastic effects
in-vitro [
4,
5] and human studies have demonstrated their use to be associated with a reduced incidence of colorectal neoplasia [
6,
7]. While more recent studies have confirmed the chemo-preventive activity of COX-2 selective NSAIDs [
8‐
10], it is also clear that long term therapy with COX-2 inhibitors is associated with an unacceptable increase in the risk of cardiovascular events [
9,
11].
The anti-neoplastic properties of NSAIDs result from the inhibition of prostaglandin generation, particularly prostaglandin E
2 (PGE
2), the most abundant
in-vivo product of COX-2 activity in colorectal cancer cells [
12,
13]. The biological activity of PGE
2 is mediated by binding to cell surface receptors. There are four subtypes of EP receptor (EP1, EP2, EP3, and EP4) with the majority localised to the plasma membrane. The binding of prostaglandins to cell surface receptors triggers changes in second messengers [
14].
PGE
2 modulates processes fundamental to tumour cell survival such as enhanced proliferation and resistance to apoptosis [
4,
15‐
17], however, the precise molecular mechanisms remain unclear. There is therefore a strong rationale to seek a more profound understanding of the downstream targets of COX-2 activity. Selective COX-2 inhibitors have shown promise as chemo-preventive agents [
18], but their adverse cardiovascular effects have undermined their suitability for long term use [
9,
11]. Renewed attention must now therefore focus on the altered signalling occurring downstream of COX-2 in cancer as a source for new refined therapeutic targets.
Methods
Cell culture
HT-29 cells were purchased from the ATCC (Rockville, MD) and maintained in McCoy's 5 A medium containing 1.5 mM L-glutamine, 10% FBS, penicillin 100 U/ml and streptomycin 100 μg/ml. HCA7 cells were kindly donated by Susan Kirkland (ICRF, London, UK) and were cultured in DMEM with 10% FBS, supplemented with 1 mM sodium pyruvate and 100 μg/ml kanamycin. SC236, a selective COX-2 inhibitor was a gift from Dr. Peter Isakson (Searle, Skokie, IL). PGE
2 was purchased from Cayman (St. Louis, MO). L-161982 (EP4A), a selective antagonist of the EP4 receptor was a kind gift of Merck Frosst, Canada [
19]. PD153035 (EGFR tyrosine kinase inhibitor) was purchased from Calbiochem (La Jolla, CA). Wortmannin was purchased from Sigma Aldrich (Dublin, Ireland).
Quantitative RT-PCR
Total RNA was isolated from cells and tissue following homogenisation in RNA lysis buffer (Qiagen Ltd. GmbH, Germany) supplemented with 1% β-mercaptoethanol. Extraction was performed using RNeasy™ Mini Kits (Qiagen Ltd. GmbH, Germany). Total RNA (1 μg) was reverse transcribed using Moloney Murine Leukaemia Virus (MMLV) reverse transcriptase (Promega, Madison, WI) according to the manufacturer's instructions. Gene expression was quantified by RT-PCR using SYBR Green Universal Master Mix (Roche Diagnostics Corp., Indianapolis, IN). Reactions were carried out in a 96 well format in the ABI 7700 Sequence Detector (Perkin Elmer/Applied Biosystems, UK). Results were then normalized to 18S rRNA amplified from the same cDNA mix. Sequences of the primer pairs used are listed below.
EP1- F ATG GTG GGC CAG CTT GTC
EP1- R GCC ACC AAC ACC AGC ATT G
EP2- F TGC CTT TCA CGA TTT TTG CA
EP2- R TTA ATT GAT AAA AAC CTA AGA GCT TGG A
EP3- F TCT CCG CTC CTG ATA ATG ATG TT
EP3- R TCT GCT TCT CCG TGT GTG TCT T
EP4-F CGA CCT TCT ACA CGC TGG TAT G
EP4-R CCG GGC TCA CCA ACA AAG T
Amphireg-F CTC GGG AGC CGA CTA TGA CTA
Amphireg-R GCT TAA CTA CCT GTT CAA CTC TGA CTG A
CyclinD1-F CTG GAG GTC TGC GAG GAA CA
CyclinD1-R TGC AGG CGG CTC TTT TTC
CDK4-F TGT TGT CCG GCT GAT GGA
CDK4-R AAA CAC AGG GTT ACC TTG ATC TC
CDK6-F CAA CTA GGA AAA ATC TTG GAC GTG AT
CDK6-R TTG GTT GGG CAG ATT TTG AAT
p21-F GCA GAC CAG CAT GAC AGA TTT CTA
p21-R GCG GAT TAG GGC TTC CTC TT
p27-F CCT GCA ACC GAC GAT TCT TC
p27-R TCT TAA TTC GAG CTG TTT ACG TTT GA
Immunohistochemical staining for COX-2 and amphiregulin
Samples of formalin fixed, paraffin embedded tissue were deparaffinised and rehydrated in Xylene and Methanol. Detection of COX-2. Endogenous peroxidase activity was quenched with 0.3% H2O2 in Methanol. Specimens were blocked in 1.5% normal serum and then incubated with antibody to COX-2 (Rabbit polyclonal anti-human COX-2, Cayman Chemical Co.) diluted 1/200, followed by secondary antibody and ABC complex from Vectastain Elite kit (Vector Laboratories, Burlingame, CA). Sections were exposed to diaminobenzidine (DAB) (Sigma Aldrich, Dublin, Ireland) and counterstained with hematoxylin (Sigma Aldrich, Ireland) and mounted using DPX (BDH, Poole, UK). Detection of Amphiregulin. Antigen retrieval was performed by heating slides in 10 mM citrate buffer (pH6.0) for 4 minutes in a pressure cooker. Blocking was performed with goat serum for 30 minutes. Slides were then incubated with primary antibody to amphiregulin (rabbit polyclonal to Amphiregulin by Abcam, Cambridge, MA) diluted 1/200 in ChemMate™ antibody diluent (DakoCytomation, Galway, Ireland) for 1 hour and staining completed using the ChemMate™ DAKO Envision™ detection kit (DakoCytomation, Galway, Ireland) visualisation, counterstaining and mounting were performed as outlined above.
Prostaglandin analysis and cAMP detection by enzyme immunoassay
Competitive enzyme immunoassay (EIA) was used to determine PGE2 levels in culture media in 96-well format by Assay Designs (Ann Arbor, MI, USA). Prostaglandin concentration was calculated using the optical density of the samples in relation to a standard curve generated by dilutions of a standard provided. Quantitative determination of cAMP concentration in cell lysates was performed using the cAMP (low pH) immunoassay by R and D Systems (Abingdon, Oxon, UK). Adherent cells were lysed in 0.1 N HCl for 10 min at 37°C and supernatants were assayed according to manufacturer's instructions in a 96 well plate. cAMP concentrations were calculated using a similar standard curve method.
Cell proliferation (BrdU incorporation)
Cell proliferation assays were carried out using the BrdU (colorimetric) cell proliferation ELISA (Roche Applied Science, Dublin, Ireland) according to manufacturer's instructions. Cells were seeded at 5 × 103/well in 96 well plates and following overnight serum starvation were treated with vehicle or drug. Amphiregulin neutralisation was with anti-human neutralising AR antibody (AR-ab) from Stratech Scientific (Soham, Cambridgeshire, UK).
Cell cycle analysis
HT-29 cells were seeded at a density of 1 × 106/well in 6 well plates and treated with drug or vehicle for 24 hours. Adherent cells were suspended using trypsin-EDTA for 3–5 min at 37°C and were fixed overnight in 75% ethanol at 4°C then washed and resuspended in a solution of PBS containing 0.1% Triton X-100, 0.05 mg/ml of DNase-free RNase and a 50 μg/ml propidium iodide (Molecular Probes, Leiden, NL) in the dark for 30 min. Cells were resuspended in PBS prior to analysis in a FACScalibur flow cytometer (Becton Dickinson, Oxford, UK) with measurement of fluorescence emission at >575 nm (FL3). Analysis of cell cycle distribution (DNA index) was performed using CellQuest™ (Becton Dickinson, Oxford, UK).
Tumour collection
The protocol was approved by the Ethics (Medical Research) Committee of Beaumont Hospital, Dublin and all patients provided written, informed consent. Samples of colorectal tumour/normal were obtained from patients at the time of surgery and were immediately placed in RNAlater solution (Qiagen GmbH, Germany) or fixed in 10% formalin.
Statistical analysis
A one-way analysis of variance (ANOVA) was used to examine overall differences between multiple groups, with Bonferroni multiple comparisons test. Two tailed paired students T-test was used to compare the means of paired samples. The statistical packages GraphPad InStat and GraphPad Prism were used. Statistical significance was set at a P value of less than 0.05, whereas a P value less than 0.005 was considered highly significant.
Discussion
COX-2 expression in colorectal tumours is biologically and clinically important [
3,
16] and PGE
2 is the major product of COX-2 activity in cancer cells [
12,
13]. Four subtypes of membrane PGE
2 receptor have been characterised (EP1–4), although the relative contribution of each of these to key signalling events in cancer has not been fully elucidated. We show that all of the EP receptor subtypes are expressed in human colorectal cancers and that EP4 receptor is predominant. We also demonstrate that the HT-29 cells share this relative distribution of receptors, validating its use as an
in-vitro model.
Animal studies demonstrate that all four EP receptors are expressed in azoxymethane (AOM) induced tumours in rats [
23]. Forced expression of COX-2 in murine mammary epithelial cells (with generation of PGE
2 as the principal product) is associated with induction of EP 1, 2 and 4 receptors and down regulation of EP3 [
24]. Knockout mice deficient in all four subtypes of EP receptor and with deletions of the DP, FP, IP or TP receptors have been generated. The formation of aberrant crypt foci (ACF) following AOM treatment was only different in animals with deletions of the EP1 [
25] and the EP4 receptor [
26]. While no difference in ACF formation with deletion of the EP2 receptor was detected; decreased polyp formation has been observed with deletion of the EP2 receptor in Apc
Δ716 mice [
27].
There are limited data on the relative expression of these receptors in CRC in humans. A down-regulation of the EP3 receptor has recently been reported in human colorectal tumours [
28]. Paradoxically, a study combining immunocytochemistry and in-situ hybridisation showed that EP3 and EP4 were the major receptors expressed (in association with COX-2 induction) in adenomatous polyps in patients with FAP [
29]. A further study failed to demonstrate EP4 receptor or COX-2 mRNA induction in tumour specimens [
30]. However, these observations are at variance with our findings and those of others [
31]. EP4 receptor signalling modulates a tumourigenic phenotype in cancer cell lines and promotes metastatic potential
in-vivo [
31‐
33]. It has also also been demonstrated as important in the pro-neoplastic effects of PGE
2 in a range of other human cancers; notably breast cancer where EP4 has been related to mediation of proliferation, invasion and metastasis [
34,
35]. Given its' relative abundance and functional activity, it seems reasonable to conclude that EP4 receptor mediates at least some of the important pro-neoplastic effects of PGE
2.
Despite the ability of PGE
2 to stimulate cancer cell growth [
16], early data suggested that COX-2 over-expression in intestinal epithelial cells was associated with a paradoxical G
1 delay [
36]. Subsequent data suggest this occurs via prostaglandin independent mechanisms [
37], perhaps representing an artefact of ectopic COX-2 expression. G
0/G
1 cell cycle arrest in cancer cells associated COX-2 inhibition has also been noted [
38] and seems more plausible given the growth inhibitory effects of NSAIDs. We confirm the observation of G
0/G
1 arrest with COX-2 inhibitor treatment and demonstrate that the effect is PGE
2 dependent. We observe that this effect is also produced by the EP4 receptor using a selective receptor antagonist (which shows similar effects on cellular cAMP concentration). Our observations are consistent with previous reports of modulation of cell growth in colon cancer cells through EP4 and of changes in susceptibility to apoptosis
via EP4 receptor activation [
16,
39,
40].
p21
WAF1/CIP1 is a cyclin dependent kinase inhibitor which indirectly regulates pRb phosphorylation and thus the G
1 to S phase transition. Induction of p21
WAF1/CIP1 expression has been described in colon cancer cells following treatment with COX-2 selective inhibitors [
38,
41,
42] and recent observations from other disease models suggest this is truly a prostanoid dependent event [
43]. We demonstrate that selective induction of p21
WAF1/CIP1 expression is associated with EP4 receptor mediated cell cycle arrest. We also note repression of p21
WAF1/CIP1 expression in colorectal tumours (the majority of which express COX-2) samples in public expression datasets (Additional File
1). p21
WAF1/CIP1 is one of the few genes which shows consistent induction in expression in the rectal mucosa of patients treated with sulindac and deletion of p21
WAF1/CIP1 in a mouse model abolished the ability of sulindac to inhibit Apc-initiated tumourigenesis [
44], observations which reinforce the hypothesis that p21
WAF1/CIP1 acts as a possible downstream effector of COX-2/PGE
2/EP4 activity in CRC.
The EP4 receptor generates intracellular cyclic AMP (cAMP) via coupling to Gs proteins leading to activation of protein kinase A (PKA), phosphorylation of cAMP-response element binding protein (CREB) and PKA-dependent activation of extracellular signal-related kinase (ERK)[
32]. However, in contrast to EP2 receptors (which also increase cAMP), EP4 receptors also activate phosphatidylinositol 3-kinase (PI3K) dependent signalling [
45]. We did not observe p21
WAF1/CIP1 induction in HT-29 cells treated with the PI3K inhibitor wortmanin, however, p21
WAF1/CIP1induction was seen with an EGFR tyrosine kinase inhibitor. PGE
2 transactivates the epidermal growth factor receptor (EGFR) in HT-29 cells through c-Src-mediated release of the EGFR ligands [
46]. This therefore seems the likely mechanism for PGE
2/EP4 mediated changes in p21
WAF1/CIP1 expression.
To further clarify the role of EGFR transactivation, we focussed on the EGFR ligand amphiregulin (AR). AR expression in colon cancer cells (in culture) is increased by PGE
2 via a cAMP/PKA dependent pathway [
47], an effect therefore mediated through via EP2 or EP4 receptors. AR is the major EGFR ligand produced by HCA7 cells, where it acts as an autocrine growth factor [
21]. L-161984 (EP4A) has also recently been shown to inhibit HCA7 proliferation [
48]. We demonstrate the ability of COX-2 inhibition and AR neutralisation to inhibit HCA7 cell proliferation with an additive effect seen with a combination of both. This supports observations of the ability of PGE
2 to synergistically enhance EGFR receptor tyrosine kinase signalling [
49] and suggests a novel therapeutic approach. It has been recognised that EGF signalling may be important in sustaining elevated COX-2 expression[
50], suggesting a positive feedback loop re-inforcing the increased activity of both pathways. Combined inhibition of COX-2 and EGFR may be a rational means to attempt to break this cycle and has shown promise in animal models [
51,
52]. Specific targeting of human EGFR with the monoclonal antibody cetuximab is already showing promise in trials in patients with metastatic CRC [
53].
We also studied AR expression in human CRC and explored its relationship with COX-2 expression. In contrast to COX-2, AR showed significant expression in normal colonic mucosa. Interestingly, a pattern of differential expression along the colonic crypt is noted, similar to that previously described for the EP4 receptor [
54]. Prior studies have shown increased AR expression in 50–70% of primary or metastatic colorectal tumours [
55,
56]. We observed a similar trend with 70% of our samples showing significant expression of both COX-2 and AR and concordance observed in the localisation of positive immunostaining within tumours. Interestingly, AR localisation was not confined to the cytoplasm of tumour epithelium as might be expected of a secreted glycoprotein (which acts as a ligand for EGFR). The significance of a nuclear localisation for AR has not been addressed although the presence of a nuclear localisation sequence in the AR protein has been noted and AR shows the ability to interact with nuclear proteins [
57]. This raises the fascinating possibility that AR may act as a nuclear effector for PGE
2 in cancer cells, a hypothesis which merits further evaluation.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
GAD designed and performed cell culture experiments, RT-PCR, flow cytometry and immunohistochemistry, drafted and revised the manuscript. SMB helped design the study, performed cell culture experiments and reviewed the manuscript. ESM helped design the study, performed cell culture experiments and reviewed the manuscript. VM performed cell culture experiments with amphiregulin neutralisation and associated immunoassays. SCA assisted with experimental design and trouble shooting, analysis of results and reviewed the manuscript. EWK assisted with study design, optimisation and interpretation of immunohistochemistry and reviewed the manuscript. FEM assisted with experimental design, interpretation of data and review of the manuscript. DJF assisted with experimental design, interpretation of data, drafting and revision of the manuscript. All authors have read and approved the final manuscript.