Report

Prostaglandin E2 Promotes Colon Cancer Cell Growth Through a Gs-Axin-ß-Catenin Signaling Axis

See allHide authors and affiliations

Science  02 Dec 2005:
Vol. 310, Issue 5753, pp. 1504-1510
DOI: 10.1126/science.1116221

Abstract

How cyclooxygenase-2 (COX-2) and its proinflammatory metabolite prostaglandin E2 (PGE2) enhance colon cancer progression remains poorly understood. We show that PGE2 stimulates colon cancer cell growth through its heterotrimeric guanine nucleotide-binding protein (G protein)–coupled receptor, EP2, by a signaling route that involves the activation of phosphoinositide 3-kinase and the protein kinase Akt by free G protein βγ subunits and the direct association of the G protein αs subunit with the regulator of G protein signaling (RGS) domain of axin. This leads to the inactivation and release of glycogen synthase kinase 3β from its complex with axin, thereby relieving the inhibitory phosphorylation of β-catenin and activating its signaling pathway. These findings may provide a molecular framework for the future evaluation of chemopreventive strategies for colorectal cancer.

Colorectal cancer represents the third leading cause of cancer-related deaths in both men and women in the United States (1). The development of colon cancer results from the sequential accumulation of mutations or deletions in the coding sequence of a number of tumor-suppressor genes and oncogenes, together with aberrant activity of molecules controlling genomic stability (2). Patients with familial adenomatous polyposis, a disease characterized by the presence of numerous colorectal polyps, harbor germline mutations of one allele of the adenomatous polyposis coli (APC) tumor-suppressor gene and develop colon cancer upon mutational damage or loss of the wild-type allele (3). Like humans, mice with germline mutations in APC, Apcmin (multiple intestinal neoplasia) mice, are predisposed to the formation of intestinal adenomas (4). Loss of full-length APC proteins is also one of the earliest events occurring in sporadic colon cancer, suggesting that APC may act as a gate-keeper of the colonic epithelium. Nonsteroidal anti-inflammatory drugs (NSAIDs)—which inhibit two enzymes involved in prostaglandin biosynthesis, cyclooxygenase-1 (COX-1) and cyclooxygenase-2 (COX-2)—reduce the number and size of adenomas in patients with familial adenomatous polyposis and prevent colon cancer development in Apcmin mice (5). Indeed, emerging clinical and experimental evidence now supports a potent antitumorigenic efficacy of NSAIDs in colon cancer (6) and implicates the contribution of COX-2 and one of its metabolites, prostaglandin E2 (PGE2), in colon cancer development (7). How the interplay between PGE2 and APC-regulated pathways leads to colon cancer cell growth remains poorly understood.

PGE2 is a potent mitogen in colon cancer cells (8), as reflected by its ability to stimulate the synthesis of DNA to an extent similar to that provoked by serum in DLD-1 cells, a colon cancer cell line homozygous for an inactivating mutation in APC (9) (fig. S1, A and B) (10). We and others (11) obtained similar results in a panel of colon cancer–derived cells. Although PGE2 can cause the indirect activation of EGF receptors (EGFR) (12), we observed only a minimal increase in the tyrosine phosphorylation of EGFRs upon PGE2 treatment. Furthermore, EGFR kinase inhibitors, such as AG1478, diminished the mitogenic response to EGF but did not prevent the stimulation of DNA synthesis in response to PGE2 or serum (supporting online text). These results indicate that PGE2 may also stimulate EGFR-independent cell growth pathway(s). Among them, we focused our attention on β-catenin, whose cytoplasmic stabilization contributes to colon cancer progression upon APC loss (13). Increased amounts of β-catenin protein lead to complex formation between β-catenin and members of the transcription factor family that includes the T cell factor (TCF) and lymphoid enhancer factor-1 (LEF) family of DNA-binding proteins, resulting in activation of target gene promoters (13).

PGE2 stimulated expression of a β-catenin/TCF/LEF-dependent reporter gene system, TOPflash (14, 15), in colon cancer cells (Fig. 1B and fig. S1C). Similar results were recently reported by others while this study was under revision (16). Because PGE2 receptors are coupled to the G protein Gs, which causes accumulation of cyclic adenosine monophosphate (cAMP) and activates protein kinase A (PKA), we confirmed that PGE2 treatment or transfection of cells with the active catalytic subunit of PKA also stimulated the activity of a cAMP-responsive-element–driven reporter gene (CRE-luc) (Fig. 1A). However, PKA did not activate TOPflash (Fig. 1B). Similarly, accumulation of cAMP after activation of adenylyl cyclase with forskolin (17) promoted CRE-luc activation but not TOPflash reporter activity (Fig. 1, A and B). These results suggest that activation of PKA is not sufficient to stimulate the β-catenin pathway. Rather, activation of TOPflash by PGE2 correlated with the dephosphorylation of β-catenin and its accumulation and nuclear translocation (Fig. 1C). Furthermore, β-catenin was strictly required for the mitogenic activity of PGE2, because reduction of cellular concentration of β-catenin by a specific small interfering RNA (siRNA) inhibited the growth-promoting effect of this metabolic product of COX-2 (Fig. 1D).

Fig. 1.

PGE2 promotes growth of colon cancer cells through β-catenin. (A and B) Transcriptional activation of reporter plasmids expressing luciferase under the control of the cAMP-responsive element, CRE-Luc (A), and the optimal Tcf-responsive element or its mutant, TOP-Luc and FOP-Luc (B), respectively, was evaluated in DLD-1 control (–) or treated (+) cells with various concentrations of PGE2 (1 to 10 μM), forskolin (Fsk) (10 μM), or a plasmid expressing an active form of PKA. Cells were assayed for luciferase activities. The data are expressed as fold induction relative to control. Results represent the average from three independent experiments ± SEM. (C) Abundance and phosphorylation status of β-catenin (upper panel) and its localization (lower panel) were evaluated in DLD-1 cells stimulated with PGE2. β-catenin was detected by Western blot analysis or immunofluorescence microscopy with an antibody to β-catenin and phospho-β-catenin. (D) Effects of β-catenin siRNA on PGE2-induced cell proliferation. DLD-1 cells were transfected with control siRNA, GAPDH (glyceraldehyde phosphate dehydrogenase) siRNA, or β-catenin siRNA (100 nM). Protein amounts and bromodeoxyuridine (BrdU) incorporation in response to PGE2 and serum were measured 48 hours after transfection.

Because EP2 receptors are central mediators of the responses to PGE2 in colon cancer cells (7, 18), we tested whether these prostaglandin receptors promoted activation of β-catenin upon ectopic expression in human kidney embryonic epithelial HEK293T cells. We used β-adrenergic receptors, a prototypical Gs-coupled receptor (19), as a control. PGE2 and isoproterenol, a β-adrenergic agonist, stimulated both the CRE and TOPflash reporter systems (Fig. 2A). In contrast, forskolin, which increases intracellular cAMP, did not enhance TOPflash activity despite activating CRE-luc to a similar extent as did PGE2. Because EP2 and β-adrenergic receptors are coupled to Gs proteins, we tested whether the reporter activity could be increased by a constitutive active form of Gαs (Gαs Q227L, GαsQL) (Fig. 2B). GαsQL activated both CRE- and TOPflash-mediated luciferase activity in a dose-dependent manner. Activation of the β-catenin pathway required the constitutive activity of Gαs, because expression of its wild-type form, GαsWT, or an active mutant of Gαi2, Gαi2QL, failed to stimulate TOPflash activity in HEK293T and colon cancer cells (fig. S2). However, specific inhibition of PKA by expression of PKI, a heat-stable inhibitor of PKA (20), abolished the activation of CRE by GαsQL and PGE2 but not their activation of TOPflash (Fig. 2B). This result suggested that Gs-coupled receptors activate β-catenin through the Gαs protein but independently of cAMP and PKA.

Fig. 2.

Gs-coupled receptors (EP2 and β-adrenergic receptors) promote β-catenin activation independently of PKA. (A) Empty vector (–) or expression plasmids for EP2 or β-adrenergic receptor were transfected together with pGL3-CRE-Luc or pTOP and pFOP reporter plasmids. After 24 hours, cells were deprived of serum and stimulated with various concentrations of PGE2 (1 to 10 μM), isoproterenol (Iso) (10 to 100 μM), or forskolin (Fsk) (10 μM), as indicated, and assayed for dual luciferase activities. The data are expressed as fold increase relative to control ± SEM of a representative experiment that was repeated three times with nearly identical results. (B) TOP and FOP or CRE-Luc activities were also measured in EP2-expressing cells upon transfection of the vector alone or increasing concentrations (0.1 to 1 μg) of pCEFL-GαsQL or RSV-PKI in the absence (–) or presence (+) of PGE2 stimulation (1 μM).

The pathway leading to β-catenin activation by Wnt involves a complex series of events that results in the dissociation of β-catenin from axin, a scaffold protein that forms a large molecular complex with APC, the signal transducer disheveled (Dsh), and GSK-3β, a kinase that phosphorylates β-catenin, thereby promoting its ubiquitin-dependent proteolytic degradation (21). The importance of this inhibitory activity of axin is reinforced by the observation that inactivating mutations in axin are found in hepatocellular carcinomas (22). Axin binds APC through a regulator of G protein signaling (RGS) domain (23), which is similar in primary amino acid sequence and overall three-dimensional structure to other RGS proteins, whose best known function is to accelerate the guanosine triphosphatase (GTPase) activity of G proteins (24). However, the surface area of the axin RGS domain that binds APC is distinct from that used by other RGS proteins to bind G protein α subunits (25). This observation and the ability of Gs-coupled receptors to stimulate the transcriptional activity of β-catenin prompted us to explore whether the axin RGS domain may provide a direct link between Gαs and the β-catenin signaling axis. By overexpressing an epitope-tagged form of axin in HEK293T cells (fig. S3A), we observed that axin coimmunoprecipitated with GαsQL but not with the active form of Gαi2 (Fig. 3, A and B). Gαs wild-type also coimmunoprecipitated with axin, but only when cells were treated with aluminum fluoride, which promotes Gα subunits to acquire a conformation that resembles their transition state, thus favoring RGS binding and GTPase activating protein (GAP) activity (26) (Fig. 3B). We next expressed epitope-tagged forms of individual axin domains, including the RGS domain, a region including the GSK-3β phosphorylation and binding sites, a β-catenin binding region, a protein phosphatase 2A (PP2A) binding area, and a DIX domain by which axin binds dsh (Fig. 3C and fig. S3B). Upon coexpression in HEK293T cells, only the RGS domain of axin was coimmunoprecipitated with GαsQL (Fig. 3C), indicating that axin interacts with Gαs through its RGS domain. Indeed, expression of a lentivirus encoding the axin RGS domain fused to green florescent protein (GFP) in DLD1 cells inhibited the activation of TOPflash evoked by PGE2 but not activation by an active mutant of β-catenin (Fig. 3D and fig. S3C). Expression of axin (RGS) also almost completely abolished the mitogenic activity of PGE2 but not the proliferative response to serum in these colon cancer cells (Fig. 3E).

Fig. 3.

Axin coimmunoprecipitation with activated Gαs through its RGS domain. (A) pcDNA3-Myc-Axin and pCEFL-HA GαsQL were cotransfected in HEK293T cells using by lipofection. After 24 hours, cells were lysed and immunoprecipitated (IP) with an antibody to Myc. Total cell extracts (TCE) (upper panels) or IP (lower panel) were assayed by immunoblotting with either a hemagglutinin (HA) antibody or a Myc antibody. (B) Selective binding of activated Gαs to axin. HEK293T cells overexpressing Myc axin and the wild-type (WT) and activated (QL) forms of Gαs and Gαi were subjected to immunoprecipitation with an antibody to axin in the absence (–) or presence (+) of AMF. Binding of G protein α subunits to axin was detected with a mix of antibodies to Gαs and Gαi only in the samples expressing GαsQL and in cells expressing GαsWT when treated with AMF. HA GαsQL was used for these experiments, hence its slightly higher molecular weight. (C) Immunoprecipitation of axin RGS domain with GαsQL. Vector-alone (AU5), or epitope-tagged forms of individual axin domains, including the RGS domain (A), a region including the GSK-3β phosphorylation and binding sites (B), a β-catenin binding region (C), a PP2A binding area (D), and a DIX domain by which axin binds dsh (E), were cotransfected with pCEFL-HA GαsQL. After 24 hours, cells were lysed and immunoprecipitated with an antibody to HA and analyzed by Western blotting with antibody to AU5 (upper panel). Total cell extracts were immunoblotted with either AU5 or HA antibody (lower panels). A band corresponding to the anti-HA immunoglobulin G (IgG) is depicted by an empty arrowhead. The position of the molecular size markers is indicated. (D and E) Axin-RGS domain inhibits the activation of the β-catenin pathway by PGE2, as well as its mitogenic effect. (D) Luciferase activity was measured in GFP and GFP-Axin RGS–infected DLD-1 cells transfected with the TOP and FOP reporter plasmids and stimulated with PGE2 (1 μM). Luciferase expression in cell lysates is represented as in Fig. 1. Transfection of an activated mutant of β-catenin (37A β-cat) was used as a control. (E) DNA synthesis was also measured as described in Fig. 1. The data were averaged from three independent experiments ± SEM, in which at least 500 cells were counted. (F) In vitro binding of axin to activated Gαs but not Gαi. Left: Recombinant His6s was immobilized on Ni++ agarose (Ni/NTA) in the presence of GDP, aluminum magnesium fluoride (AMF), or GTPγS, and incubated with recombinant GST-Axin (RGS domain), GST-PX1 (RGS), or GST-RGS19 (GAIP). After bead washing, bound proteins were identified by immunoblotting with antibody to Gαs (top) or antibody to GST (bottom), running half of the total purified GST-fusion protein used for the experiment (input) as a control. Right: The indicated bacterially expressed GST-fusion proteins were purified and incubated in vitro with His6s or Gαi1 and affinity-precipitated with glutathione agarose. Recombinant proteins bound to beads after extensive washing were detected with the indicated antibodies. Half of the total purified His6s or Gαi1 used for the experiment (input) was run as a control. Figures represent four similar experiments.

To investigate whether the axin RGS domain binds Gαs directly, we expressed axin (RGS) as a glutathione S-transferase (GST)–fusion protein in bacteria and measured in vitro binding to recombinant six-histidine-tagged (H6)-Gαs. GST-RGS bound Gαs immobilized on nickel beads in an aluminum fluoride–dependent manner, as did an RGS protein that binds Gαs specifically, PX1 (RGS) (Fig. 3F, left panels). The guanosine diphosphate (GDP)–loaded, inactive form of Gαs also bound PX1 (RGS), albeit to a much lesser extent, as observed for other RGS proteins in vitro (27). By using GTPγS, a non-hydrolyzable GTP analog, we observed that Gαs does not bind to either RGS in its GTP-bound active state. A Gi-specific GAP, RGS19, did not bind Gαs under any condition. Nearly identical results were obtained in reciprocal experiments in which these bacterially expressed GST-RGS fusion proteins were precipitated with glutathione beads (Fig. 3F, left panels). As a control, recombinant Gαi efficiently bound GST-RGS19 in the presence of aluminum fluoride, but not to PX1 or axin (RGS) (Fig. 3F). These findings indicated that Gαs binds directly to the RGS domain of axin in a transition-state conformation.

Because axin bound Gαs in an aluminum fluoride–dependent manner, we evaluated whether axin could increase the GTPase activity of Gαs. However, neither the RGS domain of axin nor full-length axin purified from baculovirus-infected Sf9 cells promoted the GTPase activity of Gαs during a single catalytic cycle of the enzyme (fig. S3D). Thus, additional accessory molecules or other modifications of axin could be required for its GAP activity, as is the case for other RGS proteins (28). It is also possible that the RGS domain of axin might be used primarily as a structural feature by which this scaffold protein can interact with and act as an effector for Gαs, as do the RGS domain–containing RhoGEFs, which are effectors for G proteins of the Gα12/13 family (29).

Because phosphorylation of β-catenin by GSK-3β leads to its rapid ubiquitination and subsequent degradation in the proteosome, inactivation of GSK-3β is often a prerequisite for stimulating the accumulation, nuclear translocation, and functional activity of β-catenin (30). Treatment of cells with PGE2 led to the rapid phosphorylation of GSK-3β on serine 9 (Fig. 4A and fig. S4), which inhibits its kinase activity (31). PGE2 also stimulated Akt activity in a PI3K-dependent manner (Fig. 4B). Because both PKA and Akt can phosphorylate GSK-3β at this inhibitory site (32), we tested whether PKA could mediate GSK-3β phosphorylation in response to PGE2. PKI did not prevent the phosphorylation of GSK-3β provoked by PGE2, but it diminished the GSK-3β phosphorylation induced by forskolin, indicating that PGE2 induces GSK-3β phosphorylation independently of PKA (Fig. 4A). In contrast, blockade of the PI3K-Akt pathway by wortmannin (WM) abolished both basal and PGE2-induced phosphorylation of GSK-3β (Fig. 4B). Although PGE2 promotes nucleotide exchange on Gαs and subsequent dissociation of GTP-bound Gαs from Gβγ subunits, GTP-bound Gαs does not activate PI3K and Akt (33). Thus, we tested whether expression of the C-terminal domain of βark (βARK-C), which causes sequestration of Gβγ subunits (34), could inhibit phosphorylation of GSK-3β and Akt. Overexpression of βARK-C did not affect basal phosphorylation of Akt or GSK-3β but abolished the accumulation of their phosphorylated species upon PGE2 stimulation (Fig. 4B). βARK-C did not affect the phosphorylation of Akt and GSK-3β evoked by insulin, and neither WM nor βARK-C affected the activation of the CRE reporter by PGE2 (Fig. 4C), which is dependent on Gαs and cAMP.

Fig. 4.

PGE2 stimulation of β-catenin activity through a convergent mechanism initiated by Gαs and Gβγ. (A) PGE2 stimulates GSK-3β phosphorylation independently of PKA. EP2-expressing HEK293T cells were stimulated with either PGE2 (1 μM) or forskolin (10 μM) at different time points, upon transfection with an empty vector (control) or a PKI-expressing vector (PKI). Cells were lysed and immunoblotted with antibodies to phospho-GSK-3β (upper panel) or GSK-3β (lower panel). (B) PGE2-dependent GSK-3β phosphorylation is mediated by PI3K-Akt activation initiated by the βγ subunits of Gs. Control HEK293T cells pretreated with 50 nM wortmannin (WM) for 15 min or transfected with pcDNA-βARK-C (βARK-C) were lysed 30 min after PGE2 stimulation, and lysates were probed with phospho-GSK-3β, GSK-3β, phospho-Akt, or Akt antibodies. (C) TOP, FOP, and CRE-luciferase activities were assayed as described in Fig. 1. (D) PGE2 promotes the dissociation of the axin-GSK-3β complex. Lysates from HEK293T, DLD-1, and SW480 cells were immunoprecipitated with axin antibody and probed with antibodies to GSK-3β and axin. (E) Activated Gαs (QL) but not wild-type Gαs (WT) and activated Gαi2 (QL) can bind to axin and displace GSK-3β from its binding to axin. HEK293T and DLD1 cells were transfected with plasmids expressing GαsWT, GαsQL, and Gαi2QL. Cell lysates were immunoprecipitated with an axin antibody and probed with antibodies to GSK-3β and axin or a mix of antibodies to Gαs and Gαi, as indicated. Expression of the corresponding molecules was also evaluated in total cell extracts (TCE).

Inhibition of GSK-3β phosphorylation by WM or βARK-C only partially reduced the activation of the β-catenin pathway by PGE2 (Fig. 4C), even when both treatments were combined. The view that GSK-3β phosphorylation is required for β-catenin activation has been challenged by experiments using knock-in mice homozygous for both a mutant GSK-3β lacking serine 9 and a mutant for the related GSK-3α lacking serine 21. These mice displayed normal regulation of the Wnt-β-catenin pathway (35). Likewise, GSK-3β phosphorylation may not be alone sufficient to stimulate β-catenin, as suggested by the findings that insulin promotes the phosphorylation of GSK-3β but not β-catenin activation (36) and that forskolin causes the PKA-dependent phosphorylation of GSK-3β but does not stimulate the TOPflash reporter effectively. Our data suggest that β-catenin stabilization by PGE2 occurs at least through two coordinated mechanisms, one initiated by Gβγ through PI3K, Akt, and the consequent phosphorylation and inactivation of GSK-3β, and another pathway initiated by Gαs that is independent of both GSK-3β phosphorylation and PKA activation.

GSK-3β appears to phosphorylate β-catenin primarily when it is bound to axin (37). We observed that GSK-3β coimmunoprecipitated with endogenous axin in both HEK293T and colon cancer cells (Fig. 4, D and E). However, stimulation of cells with PGE2 or expression of activated forms of Gαs was associated with reduced amounts of GSK-3β bound to axin without affecting the total amount of GSK-3β (Fig. 4, D and E). Because the GSK-binding site on axin is close to the RGS domain, we tested whether Gαs interaction with axin competes for GSK-3β binding in experiments with recombinant proteins. However, a 10-fold molar excess Gαs-AMF (aluminum magnesium fluoride) did not reduce association of axin-H6 with recombinant GSK-3β, suggesting that additional factors may be required for the dissociation of axin-GSK-3β complexes.

To test whether displacing GSK-3β from axin is sufficient to stimulate the β-catenin pathway, we first reduced the cellular concentration of GSK-3β and axin by specific siRNAs. Knockdown of axin or GSK-3β activated the TOPflash reporter (Fig. 5A and fig. S5A). Furthermore, the GSK-3 binding region of axin, but not the axin RGS as a control, bound GSK-3β and displaced it from its binding to axin (Fig. 5B and fig. S5B). This resulted in an increase in β-catenin activity similar to that stimulated by PGE2. Thus, the ability of PGE2 and Gαs to promote the release of GSK-3β from axin-containing complexes may represent an alternative mechanism to inhibit the functional activity of GSK-3β independently of phosphorylation.

Fig. 5.

Knockdown of axin or GSK-3β, or displacement of GSK-3β from axin, is sufficient to stimulate the β-catenin pathway in DLD1 colon cancer cells. (A) Knockdown of GSK-3β or axin in DLD1 cells stimulates β-catenin to a similar extent as PGE2 stimulation. Cells were transfected with the indicated siRNAs, and the expression of endogenous axin or GSK-3β was evaluated by Western blot analysis (left panels). Western blot analysis of Bcl-xL was used as a control. Luciferase expression was measured in cells transfected with the TOP and FOP reporter plasmids and the indicated siRNAs, with (+) or without (–) stimulation with PGE2 (1 μM), as in Fig. 1C (right panels). (B) Expression of the isolated GSK-3β binding region of axin is sufficient to displace GSK-3β from axin and stimulate β-catenin signaling. The GSK-3β binding region of axin [depicted as (B) in fig. S3] binds GSK-3β in vivo and displaces it from its binding to endogenous axin (left panel). The axin RGS domain [depicted as (A) in fig. S3)], which does not bind GSK-3β, served as a control. Arrows point to the epitope-tagged forms of the indicated axin regions. A band corresponding to the anti-HA IgG is depicted by an empty arrowhead. Cells were transfected with the TOP and FOP reporter plasmids together with the vector control or the expression vectors for the isolated axin domains and either left untreated (–) or stimulated with PGE2 (1 μM) (+). Luciferase expression was measured and represented as in Fig. 1B (right panel).

These observations may have important implications for the study of β-catenin activation by Wnt, which involves two cell surface receptors, an LDL-containing single-pass transmembrane protein, LRP5 or LRP6, and a seven-transmembrane receptor, Frizzled (38, 39). Whereas LRP5 and LRP6 appear to bind axin directly, how Frizzled signals to the canonical β-catenin pathway is still unclear. Frizzled may activate heterotrimeric G proteins (40), thus raising the possibility that the association of G protein α subunits with the RGS domain of axin may represent a point of convergence between the Wnt and prostaglandin-initiated pathways leading to β-catenin activation. On the other hand, how this process can be regulated by APC is at the present unknown. In our in vitro experiments, a peptide containing the primary sequence of APC that binds to axin RGS (23) did not compete for Gαs binding to axin (RGS). This is consistent with a distinct binding surface area for APC and Gαs on the axin RGS domain as predicted by the crystal structure (25). In contrast, expression of APC in colon cancer cells inhibited the activation of the β-catenin reporter system by PGE2. Considering that the tumor-promoting effect of PGE2 becomes evident only in the absence of functional APC, it is conceivable that APC may act by limiting the activation of β-catenin by prostaglandins, which are normally released in the colon in response to bacterial infection and proinflammatory substances. APC may bind the axin RGS domain and hinder the access of Gα subunits to this RGS. Alternatively, as APC interacts with β-catenin through numerous binding motifs, it may sequester free β-catenin, once it has been released from axin, and promote its turnover (41), thus raising the threshold of EP2 activation that is required to stimulate β-catenin–dependent gene expression.

Our studies indicate that in the absence of a functional APC, PGE2 can stimulate the proliferation of colon cancer cells by activating the β-catenin axis through a biochemical pathway initiated by the activation of the G protein–linked PGE2 receptor, EP2 (Fig. 6). PGE2 stimulation leads to the association of the activated α subunit of Gs with the RGS domain of axin, promoting release of GSK-3β from its complex with axin. Concurrently, free Gβγ subunits liberated upon Gαs activation directly stimulate the activity of PI3K and Akt, leading to phosphorylation and inactivation of GSK-3β. Ultimately, these processes result in the stabilization and nuclear translocation of β-catenin, thereby stimulating LEF and β-catenin–dependent gene expression and the aberrant growth of colon cancer cells. These findings support the existence of a direct molecular mechanism by which COX-2 and inflammation can promote the progression of colon cancer, thus providing a molecular framework for the future clinical evaluation of NSAIDs as chemopreventive strategies for this devastating disease.

Fig. 6.

Schematic representation of β-catenin pathway activation in response to PGE2. In the basal state (left panel), a cytoplasmic protein complex containing GSK-3β and axin promotes the inhibitory phosphorylation and consequent ubiquitin-dependent degradation of β-catenin in the proteasome. Overexpression of COX-2 in colon cancer cells and during inflammatory processes leads to the production of PGE2, which can activate EP2 receptors that are coupled to heterotrimeric G proteins of the Gs family (right panel). Upon exchange of GDP for GTP, the α subunit of Gs binds the RGS domain of axin, thereby promoting the release of GSK-3β from the complex. Concomitantly, free βγ subunits stimulate the PI3K-PDK1-Akt signaling route, which causes the phosphorylation and inactivation of GSK-3β. These events lead to the stabilization and nuclear translocation of β-catenin and to the expression of growth-promoting genes regulated by the Tcf and LEF family of transcription factors.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1116221/DC1

Materials and Methods

SOM Text

Figs. S1 to S5

References

References and Notes

View Abstract

Navigate This Article