RhoA Mediates Cyclooxygenase-2 Signaling to Disrupt the Formation of Adherens Junctions and Increase Cell Motility

Introduction

Cyclooxygenases (COX) are the key enzymes in the biosynthesis of prostaglandins. COX-2, the inducible form of COX, is overexpressed in early and advanced colorectal cancer tissues and is associated with a poor prognosis ( 1, 2). COX-2 and its derived prostaglandins have been shown to promote tumorigenesis. Overexpression of COX-2 or treatment with prostaglandin prostaglandin E₂ (PGE₂) promotes transformed and invasive phenotypes, such as increased cell proliferation, motility, invasion, angiogenesis, resistance to apoptosis, and tumor growth ( 1, 2). Conversely, deletion of the COX-2 gene in Apc^Δ716 mice, a model frequently used for studying colorectal cancer, dramatically reduced intestinal polyp numbers and resulted in decreased neoplastic growth ( 1, 2). In addition, treatment with selective COX-2 inhibitors resulted in a significant decrease in colorectal adenomas in humans and in animals ( 1, 2).

During the progression to invasive and metastatic phenotypes, tumor cells undergo a series of changes that are associated with cytoskeleton rearrangements as well as alterations of cell-cell and cell-matrix adhesion, allowing cells to separate from the tumor, invade surrounding tissues, and ultimately metastasize to distant organs ( 3). Rearrangement of the actin cytoskeleton is primarily controlled by the members of Rho small GTPase family, including RhoA, Rac1, and Cdc42 ( 4). Like other small GTPases, Rho GTPases cycle between inactive GDP-bound and active GTP-bound forms. Activation of Rho GTPases is stimulated by guanine nucleotide exchange factors and inhibited by GTPase-activating proteins. On activation, Rho GTPases recruit effector proteins to regulate the actin cytoskeleton. In epithelial cells, Rho GTPases are implicated in regulating morphology and adhesion because interactions between the actin cytoskeleton and adherens junctions determine cell shape and motility ( 3– 6). Formation of adherens junctions promotes cell-cell adhesion and is important in organizing normal epithelial sheets. Adherens junctions consist of the trans-membrane protein E-cadherin, whose cytoplasmic domain interacts with β-catenin, which binds α-catenin ( 3). Because formation of adherens junctions is associated with actin dynamics, Rac1 and Cdc42 activity are required for the formation and maintenance of E-cadherin-mediated adherens junctions ( 3– 6). Although a basal level of RhoA activity is also necessary for adherens junction formation, high RhoA activity disrupts the formation of adherens junctions ( 3– 6). On the other hand, E-cadherin-mediated formation of adherens junctions strongly inhibits RhoA activity but increases activities of both Rac1 and Cdc42 ( 3– 6).

There is a growing evidence that expression levels of RhoA are significantly elevated in several human tumors, including colon cancer ( 7– 10). One possible mechanism by which RhoA promotes tumor invasiveness and metastasis is down-regulation of adherens junction formation. Microinjection of constitutively active RhoA results in disruption of adherens junctions by mislocalization of α-catenin in HCT116 colon carcinoma cells ( 11). In contrast, overexpression of an active form of Rho kinase (ROCK) I, the RhoA effector protein, disrupts adherens junctions by reducing the total levels of E-cadherin and α-catenin in HEK293 cells ( 11). Conversely, inhibition of ROCK with the ROCK inhibitor Y-27632 promotes the localization of α-catenin at cell-cell contacts of SW620 colon carcinoma cells with particularly high levels of RhoA ( 11). Furthermore, conditional activation of ROCK II removes the localization of α-catenin and γ-catenin from cell-cell junctions in LS174T colon carcinoma cells and this is inhibited by Y-27632 ( 12). Loss of adherens junctions results in altered cell shape, promotes disruption of epithelial sheet organization, and as a consequence, increases motility and invasiveness ( 3, 7).

Disruption of adherens junctions is commonly associated with tumor progression to invasive and metastatic stages and apparently results from a loss of E-cadherin or α-catenin expression ( 3). Levels of E-cadherin are reduced in rat intestinal epithelial cells that stably express COX-2 compared with the parental cells ( 13). Human non–small cell lung cancer (NSCLC) cells stably expressing COX-2 show low levels of E-cadherin, which are increased by celecoxib treatment ( 14). Furthermore, PGE₂ mediates COX-2 signaling to induce the transcription repressors ZEB1 and Snail of E-cadherin in NSCLC cells ( 14). However, the molecular mechanisms by which COX-2 regulates cell-cell junctions and adhesion in colon carcinoma cells are not well understood. In this study, we have investigated the effects of constitutive expression of COX-2 in colon carcinoma cells on RhoA and formation of adherens junctions. Our data identify a novel cross-talk between COX-2 signaling and the RhoA/ROCK pathway, which mediates COX-2 signaling to disrupt adherens junctions by reducing levels of E-cadherin and α-catenin leading to increased motility.

Materials and Methods

Antibodies and reagents. RhoA monoclonal antibody, COX-2 small interfering RNA (siRNA), RhoA siRNA, and control siRNA were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). COX-2, Rac1, Cdc42, and E-cadherin monoclonal antibodies were from BD Biosciences Pharmingen (San Diego, CA). The α-catenin rabbit polyclonal antibody (C2081) and β-actin mouse monoclonal antibody were from Sigma-Aldrich (St. Louis, MO). E-cadherin mouse monoclonal antibody (clone HECD-1) was from Zymed (Carlsbad, CA). The k-Ras monoclonal antibody (AB-1, clone 234-4.2), toxin A, and Y-27632 were from Calbiochem EMD Biosciences (San Diego, CA). Glutathione-Sepharose 4B was from GE Healthcare Amersham Biosciences (Piscataway, NJ). Goat anti-mouse and anti-rabbit antibodies conjugated with horseradish peroxidase (HRP) and SuperSignal West Pico chemiluminescent substrate were from Pierce Biotechnology (Rockford, IL). Alexa Fluor 568–conjugated goat anti-mouse antibody, Alexa Fluor 488–conjugated goat anti-rabbit antibody, and CellTracker Orange were from Molecular Probes (Carlsbad, CA). FluoroBlok inserts (8 μm) were from BD Biosciences Discovery Labware (Bedford, MA). NS-398 and SC-236 were from Alexis Biochemicals (Axxora, San Diego, CA). PGE₂ was from Cayman Chemical (Ann Arbor, MI). TransIT-TKO transfection reagent was from Mirus Bio Corp. (Madison, WI).

Cell culture and treatments. HCA-7 colon carcinoma cells were obtained from the European Collection of Cell Cultures (Salisbury, Wiltshire, United Kingdom). HT-29, DLD-1, and Caco-2 colon carcinoma cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA). HCA-7 and HT-29 cells were maintained in McCoy 5A (Invitrogen, Carlsbad, CA) and DLD-1 cells were maintained in RPMI 1640 (Invitrogen) containing 10% fetal bovine serum (FBS; Hyclone, Logan, UT) and penicillin/streptomycin (Invitrogen). Caco-2 cells were maintained in MEM containing 20% FBS and penicillin/streptomycin.

For inhibitor treatment, cells were plated at the density that they would reach 80% confluency at the following day. Cells were treated with or without NS-398 (10 μmol/L), PGE₂ (0.1 μmol/L), toxin A (50 ng/mL), or Y-27632 (10 μmol/L) in the serum-free medium for the indicated time.

For siRNA transfection, HCA-7 cells were plated at a density that they would reach 50% of confluency within 24 hours. Then, cells were transfected with control siRNA, COX-2 siRNA, or RhoA siRNA using TransIT-TKO transfection reagent following the manufacturer's instruction and incubated for 48 to 72 hours.

Adenoviruses for expression of dominant-negative RhoAN19 (AD-RhoAN19) under the control of tetracycline repressor elements and a minimal cytomegalovirus promoter (Tet-mCMV) were a gift from Dr. Daniel Kalman (Emory University, Atlanta, GA; ref. 15). HCA-7 cells were incubated with AD-RhoAN19 together with AD-VP16, which is required to activate expression of Tet-mCMV ( 15). Virus infection was done in medium containing 2% FBS at 50 plaque-forming units per cell for 1 hour followed by incubation in complete medium for additional 72 hours ( 16).

Cell-permeable TAT-MYC-C3 protein was expressed in Escherichia coli from the plasmid pGEX-KG-TAT-MYC-C3 (a gift from M.F. Olson, Beatson Institute, Glasgow, United Kingdom) and prepared as described previously ( 17). HCA-7 cells were incubated in serum-free medium with or without 0.5 μmol/L TAT-MYC-C3 for 24 hours.

Western blot analysis. Cells were washed once with ice-cold PBS and lysed in lysis buffer [50 mmol/L HEPES (pH 7.5), 150 mmol/L NaCl, 5 mmol/L MgCl₂, 1% NP40, 1 mmol/L DTT, 5 μg/mL aprotinin, 5 μg/mL leupeptin, 1 mmol/L phenylmethylsulfonyl fluoride]. Lysates were clarified by centrifugation at 13,000 × g at 4°C for 15 minutes. The protein concentration was quantified by Bradford assay using bovine serum albumin (BSA) as a standard. The cell lysates were separated on 12% SDS-PAGE and transferred to polyvinylidene difluoride membranes. Membranes were blocked with 5% nonfat dry milk in TBS plus 0.1% Tween 20 (TBST) and then incubated with the specific primary antibody diluted in TBST containing 3% BSA and 0.1% NaN₃ overnight at 4°C. The membranes were washed with TBST followed by incubation with HRP-conjugated secondary antibody (1:20,000 dilution) diluted in TBST containing 3% nonfat dry milk for 1 hour. The detection was carried out using SuperSignal West Pico chemiluminescent substrate followed by exposure to X-ray films.

Small GTPase pull-down assay. The cDNA corresponding to the p21-binding domain (PBD) of p21 activated protein kinase PAK2 (amino acids 66–146) was amplified from PAK2 cDNA ( 18) by PCR and cloned between the NdeI and XbaI sites of the pGEX-2T vector. Glutathione S-transferase (GST)–Rhotekin-Rho binding domain (RBD) and GST-Raf1-Ras binding domain were generated as described previously ( 19, 20). GST-fusion proteins were expressed in E. coli, purified by glutathione-Sepharose affinity chromatography, and stored bound to the GST-Sepharose beads in 10% glycerol at −80°C as described previously ( 19). The capacity of GST-fusion protein was determined by incubation of different amount of beads with the NIH3T3 lysates pretreated with GTPγS or GDP ( 21).

Pull-down assays were carried out using recombinant GST-Rhotekin-RBD for assaying RhoA activity, GST-PAK2-PBD for measuring activities of Rac1 and Cdc42, or GST-Raf1-Ras binding domain for assaying k-Ras activity ( 22). Cell lysates were incubated with GST-Rhotekin-RBD (200 μg), GST-PAK2-PBD (20 μg), or GST-Raf1-Ras binding domain (80 μg) at 4°C for 1 hour. The beads were washed thrice with 500 μL lysis buffer. Bound proteins were eluted by incubation in 2× Laemmli buffer at 95°C for 5 minutes. The eluted samples were analyzed by Western blot using a RhoA, Rac1, Cdc42, or k-Ras antibody.

Immunofluorescence. Cells were fixed in 4% formaldehyde followed by permeabilization in 0.2% Triton X-100. After washing, cells were blocked in 1% BSA and then incubated with α-catenin antibody or E-cadherin antibody clone HECD-1 (1:1,000 dilution) in PBS containing 1% BSA (Fraction V). Cells were incubated with Alexa Fluor 568–conjugated goat anti-mouse antibody or Alexa Fluor 488–conjugated goat anti-rabbit antibody (1:500 dilution) in PBS containing 1% BSA. Cells were mounted in immunofluore mounting medium. Fluorescence microscopy was done using a Nikon Eclipse TE2000U microscope and images were acquired by a charge-coupled device monochrome camera (CoolSnaps ES) using MetaMorph software.

Motility assay. Motility assays were carried out in triplicate using FluoroBlok inserts (8 μm) in a modified Boyden chamber. Cells were labeled with CellTracker Orange (2 μmol/L) in serum-free medium for 45 minutes followed by removal of the dye and incubation in the medium for additional 30 minutes. Cells were trypsinized, suspended (5 × 10⁴) in serum-free medium containing 0.2% BSA, and placed in upper chamber. The lower chamber was filled with serum-free medium containing vehicle or NS-398 (10 μmol/L). After an incubation period of 24 hours at 37°C, the cells on the upper surface of the filter were removed with a cotton swab. Cells adhering to the lower surface of the filter were visualized by fluorescence microscopy. A total of three image fields were counted for each sample.

Results

The activity of RhoA is elevated in HCA-7 colon carcinoma cells that constitutively express COX-2. Because both COX-2 and RhoA have been implicated to promote transformed and invasive phenotypes, we investigated if constitutive expression of COX-2 regulates RhoA activity. Several colon carcinoma cell lines, in which COX-2 is differentially expressed, were used in these studies. HCA-7 colon carcinoma cells established from a primary human colorectal adenocarcinoma are well differentiated ( 23). HT-29 and Caco-2 colon carcinoma cells derived from colorectal adenocarcinoma are also well differentiated but poorly invasive ( 24, 25). DLD-1 colon carcinoma cells also derived from colorectal adenocarcinoma are dedifferentiated and highly invasive ( 26). Western blot analysis using a mouse monoclonal anti-COX-2 antibody clearly showed that only HCA-7 cells but not HT-29, Caco-2, and DLD-1 cells constitutively contain high levels of COX-2 protein in both serum-fed and serum-starved conditions ( Fig. 1A ). We measured RhoA activity by a pull-down assay to specifically capture the active GTP-bound form of RhoA using GST-Rhotekin-RBD ( Fig. 1B). The activity of RhoA was significantly higher in HCA-7 cells than in HT-29, Caco-2, and DLD-1 cells under serum-fed or serum-starved conditions. In addition, HCA-7 cells contained higher levels of RhoA protein than the other cell lines investigated. We noticed that the RhoA antibody detected two bands in the lysate from DLD-1 cells and that only the lower band showed active GTP-bound RhoA. It has been reported that a RhoA variant with molecular weight lower than wild-type (WT) is detected in the lysates prepared from some malignant colon tissues but not normal tissues ( 27). The difference in the levels of RhoA-GTP between HCA-7 cells and the other three cell lines was higher than the difference in protein levels, suggesting that the higher levels of active RhoA in HCA-7 cells are caused by stimulated RhoA activation and not only due to more protein present. Our data showed for the first time that RhoA activity is up-regulated in colon carcinoma cells that constitutively express COX-2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Determination of protein levels of COX-2 and activity and protein levels of RhoA, Rac1, Cdc42, and k-Ras in colon carcinoma cells. A, total cell lysates (15 μg) prepared from serum-fed (SF) or 24-hour serum-starved (SS) HCA-7, HT-29, Caco-2, and DLD-1 cells were analyzed by Western blot with a mouse monoclonal anti-COX-2 antibody. The membrane was reprobed with anti-β-actin antibody. The experiments were repeated four times. Representative data. B, activity and protein levels of RhoA were analyzed by Western blot with anti-RhoA antibody using pulldowns of the active GTP-bound form of RhoA or total lysates (15 μg). RhoA-GTP was pulled down from the lysates (1 mg) prepared from serum-fed or serum-starved cells using GST-Rhotekin-RBD. The membrane containing the lysates was reprobed with anti-β-actin antibody. The experiments were repeated four times. Representative data. C, activity and protein levels of Rac1, Cdc42, or k-Ras were analyzed by Western blot with anti-Rac1, Cdc42, or k-Ras antibody using pulldowns of the active GTP-bound form of Rac1, Cdc42, or k-Ras or total lysates (15 μg). Rac1-GTP or Cdc42-GTP was pulled down from the lysates (500 μg) using GST-PAK-PBD, whereas k-Ras-GTP was pulled down from the lysates (250 μg) prepared using GST-Raf1-Ras binding domain. The experiments were repeated twice. Representative data. HCA, HCA-7 cells; HT, HT-29 cells; Caco, Caco-2 colon cells, DLD, DLD-1 cells. D, left, HCA-7 cells were incubated with adenoviruses expressing dominant-active RhoA (AD-RhoAN19) in the presence or absence of adenoviruses expressing VP16; right, HCA-7 cells were incubated with or without TAT-MY-C3. Activities of RhoA, Rac1, and Cdc42 were measured by pull-down assays by incubation of lysates with GST-Rhotekin-RBD or GST-PAK-PBD. Total lysates (15 μg) and pull-down samples were analyzed by Western blot. The experiments were repeated twice. Representative data.

We also measured the activity of other Rho family members, Rac1 and Cdc42, by pull-down assays using GST-PAK-PBD ( Fig. 1C). Although the protein levels were similar, the activity of Rac1 in HCA-7 cells was significantly lower than in HT-29, DLD-1, and Caco-2 cells. In agreement with a previous report, a splice variant of Rac1, Rac1b, was detected in both pulldown and lysate from HT-29 cells ( 28). Although the protein levels of Cdc42 were higher in HCA-7 cells than in DLD-1, Caco-2, and HT-29 cells, the activity of Cdc42 in HCA-7 cells was lower than in the other three cell lines. Additionally, we analyzed the activity of k-Ras in HCA-7 cells because k-Ras is frequently mutated in colorectal cancer ( Fig. 1C; ref. 29). Interestingly, our data show high levels of active k-Ras in HCA-7 cells, although the protein levels were lower than in DLD-1, Caco-2, and HT-29 cells. In agreement with previous findings, we also detected high levels of active k-Ras in DLD-1 cells, which was reported to harbor an oncogenic mutation in the k-Ras gene (G13D; ref. 30). We did not detect active k-Ras in HT-29 and Caco-2 cells, in agreement with reports that HT-29 and Caco-2 colorectal cells have WT k-Ras ( 30, 31).

It has been suggested that Rho family proteins antagonize each other. Rac1 leads to the inactivation of RhoA or vice versa ( 4, 32). Therefore, we examined if low activity of Rac1 or Cdc42 in HCA-7 cells is caused by elevated RhoA activity. Adenovirus-mediated overexpression of dominant-negative RhoAN19 in HCA-7 cells blocked RhoA activity but had no significant effects on Cdc42 activity and reduced Rac1 activity ( Fig. 1D, left). Similarly, treatment of HCA-7 cells with the cell-permeable C3 exoenzyme (TAT-MYC-C3) resulted in the mobility shift of RhoA and markedly reduced RhoA activity but had no significant effect on Cdc42 activity and decreased Rac1 activity ( Fig. 1D, right; ref. 17). Because inhibition of RhoA did not result in increased activity of Rac1 or Cdc42, it is suggested that elevated RhoA activity is not involved in down-regulation of Rac1 or Cdc42 activity in HCA-7 cells.

Inhibition of COX-2 activity results in decreased RhoA activity. To examine if constitutive expression of COX-2 regulates RhoA activity, HCA-7 cells were treated with the specific COX-2 inhibitor NS-398 ( 33). The levels of RhoA-GTP in HCA-7 cells were significantly reduced at 16 hours of incubation with NS-398 and little RhoA-GTP was detected at 24 hours, whereas the levels of RhoA proteins remained unchanged ( Fig. 2A ). To verify that the decrease of RhoA activity was not due to direct inhibition of RhoA activity by NS-398 treatment, DLD-1 cells were also treated with NS-398. NS-398 treatment had no significant effect on the levels of RhoA protein and RhoA-GTP in DLD-1 cells ( Fig. 2A). We also examined if inhibition of COX-2 affects k-Ras activity in HCA-7 cells. NS-398 treatment had no effects on the level of k-Ras protein and k-Ras activity in both HCA-7 and DLD-1 cells ( Fig. 2B). Furthermore, inhibition of COX-2 by another highly selective inhibitor SC-236, which is structurally similar to celecoxib, also significantly reduced RhoA activity ( Fig. 2C; refs. 34, 35). Because HCA-7 cells have been shown to produce significant amounts of PGE₂, we examined the effects of PGE₂ on RhoA activity ( 36, 37). Inhibition of RhoA activity by treatment with NS-398 was reversed by PGE₂ treatment ( Fig. 2D). Therefore, our data suggest for the first time that constitutive expression of COX-2 is involved in stimulation of RhoA activity. Furthermore, PGE₂ seems to mediate COX-2 signaling to regulate RhoA activity.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Inhibition of COX-2 results in decreased levels of the active, GTP-bound form of RhoA but not k-Ras. HCA-7 and DLD-1 cells were incubated with or without NS-398 (10 μmol/L) in serum-free medium for the times indicated. A, levels of GTP-bound RhoA were measured by pull-down assays by incubation of lysates (1 mg) with GST-Rhotekin-RBD. Total lysates (15 μg) and pull-down samples were analyzed by Western blot using an anti-RhoA antibody. Representative experiment. The relative amount of active RhoA in each sample was quantified by densitometry and corrected for the ratio with total RhoA in the lysate. Columns, mean of three independent experiments; bars, SE. *, P < 0.05, significantly different from control. The P values were analyzed by post-hoc analyses using the Bonferroni's procedure. B, the levels of k-Ras-GTP from HCA-7 or DLD-1 cells treated with or without NS-398 were examined by pull-down assays by incubation of lysates (500 μg) with GST-Raf1-Ras binding domain. Total lysates (7.5 μg) and pull-down samples were analyzed by Western blot using an anti-k-Ras antibody. C, HCA-7 cells were incubated with or without SC-236 (10 or 25 μmol/L) for 24 hours. The activity of RhoA was measured using 1 mg lysate. Total lysates (15 μg) and pull-down samples were analyzed by Western blot. D, HCA-7 cells were incubated with or without NS-398 (10 μmol/L) or NS-398 plus PGE₂ (0.1 μmol/L) in serum-free medium for 24 hours. The activity of RhoA was measured using 1 mg lysates. Total lysates (15 μg) and pull-down samples were analyzed by Western blot using an anti-RhoA antibody. Representative data. The experiments were repeated twice (C and D).

Inhibition of COX-2 increases the formation of adherens junctions and levels of adherens junction proteins. It is suggested that high levels of active RhoA signal through ROCK to disrupt the localization of the adherens junction protein α-catenin and thereby cause loss of adherens junctions ( 6, 11). We have shown that HCA-7 colon carcinoma cells have high levels of RhoA activity and high levels of COX-2 protein. Thus, we compared adherens junctions in HCA-7, HT-29, and Caco-2 cells by examining the localization of E-cadherin and α-catenin ( Fig. 3A ). HT-29 and Caco-2 cells contained intact adherens junctions as shown by prominent staining of E-cadherin and α-catenin at cell-cell contacts and the merged images revealed colocalization of both proteins. In contrast, little staining of E-cadherin was detected at cell-cell contacts in HCA-7 cells and α-catenin was diffusely distributed within cells, suggesting formation of adherens junctions were disrupted in HCA-7 cells. Interestingly, colocalization of E-cadherin and α-catenin at adherens junctions was significantly stimulated when HCA-7 cells were treated with NS-398 or SC-236. Furthermore, we analyzed levels of E-cadherin and α-catenin in these colon carcinoma cell lines by Western blot analysis ( Fig. 3B). The protein levels of E-cadherin and α-catenin in HCA-7 cells were significantly lower than in Caco-2, HT-29, and DLD-1 cells. Interestingly, levels of E-cadherin and α-catenin in HCA-7 cells were dramatically stimulated when COX-2 was inhibited by treatment with NS-398 or when RhoA activity was inhibited by treatment with toxin A from Clostridium difficile ( Fig. 3B; ref. 38). However, treatment with NS-398 or toxin A had no effect on levels of E-cadherin and α-catenin in HT-29, Caco-2, and DLD-1 cells that contain low levels of RhoA activity and express no COX-2. Treatment with PGE₂ seemed to slightly reduce levels of E-cadherin in Caco-2, HT-29, and DLD-1 cells but had no significant effect on HCA-7 cells. In contrast, PGE₂ treatment only slightly decreased the levels of α-catenin in Caco-2 and DLD-1 cells but not HCA-7 and HT-29 cells. Our data suggest that constitutive expression of COX-2 and high levels of RhoA activity promote the disruption of adherens junction formation in HCA-7 cells by reducing the levels of E-cadherin and α-catenin.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

COX-2 and RhoA regulate adherens junctions. A, HCA-7, HT-29, and Caco-2 cells were plated at 5 × 10⁴ per well of a four-well chamber slide. At the next day, cells were placed in serum-free medium with or without NS-398 (10 μmol/L) or SC-236 (10 μmol/L) for 24 hours. After fixation and permeabilization, cells were stained for E-cadherin or α-catenin. B, HCA-7, Caco-2, HT-29, and DLD-1 cells were treated with NS-398 (N; 10 μmol/L), PGE₂ (E₂; 0.1 μmol/L), or toxin A (A; 50 ng/mL) in serum-free medium for 24 hours. Total lysates (15 μg) were analyzed by Western blot with anti-E-cadherin antibody or α-catenin antibody. The membrane was reprobed with β-actin antibody. Representative data. Experiments were repeated thrice (A and B).

Silencing of COX-2 or RhoA expression promotes adherens junctions and increases levels of adherens junction proteins. We tested if silencing of COX-2 or RhoA expression with siRNA would lead to stimulation of E-cadherin or α-catenin levels. Transfection of HCA-7 cells with 10 or 100 nmol/L synthetic COX-2 or RhoA siRNA for 48 to 72 hours significantly reduced the levels of COX-2 or RhoA protein compared with nontransfected cells or cells transfected with control siRNA ( Fig. 4A ). Silencing of COX-2 expression had no effects on levels of RhoA protein as well as silencing of RhoA expression had no significant effects on levels of COX-2 protein. However, RhoA activity was significantly reduced by silencing of COX-2 expression compared with cells transfected with control siRNA ( Fig. 4B), verifying our data using the specific COX-2 inhibitors and suggesting that constitutive expression of COX-2 leads to elevated RhoA activity ( Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

COX-2 and RhoA regulate levels of E-cadherin and the adherens junction complex. A, HCA-7 cells were transfected with COX-2 siRNA for 48 or 72 hours or with RhoA siRNA or control siRNA for 72 hours. Cell lysates (15 μg) were analyzed by Western blot with anti-COX-2 or anti-RhoA antibody. The membranes were reprobed with β-actin antibody. B, HCA-7 cells were transfected with control or COX-2 siRNA (100 nmol/L) for 72 hours. The levels of RhoA-GTP were measured by pull-down assays by incubation of cell lysates (500 μg) with GST-Rhotekin-RBD. Total lysates (15 μg) and pull-down samples were analyzed by Western blot using an anti-RhoA antibody. C, cell lysates (15 μg) from HCA-7 cells transfected with COX-2 siRNA (100 nmol/L), control siRNA (100 nmol/L), or RhoA siRNA (100 nmol/L) for 72 hours were analyzed by Western blot with anti-E-cadherin or α-catenin antibody. The membrane was reprobed with β-actin antibody. D, HCA-7 cells were transfected with control siRNA or COX-2 siRNA (100 nmol/L) for 72 hours. After fixation and permeabilization, cells were stained for E-cadherin and α-catenin. Representative data. The experiments were repeated twice (A–C).

The samples prepared from cells that were transfected with 100 nmol/L siRNA were also analyzed for E-cadherin and α-catenin expression by Western blot ( Fig. 4C). Consistent with the data of pharmacologic inhibition of COX-2 or RhoA ( Fig. 3), silencing of COX-2 or RhoA expression resulted in increased E-cadherin and α-catenin levels in HCA-7 cells. Furthermore, silencing of COX-2 expression in HCA-7 cells significantly increased colocalization of E-cadherin and α-catenin at adherens junctions compared with cells transfected with control siRNA ( Fig. 4D), verifying the data using the specific COX-2 inhibitors ( Fig. 3). Our data show for the first time that elevated RhoA activity plays an important role in suppression of E-cadherin and α-catenin levels in colon carcinoma cells. Because constitutive expression of COX-2 stimulates RhoA activity, RhoA seems to mediate COX-2 signaling to down-regulate levels of E-cadherin, leading to suppression of adherens junctions.

Inhibition of ROCK stimulates the formation of adherens junctions. It has been shown that expression of an active ROCK disrupts cell-cell junctions by altering localization of α-catenin or by reducing the total levels of both α-catenin and E-cadherin ( 11). Our data showed that elevated RhoA activity down-regulates levels of E-cadherin and α-catenin and leads to suppression of adherens junctions. Therefore, we examined whether ROCK activity is involved in down-regulating expression of E-cadherin and α-catenin as well as formation of adherens junctions. HCA-7, HT-29, and DLD-1 cells were treated with the specific ROCK inhibitor Y-27632 or COX-2 inhibitor NS-398. Because RhoA activity in both HT-29 and DLD-1 cells was low, inhibition of ROCK with Y-27632 for 24 hours had no effect on levels of E-cadherin or α-catenin ( Fig. 5A ). In agreement with the result of Fig. 3A, NS-398 treatment had no effects on levels of E-cadherin or α-catenin in HT-29 and DLD-1 cells that express no COX-2. In contrast, treatment with the ROCK inhibitor Y-26732 for 24 hours significantly increased levels of E-cadherin and α-catenin in HCA-7 cells to a similar extent as NS-398 treatment ( Fig. 5A). Furthermore, treatment of HCA-7 cells with the ROCK inhibitor Y-27632 promoted colocalization of E-cadherin and α-catenin to adherens junctions compared with control cells ( Fig. 5B). Treatment of HT-29 cells with the ROCK inhibitor Y-27632 had no significant effects on colocalization of E-cadherin and α-catenin to adherens junctions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Inhibition of ROCK increases levels of E-cadherin and α-catenin and promotes the formation of adherens junctions. A, HCA-7, HT-29, and DLD-1 cells were treated with or without ROCK inhibitor Y-27632 (Y; 10 μmol/L) or COX-2 inhibitor NS-398 (10 μmol/L) for 24 hours. Cell lysates (15 μg) were analyzed by Western blot with anti-E-cadherin or α-catenin antibody. The membranes were reprobed with β-actin antibody. B, HCA-7 or HT-29 cells were treated with or without ROCK inhibitor Y-27632 (10 μmol/L) for 24 hours. After fixation and permeabilization, cells were stained for E-cadherin and α-catenin. Representative data. Experiments were repeated thrice (A and B).

COX-2 and RhoA regulate cell motility. COX-2 and RhoA have been implicated in promoting cell motility and invasion. We found that HCA-7 colon carcinoma cells with high levels of constitutive COX-2 and elevated levels of RhoA activity migrate faster than HT-29 or Caco-2 colon carcinoma cells ( Fig. 6 ). Treatment of HCA-7 cells with the specific COX-2 inhibitor NS-398 or silencing of COX-2 expression with siRNA significantly decreased the motility. Silencing of RhoA expression with siRNA also reduced the motility. Our data suggest that constitutive expression of COX-2 or elevated RhoA activity increases motility.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Inhibition of COX-2 or RhoA reduces the motility of HCA-7 cells. HCA-7 cells were transfected with or without COX-2 or RhoA siRNA (100 nmol/L). At 48 hours after transfection, cells were labeled with CellTracker Orange. Cells (5 × 10⁴) were plated onto the FluoroBlok insert of a modified Boyden chamber. The lower chamber was filled with serum-free medium containing vehicles (DMSO) or NS-398 (10 μmol/L). After 24 hours of incubation, cells that migrated to the lower surface of the filter were visualized by fluorescence microscopy (magnification, ×200) and counted. Columns, average of three independent experiments with triplicate samples; bars, SE.

Discussion

Levels of COX-2 and COX-2-derived PGE₂ are elevated in various types of cancers, including colorectal cancer, and have been shown to promote tumor cell growth, motility, invasion, resistance to apoptosis, and angiogenesis ( 1, 2). Therefore, COX-2 plays a key role in tumorigenesis. However, the mechanisms by which COX-2 promotes tumor progression are not well understood. The goal of our present studies was to understand the molecular mechanisms by which COX-2 disrupts formation of E-cadherin-mediated adherens junction, leading to increased motility. Disruption of adherens junctions is associated with progression toward tumor malignancy and promotes tumor cell invasion and metastasis.

To avoid artifacts resulting form overexpression of COX-2, we used HCA-7 human colon carcinoma cells that constitutively express endogenous COX-2 and three other human colon carcinoma cell lines HT-29, Caco-2, and DLD-1 cells that do not express detectable levels of COX-2. It is suggested that activation of the RhoA/ROCK signaling pathway leads to loss of adherens junctions by disrupting the localization of α-catenin or by reducing the levels of both E-cadherin and α-catenin ( 11, 12). Therefore, we investigated whether RhoA mediates COX-2 signaling to promote the loss of adherens junctions, leading to invasive phenotypes. Here, we show that RhoA activity in HCA-7 cells is significantly higher than in the other three cell lines. Interestingly, inhibition of COX-2 with the specific inhibitor NS-398 or SC-236 or silencing of COX-2 expression by transfection of COX-2 siRNA results in a significant decrease of RhoA activity. Furthermore, inhibition of RhoA activity by treatment with NS-398 is reversed by PGE₂ treatment. Although activation of the RhoA/ROCK pathway has been linked to the PGE₂ receptor EP3 to induce neurite retraction of PC12 cells or contraction of guinea-pig aorta, no previous report has shown that COX-2 signaling regulates RhoA activity ( 39, 40). Therefore, our data show for the first time that constitutively high levels of COX-2 lead to elevated RhoA activity in colon carcinoma cells and PGE₂ seems to mediate COX-2 signaling to regulate RhoA activity.

To determine whether HCA-7 cells maintain adherens junctions, we examined the localization of the adherens junction proteins E-cadherin and α-catenin at cell-cell contacts. In contrast to HT-29 and Caco-2 cells, which both maintain adherens junctions, formation of adherens junctions is disrupted in HCA-7 as shown by very little staining of E-cadherin at cell-cell contacts and diffused distribution of α-catenin within the cells. Loss of adherens junctions in HCA-7 cells correlates with low levels of E-cadherin and α-catenin. Interestingly, inhibition of COX-2 activity significantly stimulates the formation of adherens junctions in HCA-7 cells and results in increased levels of E-cadherin and α-catenin. Selective knockdown of COX-2 expression with siRNA also leads to increased levels of E-cadherin and α-catenin and promotes formation of adherens junctions. Therefore, our data show that constitutive expression of COX-2 results in disrupted formation of adherens junctions by decreasing levels of E-cadherin and α-catenin. Our findings are consistent with previous studies showing that levels of E-cadherin are down-regulated in rat intestinal epithelial cells and human NSCLC cells that stably express COX-2 ( 13, 14). Furthermore, we show that inhibition of RhoA activity by toxin A or silencing of RhoA expression with siRNA increases levels of both E-cadherin and α-catenin in HCA-7 cells. Inhibition of ROCK activity also increases protein levels of E-cadherin and α-catenin and formation of adherens junctions in HCA-7 cells. Therefore, our data suggest that constitutive expression of COX-2 stimulates the RhoA/ROCK pathway that results in decreased expression of E-cadherin and α-catenin, leading to loss of adherens junctions. Formation of adherens junctions at epithelial cell-cell contacts greatly limits the ability of cells to move or migrate. Therefore, loss of adherens junctions increases motility of epithelial tumor cells ( 3). Our data show that inhibition of COX-2 or RhoA leads to decreased motility of HCA-7 cells, suggesting COX-2 and RhoA signaling pathways increase tumor cell motility by disruption of adherens junction formation.

Several studies suggest that a basal level of RhoA activity is necessary for adherens junction assembly, whereas high RhoA activity disrupts adherens junction assembly ( 3– 6). The molecular mechanisms by which increased expression of RhoA and subsequent signaling through ROCK lead to the loss of adherens junctions in colon carcinoma cells involve disrupted α-catenin localization at cell-cell contacts without affecting the protein levels of E-cadherin and α-catenin ( 11). In contrast, expression of an active form of ROCK in HEK293 cells disrupts adherens junctions by reducing the total levels of both E-cadherin and α-catenin ( 11). Our data suggest that the RhoA/ROCK pathway mediates COX-2 signaling to disrupt formation of adherens junctions by decreasing the protein levels of E-cadherin and α-catenin. Therefore, loss of adherens junctions by activation of the RhoA/ROCK pathway seems to involve at least two different mechanisms, reducing the total levels of E-cadherin and α-catenin and disrupting the localization of α-catenin at cell-cell junctions.

It has been suggested that Rac1 and Cdc42 activities are essential for the formation of adherens junctions ( 3– 6). On the other hand, E-cadherin-mediated formation of adherens junctions activates Rac1 and Cdc42 ( 3– 6). Therefore, when we compared the activities of Rac1 and Cdc42 in HCA-7 cells to HT-29 and Caco-2, which maintain adherens junctions at cell-cell contacts, both Rac1 and Cdc42 activities in HCA-7 cells are lower than in HT-29 and Caco-2 cells. Low activities of Rac-1 and Cdc42 in HCA-7 cells correlate with loss of adherens junction formation. Our data show that inhibition of COX-2 or RhoA in HCA-7 cells increases levels of E-cadherin and α-catenin and promotes the formation of adherens junctions but to a lower extent than in HT-29 and Caco-2 cells. This suggests that low activities of Rac1 and Cdc42 in HCA-7 cells down-regulate adherens junctions independent of COX-2 signaling. The COX-2/PGE₂ pathway has been shown to mediate the activation of Rac1 and Cdc42 in endothelial cells in response to α_vβ₃ integrin signaling ( 41). In contrast, our data show that constitutive expression of COX-2 does not activate Rac1 and Cdc42 in colon carcinoma cells, indicating the COX-2 pathway differentially regulates Rac1 and Cdc42 activities in different cell types. It has been suggested that Rho family proteins antagonize each other by cross-talk ( 4, 32, 42). Our results show that inhibition of RhoA does not increase Rac1 or Cdc42 activity, suggesting that elevated RhoA activity is not involved in down-regulation of Rac1 or Cdc42 activity in HCA-7 cells.

Formation of adherens junctions is also affected by oncogenic Ras, thus allowing transformed cells to migrate ( 43). We compared the k-Ras activity in HCA-7 cells to HT-29 and Caco-2 cells that both carry a WT k-Ras gene as well as DLD-1 cells that contain an oncogenic k-Ras mutant. As expected, no k-Ras activity is detected in both HT-29 and Caco-2 cells and significant high levels of k-Ras activity are found in DLD-1 cells. To our surprise, elevated levels of k-Ras activity are detected in HCA-7 cells and are not affected by inhibition of COX-2. This is in contrast to a previous study showing that low Ras activity is detected in HCA-7 cells and is further increased on PGE₂ treatment ( 44). The main differences are that we used 250 μg lysate and 80 μg GST-c-Raf-1, whereas the other study used 400 μg lysate and 20 μg GST-Raf-1 in the pull-down assays for Ras activation ( 44). We determined previously that 80 μg GST-Raf-1 is required to sufficiently pull down Ras-GTP from 500 μg GTPγS-treated NIH3T3 lysate ( 22). Because it is reported that HCA-7 cells contain WT Ras ( 44), elevated k-Ras activity in HCA-7 cells seems to be caused by an unknown constitutive signal.

In summary, we show for the first time that the RhoA/ROCK pathway mediates COX-2 signaling to promote loss of adherens junctions by suppressing the levels of the two major adherens junction proteins E-cadherin and α-catenin. Disruption of adherens junction function in most epithelial cell types results in a fundamental change in cellular phenotypes, such as increased migration and invasive growth, commonly associated with tumor progression ( 3). Understanding the signaling pathways downstream of COX-2 that lead to tumor invasion and metastasis is crucial for the development of novel therapeutic strategies.