We showed that the PEA3 transcriptional factor interacted with LKB1, a serine/threonine kinase, which is somatically mutated in lung cancer. This interaction occurred through the ETS domain of PEA3 and the kinase domain of LKB1. Mutation of LKB1 in lung cancer cells stabilized PEA3. Reintroduction of wild-type (WT) LKB1 into cells induced down-regulation of PEA3 and subsequently resulted in reduced cyclooxygenase-2 RNA and protein expression, whereas germ-line and somatic LKB1 mutants were defective in this activity. LKB1 phosphorylated PEA3 and promoted its degradation through a proteasome-mediated mechanism. Cells expressing mutant LKB1 possessed greater invasive potential compared with cells expressing WT LKB1. Increased invasion of cells with mutant LKB1 was partly due to PEA3 expression, as RNA interference inhibition of PEA3 resulted in dramatic decrease of Matrigel invasion. However, forced expression of PEA3 resulted in down-regulation of epithelial markers and induction of mesenchymal markers. These results suggest that PEA3 stabilization due to LKB1 inactivation could lead to epithelial/mesenchymal transition and greater lung cancer invasion potential. (Cancer Res 2006; 66(16): 7870-9)
Germ-line mutations in the serine-threonine kinase LKB1/STK11 result in Peutz-Jeghers syndrome (PJS; refs. 1, 2). PJS is an autosomal inherited disorder characterized by intestinal hamartoma, oral mucocutaneous hyperpigmentation, and increased risk for gastrointestinal and extraintestinal malignancies (3–5). Mutations of LKB1 result in complete loss of LKB1 protein in PJS (2). LKB1/STK11 inactivation in primary lung adenocarcinomas and in lung cancer cell lines is a common event (6). Therefore, the formation of hamartomas and tumor in PJS is mediated by inactivation of the remaining wild-type (WT) allele. LKB1 heterozygous mice were recently shown to develop hepatocellular carcinoma and intestinal polyposis (7, 8). Molecular characterization of the polyps showed that cyclooxygenase-2 (COX-2) was up-regulated through activation of extracellular signal-regulated kinases 1 and 2 (9). DNA chip microarray analysis of the mouse polyps (LKB1+/+ versus LKB1+/−) has shown that LKB1 can modulate factors linked to angiogenesis, extracellular matrix remodeling, cell adhesion, and inhibition of Ras-induced transformation (7).
To elucidate the unknown molecular mechanism of LKB1-mediated tumor suppression, we applied a yeast two-hybrid approach to identify LKB1-interacting proteins. Here, we showed that LKB1 binds, phosphorylates, and down-regulates PEA3-mediated induction of COX-2 RNA and protein. In addition, we showed that RNA interference (RNAi) knockdown of PEA3 transcription decreased the invasive potential of lung cancer cells, whereas forced expression of PEA3 induced epithelial/mesenchymal transition of lung epithelial cells.
Materials and Methods
Plasmids. The cDNA for hemagglutinin (HA)-tagged WT LKB1 was a gift from Dr. M. Sanchez-Cespedes (Spanish National Cancer Center, Madrid, Spain). To generate HA-LKB1 mutants (D194A, K44Stop, and Δ175-176) with impaired kinase activity [kinase dead (KD)], we used the QuikChange Mutagenesis kit according to the manufacturer's protocol (Stratagene, La Jolla, CA). pcDNA3.1 FLAG-tagged PEA3 expression vector was a kind gift from Dr. J. Hassell (McMaster University, Hamilton, Ontario, Canada). The pGL3-COX-2 promoter-luciferase reporter plasmid was a gift from Dr. C.C. Harris (National Cancer Institute, Frederick, MD). The COX-2 promoter deletion constructs were amplified using PCR (list of primers is available on request) and then subcloned into the pGL3-basic vector using the MluI and XhoI restriction sites (Promega, Madison, WI).
Yeast two-hybrid screen and mapping. To screen for the LKB1-interacting protein candidates, we used the LexA-based system (10). The full-length LKB1 cDNA was introduced into the LexA expression vector (pEG202) using the EcoRI and NotI restriction sites in frame with and downstream to LexA-binding domain. Because of strong self-activation of the LexA-LKB1 bait, we deleted the COOH terminus of LKB1 (residues 301-433). This deletion eliminated self-activation activity, and we therefore used this COOH-terminal truncated LKB1 bait (residues 1-300) to screen a human prostate cancer two-hybrid cDNA library based on pJG4-5 yeast expression vector (OriGene Technologies, Rockville, MD). For activation domain plasmids harboring various deletions of PEA3, we used pJG4-5. The full-length PEA3 cDNA was subcloned into the EcoRI and XhoI restriction sites of pJG4-5, and serial internal deletions were made using the QuikChange Mutagenesis kit according to the manufacturer's protocol. The yeast two-hybrid screens were done essentially as described elsewhere (10).
Cell cultures, transfections, and reporter assays. Human embryonic kidney cell line HEK293, lung cancer cell lines A427, A549, H1395, H1299, and H2095, and normal human bronchiolar epithelial (NHBE) cells were purchased from the American Type Culture Collection (ATCC; Manassas, VA) and grown in the recommended medium under standard conditions.
The expression cassettes for LKB1 and PEA3 were transiently transfected into cells using FuGENE 6 reagent (Roche Molecular Biochemicals, Indianapolis, IN). For luciferase assay, cells were plated in 24-well plates at a density of 1 × 105 per well. Cells were transfected for 24 hours with 0.1 μg of the pGL3-COX-2 promoter-luciferase plasmid, 0.25 μg of the pCDNA3.1 FLAG-PEA3 expression plasmid, and 0.75 μg of pcDNA3.1-HA-LKB1 expression plasmid into each well. To normalize data for transfection efficiency, cells were also transfected with 200 ng of the pRL-TK plasmid (Renilla luciferase). The COX-2 reporter activity was determined as described elsewhere (11, 12). Twenty-four hours after transfection, luciferase activity was monitored using the Dual-Luciferase Assay kit (Promega) in a Monolight TM-20 luminometer for 10 seconds. Three independent transfections were done, and calculated mean and SD were used for data presentation.
Western blot analysis and immunoprecipitation. For protein extraction, 5 × 105 cells per well were plated into six-well plates and transiently transfected with 0.5 μg of pcDNA3.1 FLAG-PEA3 expression plasmid and 1.5 μg of pcDNA3.1-HA-LKB1 expression plasmid. Eighteen hours after transfection, cells were incubated with or without proteasome inhibitor MG-132 (40 μmol/L) for 6 hours before protein extraction. Protein extraction and Western blot analysis were done as described previously (13). Primary antibodies against PEA3 (1:500), LKB1 (1:500), and HA (1:500) from Santa Cruz Biotechnology (Santa Cruz, CA), FLAG (1:2,000) and β-actin (1:2,000), both from Sigma (St. Louis, MO), and COX-2 (1:500; Cayman Chemical, Ann Arbor, MI) were used for protein detection and immunoprecipitation. For immunoprecipitation, HEK293 cells transfected with the indicated expression cassettes were lysed in buffer [50 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 5 μg/mL aprotinin, pepstatin, 1% NP40, 1 mmol/L EDTA, 0.25% deoxycholate]. Total cell lysates (500 μL) were incubated with 5 μg of primary antibodies against LKB1 or PEA3 for 16 hours at 4°C followed by subsequent incubation with anti-HA antibody conjugated to agarose beads (Roche) for 2 hours at 4°C, and the beads were then washed with lysis buffer and protein complexes were resolved by a 4% to 20% SDS-PAGE. Endogenous protein-protein interactions between PEA3 and LKB1 were examined in total cell lysate obtained from the breast cancer cell line Sk-Br3 (purchased from ATCC). For immunoprecipitation, lysates were incubated overnight with the indicated antibodies. Immune complexes were precipitated with protein A/G-Sepharose beads and washed with lysis buffer before being resolved on SDS-PAGE.
Immunofluorescence confocal imaging. For immunofluorescence microscopy, H1299 cells were grown on coverslips, fixed with 4% paraformaldehyde, and then incubated with anti-FLAG M2 monoclonal antibody (mAb) for 2 hours. Proteins were visualized by incubation with tetramethylrhodamine isothiocyanate (TRITC)–labeled goat anti-mouse immunoglobulins (IgG; Jackson ImmunoResearch, West Grove, PA). Finally, coverslips were incubated with Hoechst 33258 for 5 minutes and inspected with a laser scanning confocal microscope (Bio-Rad, Hercules, CA) at the Johns Hopkins University School of Medicine Confocal Imaging Core (Baltimore, MD).
Electrophoretic mobility shift assay. H1299 cells were resuspended in 10 mmol/L Tris-HCl (pH 7.5)/5 mmol/L MgCl2/0.05% (v/v) Triton X-100 and lysed with Dounce homogenizer. The homogenate was centrifuged at 3,000 × g for 15 minutes at 4°C. The nuclear pellet was resuspended in an equal volume of 10 mmol/L Tris-HCl (pH 7.4)/5 mmol/L MgCl2 followed by the addition of one nuclear pellet volume of 1 mol/L NaCl/10 mmol/L Tris-HCl (pH 7.4)/4 mmol/L MgCl2. The lysed nuclei were left on ice for 30 minutes and then spun down at 10,000 × g for 15 minutes at 4°C. The supernatant (nuclear extract) was kept, and protein concentration was measured with bicinchoninic acid protein assay (Pierce Biochemicals, Rockford, IL). Nuclear proteins (10 μg) were subjected to electrophoretic mobility shift assay (EMSA). An oligomer representing −140 to −52 of the COX-2 promoter (available on request) was radioactively labeled and used in EMSA analysis.
Northern blot analysis. Total cellular RNA was isolated using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. RNA (10 μg) from cells was denatured and loaded on a 1% formaldehyde-agarose gel. The RNA was transferred onto a Nytran membrane (Schleicher & Schuell, Keene, NH) using the TurboBlotter (Schleicher & Schuell) in 20× SSC buffer and subsequently UV cross-linked in a Stratalinker (Stratagene). The membrane was prehybridized and hybridized in 7 mL of hybridization buffer (0.25 mol/L Na2HPO4 plus 7% SDS) with 32P-labeled full-length cDNA probes for either PEA3 or COX-2 overnight at 65°C in hybridization buffer at constant rotation. After hybridization, the membrane was washed twice (30 minutes each) in 20 mmol/L Na2HPO4 and 5% SDS at 65°C and then washed twice in 20 mmol/L Na2HPO4 and 1% SDS (30 minutes each) at 65°C. Finally, the membrane wrapped in plastic paper was exposed to X-ray film (Kodak, Rochester, NY).
Small interfering RNA design and manipulation. Small interfering RNA (siRNA) oligonucleotides for LKB1 and scrambled siRNA oligonucleotides were purchased from Dharmacon (Lafayette, CO) and used according to the manufacturer's recommendation. Briefly, control and LKB1 siRNA oligonucleotides (200 pmol/six-well plate) were transiently transfected into H1299 lung cancer cells using FuGENE 6 (4 μL), and 24 hours later, total cell lysates were used for Western blot analysis as described above. For LKB1-RNAi4 plasmid, the oligonucleotide 5′-TCGAAAATGTCATCCAGCTGGTGGAGGAATTCGTCCACCAGCTGGATGACATTTTTTTT-3′ and its reverse complement with overhanging XbaI site were cloned into the SalI and XbaI site of the pSuppressor/Neo vector (Imgenex, San Diego, CA). The cloned oligonucleotide was sequenced to confirm the absence of mutations occurring during the cloning process.
Cell invasion/Matrigel assay. Cells (1 × 104) in 0.5 mL of serum-free MEM were added to each well of 24-well/8-μm pore invasion membrane chambers coated with Matrigel (BD Discovery Labware, Bedford, MA). The lower chambers contained 10% fetal bovine serum (FBS) in MEM to serve as a chemoattractant. Cells were allowed to migrate or invade over the course of 48 hours. Cells that failed to penetrate the filters were removed by scrubbing with cotton swabs. Chambers were fixed with 100% methanol for 2 minutes, stained with 0.5% crystal violet for 2 minutes, rinsed in water, and examined under a bright-field microscope. Values for invasion and migration were obtained by counting five fields per membrane (20× objective) and represented the average of three independent experiments done over multiple days.
In vitro kinase assay. Active His-tagged LKB1 protein (purified from recombinant baculovirus-infected Sf21 cells) was purchased from Upstate Cell Signaling Solutions (Waltham, MA). This enzyme (His-tagged LKB1, 51 kDa) is provided as a complex with glutathione S-transferase (GST)-MO-25α protein (66 kDa) serving as a control substrate for LKB1. Both proteins were phosphorylated by incubation with a kinase mixture. For bacterial expression, the full-length PEA3 cDNA was subcloned into the BamHI and XhoI sites of the pGEX-5X-2 bacterial expression vector (Amersham/Pharmacia, Inc., Piscataway, NJ). The resulting expression cassette, pGEX-5X-2-PEA3, was used to generate PEA3 point mutants (T355A, T363A, S393A, and S395A) using the QuikChange Mutagenesis kit according to the manufacturer's protocol. WT and point mutants of GST-PEA3 were purified from Escherichia coli using a GST MicroSpin Purification Module (Amersham/Pharmacia). Purified GST-PEA3 (0.1 μg) and LKB1 (0.1 μg) were mixed and incubated for 30 minutes at 30°C in kinase buffer (50 mmol/L Tris-HCl, 0.1 mmol/L EGTA, 0.1% 2-mercaptoethanol, 5 mmol/L MgCl2, 0.4 mmol/L MnCl2, 75 μg/mL bovine serum albumin, 20 μmol/L ATP, 10 μCi of [γ-32P]ATP). Samples were separated by SDS-PAGE and visualized by autoradiography.
Chromatin immunoprecipitation assay. Chromatin immunoprecipitation (ChIP) was done using a ChIP assay kit (Upstate Cell Signaling Solutions). H1299 cells (2 × 106) were transfected with mock (pcDNA3.1), pcDNA3.1 FLAG-PEA3, or pcDNA3.1 FLAG-PEA3 plus pcDNA3.1-LKB1 and cross-linked with 1% formaldehyde, and ChIP was done according to the manufacturer's protocol. After reversing the cross-links, chromatin was purified using the QIAquick PCR Purification kit (Qiagen, Inc., Valencia, CA) and samples were eluted with 30 μL of elution buffer. Chromatin was immunoprecipitated with 1 μg of anti-FLAG antibody M2 (Sigma). Immunoprecipitates were used as templates for PCR amplification of the COX-2 promoter with the following primers: COX-2, 5′-GCTTCCTGGGTTTCCGATTTTCT-3′ (sense) and 5′-GGTAGGCTTTGCTGTCTGAG-3′ (antisense). PCR was done as follows: 94°C for 1 minute, 55°C for 1 minute, and 72°C for 30 seconds using Platinum Taq I DNA polymerase (Invitrogen) for 37 cycles.
Tissue specimens. Human tissue samples were obtained from paraffin-embedded blocks from the tissue bank of the Department of Pathology at the Johns Hopkins School of Medicine (approved by the Joint Committee on Clinical Investigation). Lung frozen tissues were obtained from Dr. William Westra (Johns Hopkins University). Lung samples were homogenized in ice-cold nuclear buffer [150 mmol/L NaCl, 1 mmol/L KH2PO4 (pH 6.4), 1 mmol/L EGTA, 5 mmol/L MgCl2, complete mini protease cocktail (Roche)] at 4°C. Cells were then exposed to 0.3% Triton X-100, and nuclear pellets were spun down at 3,500 rpm for 20 minutes at 4°C. The nuclear pellets were mixed gently with an equal volume of nuclear buffer containing 0.4 mol/L NaCl at 4°C and spun down at 10,000 rpm for 20 minutes at 4°C. Protein concentration in supernatants was determined before Western blot analysis.
LKB1 associates with PEA3 in yeast. To characterize the molecular mechanism underlying the role for LKB1 in tumor suppression, we searched for LKB1-interacting proteins using the LKB1 kinase domain (residues 1-306) as bait. Because the LexA plasmid harboring the full-length LKB1 displayed strong self-activation activity, we deleted the COOH-terminal regulatory domain of LKB1 (residues 301-433). The resulting LKB1 COOH-terminal truncated bait was used to screen the two-hybrid cDNA library from human prostate cancer cells. From this screen, we identified human prostate-derived ETS factor (PDEF) as a highly probable LKB1-interacting partner. A homology search showed that PDEF has similarities to PEA3 belonging to the ETS family of transcriptional factors (14–16) and a known regulator of COX-2. To determine whether PEA3 interacts with LKB1, we cloned PEA3 into the prey plasmid pJG4-5 (activation domain) and did a yeast two-hybrid analysis using the LKB1 COOH-terminal truncation subcloned into the pEG202 plasmid (binding domain). As shown, the LKB1 COOH-terminal truncation (containing kinase domain) indeed physically associated with PEA3 (Supplementary Fig. S1).
We further mapped the interacting domains of both LKB1 and PEA3 using serial deletion of both proteins and yeast two-hybrid analysis (Fig. 1). We found that the ETS domain of PEA3 is absolutely necessary for the interaction with LKB1 (Fig. 1B). Deletion of the activation and COOH-terminal domains of PEA3 significantly reduced its interaction with LKB1, suggesting that these domains may cooperate with the ETS domain in transactivation. We also observed that mutations in the activation domain or the COOH-terminal domain of PEA3 dramatically reduced the interaction between PEA3 and LKB1 possibly due to significant changes in the tertiary structure of these PEA3 mutants (Fig. 1B).
To further examine the association of LKB1 and PEA3, we did coimmunoprecipitation of ectopically expressed proteins. HEK293 cells were transiently transfected with the HA-tagged LKB1 and FLAG-tagged PEA3 expression cassettes. Protein complexes were precipitated with the indicated antibodies and resolved electrophoretically followed by immunoblotting with antibodies to LKB1 or PEA3 (Fig. 1A). We found that LKB1 binds PEA3 through residues 150 to 200 that include the kinase domain (Fig. 1A). To explore the endogenous protein complexes between LKB1 and PEA3, we used Sk-Br3 breast cancer cells that express both LKB1 and PEA3. We observed that LKB1 was specifically precipitated with anti-PEA3 antibody but was not precipitated with irrelevant IgG (Fig. 1C). Reciprocally, anti-LKB1 antibody was able to immunoprecipitate PEA3 (Fig. 1C). To further analyze the PEA3/LKB1 association, we examined colocalization of these two proteins using confocal fluorescence imaging. H1299 lung cancer cells were transiently transfected with the green fluorescent protein (GFP)-LKB1 and FLAG-PEA3 fusion expression cassettes. Cells were stained for both GFP and FLAG markers with the indicated antibodies (Fig. 1D). Localization of LKB1 seems to be mostly nuclear, whereas PEA3 is localized in both nuclei and cytoplasm. Merging of these images clearly indicated that LKB1 and PEA3 colocalized (Fig. 1D , merge). These results suggest that LKB1 interacts with PEA3 in vitro and in vivo.
Functional relevance of the LKB1 and PEA3 interaction. To explore the functional relevance of the LKB1 and PEA3 interaction, we studied expression of PEA3 and COX-2 using Northern blot analysis. Lung cancer cells expressing WT or mutant LKB1 were used for these experiments. Overall, we found greater expression of both COX-2 and PEA3 in cells expressing mutant LKB1 than cells expressing WT LKB1 (Fig. 2A and B, respectively). These data suggested that the transcriptional activation of the COX-2 promoter was greater in cells with mutant LKB1 compared with cells with WT LKB1. Inactivation of LKB1 is expected to result in reduced protein expression of LKB1. Therefore, we examined expression of LKB1 in WT and mutant cells that have been studied previously (6, 17) and found that indeed LKB1 protein expression was reduced in mutant cells compared with LKB1 WT cells (Fig. 2C). Finally, we examined the expression of PEA3 and COX-2 in lung cancer tissues expressing WT and mutant LKB1. These tissues have been characterized for LKB1 status previously (6). Protein was extracted from these tissues and Western blotted with PEA3 and COX-2-specific antibodies. Results showed that, overall, tissues with LKB1 mutation showed greater expression of PEA3 and COX-2 (Fig. 2D). These results further support our hypothesis that mutation in LKB1 leads to increased expression of PEA3 and COX-2 in vivo.
Therefore, we examined the effect of LKB1 on COX-2 promoter activity. Lung cancer A427 cells were transiently transfected with the pGL3-COX-2 luciferase reporter plasmid and Renilla luciferase plasmid along with LKB1 and PEA3 expression cassettes. We found that WT LKB1 significantly reduced COX-2 reporter activity (Fig. 3A). However, LKB1 mutations found in lung adenocarcinoma (K44Ter; ref. 6) and PJS polyps (Δ175-176; ref. 18) were much less effective in down-regulating the COX-2 reporter activity (Fig. 3A). Similar results were obtained in H1299 lung cancer cells expressing WT LKB1 (Supplementary Fig. S2). Moreover, we found that WT LKB1 was able to down-regulate COX-2 reporter activity induced by exogenous PEA3 (Supplementary Fig. S2). The KD LKB1 mutant was completely ineffective in down-regulating the COX-2 reporter activity in both A427 and H1299 cells (Fig. 3A; Supplementary Fig. S2).
To determine the regulatory elements in the COX-2 promoter that were responsible for PEA3-mediated induction, a series of the COX-2 promoter deletion constructs were introduced into A427 cells and luciferase reporter activity was measured (Fig. 3B). We observed that PEA3-induced COX-2 promoter activity was completely abolished with the −52/+59 construct, suggesting that the sequence(s) critical for the promoter induction probably lies between −140 and −52. Furthermore, PEA3 was able to induce luciferase activity from a plasmid containing only the sequence −140 to −52 of the COX-2 promoter (Fig. 3C). Moreover, cotransfection of the WT LKB1 expression cassette into cells significantly decreased the COX-2 promoter activity, suggesting that down-regulation of COX-2 promoter by LKB1 is probably through PEA3 (Fig. 3C). To further explore the effect of LKB1 on PEA3 activity, we did an EMSA using the sequence −140 to −52 of the COX-2 promoter. We found that LKB1 was able to reduce the DNA-binding activity of PEA3 under these experimental conditions (Fig. 3D). We also showed that the PJS-associated LKB1 mutant (SL8) with an intact kinase domain but impaired regulatory domain dramatically reduced the DNA-binding activity of PEA3, suggesting that the kinase domain of LKB1 is critical for inhibiting PEA3 activity (Fig. 3D).
To further understand the molecular interaction between LKB1 and PEA3 on the COX-2 promoter, we did ChIP analysis using A427 lung cancer cells expressing mock vector or HA-tagged LKB1 (Fig. 3E). Chromatin was precipitated with a primary antibody against PEA3, and enriched specific DNA sequences in the chromatin immunoprecipitates were visualized by PCR amplification (Fig. 3E). We found that PEA3 selectively occupied the COX-2 promoter, whereas cotransfection of WT LKB1 dramatically decreased this activity (>3-fold; Fig. 3E). Negative control (absence of antibody) confirmed the specificity of these results (Fig. 3E , bottom).
LKB1 down-regulates PEA3 and COX-2. To further elucidate the mechanism underlying down-regulation of COX-2 transcription by LKB1 through PEA3, we hypothesized that LKB1 might promote PEA3 degradation. We thus examined the effect of LKB1 on both exogenous and endogenous PEA3 protein levels. We found that introduction of WT LKB1 into A427 cells (expressing mutant LKB1) led to a marked decrease in the amount of PEA3 protein level (Fig. 4A). However, LKB1 mutants (KD and Δ175-176) were dramatically less effective in down-regulating the PEA3 protein level (Fig. 4A). We also showed that WT LKB1 induced degradation of ectopically expressed FLAG-tagged PEA3 in H1299 lung cancer cells (Supplementary Fig. S3). It is expected that cells expressing WT LKB1 have intact ubiquitin pathway, which can be activated by ectopic LKB1.
To further confirm the down-regulation of PEA3 and COX-2 by endogenous LKB1, we screened for cells with high protein levels of PEA3 and COX-2 in WT LKB1 genetic background. First, we identified that treatment of Sk-Br3 breast cancer cells with 5-amino-4-imidazolecarboxamide riboside [AICAR; a cell-permeable drug activating AMP-activated protein kinase (AMPK), the downstream target for WT LKB1] led to phosphorylation of AMPK at the Thr172 (19), confirming that Sk-Br3 cells indeed express WT LKB1 (Fig. 4B,, 1). Second, we found that Sk-Br3 breast cancer cells express abundant amounts of both proteins (COX-2 and PEA3; Fig. 4B,, 2 and 3). We observed that a pharmacologic activation of LKB1 in Sk-Br3 cells resulted in abrogation of PEA3 and COX-2 expression (Fig. 4B , 2 and 3).
Next, we examined the role of the proteasome-dependent pathway in LKB1-mediated degradation of PEA3. We found that the 26S proteasome inhibitor MG-132 dramatically abrogated down-regulation of PEA3 and possibly ubiquitination of PEA3 in the presence of WT LKB1 (Fig. 4C and D). However, the pan-caspase inhibitor failed to inhibit LKB1-induced PEA3 degradation (Fig. 4C). To further examine ubiquitination of PEA3, we used H1299 cells transiently transfected with FLAG-tagged PEA3 alone or together with HA-tagged LKB1 (Fig. 4D). Twenty-four hours after transfection, PEA3 was immunoprecipitated with anti-FLAG antibody and resolved by SDS-PAGE. Western blot analysis with antibody against ubiquitin revealed a ladder of PEA3-specific high molecular weight bands, indicating ubiquitination of PEA3 (Fig. 4D).
To further examine the role of LKB1 in regulation of levels of PEA3 and COX-2, we used siRNA specific to LKB1 to selectively knock down its expression. We found that introduction of LKB1 siRNA into H1299 cells expressing WT LKB1 resulted in 80% reduction in the LKB1 protein level after 48 hours after transfection (Fig. 5). siRNA-mediated knockdown of LKB1 was accompanied by a dramatic increase in both PEA3 and COX-2 protein levels (Fig. 5), whereas scrambled siRNA oligonucleotides (control) had no effect (Fig. 5A). These results were confirmed using a vector-based LKB1 siRNA (LKB1-RNAi4; Fig. 5B), which showed a similar effect.
Because KD LKB1 was unable to down-regulate PEA3-induced COX-2, we explored the possibility that PEA3 protein degradation is associated with phosphorylation events induced by LKB1. To test this hypothesis, we did an in vitro kinase assay with GST-PEA3 as a substrate. AMPK, a well-known LKB1 substrate, was used as a positive control. We found that indeed LKB1 phosphorylated PEA3 and AMPK (Fig. 6A). The DNA-binding domain of PEA3 seemed necessary for its association with LKB1. Analysis of this domain in PEA3 revealed two threonine (Thr, T) and two serine (Ser, S) residues at amino acid 355, 363, 393, and 395, respectively. Therefore, we mutated each of these residues to alanine and did an in vitro kinase assay with these PEA3 mutants. We found that LKB1 was able to phosphorylate PEA3 mutants in vitro (T355A, T363A, and S393A; Supplementary Fig. S4). However, LKB1 failed to phosphorylate PEA3 mutant (S395A) in vitro (Fig. 6A). To verify that phosphorylation of Ser395 of PEA3 is important for its degradation by LKB1, we introduced expression cassettes for the PEA3-mutant S395A and WT LKB1 into H1299 lung cancer cells. Western blot analysis showed that the PEA3-mutant S395A could not be degraded by LKB1. We also found that the PEA3 point mutants (S393A and S395A; Fig. 6B) had greater protein stability compared with WT PEA3.
PEA3 mediates cellular invasion and epithelial/mesenchymal transition. COX-2 overexpression in lung cancer tissue has been frequently associated with increased invasion (20). To further assess the role of LKB1 in cellular invasion, we did the Matrigel invasion assay using lung cancer cell lines A427 (expressing mutant LKB1) and H1299 (expressing WT LKB1). We studied whether and how these cells respond to chemoattractant (10% FBS) >48 hours. We observed that A427 cells showed a higher degree of invasiveness compared with H1299 cells (Fig. 7A). However, forced expression of WT LKB1 in A427 cells led to ∼8-fold reduction in Matrigel invasion after 48 hours (Fig. 7A).
Next, we inhibited PEA3 transcription by stable siRNA expression in A549 cells (expressing mutant LKB1 and high levels of PEA3). We selected one stably transfected clone (clone 8) showing a ∼60% decrease in PEA3 protein level (Fig. 7A). Interestingly, this clone also showed ∼50% decrease in COX-2 expression compared with control cells, further confirming a role for PEA3 in COX-2 expression (Fig. 7A). We also showed that clone 8 was ∼40% less invasive in Matrigel assay than control cells (Fig. 7A).
Several studies have shown a correlation between the epithelial/mesenchymal transition and cancer invasion (21). It is well known that the epithelial/mesenchymal transition is associated with reduction in epithelial markers and increase in mesenchymal markers (22). We examined whether expression of PEA3 was able to promote an epithelial/mesenchymal transition in normal mammalian cells. We thus expressed PEA3 in NHBE epithelial cells. We found that forced expression of PEA3 resulted in decrease in epithelial markers (e.g., E-cadherin, β-catenin, and γ-catenin). In contrast, the expression of fibroblast markers (e.g., fibronectin, vimentin, and smooth muscle actin), whose expression has been shown to correlate positively with the epithelial/mesenchymal transition (22), was strongly induced by PEA3 overexpression (Fig. 7B). These results suggest that overexpression of PEA3 can induce epithelial/mesenchymal transition in normal human epithelial cells and mediate increased cellular invasion of lung cancer cells.
In this study, we provide evidence that PEA3, an ETS family transcription factor, physically and functionally interacts with the PJS tumor suppressor LKB1. First, our yeast two-hybrid assay indicated to us that PDEF bound LKB1 and would likely bind similar transcription factors. Moreover, COX-2 was implicated as an important target of LKB1 inactivation in human colon polyps. In an effort to tie in these two observations, we proceeded with a homology search to PDEF and discovered that PEA3 was a closely related and a relevant transcription factor for COX-2. This approach allowed us to link an important mediator of COX-2 with LKB1 inactivation. Second, we determined that binding to LKB1 is mediated by DNA-binding domain of PEA3, whereas interaction with PEA3 is mediated by the kinase domain of LKB1. Third, we observed that the physical association with LKB1 induced phosphorylation of PEA3 at S395 and targeted PEA3 into a ubiquitination and proteasome-dependent degradation pathway. Fourth, we found that LKB1-mediated degradation of PEA3 led to transcriptional down-regulation of COX-2 gene expression. Finally, we showed that LKB1-mediated degradation of PEA3 reduced invasiveness of lung cancer cells and forced expression of PEA3 induced an epithelial/mesenchymal transition of normal epithelial cells.
This work thus provides a mechanistic link between LKB1 tumor suppressor and PEA3 transcriptional activation of COX-2 gene expression. As a member of the ETS transcription factor family, PEA3 was shown to play a role in tumor progression. PEA3 was shown to be up-regulated in human breast cancer (14) and primary and metastatic lesions of mouse mammary carcinoma (23). Moreover, PEA3 protein expression predicts worse overall survival in breast cancer (24). Forced expression of PEA3 results in enhanced motility and invasion of non–small cell lung cancer cells (25). PEA3 factors are also shown to be highly expressed in tumors from Wnt1 transgenic mice, in which COX-2 is also up-regulated (26, 27).
Deregulation of COX-2 signaling is ubiquitous in human cancers (28, 29). COX-2, an inducible enzyme involved in prostaglandin biosynthesis, is overexpressed in several epithelial malignancies, including breast, prostate, lung, colorectal, esophageal, and ovarian human cancers (30–34). It was shown that COX-2 overexpression in human cancer cells enhances cell motility and invasiveness, thus suggesting a mechanism of COX-2-mediated metastasis. COX-2 is overexpressed in PJS hamartomas (35) and lung adenocarcinomas (36). The COX-2 overexpression was also associated with shortened survival in patients with resected early-stage adenocarcinoma of the lung (36). COX-2 inhibition decreases tumor cell proliferation in vivo and has been shown to enhance tumor radiosensitivity (37).
Forced expression of COX-2 in cancer cells produces prostaglandins and proangiogenic factors [e.g., vascular endothelial growth factor (VEGF)], whereas aspirin and antibodies against VEGF abrogate angiogenesis (38). COX-2 overexpression has been shown in PJS (39) and in LKB1 heterozygous mice (9), suggesting a role for LKB1 in regulation of COX-2.
In our study, we showed that ectopic expression of LKB1 suppressed cellular invasion of lung cancer cells. In addition, cells expressing mutant LKB1 (A427) showed greater invasive potential compared with cells expressing WT LKB1 (H1299), although no difference was found in the rate of cellular proliferation. Clearly, pathways that increase cellular invasion do not necessarily increase cellular proliferation, although interactions between these pathways cannot be excluded. LKB1 has been shown to regulate mammalian target of rapamycin, which plays a role in cancer development and progression. Therefore, it is expected that mutations in LKB1 lead to activation of multiple pathways that play a role in cellular invasion. Our proposed mechanism detailing the role of LKB1 in regulating PEA3-mediated invasion is one component of this machinery as pointed out by the author. Further validation of these results will require studies in more cell lines, in vivo studies, and additional observations on LKB1 downstream effects, which are outside the scope of this present study.
Invasive potential of LKB1-mutant cells is, at least in part, mediated by PEA3 because most cells with down-regulation of PEA3 levels showed decreased cellular invasion and activation of epithelial/mesenchymal transition. PEA3 expression in human cancers (e.g., breast tumors) was found to correlate with down-regulation of E-cadherin and increased expression of matrix metalloproteinases (MMP-2 and MMP-7) and increased invasiveness of cancer cells (16). E-cadherin is critical for maintenance of cell polarity and differentiation (40). However, a decrease in E-cadherin expression leads to defects in cell-cell adhesion and, therefore, to tumor cell dissemination (41).
There is also evidence that K-Ras may regulate COX-2. Mutations in K-Ras are found in ∼30% of lung adenocarcinomas and associated with poor survival (42). Human non–small cell lung cancers with mutations in K-Ras have high expression of COX-2 (43). Expression of mutant K-Ras activates the Raf/mitogen-activated protein kinase pathway and results in increased transcription of COX-2 (44). Oncogenic K-Ras also activates the phosphatidylinositol 3-kinase pathway and results in post-transcriptional stabilization of COX-2 mRNA (45). LKB1 and K-Ras pathways are most likely additive because several lung cancer cell lines harbor mutations in both of these genes (17). Taken together, LKB1 or K-Ras mutation in lung cancer can lead to PEA3 and COX-2 up-regulation, resulting in resistance to apoptosis, increased angiogenesis, and increased invasion and metastasis. Because PEA3 is a major transcriptional regulator of COX-2 expression and probably other oncogenic targets in lung adenocarcinomas, PEA3 itself in addition to COX-2 may be a viable therapeutic target for human cancers with LKB1 mutations. In addition to in vitro and clinical studies of COX-2 inhibitors in colon polyps, nimesulide (a COX-2 inhibitor) can inhibit proliferation of non–small cell lung cancer cell lines (38). Our data suggest that similar approaches deserve further attention, particularly in lung cancers with LKB1 mutations.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: Lung Cancer Specialized Program of Research Excellence grant CA-058184-11 (D. Sidransky).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. Russell Finley for help with yeast two-hybrid assay, Dr. M. Sanchez-Cespedes for providing pcDNA3.1-HA-LKB1 plasmid, Dr. J. Hassell for the pcDNA3.1 FLAG-tagged PEA3 expression vector, Drs. Schlattner and Neumann for WT AMPK, Dr. C.C. Harris for pGL3-COX-2 promoter-luciferase reporter plasmid, and Dr. Venu Raman for helpful discussions.