|Home | About | Journals | Submit | Contact Us | Français|
Extracellular signal-regulated kinase (ERK) is important for various cellular processes, including cell migration. However, the detailed molecular mechanism by which ERK promotes cell motility remains elusive. Here we characterize epithelial protein lost in neoplasm (EPLIN), an F-actin cross-linking protein, as a novel substrate for ERK. ERK phosphorylates Ser360, Ser602, and Ser692 on EPLIN in vitro and in intact cells. Phosphorylation of the C-terminal region of EPLIN reduces its affinity for actin filaments. EPLIN colocalizes with actin stress fibers in quiescent cells, and stimulation with platelet-derived growth factor (PDGF) induces stress fiber disassembly and relocalization of EPLIN to peripheral and dorsal ruffles, wherein phosphorylation of Ser360 and Ser602 is observed. Phosphorylation of these two residues is also evident during wound healing at the leading edge of migrating cells. Moreover, expression of a non-ERK-phosphorylatable mutant, but not wild-type EPLIN, prevents PDGF-induced stress fiber disassembly and membrane ruffling and also inhibits wound healing and PDGF-induced cell migration. We propose that ERK-mediated phosphorylation of EPLIN contributes to actin filament reorganization and enhanced cell motility.
Extracellular signal-regulated kinase (ERK), a member of the mitogen-activated protein kinase (MAPK) family, plays pivotal roles in diverse cellular events, such as proliferation, differentiation, migration, growth, and survival (4, 18, 30, 44). Activation of ERK occurs in response to growth factor stimulation through the Ras-Raf-MEK pathway, and activated ERK translocates from the cytoplasm to the nucleus, where it phosphorylates several protein kinases, nuclear transcription factors, and other proteins (12, 17, 30). In addition to its role in the nucleus, recent data show that ERK is involved as an essential component in the migration of cells from many different organisms (9, 16, 21, 38). Certain substrates, such as myosin light chain kinase (25), focal adhesion kinase (14), paxillin (19), actopaxin (5), calpain (10), and vinexin (24), are known to function in ERK-mediated cell migration (13).
Cell migration requires dynamic reorganization of the actin cytoskeleton (31). Composite extracellular stimuli, including growth factors and cell-matrix adhesions, trigger signals for cell motility, which are then transduced by diverse intracellular components, such as the MAPK family (13, 39), protein kinase B/Akt (36), tyrosine kinases (6), and Rho family small GTPases (8, 34). During dynamic remodeling of the actin system for cell migration, a number of actin cross-linking/bundling proteins are crucial (1, 37, 40, 46). In addition, actin bundles and cross-linked networks play key roles in the generation of tension and flexibility in the actin cytoskeleton (2, 33). Thus, ERK might mediate cell migration via phosphorylating some actin cross-linking/bundling proteins.
Epithelial protein lost in neoplasm (EPLIN) was originally identified as the product of a gene that is transcriptionally down-regulated or lost in a number of human epithelial tumor cells, including oral, prostate, and breast cancer cell lines (3, 22). EPLIN is expressed from a single gene as two isoforms, α and β, the latter of which has an extra N-terminal sequence of 160 amino acids. Both EPLINα and -β contain a centrally located LIM domain that may mediate self-dimerization and N- and C-terminal actin-binding sites flanking the LIM domain (23). EPLIN cross-links and bundles actin filaments, thereby stabilizing actin stress fibers. Furthermore, EPLIN inhibits Arp2/3 complex-mediated branching nucleation of actin filaments. Thus, EPLIN controls actin filament dynamics by stabilizing actin filament networks (23). It is therefore assumed that the loss of EPLIN expression in cancer cells is involved in the enhanced motility of these cells.
Recently, we identified EPLIN as a candidate substrate for ERK by a proteomic approach using two-dimensional difference gel electrophoresis (2D-DIGE) combined with phosphoprotein enrichment. In this study, we show that ERK phosphorylates EPLIN in vitro and in vivo. Phosphorylation of the C-terminal region of EPLIN inhibits its actin-binding activity. Stimulation with platelet-derived growth factor (PDGF) induces stress fiber disassembly and localization of phosphorylated EPLIN to peripheral and dorsal ruffles. Furthermore, expression of a non-ERK-phosphorylatable mutant of EPLIN prevents PDGF-induced membrane ruffling as well as cell migration. These results suggest that phosphorylation of EPLIN by ERK leads to reorganization of actin filaments and stimulation of cell motility.
ΔB-Raf:ER cells (NIH 3T3 cells expressing the B-Raf kinase domain fused to the estrogen receptor ligand binding domain) (32) were cultured in Dulbecco's modified Eagle's medium (DMEM) without phenol red (Invitrogen, Carlsbad, CA) but containing 10% fetal bovine serum (FBS). NIH 3T3 cells were cultured in DMEM containing 10% calf serum (CS). 293T and HeLa cells were cultured in DMEM containing 10% FBS. Primary calvarial osteoblasts were isolated from 1-day-old Jc1:ICR mice by five sequential digestions (for 10 min each) with 0.1% collagenase and 0.2% dispase. The cells from the last four digestions were grown in α-minimum essential medium (Invitrogen) containing 10% FBS. Transfections were performed by using Lipofectamine 2000 (Invitrogen) for 293T and ΔB-Raf:ER cells and Lipofectamine LTX (Invitrogen) for NIH 3T3 and osteoblastic cells, according to the manufacturer's instructions.
The DNA fragments encoding mouse EPLINα (161-753), EPLINβ (1-753), EPLIN-N (161-387), and EPLIN-C (440-753) were amplified by PCR and cloned into pCMV-Tag3-Myc (Stratagene, La Jolla, CA), pCMV-Tag2-Flag (Stratagene), or pGEX-4T-3 vector (GE Healthcare, Buckinghamshire, United Kingdom). For expression of enhanced green fluorescent protein (EGFP)-fused EPLINα in mammalian cells, the DNA fragment encoding EGFP was amplified by PCR and cloned into the C-terminal coding region of pCMV-Tag3-Myc-EPLINα. Point mutations were introduced using a QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions.
A rabbit polyclonal antibody (PAb) against mouse EPLIN was generated against glutathione S-transferase (GST)-fused full-length EPLINα expressed in Escherichia coli BL21-CodonPlus (DE3)-RIPL (Stratagene) and was affinity purified with immobilized EPLINα, from which the GST moiety was removed by thrombin digestion. Anti-phosphorylated-EPLIN antibodies were raised by immunizing rabbits with keyhole limpet hemocyanin-conjugated synthetic phosphopeptides corresponding to 11-amino-acid sequences of EPLINα and were purified from antiserum as the bound fraction of a phosphopeptide-conjugated SulfoLink column (Pierce, Rockford, IL) and the unbound fraction of a non-phosphopeptide-conjugated column. The following antibodies were also used: anti-Myc mouse monoclonal antibody (MAb) 9E10, anti-Myc rabbit PAb A-14, anti-ERK1 rabbit PAb K-23 (Santa Cruz Biotechnology, Santa Cruz, CA), anti-p-ERK mouse MAb E10, anti-p-RSK (Thr573) rabbit PAb (Cell Signaling Technology, Danvers, MA), antiactin mouse MAb (Chemicon, Temecula, CA), anti-Flag mouse MAb M2 (Sigma, St. Louis, MO), anti-EPLIN rabbit PAb BL1141 (Bethyl, Montgomery, TX), anti-GFP rabbit PAb (Invitrogen), and antihemagglutinin (anti-HA) rat MAb 3F10 (Roche, Basel, Switzerland). PDGF and 4-hydroxy-tamoxifen (4-HT) were obtained from Sigma. U0126 was purchased from Promega (Madison, WI).
Myc-EPLINβ was transfected into ΔB-Raf:ER cells. Cells were then treated with 4-HT for 2 h, and cell lysates were immunoprecipitated with an anti-Myc (9E10) antibody. Immunoprecipitates were resuspended in a reaction buffer containing 4 units of calf intestinal alkaline phosphatase (CIAP; Takara, Shiga, Japan) and incubated at 37°C for 60 min. The reaction was stopped by adding Laemmli's sample buffer and boiling the samples. Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and subjected to immunoblotting with an anti-Myc (A-14) antibody.
EPLIN small interfering RNA (siRNA) and siCONTROL nontargeting siRNA 1 were obtained from Dharmacon (Lafayette, CO). The sequences of siRNA duplexes that target mouse EPLIN are as follows: sense, 5′-GGACGAAUCUACUGUAAGCUU-3′; and antisense, 5′-GCUUACAGUAGAUUCGUCCUU-3′. ERK1 and ERK2 Stealth siRNA duplexes were obtained from Invitrogen. The sequences of mouse ERK1 siRNAs are as follows: sense, 5′-GGAAGCCAUGAGAGAUGUUUACAUU-3′; and antisense, 5′-AAUGUAAACAUCUCUCAUGGCUUCC-3′. The sequences of mouse ERK2 siRNAs are as follows: sense, 5′-GGCUAAAGUAUAUCCAUUCAGCUAA-3′; and antisense, 5′-UUAGCUGAAUGGAUAUACUUUAGCC-3′. These siRNA duplexes were transfected into NIH 3T3 or primary osteoblastic cells by using DharmaFECT 1 reagent (Dharmacon), and cells were cultured for 72 h. For rescue assays, we constructed an RNA interference (RNAi)-refractory EPLINα cDNA (EPLINαr) and EPLINα(S360/602/692A) cDNA [EPLINαr(S360/602/692A)]. Three silent mutations were introduced into the mouse EPLINα and EPLINα(S360/602/692A) cDNAs, changing the nucleotide sequence at positions 817 to 825 of EPLINα/EPLINα(S360/602/692A) to CGCATATAT.
Phosphorylation of recombinant GST-EPLINα, GST-EPLIN-N, GST-EPLIN-C, and their Ala substitutes by ERK was performed by incubation of 50 ng of recombinant active ERK2 (New England Biolabs, Beverly, MA) with 3.0 μg each of GST-EPLINα, -N, -C, and their mutants and 50 μM [γ-32P]ATP (2.5 μCi; GE Healthcare) in 30 μl of a kinase buffer (50 mM Tris-HCl, pH 7.5, 15 mM MgCl2, 1 mM EGTA, and 2 mM dithiothreitol [DTT]) for 20 min at 30°C. The reaction was stopped by adding Laemmli's sample buffer and boiling the samples. Half of the sample was subjected to 10% SDS-PAGE, and the phosphorylation reaction was visualized by autoradiography.
GST-EPLINα phosphorylated by ERK in vitro was separated by SDS-PAGE. In-gel digestion was performed using sequencing-grade trypsin (Promega) or endoproteinase Glu-C (Roche). The resulting peptides were separated by C18 reversed-phase high-pressure liquid chromatography (LC), and each peptide was analyzed with a matrix-assisted laser desorption ionization-time-of-flight tandem mass spectrometer (model 4700 proteomics analyzer; Applied Biosystems, Foster City, CA). Detected masses and peptide sequences were subjected to database searches with the Mascot search engine (Matrix Science, London, United Kingdom).
Binding of EPLIN to F-actin was tested in a cosedimentation assay as described previously (23). Briefly, rabbit muscle G-actin (Cytoskeleton, Denver, CO) and GST-EPLIN-C or GST-EPLIN-C(S602/692A) were separately precleared by centrifugation at 100,000 × g for 30 min at 4°C. G-actin (2.5 μM) was polymerized in 5 mM Tris-HCl, pH 7.5, 100 mM KCl, 2 mM MgCl2, 0.2 mM ATP, and 0.5 mM DTT at room temperature for 30 min. Fifty microliters of F-actin was then incubated with 10 μl of ERK-phosphorylated or nonphosphorylated GST-EPLIN-C or GST-EPLIN-C(S602/692A) for 30 min at 4°C. After being centrifuged at 100,000 × g for 30 min at 4°C, the supernatant and pellet were separated and analyzed by SDS-PAGE and Coomassie brilliant blue (CBB) staining.
Cells were lysed with immunoprecipitation buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM NaF, 25 mM β-glycerophosphate, 2 mM EGTA, 2 mM MgCl2, 1% NP-40, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 20 μg/ml aprotinin, and 2 mM DTT) for 15 min on ice. Lysates were clarified by centrifugation and incubated with agarose beads conjugated with the 9E10 anti-Myc antibody for 1 h at 4°C. The beads were then washed three times with immunoprecipitation buffer and finally resuspended in Laemmli's sample buffer. Proteins were resolved by SDS-PAGE for immunoblot analysis.
Cells were grown on coverslips coated with poly-l-lysine and fixed with 3.7% formaldehyde for 10 min at room temperature. Fixed cells were then permeabilized with 0.1% Triton X-100 for 10 min. After being washed with phosphate-buffered saline, the cells were incubated with primary antibodies in phosphate-buffered saline containing 2% goat serum for 2 h, followed by incubation with Alexa Fluor-conjugated secondary antibodies (1:1,000 dilution; Invitrogen) for 1 h. F-actin was detected by staining with rhodamine-phalloidin or Alexa Fluor 647-conjugated phalloidin (Invitrogen). Samples were observed on an inverted microscope (model IX71; Olympus, Tokyo, Japan) equipped with a PlanApo 60×, 1.4-numerical-aperture (NA) oil immersion objective. Images were obtained with a cooled charge-coupled device camera (ORCA-ER; Hamamatsu Photonics, Shizuoka, Japan) controlled by Aqua-Lite software (Hamamatsu Photonics) and were processed using Adobe Photoshop CS3.
Cell migration was assayed in Boyden chambers (8.0-μm-pore-size polyethylene terephthalate membrane with Falcon cell culture insert; Becton Dickinson, Mountain View, CA) as previously described (7). NIH 3T3 cells transfected with EGFP, EPLINα-EGFP, or EPLINα(S360/602/692A)-EGFP were serum starved with DMEM containing 0.2% CS and then trypsinized and counted. Cells (5 × 104 to 10 × 104) in DMEM containing 0.2% CS (0.8 ml) were added to the upper chamber, and 1.8 ml of appropriate medium, with or without 30 ng/ml PDGF, was added to the lower chamber. When the MEK inhibitor was used in this assay, cells were treated with 20 μM U0126 for 30 min before trypsinization, and U0126 was also added to both the upper and lower chambers during migration. For RNAi rescue assays, cells were sequentially transfected with control siRNA or EPLIN siRNA and EGFP, EPLINαr-EGFP, or EPLINαr(S360/602/692A)-EGFP. For primary osteoblasts, α-minimum essential medium with 0.2% FBS was used. Transwells were incubated for 6 h at 37°C. EGFP-positive cells on both sides of the membrane were counted, and then cells on the inside of the insert were removed with a cotton swab and EGFP-positive cells on the underside of the insert were counted. The number of cells in five randomly chosen fields per filter was counted by microscopic examination.
To observe PDGF-induced membrane ruffling, NIH 3T3 cells transfected with EPLINα-EGFP or EPLINα(S360/602/692A)-EGFP were stimulated with 50 ng/ml PDGF for 15 min after time-lapse recording. Time-lapse microscopy was performed using a DeltaVision deconvolution microscope system controlled by softWoRx software (Applied Precision, Issaquah, WA) configured around an Olympus IX70 inverted microscope. Images were acquired using a UPlanSApo 20×, 0.85-NA oil immersion objective. For wound-healing assays, after scratching of a monolayer of NIH 3T3 cells transfected with EPLINα-EGFP or EPLINα(S360/602/692A)-EGFP, live cells were recorded at 37°C in a 5% CO2 atmosphere, using a confocal microscope (CSU22; Yokogawa, Tokyo, Japan) equipped with a cooled charge-coupled device camera (DV887DCS-BV; Andor Technology, Belfast, Northern Ireland). Images were acquired using a UPlanApo 20×, 0.80-NA oil immersion objective and were analyzed with MetaMorph software (Molecular Devices, Downingtown, PA).
To globally identify protein kinase substrates, we recently developed a system consisting of phosphoprotein purification by immobilized metal-affinity chromatography, fluorescent 2D-DIGE, and MS protein identification (20, 41). We applied this method to the ERK signaling pathway and identified 24 candidates for novel ERK targets, one of which was EPLIN (H. Kosako, N. Yamaguchi, M. Ushiyama, E. Nishida, and S. Hattori, submitted for publication).
To examine whether EPLIN is phosphorylated by ERK in living cells, Myc-tagged EPLIN was expressed in ΔB-Raf:ER cells. This cell line is a derivative of NIH 3T3 cells in which the protein kinase domain of mouse B-Raf is expressed as a fusion protein with the hormone-binding domain of the human estrogen receptor (32). ERK can be activated by 4-HT, an antagonist of estrogen. To suppress the ERK pathway, the MEK inhibitor U0126 was used. The lysates of these cells were subjected to immunoblotting with anti-Myc antibody (Fig. (Fig.1A,1A, left panel). Both Myc-EPLINα and Myc-EPLINβ showed mobility shifts on SDS-PAGE upon treatment with 4-HT compared to treatment with U0126. To confirm phosphorylation as the cause of these shifts, we examined the effect of phosphatase treatment. Myc-EPLINβ was immunoprecipitated from 4-HT-treated ΔB-Raf:ER cells, and the immunoprecipitates were incubated with CIAP. As shown in Fig. Fig.1A1A (right panel), the 4-HT-induced band shift of Myc-EPLINβ was completely reversed by CIAP treatment. These results suggest that EPLIN is phosphorylated by the activation of the ERK pathway. To determine the phosphorylation site on EPLIN that induces the mobility shift, HA-tagged wild-type EPLINβ and two Ser-to-Ala mutants were expressed in ΔB-Raf:ER cells, and then the cells were treated with 4-HT or U0126. HA-EPLINβ-S360A did not show the mobility shift (see Fig. S1 in the supplemental material), indicating that the shift was due to the phosphorylation of Ser360 (see below).
We generated a polyclonal antibody by immunizing a rabbit with bacterially expressed full-length mouse EPLINα fused to GST. To examine the specificity of the generated antibody, Flag-tagged EPLIN was expressed in NIH 3T3 cells. Both Flag-EPLINα and Flag-EPLINβ were detected by immunoblotting with our anti-EPLIN, commercially available anti-EPLIN (BL1141), and anti-Flag antibodies (Fig. (Fig.1B).1B). The commercial anti-EPLIN (BL1141) antibody showed very little reactivity to endogenous EPLINα (Fig. (Fig.1B,1B, middle panel). In contrast, our affinity-purified anti-EPLIN antibody specifically recognized endogenous EPLINα and also EPLINβ, as a faint band (Fig. (Fig.1B,1B, left panel, and C). Since EPLINβ includes the entire sequence of EPLINα, EPLINβ is thought to be expressed at a much lower level than EPLINα in NIH 3T3 cells. When ΔB-Raf:ER cells were treated with 4-HT or when NIH 3T3 cells were stimulated with PDGF for 30 min, endogenous EPLINα and -β were phosphorylated, and this was prevented by pretreatment with U0126 (Fig. (Fig.1C1C).
As described in the previous section, EPLIN is phosphorylated upon ERK activation. To test whether EPLIN is a direct substrate of ERK, we prepared GST fusion proteins of full-length EPLINα and the N-terminal and C-terminal portions of EPLINα, as illustrated schematically in Fig. Fig.2A.2A. An in vitro kinase assay was then performed, using recombinant active ERK, [γ-32P]ATP, and recombinant GST-EPLINα, GST-EPLIN-N, and GST-EPLIN-C as substrates. Both GST-EPLINα and GST-EPLIN-C were strongly phosphorylated by ERK, whereas GST-EPLIN-N phosphorylation was rather weak (Fig. (Fig.2B,2B, lanes 1, 3, and 6). It has been established that ERK preferentially phosphorylates Ser or Thr residues just before Pro residues (11). EPLIN has seven Ser-Pro sequences that are conserved between mouse and human EPLIN proteins (Fig. (Fig.2A).2A). To identify ERK phosphorylation sites on EPLIN, we replaced each Ser residue with Ala. As shown in Fig. Fig.2B,2B, lane 4, the S360A substitution completely abolished ERK phosphorylation of EPLIN-N. On the other hand, replacement of either Ser602 or Ser692 by Ala partially abolished phosphorylation, and replacement of both residues (EPLIN-C-S602/692A) markedly reduced phosphorylation (Fig. (Fig.2B,2B, lanes 9, 11, and 12). When full-length EPLINα was used as a substrate, replacement of Ser360, Ser602, and Ser692 by Ala (EPLINα-S360/602/692A) strongly impaired phosphorylation by ERK (Fig. (Fig.2B,2B, lane 2). This suggests that Ser360, Ser602, and Ser692 are the primary sites at which ERK phosphorylates EPLIN in vitro.
To further confirm the phosphorylation sites, EPLIN phosphorylated by ERK in vitro was digested with trypsin or V8 protease, and the resulting peptides were subjected to LC-MS/MS analysis. As shown in Fig. S2 in the supplemental material, phosphorylation of Ser360, Ser602, and Ser692 was confirmed by this analysis. Phosphorylation of Ser488 and Ser607 was also observed, and these may be minor phosphorylation sites, as suggested by the slightly reduced phosphorylation of EPLIN-C-S488A and -S607A (Fig. (Fig.2B,2B, lanes 8 and 10).
We then produced phospho-specific antibodies by using synthetic phosphopeptides that harbor phosphorylated Ser360, Ser602, or Ser692. The anti-pS360 antibody recognized wild-type GST-EPLIN-N and the S372A mutant of GST-EPLIN-N upon ERK-mediated phosphorylation but did not recognize the S360A mutant (Fig. (Fig.2C).2C). Similarly, anti-pS602 and anti-pS692 antibodies recognized wild-type and S692A mutant GST-EPLIN-C and wild-type and S602A mutant GST-EPLIN-C, respectively, only when phosphorylated by ERK. These results indicate that the anti-pS360, anti-pS602, and anti-pS692 antibodies specifically recognize EPLIN phosphorylated at Ser360, Ser602, and Ser692, respectively.
To examine whether the anti-pS360, anti-pS602, and anti-pS692 antibodies can detect endogenous EPLIN phosphorylated by physiological stimuli that activate ERK, immunoblot analysis was performed on lysates from PDGF-stimulated NIH 3T3 cells. As shown in Fig. Fig.2D,2D, after the addition of PDGF, all three antibodies reacted with bands corresponding to EPLINα and EPLINβ. The time course of Ser602 and Ser692 phosphorylation was similar to that of ERK activation, but Ser360 phosphorylation proceeded slowly and increased for up to 240 min. Since the phosphorylation of Ser360 caused the mobility shift, anti-pS602 and anti-pS692 antibodies detected both EPLINα and -β as doublets at later time periods. When immunoblot analysis was performed on lysates from PDGF-stimulated primary calvarial osteoblasts, all three antibodies reacted with bands corresponding to EPLINα (see Fig. 10A). Phosphorylation of these three residues was strongly inhibited by pretreatment with U0126 (Fig. (Fig.2D;2D; see Fig. 10A) or transfection with siRNA for ERK2 or ERK1 plus ERK2 (see Fig. Fig.6A).6A). Because the level of ERK2 expression in NIH 3T3 cells is significantly higher than that of ERK1 (note that comparable amounts of ERK1 and ERK2 bands were detected by immunoblotting with the K-23 anti-ERK1 antibody in all figures), it may be reasonable that phosphorylation of EPLIN as well as p90 ribosomal S6 kinase, a well-known ERK substrate, was not clearly inhibited by ERK1 depletion (see Fig. Fig.6A).6A). We therefore concluded that EPLIN is phosphorylated at Ser360, Ser602, and Ser692 by ERK in living cells.
Because EPLIN has two actin-binding domains and cross-links actin filaments into bundles (23), we next examined whether phosphorylation of EPLIN regulates its association with F-actin. First, F-actin-binding properties of the nonphosphorylated and ERK-phosphorylated C-terminal region of EPLIN (GST-EPLIN-C) were compared by the F-actin cosedimentation assay in vitro (Fig. (Fig.3).3). In this assay, GST-EPLIN-C preincubated with or without ERK was mixed with purified actin filaments, the sample was ultracentrifuged, and the distribution of GST-EPLIN-C in the supernatant and pellet was examined. GST-EPLIN-C alone did not sediment, indicating that the sedimentation was due to binding to F-actin. The recovery of ERK-phosphorylated wild-type GST-EPLIN-C in the pellet (35%) was significantly less than that of the nonphosphorylated form (59%), whereas the nonphosphorylatable mutant (S602/692A) of GST-EPLIN-C did not show ERK-dependent changes (63% recovery in the absence of ERK and 62% recovery in its presence) (Fig. (Fig.3A).3A). To determine the stoichiometries of binding and dissociation constants (Kds), various concentrations of these proteins were assayed for cosedimentation with a fixed amount of F-actin. As shown in Fig. Fig.3B,3B, the Kds of nonphosphorylated and phosphorylated forms of wild-type GST-EPLIN-C for F-actin were calculated to be 0.65 μM and 1.2 μM, respectively. On the other hand, GST-EPLIN-C(S602/692A) preincubated with or without ERK showed similar binding properties for F-actin, with Kds of ~0.6 μM. These results suggest that phosphorylation of the C-terminal region of EPLIN by ERK reduces its affinity for F-actin. In contrast, phosphorylation of full-length EPLIN and the N-terminal region of EPLIN by ERK did not significantly reduce their affinity for F-actin in a similar in vitro actin cosedimentation assay (data not shown).
To test the effect of ERK phosphorylation of EPLIN on its affinity for actin in vivo, Myc-tagged full-length EPLIN or the N-terminal or C-terminal region of EPLIN was coexpressed with constitutively active (SDSE) or dominant-negative (SASA) HA-tagged MEK in 293T cells. Myc-EPLIN was immunoprecipitated with anti-Myc antibody, and immunoprecipitates were subjected to immunoblotting with anti-actin (Fig. (Fig.4).4). The coexpression of MEK-SDSE markedly decreased the binding of EPLIN-C to actin compared to coexpression of MEK-SASA, while the binding of EPLIN-C(S602/692A) did not change upon ERK activation (Fig. (Fig.4,4, right panels). Full-length Myc-EPLINα and Myc-EPLIN-N did not show such a reduction (Fig. (Fig.4,4, left and middle panels). EPLIN may homodimerize through a LIM domain and bind to the side of an actin filament through two actin-binding domains (23). Therefore, it may be reasonable that a reduction in the actin-binding activity of the C-terminal region does not necessarily result in a significant decrease in that of full-length EPLIN (see Fig. Fig.9F9F).
To examine whether phosphorylation by ERK regulates EPLIN function in living cells, we first investigated the subcellular localization of EPLIN during ERK activation. Serum-starved NIH 3T3 cells were stimulated with PDGF and then immunostained with anti-EPLIN (Fig. (Fig.5A).5A). EPLIN colocalized with actin stress fibers in quiescent cells (Fig. (Fig.5A,5A, left panels). PDGF stimulation induced disassembly of stress fibers and formation of peripheral and dorsal ruffles, where EPLIN was relocalized (Fig. (Fig.5A,5A, middle panels). When cells were treated with PDGF in the presence of U0126, stress fiber disassembly was partially inhibited, and a fraction of EPLIN localized on the remaining stress fibers (Fig. (Fig.5A,5A, right panels). Similar results were obtained with primary osteoblasts (see Fig. 10B).
We then examined the localization of phosphorylated EPLIN in PDGF-stimulated NIH 3T3 cells by indirect immunofluorescence microscopy using anti-pS360 and anti-pS602 antibodies (Fig. 5B and C; the anti-pS692 antibody did not show specific staining). In quiescent cells, phosphorylation of Ser360 was hardly detected. When cells were stimulated with PDGF, Ser360-phosphorylated EPLIN appeared first in peripheral and dorsal ruffles and gradually increased with time (Fig. (Fig.5B),5B), consistent with the results of immunoblot analysis (Fig. (Fig.2D).2D). U0126 pretreatment or siRNA-mediated depletion of ERK2 or ERK1 plus ERK2 completely abolished the staining by the anti-Ser360 antibody (Fig. (Fig.5B5B and and6B6B).
Immunostaining with the anti-pS602 antibody revealed that phosphorylation of EPLIN at Ser602 proceeded earlier than that at Ser360 (Fig. (Fig.5C).5C). After 5 min of stimulation with PDGF, phosphorylation signals clearly appeared in dorsal ruffles (Fig. (Fig.5C,5C, arrowheads). The time course of anti-pS602 staining intensity also correlated well with the results of immunoblot analysis. Phosphorylated ERK (p-ERK) was observed throughout the cell body after 5 min, translocated into the nucleus after 30 min, and returned to the cytoplasm after 120 min (Fig. (Fig.5C).5C). These staining patterns with anti-pS602 and anti-p-ERK antibodies were abolished by U0126 pretreatment or siRNA-mediated depletion of ERK2 or ERK1 plus ERK2 (Fig. (Fig.5C5C and 6B and C). Nuclear staining with the anti-pS602 antibody was observed in quiescent cells and in cells pretreated with U0126, suggesting that it may be nonspecific staining.
The localization of phosphorylated EPLIN at membrane ruffles prompted us to test whether the phosphorylation of EPLIN occurs during cell migration. Wound healing of fibroblasts causes a rapid and transient activation of ERK at the leading edge, which can be inhibited by U0126 (21, 26). Six hours after wounding of a confluent monolayer of NIH 3T3 cells, cells were immunostained with the anti-pS360 or anti-pS602 antibody (Fig. (Fig.7).7). Both phosphorylated Ser360 and Ser602 were evident in cells at the leading edge. As expected, pretreatment with U0126 completely abolished these staining patterns. These results indicate that EPLIN is phosphorylated by ERK at the leading edge of migrating fibroblasts.
To examine whether the phosphorylation of EPLIN by ERK is involved in membrane ruffling, we constructed EGFP-fused full-length EPLINα (EPLINα-EGFP) and a nonphosphorylatable mutant EPLINα protein [EPLINα(S360/602/692A)-EGFP] in which three major phosphorylation sites were replaced with Ala. Expression of both types of EPLIN increased the number and size of actin stress fibers in quiescent NIH 3T3 cells (Fig. (Fig.8A,8A, left panels), as reported previously for MCF-7 cells (23). After stimulation with PDGF, EPLINα-EGFP-expressing cells lost their stress fibers and formed prominent lamellipodia/membrane ruffles, but EPLINα(S360/602/692A)-EGFP-expressing cells still retained stress fibers and formed fewer membrane ruffles (Fig. (Fig.8A,8A, right panels; see Videos S1 and S2 in the supplemental material).
EGPF-positive cells were categorized into four classes according to their degree of ruffling (Fig. (Fig.8B).8B). Among quiescent cells, most wild type- and Ala mutant-transfected cells were classified as type I (without ruffles). When cells were stimulated with PDGF for 5 min, EPLINα-EGFP-expressing cells were mostly classified into types III and IV, with marked lamellipodia/membrane ruffles, but in EPLINα(S360/602/692A)-EGFP-expressing cells ruffle formation was significantly impaired. These data suggest that phosphorylation of EPLIN by ERK is involved in PDGF-induced lamellipodium/membrane ruffle formation.
Dynamic phosphorylation and dephosphorylation of cytoskeletal proteins are essential for effective cell motility. To evaluate the potential role of EPLIN phosphorylation in cell migration, wound-healing assays were performed using EPLINα-EGFP- and EPLINα(S360/602/692A)-EGFP-transfected NIH 3T3 cells. The proportion of EGFP-positive cells at the wound edge was assessed over an 8-h time period. The ratio of EPLINα-EGFP expression at the wound edge (approximately 20%) did not change during this period, indicating that EPLINα-EGFP-expressing cells and surrounding untransfected cells migrated at similar velocities (Fig. (Fig.9A;9A; see Video S3 in the supplemental material). In contrast, the EPLINα(S360/602/692A)-EGFP-expressing cells showed a marked decrease in motility and gradually fell behind the wound edge during recovery (Fig. (Fig.9A;9A; see Video S4 in the supplemental material). These results indicate that EPLIN phosphorylation by ERK is required for cell migration during wound healing.
The roles of EPLIN phosphorylation by ERK in cell motility were evaluated in another experiment, a modified Boyden chamber assay. The addition of PDGF to the lower chamber induced migration of NIH 3T3 cells expressing EGFP or EPLINα-EGFP, and pretreatment with U0126 inhibited PDGF-induced migration of these cells (Fig. (Fig.9B).9B). In contrast, expression of EPLINα(S360/602/692A)-EGFP significantly inhibited PDGF-induced migration compared with expression of EGFP or EPLINα-EGFP (Fig. (Fig.9B).9B). Similar results were obtained with primary osteoblasts (Fig. 10C). These results suggest that EPLIN phosphorylation by ERK is also required for PDGF-induced cell migration.
To further investigate the functions of EPLIN in cell motility, siRNA-mediated depletion of EPLIN and rescue assays were performed. We prepared RNAi-refractory versions of EPLIN proteins [EPLINαr-EGFP and EPLINαr(S360/602/692A)-EGFP] harboring silent mutations (Fig. (Fig.9C).9C). EPLIN depletion significantly enhanced the ability of NIH 3T3 cells to migrate in both the wound-healing assay and the modified Boyden chamber assay (Fig. 9D and E), suggesting that EPLIN functions to negatively influence cell motility. While expression of EPLINαr-EGFP efficiently restored the enhanced migratory activity to the level seen with control siRNA treatment, EPLINαr(S360/602/692A)-EGFP-expressing cells showed a significant decrease in motility compared with control siRNA-treated or EPLINαr-EGFP-expressing cells (Fig. 9D and E). We also confirmed these findings by performing modified Boyden chamber assays with primary osteoblasts (Fig. 10D). Taken together, these results indicate that EPLIN phosphorylation by ERK is required for cell migration.
In the present study, we have characterized an F-actin cross-linking protein, EPLIN, as a novel ERK MAPK substrate. First, ERK phosphorylates EPLIN on Ser360, Ser602, and Ser692 in vitro and in living cells. Second, ERK phosphorylation of EPLIN decreases the affinity of its C-terminal region for actin filaments. Third, EPLIN localizes to actin stress fibers in quiescent cells, and stimulation with PDGF induces relocalization of EPLIN to lamellipodia/membrane ruffles. Fourth, phosphorylated EPLIN localizes to membrane ruffles both upon PDGF stimulation and during wound healing. Fifth, a non-ERK-phosphorylatable mutant of EPLIN inhibits PDGF-dependent actin stress fiber disassembly, membrane ruffling, and cell migration, while RNAi-mediated silencing of EPLIN enhances cell motility. ERK thus controls actin organization and cell motility by phosphorylating EPLIN.
ERK phosphorylation sites within EPLIN were identified by site-directed mutagenesis. We determined that the Ser360, Ser602, and Ser692 residues of EPLIN are the major phosphorylation sites for ERK. These phosphorylation sites were confirmed by LC-MS/MS analysis of in vitro-phosphorylated EPLIN (see Fig. S2A, C, and D in the supplemental material). Immunoblot analysis using phospho-specific antibodies revealed that these three sites are indeed phosphorylated by ERK in intact cells (Fig. (Fig.2D2D and 10A). Although recent phosphoproteomic studies detected intracellular phosphorylation of mouse EPLIN at Ser360 (43) and of human EPLIN at Ser604 and Ser698 (corresponding to Ser602 and Ser692 of mouse EPLIN) (27, 28), spatiotemporal changes had not been reported. Interestingly, PDGF-induced phosphorylation of Ser360 occurred rather slowly compared to the rapid phosphorylation of Ser602, Ser692, and ERK (Fig. (Fig.2D).2D). This raises the possibility that cellular phosphatase activity toward Ser360 is high in the early phase or that ERK indirectly phosphorylates Ser360 through a downstream kinase.
EPLIN contains two actin-binding sites, in the N- and C-terminal halves, and a LIM domain between these sites may allow EPLIN to homodimerize. EPLIN therefore cross-links and bundles actin filaments, but the two actin-binding domains may have different functions in the cell (23). In cosedimentation assays with F-actin, we found that the C-terminal half of EPLIN, but neither full-length EPLIN nor the N-terminal half of EPLIN, reduces its association with F-actin upon ERK-mediated phosphorylation. This observation was confirmed by an in vivo experiment showing that the amount of actin coimmunoprecipitated with the C-terminal half but neither full-length EPLIN nor the N-terminal half of EPLIN was reduced by activation of ERK. Since EPLIN is supposed to bind to the side of an actin filament through two actin-binding domains (23), it may be reasonable that a reduction in the actin-binding activity of the C-terminal region does not necessarily lead to a significant decrease in that of full-length EPLIN (Fig. (Fig.9F9F).
The phosphorylation-dependent reduction of the affinity of the C-terminal region for F-actin may affect the actin-bundling activity of EPLIN to facilitate dynamic remodeling of actin filament networks. Thus, we investigated the effects of EPLIN phosphorylation on its localization, actin dynamics, and cell motility. It has been reported that endogenous EPLIN is distributed predominantly along actin stress fibers in U2OS cells (35). Consistent with this finding, immunostaining showed that EPLIN colocalized with stress fibers in quiescent NIH 3T3 cells (Fig. (Fig.5A)5A) and primary osteoblasts (Fig. 10B). Stimulation with PDGF induced stress fiber disassembly and relocalization of EPLIN to membrane ruffles within 15 to 30 min. When cells were treated with PDGF in the presence of U0126, stress fiber disassembly was partly inhibited by blocking the ERK pathway (29), and a fraction of EPLIN remained localized on the resultant stress fibers.
We further demonstrated by indirect immunofluorescence microscopy that both Ser360 and Ser602 are phosphorylated in specific subcellular areas by PDGF stimulation or during cell migration, suggesting the physiological significance of these phosphorylation sites in cellular processes. Both staining patterns were not detectable when the cells were pretreated with U0126 (Fig. 5B and C and Fig. Fig.7)7) or transfected with siRNA for ERK2 or ERK1 plus ERK2 (Fig. 6B and C), indicating ERK-dependent phosphorylation. PDGF treatment induced the phosphorylation of EPLIN at peripheral and dorsal ruffles. In migrating NIH 3T3 fibroblasts, phosphorylated EPLIN preferentially localized to the leading edge, which is consistent with previous observations that activated ERK is also localized at the leading edge during migration of rat embryo fibroblasts and 3Y1 cells (21, 26). These findings support the possible involvement of EPLIN phosphorylation by ERK in actin reorganization and cell migration (see below).
To clarify the effects of EPLIN phosphorylation on actin organization and cell motility, we used wild-type EPLIN and a non-ERK-phosphorylatable mutant EPLINα fused to EGFP. The nonphosphorylatable mutant inhibited both cellular processes. The precise molecular mechanism by which ERK promotes ruffle formation and cell migration via phosphorylating EPLIN remains unclear. Since the mutant contains substitutions in both the N- and C-terminal regions, there remained the possibility that the reduction of the affinity of the C-terminal region for F-actin may not participate in this mechanism. However, in a modified Boyden chamber assay, two substitutions in the C-terminal region [EPLINα(S602/692A)-EGFP migration index, 3.53 ± 0.46] showed a similar inhibitory effect to that by three substitutions [EPLINα(S360/602/692A)-EGFP migration index, 3.03 ± 0.35] compared with the wild type (EPLINα-EGFP migration index, 5.17 ± 0.47) or a protein with one substitution in the N-terminal region [EPLINα(S360A)-EGFP migration index, 4.67 ± 0.52], indicating the importance of a phosphorylation-dependent reduction in the C-terminal binding activity. Phosphorylation of Ser360 in the N-terminal region causes an electrophoretic mobility shift (see Fig. S1 in the supplemental material), suggesting conformational and functional changes that should be addressed in future studies. Since nonphosphorylated EPLIN dimers can form thick actin bundles through the N- and C-terminal actin-binding sites, EPLIN in quiescent cells may stabilize stress fibers and inhibit cell migration (Fig. (Fig.9F,9F, upper panel). Phosphorylated EPLIN dimers can cross-link actin filaments through only the N-terminal actin-binding sites, and thereby EPLIN in migrating cells may form a dynamic actin meshwork in membrane ruffles (Fig. (Fig.9F,9F, lower panel). Taken together, the data show that PDGF stimulation activates ERK, which phosphorylates EPLIN to reduce the affinity of its C-terminal region for actin filaments, and then phosphorylated EPLIN causes destabilization of stress fibers and reorganization of the actin cytoskeleton to form membrane ruffles and to enhance cell migration.
ERK is known to regulate actin organization and cell motility by phosphorylating a number of proteins, including MLCK, FAK, paxillin, actopaxin, and vinexin. We demonstrate in this study that EPLIN is also a mediator of ERK-regulated cytoskeletal dynamics. Because the expression of phosphomimetic mutants of EPLIN had weak effects on these processes (data not shown), many actin-binding proteins phosphorylated by ERK are likely to act in concert to regulate actin dynamics. Furthermore, various extracellular stimuli induce actin reorganization and cell migration through other ERK-independent pathways. For example, it was recently reported that Akt regulates these processes via phosphorylation of girdin, an F-actin cross-linking protein (7). Other actin cross-linking proteins, such as fascin (42, 45) and L-plastin (15), were also shown to be regulated by phosphorylation to control actin cytoskeletal assembly and cell motility.
It has been reported that EPLIN is down-regulated or lost in a number of oral, prostate, and breast cancer cell lines (3, 22). Since siRNA-mediated depletion of EPLIN enhanced cell motility during wound healing and in PDGF-induced cell migration, the down-regulation of EPLIN expression might be relevant to migration and invasion of these cancer cells. Previously, it was reported that ectopic expression of EPLIN can suppress anchorage-independent growth of NIH 3T3 cells transformed by Cdc42V12 or EWS/Fli-1 but not by RasV12 (35). This can now be explained by actin reorganization and enhanced cell motility through the Ras-Raf-MEK-ERK-EPLIN pathway. Ras-mediated phosphorylation of EPLIN may be involved in the invasion of tumor cells with Ras mutations. EPLIN is highly conserved from zebra fish to humans and contains multiple Ser/Thr-Pro motifs that can potentially be phosphorylated by ERK. The ERK-EPLIN pathway may play important roles in diverse physiological processes in vertebrates.
We thank Michimoto Kobayashi for his help during the initial stages of this work, Yutaka Harita for help with rabbit immunization, Makoto Watanabe for help with isolating primary calvarial osteoblasts, Hiroyuki Fukuda for LC-MS/MS analysis, Miho Ohsugi and Noriko Tokai-Nishizumi for help with time-lapse video microscopy, Masanori Mishima and Max Douglas for critical readings of the manuscript, and Shiro Suetsugu and Eisuke Nishida for helpful discussions and advice. We also thank Martin McMahon for kindly providing ΔB-Raf:ER cells.
This work was supported in part by grants-in-aid for scientific research from the Japan Society for the Promotion of Science (to H.K.) and the Encouraging Development Strategic Research Centers Program, Special Coordination Funds for Promoting Science and Technology, the Ministry of Education, Culture, Sports, Science, and Technology (to S.H.). This work was also supported by grants from the Nakajima Foundation (to H.K.) and the Novartis Foundation (Japan) for the Promotion of Science (to S.H.). This work was developed and coordinated under the framework of the program for the International Research and Educational Institute for Integrated Medical Sciences (IREIIMS).
Published ahead of print on 17 September 2007.
†Supplemental material for this article may be found at http://mcb.asm.org/.