Interaction of Filamin A with the Insulin Receptor Alters Insulin-dependent Activation of the Mitogen-activated Protein Kinase Pathway*

The biological actions of insulin are associated with a rapid reorganization of the actin cytoskeleton within cells in culture. Even though this event requires the participation of actin-binding proteins, the effect of filamin A (FLNa) on insulin-mediated signaling events is still unknown. We report here that human melanoma M2 cells lacking FLNa expression exhibited normal insulin receptor (IR) signaling, whereas FLNa-expressing A7 cells were unable to elicit insulin-dependent Shc tyrosine phosphorylation and p42/44 MAPK activation de-spite no significant defect in IR-stimulated phosphorylation of insulin receptor substrate-1 or activation of the phosphatidylinositol 3-kinase/AKT cascade. Insulin-de-pendent translocation of Shc, SOS1, and MAPK to lipid raft microdomains was markedly attenuated by FLNa expression. Coimmunoprecipitation experiments and in vitro binding assays demonstrated that FLNa binds constitutively to IR and that neither insulin nor depolymerization of actin by cytochalasin D affected this interaction. The colocalization of endogenous FLNa with IR was detected at the surface of HepG2 cells. Ectopic expression of a C-terminal fragment of FLNa (FLNa CT ) in HepG2 cells blocked the endogenous IR-FLNa interaction and potentiated insulin-stimulated MAPK phosphorylation

Insulin is a pleiotropic hormone with multiple integrated metabolic and mitogenic signaling pathways. Upon binding of insulin, the cell surface insulin receptor (IR) 1 undergoes auto-phosphorylation on several tyrosine residues located in the cytoplasmic portion of the ␤-subunit, with subsequent increase in its intrinsic tyrosine kinase activity. A number of adaptor proteins, including insulin receptor substrate (IRS) 1/2, the Src and collagen homologous (Shc) molecules, Cbl, Gab1, p60 Dok , and APS, are recruited to the activated IR through their Src homology 2 (SH2) or phosphotyrosine binding (PTB) domains and become tyrosine-phosphorylated, thus allowing formation of signaling competent complexes subjacent to the inner wall of the plasma membrane (1). Specificity in signaling is achieved through differences in the ability of IR to interact with these adaptor molecules.
The submembranous actin microfilament network links various signaling proteins (e.g. phosphatidylinositol (PI) 3-kinase and Rho family of proteins) that play an important role in membrane trafficking, cellular integrity, and homeostasis (2)(3)(4). Incidentally, insulin is known to induce rapid dynamic reorganization of the actin cytoskeleton to generate the forces necessary for plasma membrane ruffling formation and a host of other cellular processes, including proper insertion of insulin-regulatable glucose transporter 4 in the cell surface (5). The stabilization of actin network at the periphery of the cytoplasm and its attachment to cellular membranes are orchestrated by actin-binding proteins. The filamin family of actin-binding proteins bind to actin filaments and to a number of macromolecules (reviewed by Stossel et al. (6)), notably small GTPases (7) and p21-activated kinase 1 (Pak1) (8). Filamins are rod-shaped proteins of ϳ280 kDa that contain an N-terminal actin-binding domain followed by 24 repeats each of 96 amino acids. Repeat 24 contains the dimerization domain of filamin. Significantly, most of the interactions between filamin and its binding partners occur through the C-terminal end of filamins, thus allowing the interwebbing of actin scaffolds with membrane-bound proteins (6).
Several proteins involved in signal transduction events are partitioned in lipid rafts, a process that allows proper compartmentalization and spatial/temporal organization of functionally competent signaling complexes (9). The IR segregates to glycolipid-enriched raft domains of the plasma membrane in a variety of cell types (10,11). Of interest, inducible association of signaling proteins with lipid rafts has been shown to depend on the actin cytoskeleton through a mechanism involving raft coalescence (12,13). Therefore, colocalization of filamin and resident raft proteins, including the scaffolding protein caveolin-1 (14), is likely to be of physiological importance in the clustering of lipid rafts and organization of multiple signaling pathways by the actin cytoskeleton. Until now, however, little is known about a role that filamins would play in the transmission of the diverse effects of insulin.
In this study, we have investigated the relative contribution of filamin A (FLNa) to the regulation of insulin signaling in human melanoma cell lines (M2 cells) that have spontaneously lost expression of FLNa, and a subline with stable expression of recombinant FLNa (A7 cells) (15). By using this cell model, we have found that FLNa expression attenuated insulin mitogenic signaling by selectively inhibiting the recruitment and tyrosine phosphorylation of Shc and subsequent activation of p42/44 MAPK. Of interest, neither early steps in insulin signaling (e.g. IR and IRS-1 tyrosine phosphorylation) nor the activation of the IRS-1/PI3-kinase/AKT pathway were affected by FLNa. MAPKs transduce a mitogenic signal by phosphorylating transcription factors such as Elk-1, which leads to regulation of critical genes (16). Our results have indicated also that FLNa binds directly to IR and that ectopic expression of a C-terminal fragment of FLNa disrupts constitutive IR-FLNa interaction in HepG2 cells, thereby inducing a marked increase in insulinstimulated MAPK phosphorylation and Elk-1 transactivation. These results indicate that FLNa has a negative role in MAPKmediated Elk-1 transcriptional activation in response to insulin, in part, by interacting directly with the IR.

EXPERIMENTAL PROCEDURES
Materials-The anti-human IR antibodies for immunoprecipitation (clones 29B4 and CII 25.3) were purchased from Calbiochem. The anti-IR ␤-subunit antibody as well as horseradish peroxidase-linked phosphotyrosine (clone RC20) antibody for Western blot were from Transduction Laboratories. The anti-IR ␤-subunit antibody (06-492) for immunofluorescence studies, and the anti-Shc, phosphotyrosine, SOS1, IRS-1, p85 subunit of PI3-kinase, and p42/44 MAP kinase (ERK1 and 2) antibodies were from Upstate Biotechnology. Inc. The monoclonal anti-FLNa antibody for immunoprecipitation and immunofluorescence studies was purchased from Chemicon International, and that used for Western blot analysis was from Research Diagnostics Inc. The anti-IR ␣-subunit and c-Src antibodies were from Santa Cruz Biotechnology. The phospho-ERK and phospho-AKT (Ser(P)-473) antibodies were purchased from Promega and Cell Signaling Technology, respectively. The influenza virus hemagglutinin (HA) epitope antibodies for Western blot and immunofluorescence studies were from BD Biosciences Clontech and Covance, respectively. The Alexa Fluor secondary antibodies, Alexa Fluor-conjugated phalloidins, and Topro-3 were from Molecular Probes. The recombinant human insulin and EGF were from Calbiochem and Upstate Biotechnology. Inc., respectively. FuGENE 6 and Lipo-fectAMINE 2000 were from Roche Applied Science and Invitrogen, respectively. Recombinant human insulin-like growth factor 1 (IGF-1), cytochalasin D, 2-mercaptoethanol, sodium orthovanadate, and Me 2 SO were from Sigma. The commercial sources for electrophoresis reagents, culture media, sera, films, horseradish peroxidase-linked secondary antibodies, and the enhanced chemiluminescence detection system for immunoblot detection have been described previously (17).
Plasmid Construction-cDNA encoding the FLNa C-terminal fragment (amino acids 2357-2647) was amplified by PCR using full-length human filamin A cDNA (kindly provided by Dr. Yasutaka Ohta (Harvard Medical School, Boston)) as the template. The following primer pairs were used: forward primer, 5Ј-TAGGATCCATGGGCTATC-CATATGATGTTCCAGATTATGCTCTGAACGGGGCCAAG-3Ј; reverse primer, 5Ј-CGACTAGTTCAGGGCACCACAAC-3Ј. The underlined nucleotides indicate the KpnI and SpeI sites in the forward and reverse primers, respectively, and the italic nucleotides indicate the HA epitope. The amplified product was digested and inserted into the KpnI/XbaI sites of pcDNA3.1 (Invitrogen). The integrity of the HA-FLNa CT construct was verified by automated sequencing.
A cDNA fragment corresponding to Arg 941 -Ser 1343 of the human IR was generated by PCR amplification to contain BamHI and EcoRI restriction sites using pCMVHIR as the template (18). The resulting 1206-bp BamHI-EcoRI cDNA fragment was inserted into pGEX-4T-1 vector (Amersham Biosciences). GST and GST-IR fusion protein were expressed in BL21, induced by 0.5 mM isopropylthio-␤-D-galactopyranoside, and purified by affinity chromatography with glutathione-Sepharose (Amersham Biosciences) according to the manufacturer's protocols.
The resulting eluates were concentrated by ultrafiltration and stored at Ϫ70°C. Translation and product size were verified by analyzing an aliquot of the samples by SDS-PAGE and Colloidal blue staining of the gel, as well as by immunoblot analysis. The integrity of the GST-IR construct was verified by automated sequencing.
Cell Treatment-M2 and A7 cells were cultured in minimum essential medium (MEM) supplemented with 10 mM Hepes, pH 7.4, 0.25% sodium bicarbonate, 2 mM L-glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, 8% newborn calf serum, and 2% fetal calf serum and maintained in a humidified atmosphere of 5% CO 2 in air at 37°C. Before treatment, cells were serum-starved for 18 h in MEM supplemented with 0.1% FBS, washed with phosphate-buffered saline (PBS), and then incubated in Krebs-Ringer phosphate (KRP) buffer. Cells were treated in the absence or the presence of 200 M orthovanadate for 30 min followed by the addition of 100 nM insulin for periods up to 10 min. In some experiments, cells were stimulated with either 25 nM insulin, 20 nM EGF, 14 nM IGF-1, or 20% FBS for 15 min. These cells were washed in PBS and immersed in liquid nitrogen. The human HepG2 hepatoma cells and HEK293 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) and McCoy's 5A medium, respectively, supplemented with 10% FBS, 2 mM L-glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin.
Transient Transfection Assays-HepG2 cells were cultured for 24 h until 60 -80% confluence was reached. Transient transfection was performed according to the manufacturer's protocol for the use of Lipo-fectAMINE 2000. In brief, empty expression vector (pcDNA3.1) or expression plasmid encoding HA-FLNa CT was mixed with the transfection reagent and directly added to the culture plates at a ratio of 6 g/60-mm dish. Twenty four to 48 h later, cells were used for various experiments. Expression of HA-FLNa CT was analyzed in total cell lysates by immunoblotting with anti-HA antibody (Clontech, Palo Alto, CA). Cells were serum-starved for 18 h and incubated for 30 min with 200 M orthovanadate prior to 100 nM insulin treatment for 10 min.
Elk-1 Transactivation Assay-Transactivation of Elk-1 was examined by the PathDetect Elk-1 trans-Reporting System (Stratagene). In brief, HepG2 cells were cotransfected with pFR-Luc, pFA-Elk-1 and 0.2 g of pCMV-␤-galactosidase and either 3 g of pcDNA3.1 or 3 g of HA-FLNa CT as indicated. Serum-starved cells were left untreated or were stimulated with 100 nM insulin for 24 h. Elk-1 luciferase and ␤-galactosidase activities were measured using assay system kits from Promega according to the manufacturer's instructions, and the luciferase values were normalized to ␤-galactosidase.
Detergent-free Isolation of Lipid Rafts-Isolation of lipid rafts was accomplished using a detergent-free sucrose gradient centrifugation method as described previously (19). In brief, cells from 150-mm culture dishes were resuspended in 1 ml of 0.5 M sodium carbonate, pH 11, supplemented with a protease inhibitor mixture (Calbiochem), and homogenized by passing cells 15 times through a 23-gauge needle and two 10-s bursts of a sonicator probe on ice. The sucrose concentration in cell extracts was adjusted to 45% (w/w) by the addition of 1.8 volume of 70% (w/w) sucrose prepared in MBS (25 mM MES, pH 6.5, 0.15 M NaCl). At the bottom of an ultracentrifuge tube, 2 ml of the extracts were placed, followed by the addition of 5 ml of 35% (w/w) and 4 ml of 5% (w/w) sucrose prepared in MBS containing 0.25 M sodium carbonate. After centrifugation at 200,000 ϫ g for 16 h at 4°C in a Beckman SW41 rotor, a total of 11 fractions (1 ml each) were collected from the top of each gradient and used for immunoblotting.

Filamin and Insulin Receptor Signaling
SDS-PAGE and electrotransferred onto polyvinylidene difluoride membranes. Detection of individual proteins was carried out by immunoblotting with specific primary antibodies and visualized by enhanced chemiluminescence. Signals were quantitated by densitometry using the ImageQuant software (Amersham Biosciences).
Measurement of Thymidine Incorporation-M2 and A7 cells were grown to 75% confluence in 6-well cluster plates. The growth medium was replaced with serum-free MEM for 18 h. Cells were stimulated with insulin at the indicated concentrations (0, 1, 10, and 100 nM) for 18 h and then pulsed with [ 3 H]thymidine (PerkinElmer Life Sciences), 2 Ci/ml, for 2 h at 37°C. The cells were washed three times in ice-cold PBS, and DNA was precipitated with 10% trichloroacetic acid for 30 min on ice. After a rapid wash, trichloroacetic acid-precipitable material was dissolved in 0.5 ml of 1 M NaOH and then neutralized with HCl. Levels of [ 3 H]thymidine incorporated into DNA were measured in a scintillation counter and expressed as counts/min. Values obtained without insulin stimulation were subtracted from the corresponding values obtained after treatment.
GST Pull-down Experiments-HepG2 cell lysates (endogenous FLNa) or 5 g of purified chicken gizzard FLNa (Research Diagnostics Inc.) were either incubated with anti-FLNa antibodies for 4 h followed by precipitation of the immune complexes with protein A/G-Sepharose beads or incubated for 16 h with 5 g of GST or 5 g of GST-IR ␤-subunit protein preimmobilized onto glutathione-agarose (Amersham Biosciences) at 4°C in TLB buffer. After a series of washes in TLB buffer, the bound proteins were eluted in Laemmli sample buffer, separated by SDS-PAGE, and immunoblotted using anti-FLNa antibody.
Fluorescence Microscopy-HepG2 cells grown on coverslips were fixed in fresh 4% paraformaldehyde in PBS for 10 min and permeabilized in 0.1% Triton X-100 in PBS for 10 min at room temperature. The cells were incubated with blocking buffer (8% bovine serum albumin in PBS) for 20 min at room temperature, washed in PBS supplemented with 0.5% bovine serum albumin and 0.05% Tween 20, and incubated with anti-IR ␤-subunit (1:100), FLNa (1:100), or HA (1:166) antibody for 16 h at 4°C. After washing, cells were stained with Alexa Fluor secondary antibody (1:1000). For immunolocalization of F-actin, fixed cells were incubated with Alexa Fluor-conjugated phalloidin (1:20). Nuclear counterstaining was performed by incubating coverslips with Topro-3 in PBS for 5 min prior to mounting slides with Vectashield (Vector Laboratories). Images were acquired using a Zeiss LSM-410 inverted confocal microscope with a 63ϫ oil-immersed objective and processed by using the Metamorph software (Universal Imaging Corp.). No fluorescent staining was observed when the primary antibody was omitted.
Statistical Analysis-Quantitative data are presented as mean Ϯ S.E., and differences between the means were determined by analysis of variance coupled to Fisher's least significance difference for multiple mean comparisons. A p value of Ͻ0.05 was considered significant. Statistical analyses were performed with the StatView statistical software program (SAS Institute, Inc., Cary, NC).

Selective Impairment in Insulin Signaling in FLNa-express-
ing Cells-Insulin elicits rapid autophosphorylation of the IR and tyrosine phosphorylation of substrates, including IRS-1 and Shc proteins, in a number of cell lines. We examined FLNa-deficient M2 cells and FLNa-expressing subline A7 cells for their insulin responsiveness, and we found that both cell lines exhibited a comparable increase in insulin-induced IR autophosphorylation (Fig. 1A). Under basal conditions, cells that were pretreated with the protein-tyrosine phosphatase inhibitor orthovanadate had very low levels of IRS-1 and Shc tyrosine phosphorylation (Fig. 1, B and C). Addition of insulin elicited a significant increase in IRS-1 phosphorylation in vanadate-pretreated M2 and A7 cells (Fig. 1B). In contrast, the combination of insulin and vanadate strongly stimulated the levels of Shc tyrosine phosphorylation in M2 cells, while being rather ineffective in the FLNa-expressing A7 cells (Fig. 1C). PI3-kinase is an important enzyme implicated in insulin signal transduction through its interaction with tyrosine-phosphorylated IRS proteins (21). It is a heterodimeric enzyme encompassing a p85 regulatory subunit and a catalytic subunit (p110). We therefore examined the cosedimentation of p85 in IRS-1 immunoprecipitates to determine whether p85 association to IRS-1 might be affected by FLNa expression. As shown in Fig. 1B, insulin was able to stimulate recruitment of PI3kinase to IRS-1 both in vanadate-pretreated M2 and A7 cells.
Two of the major insulin signaling events initiated downstream of the IR is activation of the PI3-kinase/AKT pathway and Ras/MAPK cascade. To evaluate the involvement of FLNa expression on the regulation of these pathways, we examined the ability of insulin to activate AKT and p42/44 MAPK (also known as extracellular signal-regulated kinase, ERK 1/2) by immunoblotting cell lysates with phospho-specific antibodies (Fig. 2). The levels of phosphorylation of AKT at Ser-473, a modification required for its activation (22), were increased upon stimulation of M2 and A7 cells with insulin. In contrast to AKT, there was a marked attenuation in insulin-induced ERK phosphorylation in FLNa-expressing A7 cells (Fig. 2 upper  panel, 9th versus 2nd lane). We then investigated whether the lack of FLNa plays a role in the phosphorylation of AKT and ERK in response to other stimuli. Both cell lines were equally responsive to EGF and serum; however, the increase in ERK phosphorylation levels induced by IGF-1 in M2 cells was blocked in FLNa-expressing A7 cells (Fig. 2, upper panel). Similar amounts of ERK protein were present in either cell line (Fig. 2, lower panel). Thus, expression of FLNa differentially affects ERK regulation in response to various stimuli. These results are consistent with a selective effect of FLNa in the signaling pathway used by insulin and IGF-1 to regulate p42/44 MAPK cascade.
It is well recognized that phosphorylation of Shc by IR is necessary for activation of the Ras/MAPK pathway and mitogenesis (23). In order to investigate the ability of FLNa to modulate mitogenic responses of insulin, we incubated M2 and A7 cells with increasing concentrations of the hormone for 18 h and measured DNA synthesis. FLNa-depleted M2 cells displayed an enhanced sensitivity to insulin when compared with FLNa-expressing A7 cells (Fig. 3). Both cell lines responded to serum (10%) stimulation by increasing thymidine incorporation to the same levels. The sensitivity of insulin-dependent activation of the Shc/MAPK cascade and the subsequent mito- B, serum-starved M2 (panels I and II) and A7 cells (panels III and IV) were left untreated (panels I and III) or treated with 100 nM insulin (panels II and IV) for 10 min. Cells were lysed and fractionated using a detergent-free sucrose density gradient centrifugation method as described under ''Experimental Procedures.'' Fractions were collected from the top of the gradient, separated by SDS-PAGE, and immunoblotted with anti-SOS1 antibody. C, lipid raft fractions (fractions 4 -6) were prepared from insulin-stimulated M2 and A7 cells and immunoblotted with anti-Shc, phosphoactive ERK, total ERK, and c-Src antibodies. Shown is a representative experiment that was repeated twice with similar results.

Filamin and Insulin Receptor Signaling
genic responses to FLNa expression suggest specific perturbation of the signaling pathway proximal to the IR.
Effect of FLNa Expression on Cellular Redistribution of Signaling Intermediates-Translocation of the Shc adaptor from the cytosol to lipid raft microdomains leads to MAPK activation (24). Moreover, the guanine nucleotide exchange factor SOS1 is prevalently cytosolic and must be brought to the cytosolic side of the plasma membrane in close juxtaposition to Ras to allow GDP/GTP exchange and Ras activation (25). To establish whether expression of FLNa could influence the ability of signaling intermediates to be redistributed upon insulin stimulation, lipid rafts were isolated by sucrose gradient centrifugation from lysates of M2 and A7 cells and immunoblotted with specific antibodies. Preliminary studies showed that both cell types contained similar levels of signaling molecules, such as Shc, SOS1, ERK, and c-Src (Fig. 4A). Light fractions, enriched in rafts, and heavy fractions, containing most soluble proteins, were first tested by immunoblot with anti-SOS1 antibodies. As shown in Fig. 4B (panels I and III), endogenous SOS1 was found only in the heavy fractions of unstimulated cells, demonstrating that SOS1 is predominantly in the cytosol. However, addition of insulin resulted in inducible localization of SOS1 in the light fractions of FLNa-depleted M2 cells but not from FLNa-expressing A7 cells (Fig. 4B, panel II versus IV). The latter cells were found to have also impaired translocation of Shc, ERK, and phosphoactive ERK to the membrane rafts following insulin treatment (Fig. 4C). Of importance, FLNa expression did not abrogate the levels of c-Src, and presumably of other raft-associated proteins, in these specialized membrane microdomains (Fig. 4C, bottom panel). These results are consistent with the hypothesis that FLNa plays an important role in insulin-dependent Shc/MAPK cascade signaling at a step proximal to the IR.
The IR Is an FLNa-interacting Protein-FLNa interacts with a number of cell surface proteins and intracellular signaling molecules (6). We examined whether IR could form a complex with FLNa in intact cells. Our data demonstrated that IR was associated constitutively with FLNa in A7 cells but not in M2 cells that lack FLNa (Fig. 5A, right panels). Insulin treatment did not cause a significant change in the formation of the IR-FLNa complex in A7 cells. Similar results were obtained in HEK293 cells and HepG2 hepatoma whereby IR-FLNa association was detected under basal and insulin-stimulated conditions (Fig. 5A, left panels). In light of previous evidence supporting the actin-binding properties of FLNa, we sought to examine the role of cytoskeletal organization in the regulation of IR-FLNa association. To this end, HepG2 hepatoma were treated with cytochalasin D, an agent that causes depolymerization of actin (26), followed by reciprocal immunoprecipitation/immunoblotting assays as well as fluorescence microscopy. The inhibitor failed to block FLNa association to the IR, although it blocked the filamentous pattern of actin staining along the longitudinal axis of the cell (Fig. 5B, left panels). Thus, it would appear that the functional pattern of filamentous actin does not regulate IR-FLNa interaction.
To verify independently that a population of the IR is colocalized with FLNa, HepG2 hepatoma were stained with antibodies to FLNa and IR ␤-subunit. As shown in Fig. 5C, both FLNa and IR in the basal state exhibited a prominent plasma membrane staining. Next, to determine whether IR could be a direct FLNainteracting partner, lysates from HepG2 cells were incubated with a GST fusion protein containing the cytoplasmic domain of the human IR (amino acids 941-1343) (GST-IR␤) and analyzed by immunoblotting for the presence of FLNa. GST-IR␤ associated with FLNa, whereas GST protein alone was unable to interact (Fig. 6, lane 6). Incubation with GST-tagged SH2SH2SH3 domain of PLC␥1 did not result in the recruitment of FLNa (data not shown). Purified chicken gizzard FLNa was then tested for its ability to bind GST-IR␤ and was found also to cosediment with recombinant IR (Fig. 6, lane 8). Taken together, these data indicate that IR interacts with FLNa and may be physiologically relevant.
Expression of the FLNa C-terminal Region Potentiates Insulin-mediated MAPK Activation-In order to determine whether the C-terminal region of FLNa is required for the interaction with IR and the regulation of insulin downstream signaling, an FLNa construct comprising the C-terminal repeat regions Arg 22 -Arg 24 (amino acids 2357-2647) was fused to the HA epitope (HA-FLNa CT ) and transiently expressed in HepG2 cells (Fig. 7A). Although ectopic expression of FLNa CT blocked the association of IR with endogenous FLNa (Fig. 7B, upper panel), it did not affect the extent of IR autophosphorylation elicited by insulin (Fig. 7B, middle panel). Shown in Fig. 7C is the quantitative data of four independent observations. We then explored the role of FLNa CT in the activation of IR-mediated downstream signaling events. HepG2 cells that were transfected with either pcDNA or FLNa CT displayed no detectable difference in basal and insulin-induced AKT phosphorylation, whereas the phosphorylation levels of ERK were significantly increased in the FLNa CT -transfected cells (Fig. 7D). Similarly, expression of FLNa CT in HEK293 cells consistently led to far greater insulin-induced phosphorylation of ERK, but not AKT, compared with the corresponding control vector-expressing cells (data not shown). It is therefore unlikely that FLNa is involved in regulating the activation of the PI3-kinase/AKT pathway by insulin.
Engagement of the IR signaling pathway results in the activation of p42/44 MAPK (ERK) and subsequent increase in Elk-1 transcriptional activity (27). To determine the relative contribution of FLNa CT in the regulation of Elk-1 transcriptional activity by insulin, we used a transient transfection assay in which HepG2 cells were cotransfected with pFA-Elk (the activation domain of Elk-1 fused to the DNA-binding domain of Gal4), pFR-Luc (luciferase reporter driven by Gal4 DNA-binding element repeats), pCMV-␤-galactosidase as a control, and either pcDNA or HA-FLNa CT . Serum-starved HepG2 cells were treated with or without insulin, and the activity of Elk-1-mediated luciferase activity was determined. Insulin stimulation of pcDNA-transfected cells resulted in a 1.5-fold increase in Elk-1 transactivation (Fig. 8; p Ͻ 0.05). However, insulin-induced Elk-1 transactivation was potentiated 2.2-fold in HepG2 cells that expressed HA-FLNa CT (Fig. 8; p Ͻ 0.01), whereas the basal level of luciferase activity was not significantly altered (p Ͼ 0.2). Taken together, our data suggest that FLNa CT does influence physiologically relevant MAPK-dependent effects of insulin such as Elk-1 transcriptional activation.
Subcellular Distribution of FLNa CT in HepG2 Cells-Recent evidence points to the role of FLNa in the control of actin cytoskeleton and ruffle formation (6). We tested whether ectopic expression of FLNa CT could cause changes in cell morphology through destabilization of cell surface actin cytoskeleton. To this end, control and HA-FLNa CT -transfected HepG2 cells were stained with an antibody to HA and Alexa 568conjugated phalloidin to detect changes in the pattern of filamentous actin. Whereas F-actin staining in cells expressing the empty vector exhibited a filamentous pattern, a marked attenuation of actin staining throughout the cell was observed upon HA-FLN CT expression (Fig. 9, D and F versus B). Of significance, expression of HA-FLN CT in unstimulated cells appeared to induce a change in cell morphology and an increase in membrane ruffling at sites where FLN CT accumulated (Fig.  9C). The pattern of HA-FLN CT staining was detected both subjacent to the plasma membrane and as a diffuse cytoplasmic distribution (Fig. 9E). Finally, overlay of anti-HA and IR staining in transfected HepG2 cells showed constitutive colocalization of FLN CT with endogenous IR at the cell membrane (Fig. 9I).

DISCUSSION
The present data provide evidence that expression of the actin-binding protein FLNa promotes selective deactivation of insulin signaling events that regulate MAPK phosphorylation and activation. The inhibition of insulin-stimulated MAPK phosphorylation was deemed to be specific because FLNa expression did not interfere with other insulin-induced signals, such as those that lead to IR and IRS-1 tyrosine phosphorylation and activation of the PI3-kinase/AKT cascade. Various growth factors such as IGF-1, EGF, and serum have been used to demonstrate the relative selectivity of FLNa expression in insulin signaling. Consistent with its role as a negative regulator of insulin-mediated activation of MAPK, FLNa expression potently inhibited the ability of insulin to induce Shc tyrosine phosphorylation and translocation of various signaling intermediates (e.g. Shc, SOS, and MAPK) to membrane rafts of A7 cells. Physical association between FLNa and the IR was demonstrated but was found to be blocked by overexpression of a C-terminal fragment of FLNa (FLNa CT ) in the human liverderived HepG2 cells. Finally, our results show a selective potentiation of insulin-mediated phosphorylation of MAPK and Elk-1 transcriptional activity in HepG2 cells expressing FLNa CT . Increase in Elk-1 transcriptional activity is thought to be an essential event for the initiation of numerous physiological responses (28,29). These observations and those showing that FLNa CT expression leads to alteration in cell morphology and increased membrane ruffle formation would suggest that FLNa is a possible modulator of insulin action.
In this study, we have demonstrated that human M2 cells lacking FLNa exhibit normal insulin-mediated phosphorylation of p42/44 MAPK, whereas the FLNa-expressing A7 cells are unable to elicit this phosphorylation event upon stimulation with insulin. In contrast, FLNa expression did not interfere with MAPK phosphorylation in response to EGF or serum but not IGF-1. These results would underscore the specificity of FIG. 6. FLNa interacts directly with the IR in GST pull-down assays. Equal amounts of HepG2 cell extracts or purified chicken FLNa (cFLNa) were incubated with either preimmobilized GST or GSTtagged cytoplasmic domain of the IR ␤-subunit (residues 941-1343) as described under ''Experimental Procedures.'' Eluates from GSH-agarose (GSH-q) were resolved by SDS-PAGE and immunoblotted with anti-FLNa antibody (lanes 5-8). HepG2 cell extracts and cFLNa were also immunoprecipitated (Ip) and immunoblotted with anti-FLNa before they were added to the GST-and GST-IR beads (lanes 1-4). These are representative experiments that were repeated at least twice.
the FLNa effects toward the insulin/IGF-1 responses and support our observations indicating that FLNa expression selectively impairs insulin-dependent thymidine incorporation, while having no effect on serum-stimulated DNA synthesis. There is evidence to suggest that TNF-stimulated MAPK activation is potently inhibited in M2 cells (30), whereas heregulin efficiently stimulates MAPK in both M2 and A7 cells (8). On the other hand, the G protein-coupled extracellular calcium-sensing receptor can activate MAPK very efficiently as a result of specific interaction between FLNa and the calcium-sensing receptor (31,32). These observations raise the interesting possibility that although FLNa is dispensable for MAPK activation in response to EGF or serum, it might be required at some regulatory step proximal to the highly homologous receptors for insulin and IGF-1, possibly in cooperation with a subset of signal transducing molecules converging to MAPK (33).
The role of Shc in insulin-induced mitogenic signaling follows a path that has been well characterized. The adaptor Shc binds via its PTB domain to the juxtamembrane region of the activated IR (34), which then leads to Shc tyrosine phosphorylation and subsequent association with preformed Grb2-SOS complex to the plasma membrane where Ras is known to reside (35). Because of the role of Shc in linking the activated IR to the Ras/MAPK pathway via the guanine nucleotide exchange factor SOS, elucidation of the mechanisms by which FLNa impairs Shc tyrosine phosphorylation is key for understanding the mitogenic signaling of insulin. Localization to membrane rafts is required for Shc activity (36). Indeed, addition of the membrane localization sequence for Ras leads to constitutive Shc targeting to the plasma membrane, thus resulting in ligandindependent activation of the Ras/MAPK cascade (24). This is somewhat consistent with our observation that FLNa-dependent interactions might inhibit insulin-induced Shc tyrosine phosphorylation due to impaired translocation of Shc to membrane rafts where the IR is known to reside. It is noteworthy that the ability of insulin to promote redistribution of SOS and MAPK to lipid rafts is also blocked as a consequence of FLNa expression in A7 cells. This argues well for the role of FLNa in governing insulin-dependent signaling events via regulation of protein trafficking.
FLNa is an actin-binding protein expressing a number of repeats that have been shown to associate with receptors or signaling proteins (reviewed by Stossel et al. (6)). For instance, FLNa binds Smads, a group of proteins necessary for the regulation of TGF-␤ signaling (37). Many proteins known to bind FLNa appear to be functionally linked to downstream signaling components via the actin cytoskeleton. As such, FLNa is a determinant of the submembranous cytoskeletal architecture of cells, and consistent with this role, it has an important function in the endocytic sorting and recycling pathways (38,39). The presence of filamin has been reported to be necessary for translocation of the 5-phosphatase SHIP-2 at the membrane ruffles and the cortical actin rim at the periphery of the COS-7 cells after EGF stimulation (40). Similarly, FLNa may

FIG. 7. Expression of FLNa C-terminal fragment leads to specific increase in insulin-stimulated ERK activation in HepG2 cells.
Cells were transiently transfected with empty vector or HA-tagged FLNa CT for 24 h, serum-starved for 18 h, and then left untreated or stimulated with 100 nM insulin for 10 min. Cell lysates were directly analyzed by SDS-PAGE and immunoblotted with anti-HA antibody (A) or immunoprecipitated (Ip) with control IgG (NI) or anti-IR antibody followed by immunoblot analysis using anti-FLNa, RC20 and IR ␤-subunit antibodies (B). C, bands corresponding to FLNa were quantitated by laser densitometry, and a value of 1.0 was assigned to pcDNA-transfected, untreated cells. be required for proper cell surface expression of the D2 dopamine receptors (41). On the other hand, ligand-induced movement of the human androgen receptor from the cytoplasm to the nucleus is a process that requires FLNa (42). Unlike these FLNa-binding proteins, the association between the IR to FLNa does not appear to be obviously related to intact actin cytoskeletal assembly as actin depolymerization with cyto-chalasin D failed to disrupt the constitutive IR-FLNa interaction.
Immunofluorescence studies revealed that transiently expressed FLNa CT colocalizes with the IR at the surface of HepG2 cells. Whereas FLNa CT expression blocks the constitutive association between endogenous FLNa and the IR, it potentiates insulin-induced MAPK phosphorylation and Elk-1 transcriptional activity when compared with pcDNA-transfected cells treated with insulin. However, FLNa CT had no effect on insulin-induced IR or IRS-1 tyrosine phosphorylation and an increase in AKT phosphorylation elicited by insulin. It is therefore likely that overexpressed FLNa CT titrates inhibitory proteins out of endogenous signaling complexes involved in MAPK cascade activation, thereby potentiating insulin signal transduction. In light of these results, what role might FLNa play in constitutive and IR-mediated activation of MAPK? With the exception perhaps of a weak binding of MEK-1 to FLNa in two-hybrid assays (30), no component of MAPK-activating pathways other than the IR, as demonstrated here, has been shown to bind to FLNa to date. Thus, it is possible that alteration in FLNa expression might affect the signaling potential of the IR at a step proximal to the receptor, namely the tyrosine phosphorylation of a subset of protein substrates and their recruitment to appropriate location. Alternatively, a subpopulation of the IR could potentially couple to the MAPK pathway, at least through its interaction with FLNa, whereas the receptor in other subcellular locations might interact with additional scaffold proteins that enable it to modulate the IRS-1/PI3kinase/AKT pathway.
The localization of FLNa CT to membrane ruffles in HepG2 hepatoma may enable it to function as a scaffold upon which reorganization of membrane actin cytoskeleton and cell morphology can take place. Activation of the small GTPases of the Rho family and their downstream targets, Paks, produces phenotypic changes consistent with actin cytoskeletal rearrangements. Notably, FLNa interacts directly with Trio, a unique

FIG. 8. Insulin-induced Elk-1 transactivation in HepG2 cells.
HepG2 cells were transiently transfected with empty vector or HAtagged FLNa CT together with pFR-Luc, pFA-Elk-1, and pCMV-␤-galactosidase and then rendered quiescent by serum deprivation. Following stimulation with 100 nM insulin for 24 h, the lysates were prepared from these cells, and the activities of luciferase and ␤-galactosidase were determined. Normalized transactivation value (luciferase/␤-galactosidase activities) for the unstimulated vector control was assigned a value of 1. Data represent the means Ϯ S.E. from a representative experiment performed in triplicate dishes. Similar results were obtained in a second separate experiment. *, p Ͻ 0.05 versus unstimulated control; **, p Ͻ 0.01 versus insulin-stimulated pcDNA control. Rho GEF (43), and Pak1 (8). It remains to be determined if FLNa regulates the assembly of signaling complexes after activation of the IR. Although insulin treatment did not appear to change the binding of endogenous and transfected FLNa to the IR in HepG2 hepatoma and A7 cells, respectively, these experiments do not rule out a signal-dependent posttranslational modification of FLNa. In this regard, p56 lck has been proposed to phosphorylate filamin, thus controlling its association with cell surface receptors such as ␤ 2 integrins and actin filament cross-linking (44). Moreover, Pak1 controls the actin cytoskeletal and ruffle formation, in part by phosphorylating FLNa (8). FLNa has also been found to undergo phosphorylation in situ in response to serum, lysophosphatidic acid, and other stimuli (45). Thus, it is likely that functional alterations in FLNa might occur as a consequence of phosphorylation events. Incidentally, phosphorylation of IRS-1 on tyrosine residues allows the recruitment of signaling-competent molecules (e.g. PI3kinase), whereas its phosphorylation by Ser/Thr kinases adversely affects the ability of IRS-1 to interact with the activated IR and several signal-transducing proteins, resulting in the development of insulin-resistant states (46,47). The possibility that insulin initiates the phosphorylation of FLNa will be the subject of future investigations. Finally, the importance of FLNa in modulating insulin signal transduction in other insulin-sensitive cells (e.g. muscle and adipose tissue) remains to be determined. The IR-binding properties of FLNa may represent only one of many possible pathways leading to the control of MAPK activation by insulin.