PYK2 as a Mediator of Endothelin-1/Gα11Signaling to GLUT4 Glucose Transporters*

Endothelin-1 (ET-1) signaling through Gαq/11 stimulates translocation of intracellular GLUT4 glucose transporters to the plasma membrane of 3T3-L1 adipocytes by an unknown mechanism that requires protein tyrosine phosphorylation and ADP-ribosylation factor 6 (ARF6) but is independent of phosphatidylinositol 3 (PI3)-kinase. In contrast, insulin action on this process requires PI3-kinase but not ARF6. Here we report the identification of two proteins selectively tyrosine-phosphorylated in response to ET-1 but not insulin: the Ca2+-activated tyrosine kinase PYK2 and its physiological substrate, the adhesion scaffold protein paxillin. Endogenous paxillin as well as expressed Myc-tagged PYK2 or a Myc-tagged kinase-deficient PYK2 protein were acutely directed to F-actin-rich adhesion sites from the adipocyte cytoplasm in response to ET-1 but not insulin. CADTK-related non-kinase (CRNK) is a dominant negative form of PYK2 containing the C-terminal portion of the protein, which binds paxillin but lacks the PYK2 autophosphorylation site (Tyr402). CRNK expression in 3T3-L1 adipocytes inhibited ET-1-mediated F-actin polymerization and translocation of Myc-tagged GLUT4-enhanced green fluorescent protein (EGFP) to the plasma membrane without disrupting insulin action on these processes. These data reveal the tyrosine kinase PYK2 as a required signaling element in the regulation of GLUT4 recycling in 3T3-L1 adipocytes by ET-1, whereas insulin signaling is directed through a different pathway.

Insulin acts through a receptor tyrosine kinase to stimulate the uptake of glucose in fat and muscle cells by a mechanism that involves the translocation of specialized intracellular vesicles containing GLUT4 glucose transporters to the plasma membrane (1,2). Although the signaling mechanism by which insulin mobilizes GLUT4 is not clear, it requires the p85/p110 type phosphatidylinositol 3-kinase (PI3-kinase) 1 and the 3Ј-polyphosphoinositide products it generates (3)(4)(5). GLUT4 translocation can also be elicited by PI3-kinase-independent pathways, as exemplified by the action of endothelin-1 (ET-1 (6 -8)). ET-1 receptors couple to G␣ q and G␣ 11 heterotrimeric G proteins, the latter being abundantly expressed in insulinsensitive 3T3-L1 adipocytes (9,10). Insulin, ET-1, and G␣ 11 (Q209L), the constitutively active mutant of G␣ 11 , all trigger GLUT4 translocation as well as the formation of cortical F-actin (9), thought to be required for optimal GLUT4 recycling to the plasma membrane. However, insulin and ET-1 apparently regulate these processes via discrete signaling pathways. In addition to PI3-kinase independence, ET-1 and G␣ 11 (Q209L), but not insulin, stimulate cortical F-actin polymerization and GLUT4 translocation through a mechanism that requires ADP-ribosylation factor 6 (ARF6) (9,11).
Interestingly, the actions of both insulin and ET-1/GTP␥S are blocked by microinjection of anti-phosphotyrosine antibody into 3T3-L1 adipocytes (12) or by treatment of cells with tyrosine kinase inhibitors (6). However, different sets of proteins are tyrosine-phosphorylated by ET-1 and insulin. In the case of insulin stimulation, many of these proteins have been identified, including the insulin receptor itself and insulin receptor substrate proteins (6), whereas the identity and the function of the tyrosine-phosphorylated proteins in ET-1-treated cells have not yet been characterized. We identify here two such proteins, the Ca 2ϩ -sensitive protein tyrosine kinase PYK2 and its substrate paxillin, and show that PYK2 functions selectively in the ET-1 signaling pathway that leads to GLUT4 glucose transporter translocation to the plasma membrane.

EXPERIMENTAL PROCEDURES
Materials-Endothelin-1 was purchased from the American Peptide Co. pcDNA-CADTK-WT-Myc, pcDNA-CADTK(K457A)-KD-Myc, and pcDNA-CRNK-Flag were kindly provided by Dr. H. S. Earp at the University of North Carolina at Chapel Hill. The GLUT4 cDNA with a C-terminal EGFP and an exofacial Myc tag was cloned into pGreen Lantern vector as described (13). Anti-PYK2 polyclonal (Santa Cruz Biotechnology), anti-paxillin monoclonal, and 4G10 monoclonal (Upstate Biotechnology, Inc.) antibodies were used. All other chemicals were purchased from Sigma.
Cell Culture and Electroporation-3T3-L1 fibroblasts were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, 50 g/ml streptomycin, and 50 units/ml penicillin for 7 days. Cells were then differentiated by adding 0.5 nM isobutylmethylxanthine, 0.25 M dexamethasone, and 5 g/ml insulin for 3 days. Cultures were maintained in regular medium thereafter. All experiments were performed with cells at least 90% differentiated. Four days after the onset of differentiation, 3T3-L1 adipocytes were electroporated (0.18 kV and 960 microfarads) with 50 g of DNA. After 48 h, cells were used for experiments.
Identification of Tyrosine-phosphorylated Proteins-Six days after differentiation, 3T3-L1 adipocytes were starved overnight in Dulbecco's modified Eagle's medium with 0.5% bovine serum albumin and either untreated or stimulated with 10 nM ET-1 or 100 nM insulin for 10 min. Cells were harvested in lysis buffer (20 mM HEPES, pH 7.2, 1% Nonidet P-40, and 1 mM sodium vanadate) supplemented with protease inhibitors and centrifuged for 30 min at 12,000 ϫ g. The supernatants (5 mg of protein each) were incubated for 4 h at 4°C with 100 l 50% slurry of agarose beads cross-linked to the 4G10 monoclonal antibody. After * This work was supported by National Institutes of Health Grant DK30648 (to M. P. C.). 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.
washing, samples were loaded onto 10% polyacrylamide gels and silverstained (Bio-Rad). Selected protein bands were excised and analyzed by trypsin digestion and mass spectrometry as described (14).
Immunoprecipitation and Immunoblotting-The cell lysates (1 mg each) were incubated with 2 g of antibodies for 2 h and with 50 l of 20% agarose beads for an additional hour at 4°C. After washing, samples were loaded onto 10% polyacrylamide gels and analyzed by Western blotting.
Immunofluorescent Microscopy and the Myc-GLUT4-EGFP Translocation Assay-3T3-L1 adipocytes were fixed with 4% formaldehyde in phosphate-buffered saline (PBS), permeabilized, and blocked with 0.5% Triton X-100 and 1% fetal bovine serum in PBS for 20 min. Coverslips were incubated with primary antibodies for 2 h and with FITC-conjugated secondary antibodies for 30 min. For actin staining, rhodaminephalloidin was added at the later step. Images were taken with an Olympus IX-70 inverted microscope with the cooled, thinned, backilluminated CCD camera. Images were processed using Metamorph software (Universal Imaging). For the GLUT4 rim assay, cells were fixed and incubated with anti-Myc 9E10 monoclonal antibody for 2 h and Alexa350-conjugated secondary antibody (Molecular Probes) for 1 h. After another fixation, cells were permeabilized and processed for immunofluorescent microscopy as described above. The specific plasma membrane content of Myc-GLUT4-GFP was estimated using Metamorph and NIH Image, version 1.62 software as described (13). Briefly, the cell surface Myc-GLUT4 content was measured by subtracting the internal Myc signal from total Myc signal of electroporated cells. The total cell EGFP-GLUT4 content was also measured. The specific cell surface Myc-GLUT4 was calculated as: cell surface Myc-GLUT4 ϭ (total Myc Ϫ internal Myc)/total EGFP.

RESULTS AND DISCUSSION
To identify proteins that are tyrosine-phosphorylated in response to ET-1, 3T3-L1 adipocytes were left untreated or treated with either ET-1 or insulin, and then cell lysates were immunoprecipitated with the monoclonal anti-phosphotyrosine antibody, 4G10. As shown in Fig. 1B, insulin and ET-1 stimulated tyrosine phosphorylation of distinct sets of proteins. As expected, the major tyrosine-phosphorylated bands in the insulin-treated sample migrated at positions identical to those of insulin receptor substrates 1/2 (Fig. 1B, IRS-1/2) and the insulin receptor (IR), whereas two major tyrosine-phosphorylated proteins with apparent molecular sizes of 120 and 70 kDa were immunoprecipitated from the lysates of ET-1-treated cells. The 120-kDa protein was clearly visible when the 4G10 immunoprecipitates were analyzed by SDS-PAGE and silver staining (Fig. 1A). Upon tryptic digestion followed by mass spectrometry, the band was shown to contain a peptide (NLL-DAVQAK) corresponding to the calcium-dependent protein ty-rosine kinase PYK2/CADTK/RAFTK/CAK␤/FAK2 (15). Fig. 2A shows that the PYK2 protein specifically immunoprecipitated with PYK2 antibody is indeed tyrosine-phosphorylated in 3T3-L1 adipocytes treated with ET-1 but not in control cells. Insulin had little or no effect on PYK2 tyrosine phosphorylation in several experiments. PYK2 is activated mainly by autophosphorylation at the tyrosine residue Tyr 402 , which can be induced by intracellular calcium (16,17). Tyrosine phosphorylation of PYK2 in intact cells has been reported in response to various stimuli, including G protein-coupled receptor ligands (16), stress signals (18), and cell adhesion (19). Interestingly, this protein kinase was also identified as the major protein tyrosine phosphorylated upon incubation of permeabilized 3T3-L1 adipocytes with GTP␥S (12), suggesting that GTP␥S may also stimulate the PYK2 tyrosine phosphorylation via activation of G␣ 11 .
The protein migrating at 70 kDa (Fig. 1B), selectively tyrosine-phosphorylated in response to ET-1 treatment of adipocytes, was not observed as a unique ET-1-specific band in the silver-stained gels (Fig. 1A). Western blotting suggests that this tyrosine-phosphorylated protein undergoes a significant mobility shift upon stimulation with ET-1, perhaps because of chemical modifications such as tyrosine and serine/threonine phosphorylation. The reduced signal of the silver-stained band at 70 kDa supports this idea (Fig. 1A, lower asterisk). Paxillin, a 70-kDa adapter protein and regulator of the actin cytoskeleton, has been reported to be tyrosine-phosphorylated by PYK2 and displays a similar dispersed banding pattern on SDS-PAGE (20). Therefore, we examined whether paxillin is tyrosine-phosphorylated by ET-1 in 3T3-L1 adipocytes by immunoblotting paxillin precipitates with anti-phosphotyrosine antibody. Fig. 2B shows that ET-1 but not insulin stimulates the tyrosine phosphorylation of paxillin in cultured adipocytes as detected with this method. The same result was obtained by blotting 4G10 immunoprecipitates with anti-paxillin antibody (Fig. 2B). The similarly smeared banding patterns of this protein in Figs. 1B and 2B further imply that the 70-kDa protein is paxillin.
Based on these results, the localization of PYK2 and paxillin proteins in 3T3-L1 adipocytes were analyzed by immunofluorescence microscopy, and their responses to ET-1 and insulin were investigated. Fig. 2C shows that endogenous paxillin displays a rather diffuse distribution near the cell surface of unstimulated cultured adipocytes, and little F-actin is observed. ET-1 treatment of the adipocytes caused a marked recruitment of paxillin to focal adhesions, where paxillin decorated the tips of ET-1-induced cortical F-actin seen at the bottom surface of the cells. In contrast, insulin caused the formation of very thin and shorter actin fibers, with no change in paxillin localization. Endogenous PYK2 could not be detected by immunofluorescence microscopy in 3T3-L1 adipocytes using commercial antibodies. However, when the Myc epitopetagged protein was expressed in adipocytes, PYK2 also dramatically localized to focal adhesions in response to ET-1 but not insulin (Fig. 3).
To test whether PYK2 is involved in F-actin polymerization downstream of ET-1, two dominant inhibitory mutants were utilized. The kinase-deficient PYK2 (PYK2-KD) bears a point mutation (K457A) in the kinase domain and displays minimal tyrosine kinase activity (17). A second mutant (CRNK for CADTK-related non-kinase) contains only the C-terminal portion (Met 865 -Glu 1009 ) of PYK2, which spans the putative focal adhesion targeting domain. The mutant lacks the major tyrosine phosphorylation site (Tyr 402 ) but retains the ability to bind paxillin in vitro. CRNK expression has also been shown to effectively abolish the autophosphorylation of PYK2 in intact cells (17). When electroporated into 3T3-L1 adipocytes, the wild type PYK2, PYK2-KD, and CRNK all showed similar diffuse cytoplasmic localization in the basal state (Fig. 3). In response to ET-1 stimulation, PYK2-KD, like PYK2-WT, localized to focal adhesions and did not inhibit cortical F-actin formation.
The cortical F-actin staining actually appeared to increase in the majority of cells expressing either the wild type or the kinase deficient mutant. Insulin had no effect on the localization of either PYK2 or PYK2-KD. Unlike the full-length proteins, CRNK distribution was mainly cytoplasmic even when cells were stimulated by ET-1. CRNK completely and specifically blocked cortical F-actin polymerization stimulated by ET-1, whereas it had no effect on F-actin formations in response to insulin (Fig. 3). In addition, focal adhesions were not observed in ET-1-treated adipocytes expressing CRNK.
Taken together, the data in Fig. 3 show that ET-1 action in 3T3-L1 adipocytes mobilizes PYK2, independent of its kinase activity, to F-actin-rich adhesion sites at the surface membrane, whereas insulin stimulation does not. That PYK2 is actually required to mediate formation of F-actin in response to ET-1 but not insulin is indicated by the ability of dominant negative CRNK to block cortical F-actin polymerization by ET-1 but not by insulin. PYK2 expression does not itself cause F-actin formation in the basal state, suggesting that its tyrosine phosphorylation in response to ET-1 stimulation is necessary for F-actin formation.
Recent work by our laboratory (9,21) has demonstrated a requirement of intact F-actin for optimal GLUT4 recycling to the plasma membrane in response to either insulin or ET-1 in 3T3-L1 adipocytes. Because we observed that PYK2 is required for F-actin polymerization in response to ET-1 in the cultured adipocytes (Fig. 3), we tested the hypothesis that ET-1 signaling to GLUT4 requires PYK2-mediated F-actin formation. Adipocytes were electroporated with a GLUT4 construct tagged with both a C-terminal EGFP and a Myc epitope (13,22). Translocation of Myc-GLUT4-EGFP was then estimated by measuring the formation of cellular rims in two ways. First, cells with a clearly demarcated EGFP signal around the cell rims were counted (Fig. 4, red bars). Second, the cell surface Myc signal was assessed with anti-Myc antibody in unperme- In addition, the ratio of Myc signal/total EGFP (blue bars, 15ϳ20 cells each) was calculated as described under "Experimental Procedures." The graphs represent values from at least three independent experiments, and the error bars show standard deviations. *, p Ͻ 0.001 over ET-1-stimulated cells not expressing CRNK PYK2 in ET-1-stimulated GLUT4 Translocation 47753 abilized cells (Fig. 4, blue bars). In the basal condition, Myc-GLUT4-EGFP was detected mostly in the perinuclear region in adipocytes, whereas the Myc staining was at a minimal background level. Expression of CRNK in the 3T3-L1 adipocytes had no effect on the intracellular distribution of Myc-GLUT4-EGFP in this condition. When cells were stimulated by insulin, Myc-GLUT4-EGFP rims at the plasma membrane were clearly visible in the majority of cells both by the EGFP signal and Myc staining (ϳ80%, Fig. 4). ET-1 also stimulated the translocation of Myc-GLUT4-EGFP to the plasma membrane, yet to a lesser extent than insulin in about 50% of the Myc-GLUT4-EGFP transfected cells. In both basal and insulin-stimulated conditions, CRNK expression exerted no effect on Myc-GLUT4-EGFP distribution or the numbers of cells with signal intensity at cell rims. In contrast, CRNK expression almost completely blocked Myc-GLUT4-EGFP translocation in response to ET-1 (Fig. 4). The inability of CRNK to repress insulin action on cortical F-actin formation and GLUT4 translocation indicates that the inhibitory effects of CRNK on ET-1 signaling are not due to nonspecific interference of cellular functions. Rather, these data support the hypothesis that PYK2 is required selectively for ET-1 but not insulin action and that cortical F-actin polymerized in response to insulin and ET-1 plays a crucial role in the membrane recycling of GLUT4. The findings reported here indicating a requirement for PYK2 in ET-1 action on GLUT4 are consistent with previous data suggesting PYK2 is itself a target of ET-1 signaling (23). A logical mechanism for such an effect of ET-1 is through its action on G␣ 11 , which activates phospholipase C, leading to the release of intracellular Ca 2ϩ , a known activator of PYK2 (16). However, recently published data are not consistent with a requirement for phospholipase C pathway in ET-1 signaling to adipocyte glucose transport (7,10). Another possible mechanism is the activation of Src kinase by G␣ 11 following its regulation by ET-1. Src kinase has been reported to be associated with PYK2 (24,25) and is required for G protein-coupled receptor-mediated PYK2 regulation (26,27). According to this scheme, Src kinase could then interact with and phosphorylate PYK2 (26,27), thereby leading to the recruitment of other proteins to this signaling complex. The finding that kinasedeficient PYK2 does not interfere with ET-1 signaling to cortical F-actin (Fig. 3) is consistent with a role for phosphorylation of PYK2 by another kinase such as Src. It should also be noted that although ET-1 stimulates glucose transport in primary neonatal rat cardiomyocytes (8), it appears to actually inhibit insulin-stimulated glucose transport in primary adipocytes (28). Thus, the 3T3-L1 adipocyte system used here appears to be a good model for the cardiomyocyte response to ET-1 rather than the response of primary adipocytes.
The present work solidifies our previous conclusion that two quite independent mechanisms account for the actions of insulin versus ET-1 on GLUT4 recycling. A previous component shown to be required specifically for GLUT4 regulation by ET-1 but not insulin action is ARF6 (9), a GTPase involved in actin rearrangements (29,30) and membrane recycling (31,32). Here we implicate PYK2 and paxillin as unique targets of ET-1 in this signaling pathway, whereas insulin signaling to GLUT4 is independent of these proteins. Interestingly, recent reports identified a PYK2-and paxillin-binding protein family that contains an ARF-GAP (GTPase-activating protein) domain, linking PYK2 and paxillin functions to potential regulation of ARF6 (33). ARF6 has recently been shown to play a role in the acute synthesis of PtdIns(4,5)P 2 in membrane ruffles through its ability to activate inositol 4-phosphate 5-kinase ␣ (34).
ARF6 also activates phospholipase D, generating phosphatidic acid, another activator of inositol 4-phosphate 5-kinase ␣ (35). PtdIns(4,5)P 2 in turn regulates multiple proteins involved in actin dynamics (36). Thus, ARF6 may be downstream of PYK2 and paxillin in the ET-1 signaling pathway, perhaps causing cortical actin filament formation, thought to be required for optimal GLUT4 exocytosis (9,21). Further experiments will be required to test this important hypothesis and to clarify the molecular details related to the role of F-actin in GLUT4 exocytosis.