Differential Inhibitor of G Signaling to AKT and ERK Derived from Phosducin-like Protein EFFECT ON SPHINGOSINE 1-PHOSPHATE-INDUCED ENDOTHELIAL CELL MIGRATION AND IN VITRO ANGIOGENESIS *

María Luisa Guzmán-Hernández, Aleida Vázquez-Macías, Jorge Carretero-Ortega, Ricardo Hernández-García, Alejandro García-Regalado, Ivette Hernández-Negrete, Guadalupe Reyes-Cruz, J. Silvio Gutkind, and José Vázquez-Prado From the Departments of Pharmacology and Cell Biology, Center for Research and Advanced Studies of the National Polytechnic Institute (CINVESTAV-IPN), Apartado Postal 14-740, D. F. 07000 Mexico and the Oral & Pharyngeal Cancer Branch, NIDCR, National Institutes of Health, Bethesda, Maryland 20892

Heterotrimeric G protein signaling depends on the actions of GTP-loaded G␣ and free G␤␥, the two functional components of the heterotrimer, leading to the generation of second messengers and cell specific functional events (1,2). Differential inhibitors of G␤␥ are required to dissect the biological impact of different G␤␥-dependent effectors. G␤␥ actions can be blocked by competition with peptides derived from its effectors. For example, the effect of G␤␥ on adenylyl cyclase II, G protein-activated inward rectifier K ϩ channel, G protein-cou-pled receptor kinase 2, and phospholipase C␤3, is attenuated by a peptide from adenylyl cyclase II (3). In addition, RACK1 (receptor for activated C kinase 1) selectively inhibits the effect the chemokine receptor CXCR2 on the activation of phospholipase C␤2 and adenylyl cyclase II in HEK293 cells, without affecting other functions of G␤␥ (4). Recently, Smrcka and colleagues characterized the effect of small molecule inhibitors of G␤␥, suggesting their potential application in therapeutic strategies targeting particular G␤␥-dependent pathways (5). Emerging possibilities to target this heterodimer in pathological situations such as inflammation and angiogenesis are based on the role of G␤␥ in cell survival and chemotaxis. To the best of our knowledge, no molecular tool is yet available to differentially inhibit G␤␥ signaling to AKT. 3 G␤␥ is a key transducer of sphingosine 1-phosphate (S1P)elicited angiogenic signals promoting endothelial cell migration, proliferation, and survival (6 -12). Multiple G␤␥-dependent effectors are potentially involved in the molecular events required for endothelial cell migration. These include lipid kinases such as PI3K␥ and PI3K␤ (13), and a novel family of Rac guanine nucleotide exchange factors, represented by P-REX1, which is activated by G␤␥ and phosphatidylinositol 3,4,5trisphosphate (14 -16). G␤␥ signaling is frequently attributed to pertussis toxin-sensitive G i coupled receptors, and it has been consistently revealed by the antagonistic effect of the carboxyl-terminal region of G protein-coupled receptor kinase 2, which sequesters G␤␥ thereby inhibiting all its intracellular actions (17). In addition, mutational analysis of G␤ revealed that different residues, all of them mapping to the interface of contact between G␤␥ and G␣, are important for the activation of distinct G␤␥ effector molecules (18).
Phosducin was originally identified as a phosphoprotein restricted to the retina and pineal gland forming a complex with G␤␥ (19,20). It was considered a protein kinase A-sensitive regulator of G protein-mediated signaling (21,22). Further studies identified a family of phosducin-like proteins (PhLPs) (23,24). Phosducin and G␣ share affinity for the same region of G␤␥, as revealed by the structural analysis of G␤␥ in complex with G␣ or phosducin and by in vitro binding experiments (25). This area of interaction includes some of the residues considered necessary for the activation of G␤␥-dependent effectors (18,26). It was initially postulated that phosducin and related proteins, by interfering with the availability of free G␤␥, exert an inhibitory role on G␤␥ signaling. However, recent genetic evidence raised an apparently conflicting situation; the knockout of PhLP in fungi resulted in a phenotype equivalent to the absence of G␤␥, contrary to its expected role as an inhibitor (27). Novel experimental evidence indicated that PhLP has a positive effect on G␤␥ signaling due to its participation in the assembly of the heterodimer, helping to stabilize free G␤ subunits leaving the ribosome after synthesis (28 -31).
Despite the positive role of full-length PhLP in the assembly of G␤␥ heterodimers, it is still possible that different fragments of this protein, which could retain their interaction with distinct regions of G␤␥, might function as inhibitors of G␤␥ signaling. Accordingly, we characterized here the effect of different PhLP-derived constructs on the signaling pathways elicited by S1P or HGF in endothelial cells. In addition, we explored the mechanism by which PhLP-M1-G149 interferes with G␤␥ preventing the activation of AKT.
Phosducin-like Protein-derived Constructs-Based on the structure of the complex formed by G␤␥ and Phd (25,32,33), and considering the homology between Phd and PhLP, five PhLP-derived constructs were obtained by PCR using as template the cDNA of human PhLP (24) (NCBI accession AF076463, kindly provided by Cheryl Craft, University of Southern California, Los Angeles). According to the structural analysis, these constructs would interact with different amino acids at the effector region of G␤␥. The corresponding cDNAs were cloned into the mammalian expression vectors pCEFL-EGFP, pCEFL-GST, or pCEFL-EGFP-CAAX. EGFP or GST was expressed at the amino terminus of the constructs, whereas in the case of the EGFP-CAAX, the CAAX box was at the carboxyl terminus. All the constructs were cloned with BamHI and EcoRI restriction sites (5Ј and 3Ј, respectively) that were included in the corresponding primers. The sequences of the oligonucleotides are as follows: PhLP (M1-E283) sense, ataGG-ATCCatgcccgagagcctggacagc and antisense, ataGAATTCtcattcaacatcttcttcttc; (PhLP-M1-G149) sense, ataGGATCCatgcccgagagcctggacagccc and antisense, ataGAATTCcccatatctaggcccaaaactc; (PhLP-Q94-E283) sense, ataGGATCCcagagtaggaatggcaaagattc and antisense, ataGAATTCtcattcaacatcttcttcttct; (PhLP-Q94-G149) sense, ataGGATCCcagagtaggaatggcaaagattc and antisense, ataGAATTCcccatatctaggcccaaaactcagc; (PhLP-M1-K102) sense, ataGGATCCcagagtaggaatggcaaagattc and antisense, ataGAATTCtcattcaacatcttcttcttct; and (PhLP-R147-E283) sense, ataGGATCC-agatatgggtttgtgtatgagc and antisense, ataGAATTCtcattcaacatcttcttcttct. For the EGFP-CAAX constructs, the stop codon in the antisense oligonucleotides was omitted.
Cell Lines and Transfections-Human embryonic kidney (HEK) 293T cells were maintained in Dulbecco's modified Eagle's medium (DMEMç Sigma) supplemented with 10% fetal bovine serum (Terra Cell International). For transient transfections, tissue culture plates were treated for 10 min with phosphate-buffered saline (PBS) containing 5 g/ml poly-Dlysine before seeding the cells to prevent them from detaching from the plates during the transfection procedure and thereafter. Cells (2.2 ϫ 10 6 ) were seeded the day before transfection in 100-mm dishes in DMEM supplemented with 10% fetal bovine serum. HEK 293T cells were routinely transfected with Lipofectamine Plus reagent (Invitrogen). Cells were incubated for 5 h in serum-free DMEM with DNA plasmids premixed with Lipofectamine-Plus reagent according to the manufacturer's instructions and were then incubated overnight at 37°C in DMEM containing 10% fetal bovine serum. The total amount of DNA in all transfections was 4 g/plate. When required, empty pCEFL vector was used to maintain a constant amount of DNA.
Interactions between G␤␥ and PhLP Constructs-The interaction between G␤␥ and PhLP-derived constructs was assessed in HEK293T as indicated in the figure legends. G␤␥ pulldown experiments were done using both subunits fused to His 6 amino-terminal tags. Briefly, cells were washed with ice-cold PBS and lysed at 4°C in TBS-T buffer containing 50 mM Tris, pH 7.5, 0.15 M NaCl, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 10 g/ml leupeptin, and insoluble material was removed by centrifugation at 14,000 rpm for 10 min at 4°C. His 6 -tagged G␤␥ was isolated by incubating with 40 l of Talon beads (Metal Affinity Resin, Clontech) for 2 h at 4°C. Beads were washed three times with lysis buffer and boiled for 5 min in 1ϫ Laemmli sample buffer. Expression of proteins and the putative interactions between them were revealed by Western blot using anti-His 6 monoclonal antibodies (Sigma) or GFP monoclonal antibodies Covance). Immunoblots were visualized by enhanced chemiluminescence detection (Pierce). Similar experiments were done to investigate the effect of PhLP-Q94-G149 construct in the interaction between PhLP-M1-G149 and G␤␥. Briefly, HEK 293T cells in 10-cm dishes were transfected with 1.0 g of PhLP-M1-G149, 0.3 g each His 6 -G␤ 1 ␥ 2 and increasing concentration of the plasmid coding for PhLP-Q94-G149 construct (0.3, 1, and 3 g). Pulldown assays were performed 2 days after transfection.
The influence of PhLP constructs on the expression of different G␤ subunits was assessed in HEK293T cells cotransfected with FLAG-tagged G␤ subunits, G␥ 2 , and PhLP, PhLP-M1-G149, or PhLP-Q94-E283. Two days after transfection, cells were harvested, 10 g of total cell lysates was used for SDS-PAGE, and the expression of the different G␤ subunits was detected by Western blot using anti-FLAG antibodies.
Influence of PhLP Constructs on the Interaction between G␤␥ and PI3K␥-The interaction between G␤␥ and PI3K␥ was assessed by coimmunoprecipitation experiments in HEK 293T cells transfected with plasmids expressing 3X-FLAG G␤ 1 , G␥ 2 , PI3K␥ (EE-tagged p110 and EE-tagged p101), kindly provided by L. R. Stephens (34) and PhLP constructs as indicated in Fig. 9. Briefly, 2 days after transfection, cells, grown in 10-cm dishes, were placed on ice, rinsed once with phosphate-buffered saline, and lysed in 1 ml of ice-cold TNTE buffer containing protease inhibitors; insoluble material was removed by centrifugation. After clearing, 3 l of EE antibody (Covance) was added to the supernatant and rotated overnight at 4°C. Next day, 25 l of protein A-Sepharose was added and rotated again for 2 h at 4°C. Immunoprecipitates were washed three times with 1 ml of lysis buffer and resuspended in 1ϫ protein sample buffer. Lysates containing ϳ50 g of total cellular protein or immunoprecipitates were analyzed by Western blotting after SDSpolyacrylamide gel electrophoresis and visualized by enhanced chemiluminescence detection (Millipore).
The influence of PhLP constructs on the recruitment of PI3K␥ to the membrane in response to G␤␥ overexpression was determined by immunofluorescence in PAE cells. Cells seeded on fibronectin (20 g/ml)-coated coverslips were transfected with 100 ng each, of His 6 -G␤ 1 ␥ 2 , EE-PI3K␥ (p101 and p110), and 200 ng of PhLP constructs or 100 ng of EGFP-CAAX as indicated in Fig. 9. Serum-starved cells were processed for immunofluorescence, 2 days after transfection, using previously described conditions (35). Briefly, cells were washed with PBS, fixed with fresh 4% paraformaldehyde in PBS, pH 7.4, for 20 min, washed five times with PBS and permeabilized with 100% methanol for 6 min at Ϫ20°C. and incubated with 1% bovine serum albumin for 30 min at 37°C. PI3K␥ was detected with EE monoclonal antibody (Covance) used at 1/500 in 0.5% bovine serum albumin for 1 h at 37°C, washed, and followed by incubation with Cy5 secondary antibody (Jackson) used at 1/200 in 1% bovine serum albumin for 40 min at room temperature. Cells were mounted with Vectashield (Vector Laboratories), and images were acquired with a DMIRE2 confocal laserscanning microscope (Leica Microsystems, Deerfield, IL) by the use of a 63ϫ, numerical aperture 1.4 oil immersion objective and a zoom of 2.
Stably Transfected Endothelial Cells-All EGFP-tagged constructs derived from PhLP were transfected into porcine aortic endothelial (PAE) cells using Polyfect reagent (Qiagen); control cells were transfected with pCEFL-EGFP. Clones were selected 48 h after transfection with 2.0 mg/ml G418 disulfate salt (Geneticin 418, Sigma) in DMEM supplemented with penicillin/streptomycin (BIOSOURCE, International) and 10% fetal bovine serum. Expression of the transfected constructs was determined by fluorescence-activated cell sorting, immunoblotting, and fluorescence microscopy.
The effect of G␤␥ overexpression on AKT and ERK phosphorylation was assessed in COS7 cells. Briefly, cells were cultured in 6-well plates in DMEM with 10% fetal bovine serum until they reached a 70% of confluence. Cells were then transfected with 0.5 g of plasmids coding for different G␤ subunits (1)(2)(3)(4)(5), G␥ 2 , and PhLP constructs as indicated in the figure legends. Next day, cells were serum-starved, and the day after they were lysed in 200 l of ice-cold TBS-T buffer containing the previously indicated protease inhibitors and phosphatase inhibitors (20 mM ␤-glycerophosphate and 1 mM sodium vanadate), insoluble material was removed by centrifugation. Phos-phorylation of AKT or ERK1/2 was detected as described above.
Effect of PI3K Inhibitors in Endothelial Cells-The effect of PI3K␥ and PI3K␤ inhibitors (36) on the activation of AKT in PAE cells was assessed in serum-starved confluent cultures grown in DMEM in 6-well plates. In brief, cells were serumstarved the night before the experiment. Next day, media was replaced before incubation with the inhibitors, either PIK93 for   (25). B, phosducin-interacting amino acids (within 3-3.5 Å) highlighted in white in G␤␥ as detected with the NCBI's CN3D program. Those interacting with the different regions of phosducin are shown as white residues in structures of G␤␥ next to the corresponding interacting region of Phd. C, based on the structure of phosducin, constructs corresponding to the different regions of PhLP were generated. The amino-terminal domain (M1-K102) corresponds to the dark gray region, the central region (Q94-G149) corresponds to the light gray region and the carboxyl-terminal domain (R147-E283) corresponds to the black part of the structure; right panel. The interaction between PhLP-derived constructs and G␤␥ was assessed in HEK-293T transiently transfected with His 6 -tagged G␤ 1 ␥ 2 and EGFP-tagged PhLP constructs. His 6 -G␤ 1 ␥ 2 was isolated with a cobalt-based metal affinity resin and interacting PhLP derived constructs were detected by Western blotting using anti-GFP antibodies. As a control, a construct corresponding to the carboxyl-terminal region of GRK-2 fused to GFP was used. The expression of transfected EGFP-tagged PhLP or the indicated constructs is shown in total lysates and those interacting with His 6 -G␤ 1 ␥ 2 are shown in the pulldown assay (PD: His). His 6 -G␤ 1 ␥ 2 is shown in the bottom panel.

PhLP-M1-G149, a Differential Inhibitor of G␤␥ Signaling to AKT
using PI3K inhibitors were done by MLGH in Tamas Balla's laboratory at NICHD, NIH using inhibitors kindly provided by Zacchary Knight and Kevan Shokat (University of California, San Francisco).
Cell Migration Assays-Migration assays of stably transfected endothelial cells were performed in Boyden chambers. Filters with 8-m pores were coated overnight with 10 g/ml fibronectin. The bottom wells of the chamber were filled with 150 l of media containing (or lacking) chemoattractant. Stably transfected endothelial cells were harvested by trypsinization, washed by centrifugation, and resuspended 1 ϫ 10 5 cells in 100 l of fresh DMEM. After the cells were inoculated into the upper chamber and incubated for 6 h at 37°C. The migrated cells on the bottom of the chamber were fixed with methanol and stained with 10 mg/ml crystal violet. The cells that remained on the top side of the filter (those that did not migrate) were removed with a cotton swab. The level of migration was determined by densitometric quantitation of the stained filters, using ImageJ software.
In Vitro Angiogenesis Assay-Stably transfected endothelial cells were serum-starved 15 h before the experiment, detached by gently trypsinization, washed, counted, and resuspended in DMEM. Previously, Matrigel basement membrane matrix (BD Biosciences, 354230) diluted 1:2 in cold serumfree DMEM was plated into 96-well tissue culture plates and incubated for 1 h at 37°C before seeding the cells. Then, 10 ϫ 10 3 cells were added on top the Matrigel in the absence or presence of different stimuli (1 M S1P or 0.5 ng/ml HGF). After 9 h of incubation, the experiment was stopped, and the results were assessed morphologically using light microscopy. The length of the tubes was measured using Image J software.
Statistical Analysis-Test and control samples in the functional assays were compared for statistical significance by using Student's t test, a p Ͻ 0.05 was considered significant.

Structural Analysis of G␤␥-Phosducin Complex and Interaction between G␤␥ and Constructs
Derived from PhLP-In this study, we tested the possibility that constructs derived from PhLP, a G␤␥-interacting protein, differentially inhibit G␤␥ signaling to AKT and ERK. To identify the sites of interaction between different regions of phosducin and G␤␥, we analyzed the crystal structure of phosducin-G␤␥ shown in Fig. 1A (25,32,33). According to the analysis of three different structures of G␤␥-phosducin complexes: 1B9Y (33), 2TRC (25), and 1AOR (32), and assuming that phosducin and PhLP interact with the same residues in G␤␥, the amino acids shown in white color in Fig. 1B, bottom right, are those in G␤␥ FIGURE 2. Analysis of the interaction between G␤␥ mutants and constructs derived from PhLP. Upper panel: G␤ 1 with the indicated point mutations at the interface known to interact with G␣ or effectors (18) was cotransfected with G␥ 2 and PhLP constructs in HEK293T cells. Interaction between G␤␥ (wild-type G␤ or the indicated mutant) and PhLP constructs, fused to GST was detected by pulldown assays. The presence of G␤ or the indicated mutants interacting with PhLP or the constructs PhLP-M1-G149 or PhLP-Q94-E283 was detected by Western blot using HA antibodies that recognized this tag fused at the aminoterminal of G␤. As a negative control, a construct corresponding to GST and wild-type G␤␥ was used. The expression of transfected GST-tagged PhLP is shown in the pulldown assays, and the expression of G␤ and the indicated mutants is shown in total cell lysates (TCL). Bottom panel: the importance of G␤ 1 Ser98 and G␤ 1 W332, analyzed with NCBI's CN3D program, is highlighted in the structure of the complex formed by phosducin and G␤␥ (25); as shown, these amino acids establish two interacting spots with the central region of phosducin. According to the pulldown experiments shown in the upper panel, the first interacting contact, depending on G␤ 1 Ser98, is critical for the interaction with full-length PhLP and is also important for PhL-M1-G149. G␤ point mutants revealed that some of these residues are required for the activation of different effectors (13). In particular, the residues of G␤␥ involved in the interaction with phosphoinositide 3-kinase (PI3K), a known effector of G␤␥ leading to AKT activation and other downstream effectors, remain to be determined. To begin addressing this issue, we engineered five PhLP-derived constructs that, according to the structure, are most likely to maintain their conformation and ability to interact with G␤␥ ( Fig. 1, B (middle panel) and C). These constructs covered the amino-(PhLP-M1-K102), central-(PhL-Q94-G149), and carboxyl-(PhLP-R147-E283) regions of PhLP (Fig. 1, B (middle panel, indicated in different intensities of  gray) and C). All the constructs were expressed as amino-terminal EGFP-tagged (GFP) fusion proteins. Their expression was confirmed in HEK-293T cells (Fig. 1C, left panel).

PhLP-M1-G149, a Differential Inhibitor of G␤␥ Signaling to AKT
Analysis of the Interaction between G␤␥ and Constructs Derived from PhLP-The interaction between G␤␥-and PhLPderived constructs was assessed in HEK-293T. His-tagged G␤␥ was isolated under non-denaturing conditions using a cobaltbased metal affinity resin. EGFP-tagged PhLP constructs present in the G␤␥ pulldown assays were detected by Western blot. All the constructs that included the central region of PhLP (Q94-G149) showed a strong interaction with G␤␥, comparable to the interaction detected between G␤␥ and the carboxyl-terminal region of GRK2 used as control. In contrast, the constructs lacking this region (PhLP amino-terminal region, PhLP-M1-K102) and (PhLP carboxyl-terminal region, PhLP-R147-E283) showed a reduced ability to interact with G␤␥ (Fig.  1C). The expression of His 6 -tagged G␤␥ was demonstrated by Western blot with anti-His antibody (Fig. 1C, lower panel). These results showed that the different regions of PhLP, expressed as EGFP-tagged derived peptides, interact with G␤␥, suggesting that they might modulate specific G␤␥-downstream pathways and therefore could be used to dissect G␤␥ signaling events.
To assess the contribution of key residues at the interface of G␤ 1 that are known to interact with G␣ or effectors (18) to the interaction between G␤␥ and PhLP constructs, we assessed the interaction of different G␤ 1 mutants, expressed with G␥ 2 , in PhLP pulldown experiments. As shown in Fig. 2, wild-type G␤␥ interacted with PhLP and the constructs PhLP-M1-G149 or PhLP-Q94-E283 in pulldown experiments using PhLP constructs fused to GST. Interestingly, G␤ 1 Ser98Ala was unable to interact with fulllength PhLP, indicating that the interacting interface formed by G␤Ser98 with the central region of PhLP is critical for the interaction. In Phd, the amino acids that interact with G␤ 1 Ser98 are Phd-E85, Phd-L90, and Phd-R94, establishing a region of electrostatic interactions   JULY 3, 2009 • VOLUME 284 • NUMBER 27

JOURNAL OF BIOLOGICAL CHEMISTRY 18339
between G␤␥ and Phd (25,32,33). The importance of G␤ 1 Ser98 and G␤ 1 W332 was analyzed with the CN3D program (NCBI), and the predicted interface is highlighted in the structure of the complex formed by Phosducin and G␤␥ (25) shown in Fig. 2, lower panel.
The experiments with G␤ point mutants and those comparing the interaction between G␤␥ and different PhLP constructs suggested the importance of the central region of PhLP on the interaction with G␤␥. Thus, we tested the ability of the construct corresponding to this region, PhLP-Q94-G149, to compete with the construct PhLP-M1-G149. This includes the central region and extends toward the amino terminus. As shown in Fig. 3, top panel, the central region of PhLP competed with PhLP-M1-G149 for the interaction with G␤␥, confirming the importance of the PhLP fragment Q94-G149 in the interaction with G␤␥ (Fig. 3).

Effect of PTX and PhLP-derived Constructs on S1P and HGF Signaling to AKT and ERK in Endothelial
Cells-We next tested whether PhLP constructs can differentially interfere with G␤␥ signaling in endothelial cells stimulated with S1P or HGF. Both agonists are known to induce activation of PI3K-AKT, ERK, and chemotactic migration in endothelial cells (10,12,37). The action of S1P activating these pathways is known to be mediated through G i -coupled receptors (10,12,38). Here, using antibodies against the phosphorylated forms of

. Effect of PTX and PhLP-derived constructs on S1P or HGF-induced activation of AKT and ERK in endothelial cells.
A, upper panels: endothelial cells were stimulated with 1 M S1P or 10 ng/ml HGF for the indicated time. After stimulation, phosphorylated AKT and ERK (pAKT and pERK) were detected by Western blotting using phospho-specific antibodies. As a reference, the expression of AKT and ERK was detected in parallel using antibodies that recognize the kinases regardless of their phosphorylation state (bottom panels). B, effect of pertussis toxin (PTX, 100 ng/ml) on S1P or HGF signaling to AKT and ERK in EGFP-transfected endothelial cells. C, effect of PI3K␥ and PI3K␤ inhibitors on the activation of AKT elicited by S1P or HGF. D and E, effect of PhLP-derived constructs on S1P or HGF signaling to AKT and ERK in endothelial cells. The activation of AKT and ERK (pAKT and pERK) was examined with phospho-specific antibodies, and the expression of AKT and ERK was detected in parallel. EGFP-transfected cells were used as control. Bars represent the mean densitometric values of three independent experiments, as the one shown below the graph, normalized to the maximal agonist dependent effect (100%) obtained in HGFstimulated cells; in the case of PhLP-M1-G149-transfected cells, the results correspond to six independent experiments; vertical lines represent the Ϯ S.E. (*, p Ͻ 0.05 respect to control). F, effect of pertussis toxin (PTX) on S1P and HGF signaling to AKT and ERK in PhLP-M1-G149-transfected endothelial cells.

PhLP-M1-G149, a Differential Inhibitor of G␤␥ Signaling to AKT
AKT and ERK, we demonstrated that 1 M S1P and 10 ng/ml HGF induced a rapid phosphorylation of AKT and ERK in PAE cells (Fig. 5A). Pertussis toxin (PTX) differentially inhibited the effect of S1P, but not of HGF, on AKT and ERK phosphorylation (Fig. 5B). These data indicated that S1P regulates signaling to AKT and ERK through a G i heterotrimeric protein in PAE cells. In addition, S1P and HGF signal to AKT via the intervention of different PI3K isoforms as demonstrated by a differential effect of specific inhibitors. As shown in Fig. 5C, S1P signaling to AKT depends on the activity of PI3K␥, and the effect of HGF is independent of this isoform. The potential effect of the different PhLP constructs as modulators of G␤␥ signaling to AKT and ERK was determined in stably transfected PAE cells stimulated with S1P or HGF. PhLP-M1-G149, the construct corresponding to the amino plus the central regions of PhLP, prevented the activation of AKT in cells stimulated with S1P while having no effect on the activation of ERK (Fig. 5, D and E). Other PhLP-derived constructs did not interfere with the ability of S1P to promote the phosphorylation of AKT and ERK (Fig. 5,  D and E). We noticed a slight increase in the basal phosphorylation of ERK in cells expressing PhLP-M1-G149 and PhLP-Q94-E283, but these constructs did not affect the agonist-dependent activation of ERK. On the other hand, cells expressing the smaller constructs, corresponding to the amino-, central-, and carboxyl-terminal regions of PhLP did not show differences, with respect to the control in the activation of AKT and ERK (supplemental Fig. S1). The effect of S1P on AKT and ERK in cells expressing PhLP-M1-G149 was still sensitive to inhibition by PTX (Fig. 5F), indicating that the coupling properties of S1P receptors in these cells were not modified by expression of the PhLP-M1-G149 construct. Neither PhLP-derived constructs nor PTX inhibited the effect of HGF (Fig. 5, D-F).
Effect of PhLP-M1-G149 on GSK3␤ Phosphorylation in Endothelial Cells-Glycogen synthase kinase 3␤ (GSK3␤) is a target of the PI3K signaling pathway, being phosphorylated by AKT and p70S6 kinase (39,40). To assess the role of G i in the phosphorylation of GSK3␤ in endothelial cells responding to S1P or HGF, PTX-treated cells were stimulated with these agonists as indicated in Fig. 6A. Both, S1P and HGF promoted the phosphorylation of GSK3␤, an effect that was sensitive to PTX in S1P-stimulated cells (Fig. 6A, upper panel). To further assess whether PhLP-M1-G149 prevents the phosphorylation of AKT substrates, we evaluated the phosphorylation of GSK3␤ in cells expressing PhLP-M1-G149 and stimulated with S1P or HGF. A, effect of pertussis toxin (PTX, 100 ng/ml) on S1P-or HGF-induced phosphorylation of GSK3␤ (pGSK3␤, an AKT substrate) in EGFP-transfected endothelial cells. EGFP-expressing PAE cells were stimulated for 5 min either with 1 M S1P or 10 ng/ml HGF, phosphorylation of GSK3␤ was detected in total cell lysates using antibodies recognizing the kinase phosphorylated in Ser-9. As a reference, the expression of GSK3␤ was detected in parallel using antibodies that recognize the kinase regardless of its phosphorylation state (bottom panel). B, effect of PhLP, PhLP-M1-G149, or PhLP-Q94-E283 on S1P-or HGF-induced phosphorylation of GSK3␤ in stably transfected endothelial cells; PAE cells expressing EGFP were used as control. After stimulation, total cell lysates were analyzed by Western blotting using phospho-specific antibodies against GSK3␤ or antibodies recognizing the kinase regardless of its phosphorylation state. C, bars represent the mean densitometric values of phospho-GSK3␤ from three independent experiments, as the one shown in B, normalized to the maximal effect (100%) obtained in HGF-stimulated cells; vertical lines represent the Ϯ S.E. (*, p Ͻ 0.05 respect to control).

PhLP-M1-G149, a Differential Inhibitor of G␤␥ Signaling to AKT
The results, shown in Fig. 6 (B and C), indicated that PhLP-M1-G149 prevented the functionality of AKT (assessed by its ability to phosphorylate GSK3␤) in cells responding to S1P. Fulllength PhLP or PhLP-Q94-E283 did not interfere with the phosphorylation of GSK3␤ in cells responding to S1P or HGF (Fig. 6, B and C).
Effect of PhLP-M1-G149 on Endothelial Cell Migration-To examine the effect of PhLP or PhLP-M1-G149 on endothelial cell migration, we assessed the chemotactic response to S1P or HGF of endothelial cells expressing these constructs or control cells expressing EGFP. Interestingly, cells expressing PhLP-M1-G149 showed a notably reduced migration in response to S1P, while the effect of HGF was not affected (Fig. 7, A and B). Full-length PhLP did not interfere with the migration of endothelial cells in response to S1P or HGF, in fact, an increase in the effect of S1P was detected (Fig. 7, A and B).
Effect of PhLP-M1-G149 on in Vitro Angiogenesis-Considering the angiogenic properties of S1P, we tested if the expression of PhLP-M1-G149 affects this response. We assessed the effect of S1P and HGF on the in vitro angiogenic response of PhLP-M1-G149-transfected endothelial cells. Our results indicated that PhLP-M1-G149 was able to interfere with S1P-induced angiogenesis, whereas the effect of HGF was not affected. In parallel, cells transfected with EGFP or EGFPtagged PhLP showed normal angiogenic responses to S1P or HGF (Fig.  8, A and B).
Molecular Mechanism of PhLP-M1-G149-mediated Inhibition-To assess the molecular mechanism by which PhLP-M1-G149 affects G␤␥ signaling to AKT, we tested the hypothesis that this construct interferes with the ability of G␤␥ to interact with PI3K␥ and the ability of this heterodimer to recruit PI3K␥ to the plasma membrane. As shown in Fig. 9A, FLAG-tagged G␤ 1 ␥ 2 interacted with PI3K␥ in coimmunoprecipitation experiments. This interaction was significantly reduced in cells expressing PhLP-M1-G149, but not in cells expressing full-length PhLP or PhLP-Q94-E283. In addition, the association of PI3K␥ to the plasma membrane in response to the overexpression of G␤␥ was attenuated by PhLP-M1-G149, expressed in endothelial cells as a GFP-tagged protein with a CAAX box at the carboxyl terminus to restrict its expression to the membrane. Full-length PhLP or the carboxyl-terminal construct did not interfere with the recruitment of PI3K␥ to the membrane (Fig. 9B). In experiments using higher amounts of transfected G␤␥, PhLP-M1-G149 lost its ability to compete with PI3K␥ (not shown), further indicating that it recognizes the same interacting interface in G␤␥.
Intriguingly, our results showed that full-length PhLP did not interfere with G␤␥ signaling. Because it has been recently reported that PhLP contributes to G␤ expression (31), we tested whether PhLP, PhLP-M1-G149, or PhLP-Q94-E283 could affect the expression of different G␤␥ heterodimers. As shown in Fig. 10, with the exception of G␤ 5 , the expression of all other G␤ subunits, in particular G␤ 1 and G␤ 2 , increased in cells transfected with PhLP. Also, cells expressing PhLP-M1-G149 or PhLP-Q94-E283 showed a slight effect compared with the respective controls.

Effect of PhLP-M1-G149 on the Activation of AKT and ERK in Response to the Overexpression of Different G␤␥
Heterodimers-To test the effect of PhLP-M1-G149 on the action of different G␤␥ heterodimers, COS7 cells were transfected with different G␤ subunits and G␥ 2 . The results shown in Fig. 11A indicated that, by overexpression, only G␤ 1 ␥ 2 and G␤ 2 ␥ 2 were able to provoke a significant increase in the activation of both AKT and ERK. Thus, we tested the effect of PhLP-M1-G149 or full-length PhLP on the phosphorylation of AKT and ERK in response to G␤ 1 ␥ 2 or G␤ 2 ␥ 2 overexpression. As shown in Fig. 11 (B and C), PhLP-M1-G149 significantly attenuated the ability of G␤ 1 ␥ 2 or G␤ 2 ␥ 2 to promote the phosphorylation of AKT, whereas full-length PhLP showed just a, nonsignificant, slight attenuation on the effect of G␤␥ overexpression on AKT phosphorylation. Neither construct showed a significant reduction on the activation of ERK in response to G␤␥ overexpression.

DISCUSSION
Differential inhibitors of G␤␥ could help to determine the cellular roles of different G␤␥ effectors, thus providing the basis for future therapeutic strategies. Here, we demonstrate that PhLP-M1-G149, a construct corresponding to the amino-terminal region of PhLP1, has a differential effect in endothelial cells stimulated with S1P, attenuating the activation of AKT without affecting signaling to ERK, or the activation of both pathways in response to HGF. Both angiogenic factors known to activate AKT, ERK, and endothelial cell migration (41)(42)(43)(44)(45). The effect of PhLP-M1-G149 seems to be due to a blockade of G␤␥ signaling by a direct interference with the effector interface of G␤␥ involved in the activation of PI3K␥. Accordingly, this construct interfered with the interaction between G␤␥ and PI3K␥ and the recruitment of this phosphoinositide kinase to the membrane in response to G␤␥ overexpression. This effect is diminished when the amount of transfected G␤␥ increases, supporting the idea that the mechanism of inhibition is due to a competition for the effector interface. Considering these findings, an inherent limitation on the use of PhLP-M1-G149 as an inhibitor of G␤␥ is the fact that this heterodimer may utilize additional mechanisms to activate AKT in a cell-specific fashion. These include an effect mediated by PI3K␤ (46), growth factor receptor transactivation (47,48), and a recently described ␤-arrestin-dependent mechanism (49). In this regard, PhLP-M1-G149 prevents the activation of AKT in response to S1P in endothelial cells, but it has no activity in the response to this agonist in S1P1-transfected HEK293 cells (not shown), in which G␤␥ mainly uses PI3K␤ to lead to AKT activation (46). A similar, cell-specific effect for a G␤␥ inhibitor has been reported by Hamm's group showing the inhibitory potential of RACK1, a G␤␥ partner that is unable to inhibit CXCR-2-dependent cell migration in HEK293 (50) but attenuates CXCR-4-dependent migration of Jurkat cells (51).
Our findings suggest that the effector interface of G␤␥ linking to PI3K␥ lies at the region of G␤␥ interacting with the amino-terminal part (amino acids M1-K102) of the PhLP-M1-G149 construct. This possibility comes from the different effects of PhLP-M1-G149 and PhLP-Q94-E283; both constructs share the amino acids Q94-G149, but they do not have the same inhibitory effects. According to the available structures of Phd-G␤␥ (25,32,33) the residues of G␤ that interact with the amino-terminal region of Phd, within a distance of 3 Å, are Lys-57, Asn-230, Thr-274, and Asp-290. At this point, the hypothesis linking these residues to PI3K␥ activation is speculative. However, because the crystal structure of this phosphoinositide kinase in complex with G␤␥ is not available, it will be interesting to explore this possibility in future studies. In agreement with these possibilities, Hamm's group recently published that RACK1 interferes in in vitro assays with the ability of G␤␥ to activate PI3K␥ (51). The residues of G␤␥ interacting with RACK1 are not known at the structural level, however, the ability of RACK1 to interact with both G␤ 1 ␥ 2 and G␤ 5 ␥ 2 (52) suggests that the interacting residues correspond to those shared between these G␤ subunits. In addition, because G␤ 5 ␥ 2 , contrary to the action of other heterodimers, is not able to activate PI3K␥ (53), it can be speculated that, although the RACK1-G␤ interface corresponds to residues shared between G␤ 1 and G␤ 5 (52), the effector interface linking G␤ 1 to PI3K␥ (putatively recognized by both RACK1 (51) and PhLP-M1-K102) corresponds to non-conserved residues. In this regard, it is interesting to notice that the amino acids of G␤ 1 that interact with PhLP-M1-K102 are either not conserved between G␤ 1 and G␤ 5 (Thr-274 and Asp-290) or located in a region of G␤ 5 in which the adjacent residues are not conserved (Lys-57 and Asn-230 are conserved in G␤ 5 but Ala-56, Ile-58, and Ala-231 are not). The putatively critical residues for the interaction between G␤␥ and PI3K␥ are conserved in G␤ 1 , G␤ 2 , G␤ 3 , and G␤ 4 , all of them able to activate PI3K␥ (supplemental Fig. S2) (53).
Our results also indicate that the central region of PhLP (amino acids 94 -149) is important to support a stable interaction between PhLP and G␤␥. Independent evidences pointing to this conclusion are derived from the characterization, in transfected cells, of the interaction between PhLP-derived constructs and G␤␥. The shortest PhLP-derived construct that shows an important interaction with G␤␥ corresponds to this central region, which also competes in the interaction between PhLP-M1-G149 and G␤␥. In addition, according to the structure of the complex between G␤ 1 ␥ 2 and Phd (25,32,33), G␤ 1 (Ser-98) and G␤ 1 (Trp-332) are residues involved in the interaction between G␤␥ and the central region of phosducin. In agreement, we found that G␤ 1 (S98A)␥ 2 does not interact with PhLP and decreases its interaction with PhLP-M1-G149, whereas it interacts with PhLP-Q94-E283 as effectively as wildtype G␤ 1 ␥ 2 . These data also suggest that the PhLP-derived constructs have conformational adjustments that allow them to interact with G␤ mutants or isoforms with a different affinity compared with full-length PhLP. Another interesting difference regarding the interacting properties of PhLP-derived constructs is related to their ability to recognize different G␤␥ heterodimers. Although both full-length PhLP and PhLP-M1-G149 interact with heterodimers containing G␤ 1 , G␤ 2 , G␤ 3 , and G␤ 4 , they lack the ability to interact with G␤ 5 . In contrast, PhLP-Q94-E283 recognizes all the heterodimers, including G␤ 5 ␥ 2 . These data also suggest the existence of conformational differences between G␤5 and other G␤ subunits, which, as mentioned, share the ability to activate PI3K␥ (53). In addition, the ability of PhLP-M1-G149 to interact with different G␤␥ heterodimers containing the G␤ subunits known to be able to activate PI3K␥ (53), and interfere with signaling to AKT by G␤ 1 ␥ 2 or G␤ 2 ␥ 2 , support the possibility that PhLP-M1-G149 can be used to characterize the signaling from different G protein-coupled receptors that utilize these different G␤␥ heterodimers to activate the PI3K␥/ AKT signaling pathway. Because it has been recently reported that PhLP acts as a chaperone contributing to the expression of G␤ (31), we tested if PhLP, PhLP-M1-G149, or PhLP-Q94-E283 affected the expression of different G␤␥ heterodimers. The indicated FLAG-tagged G␤subunits were cotransfected with G␥ 2 in HEK-293T cells, and their expression was detected in total cell lysates using FLAG antibodies that recognized this tag fused at the amino-terminal of G␤. Untransfected cells or cells transfected with GST, and the different G␤␥ heterodimers were used as negative controls. Total AKT was detected in the same lysates as a loading control.

PhLP-M1-G149, a Differential Inhibitor of G␤␥ Signaling to AKT
Intriguingly, whereas PhLP-M1-G149 inhibits S1P signaling to AKT, full-length PhLP is not inhibitory. Based on our results, two alternative possibilities can be proposed. First, we observed that full-length PhLP has a positive effect contributing to the expression of different G␤␥ heterodimers, in particular those containing G␤ 1 or G␤ 2 , suggesting that cells that express full-length PhLP also express a higher content of G␤␥ heterodimers that are then available to interact with PhLP but also with PI3K␥ and other effectors. Previous reports assigning an inhibitory role to phosducin and PhLP on G␤␥ signaling were mainly based on in vitro assays or overexpression experiments (54). However, recent findings assign a positive role for PhLP in G␤␥ function (31). In this scenario, G␤␥ folding and heterodimer formation is positively influenced by PhLP (28,55). Our results confirm these findings and extend the observations to different G␤␥ heterodimers, suggesting that, excluding G␤5, the expression of all other G␤ subunits is, to some extent, positively regulated by full-length PhLP. The second possibility that helps to explain why full-length PhLP is not inhibitory emerges from the observed differences in the interacting properties of full-length PhLP and PhLP-M1-G149 with G␤ mutants or different heterodimers. These differences can be interpreted as an indication of conformational differences that, in theory, can confer to the PhLP-M1-G149 construct a better ability to recognize and interfere with the G␤␥-PI3K␥ effector interface.
All together, our results indicate that PhLP-M1-G149 interacts with G␤␥ thereby inhibiting its effector region involved in the activation of PI3K␥/AKT, cell migration, and tubulogenesis, whereas the effector region leading to ERK activation remains available but does not participate in these S1P-dependent cellular events. In conclusion, PhLP-M1-G149 may represent a suitable tool to differentially modulate G␤␥ in the pathway linking this het- FIGURE 11. Effect of PhLP-M1-G149 on the activation of AKT and ERK in response to the overexpression of different G␤␥ heterodimers. A, the effect of overexpressing different G␤␥ heterodimers on the activation of AKT and ERK in COS7 cells was detected by Western blot using antibodies that recognized the phosphorylated forms of these kinases. The expression of total AKT and ERK was detected with antibodies that recognize these kinases regardless of their phosphorylation status. B and C, the effect of PhLP or PhLP-M1-G149 on the activation of AKT and ERK in response to G␤ 1 ␥ 2 or G␤ 2 ␥ 2 overexpression (those that were able to activate both AKT and ERK in the overexpression experiments shown in A) was determined by Western blot using phosphospecific antibodies as described in A. Cells transfected with EGFP and the indicated G␤␥ heterodimers were used as controls. Bars represent the average results of eight independent experiments, vertical lines represent the Ϯ S.E. (*, p Ͻ 0.05 respect to control).

PhLP-M1-G149, a Differential Inhibitor of G␤␥ Signaling to AKT
erodimer to AKT signaling, polarized cell migration, and in vitro angiogenesis.