Gβγ Mediates the Interplay between Tubulin Dimers and Microtubules in the Modulation of Gq Signaling*

Agonist stimulation causes tubulin association with the plasma membrane and activation of PLCβ1 through direct interaction with, and transactivation of, Gαq. Here we demonstrate that Gβγ interaction with tubulin down-regulates this signaling pathway. Purified Gβγ, alone or with phosphatidylinositol 4,5-bisphosphate (PIP2), inhibited carbachol-evoked membrane recruitment of tubulin and Gαq transactivation by tubulin. Polymerization of microtubules elicited by Gβγ overrode tubulin translocation to the membrane in response to carbachol stimulation. Gβγ sequestration of tubulin reduced the inhibition of PLCβ1 observed at high tubulin concentration. Gβ1γ2 interacted preferentially with tubulin-GDP, whereas Gαq was transactivated by tubulin-GTP. Prenylation of the γ2 polypeptide was required for Gβγ/tubulin interaction. Both confocal microscopy and coimmunoprecipitation studies revealed the spatiotemporal pattern of Gβγ/tubulin interaction during carbachol stimulation of neuroblastoma SK-N-SH cells. In resting cells Gβγ localized predominantly at the cell membrane, whereas tubulin was found in well defined microtubules in the cytosol. Within 2 min of agonist exposure, a subset of tubulin translocated to the plasma membrane and colocalized with Gβ. Fifteen min post-carbachol addition, tubulin and Gβ colocalized in vesicle-like structures in the cytosol. Gβ/tubulin colocalization increased after pretreatment of cells with the microtubule-depolymerizing agent, colchicine, and was inhibited by taxol. Taxol also inhibited carbachol-induced PIP2 hydrolysis. It is suggested that Gβγ/tubulin interaction mediates internalization of membrane-associated tubulin at the offset of PLCβ1 signaling. Newly cytosolic Gβγ/tubulin complexes might promote microtubule polymerization attenuating further tubulin association with the plasma membrane. Thus G protein-coupled receptors might evoke Gα and Gβγ to orchestrate regulation of phospholipase signaling by tubulin dimers and control of cell shape by microtubules.

Tubulin translocates to the plasma membrane to regulate these G proteins. Activation by G q -coupled muscarinic receptors (3,9) or glutamate receptors (mGluR1␣) (10) causes microtubule depolymerization and association of tubulin with plasma membrane proteins in living cells. Upon binding to the plasma membrane tubulin transactivates G␣ q , which, in turn, activates PLC␤1 (7). However, continued accumulation of tubulin at the membrane inhibits PLC␤1 as that tubulin binds to the PLC␤1 substrate, phosphatidylinositol 4,5 bisphosphate (PIP 2 ) (9). Because the product of PLC␤1-directed PIP 2 hydrolysis, inositol 1,4,5-trisphosphate (IP 3 ), mobilizes stored calcium that evokes microtubule depolymerization (11), feedback inhibition of PLC␤1 may be caused by increased concentration of tubulin dimers in regions close to the plasma membrane (9).
Both G protein ␣ and ␤␥ subunits regulate the assembly of the microtubule cytoskeleton in vitro. Although G␣ activates tubulin GTPase and increases microtubule dynamics (12), G␤␥ promotes microtubule polymerization and stabilizes microtubules in vitro (13). Post-translational isoprenylation of G␥ appears important to this process, because the mutant ␤1␥ 2 (C68S), which is isoprenylation-deficient, and ␤1␥1, which is farnesylated, rather than geranylgeranylated, do not support microtubule formation.
Although this G␤␥ effect on in vitro microtubule assembly has been shown, the potential impact of a tubulin-G␤␥ interaction in the cell and its effect on intracellular signaling has not been investigated. Interaction of G␤␥ and PIP 2 with the G protein receptor kinase 2 (GRK2) assists GRK2 membrane translocation and the subsequent phosphorylation of the activated ␤-adrenergic receptor (14). G␤␥ might be similarly instrumental in translocation of tubulin to its membrane-associated signaling partners. Regulated interactions of tubulin with either G␣ or G␤␥ might differentially affect cellular signaling, because these subunits reciprocally regulate tubulin polymer-ization. By promoting microtubule polymerization in response to a signal, G␤␥ might sequester cytosolic tubulin and inhibit tubulin involvement in the regulation of intracellular signaling.
This report investigates the interaction of tubulin with G␤␥ and the effect of this interaction on the interplay between agonist-evoked PLC␤1 signaling and microtubule assembly. It is suggested that agonist stimulation recruits tubulin to the membrane, where it transactivates G␣ q . Subsequent to PLC␤1 activation, tubulin sequesters with G␤␥, and this complex internalizes to the cytosol. In vitro membrane association/microtubule polymerization experiments suggest that such complexes might seed microtubule polymerization. Data from transactivation, coimmunoprecipitation and cross-linking experiments in vitro, as well as those from confocal immunofluorescence microscopy or coimmunoprecipitation in cells, indicate a direct functional relationship between G protein-coupled receptor signaling and the dynamics of the microtubule cytoskeleton. This report provides first direct evidence for agonist-evoked and G␤␥-mediated cross-regulation between G protein signaling and the dynamics of the microtubule cytoskeleton. Such a mechanism could be instrumental in the regulated reorganization of cytoskeletal elements or other cellular rearrangements leading to synapse formation, cell motility, and cell shape.

EXPERIMENTAL PROCEDURES
Cell Culture-Sf9 cells were maintained in Sf-900 II SFM media as described previously (7). SK-N-SH neuroblastoma cells were grown in Dulbecco's modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum according to standard procedures (7).
Baculovirus-directed Protein Expression in Sf9 Cells-Sf9 cells were simultaneously infected with baculoviruses bearing m 1 muscarinic receptor, G␣ q , and/or PLC␤1 cDNA as described previously (7). The construction of these recombinant baculoviruses was described earlier (15)(16)(17). Membranes were prepared from cells collected 60 h post-infection.
Membrane Preparation and Western Blotting-Sf9 or SK-N-SH cells were sonicated in ice-cold 20 mM Hepes, pH 7.4, 1 mM MgCl 2 , 100 mM NaCl, 1 mM DTT, 0.3 mM phenylmethylsulfonyl fluoride, and membrane pellets were prepared as described (7). Protein concentration was measured by the Bradford dye-binding assay (18) with bovine serum albumin as a standard. Expression of receptors, G proteins, and PLC␤1, was determined by immunoblotting. Membrane proteins transferred to polyvinylidene difluoride membranes (0.45 m, Millipore Corp.) were probed with antisera specific for the m 1 muscarinic receptor (#71, from G. Luthin), G␣ q/11 (#0945, from D. Manning, Philadelphia, PA), or PLC␤1 (K-32-3, monoclonal, from S. G. Rhee, Bethesda, MD) at a dilution of 1:500. Goat anti-rabbit or anti-mouse IgGs were used as secondary antisera, respectively, followed by ECL detection of the corresponding protein bands. Expression levels were estimated by densitometry of the bands (Storm 840, Amersham Biosciences). They varied by no more than 10% for a given recombinant protein. Receptor binding studies using [ 3 H]QNB as a ligand were performed to monitor m 1 muscarinic receptor expression (7). When coexpressed with G␣ q and PLC␤1 in the Sf9 cells, m 1 muscarinic receptor density was estimated at 240 fmol/mg of membrane protein (7). SK-N-SH cells have a high density of m 3 muscarinic receptors (500 fmol/mg of membrane protein) coupled to phosphoinositide turnover (19 -21).
Purification of Proteins-Microtubule proteins were isolated as described previously (22). Microtubule-associated proteins were removed by phosphocellulose chromatography, and the remaining pure tubulin fraction was termed PC-tubulin (23). PC-tubulin was aliquoted and stored in liquid nitrogen until used.
Replacement of GTP in ␤-tubulin was performed as described previously (5). Briefly, GTP was removed from PC-tubulin by charcoal pretreatment followed by incubation of tubulin with 150 M guanine nucleotide (GDP or [ 32 P]AAGTP) for 30 min on ice. Prior to use, these samples were passed through Bio-Gel P6DG desalting columns (Bio-Rad) twice to remove the unbound nucleotide. This procedure yields 0.4 -0.6 mol of guanine nucleotide bound per mol of tubulin dimer. Tubulin-guanine nucleotide concentrations used throughout the study were based on the protein concentration.
Microtubule Assembly-PC-tubulin (1.5 mg/ml) was incubated with SK-N-SH membranes (0.3 mg/ml) and/or G␤␥ (0.13 mg/ml), both in the presence and absence of 100 M carbachol in polymerization buffer (100 mM Pipes, 2 mM EGTA, 3 mM MgCl 2 , 1 mM GTP, pH 6.9). (When tested, G␤␥ was preincubated with tubulin for 30 min on ice.) The assembly reaction was carried out for 30 min at 37°C in a shaking water bath. Centrifugation at 150,000 ϫ g for 30 min at 37°C was performed, followed by separation of the pellets and the supernatants. Pellets were resuspended in identical amounts of cold polymerization buffer, kept on ice for 1 h, and centrifuged at 20,000 ϫ g for 15 min at 4°C, to separate membranes from depolymerized tubulin polymer. Samples were subjected to SDS-PAGE and immunoblotting with a monoclonal anti-␣tubulin antibody (DM1A, Sigma) and ECL detection (Amersham Biosciences). The results were analyzed in a Storm 840 imaging system (Amersham Biosciences). When tested, carbachol had no effect on tubulin polymerization in the absence of SK-N-SH membranes.
Photoaffinity Labeling with AAGTP-Cell membranes were incubated with the indicated concentrations of tubulin-[ 32 P]AAGTP in the presence or absence of carbachol and/or G␤␥ (at a protein ratio with tubulin of 1:5) and/or PIP 2 in buffer A as described previously (26). (When tested, PIP 2 was evaporated under a stream of nitrogen, sonicated for 5 min in buffer A on ice, and preincubated with tubulin for 15 min at 4°C in a bath sonicator.) Following incubation, tubes were UV-irradiated, and the reaction was quenched with ice-cold Pipes buffer, 1 mM MgCl 2 , 4 mM DTT. After centrifugation at 20,000 ϫ g for 15 min, the membrane pellets were washed with buffer A and dissolved in SDS Laemmli sample buffer with 50 mM DTT as described (27). SDS-PAGE of the samples was performed, and the gels were either stained (Coomassie Blue) or subjected to Western blotting, followed by autoradiography (Kodak XAR-5 film) or phosphor image analysis (Storm 840, Amersham Biosciences).
Membrane Association of Tubulin in SK-N-SH Cells-SK-N-SH cells were pretreated with 330 nM colchicine or 330 nM taxol, as described. Cells were collected and washed three times with PBS, and aliquots of 1 ϫ 10 7 cells were distributed in plastic tubes on ice. Carbachol (100 M) was added, and the samples were incubated for the indicated time periods at 37°C with constant shaking. Samples were transferred on ice and immediately sonicated, as described (3). Each sample was centrifuged at 600 ϫ g at 4°C (P 1 pellets). Supernatants were centrifuged at 20,000 ϫ g at 4°C (P 2 pellets), and the P 2 pellets were washed thoroughly with buffer. SDS-PAGE of the P 2 pellet and cytosolic fractions was followed by immunoblotting, using anti-␣-tubulin antibody (DM1A, Sigma) and ECL detection (Amersham Biosciences). The results were analyzed in a Storm 840 image system (Amersham Biosciences).
Immunoprecipitation-Purified G␤␥ subunits (G␤ 1 ␥ 2 , G␤ 1 ␥ 1 , and G␤ 1 ␥ 2 (68S)) and PC-tubulin in a GTP-or GDP-bound form, or devoid of nucleotide, as well as extracts from SK-N-SH neuroblastoma cells, were tested for tubulin/G␤␥ coimmunoprecipitation. Purified G␤␥ subunits and the different PC-tubulin species were tested at a protein ratio of 1:2. When tested, SK-N-SH membranes were extracted with 1% sodium cholate in buffer A for 1 h at 4°C with constant stirring. Samples were centrifuged at 20,000 ϫ g for 15 min at 4°C, and the isolated membrane extracts (0.5 mg/ml membrane protein) were incubated with the indicated tubulin species (1 M). Incubations were carried out in buffer A for 30 min at 24°C and constant shaking. After preclearing (Pansorbin, Calbiochem), each sample was incubated overnight with appropriate specific antiserum or preimmune serum (1:20 dilution) at 4°C with constant shaking. Note, that dimeric tubulin has a very low GTPase activity, which is activated upon polymerization at temperatures higher than 24°C. In addition, no tubulin (or possible aggregates), or G␤␥, were immunoprecipitated without specific antibodies. Immune complexes were precipitated with Pansorbin, and each immunoprecipitate was subjected to SDS-PAGE, followed by immunoblotting and ECL detection of the protein bands. Polyclonal anti-tubulin (raised against the ␤-tubulin C-terminal region of 422-431 amino acids) (3) and anti-G␤ 1 (Santa Cruz Biotechnologies) antisera were utilized. No crossreactivity between these antisera was observed.
Chemical Cross-linking-The method of Tucker and Goldstein (28) was used. Briefly, PC-tubulin, in its GDP-bound form, and purified G␤ 1 ␥ 2 subunits (at a protein ratio of 2:1) were incubated in buffer A for 30 min at 25°C. The cross-linking agent EDC (Pierce) was added to the reaction mixture at the final concentration of 0.20 mM (concentration that limits nonspecific cross-linking), and the incubation was carried out for 2 h at 25°C (28). The reaction was quenched by the addition of 2ϫ SDS sample buffer. Proteins were separated by SDS-PAGE, followed by immunoblotting and ECL detection of the protein bands. Monoclonal anti-␣-tubulin antibody (DM1A, Sigma) and polyclonal anti-G␤ antiserum (Santa Cruz Biotechnologies) were used for immunodetection. PLC␤ 1 Assay-20 g of Sf9 membrane protein was incubated with a [ 3 H]PIP 2 substrate mixture (30 M final concentration) as described (7). 10 l of each GTP␥S, tubulin-GDP, G␤ 1 ␥ 2 , and carbachol was added at appropriate concentrations to a final volume of 120 l as indicated. When tested, GTP␥S was preincubated with the membranes for 30 min on ice. G␤ 1 ␥ 2 subunits were also preincubated with tubulin for 30 min on ice before the experiment. The tubes were incubated for 15 min at 37°C with constant shaking, as described previously (7). [ 3 H]inositol trisphosphate ([ 3 H]IP 3 ) production was measured as described (29).
Analysis of Phosphoinositide Hydrolysis in SK-N-SH Cells-SK-N-SH neuroblastoma cells were grown in 6-well plates in DMEM supplemented with 10% fetal bovine serum and 50 units/ml penicillinstreptomycin. 24 h before the experiment, inositol-free DMEM supplemented with 2 Ci/well myo-[ 3 H]inositol was added. The cells were washed three times with Locke's buffer, containing 10 mM LiCl, and incubated for 1 h with or without 330 nM taxol in the same buffer. After triplicate wash with Locke's buffer, 100 M carbachol was added and the cells were incubated for 30 min at 37°C. Carbachol effects were routinely controlled for by addition of 10 M atropine. The reaction was stopped with ice cold 10% trichloroacetic acid, and the cells were scraped from wells with a rubber policeman and transferred to tubes. After sonication (as described above) and centrifugation at 20,000 ϫ g for 15 min (4°C), the supernatants were extracted with water-saturated ether and neutralized with 1 M NH 4 HCO 3 . Ion exchange chromatography (Dowex AG 1-X8 resin, formate form, Bio-Rad) of the samples was performed as described (29). Total [ 3 H]inositol phosphates were quantified by liquid scintillation counting. The inositol phosphate content of SK-N-SH cells at the start of the experiment (0% increase) was 0.98 Ϯ 0.31 ϫ 10 3 dpm per 10 6 cells.
Microscopy-SK-N-SH neuroblastoma cells were plated onto glass coverslips in 12-well culture plates at a density of 1 ϫ 10 5 . Where indicated, cells were pretreated for 1 h with 330 nM colchicine or taxol. After a PBS wash the cells were treated for the indicated times with 100 M carbachol, 10 M atropine, or both. After washing with PBS buffer the cells were immediately fixed in Ϫ20°C methanol for 3 min and washed three times, 10 min each, in PBS, containing 0.1% Triton X-100. The cells were blocked for 40 min in PBS, containing 5% milk, and washed in PBS. Subsequently, the cells were incubated for 1 h with a polyclonal rabbit anti-G␤ antiserum (Santa Cruz Biotechnologies) at a dilution of 1:100. Following a PBS wash, a secondary FITC (fluorescein isothiocyanate)-conjugated goat anti-rabbit antiserum (EY Labs) was applied for 1 h at a dilution of 1:100. The cells were washed and incubated for 1 h in blocking buffer, containing 10 g/ml rabbit IgG. Subsequently a monoclonal anti-␣-tubulin antibody (DM1A, Sigma) was applied for 1 h at a dilution of 1:500. Following a PBS wash, secondary Texas Red-conjugated goat anti-mouse antibody (Jackson ImmunoResearch Laboratories) (1:100 dilution) was applied for 1 h, followed by washing and mounting of the coverslips. Images were acquired using a laser scanning confocal microscope (Ziess LSM 510) equipped with a ϫ63 water immersion objective. A 488-nm beam from an argon-krypton laser was used for the excitation of FITC, whereas a 543-nm beam was used for Texas Red excitation. Emission from FITC was detected through a BP505 filter, whereas emission from Texas Red was detected through a LP560 filter. Areas of antibody colocalization appeared in yellow. Differential interference contrast images of the cells were regularly acquired as well. Coverslips were examined at random. For each experimental condition a total of 80 randomly selected cells over six consecutive experiments were evaluated for G␤ and tubulin distribution and colocalization. Final image composites were created using Adobe Photoshop 6.0. No specific FITC or Texas Red labeling was observed in cells treated with mouse or rabbit preimmune serum instead of anti-tubulin or anti-G␤ antisera, respectively. Texas Red labeling was not observed when the anti-tubulin antibody was preincubated overnight at 4°C with PC-tubulin (1:1 ratio), and FITC labeling was not detected when the anti-G␤ antiserum was preincubated with purified G␤␥, both conditions tested at the same antibody dilutions (1:100) afterward.
Materials-[␣- 32 1 Activation-To verify the physiological significance of tubulin regulation of PLC␤1 signaling in vivo, SK-N-SH neuroblastoma cells were treated with the microtubule-stabilizing agent taxol before stimulating the cells with the muscarinic receptor agonist, carbachol (Fig. 1). Because taxol increases the formation of microtubule polymers and inhibits microtubule dynamics, the pool of dimeric tubulin available for regulation of PLC␤1 is decreased. As seen in Fig. 1, the normal 6-fold increase in phosphoinositide hydrolysis elicited by carbachol (610 Ϯ 67% (ϮS.D.)) was reduced by half in taxol-pretreated cells (350 Ϯ 26% (ϮS.D.)).

Microtubule Stabilization Inhibits Agonist-induced PLC␤
Agonist-evoked Membrane Association of Tubulin Is Not Assisted by G␤␥ Subunits-It has been previously shown that both G␣ and G␤␥ subunits interact with tubulin and these interactions reciprocally regulate microtubule polymerization in vitro (12,13). Although we have shown that interaction of tubulin with certain G␣ subunits, including G␣ q , regulates intracellular signaling (4 -8), the role of G␤␥ in such membrane-located signaling events has not been evaluated. Because G␤␥ subunits appear to regulate the translocation of signaling enzymes to their membrane-located signaling targets (30 -32), we initially aimed to determine whether G␤␥ also modifies the recruitment of tubulin to the plasma membrane in response to carbachol stimulation (3).
Membranes from SK-N-SH neuroblastoma cells were incubated with tubulin-[ 32 P]AAGTP in the presence or absence of carbachol and/or purified G␤␥ subunits (Fig. 2). As seen previously (7), association of tubulin with SK-N-SH membranes increased by 133 Ϯ 11% (ϮS.D.) after carbachol stimulation. However, when G␤␥ complexes were present in the medium, the carbachol-induced increase in tubulin association with the membrane was blocked. At the concentration used, G␤␥ had no effect on the "basal" association of tubulin with the membrane.

Concomitant Binding of G␤␥ and PIP 2 to Tubulin Prevents
Tubulin Association with the Plasma Membrane-Tubulin binds PIP 2 (7), which appears to act as a membrane anchor (9). Coordinated binding of PIP 2 and G␤␥ to the pleckstrin homology domain of GRK2 synergistically enhances GRK2 membrane association in response to agonist stimulation (14). To test if PIP 2 and G␤␥ similarly affected tubulin, membranes from Sf9 cells containing recombinant m 1 muscarinic receptors, G␣ q and PLC␤1, were incubated with tubulin-[ 32 P]AAGTP in the presence or absence of carbachol, G␤␥, and/or PIP 2 (Fig. 3). Because G␣ q is overexpressed in this system, we initially tested whether exogenous G␤␥ associated with the membranes. Less than 10% of the added G␤␥ subunits bound to the membranes. Carbachol stimulation did not alter this.
Stimulation of m 3 muscarinic receptors by carbachol caused tubulin to associate with the membrane and transactivate G␣ q (182 Ϯ 35% (ϮS.D.) increase in [ 32 P]AAGTP-labeled G␣ q ) (Fig.  3A). G␤␥ inhibited this process by 77 Ϯ 15% (ϮS.D.), and G␣ q transactivation was completely blocked in the presence of PIP 2 and G␤␥. PIP 2 also decreased the amount of purified G␤␥ that associated with the membrane (Fig. 3B). Thus, unlike GRK2/ PIP 2 /G␤␥ (14), tubulin/PIP 2 /G␤␥ complexes did not translocate to the membrane in response to agonist stimulation. G␤ 1 ␥ 2 Subunits Decrease Agonist-evoked Membrane Association of Tubulin by Promoting Tubulin Polymerization-G␤␥ subunits attenuate the carbachol-induced association of tubulin with the plasma membrane. Because carbachol stimulation also appears to cause microtubule depolymerization (3, 7), a possible mechanism for G␤␥ inhibition of these processes is G␤␥-induced microtubule assembly (13). To test this, tubulin was incubated first with G␤ 1 ␥ 2 subunits and, then, with SK-N-SH membranes under conditions that allow microtubule polymerization. Microtubule polymerization increased more than 2-fold (120 Ϯ 13% (ϮS.D.) increase above control) in the presence of G␤ 1 ␥ 2 subunits (Fig. 4). The majority of tubulin was either in microtubules or soluble tubulin dimers. Small amounts of tubulin associated with the SK-N-SH membranes. This might be due to nonspecific membrane association of G␤␥-bound tubulin dimers, because small amounts of G␤␥ routinely associated with the membranes independent of carbachol stimulation. Carbachol stimulation tripled the amount of membrane-associated tubulin (199 Ϯ 37% (ϮS.D.) increase above control), whereas the microtubule mass correspondingly decreased (254 Ϯ 43% (ϮS.D.)). However, microtubule polymerization, but not membrane association of tubulin, was observed in the presence of carbachol and G␤ 1 ␥ 2 subunits. Thus, G␤ 1 ␥ 2 appeared to drive tubulin toward polymerization and away from association with the plasma membrane in response to agonist stimulation. This is consistent with the microtubulepolymerizing role of G␤␥ seen with pure components (13). It is also consistent with the observation that taxol-induced microtubule polymerization inhibits membrane-located PLC␤1 signaling (Fig. 1).
G␤␥ Reverses the Inhibition of PLC␤ 1 Caused by High Concentration of Tubulin Dimers-High, millimolar, concentrations of tubulin inhibit PLC␤1 presumably through receptorindependent association of tubulin with PIP 2 /PLC␤1 sites at the cell membrane (7,9). Diminished access of PLC␤1 to its substrate, PIP 2 , and/or obstruction of G␣ q coupling to PLC␤1, appear to be responsible for this inhibition (9). We wanted to determine whether G␤␥ binding of tubulin dimers would also block this process. Tubulin-GDP was used in these experiments, because it did not transactivate G␣ q and, thus, did not activate PLC␤1 (3).
It has also been shown that G␤ 1 ␥ 2 subunits in which the ␥ 2 polypeptide is isoprenylated promote microtubule assembly (13). To evaluate if this lipid modification is important for G␤␥ interaction with tubulin-GDP dimers, as it is for their microtubule binding, coimmunoprecipitation experiments of tubulin-GDP with G␤ 1 ␥ 2 , G␤ 1 ␥ 1 , and G␤ 1 ␥ 2 (68S) were performed (Fig.  6D). The mutant G␤ 1 ␥ 2 (68S) is unable to undergo post-translational modification and, thus, does not carry a geranylgeranyl moiety (33). G␤ 1 ␥ 2 , but not G␤ 1 ␥ 2 (68S) or G␤ 1 ␥1 (which is farnesylated rather that geranylated), coimmunoprecipitated with the GDP-bound tubulin dimers. These results suggested that geranylgeranylation of the ␥ subunit is important for its interaction with the GDP-bound form of dimeric tubulin.
Chemical Cross-linking of G␤␥ and Tubulin-The interaction between G␤ 1 ␥ 2 and tubulin-GDP was verified by chemical cross-linking (Fig. 7). Tubulin-GDP was incubated with G␤ 1 ␥ 2 or G␤ 1 ␥1, followed by cross-linking of the proteins with the agent EDC, SDS-PAGE, and immunoblotting with anti-G␤ or anti-tubulin antisera. A band consistent with the electrophoretic mobility of cross-linked G␤ 1 ␥ 2 dimers (molecular mass of ϳ40 -42 kDa), and a band, consistent with that of tubulin-G␤ 1 ␥ 2 complexes (molecular mass of ϳ140 kDa), were detected FIG. 4. G␤ 1 ␥ 2 inhibits agonist-evoked membrane recruitment of tubulin due to increased microtubule polymerization. Redistribution of tubulin between microtubules and membranes was studied during the course of carbachol stimulation of SK-N-SH membranes both in the presence and absence of purified G␤ 1 ␥ 2 . PC-tubulin (1.5 mg/ml) was incubated for 30 min at 37°C (conditions that allow microtubule assembly) in the presence of SK-N-SH membranes (0.3 mg/ml), with or without carbachol (100 M), and/or G␤␥ (0.13 mg/ml), as described under "Experimental Procedures." Membranes and microtubules were pelleted by centrifugation at 37°C. Pellets were resuspended in cold polymerization buffer and kept on ice for 1 h to depolymerize microtubules. Tubulin was then separated from the membranes by centrifugation at 4°C. Both tubulin supernatants and pelleted membranes were subjected to SDS-PAGE, immunoblotting with anti-tubulin antibody, and ECL detection of the protein bands. A representative of three identical experiments with similar results is shown. Numerical values are given under "Results." Note that, although carbachol stimulation triggers tubulin association with membranes, the addition of G␤␥ causes microtubule polymerization and an increase in the microtubule pellet even after carbachol exposure. with the anti-G␤ antibody (Fig. 7, lane a). Immunoblotting with the anti-tubulin antibody revealed the presence of both tubulin monomers (at ϳ50 kDa) and cross-linked tubulin dimers (at ϳ100 kDa) (Fig. 7, lane c). A band at 140 kDa was also recognized, confirming the interaction between tubulin and G␤ 1 ␥ 2 . Under the chosen experimental conditions, (28) higher order molecular weight complexes were not detected by either antiserum. When G␤ 1 ␥ 1 subunits were tested, labeling at 140 kDa was barely visible (Fig. 7, lanes b and d). These experiments confirmed tubulin-GDP interaction with G protein ␤1␥ 2 subunits. It also indicated that the stoichiometry of this interaction was 1:1.
Agonist-regulated Spatiotemporal Pattern of Tubulin-G␤ Colocalization in Neuroblastoma SK-N-SH Cells-Because tubulin and G␤␥ interact with each other, they would be expected to colocalize at specific cellular locations. In addition, if tubulin/ G␤␥ interaction is signal-regulated, spatiotemporal colocalization should exist in the living cell. To test this, SK-N-SH cells were studied by confocal microscopy before and after carbachol stimulation (Fig. 8). Tubulin is shown to regulate PLC␤1 signaling in this neuroblastoma cell line (3).
In unstimulated SK-N-SH cells, tubulin and G␤ colocalized sporadically in the cytoplasm (Fig. 8, 0 min). G␤, but not tubulin, was seen at the membrane of the resting cells. Two minutes after carbachol stimulation, microtubules depolymerized and tubulin translocated to the cell membrane (Fig. 8, 2  min). G␤ and tubulin colocalized at the membrane. Fifteen minutes subsequent to carbachol exposure, tubulin and G␤ were colocalized in vesicle-like structures in the cytosol (Fig. 8,  15 min). Diminished association of tubulin with the plasma membrane was seen at this time point. As previously reported (3,7,9), carbachol-induced redistribution of tubulin was inhibited by atropine.
These findings were corroborated by coimmunoprecipitation studies of carbachol-treated SK-N-SH cells (Fig. 9). Two minutes after carbachol stimulation G␤/tubulin complexes were immunoprecipitated from detergent-extracted SK-N-SH membranes (increase by 61 Ϯ 10% (ϮS.D.) compared with control untreated cells). Fifteen minutes later such complexes were observed in the cytosol (increase by 44 Ϯ 17% (ϮS.D.) compared with control cells). Thus, both confocal microscopy and coimmunoprecipitation experiments in cells indicated agonistevoked interaction between tubulin and G␤ at the plasma membrane as well as translocation of these complexes to the cell cytosol at the offset of signaling.
Microtubule Depolymerization Increases Membrane Recruitment of Tubulin and Its Colocalization with G␤-We used compounds that would either increase or decrease tubulin dimer concentrations to evaluate tubulin translocation and colocalization with G␤ in response to agonist stimulation. As previously observed (9), colchicine depolymerized SK-N-SH cell microtubules and increased the amount of membrane-associated tubulin by 44 Ϯ 11% (ϮS.D.) (Fig. 10B). These biochemical data correlated well with images showing colchicine-induced redistribution of tubulin toward the plasma membrane regardless of carbachol stimulation (Fig. 10, A and C, center panels). Tubulin/G␤ colocalization in the cytosol was also readily observed. To the contrary, taxol stabilized the microtubule network and decreased membrane-associated tubulin by 27 Ϯ 17% (ϮS.D.) (Fig. 10B). Images from these experiments showed taxol-bundled microtubules and no tubulin colocalization with FIG. 6. Prenylated G␤ 1 ␥ 2 binds preferentially tubulin-GDP. A and B, PC-tubulin with the indicated nucleotide were incubated with purified G␤␥ (at a protein ratio of 2:1) as described under "Experimental Procedures." Immunoprecipitation was carried out with anti-G␤ 1 (A) or anti-tubulin antiserum (B). Samples were subjected to SDS-PAGE, immunoblotting with either anti-tubulin (A) or anti-G␤ 1 (B) antisera, and ECL detection of the protein bands. C, extracts of SK-N-SH membranes (0.5 mg/ml) were incubated with 1 M PC-tubulin in either GDPor GTP-bound forms as indicated. Immunoprecipitation with anti-tubulin antiserum and immunodetection with anti-G␤ 1 antiserum were performed as in A and B. As also seen in A and B, preferential interaction of G␤ 1 ␥ 2 with the GDP-bound form of tubulin is apparent. D, tubulin-GDP was incubated with either G␤ 1 ␥ 2 , G␤ 1 ␥ 1 , or G␤ 1 ␥ 2 (68S) as indicated. Immunoprecipitation was performed with anti-tubulin antiserum and immunodetection with anti-G␤ 1 antiserum. Tubulin binds to G␤ 1 ␥ 2 but not to G␤ 1 ␥ 1 or the prenylation-deficient G␤ 1 ␥ 2 (68S). The examples shown in A-D are representative of six identical experiments with similar results. Although not shown, each experimental condition described (A-D) included control samples containing preimmune serum instead of the antiserum used for immunoprecipitation. Neither G␤ 1 ␥ 2 nor tubulin precipitation was observed in the absence of antibody. FIG. 7. G␤ 1 ␥ 2 cross-linking to tubulin-GDP. To test the interaction of G␤ 1 ␥ 2 with tubulin-GDP, purified G␤ 1 ␥ 2 subunits (0.63 mg/ml) were incubated with PC-tubulin in the GDP-bound form (protein:protein ratio of 1:2) for 30 min at 25°C. Interacting partners were crosslinked with the agent EDC as described under "Experimental Procedures." Purified G␤ 1 ␥ 1 subunits were employed as a control. Protein complexes were separated by SDS-PAGE, followed by immunoblotting and ECL detection of the protein bands. Monoclonal anti-␣-tubulin antibody (DM1A) and polyclonal anti-G␤ antiserum were used for immunodetection. A representative experiment of four identical ones with similar results is shown. Suggested identity of the cross-linked protein species is indicated. It is evident that, although G␤ 1 ␥ 2 forms complexes with tubulin-GDP, G␤ 1 ␥ 1 does not. membrane G␤ (Fig. 10, A and C, bottom panels). In addition, the G␤-containing endocytic vesicles, observed after 2 min of carbachol stimulation, also lacked tubulin. These findings corroborated the notion that tubulin dimers translocated to the plasma membrane in response to carbachol stimulation might subsequently sequester through G␤-containing endocytic vesicles. DISCUSSION G␤␥ heterodimers perform many diverse intracellular regulatory functions. Adenylyl cyclases and PLC␤ isozymes, ion channels, and kinases are directly regulated by G protein ␤␥ subunits (for reviews see Refs. 34 and 35). The membrane association of G␣ s , G␣ q , and G␣ z requires G␤␥ subunits (36 -38). G␤␥ subunits may also act as "molecular levers" for activated receptors to pry open the G␣ guanine nucleotide binding pocket and thus, release the GDP (38). G␤␥ is involved in the targeting of cytosolic GRKs, phosphoinositide 3-kinase (PI3K), as well as PLC␤1, to the membrane of the cell (30 -32). G␤␥ isoform selectivity for this process has been demonstrated (36,39). G␤ 1 ␥ 2 , but not G␤ 1 ␥ 1 subunits, bind to microtubules and promote microtubule assembly (13) (Fig. 4). G␤␥ subunits regulate the migration of the centrosome around the nucleus and hence, the orientation of the mitotic spindle, in embryos of Caenorhabditis elegans (40). This report suggests that, in the cell, G␤␥ interaction with tubulin is signal-regulated and responsible for both down-regulation of tubulin involvement in PLC␤1 signaling and remodeling of the microtubule network. Cross-regulation between intracellular signaling and cytoskeletal reorganization is proposed.
The ability of G␤ 1 ␥ 2 subunits to affect tubulin regulation of PLC␤1 appeared related to their preferential interaction with tubulin-GDP. Both coimmunoprecipitation and chemical crosslinking experiments showed the specificity of this interaction (Figs. 6 and 7). This was in contrast to the G␣ subunits G␣ i1 , G␣ s , and G␣ q , which interacted preferentially with tubulin-GTP (3-5, 7-9, 25, 41). Such preferred interactions suggest a mechanism by which receptor activation evokes membrane association of tubulin-GTP leading to G␣ q transactivation and initiation of PLC␤1 signaling. This is followed by internalization of tubulin-GDP/G␤␥ complexes at the offset of signaling. They also explain the accelerated microtubule dynamics caused SK-N-SH cells were pretreated for 1 h with or without 330 nM of the microtubule-depolymerizing agent, colchicine, or the microtubule-stabilizing agent, taxol. A, representative images of control and drug-treated cells, as indicated. After the treatment the cells were fixed and incubated with DM1A anti-␣-tubulin antibody and secondary Texas Red-conjugated antibody. Cells were examined by confocal microscopy as described under "Experimental Procedures." Note the extensive microtubule network of control cells, depolymerized tubulin along the cell membrane in colchicine pretreated cells and microtubule bundles in taxol-pretreated cells. B, membrane fractions were prepared from cells similar to those shown in A and membrane-associated tubulin was detected by immunoblotting. A representative experiment of three with similar results is shown. Colchicine by purified G␣ i1 (12) and the increased microtubule polymerization promoted by G␤ 1 ␥ 2 (13) (Fig. 4). Chemical cross-linking also indicated that the ratio of tubulin/G␤␥ binding was 1:1 (Fig. 7). EMAP, an echinoderm MAP, which has significant sequence homology with G␤ (42), binds to both microtubules and tubulin dimers (43). The ratio of dimeric tubulin/EMAP binding is 1:1.
Down-regulation of tubulin involvement in PLC␤1 signaling by G␤␥ subunits was supported by the following observations. G␤␥ subunits attenuated agonist-evoked membrane association of tubulin and G␣ q transactivation by tubulin, both in the absence and presence of PIP 2 (Figs. 2 and 3). Moreover, when membrane association of tubulin and microtubule assembly were tested simultaneously during the course of carbachol stimulation, G␤␥ did not facilitate the membrane binding of tubulin but promoted microtubule polymerization (Fig. 4). In addition, G␤␥ binding of tubulin-GDP reversed the inhibition of PLC␤1 caused by this tubulin (3) (Fig. 5). These results argued against the well-recognized membrane-translocating function of G␤␥ in the case of tubulin. They also suggested that, in the cell, cytosolic G␤␥ subunits could attenuate local increases in tubulin concentration, caused by microtubule depolymerization in response to Ca 2ϩ (11). We have observed that increases in cellular concentrations of tubulin-GDP evoked by colchicine allowed binding of that tubulin to the membrane and inhibited PLC␤1 (9). Adjacent G␤␥ subunits might thus block the inhibition of PLC␤1 by increased local concentrations of tubulin-GDP (7,9).
Contrary to colchicine, taxol potentiates microtubule assembly and stabilizes the microtubule network. However, it also inhibited PLC␤1 signaling (Fig. 1). This effect was not surprising, because dimeric tubulin-GTP transactivated G␣ q (3). Decrease in tubulin-GTP concentration due to microtubule stabilization was apparently responsible for the inhibition of PLC␤1 observed after taxol pretreatment of the cells (Fig. 1). These experiments also confirmed the physiological relevance of tubulin regulation of PLC␤1 signaling.
In unstimulated SK-N-SH cells, tubulin showed minimal colocalization with G␤ at the membrane and in the cytoplasm. Two minutes after carbachol stimulation, microtubule depolymerization, tubulin translocation to the plasma membrane, and colocalization of tubulin with membrane G␤ were observed (Figs. 8 and 10). Fifteen minutes after carbachol stimulation, tubulin and G␤ colocalized in vesicle-like structures in the cytosol (Fig. 8). These observations were supported by coimmunoprecipitation studies utilizing SK-N-SH cells before and after carbachol treatment. Although G␤␥/tubulin complexes were immunoprecipitated from cell membrane extracts at 2 min of carbachol stimulation, at 15 min they were immunoprecipitated from the cell cytosol. When SK-N-SH cells were pretreated with the microtubule-stabilizing agent taxol and studied by confocal microscopy, no tubulin translocation to the plasma membrane or colocalization with membrane G␤ was observed after agonist stimulation (Fig. 10, A and C). Moreover, tubulin was not present in the G␤-containing endocytic vesicles. Pretreatment of the cells with the microtubule-depolymerizing agent colchicine increased tubulin/G␤ colocalization, but this was insensitive to carbachol (Fig. 10C). Colchicine treatment increases cytosolic tubulin-GDP dimers, and we have previously demonstrated that these inhibit PLC␤1, because tubulin-GDP cannot transactivate G␣ q (9). (Note that it ap-pears that more tubulin is associated with membranes from unstimulated cells in immunoblots than appears in those same cells viewed by confocal microscopy. This is probably because tubulin association with membranes increases subsequent to cell disruption (44).) These observations also support the idea that tubulin dimers translocate to the membrane to regulate cellular signaling in response to agonist stimulation.
Coimmunoprecipitation studies and confocal microscopy of carbachol-stimulated SK-N-SH cells, as well as data with taxol and colchicine, corroborate the notion that, although G␤␥ subunits do not influence tubulin translocation to the plasma membrane, they might help sequester tubulin once it becomes membrane-associated. Recent data support this idea by demonstrating that a clathrin-mediated endocytic mechanism is involved in this process. 2 Clathrin-coated vesicles of neuronal cells contain tubulin and G␤␥ (45)(46)(47), and each of them plays a distinct role in vesicle function (48,49). In addition, G␤␥ subunits bind to the i3 loop of m 3 muscarinic receptors (50), known to internalize through a clathrin-mediated endocytic mechanism (51)(52)(53). Although tubulin binds to the mGluR type 1␣ in response to agonist stimulation (10), it does not directly bind to the m 1 muscarinic receptor (7). However, tubulin might sequester with muscarinic receptors through G␤␥ subunits or other scaffold proteins. G␤␥ has been suggested to function as docking unit within larger protein assemblies allowing the interface of selected G protein-coupled receptors with diverse signaling pathways (50). Endogenous G␤␥ has been observed in the cytosol (54) and is associated with microtubules prepared from sheep brain (13).
Although data in this report show that G␤␥ subunits do not directly facilitate membrane translocation of tubulin in response to agonist stimulation (Figs. 2-4), indirect involvement of G␤␥ in this process cannot be excluded. For instance, G␤␥ has been shown to regulate the agonist-dependent translocation of GRK2 and phosphoinositide 3-kinase ␥ (PI3K␥) (31,55,56), but GRK2 is directly responsible for PI3K recruitment to the membrane of the cell (30). Because tubulin, GRK2, and G␤␥ bind to each other (13,57), they might form a signaling complex that translocates to the membrane in response to agonist stimulation. It seems likely that both signal processing and sequestration of signaling molecules involve translocation of distinct sets of signal transduction complexes, where components are physically separated onto distinct protein scaffolds (58).
A simple "antagonist" relationship between G␣ and G␤␥ in microtubule-dependent processes is unlikely to exist. G␣ has been observed to associate with microtubules in developing neurites from PC12 pheochromocytoma cells (59) and a dominant-negative G␣, which binds to microtubules but prevents transactivation of G␣ by tubulin, prevents COS1 cells from sending out cellular processes (60). It is noteworthy in this regard that developing cellular extensions, such as neuronal growth cones, contain microtubules that are more dynamic. Thus, a syncopated interaction between G␣ and G␤␥ may exist in which dynamic microtubules in a developing neurite give way to G␤␥-stabilizing microtubules in a more mature structure (61).
In summary, we have demonstrated that purified G␤␥ subunits do not translocate tubulin to the plasma membrane for the regulation of PLC␤1 signaling but instead promote microtubule assembly. Because G␤ 1 ␥ 2 associates preferentially with tubulin-GDP, this may be relevant for regions of the cell where microtubules depolymerize in response to agonist-evoked elevation in calcium. In these areas, receptor-independent inhibition of PLC␤1 by high tubulin concentrations may be reversed by tubulin interaction with local G␤␥ subunits and subsequent "diversion" of that tubulin into microtubules. Furthermore, during the course of agonist stimulation, membrane G␤␥ subunits could assist sequestration of membrane-associated tubulin and its subsequent internalization. Taken together, this may constitute a novel mechanism for agonist-mediated crossregulation between intracellular signaling and the reorganization of the microtubule cytoskeleton. Such consequence of events could prove important to the regulation of neuronal plasticity and development, cell motility, and changes in cell shape.