RACK1 Regulates Specific Functions of G (cid:1)(cid:2) *

, We showed previously that G (cid:1)(cid:2) interacts with Receptor for Activated C Kinase 1 (RACK1), a protein that not only binds activated protein kinase C (PKC) but also serves as an adaptor/scaffold for many signaling pathways. Here we report that RACK1 does not interact with G (cid:3) subunits or heterotrimeric G proteins but binds free G (cid:1)(cid:2) subunits released from activated heterotrimeric G proteins following the activation of their cognate receptors in vivo . The association with G (cid:1)(cid:2) promotes the translocation of RACK1 from the cytosol to the membrane. Moreover, binding of RACK1 to G (cid:1)(cid:2) results in inhibition of G (cid:1)(cid:2) -mediated activation of phospholipase C (cid:1) 2 and adenylyl cyclase II. However, RACK1 has no effect on other functions of G (cid:1)(cid:2) , such as activation of the mitogen-activated protein kinase signaling pathway or chemotaxis of HEK293 cells via the chemokine receptor CXCR2. Similarly, RACK1 does not affect signal transduction through the G (cid:3) subunits of G i , G s , or G q . Collectively, these findings suggest a role of RACK1 in regulating specific functions of G (cid:1)(cid:2) . Heterotrimeric G proteins transduce extracellular signals from a large family of

Heterotrimeric G proteins transduce extracellular signals from a large family of G protein-coupled receptors and mediate intracellular responses critical for many cellular processes, such as vision, taste, metabolism, and neuronal and cardiovascular functions (1). G proteins consist of three subunits, ␣, ␤, and ␥. The signaling function of G proteins was once attributed totally to the G␣ subunit. The G␤␥ subunit was considered merely a membrane anchor and a negative regulator of the G␣. However, it is now clear that it itself plays a prominent role in signal transduction. G␤␥ has a long list of effector and interacting proteins and has been shown to play a dominant role in certain cellular functions. For example, in yeast, G␤␥ is the principal transducer of the mating signal for cell cycle arrest and differentiation (2). Chemotactic responses of leukocytes (3,4) and Dictyostelium discoideum amoeba (5,6) are mediated through G␤␥. Recently, G␤␥ has also been implicated in smooth muscle cell proliferation and arterial restenosis (7).
The diversity of G␤␥ target proteins raises the question of how the specificity and efficiency of G␤␥ signaling are regulated. Several G␤␥-interacting proteins have been shown to regulate its function. For example, phosducin and phosducin-like proteins (PhLPs) 1 bind G␤␥ with high affinities and are regarded as scavengers of G␤␥ (8). Binding of these proteins to G␤␥ limits the amount of G␤␥ available to interact with G␣ and to form functional G protein heterotrimers, resulting in inhibition of signal transmission from receptors to G proteins (9,10). Moreover, they block the activation of effectors by G␤␥, because their binding sites on G␤␥ overlap with the contact residues for G␤␥ effectors (11,12). Some G␤␥ effectors, such as G proteincoupled receptor kinases (GRK) 2 and 3, could also exert regulatory roles in G␤␥ signaling, because binding of these proteins impede the access of other effectors to G␤␥ (13). Notably, these G␤␥-interacting proteins appear to affect various functions of G␤␥. However, whether they could modulate the signaling specificity of G␤␥ remains unknown.
Recently, through the use of a yeast two-hybrid screen, glutathione S-transferase (GST)-pull-down and co-immunoprecipitation studies, we identified a new G␤␥-binding protein, RACK1 (14). RACK1 is a WD40 repeat protein that is proposed to form a seven-bladed ␤ propeller, similar to that formed by G␤ (15). Although originally identified as a receptor for binding and translocating activated PKC to particular cell fractions, subsequent studies indicate that it can also act as an adaptor/ scaffold protein to recruit or assemble many other signaling proteins into membrane-associated complexes (15).
We showed previously that RACK1 interacts not only with the G␤ 1 ␥ 1 subunit but also heterotrimeric G␣ t ␤ 1 ␥ 1 (14). Moreover, G␤ 1 ␥ 1 competes with several RACK1-interacting proteins such as PKC␤II and dynamin-1 for binding to RACK1 (14). However, the functional significance for the interaction between RACK1 and G␤␥ on G protein signaling remains unknown.
In this study, we provide evidence that RACK1 interacts only with free G␤␥ subunits released from activated heterotrimeric G proteins in cells. Moreover, we show that the binding of G␤␥ to RACK1 results in the translocation of RACK1 from the cytosol to the plasma membrane, and the association of RACK1 with G␤␥ specifically inhibits G␤␥ subunit-but not G␣ subunitmediated signal transduction. Particularly, only a subset of G␤␥ functions is affected by RACK1. To our knowledge, this is the first protein identified to be able to regulate specific functions of G␤␥.

EXPERIMENTAL PROCEDURES
Materials-Cyclic AMP assay kits were from R&D systems, myo-[ 3 H]inositol (80 Ci/mmol) and phosphatidylinositol-4,5-biphosphate (5.5 * 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  Ci/mmol) were from PerkinElmer Life Sciences. AG1-X8 (100 -200 mesh, formate form) was from Bio-Rad. Mouse anti-RACK1 antibody was from BD Transduction Laboratories, mouse anti-V5 from Invitrogen, rabbit anti-G␤ (T20) from Santa Cruz Biotechnology, Inc., and p42/44 MAPK and phospho-p42/44 MAPK antibodies from Cell Signaling Technology, Inc. Nickel-nitrilotriacetic acid-agarose was from Qiagen, amylose resin from New England Biolabs, and glutathione-Sepharose from Amersham Biosciences. Other chemicals were of the highest grade available commercially.
Purification of Other Proteins-Maltose-binding protein (MBP), MBP-RACK1, and G␣ s short isoform were expressed in Escherichia coli BL21 cells and purified as described (14,18). PLC␤2 was expressed by baculovirus infection of Sf9 cells and purified as described (17).
Vectors and DNA Constructs-The hamster ␣ 1B -, ␣ 2A -, and ␤ 2 -AR cDNAs in pMT2, and G␣ q Q209L in pCMV were kindly provided by Dr. Robert M. Graham (Victor Chang Cardiac Research Institute, Australia). To construct the eukaryotic expression vector for expression of RACK1 with the GST epitope at the N terminus or V5 epitope and His 6 tag at the C terminus, RACK1 was first amplified by PCR and ligated into the pENTR/SD/D-TOPO entry vector using a GATEWAY cloning system (Invitrogen). Homologous recombination reactions were then conducted to transfer RACK1 cDNA from the entry vector into the expression vectors, pDEST27 containing the N-terminal GST epitope and pcDNA3-DEST40 containing the C-terminal V5 epitope and His 6 tag (Invitrogen).
Cell Culture and Transfection-COS-7 and HEK293 cells (American Type Culture Collection, Rockville, MD) were cultured and transiently transfected with the indicated constructs using the LipofectAMINE 2000 reagent (Invitrogen), as described previously (19). Cells were harvested 48 -60 h post-transfection.
GST Binding Assays in Vivo-To determine the association of RACK1 with G proteins in cells, GST fusion RACK1 was used, because it expressed at a much higher level than other RACK1 tagged with other epitopes. COS-7 cells were transiently transfected with receptors and G proteins, together with GST or GST-RACK1 as described above. After serum-starvation overnight and agonist stimulation, cells were incubated with the cross-linking reagent dithiobis(succinimidylpropionate) (DSP) (2 mM) in phosphate-buffered saline (PBS) for 20 min at room temperature and then lysed in Tris buffer containing 10 mM Tris-HCl, pH 7.4, 200 mM NaCl, 1 mM EDTA, 1% Triton, 0.5% deoxycholate, 0.1% SDS, and protease inhibitors. GST-RACK1 was precipitated from the cell lysates using 100 l of 50% slurry of glutathione-Sepharose beads, and the protein complex was subjected to SDS-PAGE and immunoblot analysis (14).
Immunofluorescence and Confocal Microscopy-Sixty hours post transfection, COS-7 cells grown on coverslips were fixed in 4% paraformaldehyde and permeabilized with 0.5% Triton X-100. After staining with a monoclonal antibody against RACK1 or the V5 epitope (1:100 dilution), a secondary Alexa 568-conjugated anti-mouse antibody (Molecular Probes) was applied at 1:100 dilution for 1 h at room temperature. Fluorescence was analyzed using a LSM510 confocal scanning microscope (Carl Zeiss, Germany) with an argon/krypton laser and a Plan Apo 63 ϫ 1.3 numerical aperture oil immersion lens. YFP and Alexa-568 were excited at 488 and 568 nm and monitored with 515-540 nm and 590 -610 nm band-pass filters, respectively. Specificity of labeling and absence of signal crossover were established by examination of single-labeled samples.
Binding of RACK1 to Lipid Vesicles-Lipid vesicles were prepared by drying a mixture of lipids (1.7 mg/ml) containing 95% phosphatidylcholine and 5% phosphatidylinositol under an N 2 stream (20). The lipids were then resuspended in PBS containing protease inhibitors and sonicated in a water bath sonicator for 20 min at room temperature. 10 pmol of MBP or MBP-RACK1 was then added to the lipid vesicles (30 l) in the absence or presence of G␤␥ (10 pmol). After incubation on ice for 1 h, lipid vesicles were pelleted by centrifugation at 100,000 ϫ g for 30 min at 4°C. The pellet was washed once with PBS. Proteins in the supernatant and the pellet were then resolved by SDS-PAGE and subjected to immunoblot analysis.
Immunoblotting-To determine the expression of GST-RACK1 in COS-7 cells, cell lysates were prepared from the transfected cells, and 100 g of protein was used for the detection of RACK1 using a monoclonal antibody (14).
Phosphatidylinositol Hydrolysis in Intact Cells-Phosphatidylinositol (PI) hydrolysis in intact, transfected COS-7 cells was determined as described previously (21,22). To determine the effect of pertussis toxin on PI hydrolysis, cells were preincubated with pertussis toxin (0.5 g/ml) overnight before PI measurement.
Measurement of PLC␤2 Activity-The G␤␥-mediated PLC␤2 activation was determined essentially as described (17), except that purified G␤ 1 ␥ 2 from Sf9 cells instead of Sf9 cell lysates was used in the assay. To determine the effect of RACK1, MBP-RACK1 was preincubated with G␤␥ for 30 min before the addition of PLC␤2 and lipid vesicles.
AC Assays-The G␤␥-mediated ACII activation was determined using cell membranes prepared from ACII baculovirus-infected Sf9 cells as described, except that GTP␥S-bound wild-type G␣ s instead of the constitutively active mutant G␣ s (Q227L) was used (17). To determine the effect of RACK1, MBP-RACK1 was preincubated with G␤␥ for 30 min before addition into the reaction. The reaction was terminated by the addition of 0.1 M HCl. cAMP levels were determined using the cAMP immunoassay system (R & D Systems) for nonacetylated cAMP according to the manufacturer's instructions.
MAPK Activity-HEK293 cells transiently transfected with the ␣ 2A -AR, GST, GST-RACK1, or G␣ t DNAs as described above were stimulated with (Ϫ)-epinephrine in the presence of 10 M dl-propranolol for 5 min at 37°C. Cell lysates were then prepared and used for detection of MAPK activity by Western blotting (23).
Measurement of Cyclic AMP in Intact Cells-HEK293 cells stably transfected with CXCR2 chemokine receptor were used in this assay. Forty-eight hours after transfection with GST, GST-RACK1, or RACK1-V5-His, HEK293 cells were incubated with 0.1% bovine serum albumin in serum-free medium overnight. To determine the endogenous ␤ 2 -ARmediated cAMP accumulation, cells were preincubated with 250 M 3-isobutyl-1-methylxanthine for 20 min before the addition of (Ϫ)-isoproterenol (22). For CXCR2-mediated inhibition of cAMP accumulation, cells were pretreated with 250 M 3-isobutyl-1-methylxanthine and 20 nM interleukin 8 (IL-8) for 20 min before the addition of 10 M forskolin. The reaction was terminated after 20 min by removal of the medium and addition of 0.1 M HCl. cAMP levels were determined using the cAMP immunoassay system described above.
Data Analysis-Data were representative of at least three independent experiments. Results are expressed as the mean Ϯ S.E. Student's t tests were used to determine significant differences (two-tail p Ͻ 0.05).

Interaction of RACK1 with G Proteins in Vivo-
We showed previously that RACK1 not only binds purified retinal G␤ 1 ␥ 1 but also G␣ t ␤ 1 ␥ 1 in vitro (14). To determine if these observations could be extended to other G protein subunits in vivo, we co-expressed G␤ 1 ␥ 2 with GST or GST-RACK1 in COS-7 cells, and precipitated it from cell lysates using glutathione-Sepharose. As shown in Fig. 1A, G␤ was detected in the precipitate containing GST-RACK1, but not GST, indicating a specific association of RACK1 with G␤␥. Intriguingly, when G␤␥ was co-expressed with G␣ subunits, G␣ i-2 , G␣ q , or G␣ s , binding of G␤␥ to RACK1 was significantly inhibited (Fig. 1, B-D). Coimmunoprecipitation studies indicated that the co-expressed G␣ subunits form heterotrimers with G␤␥ (data not shown). This suggests that heterotrimeric G proteins do not bind or have lower binding affinity to RACK1. In support of this contention, when heterotrimeric G protein were activated to release G␤␥ subunits by co-expressed receptors, the ␣ 2A -, ␣ 1B -, and ␤ 2 -ARs, respectively, the association of G␤␥ with GST-RACK1 was restored (Fig. 1, B-D). However, no binding of G␣ subunits to RACK1 was detected despite their significant expression in cell lysates (Fig. 1, B-D). These results indicate that RACK1 binds only the free G␤␥ subunit but not the heterotrimeric G protein nor the G␣ subunit in vivo.
G␤␥ Targets RACK1 to Cell Membranes-We next examined if interaction with G␤␥ alters the cellular distribution of RACK1. As shown in Fig. 2A (panels a and b), in the absence of overexpressed G␤ 1 , endogenous RACK1 and RACK1-V5-His were evenly distributed in the cytosol of COS-7 cells. In con-trast, a majority of YFP-G␤ 1 ␥ 2 expressed in COS-7 cells was located in the cell membranes, although a small fraction was also detected in the cytosol ( Fig. 2A, panel c). When YFP-G␤ 1 ␥ 2 was co-expressed with RACK1-V5-His, the cellular distribution of YFP-G␤ 1 ␥ 2 was not significantly altered ( Fig. 2A, panels c-e). However, it resulted in a punctate distribution of RACK1-V5-His both in the cell membranes and in the cytosol where it co-localized with YFP-G␤ 1 (Fig. 2A, panels c-e). These results suggest that G␤␥ may target RACK1 to the cell membranes. To confirm these findings, we further assessed the effect of G␤␥ on the association of an MBP-RACK1 fusion protein with lipid vesicles in vitro. As shown in Fig. 2B, after incubation with the lipid vesicles followed by centrifugation to separate the bound from the free MBP-RACK1, MBP-RACK1 was found in both the supernatant (45%) and the lipid membrane (55%). The association of MBP-RACK1 with the lipid membrane is most likely due to MBP, because MBP alone was predominantly associated with the lipid (75% in the lipid membranes). Notably, the addition of either G␤ 1 ␥ 1 or G␤ 1 ␥ 2 to the lipid vesicles significantly enhanced the association of MBP-RACK1 with the lipid (70 and 80% for ␤ 1 ␥ 1 and ␤ 1 ␥ 2 , respectively), although it has little effect on MBP alone (Fig. 2B). Taken together, these findings suggest that the physical interaction between G␤␥ and RACK1 facilitates the translocation of RACK1 from the cytosol to the membrane.
RACK1 Inhibits Specific G␤␥ Functions-Because RACK1 binds to G␤␥ in cell membranes where the interaction of G␤␥ with effectors occurs, we questioned if RACK1 could affect the ability of G␤␥ to activate effectors such as PLC␤2. As shown in Fig. 3A, co-expression of G␤ 1 ␥ 2 with PLC␤2 in COS-7 cells resulted in a 2-to 3-fold increase in inositol phosphate (IP) accumulation. Co-expression of GST-RACK1 had little effect on the basal IP turnover but significantly inhibited the G␤ 1 ␥ 2mediated IP signaling. This effect was specific to RACK1, because co-expression with GST had no effect on the G␤ 1 ␥ 2stimulated IP generation (Fig. 3A). Moreover, like GST-RACK1, overexpression of RACK1-V5-His inhibited G␤ 1 ␥ 2mediated PLC␤2 activation (data not shown). The magnitude of the inhibitory effect of GST-RACK1 depended on the level of its expression (data not shown). The transfection of 4 g of GST-RACK1 plasmid relative to 1 g of G␤ 1 ␥ 2 plasmid inhibited G␤ 1 ␥ 2 -mediated IP accumulation by ϳ50% (Fig. 3A).
We then evaluated if RACK1 could interfere with receptor- mediated PLC activation through endogenous G␤␥. The ␣ 2A -AR was chosen for this study, because it was known to activate phosphatidylinositol (PI) hydrolysis through the G␤␥ released from the activated heterotrimer G i . As reported previously (22,24) and shown in Fig. 3B, stimulation of COS-7 cells transiently expressing the ␣ 2A -AR with the ␣ 2A -AR agonist UK14,304, resulted in a small but significant increase in IP accumulation, which was completely abolished by pertussis toxin pre-treatment. Co-expression of GST-RACK1 blocked the ␣ 2A -AR-mediated IP response (Fig. 3B). Similarly, the IP response initiated by the chemokine receptor CXCR2 transiently expressed in COS-7 cells via the pertussis toxin-sensitive pathway was also inhibited by RACK1 (data not shown).
Because RACK1 can interact with many target proteins other than G proteins, it is possible that the inhibition of G␤␥ signaling by RACK1 was mediated indirectly through the perturbation of other signaling pathways regulated by RACK1. To evaluate this possibility, the effect of RACK1 on G␤␥-mediated PLC␤2 stimulation was further evaluated in vitro using heterologously expressed and purified proteins. As shown in Fig. 3C, 10 M MBP-RACK1 specifically inhibited the G␤ 1 ␥ 2 but not the Ca 2ϩ -mediated increase in 3 H-labeled phosphatidylinositol 4,5-biphosphate hydrolysis. The inhibitory effect of MBP-RACK1 on G␤ 1 ␥ 2 signaling was dose-dependent, with an IC 50 of ϳ4.9 M (Fig. 3D). This effect was specific, because neither G␤ 1 ␥ 2 nor Ca 2ϩ -mediated PLC␤2 stimulation was affected by MBP alone. Moreover, because MBP-RACK1 did not affect Ca 2ϩmediated PLC activation, the inhibition of G␤ 1 ␥ 2 signaling is unlikely due to a direct inhibition of PLC enzyme activity. Therefore, the inhibition of RACK1 on G␤␥ signaling is directly related to its binding to G␤␥.
To determine if RACK1 functions as a general inhibitor of G␤␥-dependent signaling processes, we evaluated the effect of MBP-RACK1 on G␤␥-mediated ACII activation. As shown in Fig. 4A, the activity of baculovirus-expressed ACII in Sf9 cell membranes was stimulated 100-and 150-fold by GTP␥S-bound G␣ s (10 nM) and forskolin (10 M), respectively. G␤ 1 ␥ 2 (50 nM) itself had little effect on the activity of ACII (data not shown), but it enhanced G␣ s -mediated ACII activation by 2.5-fold (Fig.  4A). 10 M MBP-RACK1 inhibited basal ACII activity by 50% and G␤ 1 ␥ 2 -mediated stimulation by 80% but had little effect on forskolin-or G␣ s -stimulated activity. At the same concentration, MBP did not affect the basal activity of ACII, but caused about 40% inhibition of the G␤ 1 ␥ 2 -mediated stimulation (Fig.   FIG. 3. RACK1  4A). These data suggest that MBP may have nonspecific effects on the G␤ 1 ␥ 2 -mediated ACII activation. However, because the effect of MBP-RACK1 was significantly greater than that of MBP alone, the additional inhibition of the G␤ 1 ␥ 2 -mediated stimulation by MBP-RACK1 is likely specific to RACK1. The reason that MBP-RACK1 also inhibited the basal activity of ACII is unknown, but it seems to be specific to RACK1, because MBP did not have an effect. Moreover, it is unlikely that MBP-RACK1 directly inhibited the enzyme activity of ACII, because the activity stimulated by forskolin or G␣ s was unaltered by MBP-RACK1 (Fig. 4A). Additional studies indicate that, as with the inhibition of the G␤ 1 ␥ 2 -mediated PLC␤2 activation, MBP-RACK1 abolished the G␤ 1 ␥ 2 -mediated ACII stimulation in a dose-dependent manner (Fig. 4B). Notably, the IC 50 of MBP-RACK1 for inhibiting G␤ 1 ␥ 2 -mediated ACII activation (2.5⌴) was close to that for inhibiting PLC␤2 (4.9 M), suggesting that inhibition occurs through a similar mechanism.
To determine if other G␤␥ functions were inhibited by RACK1, we tested the ␣ 2A -AR-mediated MAPK activation in HEK293 cells. As shown in Fig. 5 (A and B), activation of the ␣ 2A -AR induced phosphorylation of p42 and p44 MAPKs in a dose-dependent manner. These responses were largely abolished by pertussis toxin pre-treatment or co-transfection with G␣ t , suggesting that they were mediated by G␤␥. However, despite a 2-fold increase over the level of the endogenous RACK1, overexpression of GST-RACK1 did not affect the ␣ 2A -AR-initiated MAPK activation. Similarly, overexpression of RACK1 does not affect the ␣ 2A -AR-mediated MAPK activation in COS-7 cells (data not shown). In addition, similar results were obtained when the CXCR2-mediated MAPK activation was examined (data not shown).
We also evaluated the consequence of GST-RACK1 overexpression on chemotaxis of HEK293 cells transiently expressing CXCR2, a response that was mediated by G␤␥. As shown in Fig. 5C, IL-8 induced chemotaxis of HEK293 cells expressing the CXCR2, which was inhibited by co-expressed G␣ t , supporting the involvement of G␤␥. Overexpression of GST-RACK1 at a level similar to that shown in Fig. 5A, however, had no effect on CXCR2-induced chemotaxis. These data suggest that RACK1 regulates specific G␤␥ functions.
RACK1 Does Not Affect G␣ q -, G␣ i -, or G␣ s -mediated Signaling-To further explore the specificity of RACK1 on G protein signal transduction, we evaluated the effect of RACK1 on signaling mediated by the G␣ q , G␣ i , and G␣ s subunits. For evaluating the G␣ q signaling pathway, we transiently expressed the ␣ 1B -AR or the constitutively active G␣ q Q209L mutant in COS-7 cells. As seen in Fig. 6 (A and B), activation of the ␣ 1B -AR resulted in G␣ q -mediated PLC activation, whereas overexpression of the constitutively active G␣ q Q209L mutant caused an increase in basal IP production. Neither the ␣ 1B -ARnor the G␣ q Q209L mutant-mediated IP generation was affected by GST-RACK1 when it was expressed at a level that was not significantly different from that seen in Fig. 3A. To evaluate the effect of RACK1 on cAMP signaling mediated by the G␣ i or G␣ s subunit, we used HEK293 cells that stably express the chemokine receptor CXCR2 and endogenously express the ␤ 2 -AR. Stimulation of the cells with forskolin resulted in a marked increase in cAMP levels (Fig. 6C). Pretreatment with the CXCR2 agonist IL-8, inhibited the forskolin-stimulated cAMP accumulation by 60% through the activation of G␣ i (Fig. 6C). As compared with the GST expression control, overexpression of either GST-RACK1 or RACK1-V5-His had no effect on the inhibition of cAMP accumulation by CXCR2. Similarly, overexpression of GST-RACK1 or RACK-V5-His did not affect the ␤ 2 -AR-stimulated increase in cAMP production (Fig.  6C). The inability of RACK1 to affect the ␤ 2 -AR-and CXCR2mediated cAMP signaling is not due to its expression level, because it was expressed at a 2-fold increase over the level of the endogenous RACK1. Taken together, these data indicate that RACK1 does not interfere with signaling through G␣ subunits. DISCUSSION RACK1 was first implicated in signal transduction when it was found to stabilize PKC in the active state and aid in its translocation to specific cellular fractions (15). Recent studies indicate that RACK1 also participates in the regulation of other signaling pathways, including interferon receptor-mediated activation of signal transducer and activator of transcription 1 (25), insulin-like growth factor-dependent signaling (26,27), and Src activation (28,29). Here we provide evidence for a new role of RACK1 in the regulation of G protein signaling mediated specifically by G␤␥.
Our data show that overexpression of RACK1 selectively inhibits both G i -coupled receptor-stimulated and direct G␤␥mediated PLC signaling but not that mediated by G␣ q or by receptors that act through G␣ q . Furthermore, RACK1 inhibits ␤␥-mediated ACII activation, but not cAMP signaling mediated by receptors acting through G␣ i or G␣ s . These findings indicate that RACK1 is a specific regulator of G␤␥ signaling. In line with the role of RACK1 in regulating G␤␥ signaling, RACK1 was found to bind only free G␤␥ subunits liberated from G protein activation, but not G␣ subunits nor heterotrimeric G proteins. In particular, G␤␥ can recruit RACK1 from the cytosol to the membrane where G␤␥-mediated signal transduction occurs. The ability of RACK1 to discriminate G␣ and G␤␥ subunits is consistent with our previous findings that RACK1  ). B, the density of phospho-p42/44 bands was determined by densitometric scanning. Data were expressed as -fold increase over the basal. C, HEK293 cells were transiently transfected with CXCR2 together with GST, GST-RACK1, or G␣ t cDNAs. Chemotaxis in response to IL-8 (1-250 ng/ml) stimulation was performed as described under "Experimental Procedures." binds the G␤ 1 ␥ 1 but not the G␣ t subunit in vitro (14). Although RACK1 was shown to bind heterotrimeric G t in vitro (14), it does not seem to interact with other classes of heterotrimeric G proteins in vivo. The reason for the differential binding of RACK1 to different classes of G proteins is unknown but may stem from the difference in the structure of G␣ subunits or the composition of G␥ subunits. However, our in vitro GST-pulldown studies did not reveal any difference in the binding of GST-RACK1 to either G␤ 1 ␥ 1 or G␤ 1 ␥ 2 (data not shown), suggesting that G␥ subunits do not play a major role in the interaction of G␤␥ with RACK1.
Because RACK1 is a scaffolding protein, theoretically it can recruit other proteins to indirectly modulate the function of G␤␥. However, our findings that RACK1 also inhibits G␤␥mediated effector activation in in vitro assays that consist of purified proteins without the presence of other RACK1 partners, indicate that the binding of RACK1 to G␤␥ itself is sufficient to modulate the interaction of G␤␥ with effectors. Because conformational changes on G␤␥ are not required for its activity, binding of RACK1 likely sterically hinders the access of effector proteins to G␤␥. However, we can not exclude the possibility that other RACK1-binding proteins may play a role in regulating G␤␥ function in vivo. Indeed, because PKC-mediated phosphorylation can regulate the activity of many G␤␥ effectors, such as Ca 2ϩ channels (30), PLC␤ (31,32), and AC (33), it is conceivable that RACK1 might regulate G␤␥ signaling via modulating the activity of PKC in cells. However, we found that the addition of PKC inhibitors immediately after transfection did not affect the ability of RACK1 to inhibit G␤␥-mediated PLC␤2 activation (data not shown), suggesting that inhibition of G␤␥ signaling by RACK1 does not require PKC activation.
Remarkably, RACK1 appears to regulate only a subset of G␤␥ functions. Thus it inhibits G␤␥-mediated PLC␤2 and ACII activation, but not MAPK signaling and chemotaxis of HEK293 cell via CXCR2. Although the effector proteins that are directly regulated by G␤␥ to mediate MAPK activation and chemotactic response of CXCR2 in HEK293 cells are not fully elucidated, both responses probably involve phosphoinositide 3-kinase ␥ (23,34). Previous studies indicate that the functional domains on G␤␥ for regulation of different effectors are overlapping but not identical (17). A simple explanation for the ability of RACK1 to discriminate different G␤␥ effectors may be due to the binding of RACK1 to a unique contact region of G␤␥ that is critical for activation of some but not other effectors. An alternative explanation is that RACK1 may have differential activities to different G␤␥ dimers, which likely co-exist in COS-7 and HEK293 cells and may regulate distinct effectors. However, we have evidence that RACK1 can interact with multiple isoforms of G␤␥ with similar binding affinities (data not shown). In addition, we found that overexpression of RACK1 inhibits the ␣ 2A -AR-mediated PLC␤ activation but has no effect on the activity of MAPK stimulated by the ␣ 2A -AR in COS-7 cells. The activation of both PLC␤ and MAPK by the ␣ 2A -AR is mediated through G␤␥ released from G i . Although specific G␤␥ pairs have been reported to couple to particular receptors and G␣ subunits, there is not evidence that the same receptor uses different G␤␥ dimers to stimulate particular effectors in the same cells (35,36). Thus, the ability of RACK1 to distinguish different effectors of G␤␥ is unlikely due to its differential interactions with different G␤␥ dimers. This characteristic of RACK1 is different from those of several recently described G␤␥ interacting proteins, including phosducin, PhLPs, and GRKs (8,10,37). These proteins interact with the G␣-binding regions of G␤ and inhibit all known aspects of G␤␥ functions.
The unique ability of RACK1 to differentially regulate different G␤␥ functions may have important implications for signaling specificity of G␤␥. G␤␥ becomes activated when it is released from activated heterotrimeric G proteins, and it has a long list of effectors, including many ion channels and enzymes, such as PLC, AC, phospholipase A 2 , phosphoinositide 3-kinase, Bruton and T cell-specific kinase tyrosine kinases, and select members of the GRK family (38). In addition, G␤␥ interacts with many other proteins, including dynamin-1 (39), ADPribosylation factor (40), calmodulin (41), phosducin (9), PhLPs (10), syntaxin 1A and SNAP25B (42), and pleckstrin homology domain-containing proteins (40). However, it is clear that not FIG. 6. Effect of RACK1 on IP and cAMP signaling pathways activated through G␣ q , G␣ s , and G␣ i . A, PI hydrolysis mediated by the ␣ 1B -AR co-expressed with GST or GST-RACK1 was determined in the absence (open bars, Basal) or presence of (Ϫ)-epinephrine (100 M) (filled bars, Epi), as described in Fig. 3 all of G␤␥ effects are activated under a given stimulus, although what constraining factors contribute to the signaling specificity of G␤␥ remains elusive. The ability of RACK1 to inhibit specific functions of G␤␥ thus may provide one mechanism to tune G␤␥ activation of effectors. In addition, because RACK1 undergoes translocation upon binding G␤␥, it may contribute to signaling specificity of G␤␥ by localizing particular G␤␥ effectors within signaling complexes. Studies are currently underway to address these hypotheses.