High Affinity Binding of β-Adrenergic Receptor Kinase to Microsomal Membranes

The β-adrenergic receptor kinase (βARK) modulates β-adrenergic and other G protein-coupled receptors by rapidly phosphorylating agonist-occupied receptors at the plasma membrane. We have recently shown that βARK also associates with intracellular microsomal membranes both “in vitro” and “in situ” (García-Higuera, I., Penela, P., Murga, C., Egea, G., Bonay, P., Benovic, J. L., and 5or, F., Jr. (1994) J. Biol. Chem. 269, 1348-1355), thus suggesting a complex modulation of the subcellular distribution of βARK. In this report, we used recombinant [35S]methionine-labeled βARK to show that this kinase interacts rapidly with a high affinity binding site (Kd of 20 ± 1 nM) present in salt-stripped rat liver microsomal membranes. Although βARK binding is not modulated by membrane preincubation with G protein activators, the activity of bound βARK toward rhodopsin or a synthetic peptide substrate was markedly enhanced upon stimulation of the endogenous heterotrimeric G proteins present in the microsomal membranes by AlF4− or mastoparan/guanosine 5′-(3-O-thio)triphosphate, thus strongly suggesting a functional link between these proteins and membrane-associated βARK. Interestingly, βARK association with microsomal membranes is not significantly affected by a fusion protein derived from the carboxyl terminus of βARK1 (the proposed location of the βγ subunit binding site), whereas it is markedly inhibited by fusion proteins corresponding to the amino-terminal region of the kinase. The main determinants of binding appear to be localized to an ∼60-amino acid residue stretch (residues 88 to 145). Our results further indicate a functional relationship between βARK and heterotrimeric G proteins in different intracellular organelles, and suggest that additional proteins may be involved in modulating the cellular localization of the kinase through a new targeting domain of βARK.

The ␤-adrenergic receptor kinase (␤ARK) modulates ␤-adrenergic and other G protein-coupled receptors by rapidly phosphorylating agonist-occupied receptors at the plasma membrane. We have recently shown that ␤ARK also associates with intracellular microsomal membranes both "in vitro" and "in situ" (García-Higuera, I., Penela, P., Murga, C., Egea, G., Bonay, P., Benovic, J. L., and Mayor, F., Jr. (1994) J. Biol. Chem. 269, 1348 -1355), thus suggesting a complex modulation of the subcellular distribution of ␤ARK. In this report, we used recombinant [ 35 S]methionine-labeled ␤ARK to show that this kinase interacts rapidly with a high affinity binding site (K d of 20 ؎ 1 nM) present in saltstripped rat liver microsomal membranes. Although ␤ARK binding is not modulated by membrane preincubation with G protein activators, the activity of bound ␤ARK toward rhodopsin or a synthetic peptide substrate was markedly enhanced upon stimulation of the endogenous heterotrimeric G proteins present in the microsomal membranes by AlF 4 ؊ or mastoparan/ guanosine 5-(3-O-thio)triphosphate, thus strongly suggesting a functional link between these proteins and membrane-associated ␤ARK. Interestingly, ␤ARK association with microsomal membranes is not significantly affected by a fusion protein derived from the carboxyl terminus of ␤ARK1 (the proposed location of the ␤␥ subunit binding site), whereas it is markedly inhibited by fusion proteins corresponding to the amino-terminal region of the kinase. The main determinants of binding appear to be localized to an ϳ60-amino acid residue stretch (residues 88 to 145). Our results further indicate a functional relationship between ␤ARK and heterotrimeric G proteins in different intracellular organelles, and suggest that additional proteins may be involved in modulating the cellular localization of the kinase through a new targeting domain of ␤ARK.
A general feature of G protein-coupled receptors is that exposure to agonists often leads to a rapid loss of receptor responsiveness, a process termed desensitization or tolerance. Such regulatory mechanisms are triggered every time a receptor is activated and play a key role in signal integration and plasticity at the cellular level. The ␤-adrenergic receptor (␤AR) 1coupled adenylyl cyclase system has provided an important model for the study of the molecular mechanisms of desensitization (Benovic et al., 1988;Dohlman et al., 1991;Kobilka, 1992;Lohse, 1993). Work from several laboratories has shown that rapid agonist-specific desensitization of the ␤AR is due to functional uncoupling from the transducing G protein, which is initiated by phosphorylation of the ␤AR by ␤-adrenergic receptor kinase (␤ARK), a serine-threonine kinase that specifically phosphorylates the agonist-occupied form of the receptor (Palczewski and Benovic, 1991;Lefkowitz, 1993;Haga et al., 1994a). ␤ARK-mediated phosphorylation allows the interaction with the ␤AR of an additional regulatory protein, ␤-arrestin, which precludes receptor interaction with G proteins and blocks signal transduction (Lohse et al., 1990;Kobilka, 1992). The uncoupling process is followed by transient receptor internalization away from the plasma membrane. Interestingly, emerging evidence indicates that ␤ARK may be able to modulate a variety of G protein-coupled receptors Kim et al., 1993b;Kwatra et al., 1993;Richardson et al., 1993). This, together with the recent characterization of different ␤ARK-related kinases, which constitute the G proteincoupled receptor kinase (GRK) multigene family, strongly suggests that very similar mechanisms of regulation operate for all G protein-coupled receptors Haga et al., 1994a;Loudon and Benovic, 1994;García-Higuera et al., 1994b).
However, there is little understanding of the protein-protein interactions of this complex regulatory network, and of the changes in the subcellular distribution of these proteins that take place upon receptor activation. ␤ARK has been described as a soluble, cytosolic enzyme that transiently translocates to the plasma membrane when the receptor is occupied by an agonist (Strasser et al., 1986;Mayor et al., 1987;García-Higuera and Mayor, 1992;Chuang et al., 1992). Recent data indicate that ␤␥ subunits of heterotrimeric G proteins are able to enhance ␤ARK activity toward different G protein-coupled receptors (Haga and Haga, 1992;Pitcher et al., 1992;Kam-eyama et al., 1993;Kim et al., 1993b), and both purified G proteins and isolated ␤␥ subunits can interact with ␤ARK in vitro (Pitcher et al., 1992;Kim et al., 1993b). Therefore, it has been proposed that the interaction of ␤␥ subunits with ␤ARK would help to target the kinase to the periphery of the plasma membrane and to increase its activity toward the activated receptor. The ␤␥-binding domain of ␤ARK has been localized to a 125-amino acid stretch in the COOH-terminal region of the enzyme (Kameyama et al., 1993;Koch et al., 1993); this region partially overlaps with a pleckstrin homology domain present in ␤ARK . Moreover, we have recently reported that, in addition to being a soluble enzyme that transiently translocates to the plasma membrane, a significant amount of ␤ARK is associated with internal microsomal membranes in a variety of tissues and cell lines (García-Higuera and García-Higuera et al., 1994a), thus raising new questions regarding the functional role of this kinase. Subcellular fractionation studies and indirect immunofluorescence and immunogold electron microscopy localization in cultured cells confirmed the association of ␤ARK with microsomal structures in situ (García-Higuera and García-Higuera et al., 1994a). In our previous report, cell-free association experiments indicated that ␤ARK peripherally associates with a protein component of the microsomal membranes by means of electrostatic interactions (García-Higuera et al., 1994a), in a way reminiscent of its transient interaction with the plasma membrane.
In this paper, we have used a direct binding assay with [ 35 S]methionine-labeled ␤ARK (Kim et al., 1993b) to further characterize the kinase binding site in the microsomes and investigated the functional consequences of such interaction. Our data support a functional link between ␤ARK and microsomal heterotrimeric G proteins and suggest the existence of a new targeting domain of ␤ARK involved in the modulation of the complex subcellular distribution of this important regulatory kinase.

EXPERIMENTAL PROCEDURES
Preparation of Rat Liver Subcellular Fractions-Subcellular fractionation of rat liver was performed essentially as described (García-Higuera and Mayor, 1994;García-Higuera et al., 1994a). Briefly, adult male Wistar rats were sacrificed by decapitation, and their livers were rapidly excised, weighed, minced and homogenized with a motorized Teflon pestle in 4 volumes of ice-cold buffer A (20 mM Tris-HCl, pH 7.4, 5 mM EDTA, 5 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, 20 g/ml benzamidine) plus 250 mM sucrose. All subsequent manipulations were performed at 4°C. The supernatant of a low speed centrifugation (250 ϫ g, 4 min) was centrifuged at 3,000 ϫ g for 10 min to obtain a plasma membrane pellet. After centrifugation of the supernatant at 10,000 ϫ g for 20 min to obtain the crude mitochondrial pellet, the microsomal membrane pellet was obtained by centrifuging the postmitochondrial supernatant for 60 min at 250,000 ϫ g in a Beckman TL-100 centrifuge. The plasma membrane and the microsomal membranes were washed once in buffer A, and the final pellets resuspended in buffer A supplemented with 200 mM NaCl and incubated for 30 min at 4°C in order to extract ␤ARK activity (García-Higuera and Mayor, 1992;García-Higuera et al., 1994a). After centrifugation as above, ␤ARK-stripped plasma or microsomal membrane aliquots were resuspended to a protein concentration of ϳ15 mg/ml in buffer A and stored at Ϫ70°C. Electron microscopy and enzymatic marker characterization of these subcellular fractions indicated that crude plasma membranes display a 6.7-fold increase in the activity of the specific marker 5Ј-nucleotidase with respect to the homogenate, whereas microsomal fractions are enriched in endoplasmic reticulum, lysosomal and Golgi apparatus enzymatic activities, with little contamination by plasma membrane (García-Higuera et al., 1994a;García-Higuera and Mayor, 1994).
Purification of ␤ARK and Radiolabeling with [ 35 S]Methionine-Bovine ␤ARK was overexpressed in Sf9 cells using the baculovirus expression system and purified from high-speed supernatant fractions of cell homogenates by sequential chromatography on SP-Sepharose and heparin-Sepharose columns exactly as described previously (Kim et al., 1993a). For radiolabeling, infected Sf9 cells were incubated for 20 h in methionine-free Graces media containing 1-2 mCi of [ 35 S]methionine (1190 Ci/mmol) and then harvested and purified exactly as previously reported (Kim et al., 1993b). Purity of radiolabeled ␤ARK was Ͼ95%, as judged by SDS-polyacrylamide gel electrophoresis. The radiolabeled ␤ARK was stored at Ϫ20°C and used at a specific activity of 350-1000 cpm/pmol.
Purification of G Protein ␤␥ Subunits-After purification of G i and G o from bovine brain membranes as described (Sternweis and Robishaw, 1984), ␤␥ subunits were isolated by chromatography on heptylamine-Sepharose in the presence of AMF (30 M AlCl 3 , 6 mM MgCl 2 , and 10 mM NaF) (Katada et al., 1984) followed by anion exchange chromatography on a Mono Q column to obtain subunits free of AMF and cholate (Kim et al., 1993b). The purified ␤␥ was stored in 20 mM Tris-HCl, pH 8.0, 1 mM EDTA, 1 mM dithiotreitol, 150 mM NaCl, and 0.05% Lubrol at Ϫ80°C.
Analysis of G Protein Subunits in Rat Liver Membrane Fractions-The presence of heterotrimeric G proteins in the plasma and microsomal membrane fractions used in our ␤ARK binding and activity experiments was investigated by [ 35 S]GTP␥S binding as previously reported (Northup et al., 1982). The presence of ␣ and ␤␥ subunits was also investigated by immunoblot analysis after resolving the membrane fractions (30 -70 g of protein) by electrophoresis on 11% SDS-polyacrylamide gels and electroblotting to nitrocellulose filters, using as probes the G protein ␣ subunit ␣ common antibody AS8 (1:300 dilution, a generous gift from Drs. G. Shü ltz and K. Spicher, Free University, Berlin), the ␤␥ subunit ␤ common antibody M-14 (5 g/ml, Santa Cruz Biotechnology) or the ␤ subunit antibody MS/1 (1:300, DuPont NEN) as described . Similar results were obtained with the different ␤ subunit antibodies. The blots were developed using a chemiluminescence method (ECL, Amersham Corp.). In the latter case, an approximate quantitation of the presence of ␤␥ subunits was performed by simultaneously probing known amounts (20 -40 ng) of purified bovine ␤␥ subunits (see above) resolved in the same gel, and densitometric analysis using a Molecular Dynamics laser densitometer.
Preparation of GST-␤ARK Fusion Proteins-Fusion proteins containing amino acids 50 -145 (FP1) or 437-689 (FP2) of bovine ␤ARK1 were prepared essentially as reported . Specific 5Ј and 3Ј polymerase chain reaction primers (containing BamHI and EcoRI restriction sites, respectively) were used to amplify the desired cDNA fragments. The amplified fragments were ligated in frame between the BamHI and EcoRI sites of the glutathione S-transferase (GST) gene fusion vector pGEX-2T (Pharmacia). Additional GST-␤ARK1 constructs comprising amino acids 1-63, 1-88, 1-122, or 1-147 were prepared in a similar way. Selected clones were verified by dideoxy sequencing using T7 DNA-Polymerase (Sequenase, Stratagene). The fusion protein constructs were used to transform the Escherichia coli strain AG1 induced with isopropyl-1-thio-␤-D-galactopyranoside, and proteins were purified by affinity chromatography on glutathione-Sepharose 4B (Pharmacia). The integrity of the purified fusion proteins (or GST alone) was analyzed by SDS-PAGE and Coomassie Blue staining.
Analysis of ␤␥ Binding to GST-␤ARK Fusion Proteins-Detection of possible interactions between purified G protein ␤␥ subunits and our fusion proteins was performed as described . Briefly, purified bovine brain ␤␥ subunits (150 nM) were incubated for 30 min at 4°C with the desired ␤ARK fusion proteins (or GST as a negative control) at a final concentration of 1 M in 50 l of 20 mM Tris-HCl, pH 7.4, 1 mM MgCl 2 , 0.01% Lubrol. Then, 20 l of a 50% slurry of glutathione-Sepharose 4B were added, and incubation continued for 90 min in ice with gentle mixing. The beads were washed three times with 1 ml of buffer containing 0.01% Lubrol, and retained proteins were analyzed by SDS-PAGE on 11% acrylamide gels, followed by transfer to nitrocellulose and Western blot analysis with antibodies against ␤ subunits as described above, using purified ␤␥ subunits run in the same gel as a positive control.
␤ARK Binding to Microsomal Membranes-35 S-Labeled ␤ARK (5-80 nM) was incubated in buffer A or in 20 mM Tris-HCl, pH 7.4, in reaction volumes of 100 -300 l (Kim et al., 1993b) with 0.4 -0.75 g/l saltstripped microsomal membranes (which do not contain any residual ␤ARK). In some experiments, microsomal membranes were preincubated for 15-30 min at 37°C in 20 mM Tris-HCl, pH 7.4, 2 mM MgCl 2 alone, or in the presence of different concentrations of MgCl 2 , cAMP, heparin, AlF 4 Ϫ , GDP, mastoparan, and GTP␥S (as indicated in the figure legends), or different concentrations (5-30 M) of bovine serum albumin, GST, or GST-␤ARK fusion proteins. The binding mixtures were incubated at 37°C for 0.5-15 min, and membrane-associated ␤ARK was collected by centrifugation at 250,000 ϫ g for 10 -30 min at 4°C in a Beckman TL-100 centrifuge. The microsomal membrane pel-lets were carefully washed with buffer and then solubilized with 1% SDS and counted with scintillation fluid. In some experiments, the radioactivity of the supernatant containing unbound ␤ARK was also quantitated. Nonspecific binding was defined by performing the binding reaction in the presence of 200 mM NaCl, an experimental condition in which ␤ARK association with a protein component of the microsomal membrane has been shown to be completely inhibited (García-Higuera et al., 1994a). 35 S-Labeled ␤ARK binding can also be competed by unlabeled recombinant ␤ARK (data not shown). In other experiments, unlabeled recombinant ␤ARK was used as the ligand, and bound ␤ARK was determined as described by immunoblot analysis with specific antibodies raised against ␤ARK (García-Higuera and García-Higuera et al., 1994a). Alternatively, the extent of ␤ARK binding under different experimental conditions was assessed by quantitating the kinase activity remaining in the soluble fraction after highspeed centrifugation by using the rhodopsin phosphorylation assay as described (García-Higuera et al. (1994a), also see below).
Effect of Na 2 CO 3 Treatment on ␤ARK Association with Microsomal Membranes-Aliquots of stripped microsomal membranes were resuspended in equal volumes of 100 mM Na 2 CO 3 , pH 11, or buffer A and incubated for 30 min at 4°C in order to extract peripheral membrane proteins, as described (Fujiki et al., 1982). The amount of ␤ subunits of heterotrimeric G proteins remaining in the membranes or released by the treatment was determined by immunoblot analysis as described above. Control and treated membranes were subsequently washed in buffer A and its ability to interact with ␤ARK was assessed as described (García-Higuera et al., 1994a). The presence of ␤ARK in the soluble and particulate fractions was determined by the rhodopsin phosphorylation assay as reported  and García-Higuera et al. (1994a), also see below), by using the direct 35 S-labeled ␤ARK binding assay as detailed above, or by utilizing recombinant unlabeled ␤ARK followed by immunoblot analysis as described (García-Higuera et al., 1994a).
Determination of ␤ARK Activity-␤ARK activity was assessed by using either a synthetic peptide substrate or purified urea-treated rod outer segments, as described previously in our laboratories Mayor et al., 1987;Kim et al., 1993b). Rhodopsin kinase-free purified rod outer segments consisting on Ͼ90% rhodopsin as indicated by Coomassie Blue staining of polyacrylamide gels were prepared in the dark as reported , and aliquots containing 120 pmol of rhodopsin were incubated for 20 -30 min at 30°C in a buffer containing 27 mM Tris, pH 7.5, 1.4 mM EDTA, 1 mM EGTA, 5.5 mM MgCl 2 , 4.5 mM NaF, 57 M [␥-32 P]ATP (2-3 cpm/fmol) plus 20 nM recombinant bovine ␤ARK in the presence or absence of 17-70 g of microsomal or plasma membrane proteins (final volume 60 l). Before testing its effect on ␤ARK activity toward rhodopsin, the desired amount of stripped microsomal or plasma membranes were preincubated (10 min at 37°C in 20 mM Tris-HCl, pH 7.5, 1 mM MgCl 2 ) alone or in the presence of the heterotrimeric G protein activators AlF 4 Ϫ (5 mM NaF, 50 M AlCl 3 , 100 M GDP) or mastoparan (100 M) plus GTP␥S (25-50 M). Reactions were started by exposure to fluorescent laboratory lighting and stopped by 20-fold dilution with ice-cold 20 mM Tris-HCl followed by centrifugation. The resultant rhodopsin-containing pellets were resuspended in SDS-PAGE sample buffer and resolved by electrophoresis on 12% polyacrylamide gels followed by autoradiography. ␤ARK activity was quantitated by measuring the radioactivity associated with the excised rhodopsin band in the dried gel by Cerenkov spectroscopy . ␤ARK activity was also investigated using a synthetic peptide as substrate. The peptide RRREEEEESAAA was synthesized on a Applied Biosystems 431A synthesizer using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. The peptide (1 mM) was incubated for 30 min with ␤ARK (20 nM) in the absence or presence of 0.6 g/l stripped microsomal or plasma membrane proteins, in a buffer containing 20 mM Tris-HCl, pH 7.5, 2 mM EDTA, 7.5 mM MgCl 2 , 100 M [␥-32 P]ATP (0.5-1 cpm/fmol) in a final volume of 55 l at 30°C. Before starting the experiment, the stripped microsomal or plasma membranes were preincubated (15 min at 37°C in 20 mM Tris-HCl, pH 7.5, 1 mM MgCl 2 ) alone or in the presence of the heterotrimeric G protein activators AlF 4 Ϫ (5 mM NaF, 50 M AlCl 3 , 100 M GDP) or mastoparan (20 -100 M) plus GTP␥S (50 M). Reactions were stopped by addition of 30 l of 30% trichloroacetic acid. The quenched reactions were centrifuged at 48,000 ϫ g for 10 min and the resulting supernatants were transferred to a 2 ϫ 2-cm square of P-81 paper followed by washes in 75 mM phosphoric acid and radioactivity quantitation as described (Kim et al., 1993b).

RESULTS
Previous studies from our laboratory have shown that ␤ARK is associated with microsomal membranes in rat liver and many other tissues and cell lines (García-Higuera et al., 1994a). The kinase can be completely extracted from the membranes by mild salt treatment (200 mM NaCl), thus suggesting a peripheral association based on electrostatic interactions. Moreover, we reported (by using activity measurements and immunoblot analysis), that recombinant purified ␤ARK1 was able to interact with a protein component of salt-stripped rat liver microsomal membranes in a cell-free system (García-Higuera et al., 1994a). In these experiments, membrane-associated and unbound ␤ARK were separated by centrifugation at 250,000 ϫ g followed by kinase quantitation in the pellet or the supernatant. In order to gain further insight into the mechanisms of association of ␤ARK with microsomal membranes, we have utilized a direct binding assay using purified recombinant [ 35 S]methionine-labeled ␤ARK1, which has recently been used for investigating the interaction of ␤ARK with phospholipid vesicles containing purified G protein subunits (Kim et al., 1993b). In the assays shown below, 35 S-labeled ␤ARK was incubated under different experimental conditions with stripped rat liver microsomal membranes, and the amount of membrane-associated ␤ARK was then determined by measuring the radioactivity in the high-speed pellets. Nonspecific binding of ␤ARK was estimated by performing parallel experiments in the presence of 200 mM NaCl, a condition known to completely inhibit the kinase binding to the microsomal vesicles (García-Higuera et al., 1994a). This has enabled a quantitative and detailed characterization of the interaction of ␤ARK with this physiological membrane preparation. Fig. 1A shows that ␤ARK binding to stripped microsomal membranes is very rapid (half-maximal binding achieved in less than 1 min) and results in the association of most (ϳ75%) of the kinase present in the assay. Scatchard analysis (Fig. 1B) indicates that the interaction process is saturable and appears to involve one population of high affinity binding sites (K d of 20 Ϯ 1 nM). Similar results are obtained using lower concentrations of membranes. Interestingly, such kinetic and affinity data are similar to those reported for the interaction of ␤ARK1 with purified heterotrimeric G proteins or ␤␥ subunits (58 Ϯ 14 and 32 Ϯ 5 nM, respectively) (Kim et al., 1993b). In order to Labeled ␤ARK (20 nM, 500 cpm/pmol) was incubated in buffer A with stripped microsomal membranes (0.56 g of protein/l) at 37°C for the times indicated and membrane-associated ␤ARK was determined after high-speed centrifugation as detailed under "Experimental Procedures." Specific binding was the difference between counts in the absence and presence of 200 mM NaCl. Data points are the mean Ϯ S.E. of two independent experiments done in duplicate. B, representative Scatchard analysis of ␤ARK binding. Labeled ␤ARK (5-80 nM, 350 cpm/pmol) was incubated in buffer A with stripped microsomal membranes (0.4 g of protein/l) for 15 min at 37°C and specific binding determined as in panel A. K d and B max values are the mean Ϯ S.E. of three independent experiments performed in duplicate. A representative saturation curve is shown in the inset (E, total binding; q, nonspecific binding; छ, specific binding).
better understand ␤ARK association with microsomal membranes, we tried to obtain further information on the type of proteins(s) responsible for the interaction, on the effect of binding on ␤ARK activity, and on possible modulators of ␤ARK association and function.
Since ␤ARK binding to G protein subunits has been shown in vitro (Pitcher et al., 1992;Koch et al., 1993;Kim et al., 1993b), and the presence of heterotrimeric G proteins in intracellular organelles is mentioned in an increasing number of reports (see references in Balch (1992), García-Higuera et al. (1994a), and Neubig (1994)), we investigated the presence of endogenous G proteins in the salt-stripped microsomal membranes used in our studies. Specific ␣ common and ␤ common subunit antibodies detect major bands of ϳ40and 35-kDa, respectively (Fig. 2), thus confirming the presence of heterotrimeric G proteins in this preparation. G proteins can also be detected in the microsomal membranes by a 35 S-labeled GTP␥S binding assay (data not shown). The identity of the subunit isoforms has not been investigated in detail, although additional protein bands in the 40 -46 kDa range can be detected with the ␣ common antibody at longer times of exposure. With regard to ␤␥ subunits, comparison of the signal obtained in microsomal membranes with that of a known amount of ␤␥ subunits purified from bovine brain (Fig. 2, lane 1) allows an approximate quantitation of at least 10 pmol of microsomal ␤␥ per mg of stripped membrane protein, in the same order of magnitude as the B max detected for ␤ARK binding (see Fig. 1B).
We next tested the ability of different compounds to modulate ␤ARK association with microsomal membranes using the direct binding assay. Particularly, we investigated whether ␤ARK interaction would be modulated by known G protein activators. In initial experiments, preincubation of stripped microsomal membranes with the selective heterotrimeric G protein activator AlF 4 Ϫ (Finazzi et al., 1994) in the presence of excess Mg 2ϩ and GDP promoted a marked increase in subsequent ␤ARK binding (Fig. 3A). However, this effect seems to be due only to the presence of Mg 2ϩ and not to G protein activation, since a similar increase in ␤ARK interaction is observed when only Mg 2ϩ is added during the preincubation (Fig. 3A); GDP alone has no effect on ␤ARK binding (data not shown). A more detailed investigation on the effect of Mg 2ϩ was performed. Fig. 3B shows a dose-dependent effect of submillimolar concentrations of Mg 2ϩ on ␤ARK association with the microso-mal membranes, which reaches an ϳ4-fold increase over control binding at ϳ1 mM Mg 2ϩ . Such fold-increase is similar to that observed in Fig. 3A at 2.5 mM Mg 2ϩ , thus suggesting that the Mg 2ϩ effect attains a maximum in such concentration range. The presence of Mg 2ϩ promotes an increase in the apparent B max (65 pmol/mg of protein with 1 mM Mg 2ϩ ) without changing the affinity of ␤ARK for its binding sites. The divalent cation Mn 2ϩ can substitute for Mg 2ϩ (70% of Mg 2ϩ effect at 1 mM Mn 2ϩ ). In order to optimize ␤ARK binding and to approach physiological intracellular concentrations, 1 mM MgCl 2 was routinely included in all subsequent experiments.
The fact that ␤ARK binding to microsomal membranes is not modulated by known G protein activators is further confirmed by the lack of effect on ␤ARK association of mastoparan, a tetradecapeptide which stimulates guanine nucleotide exchange and activates heterotrimeric G proteins (Higashijima et al., 1988;Colombo et al., 1992) or GTP␥S, a non-hydrolyzable GTP analog which is a general activator of G proteins (Colombo et al., 1992;Tsai et al., 1992) (data not shown). The presence of other nucleotides such as GDP, ATP, GTP, or App(NH)p (0.5-1 mM) did not have any effect on ␤ARK binding, and did not promote the release of previously bound ␤ARK from the microsomal membranes (data not shown). Other possible modulatory compounds, such as cAMP or heparin, a ␤ARK inhibitor (Benovic et al., 1989a), do not significantly modulate ␤ARK interaction. The association of the kinase with the stripped microsomal membranes is reduced by ϳ50% when tested in the presence of 100 mM potassium gluconate, an ionic condition similar to the intracellular medium (not shown, see also García-Higuera et al. (1994a)). In summary, our data indicate that despite its presence in microsomal membranes and the previously reported interaction with ␤ARK in vitro (Pitcher et al., 1992) G protein activation was not required and did not modulate ␤ARK binding in our model.
We have previously shown (García-Higuera et al., 1994a) that ␤ARK association to microsomal membranes is reversible, i.e. that previously extracted or recombinant ␤ARK1 can interact with salt-stripped microsomes, thus indicating that the microsomal component involved in the interaction is not removed during the mild stripping procedure. Fig. 4A shows that when salt-stripped membranes are further treated with 0.1 M Na 2 CO 3 at pH 11, a commonly used method for removing FIG. 2. Presence of ␣ and ␤␥ subunits of heterotrimeric G proteins in rat liver stripped microsomal membranes. Stripped microsomal membranes (60 g of protein, lanes 1 and 3) or 20 ng of purified ␤␥ subunits obtained from bovine brain (lane 2) were resolved by 11% SDS-PAGE, blotted to nitrocellulose membranes, and probed with antibodies AS8 (␣ common subunit antibody, 1/300, lane 3) or M14 (␤ subunit antibody, 5 g/ml, lanes 1 and 2). Results are representative of three experiments. strongly-attached peripheral membrane proteins (Fujiki et al., 1982;Yu et al., 1992), ␤ARK association (as assessed by an activity assay) is not affected, thus suggesting that ␤ARK interacts with a microsomal component which behaves as an integral membrane protein. A more detailed and quantitative analysis of the interaction of ␤ARK with Na 2 CO 3 -stripped membranes was performed. Immunoblot analysis indicates that the interaction of recombinant ␤ARK with a defined amount of salt-stripped microsomal membranes (Fig. 4B, lane  1) is not altered after removal of additional peripheral proteins with the Na 2 CO 3 pH 11 treatment (lane 2); the association with the Na 2 CO 3 -stripped membranes is blocked in the presence of 200 mM NaCl (lane 3). Direct binding studies using 50 nM 35 S-labeled ␤ARK show that a given amount of microsomal membranes retains 86 Ϯ 9% (mean Ϯ S.D. of two independent experiments performed in triplicate) of specific kinase binding after the Na 2 CO 3 pH 11 treatment, thus leading to an ϳ1.5-fold increase in binding specific activity (up to 82 Ϯ 5 pmol/mg protein). Similar results were obtained in other sets of assays using 25 nM labeled ␤ARK. It is worth noting that the Na 2 CO 3 treatment leads to the loss of ϳ50% of the G protein ␤ subunits from the membranes, as assessed by immunoblot analysis (Fig.  4C). The fact that such ϳ50% decrease in G protein subunits does not alter ␤ARK association suggests that G proteins may not be the main anchor for ␤ARK in the microsomal membranes (also see below).
We next investigated the functional consequences of ␤ARK binding to the microsomal membranes by comparing the activity of the free and bound kinase toward exogenous substrates. Fig. 5 shows that, compared to free ␤ARK (lane 1), a marked inhibition of light-dependent rhodopsin phosphorylation is observed in the presence of increasing amounts of microsomal membranes (lanes 3-5), which do not show any residual ␤ARK activity (lane 2). This result indicates that bound ␤ARK is less able to interact with rhodopsin and/or to phosphorylate this membrane-bound specific substrate. It is worth noting that the same effect is observed when a synthetic peptide substrate is used to test the activity of ␤ARK (Fig. 7A), thus showing that the observed effect is not restricted to the rhodopsin phosphorylation assay. Interestingly, it has been reported that rhodopsin phosphorylation by the homologous, if not identical, muscarinic acetylcholine receptor kinase was strongly inhibited by heterotrimeric G proteins in the absence of guanine nucleotides in a dose-dependent way (Haga and Haga, 1992).
Since purified ␤␥ subunits or stimulators of G protein activation and dissociation such as GTP␥S have been shown to stimulate ␤ARK activity toward different substrates (Haga and Haga, 1992;Pitcher et al., 1992: Mü ller et al., 1993Pei et al., 1994), we next investigated whether preincubation of stripped microsomal membranes with specific agents that would activate endogenous heterotrimeric G proteins may modulate the activity of bound ␤ARK. Fig. 6A shows that the marked inhibition of ␤ARK activity observed in the presence of microsomal membranes (compare lanes 1 and 3) is relieved when the mi- FIG. 4. Influence of microsomal membrane pretreatment with Na 2 CO 3 on ␤ARK association and on the presence of G protein ␤ subunits. A, study of ␤ARK association by using an activity assay. Stripped microsomal membranes were resuspended in buffer A or in 100 mM Na 2 CO 3 , pH 11, and incubated for 30 min at 4°C in order to extract remaining peripheral proteins. After one wash in buffer A, control and treated membranes were incubated for 10 min at 37°C with a similar amount of a soluble extract containing ␤ARK activity exactly as described (García-Higuera, et al., 1994). Soluble and membraneassociated ␤ARK were separated by centrifugation and ␤ARK activity in the supernatant (S) and microsomal membrane (M) fractions was determined by using a rhodopsin phosphorylation assay as detailed under "Experimental Procedures." The arrow indicates the position of phosphorylated rhodopsin. B, immunoblot analysis of ␤ARK association. Aliquots of salt-stripped microsomal membranes were treated in the absence (lane 1) or presence (lanes 2 and 3) of Na 2 CO 3 pH 11 as indicated under "Experimental Procedures" and extracted proteins separated by high-speed centrifugation. The same volume of membrane pellets (corresponding to 50 g of protein in the initial, non Na 2 CO 3treated membranes) were resuspended in 20 mM Tris-HCl, pH 7.5, 2 mM MgCl 2 and incubated for 5 min at 37°C with 30 nM recombinant ␤ARK in the absence (lanes 1 and 2) or presence (lane 3) of 200 mM NaCl. After centrifugation at 250,000 ϫ g for 10 min, the amount of bound ␤ARK in each sample was assessed by immunoblot analysis as detailed under "Experimental Procedures." 80 ng of recombinant ␤ARK were directly resolved and analyzed in the same gel as a control (lane 4). Results are representative of two independent experiments. C, presence of G protein ␤ subunits in microsomal membranes subjected to different stripping procedures. Salt-stripped purified microsomal membranes (800 g of protein) were treated in the absence (lanes 2 and 3) or presence (lanes 4 and 5) of Na 2 CO 3 pH 11 as indicated under "Experimental Procedures" in a final volume of 800 l. After high-speed centrifugation to separate membranes and extracted proteins, equivalent aliquots of the membrane pellets (60 g of proteins, lanes 2 and 4) and extracted supernatants (60 l, lanes 3 and 5) were resolved by 11% SDS-PAGE, blotted to nitrocellulose membranes, and probed with a ␤ subunit antibody (M14, Santa Cruz Biotechnology, 5 g/ml) as described under "Experimental Procedures." 60 g of unstripped microsomal membranes (lane 1) and 20 ng of purified ␤␥ subunits obtained from bovine brain (lane 6) were resolved and blotted in the same gel and probed as above. crosomes were previously preincubated with the G protein activators AlF 4 Ϫ (lane 4) or mastoparan plus GTP␥S (lane 5). This effect cannot be simply ascribed to a release of bound ␤ARK, since such compounds do not show any effect on ␤ARK binding to the microsomal membranes ( Fig. 3 and data not shown), and the same result is obtained when the microsomal membranes are pelleted and then resuspended in phosphorylation buffer in the presence of rhodopsin (data not shown), thus indicating a direct effect on bound ␤ARK as a consequence of G protein activation and subunit dissociation. The same effect can be observed when ␤ARK functionality is assessed using a synthetic peptide substrate (Fig. 7A). It is interesting to note that similar results were apparent when we used stripped plasma membranes instead of microsomal membranes for these experiments: bound ␤ARK activity toward either rhodopsin or a peptide substrate was marked inhibited and was only relieved in the presence of heterotrimeric G protein activators (Figs. 6B and 7B). These data suggest a general feature of ␤ARK interaction with cellular membranes and indicate, for the first time using physiological membrane preparations, that the activity of ␤ARK is regulated by endogenous G proteins in different intracellular locations.
The fact that G protein activation can modulate the activity of bound ␤ARK clearly demonstrates a functional link between these proteins in the microsomal membranes. Such a functional link could be the consequence of a direct association of ␤ARK to G proteins under basal conditions, which would result in kinase activation in the presence of specific G protein stimulators. Alternatively, or in addition, ␤ARK could be binding to a different protein in the stripped microsomes, but its activity may be modulated by additional interactions with G protein ␤␥ subunits released upon G protein activation.
Since attempts to directly identify the microsomal protein(s) involved in ␤ARK interaction using different cross-linkers were unsuccessful, we tried another approach based on investigating the ability of fusion proteins containing different ␤ARK domains to displace 35 S-labeled kinase association with microsomal membranes. It has been described that the ␤␥ binding site domain of ␤ARK is localized in the COOH-terminal portion of the kinase Kameyama et al., 1993) and GST fusion proteins containing the last 222 amino acids of ␤ARK have been recently used to characterize ␤ARK interactions with ␤␥ subunits (Pitcher et al., 1992;Inglese et al., 1994;Touhara et al., 1994). Therefore, we prepared a similar fusion protein construct (FP2) containing amino acids 437-689 of bovine ␤ARK1 and an additional construct containing amino acids 50 -145 of the kinase (FP1) (see Fig. 8A), which contains a highly charged region that could be involved in electrostatic interactions (Benovic et al., 1989b). The ␤ARK fusion proteins were expressed in E. coli and purified by affinity chromatography on gluthatione-Sepharose columns; purified GST alone was also prepared as a negative control. As expected, FP2 (encompassing the COOH-terminal portion of ␤ARK), was able to associate with purified G protein ␤␥ subunits isolated from bovine brain, whereas GST alone or FP1 did not associate or showed a very weak interaction (Fig. 8B). Consistently, FP2 blocked ␤␥ subunit activation of rhodopsin phosphorylation by ␤ARK, whereas FP1 had no an inhibitory effect (Fig. 8C). We next investigated whether these proteins could inhibit ␤ARK association with stripped microsomal membranes by meas-uring the binding of 35 S-labeled kinase to microsomes preincubated under control conditions or in the presence of different concentrations of GST or GST-␤ARK fusion proteins (Fig. 8D). Interestingly, GST-FP2 showed only a very slight effect on ␤ARK binding over a range of concentrations that have been shown to strongly inhibit ␤ARK association to purified G protein ␤␥ subunits in vitro or to block ␤␥ subunit modulation of type II adenylyl cyclase in reconstituted or permeabilized systems (IC 50 in the range 10 -20 M) (Pitcher et al., 1992;Touhara et al., 1994;Inglese et al., 1994). Similar results were obtained in the presence of G protein activators (data not shown). On the contrary, the fusion protein GST-FP1, containing an amino-terminal region of the kinase, markedly decreased labeled ␤ARK interaction with the microsomal membranes, with a potency (ϳ60% of binding inhibition at 15 M) similar to that of the COOH terminus on ␤␥-related functions. Interestingly, similar results were obtained when the effects of these fusion proteins on 35 S-labeled ␤ARK binding were tested in Na 2 CO 3 -stripped membranes, which are partially devoid of G proteins (Fig. 4C and Kehlenbach et al. (1994)) but retain the ability to interact with the kinase (60 Ϯ 5% and 106 Ϯ 8% of control binding of 40 nM 35 S-labeled ␤ARK in the presence of 20 M GST-FP1 and GST-FP2, respectively). Taken together, our data indicate that the mechanism of ␤ARK association with microsomal membranes is different from that of its interaction with G protein ␤␥ subunits, and suggest that ␤ARK is primarily interacting with an additional protein component of the membranes.
In order to confirm these results and start to delineate the region of ␤ARK involved in the association with microsomal membranes, we tested the effects of additional fusion proteins. In these experiments, microsomal membranes were incubated with recombinant ␤ARK in the presence of different GST-␤ARK constructs (Fig. 9A) and either GST alone or 200 mM NaCl as controls. Membrane-bound and soluble ␤ARK were separated by high-speed centrifugation, and the ␤ARK activity remaining in the supernatant was quantitated. Control experiments were performed to show that remaining fusion proteins per se did not have any effect on the phosphorylation assay (data not shown). Fig. 9B show that, as expected, NaCl strongly inhibits ␤ARK binding to the membranes (i.e. high ␤ARK activity present in the supernatant), whereas binding is high in the presence of GST (i.e. low ␤ARK activity remaining in the supernatant), which does not inhibit ␤ARK association (see Fig. 8D). The presence of GST-␤ARK fusion proteins comprising amino acids 1-147 or 50 -145 (FP1) clearly inhibits ␤ARK binding to microsomal membranes, whereas a construct encompassing amino acids 1-88 does not, thus suggesting that the FIG. 8. Effect of GST-␤ARK fusion proteins on 35 S-labeled ␤ARK binding to microsomal membranes. A, domain structure of bovine ␤ARK1 indicating the regions from which fusion protein containing amino acids 50 -145 (FP1) or 437-689 (FP2) were derived; the proposed location of the pleckstrin homology domain (PH) and the ␤␥ subunits binding region are also indicated. B, ␤␥ binding properties of GST-␤ARK fusion proteins. Purified bovine brain ␤␥ subunits were incubated with identical concentrations of GST, FP1, and FP2 and protein complexes pelleted using glutathione-Sepharose beads as indicated under "Experimental Procedures." Bound proteins were resolved by SDS-PAGE, transferred to nitrocellulose, and detected with a specific antibody; purified brain ␤␥ subunits (G␤␥) were used as a positive control. C, effect of GST-␤ARK fusion proteins FP1 and FP2 on rhodopsin phosphorylation by ␤ARK in the presence of G protein ␤␥ subunits. The rhodopsin phosphorylation assay was performed as described under "Experimental Procedures"; after incubation for 40 min at 30°C in the presence of 20 nM recombinant ␤ARK and 80 nM ␤␥ subunits alone or in the presence of 10 M GST or the indicated GST-␤ARK fusion proteins the reaction was stopped by addition of SDS-PAGE buffer, and phosphorylated rhodopsin was resolved by 12% SDS-PAGE followed by autoradiography. D, dose-dependent inhibition of 35 S-labeled ␤ARK binding to microsomal membranes by GST-␤ARK fusion proteins. Stripped microsomal membranes (0.75 g of protein/l) were preincubated for 30 min at 37°C in 20 mM Tris-HCl, pH 7.5, 2 mM MgCl 2 with the indicated concentrations of GST or the fusion proteins FP1 and FP2 or an equivalent amount of bovine serum albumin in the same vehicle (control conditions). Labeled ␤ARK was then added at a final concentration of 15 nM. After 5 min at 37°C ␤ARK specific binding was determined as detailed in Fig. 1 and under "Experimental Procedures." Data are expressed as percentage of the binding detected in control conditions, and are means Ϯ S.E. of 3-4 independent experiments performed in duplicate.
FIG. 9. Analysis of the effect of amino-terminal GST-␤ARK fusion proteins on ␤ARK association with microsomal membranes. A, schematic representation of the GST-␤ARK constructs used in the experiments. B, effect of GST-␤ARK fusion proteins on ␤ARK binding to microsomal membranes. Stripped microsomal membranes (0.75 g of protein/l) were preincubated for 30 min at 37°C as detailed in Fig. 8D with the indicated fusion proteins (15 M) or NaCl 200 mM, and recombinant ␤ARK was then added at a final concentration of 10 nM. After 5 min at 37°C, bound and soluble ␤ARK were separated by high-speed centrifugation as detailed under "Experimental Procedures," and unbound ␤ARK was quantitated in 5 l of the supernatants using the rhodopsin phosphorylation assay. Results are representative of five independent experiments. main determinants of ␤ARK association are located in the region 88 -145. Taken together, these results indicate that the interaction of this regulatory kinase with the microsomes preferentially involves a domain located within the amino-terminal portion of ␤ARK, and suggest that G protein ␤␥ subunits are not the main anchor for ␤ARK in the microsomal membranes. DISCUSSION An early step in the regulation of ␤AR is the phosphorylation of the agonist-occupied receptor by ␤ARK, which was initially described as a soluble enzyme that transiently translocates to the plasma membrane upon receptor activation (Lefkowitz, 1993;Inglese et al., 1994;Haga et al., 1994a). Recent data have shown that ␤ARK can interact with purified ␤␥ subunits or whole heterotrimeric G proteins in vitro (Pitcher et al., 1992;Kim et al., 1993b), thus suggesting possible anchors for the kinase in the plasma membrane. In addition, we have recently reported that ␤ARK associates with internal microsomal membranes both in vitro and in situ (García-Higuera et al., 1994a). These data indicate that several ␤ARK pools (microsomebound, plasma membrane-bound, and cytosolic) exist inside the cell and suggest that complex mechanisms (probably involving specific interactions with different proteins) may operate in order to regulate the subcellular distribution and activation of ␤ARK. In this report, we have further characterized the association of ␤ARK with microsomal membranes using a direct 35 S-labeled ␤ARK binding assay, and investigated the modulation of the functionality of the microsome-bound kinase.
Our data indicate that ␤ARK binds very rapidly to stripped microsomal membranes with a K d of ϳ20 nM. The kinase appears to associate with only one population of high affinity binding sites, although the possibility of more than one population of sites with similar binding affinities cannot be completely ruled out. Such binding sites are suggested to be protein components of the microsomal membrane, since previous data from our laboratory showed that protease or heat pretreatment of the microsomes strongly inhibits ␤ARK association (García-Higuera et al., 1994a). Furthermore, the lack of effect on ␤ARK binding of Na 2 CO 3 pH 11 treatment shown in the present report indicates that the kinase binding sites behave as integral membrane proteins. It is worth noting that the affinity of ␤ARK interaction with microsomal membranes is higher than that reported for its association with agonist-activated ␤ 2 AR (K d Ͼ 100 nM) and similar to that obtained with purified G proteins (K d ϳ 58 nM) or isolated ␤␥ subunits (K d ϳ 32 nM) (Kim et al., 1993b), consistent with a functionally relevant role. The present study provides the first report of a high affinity interaction of ␤ARK with a physiological membrane preparation. Association with microsomal membranes is not unique for ␤ARK since a number of other proteins (pp60 src , PKC-, ADPribosylation factors, kinesin, and dynein) have been reported to interact in different ways with such types of preparations (Resh, 1989;Chida et al., 1994;Tsai et al., 1992;Yu et al., 1992;Thissen and Casey, 1993).
Given the previously reported interaction of ␤ARK with purified G proteins and ␤␥ subunits and the stimulation of ␤ARK activity by free ␤␥ subunits (see above), these proteins were obvious candidates for participating in ␤ARK binding and for modulating the activity of microsomal ␤ARK, since their presence in intracellular membranes has been increasingly appreciated (see references in García-Higuera et al. (1994a), Neubig (1994, and Nuoffer and Balch, 1994). We confirmed by [ 35 S]GTP␥S binding and immunoblot analysis the presence of heterotrimeric G proteins in our preparations, and its preliminary quantitative analysis was compatible with a role in ␤ARK association, taking into account the apparent B max of ␤ARK binding. However, our results clearly show that the interaction of ␤ARK with the microsomal membrane is not dependent on heterotrimeric G protein activation, since no significant changes can be observed under experimental conditions routinely used for G protein stimulation (Colombo et al., 1992;Tsai et al., 1992;De Almeida et al., 1993;Kim et al., 1993b). The only apparent modulators of ␤ARK binding were low millimolar concentrations of divalent cations such as Mg 2ϩ . Since Mg 2ϩ can modulate a variety of enzymatic reactions (including phosphorylation/dephosphorylation processes, although the involvement of proteins kinases as mediators of its effect is unlikely in our experimental conditions), the mechanisms of the Mg 2ϩ effect on ␤ARK interaction and its possible functional relevance remain to be established.
Our studies on the modulation of the activity of microsomebound ␤ARK put forward two interesting findings: (i) the bound kinase is in an inactive state, as shown by its reduced ability to phosphorylate rhodopsin or a synthetic peptide substrate; and (ii) although ␤ARK binding is not affected by stimulators of heterotrimeric G proteins such as mastoparan/ GTP␥S or AlF 4 Ϫ , these agents promote a marked increase in the activity of bound kinase, indicating a functional link between endogenous microsomal G proteins and ␤ARK.
It has been previously shown that the phosphorylation of rhodopsin or the muscarinic acetylcholine receptor by rhodopsin kinase or ␤ARK-like kinases was strongly inhibited by phospholipid vesicles containing purified, non-activated heterotrimeric G proteins (Kelleher and Johnson, 1988;Haga, 1990, 1992), and that this effect was relieved in the presence of GTP␥S (i.e. subunit dissociation) (Haga and Haga, 1992). It has been suggested that such an inhibitory effect is due to a competition between trimeric G proteins and receptor kinases for overlapping sites on the activated receptor. However, the fact that we observed inhibition of ␤ARK activity toward both an activated receptor (rhodopsin) and a synthetic peptide substrate suggest that, in addition, the interaction of ␤ARK with the membranes temporarily masks or alters functionally relevant domains of the kinase. The activation of microsomal heterotrimeric G proteins would favor the interaction of free ␤␥ subunits with ␤ARK, which will then adopt an active conformation. Interestingly, the same pattern of modulation of ␤ARK activity can be detected when stripped plasma membranes were used instead of microsomal membranes, thus suggesting a general feature of ␤ARK interaction with cellular membranes. Our results confirm, for the first time using physiological membrane preparations, that ␤␥ subunits activate ␤ARK-mediated phosphorylation Haga, 1990, 1992;Pitcher et al., 1992;Kameyama et al., 1993;Kim et al., 1993b). Furthermore, we show that ␤ARK activity can be modulated by endogenous G proteins in different intracellular locations. The fact that G protein stimulation also increases ␤ARK activity toward a soluble peptide substrate is in line with recent in vitro results showing that phosphorylation of soluble substrates by ␤ARK or related kinases is enhanced in the presence of ␤␥ subunits Kim et al., 1993b). Interestingly, such results are obtained in the presence of putative synergistic modulators of the kinase (activated receptors, synthetic receptor fragments, mastoparan) but not in its absence (Pitcher et al., 1992;Kim et al., 1993b). Taken together, these data suggest that G protein ␤␥ subunits stimulate the enzymatic activity of ␤ARK, in addition to its possible role as a membrane anchor for the kinase (Haga et al. 1994a; also see below).
Our results using either plasma or microsomal membranes support a clear distinction between the process of ␤ARK association with cellular membranes and that of kinase activation. Since the association of ␤ARK with cellular membranes does not require G protein stimulation, it could take place under resting, basal conditions, whereas ␤ARK activation leading to substrate phosphorylation would be dependent on the stimulation of G proteins by specific signals. This is consistent with previous data showing that ␤ARK associates equally well to the heterotrimeric G protein or to the ␤␥ dimer alone, thus leading to the suggestion that ␤ARK targeting to the plasma membrane may take place prior to receptor activation (Kim et al., 1993b).
A key issue to be addressed is the identity of the protein(s) involved in the association of ␤ARK with the microsomal membranes. To date, ␤ARK1 has been reported to interact with the activated ␤ 2 AR and with ␤␥ dimers, which have been proposed to play a key role in the kinase targeting to the periphery of the plasma membrane (Pitcher et al., 1992;Kim et al., 1993b). The presence of ␤␥ subunits in our salt-stripped microsomal membrane preparation argued for a role of these proteins in ␤ARK association. The fact that G protein stimulators did not have any effect on 35 S-labeled ␤ARK binding did not rule out its participation, since it has been previously shown that the activation of purified G i /G o proteins does not modulate the kinase binding to heterotrimeric G proteins, which could be an entity recognized by ␤ARK (Kim et al., 1993b). On the other hand, the functional link between membrane-bound ␤ARK and activated G proteins discussed above does not necessarily imply G proteins as the only anchors of ␤ARK, since (as mentioned under "Results") ␤ARK could also be binding to a different protein in the stripped microsomes, and its activity modulated by additional interactions with ␤␥ subunits released upon G protein stimulation.
Two main lines of evidence support the notion of a binding site for ␤ARK other than G protein subunits. First, the lack of effect on ␤ARK association of membrane pretreatment with Na 2 CO 3 pH 11, which leads to an ϳ50% reduction in immunoreactive G proteins (Kehlenbach et al. (1994) and Fig. 4C). Second, the effect of different GST-␤ARK fusion proteins on ␤ARK binding. It has been described that the carboxyl-terminal portion of ␤ARK-1 is responsible for the activation by G protein ␤␥ subunits and that the minimal ␤␥ binding domain is localized to a 125-amino acid stretch Kameyama et al., 1993). This region partially overlaps with a pleckstrin homology domain present in ␤ARK . It is worth noting that, although pleckstrin homology domains have been recently reported to interact with phospholipid vesicles rich in phosphatidylinositol 4,5-bisphosphate (Harlan et al., 1994), this does not appear to play a predominant role in the interaction of ␤ARK with microsomal membranes, since we have previously shown that the association is heat and protease-sensitive and therefore involves a protein component of the microsomes (García-Higuera et al., 1994a). Interestingly, our results show that GST-FP2, a fusion protein containing the COOH terminus of ␤ARK which has been used by others to characterize ␤ARK interactions with ␤␥ subunits (Pitcher et al., 1992;Inglese et al., 1994;Touhara et al., 1994), does not inhibit ␤ARK1 binding to microsomal membranes, whereas it is able to interact with purified ␤␥ subunits (Fig. 8B) and to inhibit the ␤␥ effect on rhodopsin phosphorylation by ␤ARK (Fig. 8C). In agreement with these results, recent experiments 2 indicate that purified phosducin, which is able to interact with G protein ␤␥ subunits (DebBurman et al., 1995) does not inhibit ␤ARK binding to microsomal membranes in our experimental conditions (94 Ϯ 12% of control binding at 300 -450 nM phosducin, mean Ϯ S.E. of four experiments). On the contrary, a fusion protein containing an NH 2 -terminal portion of the kinase (residues 50 -145) which does not bind purified ␤␥ subunits (FP1), strongly inhibits ␤ARK association with a potency similar to other reported effects of ␤ARK fragments Inglese et al., 1994;Touhara et al., 1994). A similar inhibitory effect on ␤ARK binding can be observed with a GST-␤ARK fusion protein comprising amino acids 1-147, whereas a 1-88 fragment is without effect, thus suggesting the ␤ARK anchoring domain may reside within residues 88 -145.
Although at present we cannot totally exclude the possibility that intact G proteins or specific ␤␥ dimers play a role in ␤ARK association with microsomal membranes, we feel that our data strongly suggest that ␤ARK binding is preferentially mediated via a high affinity interaction with a currently unidentified microsomal protein. As shown in the model depicted in Fig. 10, such protein component (X) appears to interact with a region of the amino terminus of the kinase, thus suggesting a new targeting domain of ␤ARK. Bound ␤ARK would be inactive until G protein stimulation leads to additional interactions of the COOH terminus of ␤ARK with ␤␥-subunits. In line with other recent results, our data suggest that the regulation of ␤ARK activity and subcellular distribution will likely involve multiple interactions with G protein subunits, phospholipids (DebBurman et al., 1995), different domains of G protein-coupled receptors Ruiz-Gómez et al., 1994), and additional anchoring proteins. In this regard, the existence of anchor proteins has been previously reported for other protein kinases such as protein kinase A (Hausken et al. (1994) and references therein) and protein kinase C (Ron et al., 1994), and emerges as a general mechanism for regulating the subcellular distribution and activity of protein kinases (Mochly-Rosen, 2 C. Murga, M. Lohse, and F. Mayor, Jr., unpublished results. 1995). Future research will strive to identify the ␤ARK anchor protein in the microsomes, and to more precisely localize the ␤ARK binding domain. Although it has been suggested that the NH 2 terminus of GRKs may interact with activated receptors (Palczewski et al., 1993), 3 there is little knowledge on the role of this domain, which does not have significant sequence homology with other known GRKs, including GRK1, GRK6, and GRK5, which have been shown to display a different mechanism of membrane association (Kunapuli et al., 1994). Further investigation using different experimental approaches would be needed to ascertain the factors (expression of G proteincoupled receptors, specific combinations of heterotrimeric G proteins, kinase anchors . . . ) and mechanisms governing the complex subcellular distribution of this key regulatory kinase and the possible function(s) of ␤ARK in microsomal membranes (García-Higuera et al., 1994a).