Endogenous ADP-ribosylation of the G protein beta subunit prevents the inhibition of type 1 adenylyl cyclase.

Mono-ADP-ribosylation is a post-translational modification of cellular proteins that has been implicated in the regulation of signal transduction, muscle cell differentiation, protein trafficking, and secretion. In several cell systems we have observed that the major substrate of endogenous mono-ADP-ribosylation is a 36-kDa protein. This ADP-ribosylated protein was both recognized in Western blotting experiments and selectively immunoprecipitated by a G protein beta subunit-specific polyclonal antibody, indicating that this protein is the G protein beta subunit. The ADP-ribosylation of the beta subunit was due to a plasma membrane-associated enzyme, was sensitive to treatment with hydroxylamine, and was inhibited by meta-iodobenzylguanidine, indicating that the involved enzyme is an arginine-specific mono-ADP-ribosyltransferase. By mutational analysis, the target arginine was located in position 129. The ADP-ribosylated beta subunit was also deribosylated by a cytosolic hydrolase. This ADP-ribosylation/deribosylation cycle might be an in vivo modulator of the interaction of betagamma with specific effectors. Indeed, we found that the ADP-ribosylated betagamma subunit is unable to inhibit calmodulin-stimulated type 1 adenylyl cyclase in cell membranes and that the endogenous ADP-ribosylation of the beta subunit occurs in intact Chinese hamster ovary cells, where the NAD(+) pool was labeled with [(3)H]adenine. These results show that the ADP-ribosylation of the betagamma subunit could represent a novel cellular mechanism in the regulation of G protein-mediated signal transduction.

Endogenous mono-ADP-ribosylation has also been described in eukaryotic cellular systems; ADP-ribosyltransferases that catalyze ADP-ribosylation of arginine residues of G proteins (similar to cholera toxin) have been described in many cells and tissues (9 -12). The endogenous ADP-ribosylation of cysteine residues of membrane G proteins (similar to pertussis toxin) has also been suggested to occur in erythrocytes (13,14). Thus, some of these enzymes are able to modify G proteins (15)(16)(17) and presumably play a role in signal transduction, although their substrates have been poorly characterized and their functional significance is even less understood.
It has also been proposed that an ADP-ribosyltransferase may be coupled to an ADP-ribosylarginine hydrolase that is able to remove the ADP-ribose group and hence regenerate free arginine, completing an ADP-ribosylation cycle that can reversibly regulate the functions of substrate proteins (18,19). An example of this cycle in eukaryotes is given by desmin, the muscle-specific intermediate filament protein as follows: ADPribosylation blocks the assembly of desmin into 10-nm filaments in vitro, and an incubation with ADP-ribosylarginine hydrolase restores the self-assembly properties of desmin (20 -22).
Here we report the direct demonstration of endogenous mono-ADP-ribosylation of the G protein ␤ subunit, and we provide evidence that this modification can modulate ␤␥ activity in a similar way to the regulation of some G protein ␣ subunits. Thus we propose that the ADP-ribosylation/deribosylation cycle of the ␤␥ subunit might represent a novel cellular mechanism to regulate G protein-mediated signal transduction. ethyl ketone-treated trypsin was from Worthington. Other chemicals used were obtained from Sigma at the highest available purities. Part of the purified bovine brain ␤␥ and the antibodies raised against the carboxyl terminus and the amino terminus of ␤ subunit were generously supplied by Dr. W. F. Simond (National Institutes of Health, Bethesda). Baculovirus encoding His 6 -␣ i1 was a generous gift of Dr. P. Gierschik (University of Ulm, Germany).
Cell Culture, Plasma Membrane, and Cytosol Preparation-Chinese hamster ovary (CHO) cells were grown in monolayers at 37°C in 95% air, 5% CO 2 in DMEM supplemented with 10% fetal calf serum, 34 g/ml proline, 100 units/ml penicillin, 100 mg/ml streptomycin. Plasma membranes were prepared as described previously (23), with the following modification: cells (6 ϫ 10 8 cells for each preparation) were washed with HBSS, detached with hypotonic buffer containing 10 mM TES (pH 7.0) and 1 mM EDTA, and then lysed with a Teflon/glass Potter homogenizer. CHO cytosol was prepared as described previously (24), with the following modification: CHO cells (detached in HBSS without Ca 2ϩ and Mg 2ϩ , plus 5 mM EGTA), were broken (10 9 cells/ml) by sonication in 5 mM Tris-HCl (pH 8.0), 1 mM EGTA, and protease inhibitors.
ADP-ribosylation Assay-Samples (4-g plasma membranes) were incubated at 37°C for 60 min in 50 l of buffer containing 50 mM potassium phosphate (pH 7.4), 0.5 mM MgCl 2 , 4 mM dithiothreitol (DTT), 5 mM thymidine, 30 M ␤-NAD, 1-2 Ci of [ 32 P]NAD (specific activity, 1000 Ci/mmol), and (where specified) 0.05-0.5 g of purified bovine brain ␤␥ (either generously provided by Dr. W. F. Simonds, National Institutes of Health, Bethesda, or purified as described previously (25)). Pertussis and cholera toxin ADP-ribosylations were performed as described previously (26). The reactions were terminated by diluting the samples with 50 l of Laemmli sample buffer, followed by 2 min boiling, and analysis by 10% SDS-PAGE without or with 4 M urea. Proteins were electroblotted (4 h at 500 mA) onto nitrocellulose membranes, and the filters were exposed for about 12 h to Kodak X-Omat film using an intensifying screen. For quantitative analysis an Instantimager (Packard Instrument Co.) was employed.
Deribosylation Assay-Following the ADP-ribosylation assay, 32 Plabeled plasma membranes were washed twice with 5 mM Tris-HCl (pH 8.0) and incubated for 30 min at 37°C with 50 g of CHO cell cytosol, without or with 10 mM MgCl 2 (in 50 l of 5 mM Tris-HCl (pH 8.0) and protease inhibitors). The analysis of protein samples was performed as described above for the ADP-ribosylation assay, whereas the labeled compounds released in the supernatant were separated by HPLC with a Partisil 10 SAX column (4.6 mm ϫ 25 cm; Whatman) using a nonlinear gradient of 0 -1 M ammonium phosphate (pH 3.35) at a flow rate of 1 ml/min (a linear gradient of 0 -15 mM ammonium phosphate for the first 45 min, 15-24 mM for 1 min, 24 -45 mM from 46 to 80 min, and 45 mM to 1 M for the last min). Fractions of 1 ml were collected, and the 32 P radioactivity was evaluated by scintillation counting.
Protein Cleavage-Membrane proteins were ADP-ribosylated for 2 h at 37°C. Trypsin, dissolved in 1 mM HCl, was added at a concentration of 0.01 mg/ml (the final HCl concentration of 0.1 mM did not affect the ADP-ribosylation). The samples were incubated in the presence of trypsin or HCl for 30 min at 37°C, as described previously (29). Proteolysis was terminated by the addition of Laemmli sample buffer.
Size Exclusion Chromatography-200 g of washed, [ 32 P]ADP-ribosylated plasma membranes were stirred at 4°C for 60 min in 0.6 ml of PBS buffer (pH 7.4) containing 1% sodium cholate and 1 mM DTT. The samples were centrifuged at 100,000 ϫ g for 60 min, and 0.2 ml of the cholate extract (supernatant) were applied to a Sepharose 12 HR 10/30 size exclusion column (Amersham Pharmacia Biotech) at a flow rate of 0.4 ml/min. The column was equilibrated and eluted with PBS buffer containing 1% cholate and 1 mM DTT. Fractions of 0.2 ml were collected, with the protein elution pattern being evaluated by the absorbance at 254 nm. The column was calibrated with the following molecular mass standard proteins (Amersham Pharmacia Biotech): aldolase (158 kDa), albumin (67 kDa), ovalbumin (43 kDa), chymotrypsinogen A (25 kDa), and ribonuclease (13.7 kDa).
Metabolic Labeling of Intact CHO Cells with [ 3 H]Adenine-Confluent CHO cells (2.5 ϫ 10 6 cells/35 mm dish) were metabolically labeled as described previously (30,31), with the following modifications: the cells in each well were incubated for 16 h at 37°C with 100 Ci of [ 3 H]adenine (specific activity, 20 -40 Ci/mmol) in 1 ml of DMEM without fetal calf serum, containing 20 g/ml actinomycin D to avoid [ 3 H]adenine incorporation into RNA. The labeled medium was replaced with fresh medium, and the incubation was carried out for an additional 4 h in the absence or in the presence of pertussis toxin (100 ng/ml). Then the supernatant was removed, and the incubation was continued for 10 min at 37°C with RNase (100 g/ml) in 100 l of 20 mM Tris-HCl (pH 7.5). The reaction was terminated by diluting the samples with 50 l of Laemmli sample buffer, followed by 2 min boiling and analysis by 10% SDS-PAGE. The protein-associated radioactivity was evaluated by a Bio-Imaging Analyzer (FUJI Film) or by fluorography with the gels exposed for at least 20 and 90 days, respectively. To verify the [ 3 H]adenine incorporation into the cellular NAD ϩ pool, radiolabeled CHO cells were extracted with methanol/chloroform/water/12 N HCl (1:1:0.5:0.01) and analyzed by HPLC as described above, with the following modifications: H 2 O for the first 5 min, followed by a linear gradient of 0 -30 mM ammonium phosphate (5-55 min), and of 30 mM to 1 M (55-115 min). [ 3 H]NAD represented ϳ8% of the total [ 3 H]adenine metabolites. Moreover, the CHO intracellular NAD concentration was determined, considering that 10 6 cells have a volume of 1.2 l (as measured by a Coulter Counter ZM linked to a Coulter Channelizer 256), and was found to be 785 Ϯ 10 M. This allows the calculation of the specific radioactivity of the NAD ϩ pool as 380 Ci/mmol, the value used to estimate the amount of the endogenously mono-ADP-ribosylated ␤ subunit.
Affinity Purification of the Endogenously 3 H-Labeled ␤␥ Subunit-[ 3 H]Adenine-labeled CHO cells (6 ϫ 10 7 for each experiment) were washed with HBSS and then broken with a Teflon/glass Potter homogenizer in hypotonic buffer containing 5 mM Tris-HCl (pH 8.0), 1 mM EGTA, and protease inhibitors. Unbroken cells and nuclei were removed by low speed centrifugation (10 min at 600 ϫ g), and the crude membranes (300 g) were collected by centrifuging the supernatant for 15 min at 25,000 ϫ g. The 3 H-labeled ␤␥ subunit were extracted and analyzed as described previously (32), with the difference that His 6 -␣ i1 was employed to reassociate with ␤␥. After washing with NaSCN, the bound ␤␥ subunits were solubilized in Laemmli sample buffer and analyzed as described above. About 2 g of ␤␥ were recovered from this affinity purification (as estimated by densitometric analysis of the immunoblot) with ϳ0.2% being modified in intact cells.
Adenylyl Cyclase Assay-Membranes (25 g/25 l) prepared from freshly dissected rat brain (33,34) were assayed for adenylyl cyclase activity as described previously (35), with the exception that cAMP was separated by thin layer chromatography using silica gel G plates (Kieselgel 60 F 254 , Merck) pretreated with 1% potassium oxalate and 2 mM EDTA, with chloroform/methanol/4 M NH 4 OH (54:42:12) as the solvent system. The effect of the ADP-ribosylated ␤␥ subunit was evaluated by adding to the reaction mixture 7 g of plasma membranes from CHO cells preincubated for 6 h at 37°C with different concentrations of purified bovine brain ␤␥ subunit (100 -500 nM), in the absence ("unmodified") or presence ("modified") of 1 mM NAD ϩ (under these experimental conditions about 60 -80% of the added ␤␥ subunit was ADP-ribosylated). After the incubation, CHO membranes containing either unmodified or modified ␤␥ subunit were centrifuged (15 min at 12,000 ϫ g) and resuspended in 5 l of 10 mM HEPES (pH 8.0) and 0.1% Lubrol, and then added to the adenylyl cyclase assay mixture.
Other Methods-Point mutations were generated using the Quickchange, site-directed mutagenesis kit (Stratagene), and the sequence of all constructs was confirmed by automated DNA sequencing. Snake venom phosphodiesterase digestion of ADP-ribosylated proteins (33,36), production of [ 32 P]ADP-ribose (36), sensitivity of the ADP-ribosylated protein to hydroxylamine (NH 2 OH) and HgCl 2 (36), and phosphoinositide-specific phospholipase C assay (37) were performed as described previously.

RESULTS AND DISCUSSION
Identification of a 36-kDa ADP-ribosylated Protein as the G Protein ␤ Subunit-Substrates of endogenous ADP-ribosylation can be identified by supplying cell extracts with radiolabeled NAD ϩ . In enriched plasma membrane preparations from different cell lines, including Swiss 3T3, CHO and HL60 cells, NAD ϩ prominently labeled a 36-kDa protein (Fig. 1A). In the CHO preparation, the labeled protein co-migrated with the purified G protein ␤ subunit on SDS-PAGE and was recognized on Western blots by a polyclonal antibody raised against the carboxyl-terminal decapeptide of the ␤ subunit (SW28, which stains a doublet representing the ␤ 1 and ␤ 2 isoforms) ( Fig. 1, B and C). Purified bovine brain ␤␥ subunit added to the assay mixture was also ADP-ribosylated (Fig. 1A, lanes 2, 4, 6, 8, and 10 and Fig. 1, B and C, lane 2) and precisely co-migrated with the ADP-ribosylated 36-kDa endogenous protein on SDS-PAGE (Fig. 1A, lanes 1, 3, 5, 7, and 9 and Fig. 1, B and C, lanes 1, and 3-6). The labeled 36-kDa band contained both the ADPribosylated ␤ 1 and ␤ 2 subunits, and the doublet could be fully resolved employing extra-long gels for the protein separation ( Fig. 1, B and C). It has been reported that the electrophoretic mobility of the G protein ␤ subunit is decreased when urea (4 M) is introduced into the SDS-PAGE (38). Under these conditions, the ADP-ribosylated protein recognized by the SW28 antibody had a higher apparent molecular mass (39 kDa) and, again, precisely co-migrated with pure ␤ subunit (Fig. 1C), whereas the mobility of other proteins, such as the G protein ␣ subunit (␣ i , ADP-ribosylated by pertussis toxin, lane 4, and ␣ s , ADPribosylated by cholera toxin, lane 6, of Fig. 1, B and C), remained unchanged. Thus the major substrate of endogenous ADP-ribosylation in these enriched plasma membrane preparations appears to be the G protein ␤ subunit. Moreover, to exclude the possibility that this is another ADP-ribosylated protein that can co-migrate on SDS-PAGE (plus or minus urea) with the ␤ subunit, we immunoprecipitated the solubilized, ADP-ribosylated CHO membranes with the SW28 antibody. Both the labeled endogenous 36-kDa protein ( Fig. 2A, lane 3) and the labeled purified ␤ subunit ( Fig. 2A, lane 2) were im-munoprecipitated by the SW28 antibody; as a control, nonimmune rabbit serum did not precipitate any radioactivity ( Fig. 2A, lane 1). Moreover, when solubilized from membranes and analyzed by gel filtration chromatography, the ADP-ribosylated 36-kDa protein eluted as a single peak of ϳ50-kDa (Fig.  2B), which completely overlapped with the peak of the ␤ subunit as revealed by the SW28 antibody (Fig. 2C). Thus we provide compelling evidence indicating that the substrate of endogenous ADP-ribosylation is the ␤ subunit of the heterotrimeric G protein.
The G Protein ␤ Subunit Is ADP-ribosylated by an Endogenous Arginine-specific Mono-ADP-ribosyltransferase, and Its ADP-ribosylation Is Reversed by an ADP-ribosylarginine Hydrolase-To verify that the modification of the ␤ subunit is indeed enzymatic mono-ADP-ribosylation, the labeled protein was digested with snake venom phosphodiesterase. This caused the release of radiolabeled 5Ј-AMP (ϳ70% of total radioactivity, as analyzed by HPLC; data not shown), which is considered diagnostic of mono-ADP-ribosylation (33,36). Nonenzymatic mono-ADP-ribosylation, which might be caused by the formation of adducts with ADP-ribose generated from NAD ϩ by cellular NAD-glycohydrolases (NADases) (39), was ruled out because the ␤ subunit was not labeled by free ADPribose (Fig. 3A, lane 2).
The modified amino acid of the ␤ subunit was further investigated by means of a characterization of the chemical stability of the ADP-ribosyl linkage. Treatment of the [ 32 P]ADP-ribosylated ␤ subunit for 12 h with NH 2 OH (which hydrolyzes the ADP-ribose of arginine), as opposed to HgCl 2 or HCl (which act on ADP-ribosylated cysteine and serine/threonine residues, respectively), completely removed the ADP-ribose bound to the ␤ subunit (Fig. 3B, lane 2), as we also observed with the [ 32 P]ADP-ribosylated ␣ subunit of G s induced by cholera toxin (Fig. 3B, lane 4). The possibility that the ADP-ribose is linked to glutamate could be ruled out since a 20-min treatment with NH 2 OH (a time sufficient to hydrolyze the ADP-ribose linked to glutamate) removed only 22% of the label from the ADP-ribo-  1 and 3) and in the presence (lane 2) of purified bovine brain ␤␥ subunit (2.5 g/ml) were solubilized and subjected to immunoprecipitation using a non-immune serum (lane 1) or the ␤-specific SW28 antiserum (lanes 2 and 3). B and C, gel filtration of the ADPribosylated G protein ␤ subunit. B, autoradiography of the nitrocellulose filter containing the indicated fractions. C, immunoblot of the same fractions with the SW28 polyclonal antibody. Fractions 33-35 correspond to a molecular mass of 53 kDa. The data shown are from a single experiment performed in duplicate, which is representative of at least three independent experiments. sylated ␤ subunit (a 55% loss of label was observed after 2 h and, as mentioned above, the complete loss of label occurred at 12 h). These results indicate that the mono-ADP-ribosylation occurs at an arginine residue of the ␤ subunit.
In line with these data, the mono-ADP-ribosylation of the ␤ subunit was inhibited by agmatine (data not shown) and by meta-iodobenzylguanidine (MIBG) in a dose-dependent manner (Fig. 3C), both well characterized inhibitors of argininespecific mono-ADP-ribosyltransferases (40,41). Thus, a plasma membrane-associated, arginine-specific, mono-ADP-ribosyltransferase can ADP-ribosylate the ␤ subunit of heterotrimeric G proteins. Notably, the enzyme displayed a good degree of specificity for the ␤ subunit, indicated by the fact that this protein is the most intensely labeled in membrane preparations (as seen in Fig. 1, B and C). A kinetic investigation of the ADP-ribosylation assays employing purified bovine brain ␤␥ subunit and enriched plasma membranes as the enzyme source gave a K m value for NAD ϩ of 350 Ϯ 20 M, a value compatible with the physiological concentrations of NAD ϩ (see "Experimental Procedures" for calculation of the CHO intracellular NAD concentration and Ref. 42). The V max was 500 Ϯ 80 pmol/h/mg membrane protein. This rate might conceivably be regulated by activators (and/or co-factors) yet to be identified and, hence, might well be faster in the natural intracellular milieu.
Few enzymes that catalyze ADP-ribosylation reactions have been purified and characterized, but among these the best known are the arginine-specific mono-ADP-ribosyltransferases. Most of the NAD:arginine ADP-ribosyltransferases purified so far are glycosylphosphatidylinositol (GPI)-anchored proteins located on the extracellular side of the plasma membrane or on the luminal face of intracellular organelles, thus being physically separated from intracellular substrates (43).
The enzyme involved in ␤ subunit mono-ADP-ribosylation was markedly enriched in plasma membranes, from which it could not be separated by high salt, suggesting that it is an integral membrane protein. Moreover, phosphoinositide-specific phospholipase C, which hydrolyzes GPI-anchored membrane proteins (37), did not release the ADP-ribosyltransferase (whereas, as expected, it did release a GPI-anchored NADase in parallel experiments; data not shown). More importantly, intact cells were not able to induce ADP-ribosylation of the purified ␤␥ subunit, whereas, under the same conditions, broken cells could (data not shown). These data indicate that the enzyme involved in the ␤ subunit modification is not GPIanchored and has a cytoplasm-oriented catalytic site.
This ADP-ribosyltransferase is present in most cell types, but the extent of ␤ subunit labeling differed significantly among different cell lines (Fig. 1A). These differences might be due at least in part to the different rates of NAD ϩ metabolism in the cells examined, which in turn depend on the presence of membrane-associated NAD ϩ -utilizing enzymes such as ADPribosyltransferases, NADases, and NAD-pyrophosphatases. For instance, in FRTL-5 plasma membranes, which possess a NADase that rapidly and completely hydrolyzes NAD ϩ to ADPribose and nicotinamide (26), the ADP-ribosylation of the ␤ subunit was barely detectable, whereas in CHO or HL60 plasma membranes, where 80% of the added NAD ϩ is still available after a 2-h incubation, the labeling of the ␤ subunit was very pronounced (Fig. 1A).
It has been postulated that ADP-ribosylation is a reversible process in animal cells. ADP-ribosylarginine hydrolases are ubiquitous cytosolic enzymes that have been purified and cloned from rat brain (44) and subsequently from mouse and human brain (45). The rat hydrolase activity is Mg 2ϩ -and thiol-dependent, whereas the recombinant human hydrolase activity is Mg 2ϩ -but not thiol-dependent. In order to investigate whether the ADP-ribosylated ␤␥ subunit could be a substrate of an arginine hydrolase, plasma membranes containing 12 ng of [ 32 P]ADP-ribosylated ␤ subunit were incubated with CHO cell cytosol in the presence of protease inhibitors (see "Experimental Procedures"). Fig. 4A shows that the cytosol was able to decrease the labeling of the ␤ subunit (compare lanes 3 and 4) and that the concomitant addition of MgCl 2 (lane 2) caused a further decrease; samples incubated with control buffer without or with MgCl 2 (lanes 1 and 4, respectively) still contained 12 ng of modified ␤ subunit, whereas the amount of modified ␤ subunit was reduced to ϳ9 ng after the incubation with cytosol (lane 3) and to ϳ6 ng after the incubation with cytosol plus MgCl 2 (lane 2) (as evaluated by Instantimager, Fig.  4D). Since the total amount of the ␤ subunit was not affected during these de-ribosylation experiments (as demonstrated by Western blot analysis, Fig. 4B), we can conclude that a cytosolic protein, which could be Mg 2ϩ -dependent, is able to decrease the ADP-ribosylation of the ␤ subunit. The addition of DTT in the de-ribosylation experiments (not shown) did not affect the level of modified ␤ subunit. HPLC analysis of the labeled compounds released from the [ 32 P]ADP-ribosylated ␤ subunit upon incubation with cell cytosol showed a peak of [ 32 P]ADP-ribose (Fig.  4C, dashed line, E), indicating the activity of a cytosolic, ADPribosylarginine hydrolase. In the presence of MgCl 2 , which caused a further decrease in the labeling of the ␤ subunit, there was not a parallel increase in [ 32 P]ADP-ribose (Fig. 4C, 2). As a control, the cholera toxin-dependent ADP-ribosylation of ␣ s , modified on arginine, is shown (lanes 3 and 4). The labeled proteins, after blotting, were cut out and treated for 12 h with 1 M NH 2 OH, which specifically hydrolyzes the covalent bond of ADP-ribose to arginine (lanes 2 and 4), or, as a control, with 1 M NaCl (lanes 1 and 3). C, inhibition of ␤ subunit ADP-ribosylation by decreasing concentrations of the arginine-specific ADP-ribosyltransferase inhibitor, MIBG: lane 1, buffer alone; lanes 2-6, decreasing concentrations of MIBG (50, 25, 15, 5, and 2 M). The data shown (A-C) are from a single experiment performed in duplicate, which is representative of at least three independent experiments.
[ 32 P]AMP would account for the 32 P i formed. Indeed, it has been reported that pyrophosphatases able to act on free ADPribose are present in the cytosol of different cells (46). Altogether these results are consistent with the proposal that the ␤␥ subunit can undergo an endogenous ADP-ribosylation/deribosylation cycle. However, the possibility that the ADP-ribosylated ␤ subunit might be processed not only by a hydrolase but also by a pyrophosphatase, as described in muscle cells for ADP-ribosylated integrin (47), cannot be excluded.
The G Protein ␤ Subunit Is ADP-ribosylated in Intact Cells-Due to the well recognized role that the G protein ␤␥ subunit plays in cell regulation (48), it was important to analyze whether its mono-ADP-ribosylation could occur under physiological conditions. Since NAD ϩ is a membrane-impermeant compound, it must either be injected into the cell or metabolically radiolabeled in order to study endogenous ADP-ribosylation. It has been previously shown that the labeling with [ 3 H]adenine permits the study of toxin-catalyzed G proteins ADP-ribosylation in intact cells (30) and the in vivo ADPribosylation of GRP78/BiP protein (49). Thus the mono-ADPribosylation of the ␤ subunit was analyzed by in vivo labeling of intact CHO cells employing [ 3 H]adenine to label metabolically the cellular NAD ϩ pool, followed by separation of the labeled ␤ subunit with an affinity column (Fig. 5).
It is known that [ 3 H]adenine enters cells by carrier-mediated transport and is then phosphoribosylated to 5Ј-AMP which is in turn phosphorylated to ATP that enters the NAD ϩ pool (50). Fig. 5A shows that indeed [ 3 H]adenine is incorporated into intracellular NAD ϩ pool during the 16 h of labeling (see "Experimental Procedures), being the NAD ϩ -associated cpm corresponding to 8% of the total cpm incorporated in the cells. The SDS-PAGE of 3 H-labeled CHO cells extract (Fig. 5B, lane1) revealed a major labeled band with a mass of approximately 70 kDa that may correspond to the modified GRP78/BiP protein (49), and some minor labeled bands. Among these, a 36-kDa band co-migrating with the ADP-ribosylated ␤ subunit was observed. Moreover, as expected, treatment of [ 3 H]adeninelabeled CHO cells with pertussis toxin before extraction caused the labeling of the G protein ␣ i/o subunits (Fig. 5B, lane 2). Thus, to explore the possibility that the labeled band of 36 kDa was the ␤ subunit, crude membranes were prepared from [ 3 H]adenine-labeled CHO cell extract, and the ␤ subunit they contain was affinity purified employing His 6 -␣ i1 bound to the nickel-nitrilotriacetic acid-agarose resin (see "Experimental Procedures"). Fig. 5C shows the fluorography of 3 H-labeled ␤ subunit obtained after its affinity purification (lane 1); as a control, 300 g of [ 3 H]adenine-labeled CHO crude membranes were incubated with nickel-nitrilotriacetic acid-agarose resin only (lane 2), whereas lane 3 shows the labeling of 30 g of the [ 3 H]adenine-labeled CHO crude membranes. From the densitometric analysis of the Western blot (Fig. 5D), we can conclude that ϳ2 g of ␤␥ were recovered from this affinity purification with ϳ0.2% being modified in intact cells under basal condition. The possibility that the percentage of the modified ␤ subunit will increase upon receptor activation is presently under investigation.
From these data it can be concluded that the substrate of endogenous ADP-ribosylation in CHO cells metabolically la- The ␤ Subunit ADP-ribosylation Blocks the Inhibitory Effect of ␤␥ on Type 1 Adenylyl Cyclase-Free ␤␥ subunit can directly activate several effectors (48), and it is generated upon ligand binding, when an activated heptahelix receptor causes GDP/ GTP exchange on the G protein ␣ subunit, which then dissociates from both the receptor and the ␤␥ dimer (48). This is not, however, the only process controlling the ␤␥ state of activation. For instance, the RGS proteins (regulators of G protein signaling) increase the GTPase activity of the G␣ subunit, resulting in the accumulation of GDP-G␣ and its re-association with the free ␤␥ subunit (51). Phosducin can bind to the ␤␥ subunit directly and translocate it to the cytosol, thus preventing its association with G␣ or with effectors (52). In the present study we propose that mono-ADP-ribosylation might represent an additional mechanism to regulate directly the activity of the ␤␥ subunit. The portions of the ␤␥ subunit involved in establishing contact with the G␣ subunit and the effector proteins are increasingly being identified (53). Recent studies indicate that the amino-terminal region is involved not only in the interaction with the ␣ subunit but also with adenylyl cyclase (54), phospholipase C-␤ 2 (54), and the muscarinic atrial potassium channel (54). A small carboxyl-terminal segment of the ␤ subunit has also been proposed to take part in the activation of phospholipase C-␤ 2 (55). In order to determine whether the mono-ADP-ribosylation of the ␤ subunit occurs on its amino-or carboxyl-terminal moiety, we used a well established tryptic assay (29). The digestion of the ␤␥ subunit by trypsin yields only two fragments with apparent molecular masses on SDS-PAGE of 14 (amino terminus) and 27 kDa (carboxyl terminus), which can also be identified by antibodies specific for the ␤ amino and carboxyl termini. Trypsin cleavage of labeled membranes resulted in a reduced labeling of the 36-kDa protein and the appearance of a labeled ϳ14-kDa peptide (Fig. 6A, compare lane 2 with lane 5 and lane 3 with lane 6), which was recognized by the antibody specific for the amino-terminal fragment of the ␤ subunit (Fig. 6B), indicating that this is the domain covalently modified by ADP-ribose. Following trypsin cleavage, purified bovine brain ␤␥ (used as a control) was completely digested, yielding the two expected fragments of 27 and 14 kDa (compare lane 4 with lane 1 of Fig. 6, B and C), whereas the ADP-ribosylated endogenous ␤ subunit (compare lane 6 with lane 3 of Fig. 6, B and C) or exogenously added purified ␤␥ (compare lane 5 with lane 2 of Fig. 6, B and C) were partially digested, as clearly shown by immunoblot analysis, agreeing with previous observations that the membrane-associated ␤␥ subunit is partially protected from trypsin degradation (56).
The amino-terminal portion of the ␤ subunit encompassing amino acids 84 -143 appears to be the region involved in the interaction with adenylyl cyclases (57). Interestingly, there are three arginine residues (96, 129, and 134) in the region predicted to interact with the 956 -982 peptide of types 1 (AC1) and 2 (AC2) adenylyl cyclases (57). It is thus possible that arginine residues 96 and/or 129, which are located on the 14-kDa peptide produced by trypsin digestion of the ␤␥ subunit, could be the target of the endogenous ADP-ribosyltransferase and that their modification might affect the interaction with adenylyl cyclase. Thus, in order to identify the site of modification on the ␤ subunit, all arginine residues in the amino-terminal portion of this protein were mutated and replaced with lysine. The mutations were introduced into the background of rat ␤ 1 tagged at the amino terminus with the influenza hemagglutinin epitope. Addition of the hemagglutinin tag was important for the ADP-ribosylation assay in transfected CHO cells since the tagged ␤ subunit, larger than the endogenous ␤ subunit, could be easily identified on SDS-PAGE.
Indeed each ␤ 1 mutant was expressed in CHO cells with ␥ 1 and tested for the ability to be endogenously ADP-ribosylated. All mutants could support some level of ADP-ribosylation, except the ␤ 1 mutant R129L, which was impaired in the ADP-ribosylation assay, thus indicating that indeed the endogenous ADPribosylation of the ␤ subunit occurs on its arginine 129, one of the residues predicted to interact with adenylyl cyclases, as discussed above (57). Therefore, we explored the possibility that this modification on arginine 129 of the ␤ subunit could affect its ability to interact with AC1. Although the AC1 activity is absent in CHO membranes (the source of ADP-ribosyltransferase in our assays), it is abundant in brain membranes, where it can be efficiently stimulated by Ca 2ϩ /calmodulin (CaM) and directly inhibited by ␤␥ subunits (58). We thus used brain membranes as an AC1 source, and as expected, the unmodified ␤␥ subunit partially inhibited CaM-stimulated AC1 in brain membranes (in agreement with previously reported data, Ref. 58), with this ␤␥-dependent inhibition being only slightly reduced after prolonged incubation of the ␤␥ subunit at 37°C (Fig. 6D). In contrast, under the same conditions the purified brain ␤␥ subunit ADP-ribosylated in the presence of NAD ϩ and CHO membranes almost completely lost its inhibitory activity (Fig. 6D). In a series of control experiments, the ␤␥ subunit incubated with CHO membranes without NAD ϩ (or with NAD ϩ without membranes) was still able to inhibit AC1 activity (Fig. 6D); obviously, neither NAD ϩ and ADP-ribose nor CHO membranes alone had effects per se on the CaM-stimulated AC1 activity; finally, NAD ϩ added at the end of the incubation of the ␤␥ FIG. 6. ADP-ribosylation of the ␤ subunit on the 14-kDa aminoterminal fragment prevents the ␤␥ inhibition of AC1. A-C, ADPribosylated CHO plasma membrane proteins were blotted onto nitrocellulose sheets after being separated on 13% SDS-PAGE. Labeled proteins were visualized by autoradiography (A), and the G protein ␤ subunit was identified using two polyclonal antibodies: MS 1, raised against the ␤ subunit amino terminus (B), and SW28, raised against the ␤ subunit carboxyl terminus (C). A, lanes 1-3 show samples identical to those in lanes 4 -6, with the exception that they were incubated with trypsin for 30 min. Lanes 1 and 4, purified bovine brain ␤␥; lanes 2 and 5, ADP-ribosylation of CHO plasma membrane proteins in the presence of purified bovine brain ␤␥ (2.5 g/ml); lanes 3 and 6, endogenous ADP-ribosylation of the ␤␥ subunit. B and C show immunoblots of the same samples as A. D and E, ␤␥ subunit-induced inhibition of AC1 is blocked by ADP-ribosylation. CaM-stimulated AC1 activity was evaluated in rat brain membranes in the presence of different added amounts (100 -500 nM) of either unmodified (open squares) or ADP-ribosylated (closed squares) purified bovine brain ␤␥ subunits associated to CHO plasma membranes. The constant amount of endogenous ␤␥ in CHO plasma membranes was 78 nM. As a control, the closed circles show the inhibitory effect of free purified bovine brain ␤␥ subunit incubated for 6 h at 37°C (the experimental conditions of the AC1 assay), whereas the open circles show the inhibitory effect of the untreated free ␤␥ subunit. The values are mean Ϯ S.E. of seven experiments. * p Ͻ 0,05. E, immunoblot of either the ADP-ribosylated (lane 1) or unmodified (lane 2) bovine brain ␤␥ subunits associated with CHO plasma membranes, as employed in the AC1 assay. subunit with CHO membranes (thus, under conditions that prevent ADP-ribosylation of the ␤␥ subunit) did not interfere with ␤␥-dependent inhibition of AC1 activity. Altogether, these data indicate that the loss of activity of ␤␥ is due to ADPribosylation of ␤␥ by the membrane-associated ADP-ribosyltransferase. Furthermore, the loss of activity of the ADP-ribosylated ␤␥ subunit was not due to a change in association to membranes, since equivalent amounts of the ␤␥ subunit were found under all the experimental conditions employed (Fig. 6E, ADP-ribosylated ␤␥ lane 1, unmodified ␤␥ lane 2).
Notably, the loss of activity of the ADP-ribosylated ␤␥ subunit on CaM-stimulated AC1 appears not to be complete; it was complete at 200 nM and decreased at higher ␤␥ concentrations (Fig. 6D). This is probably due to the fact that not all of ␤␥ is ADP-ribosylated; indeed, under our experimental conditions the amount of ADP-ribosylated ␤␥ subunit (evaluated from the expected ϳ500-Da shift due to ADP-ribose on SDS-PAGE) was about 60 -80% of the total ␤␥ subunit (Fig. 6E). Thus, when 500 nM of "ADP-ribosylated" ␤␥ was applied to the system, the unribosylated ␤␥ amounted to 100 -200 nM, a concentration sufficient to cause maximal inhibition of cAMP production. Moreover it can be demonstrated that the addition of ADPribosylated ␤␥ (actual concentrations 10 -200 nM) cannot affect the inhibition due to 200 nM of unmodified ␤␥, hence indicating that the ADP-ribosylated ␤␥ is inactive on the inhibition of cAMP production (data not shown). This indicates that the ADP-ribosylated ␤␥ does not per se affect the enzyme activity. However, whereas the precise mechanism remains to be defined, it is clear that ADP-ribosylation strongly affects the ability of ␤␥ to interact with AC1.
As noted above, the ␤␥ amino terminus is involved in interactions with the ␣ subunit as well as with several effector proteins. It is therefore possible that amino-terminal ADPribosylation might be a general switch-off mechanism for ␤␥mediated responses. Indeed, our preliminary data on the ␤␥dependent activation of phospholipase C-␤ 2 activity suggest that this is the case. 2 Heterotrimeric G proteins are subjected to multiple modulatory inputs, as dictated by their central importance in cell regulation. Our finding of an ADP-ribosylation/deribosylation cycle acting on the ␤␥ subunit delineates a novel molecular mechanism that could provide crucial control of G proteinmediated signaling pathways.