INTRODUCTION

Covalent modification of proteins with lipids is particularly common among those proteins involved in signal transduction reactions (1). Two of the three subunits of heterotrimeric G proteins are so modified, as are many of their associated receptors and effectors. Lipidation of G proteins imparts essential functional characteristics to these molecules, affecting both protein-membrane and protein-protein interactions. Two of the known modifications of G proteins (prenylation and myristoylation) are stable metabolically, whereas palmitoylation is a dynamic process.

The  subunits of signal-transducing G proteins are prenylated near their carboxyl termini in reactions catalyzed by farnesyl and geranylgeranyl transferases. These modifications do not influence the formation of dimers but are essential for interactions of with the plasma membrane,  subunits, and effectors such as adenylyl cyclases (2). G protein  subunits of the i subfamily (i, o, g, t, and z) are myristoylated at their amino-terminal glycine residues (3). Such amidation occurs cotranslationally and is catalyzed by a protein N-myristoyl transferase (5). Myristoylation also facilitates interactions of these proteins with membranes, the dimer, and effectors (6, 7).

All heterotrimeric G protein  subunits except transducin appear to be fatty acylated (S-acylated) on cysteine residues near their amino termini by formation of thioesters (4). Labeling of proteins has usually been observed with [³H]palmitate. However, arachidonate can be incorporated similarly (8), and other S-acylated proteins incorporate myristate, stearate, and oleate as well as palmitate (9). Palmitoylation of G protein  subunits is dynamic; furthermore, activation of the -adrenergic receptor increases the rate of turnover of palmitate associated with G_s, suggesting that S-acylation of these proteins may alter their signal transducing properties (10, 11). Expression of mutant proteins that cannot be palmitoylated indicates that palmitoylation of nonmyristoylated  subunits may affect their association with membranes, and thus, that receptor-induced depalmitoylation would affect the cellular distribution of these proteins (12). However, examination of the biochemical properties of palmitoylated  subunits has been hampered by the lability and unknown stoichiometry of the modification. The thioester bond is sensitive to reducing agents commonly used during protein purification, and most studies of purified G protein  subunits have probably been performed on nonpalmitoylated protein.

The putative enzymology of protein palmitoylation is largely unknown. An activity capable of palmitoylating G protein  subunits has been detected in plasma membranes (13), as has one that incorporates palmitate into myristoylated fyn (14). This kinase and several other members of the src family are also dually acylated (myristoylated and palmitoylated) at their amino termini and resemble G protein  subunits in their amino-terminal amino acid sequences (15). Unfortunately, the proteins responsible for these activities have defied purification. By contrast, several proteins and peptides can be palmitoylated in the absence of a specific acyltransferase using palmitoyl-CoA as the acyl donor (16, 17). Furthermore, such autoacylation occurs at the same sites that become palmitoylated in vivo with myelin proteolipid protein and myelin P_o glycoprotein, raising the question of the requirement for an exogenous enzyme (18, 19). We show herein that certain G protein  subunits can be autoacylated stoichiometrically in vitro and that the requirements for this reaction resemble those described for palmitoylation in vivo.

EXPERIMENTAL PROCEDURES

Purification of G_i1

Myristoylated G_i1 was purified as described for the nonmyristoylated protein by Lee et al. (20) with the following modifications. Escherichia coli strain JM109 was cotransformed with pQE60(G_i1) and pBB131, an expression vector for Saccharomyces cerevisiae N-myristoyltransferase, and grown in the presence of carbenicillin (50 µg/ml) and kanamycin (50 µg/ml) (21). A single colony was then grown in 20 ml of LB medium containing ampicillin (50 µg/ml) and kanamycin (50 µg/ml) at 37 °C for 20 h, and this culture was used to inoculate 10 liters of T7 medium containing ampicillin (50 µg/ml) and kanamycin (5 µg/ml). The culture was grown at 30 °C to an absorbance of 0.6, at which time 60 µ isopropyl---thiogalactoside was added. The cells were harvested and lysed as described after 16 h at 30 °C. The lysate was applied to two HiLoad 26/10 Q-Sepharose columns (Pharmacia Biotech Inc.) in series and eluted with a gradient of NaCl (0 to 300 m). Fractions were assayed by both immunoblotting (antiserum BO87 (4)) and incorporation of [³H]palmitate as described below. Both activities comigrated, and peak fractions were pooled and applied to 10 ml of Macroprep Ceramic HAP resin (Bio-Rad) packed in a HR 10/10 column (Pharmacia), which was eluted with a gradient of potassium phosphate (0 to 300 m). Pooled material from the HAP column was brought to 1.2  (NH₄)₂SO₄ and applied to an HR 16/10 column packed with 20 ml of High Performance Phenyl Sepharose resin (Pharmacia), which was eluted with a gradient of (NH₄)₂SO₄ (1.2 to 0 ). Two immunoreactive peaks of G_i1 were resolved; the later eluting peak contains the myristoylated protein. It is highly purified and, in its nonmyristoylated form, can be crystallized (22). This material (50 mg) was concentrated to 4 ml by ultrafiltration, dialyzed into buffer A (20 m NaHepes (pH 8.0), 2 m MgCl₂, and 1 m EDTA) plus 1 m dithiothreitol, frozen, and stored at 80 °C.

[³H]Palmitate (1 mCi; 60 Ci/mmol; DuPont NEN) was dried under N₂, resuspended in 100 µl of ethanol, and added to 900 µl of 0.05% Triton X-100, 1 m CoA, 5 m ATP, 5 m MgCl₂, 10 m Tris-HCl (pH 7.7), and 2 units of acyl CoA ligase (Sigma). After incubation for 1 h at 30 °C, 10 ml of chloroform:methanol (1:1) was added, and the mixture was centrifuged at 2000 × g for 10 min. The supernatant was separated into two phases by the addition of 5 ml of chloroform and 2.5 ml of H₂O. The aqueous phase, containing the palmitoyl-CoA, was divided into 200-µl aliquots, dried under vacuum, and stored at 80 °C. The radiopurity of [³H]palmitoyl-CoA (>99%) was analyzed by thin-layer chromatography on Whatman C₁₈ reverse phase plates with isobutanol:H₂O:acetic acid (50:30:20) as the mobile phase, followed by fluorography.

[³H]Palmitoyl-CoA was dissolved in buffer A containing 100 µ palmitoyl-CoA and 15 m CHAPS to give a stock solution of palmitoyl-CoA with a specific activity of ~2000 cpm/pmol. G protein  subunits or heterotrimeric G proteins (0.5-2.0 µ, except where indicated) were incubated in buffer A with 10-20 µ palmitoyl-CoA and 7.5 m CHAPS. Aliquots of the reaction mixture (10 µl) were removed at the indicated times and added to 30 µl of 1% SDS/2 mg/ml of bovine serum albumin. Samples were then precipitated with 100 µl of 10% trichloroacetic acid, and precipitates were collected on a Whatman GFA glass fiber filter using a Minifold I dot-blotting apparatus (Schleicher & Schuell). Each well was washed five times with 200 µl of 50% ethanol/3% trichloroacetic acid. The filter was dried, and each filter dot was punched out, placed in 4 ml of scintillation fluid, vortexed vigorously, and counted by liquid scintillation spectrometry. All of the experiments described were replicated at least twice. Data shown are averages of duplicate or triplicate determinations of representative experiments.

G protein  subunit concentrations were determined by quantification of [³⁵S]GTPS¹ binding as described (23). Incubation times were 45 min (G_s), 150 min (G_i1 and G_o), or 90 min (with 1 m GTPS; G_q). -Agarose affinity chromatography was performed as described by Pang and Sternweis (24). Protein concentrations were determined either by staining with amido black or by the method of Schaffner and Weissmann (25) and Bradford (26). SDS-polyacrylamide gel electrophoresis, fluorography, and immunoblotting were performed as described by Linder et al. (4). Recombinant G_i1, recombinant G_s, and myristoylated recombinant G_o were purified as described by Lee et al. (20). The ₁₂ subunit complex was synthesized in Sf9 cells and purified as described by Kozasa and Gilman (27).

The following reagents were kindly provided by members of our laboratory: G_z (Tohru Kozasa), hexa-histidine tagged wild-type and mutant G_i1 and -agarose (Christiane Kleuss), hexa-histidine tagged G_q (John Hepler), nonprenylated ₁₂ C68S (Bruce Posner), and antiserum BO87 (Susanne Mumby). E. coli-derived myristoylated FynSh432His₆ was a gift from Dr. Marilyn Resh (14), and Dr. Maurine Linder generously supplied recombinant GAP-43 and hexa-histidine tagged SNAP-25 (purified from E. coli).

RESULTS

Incorporation of Palmitate into G_i1

³H-Label is incorporated into purified myristoylated G_i1 when incubated with [³H]palmitoyl-CoA. To determine the nature and location of this labeling, GTPS-bound G_i1 was incubated with [³H]palmitoyl-CoA and then treated with either 1  hydroxylamine (pH 7.0), 1  Tris (pH 7.0), or trypsin (Fig. 1A). Analysis of these samples by immunoblotting with an antibody directed against the carboxy terminus of the protein indicated that trypsin removed a fragment from the amino terminus, whereas the other treatments had no effect on electrophoretic mobility or immunoreactivity. Fluorography revealed that the incorporated ³H was removed by hydroxylamine, suggesting that G_i1 contained thioesterified [³H]palmitate. Previous studies have demonstrated that trypsin cleaves GTPS-G_i1 at Arg²¹ (28), producing a stable protein core lacking amino acid residues 2 to 21. The only cysteine residue in the cleaved amino-terminal fragment is Cys³. The absence of ³H in the protected tryptic core implies that palmitate is incorporated at Cys³, the site of palmitoylation in vivo.

i1

i1

i1

i1

To confirm palmitoylation of functional G_i1, the GDP-bound protein was incubated with [³H]palmitoyl-CoA and applied to a column of -agarose (Fig. 1B). The resin was washed and then eluted by the addition of buffer containing AlF₄, which causes dissociation of GDP-G_i from . More than 90% of the palmitoylated G_i1 bound to the column and was eluted with AlF₄, indicating that functional protein had been palmitoylated and that thioacylation did not interfere significantly with the capacity of G_i1 to bind to or to be released on activation.

A filtration assay was used to quantify the stoichiometry of palmitoylation. The total incorporation of palmitate reached and maintained a maximal value of 0.8 mol palmitate per mol of G_i1 in roughly 3 h (Fig. 2). The data fit a simple first-order rate equation with a calculated t1/2 of 38 min. Apparent first-order kinetics is presumably due to the substantial excess of palmitoyl-CoA in the reaction. To rule out the possibility that palmitoyl-CoA had been exhausted, the reaction was supplemented with additional [³H]palmitoyl-CoA after 3 h, and incorporation was further monitored (Fig. 2A). To confirm that a dynamic equilibrium of palmitoylated and nonpalmitoylated G_i1 had not been reached, a 10-fold excess of unlabeled palmitoyl-CoA was added to a separate aliquot of the reaction mixture (Fig. 2B). There was no change in the amount of [³H]palmitate incorporated into G_i1 in either case. However, when additional G_i1 was added to the reaction after 180 min, this protein was also palmitoylated at a similar rate (Fig. 2C). Thus, myristoylated G_i1 can be palmitoylated with a stoichiometry of approximately one when palmitoyl-CoA is present as an acyl donor.

i1

i1

i1

Requirements for Autoacylation of G_i1

Many G protein  subunits have a conserved cysteine residue at position three, which is the site of palmitoylation of those proteins. To test whether Cys³ was required for autoacylation of G_i1, we tested the Cys³  Ala mutant of the protein (hexa-histidine tagged at residue 121) after expression in E. coli and purification (Fig. 3A). There was little incorporation of palmitate into this protein, although the hexa-histidine tagged but otherwise wild-type control protein was palmitoylated normally. Gly²  Ala mutants of G_i1 and G_o are not myristoylated in vivo, and these nonmyristoylated proteins also fail to incorporate palmitate in vivo (10). To determine whether myristoylation is required for autoacylation of G_i1, nonmyristoylated G_i1 was also tested. This protein was not autoacylated under the usual reaction conditions (Fig. 3B); it also did not inhibit the palmitoylation of myristoylated G_i1. When myristoylated and nonmyristoylated G_i1 were incubated in the same reaction and then resolved on a urea-containing SDS-polyacrylamide gel (7), fluorography confirmed that [³H]palmitate was incorporated only in the myristoylated protein (data not shown). Finally, the hexa-histidine tagged Gly²  Ala mutant of G_i1, purified from E. coli also expressing protein N-myristoyl transferase, failed to incorporate palmitate in vitro (data not shown). These results demonstrate that autoacylation of G_i1 requires both Cys³ and amino-terminal myristoylation.

i1

i1

i1

i1

i1

Both the GDP- and the GTPS-bound forms of G_i1 can be autoacylated (Fig. 4). The final stoichiometry of palmitoylation was consistently higher for GDP-G_i1 (Fig. 4). When G_i1 was denatured by boiling, autoacylation activity was largely lost (Fig. 4). The reaction was also completely inhibited by the addition of 0.4% SDS (data not shown).

i1

i1

The purified G protein ₁₂ subunit complex was added to G_i1 to assess autoacylation of the heterotrimer. The initial rate of the reaction with G_{i1 12} was four times faster than that with the free myristoylated  subunit (Fig. 5). The final stoichiometry of palmitoylation of the heterotrimer was 10 to 20% greater than that of the free  subunit, but values greater than 1 were never observed (data not shown). The maximal rate enhancement was observed at 1:1 ratios of  to (data not shown). The effect of was more dramatic with nonmyristoylated G_i1. As shown above, nonmyristoylated G_i1 is not autoacylated appreciably, but the rate of palmitoylation of nonmyristoylated G_i1 associated with approximates that of free myristoylated G_i1 (Fig. 5). Nonmyristoylated mutants of some G subunits do incorporate palmitate in cell labeling experiments, although poorly when compared to the wild-type protein (29, 30). In one of these cases, incorporation of palmitate depended on overexpression of (29). All such labeling experiments have been performed in cells that express , consistent with our findings that heterotrimeric nonmyristoylated G_i1 is capable of autoacylation in vitro. When residue 68 of the ₂ subunit is mutated from cysteine to serine, the resulting protein is not prenylated, although it still forms a dimer with ₁. Nonprenylated ₁₂ has a reduced affinity for  (2) and did not enhance the rate of palmitoylation of either myristoylated or nonmyristoylated G_i1 in vitro (data not shown). When palmitoylated heterotrimer was subjected to electrophoresis and fluorography, G_i1 was the only protein in this reaction that had incorporated ³H, ruling out incorporation of palmitate into  or  (data not shown). In addition, there was no autoacylation of G_i1 (C3A)₁₂.

The G protein subunit complex enhances the rate of G

i1

i1

i1

i1

i1

i1

i1

Further Characterization of G_i1 Autoacylation

A plot of initial reaction rate as a function of palmitoyl-CoA concentration revealed a Michaelis-Menten nonlinear relationship (Fig. 6). The apparent K_m for palmitoyl-CoA was 300 to 600 µ. The concentration of detergent micelles in these reactions was about 400 µ; when the concentration of CHAPS in the reaction was raised, there was a coordinate decrease in reaction rate (data not shown). The apparent K_m for this reaction is almost 100 times greater than the cellular concentration of palmitoyl-CoA but is only 10 times greater than total cellular acyl-CoA concentration. Myristoyl-CoA, stearoyl-CoA, and oleoyl-CoA were also tested as substrates for autoacylation. All were roughly comparable to palmitoyl-CoA (data not shown).

i1

i1

i1

i1

The pH dependence of autoacylation of free G_i1 was tested between pH values of 7 and 10, and initial rates increased roughly linearly (Fig. 7A). However, longer time courses also revealed an increase in the final stoichiometry of palmitoylation to values in excess of 1 as the pH was raised (data not shown). The C3A mutant of G_i1 was thus examined at pH 8 and pH 10 (Fig. 7B). The amount of palmitate incorporated into wild-type G_i1 at pH 10 was similar to the sum of palmitate incorporated into the C3A mutant at pH 10 plus palmitate incorporated into the wild-type protein at pH 8.0. At physiological pH, only Cys³ is autoacylated, and the rate of that reaction increases roughly 2.5-fold between pH 7 and pH 8. As the pH is raised further, other cysteine residues clearly become reactive. The rate of autoacylation of G_i1 is also dependent on temperature. The Arrhenius plot is linear between 7 °C and 37 °C, suggesting that palmitoylation occurs by a single reaction mechanism with an activation energy of 3.2 kcal/mol within this temperature range (data not shown).

i1

i1

i1

i1

A large number of cellular proteins are palmitoylated in vivo, and we tested several of these for their capacity to be autoacylated in vitro (Fig. 8A). Bacterially expressed GAP43 and SNAP25 displayed only slow rates of autoacylation. Surprisingly, a bacterially expressed, truncated Fyn kinase (14), which is myristoylated at its amino terminus, also failed to incorporate palmitate. Recombinant G_s (bacterially expressed) and Sf9 cell-derived G_q were also not autoacylated. Both myristoylated G_i1 and myristoylated G_o incorporated palmitate as free  subunits (Fig. 8A), as did recombinant G_z (from Sf9 cells; data not shown). Myristoylated G_i1, myristoylated G_o, and G_s all incorporated palmitate to a significant extent in the presence of , although the rate of autoacylation of heterotrimeric G_s was significantly slower than that of the heterotrimeric myristoylated proteins (Fig. 8B). The rate of palmitoylation of G_q was slow, even in the presence of .

q

i1

o(A)

q

DISCUSSION

Highly purified G_i1 reacts spontaneously with palmitoyl-CoA in vitro to form a thioester between Cys³ of the protein and palmitic acid. Other long chain acyl-CoAs can also serve as acyl donors, and the reaction appears to require native protein structure, as judged by the effects of denaturation with heat or SDS. Although it remains formally possible that these treatments inhibit autoacylation of the protein by other mechanisms (e.g. oxidation and sequestration of substrate, respectively), it is at least proven that native protein is a reactant. The reaction is nearly complete under the conditions used, although the state of the native protein appeared responsible for modest variations in stoichiometry. Thus, the stoichiometry of palmitoylation of GTPS-G_i1 was slightly less than that of the GDP-bound protein. Since GTPS stabilizes G protein  subunits, this observation is counterintuitive. We believe that GTPS-bound G_i1 must be more susceptible to a competing reaction that interferes with autoacylation. Oxidation of Cys³ is a likely candidate, since the autoacylation reaction is performed in the absence of reducing agents. The stoichiometry of autoacylation of heterotrimeric G_i is slightly greater than that of GDP-G_i1; this may be due to slower rates of oxidation and/or denaturation during the autoacylation reaction. Furthermore, the enhanced rate of autoacylation of the heterotrimer simply allows less time for interference from competing reactions.

The palmitoylation of G_i1 probably proceeds by a nucleophilic attack of the Cys³ sulfhydryl on the thioester bond of palmitoyl-CoA, and the deprotonated cysteine residue is likely to be the reactive nucleophile in the reaction. Although the pK_a of the free cysteine sulfhydryl group is roughly 8.5, we believe the microenvironment around Cys³ of G_i1 lowers the pK_a such that it is easily deprotonated at physiological pH. When the pH is elevated above 8, other cysteine residues become deprotonated as well, allowing them to react with palmitoyl-CoA (see Fig. 7).

enhances the rate of autoacylation of G_i1 almost 5-fold. This is not due to palmitoylation at secondary sites, and it requires prenylation of the  subunit. The geranylgeranyl group may facilitate the reaction by forming part of a palmitoyl-CoA binding site or by favoring association of the protein with substrate-containing detergent micelles. Interestingly, abrogates the requirement for amino-terminal myristoylation of G_i1. This is consistent with the observations of Degtyarev et al. (29), who noted that overexpression of permitted palmitoylation of the nonmyristoylated Gly²  Ala mutant of G_o. The crystal structures of heterotrimeric G_i (31) and G_t (32) demonstrate that the prenylated carboxy terminus of  lies very close to the myristoylated amino terminus of , which may permit their functional interaction or substitution.

Although the kinetics of the autoacylation reaction appears to obey the Michaelis-Menton equation, we question the quantitative relevance of these observations. The apparent K_m for palmitoyl-CoA is roughly equal to the concentration of detergent micelles in the reaction. In the other reported cases of autoacylation in which the apparent K_m for palmitoyl-CoA was measured, values of roughly 50 µ were observed and, in both cases, the concentration of detergent micelles was also about 50 µ (16, 18). This suggests that the rate is limited by the fraction of micelles that contains palmitoyl-CoA or that the K_m values depend primarily on protein-detergent interactions or substrate-detergent interactions, rather than on binding interactions between palmitoyl-CoA and G_i1. Quantitative analysis of protein autoacylation cannot be performed readily in the absence of detergent because detergent prevents adsorption of palmitoyl-CoA to both glass and plastic.

Although autoacylation is not unique to G protein  subunits, G_i1 and G_o are clearly more susceptible to the reaction, at least as measured here, than are certain other palmitoylated proteins, such as GAP43 and SNAP25. Perhaps these proteins are missing other requisite covalent modifications or can be autoacylated only in the presence of a hydrophobic protein partner. Of particular interest to us is the generality of autoacylation of G protein  subunits. Members of the G_i subfamily are clear participants, at least as judged by our results with G_i1, G_o, and G_z. Autoacylation of bacterially derived G_s appeared to differ in that it required . However, we have long noted (but failed to explain) substantial differences in the affinities of recombinant and natural G_s for adenylyl cyclases (33) and have speculated that bacteria fail to perform some as yet unidentified covalent modification that characterizes natural G_s. We believe that this modification adds hydrophobic character to the protein and is near its amino terminus,² consistent with the effects of myristoylation of G_i1 on both autoacylation and apparent affinity for adenylyl cyclase. We offer a few hypotheses, largely untested, for our failure to observe autoacylation of G_q. After purification of G_q from Sf9 cells, the protein contains palmitate. Although the stoichiometry is unknown, this may serve as an explanation. However, treatment of purified G_q with palmitoyl thioesterase (34) did not unmask any significant capacity to undergo autoacylation with palmitoyl-CoA. Members of the G_q and G₁₂ subclasses of G proteins contain two or more cysteine residues near their amino termini, and these may undergo rapid disulfide bond formation under the nonreducing reaction conditions used here. Finally, of course, autoacylation of G subunits may simply not apply to these proteins. Study of the possible autoacylation of G_q and G₁₂ family members is hindered both by a relative paucity of material and our inability to synthesize these proteins in bacteria, where endogenous palmitoylation is not an issue.

Dunphy et al. (13) have also studied the palmitoylation of G protein  subunits but did not detect significant autoacylation. Their experiments were performed in Triton X-100 and at pH 6.4, whereas the work described above was performed in CHAPS and at pH 8.0. We have observed significant autoacylation of myristoylated G_i1 in the presence of polyoxyethylene 10 lauryl ether, a nonionic detergent with properties similar to Triton X-100, as well as in the absence of detergent (data not shown). Furthermore, we have also shown that autoacylation is inhibited by lowering the pH, the likely explanation for the observations of Dunphy et al. (13).

It is difficult to assess the biological significance of protein autoacylation. However, we note a provocative level of consistency between our observations in vitro and those based on labeling studies performed in vivo. Myristoylated G_i1 is a much better reactant than its nonmyristoylated counterpart. Autoacylation occurs exclusively at Cys³ of G_i1 at physiological pH. Autoacylation is promiscuous with regard to fatty acid identity; this has been inadequately studied in vivo but is consistent with the few studies reported (8). Although fraught with difficulties because of lack of knowledge of free cellular acyl-CoA concentrations and other factors, the rate of the reaction observed in vitro is similar to the t1/2 for turnover of palmitate on G subunits observed in pulse-chase experiments (10, 11). This would be a necessity for a physiologically relevant reaction, since the rate of depalmitoylation at steady state must equal the rate of palmitoylation.

It is generally accepted that receptor-initiated regulation of G palmitoylation is exerted at the level of depalmitoylation of these proteins (10, 11). The simplest hypothesis is that free  subunits are substrates for a constitutively active palmitoyl thioesterase, and thus, that G proteins are depalmitoylated upon activation. Although we would like to express our own skepticism, our data then open the possibility that palmitoylation of these proteins could then simply proceed nonenzymatically, particularly upon reassociation with . Critical assessment of the physiological significance of G protein autoacylation will require more complete characterization of the enzymatic activities that have been proposed as candidates, followed by their inhibition or deletion in physiological settings. For the moment, we are left to wonder what prevents autoacylation in vivo if it is found not to be significant. Furthermore, we note that autoacylation of at least certain G protein  subunits provides a powerful tool for investigation of the biochemical significance of this lipid modification.