Binding of Regulator of G Protein Signaling (RGS) Proteins to Phospholipid Bilayers

Regulator of G protein signaling (RGS) proteins must bind membranes in an orientation that permits the protein-protein interactions necessary for regulatory activity. RGS4 binds to phospholipid surfaces in a slow, multistep process that leads to maximal GTPase-activating protein (GAP) activity. When RGS4 is added to phospholipid vesicles that contain m2 or m1 muscarinic receptor and Gi, Gz, or Gq, GAP activity increases ∼3-fold over 4 h at 30 °C and more slowly at 20 °C. This increase in GAP activity is preceded by several other events that suggest that, after binding, optimal interaction with G protein and receptor requires reorientation of RGS4 on the membrane surface, a conformational change, or both. Binding of RGS4 is initially reversible but becomes irreversible within 5 min. Onset of irreversibility parallels initial quenching of tryptophan fluorescence (t 1 2 ∼ 30 s). Further quenching occurs after binding has become irreversible (t 1 2 ∼ 6 min) but is complete well before maximal GAP activity is attained. These processes all appear to be energetically driven by the amphipathic N-terminal domain of RGS4 and are accelerated by palmitoylation of cysteine residues in this region. The RGS4 N-terminal domain confers similar membrane binding behavior on the RGS domains of either RGS10 or RGSZ1.

Regulator of G protein signaling (RGS) proteins must bind membranes in an orientation that permits the protein-protein interactions necessary for regulatory activity. RGS4 binds to phospholipid surfaces in a slow, multistep process that leads to maximal GTPase-activating protein (GAP) activity. When RGS4 is added to phospholipid vesicles that contain m2 or m1 muscarinic receptor and G i , G z , or G q , GAP activity increases ϳ3-fold over 4 h at 30°C and more slowly at 20°C. This increase in GAP activity is preceded by several other events that suggest that, after binding, optimal interaction with G protein and receptor requires reorientation of RGS4 on the membrane surface, a conformational change, or both. Binding of RGS4 is initially reversible but becomes irreversible within 5 min. Onset of irreversibility parallels initial quenching of tryptophan fluorescence (t1 ⁄2 ϳ 30 s). Further quenching occurs after binding has become irreversible (t1 ⁄2 ϳ 6 min) but is complete well before maximal GAP activity is attained. These processes all appear to be energetically driven by the amphipathic N-terminal domain of RGS4 and are accelerated by palmitoylation of cysteine residues in this region. The RGS4 Nterminal domain confers similar membrane binding behavior on the RGS domains of either RGS10 or RGSZ1.
The initial events in G protein-mediated signaling take place at the membrane surface. G proteins are peripheral membrane proteins, anchored to the membrane both by intrinsic hydrophobicity and by lipid modifications of their ␣ and ␥ subunits. G protein-coupled receptors, which activate G proteins by accelerating the binding of GTP, span the membrane bilayer and present one relatively hydrophilic face to the cytoplasmic surface. GTPase-activating proteins (GAPs), 1 which accelerate GTP hydrolysis and consequent deactivation, must also operate in this interfacial environment. G protein GAPs include at least two families of proteins, the phosopholipase C-␤s and the RGS proteins (including the RGS-related proteins such as p115 Rho-GEF and G protein-coupled receptor kinases) (1). Each group displays different modes of membrane binding. Phospholipase C-␤, which is both a G q -stimulated signaling protein and a G q GAP, is soluble in aqueous solution but binds to membranes by nonspecific ionic interactions, with contribution from a plekstrin homology domain that can bind to phosphatidylinositol 3,4-bisphosphate (2). The receptor kinases are also soluble and apparently do not bind tightly to phospholipid bilayers but rather are recruited by a combination of palmitoylation and by binding both their receptor substrates and G␤␥ subunits (see Ref. 3 for review).
The membrane binding behavior of RGS proteins is more diverse. Although RGS proteins do not contain obvious membrane-spanning domains, many are tightly membrane-associated and behave as hydrophobic molecules. At the extreme, RGSZ1 2 is tightly membrane-bound and can be extracted from natural membranes under non-denaturing conditions only using high concentrations of Triton X-100 at elevated temperatures (4). Members of the RGSZ family (GAIP, RGSZ1, RGSZ2, Ret-RGS) also contain cysteine strings that are probable sites for multiple palmitoylation (5), but RGSZ1 expressed in Escherichia coli is hydrophobic and aggregates in the absence of detergent even after exposure to 20 mM dithiothreitol, which removes thioesterified palmitate (6,7). RGSZ1 tends to form oligomers even on SDS-polyacrylamide gels. At the other extreme, RGS10 is a soluble protein found in the cytoplasm (7, but see Ref. 8). RGS3, RGS4, and RGS16 display intermediate behavior. They are found both in cytoplasm and in membrane fractions, their localization is influenced by reversible palmitoylation at two or more sites, and their N-terminal sequences can form an amphipathic helix that directs binding to anionic lipid surfaces (7, 9 -11). In addition to this behavioral heterogeneity, the specific binding sites for individual RGS proteins on membranes (lipids, G␣ or G␤␥ subunits, receptors, or other proteins) remain unknown in most cases.
The GAP activities of RGS proteins are markedly influenced by their interactions with membranes and, presumably, by the ordering influence that membrane surfaces provide. GAP activity can be readily and accurately measured in detergent solution under conditions where substoichiometric GAP promotes hydrolysis of preformed G␣-GTP complexes (12). However, low GAP activities are more sensitively measured by monitoring stimulation of steady-state GTPase activity in reconstituted phospholipid vesicles that contain trimeric G protein and an appropriate agonist-bound receptor (12). The enhanced GAP activity observed in this system may reflect interaction of the GAP with the lipid bilayer. Removal of the N-terminal region of RGS4 diminishes its activity in the membrane-based assay but not in detergent solution (9,10,13). RGSZ1, which is significantly more hydrophobic than RGS4, does not incorporate into membranes from solution and is therefore essentially inactive in the membrane-based assay unless it is incorporated into the vesicle bilayer during the reconstitution process (6).
We report here that RGS protein must be able to bind membranes with correct orientation to display optimal GAP activity. In the case of RGS4, binding to the bilayer from solution and reorientation are multiphasic phenomena that become essentially irreversible over a few minutes but that goes to completion over at least 60 min at 30°C. The low intrinsic GAP activity of RGS10 in the vesicle-based assay can be markedly enhanced by the addition of the N-terminal amphipathic domain from RGS4 while maintaining the G␣ selectivity characteristic of RGS10.
cDNA Constructs-cDNA constructs used for expression of wild-type RGS proteins (6), for the RGS box domains of RGS10 and RGS4 (18), and for RGS4⌬N57 (7) have been described. cDNAs used for expression of other truncated and chimeric RGS proteins were constructed in pQE60 (Qiagen) by polymerase chain reaction using as templates the cDNAs for human RGSZ1 (6), rat RGS4 (19), and human RGS10 (18). Receptor-G Protein Vesicles and Protein-free Liposomes-Heterotrimeric G proteins and m2AChR, with or without added GAPs, were reconstituted in to phospholipid vesicles essentially as described previously (15,16). Lipids (25 g; PE:PS:CHS, 55:35:10) were suspended in 25 l of 20 mM NaHepes (pH 8.0), 0.1 M NaCl, 1 mM EDTA, 1 mM MgCl 2 , 0.2% sodium deoxycholate, 0.02% sodium cholate and sonicated until translucent. The suspension was then mixed with G protein, m2AChR, and GAP (when used) to yield a final volume of 50 l. This mixture was then chromatographed on a 3 ϫ 150-mm column of Sephacryl AcA34 in the resuspension buffer, but without detergent. The usual fraction volume was 75 l.
To prepare small, unilamellar vesicles, lipids were dissolved in chloroform (10 mg/ml) at an appropriate molar ratio. Solvent was evaporated under argon to form a thin lipid film, and buffer A (50 mM NaHepes (pH 8.0), 50 mM NaCl, 2 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol) was added to yield the desired concentration. The lipid dispersion was vortexed and then sonicated under argon at room temperature for 30 min until an almost clear solution was obtained. Liposomes were analyzed by gel filtration under the same conditions used to prepare protein-lipid vesicles.
Fluorescence Measurements-RGS4 or RGS10 (0.5 M) was diluted in 300 l of 50 mM NaHepes (pH 7.5) plus 1 mM dithiothreitol in the presence or absence of 100 M sonicated lipid vesicles at 25°C. After 10 min of incubation, fluorescence was recorded either at 328 nm or between 305 and 405 nm with excitation at 285 nm in a Hitachi F-2000 fluorescence spectrophotometer.
GAP Assays-GAP activity was assayed in two formats. In "single turnover" assays, G␣ is first bound to [␥-32 P]GTP, and the rate of hydrolysis of [␥-32 P]GTP-G␣ is measured in a solution that contains 25 mM NaHepes (pH 7.5), 1 mM EDTA, 6.8 mM MgCl 2 (1 mM free Mg 2ϩ ; see Ref. 6), 1 mM dithiothreitol, 0.1% Triton X-100, 5 mM GTP (12). Activity is calculated as the increase in the rate constant for hydrolysis of the bound [␥-32 P]GTP. Single turnover assays, using either 1.5 nM [␥-32 P]GTP-G␣ z at 15°C or 10 nM G␣ i1 at 8°C, were performed as described (12). A more sensitive and presumably more physiological assay for GAP activity monitors the enhancement of agonist-stimulated, steady-state GTPase activity in proteoliposomes reconstituted with receptor and heterotrimeric G proteins. Reconstitution of purified m2AChR with either G z or G i and of m1AChR with G q were performed as described (6,12,15). RGS proteins were either coreconstituted into the vesicles by mixing with the detergent-dispersed receptor, G protein, and lipid prior to formation of the vesicles or added to the vesicles in detergent-free solution. Added RGS proteins were incubated with the vesicles for 1 h at 30°C prior to assay unless otherwise indicated. Steady-state assays were carried out at 30°C for 5-10 min. Data are given as increases in the steady-state GTPase activity. Interpretation of GAP assay data has been discussed elsewhere (12).
Other-Standard procedures were used for SDS-polyacrylamide gel electrophoresis (20) and staining with Coomassie Blue or silver (21). Protein was measured by Amido Black binding (22).

RESULTS
Several experimental findings initially suggested that the GAP activity of RGS proteins depends on the extent and mode with which they bind to phospholipid membranes. First, when RGSZ1 was assayed for its ability to increase agonist-stimulated, steady-state GTPase activity in receptor-G protein proteoliposomes, it was essentially inactive when added to the vesicles in buffer; it was ϳ200-fold more active when inserted into the nascent bilayer along with G protein and receptor during formation of the vesicles (Ref. 6 and Table I). GAIP, a close homolog of RGSZ1, behaved similarly (data not shown). In contrast, RGS4 and phospholipase C-␤, a non-RGS GAP, are active when added from solution (6,15). A third pattern was displayed by RGS10. It was relatively ineffective in the vesiclebased assay either when added from solution or when coreconstituted with the other proteins and lipids, even though its GAP activity with soluble G␣ i -GTP substrate is not much different from that of the other two RGS proteins ( Fig. 1, Table I).
One explanation for the low activity of RGS10 in the vesiclebased assay is that it neither incorporates into nor binds tightly to phospholipid vesicles. To test this idea, stable incorporation of RGS proteins was monitored during preparation of receptor-G protein vesicles by gel filtration. When RGS4 and RGS10 were mixed with receptor, G protein, and phospholipids prior to reconstitution, only about 5% of the RGS10 eluted with the vesicle peak (Fig. 1). In contrast, about 80% of RGS4 coeluted with the vesicle fraction. RGSZ1 and GAIP also coeluted with TABLE I GAP activities measured in detergent solution and in phospholipid vesicles The GAP activity of each RGS protein was measured in detergent solution ("micelle") in a single-turnover assay according to the increase in the rate of hydrolysis of G␣-[␥-32 P]GTP ("Experimental Procedures"). Assays were performed at 15°C for G␣ z -GTP (1.5 nM) or 8°C for G␣ il -GTP (10 nM). GAP activity is expressed as an increase in the hydrolytic rate constant (min Ϫ1 ) per pmol of RGS protein. GAP activity was also measured according to the stimulation of steady-state GTPase activity of phospholipid vesicles that contained m2AChR and trimeric G protein ("vesicle"). RGS proteins were added to the vesicles just before initiation of the assay. Steady-state GAP activity, measured at 30°C, is expressed as the increase in the GTPase turnover number (min Ϫ1 ) per pmol of RGS protein. Turnover numbers for m2-G z vesicles were 0.11 and 1.3 in the absence and presence, respectively, of 10 nM RGS4 vesicles and 1.0 and 8.7 for m2-G i vesicles in the absence or presence, respectively, of 5 nM RGS4. Concentrations of GAIP and RGSZ1 were varied between 2 and 16 M to obtain quantifiable stimulation. Vesicles used in these assays contained 1 nM G z and 0.26 nM m2AChR or 1 nM G i and 0.25 nM m2AChR. Note that GAP activities for G i and G z cannot be compared directly because the intrinsic GTPase activities of the two proteins (and their interactions with receptor) vary markedly. Interpretation of these assays has been discussed elsewhere (12 Fig. 2A, RGS4 spontaneously binds to protein-free phospholipid vesicles when added after their formation (see also Ref. 9). Association is stable over the time course of gel filtration, suggesting that the affinity of RGS4 for phospholipids is high. In contrast, RGS10 showed no evidence of stable binding to preformed phospholipid vesicles (Fig. 2). These data support the idea that the major determinant of the low GAP activity of RGS10 in vesicle-based assays is its inability to bind to the vesicle surface. This idea is supported by our previous observation that palmitoylation of RGS10 at Cys 66 markedly increases its GAP activity in vesicle-based assays (7) and allows it to bind to phospholipid vesicles with about the same efficiency as seen for RGS4 (data not shown). The importance of membrane binding is emphasized by the fact that palmitoylation of RGS10 at Cys 66 increases net GAP activity in the vesicle-based assay even though it inhibits the intrinsic GAP activity of the RGS10 protein (7).
Although the binding of RGS4 to lipid vesicles appears to be necessary for efficient GAP activity, a second and slower event is apparently also required. When RGS4 is added to receptor-G protein vesicles before a steady-state GTPase assay, some GAP activity is displayed immediately, but prolonged incubation of RGS4 with vesicles before assay further increases GAP activity up to 3-fold (Fig. 3). The rate of activation depends on temperature. Full activation required several hours at 30°C (Fig. 3) but was not complete even after 10 h at 20°C (data not shown). The extent of the increase in GAP activity was not altered by temperature, however. Slow activation of GAP activity is not limited to G i . A similar slow increase in GAP activity of RGS4 was observed with vesicles that contained G z . The activation process was not influenced by the presence of agonist. The increased in the GAP activity of RGS4 was accelerated by N-terminal palmitoylation, measured using the C95V mutant to avoid inhibiting GAP activity (see Ref. 7). Palmitoylated C95V RGS4 displayed initially high GAP activity when assayed with m2AChR-G i vesicles, and activity did not increase with prolonged incubation. RGS10 did not display a time-dependent increase in GAP activity even when added at a high enough concentration to produce an easily measured effect (Fig. 3).
Functional activation of RGS4 upon vesicle binding is preceded by a slow change in either its conformation or orientation, as determined by quenching of its intrinsic tryptophan fluorescence (Fig. 2, B and C). Phospholipid binding causes ϳ30% quenching in the fluorescence of RGS4 and a slight blue shift in the emission maximum. This process is biphasic; about 40% of the observed quenching takes place rapidly (t1 ⁄2 ϳ 30 s at room temperature), and the remainder occurs over about 30 min (Fig. 2C). We do not know the physical basis of the observed quenching, which would not be predicted if the two tryptophan residues in RGS4 were simply inserted into the hydrophobic environment of the membrane. Both tryptophan residues, Trp 59 and Trp 92 , are in the conserved RGS GAP domain, and neither faces the exterior of the molecule (23). Trp 59 is almost completely buried among helices 1, 2, 3 and 9; Trp 92 lies in the cleft between the two bundles of helices opposite the site of G␣ binding (23). It seems most likely therefore that fluorescence quenching observed upon binding to a phospholipid bilayer reflects a conformational change of the RGS4 protein, probably in the environment of Trp 92 . The fluorescence of RGS10 was also quenched slightly upon exposure to lipid vesicles (ϳ10%; Fig. 2), probably reflecting the same physical phenomenon observed for RGS4, but with weaker binding.
Based on the results above, we used competition for RGS binding between receptor-G protein vesicles and protein-free liposomes to study the lipid binding behavior of RGS4 in greater detail. Mixing RGS4 with receptor-G protein vesicles and empty vesicles causes net inhibition of GAP activity (Fig.  4). Inhibition apparently reflects competition for RGS4 between the two populations of vesicles, because GAP activity is inhibited half-maximally when the concentration of both vesicles is equal (Fig. 4A). Maximal inhibition is essentially complete at high concentrations of added liposomes, as is also consistent with competitive binding of RGS4 to available lipid surfaces. Binding of RGS4 to lipid vesicles depends significantly on the presence of anionic lipids (Fig. 4B; see also Ref. 9). Although we have not attempted to analyze the selectivity of RGS4 among phospholipids in any detail, individual or mixed neutral lipids such as PE (or phosphatidylcholine; data not shown) do not by themselves bind RGS4. Phosphatidylglycerol and phosphatidic acid are approximately as effective, as is PS. A dispersion of cholesteryl hemisuccinate alone also does not bind RGS4 tightly (Fig. 4B).
As suggested by the gel filtration experiments (Figs. 1 and 2), binding of RGS4 is at best poorly reversible. If RGS4 is first allowed to bind to receptor-G protein vesicles, subsequent addition of liposomes decreased GAP activity only slightly (Fig. 5, upper curve). Whereas binding appeared to be complete by 5 min, stabilization of binding was incomplete at 15 s. Adding liposomes at this time produced almost as much inhibition as did mixing all three components simultaneously. In the converse experiment, prior exposure of RGS4 to liposomes before addition of receptor-G protein vesicles increased inhibition over the same 5-min period (Fig. 5, lower curve). In the experiments shown in Fig. 5, mixtures of receptor-G protein vesicles and liposomes were held only 1 min before GAP assay, and there would be little opportunity for exchange of RGS4 among vesicles. To test for possible exchange, we also incubated liposomes for up to 60 min with added vesicles that had been pre-equilibrated with RGS4. The extended incubation did not further inhibit GAP-stimulated GTPase activity (data not shown). These data confirm the idea that once RGS4 binds to an anionic lipid bilayer there is an initially freely reversible interaction that is stabilized over the course of a few minutes. This stabi- lization correlates with the rapid phase of quenching of Trp fluorescence and precedes the conformational or orientational event that increases GAP activity.
The N-terminal region of members of the RGS4 family (RGS1-4, 16) is important for maintaining both correct subcellular localization (7, 9 -11) and cellular interactions with receptors and G proteins (13,24). This region, usually defined as the first 53-57 amino acid residues for RGS4, contains sites for covalent palmitoylation (7,11) and behaves as an amphipathic ␣ helix (9). To study the contribution of the N-terminal region of RGS4 to GAP activity in vesicle-based assays, we prepared and assayed several truncated and chimeric RGS proteins with distinct N-terminal domains. As shown in Fig. 6, the N-terminal domain of RGS4 is both necessary and sufficient for most of the interactions with membrane lipids described above for RGS4. Replacement of the N-terminal domain of RGS10 with that of RGS4 increased its potency as a G q GAP by about 1000-fold in the receptor-coupled, vesicle-based assay relative either to RGS10 itself or the RGS10 box. This effect can be accounted for by the ability of the RGS4:10 chimera to associate stably with receptor-G protein vesicles (Fig. 6B). RGS4:Z bound to liposomes, as did RGS4 or RGS4:10, and incubation of these chimeras with the vesicles increased GAP activity, as was seen for RGS4 itself (Fig. 3). RGS10:4 did not bind vesicles, and RGSZ:4 aggregated in aqueous solution such that vesicle binding could not be measured (data not shown). Each case mimicked the behavior of the RGS protein that donated the N terminus.
The experiments shown in Fig. 6 were performed with vesicles that contained G q and m1AChR, instead of m2AChR and G i or G z , to demonstrate the generality of the effect of the N-terminal domain and to confirm a brief report that truncation of the N terminus of RGS4 decreases G q GAP activity (13). However, similar results were obtained with the m2AChR and either G i or G z (Table II). Truncation of the N-terminal domain of RGS4 resulted in loss of GAP activity in the vesicle-based, steady-state assay with either G protein as target. Conversely, activity was significantly restored when the N-terminal region of RGS4 replaced that of RGS10, which does not drive bilayer attachment, or that of RGSZ1, which causes aggregation in aqueous buffer. GAP activities of intact RGS10, the isolated RGS domain of either RGS10 or RGSZ1, or an N-terminally truncated RGS4 were all similarly low. For RGS10, which is already relatively inactive in this assay (Fig. 3, Table I), removal of the short N-and C-terminal domains had little effect. Similarly, replacement of the N-terminal domain of RGS4 with that of RGS10 did not support activity in this assay.
To confirm that manipulation of the N termini of the RGS proteins had not inactivated them in some way, we also assayed the GAP activities of the truncated and chimeric proteins in single turnover assays in detergent solution (Table III). As suggested by the work of Popov et al. (18), alteration of the RGS4 N-terminal domain in the constructs shown here had little effect on their activities with G␣ i -[␥-32 P]GTP as substrate. Structural determinants of G z GAP activity were more complex and indicate that the N-terminal regions of RGS4 contribute to interaction with G␣ z . Full-length RGS4 displayed FIG. 5. Reversibility of RGS4 binding to liposomes or receptor-G protein vesicles. In one experiment (q), RGS4 was mixed with m2AChR-G i vesicles (ϳ2 M lipid) at 30°C, and protein-free liposomes (2.5 M lipid) were added at the times shown. After 1 min of further incubation, carbachol-stimulated GTPase activity was assayed over 5 min. Control experiments (data not shown) indicated that prolonged incubation (1-60 min) of liposomes with the vesicles and RGS4 before assay produced no greater inhibition than shown here. In the complimentary experiment (OE), RGS4 was mixed with liposomes at 30°C, and m2AChR-G i vesicles were added at the times shown, followed by GTPase assay 1 min later. The initial time points are 15 s, not zero. To correct for the increase in GAP activity as RGS4 is incubated with receptor-G protein vesicles (see Fig. 3), activities are normalized as percentages of activities in similar preparations that were mixed with buffer rather than with liposomes.
FIG. 6. The RGS4 N-terminal domain potentiates GAP activity and bilayer binding of RGS10. A, carbachol-stimulated GTPase activity of m1AChR-G q vesicles was assayed in the presence of increasing concentrations of RGS4 (q) or RGS4:10 (f). Vesicles and RGS protein were pre-incubated for 60 min at 30°C before assay. For reference, GTPase activity with no added RGS protein was 10 fmol/min, and activity with 16 M RGS10 was 70 fmol/min. Each assay contained 0.5 nM G q and 0.085 nM m1AChR. B, RGS4, RGS10, or RGS4:10 (1 M) were incubated with PE:PS:CHS liposomes for 10 min and analyzed by gel filtration on Sephacryl AcA34. Fractions were assayed for G z GAP activity in a solution-phase, single turnover assay. far more GAP activity toward G␣ z -GTP than did RGS4⌬N57 (or RGS10), indicating the importance of the N-terminal region in addition to the RGS domain. The RGS4 N-terminal domain thus contributes to interaction with G z , as is true for G q (13). Consistent with this positive role, the RGS4 N-terminal domain in the RGS4:10 chimera potentiated the G␣ z GAP activity of RGS10, which has slight activity with or without its Nterminal region. The RGS4:10 construct was reproducibly more active than RGS4. Replacement of the N-terminal domain of RGSZ1 with that of RGS4 had relatively little effect. Indeed, RGSZ1⌬N82, which lacks the entire N-terminal region up to the RGS, retained substantial activity in the solution-based assay. Thus, some interaction of RGS proteins with G␣ z is contributed by the N-terminal domain as well as by the RGS box. This situation is qualitatively similar to that described for G␣ q , where removal of the N-terminal region of RGS4 also decreased activity in solution (13). DISCUSSION To modulate a membrane-bound G protein signaling pathway, an RGS protein must act at the surface of a cellular membrane, but patterns of membrane attachment vary widely among RGS proteins. Some RGS proteins are hydrophobic and essentially integral membrane proteins (6). For others, such as RGS10, membrane binding has been hard to demonstrate convincingly except when the protein is palmitoylated (7). RGS4 also behaves as a soluble protein when purified but clearly binds to cellular membranes in a process that is dependent on its N-terminal domain and that is apparently driven by direct binding to anionic lipids (Ref. 9 and Fig. 4B). Although RGS4 is also palmitoylated, palmitoylation is not required for its binding to natural membranes or phospholipid bilayers (6,7,9,11).
RGS4 binds tightly to receptor-G protein vesicles or to protein-free liposomes such that, after a few minutes, binding is essentially irreversible. Vesicle binding is not appreciably influenced by receptor or G protein, because receptor-G protein vesicles and protein-free liposomes bind RGS4 at similar rates (Fig. 5) and compete equally for limiting RGS4 (Fig. 4). The interaction of RGS4 with membranes consists of several steps that occur after initial contact. Quenching of Trp fluorescence is observed essentially immediately upon mixing RGS4 with liposomes and occurs with a t1 ⁄2 of about 30 s at room temperature. Additional quenching occurs more slowly and roughly correlates with the loss of reversibility of binding. Onset of irreversibility of binding was measured as the ability of protein-free liposomes to sequester RGS4 and thus inhibit its interaction with subsequently added receptor-G protein vesicles. The coelution of RGS4 with lipid vesicles during gel filtration also supports the idea that binding becomes poorly reversible.
The most interesting of these slow phenomena is the timeand temperature-dependent increase in GAP activity that is observed after initial binding of RGS4 to receptor-G protein vesicles. Increased GAP activity kinetically follows both irreversible binding and the initial quenching of Trp fluorescence (Fig. 2C). Activation is apparently dependent on these prior events, because N-terminal palmitoylation increases the rates of onset of all the binding-related phenomena described. Slow activation may result from a change in the orientation of RGS4 with respect to the bilayer or from a change in its conformation. Either effect could be consistent with the slow increase in fluorescence quenching. Enhanced GAP activity might also result from slow association with receptor or G protein. In the latter case, binding could be either to G␣ or G␤␥. G␤␥ interacts functionally with RGS proteins (6,25), and they appear to bind directly to each other (26). 3 We consider this possibility less likely, because the activation process remains even when high concentrations of RGS4 are added to the vesicles. For this same reason, slow activation does not reflect redistribution of RGS4 among vesicles or from an aggregated state to vesicles.
Essentially all of the membrane association behaviors described here for RGS4 are dependent on the contribution to net lipid binding of the N-terminal region. This region binds directly to anionic lipid bilayers, apparently as an amphipathic ␣ helix (9), and contains sites for palmitoylation that further increase hydrophobicity (7,11). Based on structural similarity and the behavior of RGS16 (10), this lipid binding function of the N-terminal domain is probably conserved through the RGS4 subfamily. This domain appears to function autonomously for lipid binding (9) and functions independently of the RGS domain to which it is attached. The RGS4:10 and RGS4:Z chimeras also bound to lipid vesicles, became irreversibly bound, and underwent slow activation in a manner similar to that of RGS4 (data not shown). The RGS4 N terminus also functioned in the context of the m2AChR and either G i or G z or of the m1AChR and G q . Taken together, these data suggest that the N-terminal region of RGS4 acts primarily to support 3 Y. Tu, unpublished observation.

TABLE II
GAP activity of truncated and chimeric RGS proteins in m2AChR-G protein vesicle RGS proteins were added into proteoliposomes that contained m2AChR and either G z or G i and incubated for 60 min at 30°C. Vesicles were then assayed for GTPase activity in the presence of 1 mM carbachol. Because activities varied dramatically, different concentrations of RGS proteins were used, and relative GAP activity is shown as the GTPase molar turnover number divided by the concentration of RGS protein in the assay (20 nM for RGS4, 100 nM for RGS4:10 and RGS4:Z1, 2 M for others). Vesicles contained 0.8 -1.0 nM Gz and 0.25-0.33 nM m2AChR or 1 nM G i and 0.29 nM m2AChR. Data shown are averages from two separate experiments performed using a single batch of G i vesicles and two separate batches of G z vesicles. adsorption to the bilayer, the first step in positioning a GAP for regulating the receptor-stimulated, steady-state GTPase reaction. The N-terminal domain may also be involved with subsequent reorientation of RGS4, but such an effect cannot readily be distinguished. Independent of its effect on lipid binding, the RGS4 N-terminal domain also contributes significantly to its GAP activity with G␣ z and G␣ q , although not with G␣ i . Removal of the N terminus does not interfere with G i GAP activity (Ref. 18; confirmed in Table III), but N-terminal truncation markedly reduces activity with G␣ z (Table III) and G␣ q (13). This probably reflects direct interaction between the N terminus and G␣ z or G␣ q , because addition of the RGS4 N-terminal domain to N-terminally truncated RGS10 yielded a chimeric protein, RGS4:10, that displayed about 20-fold more G z GAP activity than did intact RGS10.
The data presented here point to the complexity of the process whereby RGS proteins must become oriented in their membrane environment to display optimal GAP activity in a receptor-coupled system. They are consistent with previous results that indicated that the amphipathic and cationic RGS4 N terminus and its palmitoylation are important to the energetics of membrane binding (Ref. 9; see also Refs. 27 and 28) but argue that proper orientation or alignment with receptor or G protein subunits or both is a necessary subsequent step. This conclusion agrees with the idea that relatively nonspecific hydrophobic or ionic interactions are frequently the major energetic components of binding of peripheral membrane proteins and that protein-protein binding is needed for specificity (27). The energetic contribution of lipid binding to affinity may in part explain the very high affinities of binding of RGS proteins to multiple partners recently described by Dowal et al. (26). We still need to understand how peripheral domains of RGS proteins contribute to these specific interactions, and the ability of the RGS4 core domain to distinguish G i from G z or G q may help us answer the question. The N-terminal regions may be generally used by RGS proteins to determine the mode of membrane attachment, as seems true for the RGS4 and RGSZ families. In addition, though, the presence of diverse N-terminal functional domains in RGS proteins indicates that membrane binding is only one of many roles for this region (1,29,30). The present work also leads to the next question of how RGS proteins, which are naturally expressed at levels below those of their G targets, are themselves directed to the specialized sites of receptor-G protein signaling.