The binding of activated Gαq to phospholipase C-β exhibits anomalous affinity

Upon activation by the Gq family of Gα subunits, Gβγ subunits, and some Rho family GTPases, phospholipase C-β (PLC-β) isoforms hydrolyze phosphatidylinositol 4,5-bisphosphate to the second messengers inositol 1,4,5-trisphosphate and diacylglycerol. PLC-β isoforms also function as GTPase-activating proteins, potentiating Gq deactivation. To elucidate the mechanism of this mutual regulation, we measured the thermodynamics and kinetics of PLC-β3 binding to Gαq. FRET and fluorescence correlation spectroscopy, two physically distinct methods, both yielded Kd values of about 200 nm for PLC-β3–Gαq binding. This Kd is 50–100 times greater than the EC50 for Gαq-mediated PLC-β3 activation and for the Gαq GTPase-activating protein activity of PLC-β. The measured Kd was not altered either by the presence of phospholipid vesicles, phosphatidylinositol 4,5-bisphosphate and Ca2+, or by the identity of the fluorescent labels. FRET-based kinetic measurements were also consistent with a Kd of 200 nm. We determined that PLC-β3 hysteresis, whereby PLC-β3 remains active for some time following either Gαq–PLC-β3 dissociation or PLC-β3–potentiated Gαq deactivation, is not sufficient to explain the observed discrepancy between EC50 and Kd. These results indicate that the mechanism by which Gαq and PLC-β3 mutually regulate each other is far more complex than a simple, two-state allosteric model and instead is probably kinetically determined.

In animals, phospholipase C (PLC) 2 hydrolyzes phosphatidylinositol 4,5-bisphosphate (PIP 2 ) to inositol 1,4,5-trisphosphate and diacylglycerol, two important second messengers that regulate diverse cellular processes in response to multiple extra-cellular stimuli that act through diverse mechanisms (1). The four PLC-␤ family members are stimulated by the G q family of G␣ subunits, G␤␥ subunits, and Rac GTPases, and PLC-␤ function is consequently central to numerous signaling pathways in all animal cells (2). PLC-␤ also contributes directly to the temporal dynamics of its activation because it is a GTPase-activating protein (GAP) for G␣ q and can accelerate deactivation of G␣ q -GTP Ͼ1000-fold (3).
G␣ q , G␤␥, and Rac all bind independently to PLC-␤s to cause activation, although G␤␥ does not activate PLC-␤4 (4), and only PLC-␤2 is stimulated by Rac (5,6). The activity of PLC-␤3 can be simulated about 400-fold by G␣ q (7). G␣ q and G␤␥ activate PLC-␤3 synergistically, and relative activation can exceed 2000-fold (7). Despite the availability of several X-ray crystal structures of G␣ q , G␤␥, PLC-␤3, and the PLC-␤3-G␣ q complex (8 -13) plus considerable enzymologic studies (3, 7, 10, 12, 14 -21), the details of this activation process remain unclear. Activation requires both binding of PLC-␤ to the surface of the membrane that contains the PIP 2 substrate and the movement of an autoinhibitory loop away from the active site (10,22). Additional intramolecular events may also be involved. Sondek and co-workers (10,22) suggested that G␣ q anchors PLC-␤ to the membrane surface and that the anionic autoinhibitory region is forced away from the active site simply by charge repulsion between this region and the membrane. Both proteins bind to distinct sites on PLC-␤ and could force the PLC-␤ against the bilayer, although the geometry of this interaction is uncertain (8, 10 -12, 20, 22). In addition, Lyon et al. (11) pointed out that binding to G␣ q may disrupt intramolecular contacts in PLC-␤ that, in the absence of activator, stabilize the autoinhibitory region over the active site. The role of the C-terminal helical domain, which is essentially required for activation by G␣ q but not by G␤␥, is not understood (8,12,14,20); nor is the way that movement of the PH domain during G␤␥ binding leads to activation (19).
At a descriptive level, G␣ q has been assumed to activate PLC-␤ by a simple allosteric mechanism in which G␣ q binding drives a structural isomerization of the PLC to a more active conformation or orientation with respect to the bilayer surface (2,23). In the simplest case of allosteric activation, the EC 50 for activating an enzyme by a ligand is equal to the equilibrium dissociation constant for ligand binding, K d , although various mechanisms can cause deviation from this pattern (24). GTPactivated G␣ q stimulates PLC-␤s with EC 50 values of ϳ1-5 nM (7,11,16). These values agree with the EC 50 for the G q GAP activity of PLC-␤, the ability of PLC-␤ to accelerate hydrolysis of G␣ q -bound GTP (3,10,16). Therefore, we expected the K d for PLC-␤3 binding to G␣ q to be in the 1-5 nM range. In fact, the first reported K d values for G␣ q and PLC-␤ binding were in the low nanomolar range, consistent with the EC 50 (25), although Waldo et al. (10) reported a K d for PLC-␤3 binding to G␣ q of about 200 nM.
To further probe the mechanism of PLC-␤ activation, we describe here two physically different but complementary measurements of the binding of GTP␥S-activated G␣ q to PLC-␤3. One is based on intermolecular FRET between the two proteins, and the other is based on fluorescence correlation spectroscopy (FCS) measurements of the decrease in G␣ q diffusion when it is bound to PLC-␤3. Both methods yielded values for K d of about 200 nM, which is 50 -100 times greater than the EC 50 for activation of PLC-␤3 by G␣ q or for acceleration by PLC-␤ of hydrolysis of G␣ q -bound GTP. The FRET assay allowed us to perform multiple controls for the validity of the determination and to show that binding affinity is relatively insensitive to minor variation in experimental conditions. This disagreement in EC 50 and K d is not explained by an essential activator mechanism (24) or by the simplest hysteretic models. Our data suggest that activation of PLC-␤ by G␣ q in the low nanomolar range is not mediated by the thermodynamically most stable state but rather reflects a fast and transient interaction of the two proteins on the bilayer surface.

Results
To study the mechanism of PLC-␤ activation by G␣ q and the G q GAP activity of PLC-␤, we measured the equilibrium binding of the two proteins and their rates of association and dissociation in comparison with their functional interactions.

Equilibrium G␣ q -PLC-␤3 binding measured by FRET
We developed a FRET-based binding assay to measure the thermodynamics and kinetics of PLC-␤3 binding to G␣ q . Both proteins were mutated to remove endogenous reactive cysteine residues and labeled at individual novel cysteine residues with donor or acceptor fluorophores. Several such FRET pairs were tested for energy transfer (see below). As an example, we first describe binding measured using a single G␣ q -PLC-␤3 FRET pair in which cysteine residues introduced in place of Glu 249 in G␣ q and Gln 871 in PLC-␤3 were labeled with Alexa Fluor 488 and 594 maleimide, respectively (see "Experimental procedures"). These labeled proteins are referred to as G␣ q -F and PLC-␤3-Q. Both proteins behave identically to wild-type, unlabeled proteins with respect to catalytic and regulatory activities (see Fig. 2). Other fluorescently labeled proteins, also checked for catalytic and regulatory activities, are described below.
When GTP␥S-bound G␣ q -F (G␣ q -F-GTP␥S) was mixed with PLC-␤3-Q, donor fluorescence was quenched, and acceptor fluorescence was increased in parallel (Fig. 1A), although FRET-driven acceptor emission is only obvious at low PLC-␤3-Q concentrations because it is masked by direct excitation. Binding of G␣ q -F to PLC-␤3-Q saturated at about 50% donor quenching when increasing amounts of PLC-␤3-Q were added to a fixed concentration of fully activated G␣ q -F-GTP␥S ( Fig. 1, B and C). Incomplete G␣ q -F activation decreased maximum quenching but did not alter affinity of binding. Donor quenching was fit to a single-site binding equation to yield the equilibrium dissociation constant, K d . Under standard phospholipase assay conditions with buffer that includes unilamellar 1-stearoyl-2-arachidonoyl-phosphatidylethanolamine (PE)/ phosphatidylserine (PS)/PIP 2 (16:4:1; 262 M total) vesicles and 60 nM free Ca 2ϩ , the K d for binding was 160 Ϯ 50 nM (n ϭ 2) (Fig. 1B). Binding was also measured under a variety of other conditions without much change in affinity (Table 1). Without Ca 2ϩ , the K d was 242 Ϯ 18 nM (n ϭ 3) (Fig. 1C); without PIP 2 but with Ca 2ϩ , the K d was 280 Ϯ 30 nM (supplemental Fig. S1A) (n ϭ 2). Phospholipid vesicles are not required for G␣ q -PLC-␤3 binding (supplemental Fig. S1, B and C). In 0.2% cholate, the K d was 140 Ϯ 30 nM (supplemental Fig. S1B), and without detergent or phospholipid, the K d was 110 Ϯ 20 nM (supplemental Fig. S1C). These values of K d , all about 200 nM, are not significantly different at a level of p ϭ 0.08 (lowest p value of 10 pairwise t tests). Hence, regardless of whether the PLC-␤ substrate PIP 2 or Ca 2ϩ was present, the K d for G␣ q binding to PLC-␤3 was ϳ200 nM. The concentration of the G␣ q -F-GTP␥S FRET donor also did not alter the value of K d over the range of 2-10 nM.
G␣ q binding to PLC-␤3 depends on G␣ q activation. G␣ q -F-GDP, which does not activate PLC-␤ at attainable concentrations, was only 8% quenched by 4000 nM PLC-␤3-Q, ϳ20 times the K d for GTP␥S-activated G␣ q -F ( Fig. 1C; see also supplemental Fig. S1 (B and C)). Quenching of G␣ q -F-GDP did not saturate, but these data are consistent with a K d of about 50 M, about 250-fold higher than that for activated G q . The behavior of partially activated preparations of G␣ q -F is also consistent with the low affinity of G␣ q -GDP for PLC-␤. When increasing concentrations of PLC-␤3-Q were added to partially activated preparations of G␣ q -F, quenching saturated with K d ϳ200 nM, but fractional quenching was reduced approximately in proportion to the amount of G␣ q that remained bound to GDP. For example, when G␣ q -F was only ϳ40% activated by GTP␥S, donor quenching was ϳ30% at saturating PLC-␤3-Q concentrations (Fig. 1B). Under the same conditions, when G␣ q -F was ϳ95% active, ϳ50% of the donor was quenched at saturating PLC-␤3-Q concentrations (Fig. 1C).
To further verify that donor quenching measures equilibrium binding, we tested the ability of unlabeled, wild-type PLC-␤3 to compete with PLC-␤3-Q for G␣ q -F binding. Activated wild-type PLC-␤3 blocked quenching of G␣ q -F-GTP␥S by PLC-␤3-Q with a K i of 180 Ϯ 30 nM, similar to the K d determined for the two labeled proteins (Fig. 1D). In contrast, a PLC-␤3 that is mutated at a G␣ q -binding interface, L859E, did not inhibit donor quenching (Fig. 1D). This mutant is not stimulated by G␣ q and does not bind G␣ q according to surface plasmon resonance measurements (10). PLC-␤1 also inhibited PLC-␤3-Q binding with a K i of 400 nM.
By standard criteria, the data described above indicate that GTP␥S-activated G␣ q -F binds PLC-␤3-Q reversibly with an equilibrium K d of about 200 nM, and that binding affinity is at most modestly altered by the presence of phospholipid bilayers or detergent micelles. In contrast, GTP␥S-activated G␣ q -F stimulates the activity of PLC-␤3-Q with an EC 50 of 2-5 nM Figure 1. Equilibrium binding of PLC-␤3 to G␣ q -GTP␥S determined by FRET. A, donor quenching and acceptor enhancement are observed when PLC-␤3-Q is added to G␣ q -F-GTP␥S. Fluorescence spectra of 10 nM G␣ q -F-GTP␥S alone (black) and with 50 nM (red), 200 nM (blue), or 1000 nM (green) PLC-␤3-Q were obtained in the presence of PE/PS/PIP 2 vesicles with no added Ca 2ϩ . Direct excitation of the acceptor has been subtracted. G␣ q -F was ϳ100% activated by GTP␥S, and maximal quenching was 42%. B, binding of PLC-␤3-Q to 2 nM G␣ q -F-GTP␥S was measured according to donor quenching in the presence of PE/PS/PIP 2 vesicles with 60 nM Ca 2ϩ . Data are the average and range of two independent experiments. The data were fit to a single-site binding equation, yielding a K d of 160 Ϯ 50 nM. In these experiments, G␣ q -F was 42% activated, and maximal donor quenching (Q max ) was 33 Ϯ 2%. C, binding of PLC-␤3-Q to 10 nM G␣ q -F-GTP␥S (closed circles) was measured in the presence of PE/PS/PIP 2 vesicles with no added Ca 2ϩ . Data are the average and range from three independent experiments. The data were fit to a single-site binding equation with K d ϭ 242 Ϯ 18 nM. In these experiments, G␣ q -F was 95% activated, and Q max ϭ 51.8 Ϯ 0.8%. Binding of PLC-␤3-Q to G␣ q -F-GDP is also shown (open circles), with a fit to a linear function. D, competitive binding of PLC-␤3 to G␣ q -F in the presence of PE/PS vesicles. Increasing concentrations of unlabeled, wild-type PLC-␤3 (closed circles) were added to a mixture of 10 nM G␣ q -F and 250 nM PLC-␤3-Q. The data shown are the average and range of two independent experiments. Data for wild-type PLC-␤3 were fit to a model of single-site competition, yielding an IC 50 of 430 Ϯ 80 nM. The G␣ q -F was 50% activated by GTP␥S, and donor quenching in the absence of inhibitor, ϳ16%, was consistent with the K d determined in B and C. A third competition experiment, with G␣ q -F that was activated 100%, yielded ϳ30% donor quenching without inhibitor (data not averaged in this figure). These three competition experiments, following correction of IC 50 values for the concentration of labeled PLC-␤3 present, yielded a K i of 180 Ϯ 30 nM, similar to the K d shown in B and C. The G q -unresponsive mutant PLC-␤3, L859E (10) (open circles), did not inhibit binding of 2.5 nM G␣ q -F-GTP␥S binding to 250 nM PLC-␤3-Q. Data are the average and range (error bars) of two independent experiments.

TABLE 1
Summary of G␣ q ؊PLC-␤3 K d measurements G␣ q binding to phospholipase C-␤3 ( Fig. 2A), which is similar to the EC 50 displayed by the unlabeled wild-type proteins (7,11). A similar EC 50 is observed for the G q GAP activity of PLC-␤3-Q (Fig. 2C). Such differences between EC 50 and K d , approaching 100-fold, were observed under identical experimental conditions, although the simplest models of allosteric regulation would predict that EC 50 and K d would be identical. We therefore set out to test the validity of the K d measurement.
The discrepancy between EC 50 and K d is not caused by the mutation or covalent labeling of G␣ q -F and PLC-␤3-Q, as shown by the activity data in Fig. 2 and the competition data in Fig. 1D. Maximal stimulation of the PLC activity of PLC-␤3-Q by G␣ q -F-GTP␥S is several hundred-fold ( Fig. 2A), as is true for the two wild-type proteins under similar conditions (7). The fraction of G␣ q -F that bound GTP␥S ranged among preparations between 50% and nearly 100%, as reported for wild-type G␣ q (17). Further, in unilamellar phospholipid vesicles reconstituted with m1 muscarinic acetylcholine receptor (m1AChR), G␣ q -F, and G␤␥, the receptor catalyzed G␣ q -F activation by GTP␥S with a rate constant of 0.5 s Ϫ1 , 3 similar to that described previously for wild-type G␣ q (16). The receptor-G q -F vesicles also activated PLC-␤3-Q in response to carbachol plus GTP, as described previously for the wild-type proteins (compare Fig.   2B with data of Biddlecome et al. (16)). Last, the G q GAP activity of PLC-␤3-Q was similar to that of wild-type PLC-␤3, with the same EC 50 (compare Fig. 2C with Biddlecome et al. (16) and Waldo et al. (10)).
To test whether the ϳ200 nM K d obtained for PLC-␤3-Q and G␣ q -F is sensitive to the specific placement of the donor and acceptor fluorophores, we prepared and studied the binding of two additional pairs of G␣ q -PLC-␤3 FRET sensors. In one pair, the G␣ q -F donor was combined with PLC-␤3 labeled with Alexa Fluor 594 maleimide at residue 720 (PLC-␤3-720 -Alx594). In the second, cerulean fluorescent protein was inserted as a FRET donor between residues Phe 124 and Glu 125 in G␣ q (G␣ q -Cer), and the acceptor Alexa Fluor 594 maleimide was attached to Cys 277 in the EF-hand domain of PLC-␤3 (PLC-␤3-277-Alx594). These two FRET pairs yielded similar values for K d , specifically 210 Ϯ 40 and 270 Ϯ 60 nM (Fig. 3). Under identical assay conditions, the three FRET pairs yielded an average K d of 250 Ϯ 40 nM. In preliminary experiments, we constructed several other FRET pairs, listed in supplemental Table  S1, and tested them at concentrations of acceptor-labeled PLC-␤3 up to 50 nM. None yielded FRET signals consistent with a K d below 50 nM.
In addition to the experiments described above, we also prepared G␣ q labeled at residue 249 with the FRET acceptor Alexa Figure 2. Fluorescently labeled G␣ q -F and PLC-␤3-Q behave biochemically like the wild-type, unlabeled proteins. A, the activation of wild-type PLC-␤3 by GTP␥S-bound wild-type G␣ q (closed circles) and PLC-␤3-Q by G␣ q -F-GTP␥S (open circles) was assayed at 60 nM free Ca 2ϩ . Duplicate measurements from three independent experiments were normalized and averaged. The range of maximum reaction rates obtained was 500 -4000 pmol of IP 3 /min/pmol of PLC-␤3. In sets of parallel experiments, EC 50 was 2.3 Ϯ 0.4 nM for wild-type PLC-␤3 activated by wild-type G␣ q (800 Ϯ 600-fold stimulation, n ϭ 3), consistent with previously reported measurements (7, 11); EC 50 was 1.7 Ϯ 1.1 nM for PLC-␤3-Q activated by G␣ q -F (500 Ϯ 300-fold stimulation, n ϭ 3); EC 50 was 4 Ϯ 3 nM for wild-type PLC-␤3 activated by G␣ q -F-GTP␥S (220 Ϯ 80-fold stimulation, n ϭ 3); and EC 50 was 1.4 Ϯ 0.1 nM for activation of PLC-␤3-Q by wild-type G␣ q (800 Ϯ 400-fold activation, n ϭ 2). Maximum activation and absolute PLC activity vary among experiments because of variation in the preparation of substrate vesicles. B, the carbachol-stimulated lipase activity of PLC-␤3-Q was assayed in reconstituted PE/PS/[ 3 H]PIP 2 /cholesterol (16:10:1:2) vesicles that contained m1AChR, G␣ q -F, and G␤␥. An EC 50 of 3.7 Ϯ 0.3 nM (n ϭ 2) was obtained, consistent with previous measurements for wild-type proteins (16). Data shown are from one of two experiments. C, the G q GAP activity of PLC-␤3-Q was assayed at steady state in PE/PS/cholesterol (9:5:1) vesicles reconstituted with 1.2 nM m1AChR, 1.4 nM wild-type G␣ q and G␤␥. The EC 50 was 4.1 Ϯ 0.9 nM, again similar to that measured previously with wild-type proteins (3,10,16). Data shown here are the average and range (error bars) of duplicates from a single experiment.

G␣ q binding to phospholipase C-␤3
Fluor 594 and PLC-␤3 labeled at residue 871 with the donor Alexa Fluor 488. This pair is identical to the G␣ q -F and PLC-␤3-Q pair, but with donor and acceptor reversed. Although we were not able to add a high enough concentration of the G␣ q acceptor to approach saturation, quenching of the PLC-␤3 donor was consistent with K d Ն 200 nM. Last, in an attempt to approximate the experiment of Runnels and Scarlata (25), we measured FRET from G␣ q -F to wild-type PLC-␤3 that was uniformly labeled (14 cysteine residues) with Alexa Fluor 594 maleimide. For this pair, K d in the presence of PC/PE/PS (1:1:1) vesicles was 560 Ϯ 110 nM (n ϭ 2; supplemental Fig. S1D). This high K d presumably reflects partial inactivation of the PLC by alkylation, as we observed in enzymatic assays. Based on all of these data, the equilibrium affinity of G␣ q binding to PLC-␤3 does not reflect the location or chemical nature of the label, and affinity is confirmed by competition with unlabeled wild-type PLC-␤3.

Equilibrium G␣ q -PLC-␤3 binding measured by FCS
We next measured PLC-␤3 binding to G␣ q using FCS (26). Unlike FRET, FCS does not depend on the relative positions of the two fluorophores in the complex. In FCS, the fluorescence intensity in a small confocal volume is recorded as a function of time. Because the number of fluorescent particles in that volume is small, translational diffusion of fluorescent molecules in and out of the observation volume causes fluorescence intensity to fluctuate significantly relative to the time-averaged fluorescence. These fluctuations are autocorrelated using Equation 1 (see "Experimental procedures") (27), and the autocorrelation function, G(), decays to 1 as approaches infinity. The rate and shape of this decay depend on the diffusion coefficients of fluorescent species in the observation volume. When wild-type PLC-␤3 (142 kDa) binds to G␣ q -F (42 kDa), the translational diffusion coefficient D of the labeled G␣ q decreases significantly due to both the 4-fold increase in size of the G␣ q -PLC-␤3 complex relative to G␣ q and the asymmetry of the resulting complex (12).
As shown in Fig. 4A, G() for GTP␥S-activated G␣ q -F increased markedly in the presence of saturating PLC-␤3. Fitting the values of G() to Equation 2 showed that the diffusion coefficient D for G␣ q decreased from 44 Ϯ 4 m 2 s Ϫ1 in the absence of PLC-␤3 to 24.8 Ϯ 1.7 m 2 s Ϫ1 at saturating PLC-␤3, a decrease of 44%. In these experiments, G␣ q -F was only 50% activated. Hence, the G␣ q -F diffusion coefficient of 24.8 Ϯ 1.7 m 2 s Ϫ1 obtained at saturating PLC-␤3 concentrations was the average diffusion coefficient obtained for a mixture of free and PLC-␤3bound G␣ q -F. Diffusion coefficients of spherical particles with equal density are inversely proportional to the cube roots of their molecular masses. If G␣ q and its complex with PLC-␤3 were spherical, the diffusion coefficient of 42-kDa G␣ q would be predicted to decrease by 37% upon binding 142-kDa PLC-␤3. The slightly larger decrease in D presumably reflects the high asymmetry of the G␣ q -PLC-␤3 complex (12).
Values of D for GTP␥S-activated G␣ q were then measured in the presence of increasing concentrations of PLC-␤3 to deter-  Table 1), no other pair that displayed FRET showed evidence of saturation Ͻ50 nM.

G␣ q binding to phospholipase C-␤3
mine their binding affinity (Fig. 4B). The decrease in D of G␣ q -F upon PLC-␤3 binding was well fit by a simple bimolecular binding function to yield K d ϭ 120 Ϯ 20 nM. This value corresponds well to the value of K d that was determined by FRET in the same aqueous buffer or in buffer with 0.2% cholate. The observed affinity of PLC-␤3 binding to G␣ q is thus the same when measured by either intermolecular FRET or by FCS.
The diffusion coefficient of GDP-bound G␣ q -F also decreased at high PLC-␤3 concentrations, but only by 23%, with an apparent K d of 240 Ϯ 60 nM (Fig. 4B). This is about half the decrease observed with the 50% activated G␣ q -F. These data might suggest a second, lower-affinity binding mode to create a complex with a different hydrodynamic radius. If this hypothetical binding site were near the PLC-␤3 C terminus, our FRET sensors would have been too far apart to detect the interaction.

Association and dissociation kinetics of G␣ q and PLC-␤3
To determine the kinetic behavior of G␣ q -PLC-␤3 binding, we used stopped-flow FRET to measure the association and dissociation rates for G␣ q -F and PLC-␤3-Q. To measure association, donor quenching was followed upon the addition of PLC-␤3-Q (150, 300, or 450 nM) to GTP␥S-activated G␣ q -F (5 or 25 nM) at 25°C in the presence of 250 M PE/PS (4:1) vesicles and no added Ca 2ϩ (Fig. 5A). Data were fit to a single exponential and a linear term to account for minor photobleaching. The pseudo-first order association rate increased with increasing PLC-␤3-Q concentration, yielding a second-order association rate constant, k on , of 1.2 Ϯ 0.2 ϫ 10 6 M Ϫ1 s Ϫ1 (Fig. 5A, inset). These data also yielded an extrapolated dissociation rate constant k off of 0.05 Ϯ 0.03 s Ϫ1 . Dissociation was measured independently under identical conditions by monitoring recovery of  A, to measure association, quenching of donor fluorescence was followed upon the addition of PLC-␤3-Q (300 nM) to GTP␥S-activated G␣ q -F (5 nM). Data were fit to a rate equation with a single exponent plus a linear term that accounts for photobleaching, which is also seen with G␣ q -F alone. Inset, pseudo-first-order association rate constants obtained at 150, 300, and 450 nM PLC-␤3-Q were fit to yield a second-order association rate constant of 1.2 ϫ 10 6 M Ϫ1 s Ϫ1 and an extrapolated dissociation rate constant of 0.05 Ϯ 0.03 s Ϫ1 . Each data point shown in the inset is the average and S.D. (error bars) of association rates derived from four independent experiments, each with 12 mixing curves of the sort shown in A at each PLC concentration. Rates did not vary with substoichiometric concentrations of G␣ q -F. G␣ q -F used in these experiments was 40 -50% activated. B, to measure dissociation, recovery of donor fluorescence was measured upon the addition of 1.8 M unlabeled wild-type PLC-␤3 to a mixture of 5 nM GTP␥S-activated G␣ q -F (25 nM in some other experiments) and 250 nM PLC-␤3-Q. The data shown are representative of 13 fluorescence recovery curves obtained from three independent experiments. The data were fit to a bi-exponential rate equation with a fast dissociation rate constant of 0.60 Ϯ 0.17 s Ϫ1 that accounts for 67 Ϯ 5% of the fluorescence recovery and a slower rate of 0.05 Ϯ 0.02 s Ϫ1 that accounts for 33 Ϯ 5% of the recovery.
donor fluorescence upon the addition of excess, unlabeled PLC-␤3 (1.8 M) to a mixture of G␣ q -F (5 or 25 nM) and PLC-␤3-Q (250 nM). These fluorescence recovery data were not fit well by a single exponent (see "Experimental procedures"). A two-exponent fit yielded a principal dissociation rate constant, k off , of 0.60 Ϯ 0.17 s Ϫ1 that accounted for 67 Ϯ 5% of the recovery and a slower dissociation rate of 0.05 Ϯ 0.02 s Ϫ1 , probable drift, that accounted for the remainder. The slower value agrees with the extrapolated k off obtained from the association data in Fig. 5A. These two rates bracket the dissociation rate of 0.3 s Ϫ1 measured in cells by Jensen et al. (28). A k on of 1.2 ϫ 10 6 M Ϫ1 s Ϫ1 and the major k off of 0.6 s Ϫ1 yield a kinetically determined K d ϭ 500 nM, but even the slower value of k off yields a K d that is still more than 10-fold higher than the EC 50 . Hence, the kinetic parameters yield a K d similar to that obtained directly from equilibrium binding measurements using either FRET or FCS.

G␣ q and PLC-␤3 binding to phospholipid vesicles
The interaction of PLC-␤ and G␣ q could be strongly influenced by their simultaneous binding to phospholipid vesicles, where relevant PLC activation occurs. We used classical orderof-addition experiments to show that the binding of these two proteins to phospholipid substrate vesicles is reversible, in that they display some component of "hopping" kinetics (29) in addition to their "scooting" on the surface of individual vesicles over the time frame of both activity and binding assays (Fig. 6).
We also used FCS to confirm that G␣ q and PLC-␤3 each bind to phospholipid vesicles and that their binding is rapidly reversible (Fig. 7). Fluorescence autocorrelations were measured for 7.5 nM PLC-␤3-Alx488 (labeled at position 871) (Fig. 7A) and 5 nM G␣ q -F (Fig. 7B) in the absence or presence of 250 M PE/PS (4:1) vesicles. Autocorrelation curves obtained for each protein were markedly right-shifted in the presence of vesicles (Fig. 7), indicating a significant decrease in the diffusion coefficient of each protein due to vesicle binding. To determine the fraction of protein bound to vesicles, autocorrelation data (Fig. 7, red circles) were fit to an equation with two sets of diffusion and triplet state terms (Equation 3), one each for the population of free protein and protein bound to vesicles (Equation 4). This analysis indicated that 56 Ϯ 6 and 44.5 Ϯ 0.4% of PLC-␤3-Alx488 and G␣ q -Alx488, respectively, were bound to vesicles. These values indicate that both proteins can freely exchange between vesicles and the aqueous medium, consistent with the hopping kinetics observed for PLC stimulation.

Relationship of K d and EC 50
Several recognized molecular mechanisms of enzyme activation can give rise to an EC 50 that is significantly smaller than the equilibrium K d for activator-enzyme binding, and we have considered three for possible involvement in the G␣ q -PLC-␤3 system.
Classically, "essential allosteric activation," in which activator is required for enzyme-substrate binding (24), can produce EC 50 Ͻ Ͻ K d but requires that substrate concentration be well above K m for this behavior. Because the concentrations of PIP 2 used here are well below K m (Biddlecome et al. (16); confirmed here), this mechanism does not apply. A second set of mechanisms that can produce EC 50 Ͻ Ͻ K d involves enzyme oligomerization with strongly negatively cooperative ligand binding. However, a disagreement of 50 -100-fold, the sort observed here, would require very high stoichiometries of oligomerization to completely obscure the initial, highest-affinity binding event. No evidence of such oligomerization has been obtained, and Waldo et al. (10) found that PLC-␤3 behaves as a monomer according to size-exclusion chromatography. Whereas we found variable formation of an apparent dimer of PLC-␤1 in pilot analytical ultracentrifugation experiments, we also found that most protein behaved as a monomer. Singer et al. (8) demonstrated formation of a dimer of the C-terminal helical domain of turkey PLC-␤2 during crystallization, and the importance of the dimer interface was supported by the effect of mutagenesis (20). Based on this work, we introduced novel cysteine residues (S1042C, H1104C, and Q1157C) in the PLC-␤3 C terminus to probe oligomerization by FRET. No changes in donor fluorescence were observed when PLC-␤3 S1042C-Alx549 (0 -500 nM) was added to PLC-␤3 H1104C-Alx488 (2.5 nM) or when PLC-␤3 Q1157C-Alx594 (0 -750 nM) was added to PLC-␤3 Q1157C-Alx488 (1 nM), with or without 100 nM G␣ q -GTP␥S. Each pair is separated by 15 Å in the turkey PLC-␤ C terminus crystal structure (8), well within the Förster distance for these fluorophores. In summary, we find no evidence for higher-order oligomerization as an origin of the high K d observed for G␣ q binding to PLC-␤3.

G␣ q binding to phospholipase C-␤3
Hysteretic PLC activation is a more likely mechanism. If PLC-␤3 remains active for some time after it dissociates from G␣ q -GTP␥S (or after G␣ q is deactivated but perhaps remains bound), EC 50 will be lower than K d . In such a mechanism, the PLC-␤3 deactivation rate, k deact , is slower than the G␣ q -GTP␥S dissociation rate (k off ). Thus, EC 50 , the ratio of the deactivation and activation rate constants k deact /k act , is lower than K d , the ratio of dissociation and association rate constants k off / k on . Because k act cannot be faster than k on , EC 50 is less than K d . Such a hysteretic PLC-␤3 activation/deactivation cycle allows G␣ q to behave formally as a catalyst of PLC activation. To determine whether PLC-␤3 displays such hysteresis, we compared the rate of dissociation of G␣ q -GTP␥S from PLC-␤3 (k off , measured as described above) with the rate of PLC-␤3 deactivation following its dissociation from G␣ q -GTP␥S (k deact ). Dissociation of G␣ q -GTP␥S and consequent dissociation of the active complex was initiated by the addition of an excess of a catalytically inactive PLC-␤3 mutant to sequester free G␣ q -GTP␥S as it dissociates. PLC-␤3 H332A,H379A, which lacks the two catalytic histidine residues, does not hydrolyze PIP 2 but competitively inhibits activation of wild-type PLC-␤3 by G␣ q -GTP␥S with a K i that is equal to the EC 50 for activation of the wildtype enzyme (Fig. 8A). G␣ q -GTP␥S-stimulated PLC-␤3 activity was measured before and after the addition of excess H332A,H379A PLC-␤3 (Fig. 8B), and the rate of PLC deactivation was used to determine the deactivation lifetime ( deact ϭ 1/k deact ), as described by Cassel et al. (30). The calculated value of deact was 0.3 Ϯ 4.7 s (n ϭ 3), which was too short to be measured accurately but which places a lower bound on k deact of ϳ3 s Ϫ1 . This value is actually faster than the dissociation rate constant of 0.6 s Ϫ1 obtained for G␣ q -GTP␥S and PLC-␤3 (Fig.  5B), indicating that PLC-␤3 hysteresis of this sort is not responsible for the discrepancy between EC 50 and K d .
In cells, however, it is likely that hydrolysis of G␣ q -bound GTP is the initial deactivating event in signal termination rather than the dissociation of GTP-activated G␣ q from PLC-␤.
Thus, slow deactivation of PLC bound to G␣ q -GDP might provide an alternative hysteretic mechanism to account for K d Ͼ EC 50 . We therefore compared the rate of G␣ q deactivation promoted by the GAP activity of PLC-␤ with the rate of consequent deactivation of PLC-␤3-catalyzed PIP 2 hydrolysis to see whether PLC deactivation is delayed, another possible source of hysteresis. We previously showed that the PLC-␤-promoted hydrolysis of G␣ q -GTP occurs with a rate of ϳ10 s Ϫ1 in a system of unilamellar phospholipid vesicles reconstituted with m1AChR and G q (3). Here, we measured the rate of PLC-␤3 deactivation following G␣ q deactivation in the same reconstituted system. PLC-␤3 activity was first allowed to reach receptor-stimulated steady-state in the presence of GTP and the agonist carbachol. To initiate deactivation, the antagonist atropine and excess GDP were added, and PLC-␤3 activity was measured for several minutes thereafter. Deactivation lifetimes were calculated as in Fig. 8B (30). As shown in Fig. 9, deactivation lifetimes are relatively long when the concentration of G␣ q is greater than that of PLC-␤3 because excess G␣ q -GTP must be sequentially turned off by a limiting amount of PLC-␤3. At PLC-␤3 concentrations that were equal to or greater than the steady-state concentration of GTP-activated G␣ q , the time constant for PLC-␤3 deactivation approached 2.7 Ϯ 1.6 s, which corresponds to a lower bound for the rate, k deact Ն 0.4 s Ϫ1 . This rate is about 25-fold slower than the previously measured rate of GTP hydrolysis. This difference alone is not slow enough to account for the nearly 100-fold difference between EC 50 and K d for the interaction of G␣ q and PLC-␤3 but could partially explain the discrepancy between the EC 50 and K d .

Discussion
We set out to measure the thermodynamics and kinetics of PLC-␤3 binding to G␣ q to define the biochemical mechanisms of their mutual regulation. Activation of PLC-␤ by G␣ q is widely thought to be a classical allosteric event, as is the G q GAP activity of PLC-␤ (7). If classical allostery were the mechanism of this bidirectional regulation, then the K d for G␣ q binding to PLC-␤ should be equal to the EC 50 values for the two activities. Both EC 50 values, for PLC-␤ activation by G␣ q and acceleration of G q -GTP hydrolysis by PLC-␤ activation, are 2-5 nM (7,10,11,16,17). Hence, the corresponding K d was expected to be in the same, low nanomolar range. Surprisingly, the data presented here show that the K d for PLC-␤3 binding to G␣ q -GTP␥S is in the range of 130 -250 nM, almost 2 orders of magnitude higher ( Table 1).
The unexpectedly high value of K d determined here appears to be correct according to multiple criteria. First, binding affin-ity was determined by two distinctly different techniques: measurement of FRET between single fluorophores covalently bound to PLC-␤3 and G␣ q and FCS measurement of retardation of diffusion of fluorescently labeled G␣ q by unlabeled PLC-␤3. Because the FRET measurement is simpler, we were also able to perform several other controls in that format. Binding was competitively inhibited by unlabeled PLC-␤3 with K i equal to the K d determined directly by FRET, and the G q -unresponsive L859E mutant PLC-␤3 did not inhibit. Only the GTP␥Sactivated form of G␣ q bound PLC-␤3 with significant affinity. Binding affinity was not changed by the choice of fluorophore bound to PLC-␤3 or by substitution on G␣ q of cerulean fluorescent protein in place of a covalently bound synthetic fluorophore. Labeling of PLC-␤3 or G␣ q at multiple positions on each molecule that are close by in the complex reported the same affinity. Last, binding affinity was not greatly altered by precise experimental conditions: the presence or absence of phospholipid vesicles with or without the PLC substrate PIP 2 , the inclusion of detergent, or the presence of Ca 2ϩ , which is required for the PLC reaction. Last, the ratio of the dissociation rate constant to the association rate constant, measured by stoppedflow FRET, was approximately equal to the equilibrium dissociation constant (i.e. K d ϭ k d /k a . Our data thus indicate that the K d for PLC-␤3 binding to G␣ q is about 200 nM and clearly much higher than EC 50 values determined here or previously. The K d value of ϳ200 nM reported here agrees with K d values of 206 and 108 nM (two experiments) that were determined by Waldo et al. (10), who used surface plasmon resonance to measure the affinity of PLC-␤3 for an Al 3ϩ /F Ϫ -activated G␣ i -G␣ q chimera, which activates PLC-␤.
In contrast, Runnels and Scarlata (25) obtained much lower K d values for G␣ q binding to three different PLC-␤ isoforms. They used intermolecular FRET between wild-type G␣ q labeled with an amine-reactive coumarin and wild-type PLC-␤ covalently labeled with a thiol-reactive quencher (dabcyl). In that study, binding measured in the presence of PC/PE/PS (1:1:1) vesicles yielded K d values of Ͻ10 nM for G␣ q -GTP␥S and PLC-␤1, -␤2, or -␤3, which are similar to values of EC 50 for PLC activation that have been measured by us and others. However, when we measured binding of G␣ q -F-Alx488 to wild-type Figure 8. PLC-␤3 deactivation following PLC-␤3 dissociation from G␣ q -GTP␥S. A, the catalytically inactive H332A,H379A PLC-␤3 inhibits G␣ q -GTP␥Sstimulated PLC-␤3 activity. The data are the average and range (error bars) of duplicates from a single experiment with a fit to a single-site competitive inhibition equation. An IC 50 of 3.2 Ϯ 1.6 nM was obtained from two independent experiments using 5 nM G␣ q -GTP␥S and 1 nM PLC-␤3. Given an EC 50 of 3 nM, IC 50 Ϸ K i . B, the PLC-␤3 deactivation lifetime, deact , was measured at 25°C by following the activity of 1 nM PLC-␤3 stimulated by 5 nM G␣ q -GTP␥S before and after quenching of activation by 600 nM H332A,H379A PLC-␤3 (red arrow). The deactivation time is given by the difference between the time at which H332A,H379A PLC-␤3 was added and the time at which the linear extrapolations of the initial and final rates intersect (black arrow) (30). These deactivation times were too fast to measure accurately. In three independent experiments, approximate values were 4 s (shown), 2 s, and Ϫ5 s. Figure 9. PLC-␤3 deactivation following GAP-catalyzed deactivation of GTP-activated G␣ q . The deactivation lifetime of PLC-␤3 following G␣ q deactivation was measured at 30°C in a reconstituted vesicle system consisting of m1AChR, G␣ q , and G␤␥. Vesicles were allowed to reach steady state in the presence of carbachol and GTP. Activation was terminated at zero time by the addition of the m1AChR antagonist, atropine, and excess GDP. PLC-␤3 activity was measured before and after termination, and the PLC-␤3 deactivation lifetime was obtained using the method described by Cassel et al. (30), similar to that shown in Fig. 8B. Deactivation lifetimes were measured at different concentration ratios of G␣ q and PLC-␤3. At higher G␣ q /PLC-␤3 ratios, deact is longer, indicating that the rate-limiting step in PLC-␤3 deactivation is the sequential deactivation of excess G␣ q -GTP molecules by the GAP activity of substoichiometric PLC-␤3. We obtained an estimate of PLC-␤3's intrinsic deactivation lifetime by plotting deact as a function of the G␣ q /PLC-␤3 ratio and extrapolating to G␣ q /PLC-␤3 ϭ 0, to yield deact ϭ 2.7 Ϯ 1.6 s. This value is an upper bound. The data points are observed deactivation times for individual experiments. Three different vesicle preparations were used to obtain the data shown. In these experiments, m1AChR concentrations varied from 0.62 to 2.7 nM, as determined by [ 3 H]quinuclidinylbenzilate binding. G␣ q concentrations varied from 0.5 to 3.5 nM as determine by carbachol-stimulated [ 35 S]GTP␥S binding (16).

G␣ q binding to phospholipase C-␤3
PLC-␤3 labeled with a thiol-reactive Alexa Fluor 594 in the presence of PC/PE/PS vesicles, we found K d ϭ 560 Ϯ 110 nM (supplemental Fig. S1D), about 2-3-fold higher than in our other measurements. We did not try to emulate the G␣ q donor used by Runnels and Scarlata (25) because we found that nonselective amine labeling inactivates G␣ q 4 and because subsequent dialysis against 0.1 M (NH 4 ) 2 SO 4 , which they used to remove unreacted fluorophore, would irreversibly inactivate G␣ q (31).
Studies of other proteins that activate PLC-␤ have also yielded physically determined values of K d much above the published EC 50 . Such disagreement may therefore be general and mechanistically informative for the PLC-␤ family. G␤␥ activates PLC-␤3 with an EC 50 of ϳ30 nM (7,19). However, FRET measurements of PLC-␤3 binding to G␤␥ in the presence of phospholipid vesicles yielded a K d of 220 Ϯ 46 nM (19), ϳ7-fold higher than the EC 50 . Similarly, Rac1 activates PLC-␤2 with EC 50 ϳ6 nM (19), but surface plasmon resonance measurements of GTP␥S-bound Rac2 binding to PLC-␤2 yielded K d ϳ5 M (32), ϳ1000-fold higher than the EC 50 . Although this latter comparison spans two laboratories, and the binding measurement used non-prenylated Rac1, the 1000-fold discrepancy is still striking. Hence, there may be a common and general mechanism by which potent activation of PLC-␤ isoforms is achieved despite much weaker equilibrium binding affinity of PLC-␤ for its activators.
Phospholipid bilayers, with or without PIP 2 or Ca 2ϩ , had little effect on the affinity of PLC-␤3-G␣ q binding as measured by intermolecular FRET (Figs. 1B and 3 (A and B), supplemental Fig. S1 (A-C), and Table 1). The addition of 0.2% cholate also had little effect (supplemental Fig. S1B). FCS measurements of PLC-␤3-G␣ q binding in aqueous buffer yielded a K d of 120 Ϯ 20 nM (Fig. 4B), consistent with the FRET measurements in the different media. Thus, PLC-␤3 binding to G␣ q in aqueous buffer or detergent solution is at least as tight as that measured in the presence of phospholipid vesicles, which suggests that permanent vesicle binding does not confine both proteins to a restricted annular membrane surface volume. This conclusion is consistent with FCS measurements of G␣ q and PLC-␤3 diffusion in the absence and presence of phospholipid vesicles, which showed that just less than half of each protein is bound to vesicles at equilibrium (Fig. 7). It is also consistent with the substantial amount of protein exchange among vesicles during the PLC assay (Fig. 6). G␣ q is palmitoylated at residues Cys 9 and Cys 10 , and knockdown of a G␣ q palmitoyl acyl-transferase causes G␣ q to relocalize from the plasma membrane to the cytoplasm and disrupts G␣ q signaling (33). Cys 9 and Cys 10 were retained in all G␣ q variants used in this work, and only the fraction of G␣ q that was both membrane-localized and attached to G␤␥ was purified (see "Experimental procedures"). We do not know whether G␣ q loses any palmitate during purification. Fluorescent maleimide labeling of G␣ q variants yielded 2 dye molecules bound per protein, suggesting that one of the cysteine residues at position 9 or 10 remains palmitoylated.
Of the multiple regulatory mechanisms that can produce an EC 50 value below the K d for the activator, the most likely in this case seemed to be slow deactivation of PLC-␤ after dissociation from G␣ q or the other protein activators mentioned above. Because the rate of activation cannot be faster than the binding rate, slow deactivation would lead to a lower than expected EC 50 . To test whether PLC-␤3 activation displays such hysteresis, we compared the rate of dissociation of the G␣ q -PLC-␤3 complex (Fig. 5B) with the rate of PLC-␤3 deactivation after dissociation (Fig. 8B). No such hysteresis was detected. The PLC-␤3 deactivation lifetime following dissociation varied between 2 and 5 s, and the dissociation lifetime itself was 1.5 Ϯ 0.3 s. The PLC-␤3 deactivation lifetime following GAP-catalyzed G␣ q deactivation was also found to have an upper limit of ϳ3 s (Fig. 9), consistent with the dissociation lifetime of 3.6 s measured in cells (28). For a simple hysteretic mechanism to produce a K d /EC 50 ratio of about 100, the PLC-␤3 deactivation rate would have to be 100-fold slower than the dissociation rate. The simplest version of this mechanism, therefore, does not explain our data.
A second mechanism that yields K d Ͼ Ͼ EC 50 is one in which the acceleration of an initially rate-limiting partial reaction in the catalytic pathway is accelerated to the point where a second partial reaction becomes limiting. Further stimulation by increased fractional binding of activator is therefore without effect, and activation saturates at ligand concentrations well below those needed to saturate binding. Classical examples are referred to as "essential" and "non-essential" activation, where the second partial reaction is substrate binding and the substrate concentration is well above K m (24). In the case of PLC-␤, however, the concentration of PIP 2 is substantially below K m (16) (confirmed in this study). Therefore, this set of mechanisms is also not applicable.
We cannot definitively propose a mechanism to account for the surprisingly high ratio of K d /EC 50 that we observe for the G␣ q -PLC-␤ interaction, but we believe it to have a kinetic basis. We have noted that G␣ q -stimulated PLC activity displays a definite lag period before maximal activity is attained. This lag can be 5-30 s in duration, depending upon assay conditions, and we have had to choose assay duration and PLC concentration carefully to ensure steady state for PLC assays where activities varied over a large range (cf. Ref. 7). We do not know the physical origin of this lag, but it may be related to the low value of k off that is obtained by extrapolating the pseudo-first-order association rate constant to zero PLC-␤3 concentration (Fig. 5, compare A (inset) with the value obtained in B and related dissociation measurements). This lower value may reflect the second phase of dissociation noted in Fig. 5B. Together, they suggest the existence of a transient, active state of the PLC that is hysteretic in its decay after G␣ q dissociation, but on a time scale shorter than the 2-s window that we are able to explore. This transient active state may relate to the two-step deactivation process proposed by Waldo et al. (10). This state may also be the one favored by G␤␥, which also activates PLC-␤ with an EC 50 lower than the observed K d (19) and which activates synergistically with G␣ q (7).

Proteins
Human PLC-␤3 with an N-terminal His 6 tag was purified from Escherichia coli as described previously (19). Solvent-exposed cysteine residues at positions 193, 516, 614, 892, 1005, and 1207 were mutated to serine by QuikChange mutagenesis (19). Additional point mutations were introduced in this background.
To create a catalytically inactive PLC-␤3 mutant, histidine residues at positions 332 and 379 in PLC-␤3 (35) were mutated to alanine by QuikChange mutagenesis. Purified PLC-␤3-L859E (10), a PLC-␤3 variant that cannot bind or be activated by G␣ q , was a kind gift from John Sondek (University of North Carolina).
Human G␣ q cDNA in which three of the protein's five cysteine codons (at positions 144, 219, and 330) had been mutated to alanine was a gift from Jürgen Wess (National Institutes of Health) (36). Cysteines at positions 9 and 10, which are subject to palmitoylation and are necessary for maximal PLC-␤ activation (37), were retained. To create G␣ q FRET sensors, a cysteine residue was introduced in place of Glu 249 in G␣ q (G␣ q -E249C) for fluorescence labeling. G␣ q with cerulean fluorescent protein (38) inserted between residues Phe 124 and Glu 125 (G␣ q -Cer) was used as a FRET donor in some experiments. G␣ q constructs were expressed and purified essentially according to Biddlecome et al. (16,39) but with some alterations. To express G␣ q FRET constructs, Sf9 insect cells were infected with baculoviruses encoding Ric8A-GST, G␤ 2 -His 6 , G␥ 2 -His 6 , and either G␣ q -E249C or G␣ q -Cer. A single baculovirus encoding both wild-type G␣ q and G␥ 2 -His 8 was made using the pFastBac-Dual shuttle vector (Invitrogen). To express wild-type G␣ q , Sf9 cells were infected with baculoviruses encoding G␣ q -G␥ 2 -His 8 , G␤ 2 -His 6 , and Ric8A-GST. Ric8A was co-expressed to improve G␣ q expression (39), but the soluble form of G␣ q that is bound to Ric8A activates PLC-␤ poorly (19,39). According to immunoblots, almost all Ric8A was removed during membrane preparation, and the remainder was removed upon adsorption of the G␣ q -G␥ 2 -His 8 -G␤ 2 -His 6 trimer to NTA-Ni 2ϩ -agarose (below). Cells were harvested at 4°C ϳ50 h after infection. Only G␣ q that was both membranelocalized and initially bound to G␤␥ was purified. All purification steps were performed on ice or at 4°C. Cells were resuspended in ice-cold lysis buffer (20 mM Tris (pH 8), 3 mM MgCl 2 , 2 g/ml leupeptin, 1 g/ml aprotinin, 200 M PMSF, and 24 g/ml DNase) at 15 ml of lysis buffer/g of cells. Following Dounce homogenization, the cell lysate was centrifuged at 30,000 ϫ g for 25 min. The resulting pellet was resuspended in 3.5 ml of lysis buffer/g of cells, homogenized, and again centrifuged at 30,000 ϫ g for 25 min. The membrane pellet was finally resuspended in 2 ml of lysis buffer/g of cells. Protein concentration of the membrane lysate was measured by the method of Bradford (40). Membranes were frozen in liquid nitrogen and stored at Ϫ80°C for future use or immediately used for protein purification.
The membrane protein extract was diluted 5-fold with buffer N1 and mixed with 15 ml of NTA-Ni 2ϩ -agarose. The mixture was rotated for 2 h and the NTA-Ni 2ϩ -agarose was transferred to a chromatography column. The resin was washed to baseline with buffer N1, N2, and N1, in that order. The resin was rinsed with 15 ml of buffer N3. G␣ q was separated from G␤ 2 -His 6 -G␥ 2 -His 6(8) and eluted using buffer N4 (41).
The eluted G␣ q contained phospholipase activity. To remove this contaminant, the pooled fractions were diluted 5-fold in 20 mM MES (pH 6), 1 mM MgCl 2 , 0.1 mM EDTA, 10 M GDP, 1 mM DTT, and 0.5% CHAPS and applied to SP-Sepharose in the same buffer. The contaminant was adsorbed to the SP-Sepharose resin. G␣ q in the unadsorbed fraction was diluted 4-fold in buffer Q (20 mM Tris (pH 8), 1 mM MgCl 2 , 0.1 mM EDTA, 10 M GDP, 1 mM DTT, 0.5% CHAPS) supplemented with 10 mM NaCl and applied to a Source Q column (1 ml) equilibrated with buffer Q. The column was washed to baseline with 10 mM NaCl, and G␣ q was eluted over a 40-ml gradient of 10 -400 mM NaCl. G␣ q eluted between 200 and 250 mM NaCl. G␣ q was concentrated in a 10,000 molecular weight cut-off Amicon 0.5-ml centrifugal filter (Millipore). Glycerol was added to 10%, and G␣ q was frozen in liquid nitrogen and stored at Ϫ80°C.
To label G␣ q -E249C, the protein was buffer-exchanged into 20 mM Tris (pH 7.5), 100 mM NaCl, 0.1 mM EDTA, 1 mM MgCl 2 , 10 M GDP, 0.5% CHAPS to remove DTT. G␣ q -E249C (10 -20 M) was incubated with Alexa Fluor 488 C 5 maleimide or Alexa Fluor 594 C 5 maleimide (250 M) for 1 h on ice. The reaction was quenched by adding 1 mM DTT. Free dye was removed by G␣ q binding to phospholipase C-␤3 successive rounds of buffer exchange into G␣ q storage buffer (20 mM Tris (pH 7.5), 100 mM NaCl, 0.1 mM EDTA, 1 mM MgCl 2 , 10 M GDP, 1 mM DTT) without detergent using a 10,000 molecular weight cut-off Amicon 0.5-ml centrifugal filter (Millipore). Fluorescently labeled G␣ q -E249C in G␣ q storage buffer supplemented with 0.5% CHAPS was frozen in liquid N 2 and stored at Ϫ80°C. Labeling ratios were determined as described for PLC-␤3. Labeling ratios of 2 dye molecules/protein were consistently obtained, suggesting that in addition to Cys 249 , one of the cysteines at positions 9 and 10 was also labeled.
Wild-type G␣ q , G␣ q -Cer (38), and G␣ q -E249C labeled with Alexa Fluor 488 (G␣ q -F) were activated by incubation at 20°C for 20 h with 1 mM GTP␥S in 50 mM NaHEPES (pH 7.5), 100 mM NaCl, 2 mM MgCl 2 , 1 mM DTT, 200 mM (NH 4 ) 2 SO 4 , 0.1 mg/ml BSA, and 0.5% CHAPS. The fraction of G␣ q that was activated was determined by [ 35 S]GTP␥S binding (16). Fractional activation ranged from 50 to 100% and is accounted for in experiments (see "Results"). G␤ 1 ␥ 2 was purified after expression with His 6 -G␣ i in Sf9 cells as described earlier (41). A human m1AChR construct with an N-terminal FLAG epitope, a C-terminal His 6 tag, and a deletion (residues 232-253) in the third intracellular loop was expressed and purified as described previously (42).

Phospholipid vesicles
Large unilamellar vesicles used in FRET, FCS, and biochemical assays were composed of PE and porcine brain PS (Avanti Polar Lipids). In some FRET and biochemical assays, vesicles also contained porcine brain PIP 2 (Avanti Polar Lipids) and inositol-[2-3 H]PIP 2 (PerkinElmer Life Sciences). The phospholipid composition of vesicles used in each experiment is specified in the text. To make vesicles, lipids were dried under argon and resuspended in 50 mM NaHEPES, 100 mM NaCl, 0.2 mM NaEGTA and sonicated to opalescence in a sonication bath. Vesicles were prepared immediately before experiments.

FRET measurements
FRET was used to measure the binding of G␣ q to PLC-␤3. Equilibrium FRET measurements were performed using a Fluorolog3-212 spectrophotometer (Horiba Jobin Yvon). For FRET between Alexa Fluor 488-and Alexa Fluor 594-labeled proteins, emission was scanned between 510 and 625 nm following excitation at 475 nm. For FRET measurement between cerulean and Alexa Fluor 594, emission was scanned between 450 and 625 nm following excitation at 410 nm. A 420-nm cut-on filter was placed in-line with the emission path to block scattered excitation light. Excitation and emission bandwidths were 5 nm. Emission scans were obtained with a 1-s integration time in 1-or 2-nm increments.
For equilibrium G␣ q -PLC-␤3 binding measurements using FRET, fluorescently labeled G␣ q was incubated for 30 min on ice with large unilamellar vesicles (either 0.26 mM PE/PS/PIP 2 (16:4:1) or 0.25 mM PE/PS (4:1)) in binding assay buffer A (50 mM NaHEPES (pH 7.5), 100 mM NaCl, 4 mM MgCl 2 , 0.2 mM DTT, 0.02% BSA, and 2 mM Ca 2ϩ /Na ϩ EGTA buffer to achieve a free Ca 2ϩ concentration of 60 nM). Following the addition of fluorescently labeled PLC-␤3, samples were incubated at 25°C for 2 min, and immediately thereafter, emission scans were obtained. The assay medium for the FRET measurements is the same as that used in steady-state PLC assays except that PIP 2 and the EGTA/Ca 2ϩ buffer were omitted in most binding experiments (see "Results" and Table 1). In some experiments, phospholipid vesicles were omitted entirely or replaced with 0.2% cholate (Table 1).
For equilibrium competitive binding experiments, GTP␥Sbound G␣ q -E249C labeled with Alexa Fluor 488 (G␣ q -F-GTP␥S) was incubated with PLC-␤3-Q871C labeled with Alexa Fluor 594 (PLC-␤3-Q) in the presence of PE/PS (4:1 molar ratio) vesicles in binding assay buffer A for 30 min on ice. Thereafter, unlabeled PLC-␤3 was added, samples were incubated at 25°C for 5 min, and emission scans were obtained.
Binding and competition data were fit as indicated using the Marquardt-Levenberg algorithm in SigmaPlot. For competitive inhibition of binding, IC 50 values from fits to a single-site competition equation were converted to K i , the equilibrium K d for the competing ligand (unlabeled PLC-␤3), using the formula, where Y is ligand whose concentration is held constant (PLC-␤3-Q), and K d,Y is its equilibrium dissociation constant for binding to G␣ q -F-GTP␥S under the same conditions.

FCS measurements
FCS (26) was used to measure both G␣ q binding to PLC-␤3 and the binding of G␣ q and PLC-␤3 to phospholipid vesicles. FCS measurements were performed using a LSM 880 microscope (Zeiss) equipped with a C-Apochromat ϫ40, water-immersion objective with a numerical aperture of 1. In the autocorrelation function G(), fluorescence fluctuations, ␦F(t), defined as the difference between the fluorescence at time t and the time-averaged fluorescence, ͗F͘, are autocorrelated with fluorescence fluctuations at time t ϩ , where is the lag time. G() decays to 1 as approaches infinity. Unless otherwise stated, time-dependent fluorescence intensities of a sample were collected over 5 s, 40 -50 times in a single experiment. G() was computed for each 5-s acquisition and averaged over the set of 40 -50 acquisitions. To obtain diffusion coefficients, average G() was fit using Origin to an equation containing one triplet state term and one diffusion term (Equation 2) (43), where F is the mean fraction of fluorophores in the triplet state; T is the triplet state lifetime in s; A is the amplitude, which is equal to the inverse of the time-averaged number of molecules G␣ q binding to phospholipase C-␤3 in the detection volume; D is the diffusion coefficient in m 2 /s; o is the lateral radius of the detection volume in m; and S is the ratio of the axial to lateral radii (S ϭ z o / o ). Rhodamine 110 was used to determine the dimensions of the FCS detection volume at the beginning of each experiment. FCS curves for rhodamine 110 were fit to Equation 1 using D ϭ 400 m 2 /s at 22°C (44), with o , S, T , and F being free parameters. Typical o values were 0.20 -0.24 m, whereas S values ranged from 6 to 10. For rhodamine 110, T , and F were 6.0 Ϯ 1.7 ϫ 10 Ϫ6 s and 0.14 Ϯ 0.03 (n ϭ 8), respectively. S and o were then fixed when FCS data of fluorescently labeled proteins or vesicles were fit to Equation 2, with D, T , and F as free parameters. For GDP-or GTP␥S-bound G␣ q -F, either alone or in the presence of PLC-␤3, T and F were found to be 2.9 Ϯ 0.8 ϫ 10 Ϫ5 s and 0.23 Ϯ 0.02 (n ϭ 12), respectively. For PLC-␤3-F, T and F were 4.8 Ϯ 0.6 ϫ 10 Ϫ5 s and 0.14 Ϯ 0.02 (n ϭ 3), respectively. DiO-labeled PE/PS (4:1) vesicles yielded T and F values of 8.4 Ϯ 0.3 ϫ 10 Ϫ6 s and 0.24 Ϯ 0.05 (n ϭ 4), respectively.
To measure equilibrium binding of G␣ q to PLC-␤3, FCS curves were obtained for G␣ q -F (5-10 nM), bound to either GTP␥S or GDP, at different concentrations of unlabeled PLC-␤3 in 50 mM NaHEPES (pH 7.5), 100 mM NaCl, 4 mM MgCl 2 , 0.1 mM DTT, 0.2 mg/ml BSA. The curves were fit to Equation 2 to obtain D, as explained above, for G␣ q -F at each unlabeled PLC-␤3 concentration. Plots of D versus PLC-␤3 concentration were fit to a single-site binding equation to obtain K d . FCS data obtained for G␣ q -F and different concentrations of PLC-␤3 were also fit to an equation with two diffusion components for free G␣ q -F and G␣ q -F bound to PLC-␤3 (similar to Equation 3). The residuals obtained from the twocomponent fit were similar to those obtained using Equation 2, and the K d values did not differ significantly.
To measure the binding of either G␣ q -F or PLC-␤3-Alx488 (labeled at residue 871) to vesicles, FCS curves of labeled proteins (5-10 nM) were obtained in the presence of PE/PS (4:1) vesicles in 50 mM NaHEPES (pH 7.5), 100 mM NaCl, and 0.2 mM NaEGTA. Ca 2ϩ and Mg 2ϩ were omitted to minimize vesicle fusion during the experiment. To account for vesicles with multiple bound proteins (45,46), autocorrelations (G m ()) obtained for a mixture of fluorescent proteins and non-fluorescent vesicles were fit to the following, where the parameters are as described above. Parameters with subscript p pertain to free protein in the mixture, and those with subscript v pertain to protein bound to vesicles. In fitting data to Equation 3, F p , T,p , and D p were fixed at values obtained for labeled protein alone. D v was fixed at 4.3 m 2 s Ϫ1 , the diffusion coefficient of DiO-labeled PE/PS (4:1) vesicles (Fig. 7). Therefore, the only free parameters were A p , A v , F v , and T,v .
The fraction, f, of protein bound to vesicles was determined using the following equation, where A p,o is the amplitude of free labeled protein at the same concentration, with no vesicles. When D v was fixed to the diffusion coefficient obtained for DiO-labeled PE/PS (4:1) vesicles, residuals of resulting fits were correlated (see Fig. 7). Better fits were obtained when D v was allowed to float (Fig. 6). Allowing D v to float yielded an average value of 6.3 Ϯ 0.7 m 2 s Ϫ1 , as opposed to 4.3 m 2 s Ϫ1 , for the diffusion coefficient of vesicle-bound G␣ q -F and PLC-␤3-Alx488. The calculated fractions of G␣ q -F and PLC-␤3-Alx488 bound to vesicles were similar regardless of whether autocorrelation curves obtained for the two labeled proteins in the presence of vesicles were fit to Equation 3 with D v as a fixed or free parameter.

Stopped-flow measurements
The kinetics of binding of G␣ q -F-GTP␥S to PLC-␤3-Q was measured by following changes in donor emission using a Bio-Logic SFM-4 stopped-flow fluorometer. Excitation was at 488 nm, and emission was measured at 520 nm. With total flow rates set at 1 ml/s, dead times of 150 ms were achieved. All measurements were performed at 25°C in binding assay buffer B (50 mM NaHEPES (pH 7.5), 100 mM NaCl, 1 mM MgCl 2 , 0.2 mM DTT, 0.02% BSA) with PE/PS vesicles. Ca 2ϩ was omitted, and Mg 2ϩ was reduced compared with binding assay buffer A to reduce vesicle fusion during the course of the measurements.
To measure the association rate, PLC-␤3-Q (150, 300, and 450 nM final) was combined with GTP␥S-activated G␣ q -F (5 or 25 nM final), and the quenching of donor fluorescence was measured for 60 s with a 20-ms integration time. These data were fit to an equation with a single exponent and a linear term to account for photobleaching that is seen with G␣ q -F alone, f ϭ f 0 ϩ fe Ϫkt ϩ ct (Eq. 5) where f 0 is the fluorescence at time 0, f is amplitude associated with rate constant k, and c is the rate of photobleaching. PLC-␤3-Q does not alter c.
To measure the dissociation rate, GTP␥S-activated G␣ q -F (5 or 25 nM) was preincubated with PLC-␤3-Q (250 nM final) and PE/PS vesicles in binding assay buffer B. This mixture was combined with an equal volume of wild-type unlabeled PLC-␤3 (1800 nM final), and the dequenching of the donor emission was measured for 60 s with a 20-ms integration time. Plots of donor emission versus time were initially fit with an equation with either one or two exponential terms, f ϭ f 0 ϩ iϭ1 n f i ͑1 Ϫ e Ϫkit ͒ (Eq. 6) where f 0 is the fluorescence at time 0 and f i is the amplitude associated with rate constant k i. The number of exponential terms is denoted by n. An F-test was performed to determine whether the two fits (n ϭ 1 and n ϭ 2) were statistically distinguishable (47). We decided that if the equation with fewer G␣ q binding to phospholipase C-␤3 parameters worsened the quality of the fit by Ͼ1, the null hypothesis would be rejected. In other words, if the following were true, the equation with more parameters would be a significantly better fit for the data. In the Equation 7, p1 2 and p2 2 are the 2 goodness-of-fit statistics for fits with one and two exponential terms, respectively. N is the number of analyzed data points, whereas p 1 and p 2 are the number of free parameters associated with fits containing one (n ϭ 1) and two (n ϭ 2) exponential terms, respectively. The number of free parameters for an equation with n exponential terms, p n , is p n ϭ 2n ϩ 1. F (N Ϫ p1),(N Ϫ p2) ␣ is the (1 Ϫ ␣) one-sided F statistic with ␣ ϭ 0.683 (1) and N Ϫ p 1 and N Ϫ p 2 degrees of freedom. In these experiments, N Ϫ p 1 and N Ϫ p 2 were 2989 and 2987, respectively, yielding an F statistic of 1.02. Fitting the data to Equation 6 with n ϭ 1 and n ϭ 2 yielded 2 statistics that satisfied the expression in Equation 7. Therefore, it was determined that the data were best described by an equation with two exponential terms.

Phospholipase C assay
PLC-␤ activity was measured by monitoring hydrolysis of [ 3 H]PIP 2 presented in unilamellar vesicles (PE/PS/PIP 2 , 16:4:1 molar ratio; 0.25 mM total phospholipid) as described previously (7). PLC-␤3 H332A,H379A, a catalytically inactive mutant, was used as a competitive inhibitor of the interaction of wild-type PLC-␤3 with G␣ q -GTP␥S (see Fig. 8). To measure the rate of PLC-␤ deactivation, G␣ q -GTP␥S was incubated with substrate vesicles for 30 min on ice in the presence of 60 nM free Ca 2ϩ . The assay was initiated by the addition of PLC-␤3, and [ 3 H]IP 3 production was measured at 10-s intervals for 50 s. At that point, 600 M PLC-␤3 H332A,H379A was added, and [ 3 H]IP 3 production was monitored for an additional 4 min. The PLC-␤ deactivation time, equal to 1/k deact , is the difference between the time at which PLC-␤3 H332A,H379A is added and the time point at which the linear extrapolations of initial and final reaction rates intersect (30).
PLC-␤ activity was also measured during steady-state receptor signaling using reconstituted vesicles that contained [ 3 H]PIP 2 , trimeric G q , and m1 muscarinic acetylcholine receptor as described by Biddlecome et al. (16). To measure the rate of PLC-␤ deactivation after receptor deactivation in this system, [ 3 H]IP 3 production was measured at 10-s intervals for a total of 40 s to establish a baseline stimulated rate. Atropine and excess GDP were then added to stop receptor signaling, and [ 3 H]IP 3 production was measured for an additional 1-2 min. The PLC-␤ deactivation time, 1/k deact , is the difference between the time at which atropine is added and the time at which the linear extrapolations of the pre-and post-atropine reaction rates intersect (16). The fraction of accessible PIP 2 that was hydrolyzed at the time of atropine addition was at most 30%; the decline of the rate did not reflect substrate depletion (7).
Author contributions-P. N. and E. M. R. designed the project and experiments, interpreted data, and prepared the manuscript. P. N. conducted the experiments and analyzed data. A. G. helped plan, analyze, and interpret FCS experiments and data.