Gβγ activates PIP2 hydrolysis by recruiting and orienting PLCβ on the membrane surface

Significance GPCRs are major mediators of transmembrane signal transduction, responding to a wide range of stimuli including hormones and neurotransmitters. Important targets of GPCR signaling, PLCβ enzymes catalyze the hydrolysis of PIP2 into IP3 and DAG, leading to increased intracellular Ca2+ levels and activation of PKC, respectively. PLCβs exhibit very low basal activity through multiple mechanisms of autoinhibition and are activated by both Gαq and Gβγ . In this study, we demonstrate that Gβγ activates PLCβ by recruiting it to the membrane where its substrate PIP2 resides and by orienting its active site. This activation mechanism permits robust and rapid activation of PLCβ upon GPCR stimulation in the setting of low background activity during GPCR quiescence.

membrane recruitment, and atomic structures on lipid membrane surfaces, to analyze the mechanism by which G activates PLCβs.

Results
To explain with accuracy our data analysis, we present a series of equations and their rationale. At least a qualitative understanding of these equations is required to fully appreciate the meaning and wider significance of the data, and what it implies about the molecular mechanisms crucial for PLCβ3 function. Some of the analysis and associated equations are, to our knowledge, unfamiliar to biochemical analysis. In particular, when analyzing both the kinetics of PIP2 hydrolysis on a membrane surface and the equilibrium binding reaction between proteins on a membrane surface, we encountered the complex issue of processes occurring in 2 dimensions that involve components in 3 dimensions. We dealt with this issue in a particular way, which we describe thoroughly to stimulate debate and invite critique. We appreciate that many readers will want to grasp the biological implications of this work without getting bogged down by equations. For this reason, we have explained the meaning of each equation in words, which should be sufficient to understand the main conclusions of this work.

Development of a Planar Lipid Bilayer Assay for PLCβ3 Function.
We developed a detergent-free, planar lipid bilayer assay to measure PLCβ3 function using a PIP2-dependent ion channel to report its concentration over time ( Fig. 1 B-D). Briefly, two chamber cups were connected in the vertical configuration by a ~250 m hole in a 100 m piece of Fluorinated ethylene propylene copolymer (30). A ground electrode was placed in the Cis chamber and a reference electrode in the Trans chamber (Fig. 1B). Lipids dispersed in decane were used to paint a bilayer over the hole separating the two chambers. We used a 2:1:1 mixture of 1, 2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE): 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC): 1palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS) lipids and included a predetermined mole fraction of long-chain PIP2 inside the membrane to set its starting concentration. Ion channels and G proteins were incorporated by proteolipid vesicle application to the bilayer, and the current from reconstituted ion channels was measured (30). We added PLCβ3 to the Cis chamber, which was subjected to continuous mixing to ensure homogeneity of the chamber.
The PIP2-dependent, G protein-dependent inward rectifier K + channel-2 (GIRK2, specified as GIRK) was used as a readout of PIP2 concentration. This channel is well characterized in vitro, strictly depends on PIP2 for channel opening, and is amenable to measuring large macroscopic currents using planar lipid bilayers (31,32). (D) Representative current recovery upon reactivation of incorporated channels with shortchain C8PIP2. This experiment was carried out under subsaturating longchain PIP2 (1.0 mol % ) in the bilayer, which correlates to ~30% of maximal GIRK current. In D, saturating C8PIP2 was added, which leads to ~3× the amount of starting current.
PIP2 , which permits the measurement of PLCβ3 catalytic activity that is not filtered by a slow channel response (31). Experiments were carried out in the presence of symmetric MgCl 2 to ensure blockage of channels with their PIP2 binding sites facing the Trans chamber, which is not accessible to PLCβ3 (Fig. 1B) (33). This ensures that when positive voltage is applied to the reference relative to the ground, the current is derived only from channels accessible to PLCβ3 added to the Cis chamber. GIRK also requires G for channel activity. To separate the effects of G on channel function and PLCβ3 activity we used the ALFA nanobody system (34) to tether soluble G to GIRK (35). We tagged GIRK with the short ALFA peptide on the C-terminus and G with the ALFA nanobody on the N-terminus in the background of the C68S mutant, which prevents lipidation of G . Nanobody-tagged G assembled normally with G and was able to bind to other effectors (35). Because the ALFA nanobody binds to the ALFA tag with ~30 pM affinity (34), at 30 nM concentration, the ALFA nanobody-tagged G fully activates ALFA peptide-tagged GIRK. In addition, the nanobody-tagged G does not activate PLCβ3 due to its lack of a lipid anchor (13,29).
Human PLCβ3 was used to establish our assay owing to its significant activation by both G and G q (14,36,37). The addition of PLCβ3 to membranes already containing lipidated G , following an equilibration period of about 2 s, led to a rapid current decay that was complete in ~20 s (Fig. 1C). Subsequent addition of 32 M C8PIP2, an aqueous-soluble, short chain version of PIP2, rescued the current to a maximum level ( Fig. 1D) (31), indicating that the current decay was due to PIP2 depletion from the bilayer by the PLCβ3 enzyme. The PLCβ3 mediated current decay was slower than when C8PIP2 is rapidly removed by perfusion (31). Furthermore, the rate of PLCβ3-mediated current decay depends on the PLCβ3 concentration (SI Appendix, Fig. S1). These findings indicate that the decay measures the rate of PLCβ3 catalytic activity rather than PIP2 unbinding from the channel. No change in the current was observed following PLCβ3 addition in the absence of CaCl 2 (2 mM EGTA), which is required for enzymatic function (SI Appendix, Fig. S1A). Repetitions of these experiments yielded consistent results with very similar time courses of current decay. These observations indicate that we can measure PLCβ3 catalytic activity using this system and that the addition of PLCβ3 does not induce artifacts to the bilayer or to reconstituted GIRK channels.
Kinetic Analysis of PIP 2 Hydrolysis by PLCβ3. The interfacial nature of PLCβ3 activity presents a challenge to the study of its function because PLCβ3 is a soluble enzyme that must associate with the membrane to carry out catalysis. To describe the reaction occurring at the two-dimensional membrane surface, which must account for the exchange of PLCβ3 with the three-dimensional water phase, we give concentrations as dimensionless mole fraction × 100 ( mf , expressed as mol % ) using square brackets, [quantity], unless specified as molar units using square brackets with subscript molar, [quantity] molar . Furthermore, to simplify expressions, we approximate mf within each solvent phase, water or lipid, as moles solute per moles solvent rather than moles solute per moles solvent plus solute. This approximation introduces into the kinetic analysis a maximum error in mf of 1.0 % for the PIP2 concentration in membranes and less than 1.0 % for all other components. For PIP2 in membranes, the initial mf is predetermined through the bilayer lipid composition. For PLCβ3, the mf in membranes is calculated from that in three-dimensional solution using its partition coefficient, which is described below.
The measured current decays can be converted to PIP2 decays using the PIP2 concentration dependence of the channel, which we determined using titration experiments. Bilayers were formed with varying concentrations of long chain PIP2 from 0.1 to 4.0 mol % , GIRK-containing vesicles were fused, the current was measured, and water-soluble C8PIP2 (32 M) was added to the Cis chamber to activate the channels maximally (SI Appendix, Fig. S1 B and C) (31). The measured current was normalized to the maximally activated current, I max , for each PIP2 concentration and fit to a modified Hill equation, Eq. 1, to determine values A, k, and r ( Fig. 2A): Eq. 1 is an empirical function whose utility is to convert GIRK current into PIP2 concentration. In subsequent experiments with PLCβ3, bilayers initially contain 1.0 mol % PIP2 , which corresponds to ~30% of the maximal current ( Fig. 2A).
The PLCβ3/Gβγ-dependent current decays were corrected by subtracting a constant current value representing nonspecific leak, then normalized to the starting PIP2 concentration (1.0 mol % ), and converted to PIP2 concentration decays using Eq. 1 with the predetermined values for k , A , and r (Fig. 2B). After an approximately 2 s delay associated with mixing of PLCβ3, PIP2 decays contained two components: an initial, approximately linear component followed by a slower, approximately exponential component (Fig. 2C). The linear component is consistent with PLCβ3 catalysis occurring as a 0th order reaction, where the catalytic rate is independent of the PIP2 concentration, suggesting that at our starting concentration (1.0 mol % PIP2 ), the active site of PLCβ3 is nearly fully occupied by substrate ( PIP2 ). The second, exponential, component is consistent with the PIP2 concentration becoming limiting to catalysis, a first-order reaction, as the decay progresses and the concentration of PIP2 decreases. In the example shown, for illustrative purpose, we estimated the rate within six intervals along the decay curve, demarcated with different colored circles (Fig. 2C), by measuring the slope to approximate d [PIP2] dt within each interval, and then plotted the slope's absolute value against the average PIP2 concentration for the corresponding interval (Fig. 2D). A Michaelis-Menten equation (Eq. 2, below) fit the data points with R 2 ~ 0.99, indicating that PLCβ3 catalytic activity can be described by this kinetic rate equation (Fig. 2D).
The graphical procedure described above and in Fig. 2 C and D was used as an example to place the PIP2 hydrolysis data into a familiar form of rate as a function of substrate concentration. For processing all data, we took a more direct approach to analyze the time-dependent decays within the Michaelis-Menten framework. Expressing the Michaelis-Menten rate equation as and integrating from t = 0, we obtain for the PIP2 concentration as a function of time [3] [PIP2(t )] = K M ProductLog e where [PIP2(0)] is the PIP2 concentration at t = 0 and K M and V max are the Michaelis-Menten parameters. Eq. 3 derived here contains a well-known function called the Lambert W function or ProductLog function (38). It describes for the PIP2 concentration an initially linear decay followed by an exponential decay. Substituting Eq. 3 into Eq. 1, we obtain an expression for GIRK current decay as a function of time due to PIP2 hydrolysis, which permits direct fitting of the normalized current decays to estimate V max and K M (Fig. 2E). [PIP2(t )] in Eq. 4 is given by Eq. 3, and a third free parameter, C , accounts for the level of background leak in bilayer experiments; this is visible as the small residual current (typically ≤ 5% of the GIRK current) at long times in Figs. 2E and 3A. [PIP2(0)], the initial PIP2 concentration, is specified by the bilayer composition and A , k, and r are predetermined through the fit of Eq. 1 to the data shown in Fig. 2A. Eq. 4 fits the current decay data accurately after ~2 s (Fig. 2E) and yields consistent results for V max (0.17 ± 0.02 mol % ∕sec ) and K M (0.42±0.05 mol % ) across repeated experiments (Fig. 3C).
The Role of Gβγ in the Function of PLCβ3. In the experiments described above, G was added to the planar lipid bilayers by equilibrating lipid vesicles containing G with the bilayer surface prior to the application of PLCβ3. When G is not added to the bilayer, PLCβ3 produces a much slower current decay, as shown ( Fig. 3A and SI Appendix, Fig. S1 D and E). Similarly, in the presence of 1 µM aqueous-soluble G without a lipid anchor, which does not partition onto the membrane surface (31), PLCβ3 catalyzed current decay is also slow (Fig. 3 B and C). Seven experiments were carried out in the absence of G and the rmsd between the current decay curves and Eq. 4 were minimized to yield V max (0.0026 ± 0.0007 mol % ∕sec ) and K M (0.43 ± 0.05 mol % ) (Fig. 3C). Thus, G in the membrane increases V max ~65fold without affecting K M (Fig. 3C).
Because PLCβ3 is soluble in aqueous solution but must localize to the membrane surface to catalyze PIP2 hydrolysis, we next examined whether G in the membrane influences PLCβ3 membrane localization. As detailed by White and colleagues, protein association with membranes cannot be considered as a simple binding equilibrium due to the fluid nature of the membrane without discrete binding sites (39). Instead, membrane association must be treated as a partitioning process between two immiscible solvents, the membrane and the aqueous solution. The equilibrium partition coefficient, K x , is the ratio of the mole fraction of PLCβ3 in the membrane (subscript m) to that in aqueous solution (subscript w) (39), To determine the value of K x , detergent-free liposomes were reconstituted using 2DOPE:1POPC:1POPS lipids to match the lipid composition of the bilayer experiments, and H + NMR was used to measure the lipid concentration at the end of the detergent removal process (SI Appendix, Fig. S2A). Large unilamellar vesicles (LUVs) were prepared from the reconstituted liposomes using freeze-thaw cycles and extrusion through a 200 nm membrane. The LUVs were incubated with PLCβ3 and pelleted using ultracentrifugation to separate the membrane-bound and aqueous protein fractions. This method allows direct measurement of both the bound and free protein using fluorescently labeled PLCβ3, which facilitates determining the partition coefficient from each experiment individually (39). The membrane-associated fraction of PLCβ3, fraction partitioned ( F p ), is [W ] molar , the molar concentration of water, is ~55 M and [L] molar , the molar concentration of lipid, is set for each experiment using a stock measured by NMR. Thus, Eq. 6 is a function of the single free parameter, K x , which we determine by fitting Eq. 6 to the partitioning data, yielding K x ~2.9 ⋅ 10 4 ( Fig. 3D, black curve).
Partitioning experiments carried out with unlabeled PLCβ3 quantified using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis yielded a similar value of K x ( ∼ 4 ⋅ 10 4 ) (SI Appendix, Fig. S2 B and C), confirming that the fluorescent label does not alter the partitioning behavior of PLCβ3. LUVs with the same lipid composition were also prepared containing G , which is exclusively membrane bound, at a protein to lipid ratio of 1:5 (wt:wt), corresponding to 0.34 mol % , to match the concentration of G in vesicles equilibrated with planar lipid bilayers in the kinetic experiments. At this concentration of G , we observe that PLCβ3 binds to vesicles much more readily than in the absence of G (Fig. 3D). This observation is explicable if, when PLCβ3 partitions onto the membrane surface, it binds to G . Writing the binding reaction on the membrane surface as 3E). (Note that subscript m indicates PLC 3 on the membrane. Since G only resides on the membrane, a subscript is not used for When equilibrium is reached, the membrane surface will contain a quantity of PLC 3 in the membrane that is not bound to G , set by K x and the aqueous solution concentration of PLC 3 , as well as a quantity of PLC 3 in the membrane that is bound to G (i.e., PLC 3 ⋅ G ) , set by the membrane concentrations of PLC 3 , G and K eq . Therefore, in the presence of a total quantity of G on the membrane, , the fraction of PLC 3 on the membrane surface, unbound plus bound to G , is given by (SI Appendix 2)  Data for 0 G were fit to Eq. 6 for K x (dashed black curve) and data for +G were fit to Eq. 7 to determine K eq (39). Error bars are range of mean from two experiments for each lipid concentration. R 2 = 0.96 in the absence of G and R 2 = 0.95 in the presence of G . (E) Cartoon representation of PLCβ3 activation by G through membrane recruitment. G significantly increases the membrane association of PLCβ, and accordingly [PLCβ] membrane , which amplifies PIP2 hydrolysis.
(F) Calculated Michaelis-Menten curves (from Eq. 2) for PLCβ3 alone (black), in the presence of 1 M soluble G (blue) or in the presence of lipidated G (pink) using the values for K M and V max determined from our fits. measurements in the absence of G (Fig. 3D), the right-hand side of Eq. 7 contains a single free parameter, K eq , for the binding of PLC 3 to G on the lipid membrane surface. The red dashed curve in Fig. 3D corresponds to K eq = 0.0090 mol % . It may seem at first surprising that the series of partitioning experiments in the presence of G , with knowledge of K x for PLC 3 in the absence of G , uncovers the equilibrium reaction between PLC 3 and G on the membrane surface. Nevertheless, the binding reaction is discernable by this approach, and the inescapable conclusion is that G concentrates PLC 3 on the membrane surface (Fig. 3E).
The PLC 3 -concentrating effect of G has obvious implications for interpreting the kinetic data reported above, which show that G increases V max by a factor ~65, without affecting K M very much (Fig. 3C). From Eq. 2, V max is the asymptotic rate of PIP2 hydrolysis when [ PIP2 ] far exceeds K M . In this limit, the maximum rate of hydrolysis, V max , is given by the total membrane concentration of PLC 3 times k cat , the turnover rate of a PLC 3 ⋅ PIP2 complex. In the bilayer chamber used for the kinetic experiments, the volume of the aqueous solution is so large compared to the small area of the lipid bilayer that surface binding does not significantly alter [PLC 3 w ] . Under this condition, we have is the membrane concentration in its pres- is a multiplier giving the fold-increase in total membrane PLC 3 concentration due to the presence of G at concentration [G tot ] . When the known quantities are entered for our experimental conditions, this factor is ~33. In the kinetic experiments, we observed a 65-fold increase in V max in the presence of G . Eq. 8 predicts a 33-fold increase through G ′ s ability to increase the local concentration of PLC 3 on the membrane surface. A mere two-fold increase in k cat produced by G binding to PLC 3 would account for the full enhancement of V max in the kinetic experiments (Fig. 3C). The important conclusion is that most of the increase in V max (within a factor of ~2) is explained by the ability of G to concentrate PLC 3 on the membrane surface. Indeed, it seems very possible that the ~two-fold shortfall is accountable by the ability of G to orient PLC 3 , in addition to concentrating it. Using a conventional Michaelis-Menten plot, with the V max and K M values derived experimentally, we observe that at concentrations in our assay, G essentially switches the PLC 3 enzyme on (Fig. 3F), and this effect is due largely to the ability of G to concentrate PLC 3 on the membrane surface.
In summary, the kinetic studies show that PLC 3 catalyzes PIP2 hydrolysis with a substrate concentration dependence like that of a Michaelis-Menten enzyme (Fig. 2 C-E). We note that K M corresponds to the mid-range of known PIP2 concentrations in cell membranes (Figs. 2D and 3F) (40,41). PLC 3 aqueousmembrane partition studies show that G concentrates PLC 3 on the membrane surface, enough to account for most of the effect on V max (Fig. 3 C and D). To a smaller extent (~twofold), G augments V max through k cat (Fig. 3C). Next, we evaluate the structural underpinnings of these functional properties.

Structural Studies of PLCβ3 in Aqueous Solution by Cryo-EM.
We next determined the structure of PLCβ3 in aqueous solution using cryo-EM. The structure, consisting of the PLCβ3 catalytic core at 3.6 Å resolution, contained the PH domain, EF hands, X and Y domains, the C-terminal part of the X-Y linker, the C2 domain, and the active site with a Ca 2+ ion bound (Fig. 4 A  and B and SI Appendix, Fig. S3 and Table S1). The autoinhibitory Hα2′ element in the proximal CTD was also resolved, bound to the catalytic core between the Y domain and the C2 domain, as proposed by Lyon and colleagues (Fig. 4 A and B) (16,21) but not the distal CTD. We also obtained several low-resolution reconstructions with varying levels of density corresponding to the catalytic core and distal CTD with differing arrangements between the two domains (SI Appendix, Fig. S3F). This observation suggests that the distal CTD is disordered rather than proteolyzed in our final reconstruction and that the two domains are flexible with respect to each other, as previously proposed (19). The catalytic core resolved by cryo-EM is very similar to the crystal structure with a Cα rmsd of 0.6 Å if the Hα2′ helix is excluded (Fig. 4C). We note that, as in the crystal structure, the autoinhibitory X-Y linker occludes the active site (Fig. 4C). We attempted to determine a structure of PLCβ3 in complex with G in solution, in the presence or absence of detergent, without success. Furthermore, we were unable to detect the formation of a complex in solution by size-exclusion chromatography (SI Appendix, Fig. S3G).

Structural Studies of PLCβ3
Associated with Liposomes. We next determined the structure of PLCβ3 bound to liposomes consisting of 2DOPE:1POPC:1POPS. PIP2 was omitted from these samples because it would have been degraded by PLCβ3 prior to grid preparation. We obtained a low-resolution reconstruction with the distal CTD associated with the membrane and the catalytic core located away from the membrane surface ( Fig. 4C and SI Appendix, Fig. S4 and Table S1). Although the map was low resolution, previously determined structures fit into the density for each domain and all reconstructions showed the same orientation of the protein on the membrane surface (SI Appendix, Fig. S4). The interaction of the distal CTD with the membrane is consistent with previous reports of its involvement in membrane association (3). The position of the catalytic core indicates that significant rearrangements of PLCβ3 with respect to the membrane must be involved in activation because the active site is too far from the membrane to access PIP2 . Activating rearrangements could be mediated by interactions of lipid-anchored G proteins with the PLCβ3 catalytic core.

The PLCβ3 · Gβγ Complex on Liposomes Reveals Two Gβγ Binding
Sites. We reconstituted G into liposomes consisting of 2DOPE:1POPC:1POPS at a protein to lipid ratio of 1:15 (wt:wt) and incubated the liposomes with purified PLCβ3 prior to grid preparation. We determined the structure of the PLCβ3· Gβγ complex to 3.5 Å and observed two Gβγs bound to the catalytic core of PLCβ3 (Fig. 5 A-C and SI Appendix, Fig. S5 and Table S1). The distal CTD is not resolved in our reconstructions, suggesting that it might adopt many different orientations on the plane of the membrane relative to the catalytic core, in agreement with previous studies showing that heterogeneity in the distal CTD increases upon G binding (42). The catalytic core is very similar to our cryo-EM structure without membranes, with a Cα rmsd of 0.7 Å. Only small rearrangements occur at the G binding sites (SI Appendix, Fig. S6A). Both autoinhibitory elements, the Hα2' and the X-Y linker, are engaged with the catalytic core (Fig. 5C) consistent with previous proposals that G does not play a role in relieving this autoinhibition (15,16,21).
One G is bound to the PH domain and the first EF hand, referred to as G 1, and the other is bound to the remaining EF hands, referred to as G 2 ( Fig. 5C and SI Appendix, Fig. S6  A and B). Both interfaces are extensive, with the G 1 interface burying ~800 Å 2 and involving 34 residues, (16 from PLCβ3 and 18 from G ) and the G 2 interface burying ~1,100 Å 2 and involving 44 residues (21 from PLCβ3 and 23 from G ) (Fig. 5 D and E and SI Appendix, Fig. S6B and Table S2). The G 1 interface is mostly composed of hydrophobic interactions, with three hydrogen bonds ( Fig. 5F and SI Appendix, Fig. S6C), whereas the G 2 interface is mostly composed of electrostatic interactions, including 10 hydrogen bonds spanning the length of the interface (Fig. 5 G and H). Both interfaces involve the same region of G that interacts with G and several residues on Gβ shown to be important for PLCβ activation are involved (43) (Fig. 5 E and F). Specifically, L117 and W99 on Gβ 1 form hydrophobic interactions with L40, I29, and V89 on PLCβ3 ( Fig. 5F and SI Appendix, Table S2) (43). On Gβ 2, W99 forms a hydrogen bond with E294 on PLCβ3, W332 forms an anion-edge interaction with D227, M101 and L117 form hydrophobic interactions with P239 and F245 on PLCβ3, and D186 forms a hydrogen bond with Y240 on PLCβ3 (Fig. 5 G and H and SI Appendix, Table S2) (43).
We also determined the structure of the PLCβ3 · Gβγ complex using lipid nanodiscs. We reconstituted G into nanodiscs formed using the MSP2N2 scaffold protein (44) and 2DOPE:1POPC: 1POPS lipids and incubated them with purified PLCβ3 prior to grid preparation. We observed only reconstructions with two Gβγs bound and determined the structure of the complex to 3.3 Å (SI Appendix, Fig. S7). The two G s are bound in the same locations as was observed in liposomes with comparable interfaces (SI Appendix, Fig. S6D). A model for this structure aligns well to the model built using the lipid vesicle reconstruction with a Cα rmsd of 0.8 Å for all proteins (SI Appendix, Fig. S6D). These structures suggest that the PLCβ3 · Gβγ complex depends on a membrane environment as we were unable to form a stable complex in solution with or without detergent, which highlights the importance of the membrane in Gβγ-dependent activation of PLCβ3.

Gβγ Mediates Membrane Association and Orientation of
the PLCβ3 Catalytic Core. Unmasked refinement of our final subset of particles from the liposome structure yielded a 3.8 Å reconstruction showing the PLCβ3 · Gβγ assembly and density from the membrane (Figs. 5B and 6A). The two Gβγs and the region of PLCβ3 between them are closely associated with the membrane and the remainder of the catalytic core, including the active site, tilts away from the membrane (Fig. 6A). Despite the tilting, the structure reveals significant rearrangement of the catalytic core with respect to the membrane compared to its position in the absence of G , where it was separated from the membrane surface by a larger distance (Fig. 6A).
Additional 2D and 3D classification without alignment revealed heterogeneity in the position of the PLCβ3 · Gβγ assembly with respect to the membrane (Fig. 6 B-D). 2D classes show large variation in the orientation of the catalytic core with respect to the membrane surface, with some classes showing the entire catalytic core engaged with the membrane (Fig. 6B). The 2D classes also reveal differences in membrane curvature originating from differences in liposome size, which do not seem to be correlated with the degree of membrane tilting (Fig. 6B). 3D classification revealed four reconstructions capturing different degrees of tilting of the catalytic core ranging from ~26° to ~36° (Fig. 6D). We note that in a locally planar membrane, as opposed to a curved vesicle membrane, the active site would be nearer the membrane surface in all classes, but the variability in orientation would presumably still exist. The protein components of these reconstructions are like in the original reconstruction, with no internal conformational changes, indicating that the whole complex tilts on the membrane as a rigid body.
The lack of conformational changes observed upon G binding and the catalytic core membrane association are consistent with our functional studies showing that activation by G is largely mediated by increasing membrane partitioning. Our structures suggest that the configuration of the two G binding sites maintains the catalytic core at the membrane and increases the probability of productive engagement with PIP2 , potentially mediated by orientation of the catalytic core observed in our  (19)] and the structure determined using cryoEM without membranes (colored by domain). Cα rmsd is 0.6 Å. Calcium ion from the cryoEM structure is shown as a yellow sphere, and the active site is denoted with an asterisk. The PH domain is pink, the EF hand repeats are blue, the C2 domain is light blue, the Y domain is green, the X domain is teal, and the X-Y linker and the Hα2' are red. (D) Unsharpened reconstruction of PLCβ3 bound to lipid vesicles containing 2DOPE:1POPC:1POPS. PLCβ3 is colored in yellow and the membrane is colored in gray.
reconstructions. Taken together, our kinetic, binding, and structural studies lead us to conclude that G activates PLC mainly by bringing it to the membrane and orienting the catalytic core so that the active site can access the PIP2-containing surface (Fig. 7).

Discussion
This study aims to understand how a G protein, G , activates the PLC 3 phospholipase enzyme. We developed and applied three new technical approaches to study this process. First, because kinetic analyses of PLC enzymes historically have been limited to relatively slow radioactivity-based or semiquantitative fluorescence assays, we have developed a new higher resolution assay using a modified, calibrated PIP2-dependent ion channel to provide a direct read out of membrane PIP2 concentration as a function of time. This assay is employed in a reconstituted system in which all components are defined with respect to composition and concentration. Second, we have used a membrane-water partition assay to study a surface equilibrium reaction between two proteins ( PLC 3 and G ) on membranes. Third, we have determined structures of a protein complex ( PLC 3 and G ) assembled on the surface of pure lipid vesicles. We also determined the structures using lipid nanodiscs; however, the lipid vesicles permitted structural analysis of the enzyme-G protein complex on lipid surfaces unperturbed by the scaffold proteins required to make nanodiscs. The membrane in our nanodisc reconstructions is poorly resolved and the complex appears to be associated at nonphysiological orientations; therefore, we cannot gain any information regarding the positioning of the complex on the membrane from those reconstructions.
We list our essential findings. 1) PLC 3 catalyzes PIP2 hydrolysis in accordance with Michaelis-Menten enzyme kinetics. 2) G modifies V max , leaving K M essentially unchanged. Under our experimental conditions, V max increases ~65-fold. 3) G increases membrane partitioning of PLC 3 , an effect accountable through equilibrium complex formation between G and PLC 3 on the membrane surface. Under our experimental conditions, partitioning increases the membrane concentration of PLC 3 ~33-fold. 4) The G -mediated increase in PLC 3 partitioning can account for most of the increase in V max , with a smaller, ~two-fold, effect on k cat . Thus, G regulates PLC 3 mainly by concentrating it on the membrane. 5) Two G proteins assemble to form a complex with PLC 3 on vesicle surfaces. One G binds to the PH domain and one EF hand of PLC 3 , while the other binds to the remaining EF hands. Both G orient their covalent lipid groups toward the membrane so that the PLC 3 catalytic core is firmly anchored on the membrane surface. 6) The PLC 3 ⋅ G assembly holds the PLC 3 catalytic core with its active site, as if on the end of a stylus, poised to sample the membrane surface. Assemblies on lipid vesicles reveal multiple orientations of the catalytic core with respect to the surface. We described the formation of a complex between PLCβ and G as a two-step process: first, partitioning of PLCβ from aqueous solution into the membrane, and second, binding to G on the membrane surface. We explicitly consider two steps rather than one in which PLCβ binds directly to G for the following reasons. We measured partitioning of PLCβ into membranes without G and measured the corresponding catalysis of PIP2 in the absence of G . Thus, we know that PLCβ partitions onto the membrane surface without G . Furthermore, we find that PLCβ and G do not form a complex in the absence of a membrane,  Fig. 7. G activates PLC by increasing its concentration at the membrane and orienting the catalytic core to engage PIP2 . Upon activation of a G i coupled receptor, GTP is exchanged for GDP in the G i subunit and free G is released to bind PLCβ, which increases the concentration of PLCβ at the membrane and orients the active site for catalysis. The k cat is limited by the X-Y linker (shown in red), which occludes the active site and is only transiently displaced from the active site to allow catalysis. The distal CTD of PLCβ was omitted for clarity. neither as evaluated by size exclusion chromatography (SI Appendix, Fig. S3G) nor on cryo-EM grids. It was also shown previously that G does not activate PLCβ in the absence of membranes (17). Taken together, this set of findings support the conclusion that PLCβ partitioning is a required first step in the two-step process of PLC 3 ⋅ G complex formation on membranes. We hypothesize that partitioning orients PLCβ with respect to G , defines a local surface concentration, and thus permits a binding equilibrium process that occurs in 2 dimensions, rather than in a three-dimensional aqueous phase.
We modeled the second step, the equilibrium reaction between PLC and G on the membrane surface, as bimolecular (1:1 stoichiometry) characterized by a single K eq . In our structural analysis, however, we discovered two binding sites for G on PLC 3 . Additional binding data, using multiple concentrations of G , for example, might reveal two distinct binding constants and whether they interact with each other (i.e., behave cooperatively). Such a finding would be important because multiple binding sites could shape the PLC 3 activity response to GPCR stimulation. But for purposes of the present study, the binding model treating a single site is sufficient. This is because using a single site model when two sites exist introduces an uncertainty in how PLC 3 is distributed over G , not how much PLC 3 is present in the membrane. The kinetics depend on how much PLC 3 is present, and this we have measured directly with experiment.
The conclusion that G concentrates PLC 3 on the membrane in our assay is unequivocal. To what extent do these conclusions apply to cell membranes? From Eq. 8, we saw that the increase in membrane PLC 3 concentration due to the fraction bound to G is proportional to total G concentration, [G tot ] . In our assay, [G tot ] is 0.34 mol % , which corresponds to ~5,000 G ∕ m 2 . In cells, we have previously estimated the concentration of G near GIRK2 channels in dopamine neurons during GABA B receptor activation at ~1,200 G ∕ m 2 (32). Applying Eq. 8, this would produce an ~nine-fold increase in the membrane concentration of PLC 3 . This is an estimate with certain unknowns, especially the cytoplasmic concentration of PLC 3 ( PLC 3 w ), but the result suggests that the conditions of our in vitro assay are applicable to cell membranes. Moreover, both G and G q have been shown to increase membrane association of PLC s in cells, consistent with our results (29).
We note that our demonstration that G increases membrane association of PLC 3 directly contradicts many previous biochemical studies and the current consensus in the field that G proteins do not increase the local concentration of PLCβs in the membrane (3,4,(22)(23)(24)(26)(27)(28). We suspect that the use of detergent solubilized G in past studies may have interfered with the control of its concentration on the membrane (22)(23)(24).
While our results and mechanism contradict the notion that G proteins do not concentrate PLC s on the membrane, they are consistent with many previous observations, some we list here. As stated above, studies with cells have led to the conclusion that G and G q increase membrane association of PLC s (29). The lipid anchor is required for the activation of PLCβs by the small GTPases and G , and G proteins do not activate PLCβs in the absence of a membrane environment (9-11, 13, 15, 17, 29). The binding of Rac1 or G q do not induce conformational changes around the active site, suggesting that activation is not mediated by obvious allosteric changes (9,18,19). Likewise, we observe no change in the PLC 3 active site conformation when G is bound, only that G recruits PLC 3 to the membrane and orients its active site.
Several properties of the G binding sites on PLC 3 offer explanations of past observations. First, it has been shown that G and G q can activate PLC 3 simultaneously (36,37,(45)(46)(47)(48)(49). We find here that the G sites do not occlude the G q binding site (18,19), and therefore both G proteins can in principle bind to PLC 3 at the same time and activate PLC 3 (3,36,37,45). Second, several amino acids on G that contact PLCβ3 in the structure were previously shown to play a role in binding to G , PLCβ, and other effectors (Fig. 5 C and D) (43). Third, the PH domain was shown to play a role in G binding and activation; however, based on our structures, G binding does not require or induce rearrangement of the catalytic core as was previously proposed (50,51). Fourth, Rac1 was also shown to bind to the PH domain of PLCβ2 (SI Appendix, Fig. S6D), and Rac1-activated PLCβ was shown to be additionally activated by G , leading to a proposal that the two binding sites did not overlap (9,10). Our structures show that Rac1 and G do indeed share an interface within the PH domain (SI Appendix, Fig. S6D); however, the second G binding site can explain the dual activation (9,10).
An intriguing aspect of PLC enzymes is that all wild-type structures show that the active site is occluded by the inhibitory X-Y linker. This includes complexes with G q , Rac1 and, now, G (9,16,18,19). It has been proposed that lipids are required to remove the X-Y linker to achieve catalysis (3,16,17). This must be true to some extent because unless the linker is displaced, even if only rarely, catalysis cannot occur. From our data, we put forth an alternative proposal that the active site is predominantly autoinhibited, accounting for a small k cat , even in the presence of lipids. Consequently, in the absence of GPCR stimulation, the baseline partitioning of PLC enzyme from the cytoplasm to the membrane, determined by K x and the cytoplasmic concentration of PLC , will produce very little PIP2 hydrolysis. Only upon GPCR stimulation, when a large quantity of PLC partitions into the membrane, determined by K eq and the G concentration generated by GPCR stimulation, is there enough PLC enzyme in the membrane, even though k cat remains low, to catalyze PIP2 hydrolysis. In other words, a small k cat combined with an ability to enact large changes in membrane enzyme concentration upon GPCR stimulation permits a strong signal when the system is stimulated and a minimal baseline when it is not.

Materials and Methods
Protein Expression, Purification, and Reconstitution. All proteins were purified according to previously established protocols using affinity chromatography and size exclusion chromatography. Detailed methods are described in SI Appendix, Materials and Methods: Protein Expression and Purification and Protein Reconstitution.
PLCβ3 Functional Assay. PLCβ activity was measured using a planar lipid bilayer setup and a PIP2-dependent ion channel to report PIP2 concentration in the membrane over time. Detailed methods are described in SI Appendix, Materials and Methods: Bilayer Experiments and Analysis.
Membrane Partitioning Experiments. Fluorescently labeled PLCβ3 was mixed with LUVs and pelleted. Protein in the pellet and supernatant were quantified using fluorescence. Detailed methods are described in SI Appendix, Materials and Methods: PLCβ3 Vesicle Partition Experiments.
PLCβ3 Structure Determination. PLCβ3 was mixed with liposomes with or without Gβγ prior to sample vitrification. Cryo-EM data were collected using a Titan Krios with a Gatan K3 direct electron detector according to the parameters in SI Appendix, Table S1 and analyzed according to the procedures outlined in SI Appendix, Figs. S3-S5 and S7. Atomic models from previously determined structures were fit into our density maps, refined using PHENIX real-space refine (52), and manually adjusted. Detailed methods are described in SI Appendix, Materials and Methods: Cryo-EM Sample Preparation and Data Collection, Cryo-EM Data Processing, and Model Building and Validation. Data, Materials, and Software Availability. Cryo-EM maps and atomic models for all structures described in this work have been deposited to the Electron Microscopy Data Bank (EMDB) and the Protein Data Bank (PDB), respectively. Accession codes are as follows: PLCβ3 in solution-8EMV and EMD-28266, PLCβ3 in complex with Gβγ on vesicles-8EMW and EMD-28267, and PLCβ3 in complex with Gβγ on nanodiscs-8EMX and EMD-28268.