Intrinsic Pleckstrin Homology (PH) Domain Motion in Phospholipase C-β Exposes a Gβγ Protein Binding Site*

Mammalian phospholipase C-β (PLC-β) isoforms are stimulated by heterotrimeric G protein subunits and members of the Rho GTPase family of small G proteins. Although recent structural studies showed how Gαq and Rac1 bind PLC-β, there is a lack of consensus regarding the Gβγ binding site in PLC-β. Using FRET between cerulean fluorescent protein-labeled Gβγ and the Alexa Fluor 594-labeled PLC-β pleckstrin homology (PH) domain, we demonstrate that the PH domain is the minimal Gβγ binding region in PLC-β3. We show that the isolated PH domain can compete with full-length PLC-β3 for binding Gβγ but not Gαq, Using sequence conservation, structural analyses, and mutagenesis, we identify a hydrophobic face of the PLC-β PH domain as the Gβγ binding interface. This PH domain surface is not solvent-exposed in crystal structures of PLC-β, necessitating conformational rearrangement to allow Gβγ binding. Blocking PH domain motion in PLC-β by cross-linking it to the EF hand domain inhibits stimulation by Gβγ without altering basal activity or Gαq response. The fraction of PLC-β cross-linked is proportional to the fractional loss of Gβγ response. Cross-linked PLC-β does not bind Gβγ in a FRET-based Gβγ-PLC-β binding assay. We propose that unliganded PLC-β exists in equilibrium between a closed conformation observed in crystal structures and an open conformation where the PH domain moves away from the EF hands. Therefore, intrinsic movement of the PH domain in PLC-β modulates Gβγ access to its binding site.

tified a helix-turn-helix extension of the C2 domain that is bound to the TIM barrel in the unliganded structure but is refolded and bound to G␣ q upon complex formation (13), supporting the notion that PLC-␤ activation requires more than simple recruitment to the membrane. Mutational analysis suggested that this helix-turn-helix extension may stabilize the X-Y linker in its inhibitory conformation. Unlike the other PLC families, PLC-␤ isoforms possess an ϳ400-amino acid helical bundle C-terminal to this extension and are separated from it by a protease-sensitive and probably disordered linker. A PLC-␤ truncation mutant that lacks this helical domain binds G␣ q with about the same affinity as wild-type PLC but displays markedly reduced sensitivity to stimulation by G␣ q (14). Multiple single amino acid mutations in the helical domain also inhibit stimulation by G␣ q (15). However, the helical bundle is dispensable for stimulation by G␤␥ or Rac1 (16).
In the absence of a co-crystal structure of the G␤␥⅐PLC-␤ complex, the mechanism of PLC-␤ regulation by G␤␥ is the least well understood. G␤␥ stimulates the phospholipase activity of PLC-␤ but inhibits its GTPase-activating protein activity (17). There is no consensus on where G␤␥ binds or how it coordinates the dual regulation of PLC-␤ activities. Two binding sites for G␤␥ on PLC-␤ have been proposed. The N-terminal PH domain is required for stimulation by G␤␥. Removal of the PH domain abolishes G␤␥ stimulation without blocking either G␤␥ inhibition of G q GTPase-activating protein activity or stimulation of PLC activity by G␣ q (18). PLC-␦, an isoform not stimulated by G␤␥, gains sensitivity to G␤␥ when its PH domain is replaced by the PLC-␤ PH domain and the first EF hand (19). G␤␥ binds the PH domains of many proteins, including some with low nanomolar affinity (20), suggesting that G␤␥ may bind to a conserved surface in PH domains. The second proposed site is a conserved helix in the C-terminal half of the TIM barrel. Smrcka and co-workers (21) showed that a peptide corresponding to this region inhibits G␤␥ stimulation of PLC-␤ and that the peptide can be chemically crosslinked to G␤␥. Point mutations in this region also diminished the response to G␤␥ (21). These two sites, the PH domain and this region of the TIM barrel surface, are not adjacent in the tertiary structure of PLC-␤, and it is hard to envision how both could be involved in G␤␥ binding unless there is substantial domain movement.
In this study, we developed a FRET-based assay for the binding of G␤␥ to PLC-␤3 and to its isolated PH domain, establishing the PH domain as the minimal G␤␥ binding site in PLC-␤. We also identify a G␤␥ binding surface on the PH domain that is consistent with other known G␤␥-PH domain interfaces. This G␤␥ binding surface is buried in the available crystal structures of PLC-␤, and movement of the PH domain is therefore required for G␤␥-PLC-␤ binding and activation. We confirm this prediction by showing that anchoring the PH domain by cross-linking it to the EF hand domain blocks PLC-␤ activation by G␤␥. Based on these data, we propose a new model for PLC-␤ interaction with G␤␥ in which the PH domain is in equilibrium between distinct positions and G␤␥ binds and stabilizes a conformation needed for enzyme activation.

Experimental Procedures
Proteins-cDNA encoding PLC-␤3 (human) with an N-terminal His 6 tag was cloned into pQE60 for expression in Escherichia coli. Cysteines at positions 193, 516, 614, 892, 1005, 1176, and 1207 were predicted to be solvent-exposed and mutated to serine (1176 to valine) by multiple rounds of QuikChange mutagenesis. The activity and regulation of this modified PLC-␤3 purified from E. coli was virtually indistinguishable from wild-type PLC-␤3 expressed in Sf9 cells, and we refer to this mutant as wild-type throughout. Point mutations to introduce cysteine residues for cross-linking or fluorescence labeling were made in this background by QuikChange mutagenesis. To quantitate the fraction of intramolecular cross-linking in the Cys 60 ,Cys 164 construct, we introduced a TEV protease recognition sequence after Gly 94 by overlap PCR (22) ( 94 Gsenlyfq-gaaagP 95 ; the inserted sequence is in lowercase, and the numbers at the ends indicate native PLC-␤3 residue numbers). We refer to this construct as PLC-TEV. All mutants were verified by sequencing over the entire PLC-␤3 open reading frame. Wild-type PLC-␤3 and variants were purified from transformed BL21DE3/pRep4 cells grown in T7 medium (2% Tryptone, 1% yeast extract, 0.2% glycerol, 0.5% NaCl, 50 mM potassium P i , pH 7.2) and induced with 60 -100 M isopropyl 1-thio-␤-D-galactopyranoside for 9 h at 25°C. Frozen cell pellets were lysed with 0.5 mg/ml lysozyme in buffer A (20 mM NaHepes (pH 7.5), 5 mM 2-mercaptoethanol, 5 g/ml leupeptin, 1 g/ml aprotinin, and 0.2 mM PMSF) plus 500 mM NaCl. After sonication, the suspension was centrifuged for 30 min at 100,000 ϫ g, and the supernatant was mixed with Ni 2ϩ -nitrilotriacetic acid resin (Qiagen) for 2 h. The resin was washed sequentially with buffer A plus 500 mM NaCl and 5 mM imidazole and buffer A plus 100 mM NaCl and 5 mM imidazole until A 280 reached baseline. PLC was eluted with buffer A plus 100 mM NaCl and 120 mM imidazole. The eluate was diluted in buffer B (20 mM NaMES (pH 6.0), 0.1 mM EDTA, 1 mM DTT, 0.2 mM PMSF, and 10% glycerol) and applied to a 1-ml SourceS column that was equilibrated with buffer B. After washing the column with 200 mM NaCl in buffer B, protein was eluted by a gradient of 200 -500 mM NaCl in buffer B. Peak fractions were pooled, exchanged into buffer C (20 mM NaHepes, 100 mM NaCl (pH 7.5), 0.1 mM DTT, and 10% glycerol) and flash-frozen to Ϫ80°C for storage. N-terminally His 6 -tagged PLC-␤2 (mouse) was purified from baculovirus-infected insect cells as described earlier (8).
The PH domain (residues 2-147) of human PLC-␤3 was amplified by PCR and cloned into a pET28 vector with an N-terminal MBP tag and a TEV protease recognition sequence (a gift from Neal Alto, University of Texas Southwestern Medical Center) (23). The resulting fusion protein had the sequence His 6 -MBP-His 6 -ENLYFQSG followed by the PLC-␤3 PH domain. TEV protease cleaves after the Gln residue in ENLY-FQSG. Point mutations were made in this background by QuikChange mutagenesis. Proteins were expressed at 16°C in BL21DE3 grown in T7 medium for 16 -20 h after induction with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside. Cells were lysed in buffer A plus 500 mM NaCl. After sonication, the lysate was centrifuged for 30 min at 100,000 ϫ g. The superna-tant was bound to nickel-nitrilotriacetic acid resin (Qiagen) for 2 h. The resin was washed sequentially with buffer A plus 500 mM NaCl and buffer A plus 100 mM NaCl. Bound protein was eluted with buffer A plus 100 mM NaCl and 150 mM imidazole. To remove the MBP tag, proteins were incubated with 1% (w/w) TEV protease for ϳ16 h at 4°C, followed by removal of the cleaved His 6 -MBP by passage over amylose resin (New England Biolabs). Proteins were further purified by size exclusion chromatography on a Superdex 75 10/30 column in 20 mM HEPES (pH 7.5), 100 mM NaCl, and 1 mM DTT. Eluted protein was concentrated to ϳ3-5 mg/ml, flash-frozen in the presence of 10% glycerol in liquid nitrogen, and stored at Ϫ80°C. All PH domains were verified to be folded by thermal denaturation using SYPRO Orange dye (24). G␣ q was purified by modifying existing methods (25,26). G␣ q and His 8 -G␥ 2 were expressed from a single baculovirus using the pFastbac-Dual shuttle vector (Invitrogen). The G␣ q -G␥ 2 baculovirus was used to infect Sf9 cells along with baculoviruses encoding G␤ 1 and GST-Ric8A (a gift from Gregory Tall, University of Rochester Medical Center). Because G␣ q isolated from a G␣ q -Ric8A complex had a 50-fold higher EC 50 value for PLC-␤ activation (8,25), we only used the G␣ q that was bound to and eluted from G␤ 2 His 8 -G␥ 2 . The eluted protein contained a contaminant with phospholipase activity that was removed using a SP-Sepharose column as described above for PLC-␤3. The G␥ 2 -CFP construct was a fusion of cerulean fluorescent protein (27) to the Ala 2 residue of human G␥ 2 . G␤ 1 g 2 and Gb 1 g 2 -CFP were purified after co-expression with His 6 -G␣ i in Sf9 cells as described earlier (26). The G␤ 1 ␥ 2 isoform was used in all experiments. G␣ i and phosducin were purified from E. coli as described previously (28,29). All protein concentrations were determined by Amido Black binding (30).
The PH domain and PLC-␤3 constructs with a single reactive Cys 97 were labeled with the thiol-reactive probe Alexa Fluor 594-C 5 -maleimide. Prior to labeling, proteins were buffer exchanged into 20 mM Hepes (pH 7.5), 100 mM NaCl to remove dithiothreitol. The protein was reacted with a 10-fold molar excess of the robe on ice for 1-2 h, and the reaction was quenched with ␤-mercaptoethanol. The dye was removed by adsorption to an appropriate ion exchange resin (Q-Sepharose for the PH domain and SP-Sepharose for PLC-␤3). The resin was washed with a buffer supplemented with 200 mM NaCl before elution of protein with a buffer supplemented with 450 mM NaCl. Peak fractions were pooled, concentrated, frozen in liquid N 2 , and stored at Ϫ80°C. The labeling efficiency was determined according to the absorption of Alexa Fluor 594 at 590 nm (⑀ ϭ 96,000 cm Ϫ1 M Ϫ1 ) and the protein concentration.
PLC Assay-PLC activity was measured by monitoring the hydrolysis of [ 3 H]PIP 2 on the surface of large unilamellar vesicles (phosphatidylethanolamine:phosphatidylserine:PIP 2 , 20:4:1 molar ratio, 0.25 mM total phospholipid) at 37°C exactly as described earlier (8). Assays were initiated by the addition of PLC. For experiments that included the isolated PH domain, assays were conducted at 25°C to prevent precipitation of the PH domain. Data are expressed as moles of inositol 1,4,5-trisphosphate produced per minute per mole PLC.
Fluorescence Measurements-Equilibrium fluorescence measurements were performed on a Fluorolog3-212 spectrophotometer (Horiba Jobin Yvon) using a 3-mm cuvette with both excitation and emission slits set at 5 mm. G␤␥-CFP was diluted in 200 l of 25 mM Hepes (pH 7.5), 50 mM NaCl, 0.1 mg/ml BSA, 2 mM MgCl 2 , and 0.2% cholate. Cholate was omitted when phospholipid vesicles were used. Phospholipid vesicles were prepared using the same procedure as for the PLC assay, except that [ 3 H]PIP 2 was not included. The final concentration of lipids in the FRET binding assay was the same as in the PLC assay. Ca 2ϩ and EGTA were omitted from all buffers for FRET measurements. After the addition of acceptor, the mixture was incubated on ice for 30 min. After equilibration to 25°C for 5 min, fluorescence was recorded between 450 and 650 nm with excitation at 432 nm. FRET was measured as the decrease of G␤␥-CFP fluorescence at 475 nm over the sum of both components at the same concentrations.
Binding and competition data were fit as indicated using the Marquardt-Levenberg algorithm in SigmaPlot. For competitive inhibition of binding or activation, IC 50 values from fits to a single-site competition equation were converted to K i , the equilibrium K d for the competing ligand, using the formula where Y is ligand whose concentration is held constant (usually PH-97Alx or PLC-97Alx), and K d , Y is its dissociation constant for G␤␥ under the same conditions. Competition by G␣ i -GDP for PH-97Alx binding to G␤␥-CFP ( Fig. 2C) was fit to a single-site model as for other competitors to yield the IC 50 value. However, the high affinity of G␣ i -GDP for G␤␥-CFP precludes simple conversion of IC 50 to K i because the K i is well below the total concentration of G␤␥-CFP. We therefore determined K i by iteratively simulating the experimental data using the total concentrations of each protein and the previously determined K d of 550 nM for G␤␥-CFP binding to PH-97Alx.
Intramolecular Cross-linking and Quantification-PLC-␤3 variants with Cys 60 and Cys 164 were cross-linked with bismaleimidoethane (BMOE, Thermo Fisher Scientific). Prior to cross-linking, the protein was buffer exchanged into 20 mM Hepes (pH 7.5), 100 mM NaCl to remove dithiothreitol. Typically, a 10-to 15-fold molar excess of BMOE was added to 10 M PLC-␤3, followed by 30 min of incubation at 0°C. The reaction was quenched with excess DTT before use.
For quantification of the fraction of PLC cross-linked, the PLC-TEV construct was reacted with BMOE as described. Following quenching, an aliquot was incubated with a 10-fold molar excess of TEV protease at 34°C for 14 h. The protein was then denatured with sample buffer, run on 6% SDS-PAGE, transferred to a nitrocellulose membrane, and detected with an antibody toward the C terminus of PLC-␤3 (B521) (5). Immunoblots were quantitated using ImageJ to measure the intensities of cross-linked and residual non-cross-linked PLC-␤3. The fraction of PLC-␤3 cross-linked was calculated as the intensity of cross-linked PLC-␤3 divided by the sum of both intensities. (19), suggesting that the PH domain is required for G␤␥ interaction with the intact enzyme. To determine whether the PLC-␤ PH domain is a G␤␥ binding site, we developed a FRET-based binding assay. We used CFP fused to G␤␥ as the donor and covalently labeled the isolated PLC-␤3 PH domain that has a single reactive cysteine (PH-A97C) with an Alexa Fluor 594 acceptor (PH-97Alx). As shown in Fig. 1, PH-97Alx quenches the fluorescence of G␤␥-CFP, accompanied by simultaneous enhancement of Alexa Fluor 594 acceptor fluorescence. We detected no quenching of CFP fluorescence when we added unlabeled PH domain or free dye to G␤␥-CFP or when the labeled PH domain was unfolded by heat denaturation prior to addition. By this assay, G␤␥ binds the PH domain with a K d of 550 nM (Fig. 1B) in detergent solution.

G␤␥ Binds the PH Domain of PLC-␤3-Deletion of the PH domain in PLC-␤ abolishes regulation by G␤␥
Termination of G␤␥ signaling in cells occurs through sequestration by G␣ i -GDP after GTP hydrolysis, and G␣ i -GDP blocks G␤␥ stimulation of PLC-␤3 (8). As predicted, G␣ i -GDP competes with PH-97Alx for G␤␥ binding, as evidenced by the recovery of CFP donor fluorescence (Fig. 1C). The IC 50 for G␣ i -GDP was ϳ40 nM, and correction for the affinity with which G␤␥ binds PH-97Alx ( Fig. 1B) (see "Experimental Procedures") indicates that the K d for G␣ i -GDP-G␤␥ binding is ϳ3 nM (range, 2-5 nM based on standard error of fit shown in Fig. 1B), consistent with the subunits existing as a constitutive heterotrimer when G␣ i is in the GDP-bound inactive state. Phosducin, another regulator of G␤␥ signaling that shares the same interface on G␤␥ as G␣ i -GDP (31), also competitively displaced PH-97Alx from the G␤␥-PH complex with an IC 50 of ϳ70 nM. From these data, the K i for G␤␥-phosducin binding is 30 nM, in good agreement with previous reports (29). These observations suggest that the assay reports on specific G␤␥-PH domain binding. PH-A97C also displaced PH-97Alx to bind G␤␥, although with a much higher K i (Fig. 1D) (see below).
To determine whether PH domain binding is necessary for G␤␥ interaction with intact PLC-␤, we asked whether the PH domain can inhibit G␤␥-stimulated PLC-␤3 activity. As shown in Fig. 2, the addition of isolated PH domain inhibited G␤␥ stimulation of PLC-␤3 with an IC 50 of 18 M and inhibition of 65% at the highest concentration tested. The PH domain did not alter the basal or G␣ q -GTP␥S-stimulated activity of PLC-␤3 at any concentration, indicating that the inhibition was an effect on G␤␥ stimulation and not on the catalytic activity of PLC-␤ itself. These data indicate that the isolated PH domain competes with the PH domain in intact PLC-␤ to specifically block the G␤␥-PLC-␤3 interaction. Thus, we propose that the PH domain defines the minimal G␤␥ binding region in PLC-␤.
The data in Figs. 1-3 measure the affinity of G␤␥ binding to the isolated PLC-␤ PH domain using several different experimental approaches: FRET-based binding measurement, the effect on stimulation of PLC-␤ catalytic activity, and competition in both of these assays. The tightest interaction is that of the direct binding of G␤␥-CFP to PH-97Alx, with K d ϳ0.5 M, and the weakest interactions are for competition by unlabeled PH domain for binding or activation in the presence of phospholipid vesicles. Disagreement among these values appears to arise both from the use of phospholipids rather than cholate in the various assay buffers and from the ability of Alexa Fluor labeling to increase affinity of binding. FRET-based binding and competition assays conducted in medium containing phospholipid vesicles report a lower-affinity interaction than when measured in cholate buffer, in agreement with the low-potency inhibition of PLC-␤ stimulation in vesicles. More significant is the observation that Alexa Fluor labeling increases affinity by 15-to 20-fold (Fig. 3B). Consistent with these observations, we observed only 10% inhibition of G␤␥-CFP-PH-97Alx binding by ϳ80 M PH-A97C in phospholipid-containing buffer. Although we do not understand the mechanism of these effects in any detail, the consistency of results from two assays, with full-length PLC and isolated PH domain, and under different conditions indicates that the variations do not undercut the interpretability of the binding interactions.
Rac1 and G␤␥ Bind Distinct Surfaces on PLC-␤-PLC-␤2 and PLC-␤3 are directly activated by the small GTPase Rac1. GTP␥S-loaded Rac1 binds intact PLC-␤ with a K d of ϳ25 M, similar to that for the isolated PH domain (32). Structural studies later confirmed the PH domain as the Rac1 binding site (12). We therefore asked whether Rac1 competes with G␤␥ for stimulation of PLC-␤ activity using PLC-␤2, the most Rac1-sensi-tive isoform. As observed previously (7), activation by Rac1 and G␤␥ was essentially additive even at near-saturating concentrations of either ligand (Fig. 4). Further, earlier studies showed that point mutations within the PH domain diminish Rac1stimulated activity without altering G␤␥-stimulated activity (12). These results indicate that both Rac1 and G␤␥ interact with the PLC-␤ PH domain through distinct binding sites that do not overlap.
Prediction of the PH Domain-G␤␥ Binding Interface by Structural Homology-Because G␤␥ binds PH domains in many proteins, it is likely that it recognizes a common PH domain surface. The GRK2 PH domain-G␤␥ interface has been extensively characterized biochemically (33) and structurally (34). The top face of the G␤ propeller makes extensive contact with strands ␤3 and ␤4 of the GRK2 PH domain core and with an unstructured, basic C-terminal extension past the PH domain, burying ϳ2200 Å 2 of surface area (34). The PH domain in PLC-␤ adopts a fold similar to the GRK2 PH domain (root mean squared distance 3.8 Å for equivalent C␣ atoms) (Fig. 5A), although it does not possess an analogous extension and is immediately followed by the EF hand domain. However, the ␤ strands of the PH domain structurally align with the equivalent GRK2 regions even though the G␤␥ contact residues are not strictly conserved (Fig. 5B). Furthermore, we noted that PLC-␤1, PLC-␤2, and PLC-␤3 display a stronger conservation of residues in this region in comparison with PLC-␤4, the only isoform that is not detectably stimulated by G␤␥ (5).
To test whether the PH domains of GRK2 and PLC-␤3 bind G␤␥ in a similar manner, we examined the role of strands ␤3 and ␤4 in the PLC-␤ PH domain by mutational analyses. The hydrophobic character of this region suggested that G␤␥-PH domain binding is driven by non-polar interactions. To increase the likelihood that a single amino acid substitution at this interface would result in significant perturbation, we chose to introduce the large polar amino acid glutamine. We found that replacement of a conserved hydrophobic residue with glutamine (F50Q) was sufficient to completely abolish binding to G␤␥ (Fig. 5C). Consistent with our results, a GRK2 mutant harboring a substitution at the same position in the PH domain (GRK2 R587Q) was insensitive to stimulation by G␤␥ (33). Because PLC-␤4 is not activated by G␤␥, we predicted that replacement of PLC-␤3 residues in this region with the corresponding PLC-␤4 residues would diminish G␤␥ binding affinity. We picked three positions, Tyr 53 , Thr 55 , and Asp 62 , which were conserved across G␤␥-responsive PLC-␤ isoforms but not in PLC-␤4 (Asp 62 is replaced by Glu in PLC-␤1 and PLC-␤2). Mutations at two of these residues, T55R and D62Q, displayed weaker binding to G␤␥ with affinities 1.5-and 3-fold weaker than the wild-type protein. These results suggest that residues from strands ␤3 and ␤4 in the PH domain are important and specific binding determinants for G␤␥.
Next, we asked what effect these mutations have on the ability of the PH domain to compete with intact PLC-␤3 for G␤␥ binding. If a mutation weakened PH domain-G␤␥ binding, then it would be predicted not to impede G␤␥ stimulation of PLC-␤3 catalytic activity. Addition of wild-type PH domain inhibited G␤␥ stimulation of PLC-␤3, as observed earlier. The PH (F50Q) mutant, which does not bind G␤␥, inhibited G␤␥-stim- ulated activity by less than 10% so that PLC-␤3 activity was within error of the control sample (Fig. 5D). The T55R mutation in the PH domain also reduced its ability to inhibit stimulation by G␤␥, although loss of inhibition was only partial. PH (T55R) inhibited G␤␥-stimulated activity by 45% at 30 M, the highest concentration tested, compared with 65% for the wildtype PH domain. This loss is consistent with the 1.5-fold reduction in its affinity for G␤␥. The PH (D62Q) mutant, which had 3-fold lower affinity for G␤␥, was similarly compromised in its ability to compete with PLC-␤3 for G␤␥ binding. At 10 M, inhibition of G␤␥-stimulated PLC-␤3 activity by PH(D62Q) was insignificant, whereas maximal inhibition was only 20% at the highest concentration tested.
Collectively, these data support our hypothesis that strands ␤3 and ␤4 in the PH domain form the G␤␥ binding face. These results also suggest that the presence of polar residues in PLC-␤4 in this region interferes with productive G␤␥ interaction, rendering it insensitive to activation by G␤␥.
The Predicted G␤␥ Binding Interface of PLC-␤ Is Not Surface-exposed-To understand how G␤␥ interacts with PLC-␤, we examined the proposed G␤␥ binding residues of the PH domain in PLC-␤ crystal structures. The interface formed by these residues is distinct from the footprint of Rac1 on the PH domain of PLC-␤2 (Fig. 6A), consistent with the lack of competition between Rac1 and G␤␥ for binding PLC-␤ (Fig. 4). Although the Rac1 binding surface is completely surface-ex-posed, the predicted G␤␥ binding face of the PH domain forms an intramolecular interface with the EF hands that buries close to 1200 Å 2 of total accessible surface area. Close inspection of the PLC-␤ structure reveals that this interface is composed of only a few hydrogen bonds and van der Waals contacts, suggesting that the interaction between the two surfaces is weak (Fig. 6B). The hydrogen bond interactions between the PH domain and EF hand residues involve residues whose side chain donor and acceptor atoms are 3-4 Å apart, significantly weaker in energy than the average hydrogen bond lengths observed in proteins (ϳ1.5-2 Å) (35). Taken together with the absence of electrostatic interactions or water bridges at this interface, we hypothesized that the PH domain is mobile in solution, existing in equilibrium between the conformation observed in crystal structures and an "open" conformation where the PH domain moves so that the residues required for G␤␥ binding become accessible.
If the PH domain of PLC-␤ has to swing away from the EF hand domain to bind G␤␥, then preventing mobility of the PH domain should obstruct G␤␥ access to its binding site. We used intramolecular cross-linking to restrict PH domain motion by introducing cysteines in the PH domain and EF hands to give PLC-␤3-Cys 60 ,Cys 164 . These residues lie at the solvent-exposed edge of the PH domain-EF hand interface (Fig. 7A) so that they can be cross-linked using BMOE, an 8-Å cysteine-reactive cross-linker. Irreversible cross-linking of these residues should  block the motion of the PH domain and stabilize it in the conformation observed in PLC-␤ crystal structures. As predicted, cross-linking PLC-␤3-Cys 60 ,Cys 164 inhibited G␤␥-stimulated PLC-␤ activity (Fig. 7B) without blocking stimulation by G␣ q -GTP␥S (Fig. 7C). Alkylation of PLC-␤3-Cys 60 ,Cys 164 with a monovalent maleimide or alkylation with BMOE of single-cysteine mutants, PLC-␤3-Cys 60 or PLC-␤3-Cys 164 , had no effect. These data indicate that intramolecular cross-linking at the PH domain-EF hand interface interferes with G␤␥-PLC-␤3 interaction uniquely without global effects on PLC-␤ structure or catalytic activity. BMOE cross-linking of PLC-␤3-Cys 60 ,Cys 164 decreased the maximal G␤␥-stimulated activity of cross-linked PLC-␤ only partially under optimized conditions, but the EC 50 for G␤␥ was unchanged (ϳ20 -30 nM, Fig. 7D). If cross-linking partially blocked a G␤␥ binding site on PLC-␤, we would predict an increase in EC 50 for G␤␥ activation because of decreased affinity rather than a decrease in maximal response.
The data in Fig. 7D suggest that BMOE cross-linking is incomplete, with the cross-linked PLC insensitive to G␤␥ and the rest retaining full sensitivity to G␤␥. Initial attempts to quantify cross-linking failed because the electrophoretic mobility of cross-linked PLC-␤3 was indistinguishable from that of the control, and mass spectrometry gave poor peptide coverage in the relevant regions. We therefore designed a mutant PLC-␤3 to quantify the fraction of cross-linked species. We introduced a TEV protease cleavage site in an exposed loop of the PH domain of PLC-␤3-Cys 60 ,Cys 164 to give a construct we refer to as PLC-TEV. Proteolytic cleavage of PLC-TEV should generate two separate polypeptides but would produce a single species in the enzyme that had been correctly cross-linked between Cys 60 and Cys 164 (Fig. 8A). The relative sizes of crosslinked and non-cross-linked species after proteolysis are such that they can be separated by gel electrophoresis. Treatment of PLC-TEV with protease yielded a species that migrated faster on a SDS-PAGE gel because of loss of the ϳ10-kDa N-terminal fragment (Fig. 8A). Treatment with the cross-linker prior to proteolysis abrogated this shift as expected, confirming that this strategy can be used to quantitate the fraction of PLC-␤ cross-linked. To test whether cross-linking proceeds to completion, we incubated PLC-TEV with the cross-linker for varying times and measured the fraction of cross-linked species. Cross-linking occurred rapidly, on the order of minutes, but plateaued with only ϳ70% of total PLC-TEV cross-linked after 30 min (Fig. 8B). Changing the pH or temperature or including a mild reducing agent, tris(2-carboxyethyl)phosphine, in the cross-linking buffer did not increase fractional cross-linking. Incubating a fixed concentration of protein with increasing amounts of BMOE gave similar results. We then analyzed the G␤␥ response of these species and observed a negative correlation between the fraction of cross-linked PLC-␤ and its stimulation by G␤␥ (Fig. 8C). This result indicates that the residual activity observed in Fig. 7D was from PLC-␤ that had not been cross-linked. Importantly, stimulation by G␣ q -GTP␥S was not dependent on the cross-linked status. Therefore, our observations are consistent with the idea that G␤␥ is unable to stimulate PLC-␤ when the PH domain is immobilized by tethering to the first EF hand.
Restraining PH Domain Motion Blocks G␤␥ Binding-Our results suggested that cross-linking PLC-␤ leads to reduced G␤␥ sensitivity by prohibiting its binding to the G protein. We developed a FRET-based assay for G␤␥ interaction with PLC-␤3 to directly test whether cross-linking blocks G␤␥-PLC-␤ binding. We labeled PLC-␤3 at Cys 97 in the PH domain with Alexa Fluor 594 (PLC-97Alx), as we had done for the isolated PH domain, and asked whether PLC-97Alx is a FRET acceptor for G␤␥-CFP. PLC-97Alx quenched G␤␥-CFP fluorescence as predicted, similar to PH-97Alx. The K d for G␤␥-PLC-␤3 binding was 220 nM (Fig. 9A). Unlabeled PLC-␤3 inhibited binding between G␤␥-CFP and PLC-97Alx with an IC 50 of 370 nM (Fig. 9B), equivalent to a K i of 170 nM, which is approximately equal to the directly determined K d for labeled PLC-␤3. These data indicate that the assay reports on specific binding between G␤␥ and PLC-␤3.
We then used this assay to ask whether cross-linked PLC-␤ binds G␤␥. Because BMOE and Alexa Fluor 594 both react with the sulfhydryl group of cysteines, we could not label crosslinked PLC-␤ with Alexa Fluor 594. Rather, we used competition to test the binding of cross-linked PLC-␤3 to G␤␥. We found that PLC-TEV treated with DMSO competes with PLC-97Alx for G␤␥ binding with an IC 50 of ϳ360 nM or K i of ϳ150 nM. (Fig. 9C), which agrees with the K i for wild-type PLC-␤3 (Fig. 9B). BMOE-treated PLC-TEV also competed for binding (Fig. 9C, inset), but PLC-TEV treated with BMOE is not completely cross-linked and competition apparently reflects only the residual non-cross-linked PLC (see above). The TEV proteolysis assay indicated that 35% of BMOE-treated PLC-TEV used in the experiment in Fig. 9C was not cross-linked, and correction for fractional cross-linking indicated that only the non-cross-linked fraction accounts for the observed inhibition (Fig. 9C). Collectively, these results indicate that PLC-␤ does not bind G␤␥ when cross-linked.

Discussion
Stimulation of PLC-␤ isoforms by G␤␥ is well established, but the identity of the binding site for G␤␥ has eluded consensus. In this study, we show that the N-terminal PH domain is the minimal FIGURE 6. The predicted G␤␥ binding face of the PH domain is not surface-exposed. A, top panel, schematic of PLC-␤. The C-terminal helical bundle is not shown. Bottom panel, crystal structure of PLC-␤2 with the PH domain depicted as a surface (PDB code 2ZKM). Residues predicted to be involved in G␤␥ binding based on homology with the GRK2-G␤␥ structure (PDB code 1OMW) are shown in red. This surface is not solvent-accessible and does not overlap with the region that binds Rac1 (green). B, details of interactions between the PH domain and EF hands in intact PLC-␤3. Residues are colored to match the schematic in A. The numbers indicate the distance between donor and acceptor atoms for residues involved in hydrogen bonds.
G␤␥ binding site. Identification of the G␤␥-binding surface led to a new structural model for G␤␥ binding that involves substantial movement of the PH domain that may help anchor and orient PLC-␤ at the bilayer surface for PIP 2 hydrolysis. This work also describes a reliable and quantitative assay for binding of Gb 1 g 2 to PLC-␤3 that should be applicable to any PLC-␤ and G␤␥ dimer.  The FRET-based binding assays indicate that G␤␥ binds similarly to both the isolated PH domain of PLC-␤3 and to intact PLC-␤. The free PH domain also inhibited PLC-␤ activation by G␤␥, but not G␣ q (Fig. 2), by competing with the intact enzyme for binding G␤␥. Binding of G␤␥ to the PH domain was inhibited by G␣ i -GDP, consistent with the idea that G␣ i sequesters G␤␥ in the G i heterotrimer until G i is activated by GTP.
We defined the G␤␥ binding surface of the PLC-␤ PH domain starting with analogy to the crystal structure of G␤␥ in complex with GRK2. PH domains in multiple proteins are implicated in G␤␥ binding (20), and, of these, G␤␥ binding to the PH domain of GRK2 is the best studied (34). The crystal structure of G␤␥ in complex with GRK2 shows that G␤␥ recognizes a hydrophobic face of the GRK2 PH domain, and this surface was readily mapped to homologous residues of the PLC-␤ PH domain (Fig. 5). The effect of mutating residues on this face of the PH domain confirms the importance of this hydrophobic surface for regulation and binding by G␤␥. Mutation of the conserved Phe 50 at this surface to a polar residue abolished G␤␥ binding and inhibition of G␤␥-stimulated PLC-␤3 activity, and milder mutations at other residues diminished binding to a lesser extent. Previous studies showed that point mutations at Ile 80 , Trp 99 , Met 101 , Leu 117 , and Trp 332 in the opposing surface of G␤ reduced stimulation of PLC-␤ (36,37). Taken together, these data suggest that non-polar rather than electrostatic interactions contribute most of the free energy of binding between G␤␥ and the PH domain.
In the GRK2-G␤␥ complex, a cluster of Lys/Arg residues in an extension past the GRK2 PH domain also interacts with the G␤␥ surface. Strikingly, the first EF hand domain (EF-1) immediately C-terminal to the PH domain in all four PLC-␤ isoforms is similarly rich in Lys/Arg residues, and Barr et al. (38) showed that a quadruple charge-reversal mutation in EF-1 is less responsive to G␤␥. They also showed that a GST fusion construct of the PH domain plus EF-1 bound G␤␥ in pulldown experiments but that the charge reversal mutation reduced binding. In that same study, G␤␥ also bound a fusion construct of GST and the PH domain alone, in agreement with our observation that the PH domain is the minimal G␤␥ binding region, but just the PH fusion was less efficient in pulldowns than the Greater quenching in the absence of competitor compared with B is due to the higher concentration of PLC-97Alx acceptor in this experiment. The solid line is a fit of the competition by DMSO-treated PLC-TEV to a single-site model with IC 50 ϭ 360 ϩ 50 nM (ϩ S.E. of fit). Similar results were obtained when this experiment was repeated using a different batch of BMOE-treated PLC-TEV with a slightly lower fraction cross-linked (53%). Inset, competition by BMOE-treated PLC-TEV but with the abscissa showing total concentrations of PLC-TEV for the same dataset as in the main panel.
PH plus EF-1 construct. These data argue that G␤␥ minimally binds the PH domain and that interactions of basic residues in EF-1 may enhance affinity.
The G␤␥-interacting, hydrophobic face of the PH domain we identified is surprisingly buried in an intramolecular interface with the EF hands in crystal structures of PLC-␤3 (Fig. 6) so that a substantial conformational change would be required to allow G␤␥ binding. We used intramolecular cross-linking of the PH domain to the EF hands to confine the PH domain in the conformation observed in PLC-␤ crystal structures. Consistent with our proposal, immobilizing the PH domain makes PLC-␤ refractory to stimulation by G␤␥ (Figs. 7 and 8). Furthermore, we showed that G␤␥ failed to bind PLC-␤3 in which the PH domain was immobilized (Fig. 9). The G␣ q response did not change when PLC-␤ was cross-linked, indicating that the impaired interaction of cross-linked PLC-␤ was G␤␥-specific. However, cross-linking-mediated immobilization also likely blocked access to the PH domain surface that faces the TIM barrel. We tested various mutants in this region and found no change in PH domain inhibition of G␤␥-stimulated PLC-␤ activity, suggesting that this surface is not involved in G␤␥ binding.
G␤␥-PLC-␤ Binding Requires Movement of the PH Domain-If our identification of the G␤␥ binding interface in PLC-␤ is correct, then there must be significant movement to expose it because it is mostly occluded in the resting PLC-␤ structure by juxtaposition to the EF hand domain (Fig. 6). Furthermore, much of the exposed surface of the PH domain is needed for the Rac1 binding site, and we showed that Rac1 and G␤␥ can bind PLC-␤ simultaneously (Fig. 4). To allow regulatory G␤␥ binding, we propose that PLC-␤ exhibits intrinsic flexibility between closed and open conformational states of the PH domain (Fig. 10). The closed state, in which the PH domain faces the EF hands, is favored under crystallization conditions and, presumably, in solution in the absence of activators. PLC-␤ can, however, transiently sample an open state in which the PH domain moves away from the EF hands to expose the hydrophobic G␤␥ binding face and, by a still not understood mechanism, stimulate catalytic activity. The need for such motion is supported by the observation that locking PLC-␤3 in the closed state by cross-linking the PH domain to EF-1 blocks both G␤␥ binding and the ability of G␤␥ to activate (Figs. 8 and 9).
Crystallography of PLC-␦ suggests that significant motion of the PH domain in PLC is plausible. In crystals of full-length PLC-␦1, the PH domain and EF-1 were both invisible, suggesting that they undergo significant motion even in crystals (39,40). Truncation of the PH domain alone stabilized PLC-␦ so that its structure could be determined. Further, Drin et al. (41) showed that G␤␥-PLC-␤ binding reduced intramolecular FRET between the PH domain and the TIM barrel, suggesting that movement of the PH domain away from the TIM barrel to an open state is coupled to G␤␥ binding (41). The transition from the closed to the open conformation in PLC-␤ may also involve loss of EF-1 structure. Unfolding EF-1 would achieve twin objectives: higher affinity G␤␥ binding from interaction with the EF-1 basic residues (38), as discussed above, and a longer tether to allow whatever PH domain movement is needed for G␤␥ to bind in an appropriately activating orientation.
The requirement for PH domain motion in PLC-␤ activation by G␤␥ does not produce a detailed model for the PLC activation process, but it is consistent with a general model for PLC activation proposed by Sondek and co-workers (2). PLC-␤ contains an anionic autoinhibitory strand that occludes the active site and that must be moved to allow activation (1). Sondek and co-workers proposed (2) that simply holding the PLC molecule tightly against the bilayer surface shoves the autoinhibitory X-Y linker away from the active site by electrostatic and/or steric repulsion by negatively charged lipids (1). Although we do not know precisely how the G␤␥-PH domain complex is oriented with respect to the bilayer, anchoring PLC-␤ at the membrane surface is an important part of the G␤␥ activation mechanism. G␤␥ with a non-prenylated G␥ does not bind to membranes and does not activate PLC-␤ (42). Less direct support for the role of bilayer anchorage comes from studies of a PLC chimera with a PIP 2 -binding PH domain (43).
A mobile PH domain can allow binding of PLC-␤ to G␤␥ and consequent activation, but this mechanism by itself does not constrain PLC-␤ orientation. Such a simple mechanism might, however, explain why G␤␥ is a less efficacious activator of PLC-␤ than G␣ q . G␣ q binds PLC-␤ by contacting regions in the EF hand domain, the C2 domain, and a helical extension immediately C-terminal to the C2 domain (11) and may thereby help orient PLC-␤ with its active site facing the bilayer to facilitate access to the PIP 2 substrate. G␣ q binding has also been proposed to relieve autoinhibition by moving the helical extension away from the autoinhibitory strand to allow it to move away from the active site (13). There are no data to suggest that G␤␥ performs either function and may therefore be a weaker activator.
Mobility of the PH domain in binding G␤␥ may also reconcile PH-G␤␥ binding with the suggestion by Smrcka and coworkers (21) that G␤␥ has a second binding site on PLC-␤ on the TIM barrel domain. This site is essentially on the side of PLC-␤ opposite from that of the PH domain in crystal structures. If the PH domain exhibits significant motion, it is possible that G␤␥ binds simultaneously to the PH and TIM barrel domains, although this proposal awaits rigorous investigation Our work suggests the presence of a previously unappreciated PLC-␤ conformation critical to G␤␥ binding and regulation of phospholipase activity. PLC-␤ is also synergistically activated by G␣ q and G␤␥ (8), and the G q GTPase-activating protein activity of PLC-␤ is inhibited by G␤␥ (17). How the physical and functional interactions of G␤␥ with PLC-␤ and G␣ q direct the dynamics of complex assembly/disassembly in this three-protein system is uncertain. Studies focused on visualizing these interactions will shed light on the mechanisms coordinating G␤␥ regulation of PLC-␤ activities.
Author Contributions-G. K. conceived the project, designed and conducted the experiments, analyzed most of the data, and prepared the manuscript. E. M. R. contributed to the data analysis and wrote the paper with G. K.