Direct observation of conformational dynamics of the PH domain in phospholipases Cϵ and β may contribute to subfamily-specific roles in regulation

Phospholipase C (PLC) enzymes produce second messengers that increase the intracellular Ca2+ concentration and activate protein kinase C (PKC). These enzymes also share a highly conserved arrangement of core domains. However, the contributions of the individual domains to regulation are poorly understood, particularly in isoforms lacking high-resolution information, such as PLCϵ. Here, we used small-angle X-ray scattering (SAXS), EM, and functional assays to gain insights into the molecular architecture of PLCϵ, revealing that its PH domain is conformationally dynamic and essential for activity. We further demonstrate that the PH domain of PLCβ exhibits similar dynamics in solution that are substantially different from its conformation observed in multiple previously reported crystal structures. We propose that this conformational heterogeneity contributes to subfamily-specific differences in activity and regulation by extracellular signals.

Phospholipase C (PLC) 2 enzymes hydrolyze phosphatidylinositol lipids at cell membranes, producing inositol phosphates and diacylglycerol, which in turn promote the release of Ca 2ϩ from intracellular stores and activate protein kinase C (PKC) (1). Of the six PLC subfamilies, PLC␤ and PLC⑀ are required for normal cardiovascular function, and dysregulation of their expression and/or activity can result in cardiac hypertrophy and heart failure (2-6). The activity of these lipases is autoinhibited by various domains and structural elements but can be stimulated up to ϳ60-fold following activation of cell surface receptors (1). PLC␤ is activated downstream of G proteincoupled receptors (GPCRs) primarily through direct interactions with the heterotrimeric G protein subunits G␣ q and G␤␥ (1,7). PLC⑀ is also activated downstream of GPCRs by G␤␥ (8,9), by the small GTPases RhoA and Rap1A, and by Ras GTPases following receptor tyrosine kinase activation (10).
Crystal structures of PLC␤ alone (22), in complex with G protein activators (12,13,23,24) and/or physiologically relevant small molecules (24) show that the PH, EF-hands, TIM barrel, and C2 domains consistently assemble into a compact, globular structure. Deletion of any of these individual domains eliminates basal activity, whereas the C-terminal extension is dispensable (7,25). Thus, interdomain contacts between the core domains appear essential for function. However, this view is being challenged by studies suggesting that the PLC␤ PH domain is flexibly connected to the rest of the core and that this flexibility may be essential for its regulation (26 -28). In support of this hypothesis, the related PLC␦ enzyme contains a PH domain that is flexibly connected to the core and dispensable for activity but increases processivity by binding PIP 2 at the membrane (29 -31).
In contrast, relatively little is known about the structure and molecular regulation of PLC⑀. The pathways leading to G protein-mediated activation of PLC⑀ have been characterized (5,10,20,(32)(33)(34), but the activator-binding sites on PLC⑀ are largely unknown. It is also not known how the accessory domains contribute to regulation, in part because soluble, catalytically active variants have not been available for study. The only structural information for this enzyme comes from NMR structures of its two RA domains and a crystal structure of the RA2 domain in complex with Ras GTPase (35). Biochemical studies of PLC⑀ have shown that its CDC25 and PH domains, its first two EF-hands, and its RA domains are dispensable for lipase activity (35)(36)(37). Thus, PLC⑀ appears more similar to PLC␦ than PLC␤ with respect to the requirement for its PH domain in lipase activity.
In this work, we used biochemical assays, small-angle X-ray scattering (SAXS), and negative stain EM to obtain structural insights into the solution architecture of PLC⑀ and probe the roles of its PH and RA domains in regulating basal activity. We show that the PH domain, likely together with the first two EF-hands (EF1/2), is conformationally heterogeneous with respect to the rest of the PLC core, which retains a compact structure. In addition, the PH and RA domains have opposing effects on the thermal stability of the enzyme that indicate that one or both RA domains are intimately associated with the core, but the PH domain is not. To compare these results with that of a structurally well-characterized PLC, we also examined the solution structure of the PLC␤3 core domains. In contrast to the structures observed across multiple different crystal forms, we show that the PLC␤ PH domain also adopts multiple conformations in solution. Using site-directed mutagenesis and disulfide bond engineering, we found that both the extended and closed conformational states of the PLC␤ PH domain retain basal activity and adsorb to liposomes, suggesting that these conformational states are functionally relevant. Thus, it seems likely that all PLC enzymes containing PH domains may be inherently dynamic in solution and that these conformational states may underlie subfamily-specific differences in their autoregulation, their association with cell membranes, and their allosteric regulation by other signaling proteins.

The PLC⑀ PH and RA domains have differential impacts on basal activity, thermal stability, and liposome binding
Given the lack of prior knowledge about the tertiary organization of PLC⑀, we investigated the respective contributions of the PH domain and the two RA domains to basal activity, the structural integrity of the enzyme, and binding to model membranes. Because the full-length enzyme presents significant challenges in expression and purification, we expressed and purified three PLC⑀ domain deletion variants: PH-COOH, which retains the core domains and both RA domains (residues 837-2282); PH-C2, which lacks both RA domains (residues 837-1972); and EF-COOH, which lacks the PH domain (residues 1035-2282) (Fig. 1A) (1,8). Using a [ 3 H]PIP 2 liposome hydrolysis assay (24,38,39), we found that the PH-COOH variant has the highest basal activity ( Fig. 1B; (Fig. 1B).
To test whether the loss in specific activity is due to protein destabilization, the melting temperature (T m ) of each variant was determined using differential scanning fluorimetry (DSF) (40). The PH-COOH and EF-COOH variants have similar T m values of 51.3 Ϯ 0.7 and 50.6 Ϯ 1.6°C, respectively (Fig. 1C), but PH-C2 has a lower T m (48.3 Ϯ 1.0°C, p Յ 0.005). This suggests that one or both RA domains directly contribute to the thermal stability of the enzyme. The resulting destabilization of the PLC core may be sufficient to account for the decreased basal activity of this variant. We next tested whether differences in basal activity were due to changes in the ability of the variants to bind PIP 2 -containing liposomes using a pelleting assay (41). We hypothesized that PH-COOH and PH-C2 would show the greatest binding to liposomes, as they retain the PH domain. However, in the presence of liposomes, all three variants were present to similar extents in the pellet, suggesting that differences in activity are not due to differences in binding (Fig. 1, D and E).

PLC⑀ variants have molecular envelopes distinct from those predicted from PLC␤ crystal structures
The PLC⑀ PH domain increases basal activity, but has no effect on thermal stability. We therefore hypothesized that the PH domain may be flexibly attached to the rest of the core. If so, then its molecular envelope should be characteristically different from that predicted by the crystal structures of PLC␤ (46% sequence identity for the region spanning EF3 through the C2 domain). We used small-angle X-ray scattering (SAXS) to determine low-resolution ab initio bead models for the PLC⑀ PH-COOH, EF-COOH, and PH-C2 variants (Fig. 2, Tables 1 and 2, and Figs. S1-S4) (42). Size-exclusion chromatography (SEC)-SAXS was used to ensure sample homogeneity (Figs. S1 and S2). The minimal variability of radius of gyration (R g ) as determined by Guinier approximation using the SAXS data ( Fig. S3) or molecular weights as determined by multi-angle light scattering (MALS) (Fig. S4) across the elution peaks indicates that the samples were monodisperse. Guinier plots for each variant were also consistent with the absence of aggregates. PH-COOH and PH-C2 have R g values of 43 Ϯ 0.54 and 40 Ϯ 0.48 Å, respectively, whereas EF-COOH had an R g of 46 Ϯ 2.5 Å (Fig. 2 (D-F) and Table 2). All three PLC⑀ variants also had pair distance distribution functions consistent with a predominantly compact shape, but with some extended conformational aspects. The maximal intramolecular dimension (D max ) for each variant was determined, with values of ϳ162, ϳ148, and ϳ161 Å, for PH-COOH, PH-C2, and EF-COOH, respectively (Fig. 2, G-I). The SAXS data were also used to estimate the molecular masses of the variants using the volume of correlation (V c ) and the Porod volume (V p ) (43). PH-COOH had molecular masses of 118 kDa (V c ) and 126 kDa (V p ), lower than the calculated molecular mass of 165.5 kDa. As an additional control, MALS was used to independently validate the molecular mass of this variant at 168 kDa (Fig. S4). PH-C2 had estimated molecular masses of 123.6 kDa (V c ) and 137 kDa (V p ), comparable with its calculated molecular mass of 130.8 kDa. Finally, EF-COOH had estimated molecular masses of 143.9

Conformational dynamics of PLC enzymes
kDa (V c ) and 138.3 kDa (V p ), again comparable with the calculated molecular mass of 143.4 kDa (44).
Comparison of the ab initio SAXS models of PH-COOH and PH-C2 reveals that they are very similar in overall shape (Fig. 2, J and K). PLC⑀ PH-COOH has a larger molecular volume (4.57 ϫ 10 5 Å 3 ) compared with PH-C2 (4.06 ϫ 10 5 Å 3 ), consistent with the presence of the RA domains in the PH-COOH variant (45). As a comparison, we fit the atomic structure of the PLC␤3 core within the density (including the PH-C2 domains from PDB entry 3OHM (12)), confirming that the PLC⑀ envelopes are substantially larger than would be predicted for PLC␤, which is consistent with their larger size (PLC␤3 PH-C2 is ϳ60 and ϳ77% of the total mass of PLC⑀ PH-COOH and PH-C2, respectively). Most interestingly, the SAXS envelopes of PH-COOH and PH-C2 both feature a well-defined, extended protrusion consistent with the extended tail observed in their pair

Conformational dynamics of PLC enzymes
distance distribution functions (Fig. 2, G, H, J, and K). This feature is consistently observed in ab initio models of these variants, as evidenced by the normalized spatial discrepancy (NSD) values of 0.968 Ϯ 0.032 for PH-COOH and 0.924 Ϯ 0.044 for PH-C2 for 15 independently determined models (46). As this feature is observed in both PH-COOH and PH-C2, it cannot correspond to the RA domains (Figs. 1A and 2) and thus most likely represents the PH domain. The ab initio model of

Conformational dynamics of PLC enzymes
the EF-COOH variant shares approximately the same long dimension as the PH-COOH variant. EF-COOH also contains a similar extended protrusion from the main body of the envelope, with an NSD of 0.627 Ϯ 0.039 for 10 independent models ( Fig. 2 (F, I, and L) and Table 2). However, electron density appears to be lost both from the protrusion and from the center of the molecule, relative to the models of PH-COOH and PH-C2 (Fig. 2, J-L). This is supported by the decreased molecular volume of the EF-COOH variant (3.48 ϫ 10 5 Å 3 ) (45). These differences are most likely a direct consequence of deleting the PH domain in EF-COOH.

Negative stain EM reconstruction of PLC⑀ variants
To obtain higher resolution and orthogonal structures of the PLC⑀ variants, we used negative stain EM ( Fig. 3 and Fig. S5) to calculate single-particle reconstructions. Preliminary analysis of the data sets revealed challenges in obtaining dominant 3D class averages for each set of particles. Together with the SAXS data, this is consistent with conformational heterogeneity in the PLC⑀ variants and necessitated a conservative approach to data processing and calculating 3D reconstructions (Fig. S5).
For the PH-COOH variant, 2D classes containing 34,155 particles were used to generate two initial 3D models, each of which was used to independently calculate two refined 25 Å 3D reconstructions (9,879 and 13,628 particles each) (  Fig. S6). These reconstructions confirm that PH-COOH and PH-C2 are monomeric. The two PH-COOH reconstructions have molecular volumes of 2.92 ϫ 10 5 and 3.40 ϫ 10 5 Å 3 , respectively, and maximum diameters of ϳ116 Å. The two reconstructions of PH-C2 have molecular volumes of 3.07 ϫ 10 5 and 2.78 ϫ 10 5 Å 3 and diameters of ϳ116 and ϳ124 Å (45). Overall, these reconstructions have smaller molecular volumes compared with the SAXS envelopes. This is most likely due to the dehydration and sample flattening that is common in negative stain (47). Although the PH-COOH variant retains the RA domains, there are no obvious differences compared with PH-C2, further supporting our hypothesis that one or both RA domains stably associate with the rest of the core under these experimental conditions (Figs. 1 and 3). Thus, distinctive features protruding from the main body of the reconstructions of both PH-COOH and PH-C2 likely reflect the conformational heterogeneity of the PH domain with respect to the PLC⑀ core ( Fig. 3 and Fig. S6).
Two refined 20 Å reconstructions of the EF-COOH variant (13,416 and 14,498 particles each) were likewise generated from 2D classes containing 53,028 particles ( Fig. 3 (C, F, I, and L) and Fig. S6). Like the other PLC⑀ variants, the EM analysis independently confirms that EF-COOH is monomeric. Similar to the D-F, Guinier analyses of low q values, ln(I) (beam intensity) versus q 2 (scattering angle) with R g of PH-COOH, PH-C2, and EF-COOH, respectively. Fitting of the linear regressions to the data is represented by residuals, shown at the bottom of the plots, demonstrating that the proteins are monomeric in solution. G-I, pair distance distribution functions (P(r)) indicating elongated envelopes for PH-COOH, PH-C2, and EF-COOH variants, respectively. Estimated maximum intramolecular distances (D max ) are provided. Ab initio envelope models (left) and equivalent envelopes rendered as volumes (right) show protrusions in PH-COOH (J) and PH-C2 (K) and lack of density, likely corresponding to the missing RA domains in PH-C2. As a reference, the crystal structure of the PLC␤3 core (colored as in Fig. 1A; PDB entry 3OHM (12)) is fit within the SAXS-derived envelopes such that the PH domain is oriented toward the extended protrusion and the C2 domain toward the additional density on the opposite side of the envelope. In L (EF-COOH), the domain is extended similarly to PH-COOH, but electron density is lost from the protrusion and the center of the molecule due to loss of the PH domain. The PLC␤3 core lacking the PH domain is fit within the density as described for J and K.

Conformational dynamics of PLC enzymes
trends observed for the PH-COOH and PH-C2 variants, the reconstructions of EF-COOH are smaller overall than its SAXS envelopes. The EF-COOH reconstructions have molecular volumes of 3.13 ϫ 10 5 and 3.67 ϫ 10 5 Å 3 and maximum diameters of ϳ126 and 130 Å (Fig. 3, G-I) (45). Importantly, the EF-COOH reconstructions further confirm that this variant has a different molecular architecture compared with the PH-COOH and PH-C2 variants, due to deletion of the PH domain. This is consistent with the SAXS analysis of these variants (Fig.  2) and supports a model in which the PH domain adopts multiple conformations with respect to the rest of the PLC core.

The PLC␤3 PH domain is also conformationally dynamic
Because our data are consistent with the PH domain adopting multiple conformations in PLC⑀, we assessed the solution structure of the PLC␤3 core by performing SAXS on a variant that includes its PH, EF, TIM barrel, and C2 domains along with an autoinhibitory portion of its C-terminal domain (CTD), PLC␤3-⌬892 (Figs. 1A and 4).
We first compared the experimentally determined SAXS data for PLC␤3-⌬892 with the calculated SAXS data derived from the PLC␤3 crystal structure (containing the PH-proximal CTD domains from PDB entry 3OHM (12); Figs. 4 and 5, Tables 1 and 3, and Figs. S1-S3). The experimental and calculated Guinier plots for PLC␤3-⌬892 generated similar R g values of 34 and 31 Å, with experimentally determined molecular masses of 87.9 kDa (V c ) and 96.2 kDa (V p ), comparable with the calculated molecular mass of 100.89 kDa. However, the Guinier plots for the two data sets have a 2 value of 7.8, indicating substantial divergence between the experimental and calculated data (Fig.  5B). The calculated pair distance distribution function for PLC␤3-⌬892 (Fig. 5C) is bell-shaped, as expected for the globular structure observed in crystal structures (Fig. 4A). However, the experimentally determined pair distance distribution function indicates the presence of extended features, as evidenced by the tail at high values of r (Fig. 5C). In addition, the calculated D max is 105 Å, compared with the experimentally determined

Conformational dynamics of PLC enzymes
D max of 122 Å. This is consistent with PLC␤3-⌬892 having a more extended structure in solution that is inconsistent with published crystal structures.
The volume of the ab initio envelope of PLC␤3-⌬892 encapsulates the crystal structure of PLC␤3 but is more elongated and features a well-defined, extended protrusion (Fig. 5D). Thus, the overall shape of the PLC␤3-⌬892 SAXS envelope is more consistent with the SAXS envelopes and 3D reconstructions of PLC⑀ PH-COOH and PH-C2 than with the low-pass-filtered 20-Å envelope of the PLC␤ 3OHM crystal structure (Figs. 2, 3, and 4A). The compact portion of the PLC␤3-⌬892 SAXS envelope also shows additional density beyond the low-pass-filtered crystal structure (Fig. 5D). This most likely corresponds to disordered regions that are not observed in the crystal structure, such as the X-Y linker within the TIM barrel domain. In PLC␤3, this linker is ϳ80 residues (ϳ9 kDa) and is disordered in all published structures (Fig. 4A).
To confirm that the extended protrusion observed in the experimentally determined PLC␤3-⌬892 envelope corresponds to the PH domain, we generated the E60C/V164C double mutant of PLC␤3-⌬892 (Fig. 4A). These two mutations, in combination with an exogenous bismaleimidoethane crosslinking reagent, were previously shown to restrict the motion of the PH domain without impacting basal activity (26). Because the distance between the C␤ atoms of these residues varies between crystal structures, we anticipated that they would be able to form a redox-dependent disulfide bond. Thus, we purified PLC␤3-⌬892 E60C/V164C under either oxidizing or reducing conditions. Using the [ 3 H]PIP 2 hydrolysis activity assay, we found that reduced PLC␤3-⌬892 E60C/V164C had a specific activity of 0.33 Ϯ 0.20 nmol of IP 3 /min/nmol of PLC␤3-⌬892 variant, comparable with that of PLC␤3-⌬892 (0.41 Ϯ 0.20 nmol of IP 3 /min/nmol of PLC␤3-⌬892 variant) (Fig. 4B). Oxidized PLC␤3-⌬892 E60C/V164C had a specific activity of 2.3 Ϯ 0.37 nmol of IP 3 /min/nmol of PLC␤3-⌬892 variant, significantly greater than that of PLC␤3-⌬892 (p Յ 0.001) and reduced PLC␤3-⌬892 E60C/V164C (p Յ 0.01, Fig. 4B). However, all three PLC␤3-⌬892 variants had essentially identical T m values as measured by DSF (Fig. 4C), and all variants bound similarly to PE/PIP 2 liposomes, as measured using the liposome pelleting assay (Fig. 4, D and E). The increased basal activity of oxidized PLC␤3-⌬892 E60C/V164C is in contrast to a recent report from Kadamur and Ross (26), who found no significant difference in basal activity when the PH domain was conformationally restricted by cross-linking. This difference is most likely due to the use of PLC␤3-⌬892 in this study as opposed to  (12)), with the domains labeled and colored as in Fig. 1A. The predicted SAXS envelope is shown as a gray 20-Å low-pass filter surrounding the crystal structure. Residues Glu-60 and Val-164 (shown in ball-and-stick representations) were mutated to cysteines for disulfide bond formation to restrict the motion of the PH domain. The X-Y linker and the loop connecting the C2 domain to the proximal CTD are disordered and omitted for clarity. The ends of the X-Y linker are denoted by hot pink asterisks, and the ends of the C2-proximal CTD are denoted by cyan asterisks (26). B, cysteine mutations in PLC␤3-⌬892 (E60C/V164C) alter basal specific activity only under oxidized conditions (***, p Յ 0.001; **, p Յ 0.01 based on one-way ANOVA followed by Tukey's multiple-comparison test). Basal specific activity data represents at least two experiments performed in duplicate Ϯ S.D. (error bars). C, cysteine mutations in PLC␤3-⌬892 (E60C/V164C) do not alter thermal stability under reducing or oxidizing conditions. Representative curves are shown for PLC␤3-⌬892 (gray diamonds), reduced E60C/V164C (red inverted triangles), and oxidized E60C/V164C (orange squares). Thermal stability data represent at least six experiments performed in triplicate Ϯ S.D. D, representative SDS-PAGE of PLC␤3-⌬892 variants following ultracentrifugation in the presence (ϩ) or absence (Ϫ) of PE/PIP 2 liposomes. Total protein samples (T) contained protein incubated in the presence (ϩ) or absence (Ϫ) of liposomes but were not subject to centrifugation. E, PLC␤3-⌬892 variants show minimal binding to liposomes. Gels were quantified as described in Fig. 1E, and no significant differences in liposome binding were observed for the PLC␤3-⌬892 variants. All data are mean Ϯ S.D. of at least four independent experiments and analyzed by one-way ANOVA followed by Tukey's multiple-comparison test.

Conformational dynamics of PLC enzymes
full-length PLC␤3. Full-length PLC␤3 retains the C-terminal extension and thus has increased membrane association and much higher basal activity (1, 7).
The oxidized and reduced PLC␤3-⌬892 E60C/V164C mutants were also analyzed by SAXS and were confirmed to be monomeric and monodisperse in solution (Fig. 5, Table 3, and Figs. S1-S4). Guinier plots of these variants generated R g values of ϳ35 Å (Fig. 5, F and J). The experimentally determined molecular mass of oxidized PLC␤3-⌬892 E60C/V164C was found to be 71 kDa (V c ) and 65.3 kDa (V p ), and reduced PLC␤3-⌬892 E60C/V164C had molecular masses of 69 kDa (V c ) and 53.3 kDa (V p ), all of which are lower than the calculated molecular mass (100.6 kDa). The discrepancy in the SAXS-derived molecular masses versus the calculated molecular masses is likely due to the lower signal/noise ratio of these data sets. Thus, MALS was used to confirm the molecular masses of these variants in solution (Fig. S4). Finally, the pair distance distribution functions of these variants are similar, with D max values of ϳ120 Å (Fig. 5, G and K). The ab initio envelopes of oxidized and reduced PLC␤3-⌬892 E60C/V164C were then compared with one another and PLC␤3-⌬892 (Fig. 5, D, H, and L). The oxidized PLC␤3-⌬892 E60C/V164C envelope has a more compact, globular structure as compared with the envelope of PLC␤3-⌬892 (Fig. 5, D and H). The molecular volume of the oxidized PLC␤3-⌬892 E60C/V164C envelope is 3.42 ϫ 10 5 Å 3 , comparable with that of PLC␤3-⌬892 (3.44 ϫ 10 5 Å 3 ) (45). A similar result was obtained upon the addition of 15 mM excess bismaleimidoethane to PLC␤3-⌬892 E60C/V164C purified under reducing conditions. The resulting envelope again has a more compact shape than that of PLC␤3-⌬892, with a molecular volume of 3.43 ϫ 10 5 Å 3 (Fig. S7 and Table S1). Thus, the disulfide bond between E60C and V164C limits the motion of PH domain similarly to the previously reported cross-linker (26). In contrast, the envelope for reduced PLC␤3-⌬892 E60C/ V164C is most similar to that of PLC␤3-⌬892, as it clearly features the extended protrusion (Fig. 5, D and L). It also shares a similar molecular volume as the other PLC␤3-⌬892 variants (3.49 ϫ 10 5 Å 3 ). Overall, these results are most consistent with the PLC␤3 PH domain, likely together with EF1/2, adopting multiple conformations in solution that contribute to the final volume of the reconstruction.

Discussion
PLC⑀ and PLC␤ regulate the intracellular Ca 2ϩ concentration and PKC activation downstream of GPCRs and, in the case of PLC⑀, downstream of receptor tyrosine kinases (1,10,11). Numerous structures of PLC␤ have been determined over the last 10 years (12-14, 18, 23, 24, 48), but the observed compactness of PLC␤ in these structures appears to be misleading, given the growing evidence that its PH domain is conformationally dynamic (26 -28), including in this study. Although PLC⑀ is anticipated to share a similar core architecture with PLC␤ (46% sequence identity for the region including EF3-C2), it was not known whether its PH domain would be situated more similarly to crystal structures of PLC␤ or PLC␦ with respect to the core or how its C-terminal RA domains would influence the overall conformation of the protein.
Using biochemical assays, SAXS, and single-particle EM, we show that the PLC⑀ variants PH-COOH, PH-C2, and EF-COOH retain activity and have molecular architectures consisting of a compact core that features a well-defined protrusion (Figs. 1-3). As both PH-COOH and PH-C2 share this feature, this protrusion is highly unlikely to correspond to the RA domains and instead is most likely a core domain (Figs. 1-3) (9,36). The loss of density on the opposite end of the SAXS envelope in PH-C2 relative to PH-COOH is consistent with deletion of the RA domains. However, fitting the structure of the PLC␤3 core (PDB entry 3OHM (12)) in the PH-COOH and PH-C2 envelopes such that the C terminus of the C2 domain extends into the additional density also fails to account for the volume of the observed envelopes. Whereas PLC⑀ PH-C2 is ϳ35 kDa larger than PLC␤3 PH-C2, this is largely due to insertions and a longer X-Y linker and is still insufficient to account for the total volume of the envelope, especially the extended protrusion. Thus, the large envelopes are most consistent with dynamic behavior at the N terminus of the PLC⑀ variants. This observation is consistent with the fact that deletion of the PH domain has no effect on thermal stability (Fig. 1C), indicating that interactions between this domain and the rest of the core are transient in nature. We propose that the PH domain in both PLC␤3 and PLC⑀ is connected to the rest of the core via the  (12)). Fitting of the experimental linear regression to the experimental data is represented by residuals, as shown in the bottom tenth of the plot, and confirms that the proteins are monomeric in solution. C, pair distance distribution function P(r), with maximum interparticle distance (D max ) for the experimentally determined and calculated PLC␤3-⌬892 data shown as in B. D, ab initio "dummy atom" envelope models (left) and equivalent envelope rendered as volumes (right) for PLC␤3-⌬892. The PLC␤3 crystal structure is fitted as a reference, with the PH domain oriented toward the extended protrusion and colored as in Fig. 1A. E-G, raw scattering curve (E), Guinier analysis (F), and pair distance distribution function (G) of oxidized PLC␤3-⌬892 E60C/V164C. H, ab initio "dummy atom" envelope models (left) and equivalent envelope rendered as volumes (right) of oxidized PLC␤3-⌬892 E60C/V164C, with the crystal structure fit as in D. I-K, raw scattering curve (I), Guinier analysis (J), and pair distance distribution function (K) of reduced PLC␤3-⌬892 E60C/V164C. L, ab initio "dummy atom" envelope models (left) and equivalent envelope rendered as volume (right) for reduced PLC␤3-⌬892 E60C/V164C, with the crystal structure fit as in D. PLC␤3-⌬892 and reduced PLC␤3-⌬892 E60C/V164C envelopes show similar protrusions, whereas oxidized PLC␤3-⌬892 E60C/V164C appears more compact.

Conformational dynamics of PLC enzymes
weakly conserved EF1/2 hands and that a hinge between EF2 and EF3 allows the PH domain to adopt multiple conformations with respect to the core. In support of this hypothesis, we used ensemble rigid-body modeling to generate a model based on PLC␤3-⌬892 SAXS data (Figs. 5 and 6). In the resulting computational model, which fits the experimental data with a 2 of 1.67, the PH domain and EF1/2 are rotated away from the rest of the PLC core and from one another. The PH domain is rotated ϳ140°from its position in the crystal structures and translated by ϳ70 Å. EF1/2 is also rotated from its position in the crystal structure by ϳ155°and translated by ϳ20 Å (Fig. 6). This flexibility would therefore allow the PH domain and EF1/2 to adopt multiple conformational changes with respect to one another and to the rest of the PLC core. Although the current model is limited by the resolution of the SAXS envelopes and the EM reconstructions, it represents the simplest model of the data. We also showed that the PLC⑀ PH domain is important for lipase activity (Fig. 1B). The C-terminal RA domains of PLC⑀ also make a substantial contribution to basal activity (Fig. 1B), but because these domains contribute to stability as measured by DSF (Fig. 1C), the loss of activity is most simply explained by decreased structural integrity of the core. As the RA2 domain is the primary binding site for activators such as Rap1A and Ras (20,21), this may provide a potential mechanism by which small GTPases could allosterically activate phosphatidylinositol hydrolysis. In future studies, the role of the N-terminal CDC25 domain will need to be similarly assessed.
Although the PLC␤3 core is compact in crystal structures (Fig. 4) (12, 24), its SAXS envelope features a similar extended protrusion observed in the PLC⑀ PH-COOH and PH-C2 envelopes (Figs. 2, 3, and 5). To determine whether this extended protrusion is in fact the PH domain, we used SAXS to study the PLC␤3-⌬892 E60C/V164C variant under reducing and oxidizing conditions, which would restrain the motion of the PH domain (Fig. 5) (26). The molecular envelope of the reduced PLC␤3-⌬892 E60C/V164C mutant is most similar to that of PLC␤3-⌬892, whereas oxidation or cross-linking results in a more compact molecular envelope lacking a defined extended protrusion ( Fig. 5 and Fig. S7). The more compact structures observed following oxidation or cross-linking are more consistent with the conformation of PLC␤3 in crystal structures. Thus, we believe that these studies are the first to provide direct structural evidence for multiple conformations of the PLC␤ PH domain and possibly EF1/2. However, the respective contributions of these conformational states to basal activity and membrane association require further study (26).
Extension of the PH domain from the catalytic core may be important for PLC␤ and PLC⑀ regulation by G␤␥ subunits (8,9,49,50). Despite multiple studies over several decades, the G␤␥binding site on PLC␤ has remained elusive (27,28,51,52). Previous studies have suggested that G␤␥-dependent regulation of PLC␤ is modulated by interactions between these two proteins and the membrane (28,53). It is therefore possible that the PH domain is extended when PLC␤ is in contact with membranes. Interdomain contacts between the PH domain and the other core domains, including their respective orientation at the membrane, may also contribute to regulation (28). More recently, the G␤␥-binding site in PLC␤ was proposed to be blocked in the closed conformation of the PH domain, suggesting that conformational flexibility of the domain is required for G␤␥ binding and activation (26). Our data are fully consistent with this model. The mechanism of G␤␥-dependent activation of PLC⑀ is only beginning to be investigated (9), but because our data are consistent with PLC⑀ also containing a conformationally dynamic PH domain, we propose that it may likewise contribute to G␤␥-dependent regulation.

PLC⑀ and PLC␤ expression and purification
cDNA encoding Rattus norvegicus PLC⑀ PH-COOH (residues 837-2282), PH-C2 (residues 832-1972), and EF-COOH (residues 1035-2282) were subcloned into the pFastBac-HTA vector (Invitrogen) and confirmed by sequencing over the entire coding region. Proteins were expressed in baculovirusinfected Sf9 cells, harvested 48 -72 h postinfection, and flashfrozen. PLC␤3-⌬892 (residues 10 -892) was expressed as Figure 6. A rigid-body ensemble model of PLC␤3-⌬892 is consistent with conformational heterogeneity of the PH domain and EF1/2. A, crystal structures of the PLC␤ subfamily reveal a compact structure, wherein the PH domain is in close proximity to the TIM barrel domain and EF-hands 1 and 2 (EF1/2). Domains are colored as in Fig. 1A, with the exception of EF1/2, which are shown in periwinkle blue. Dashed lines, regions disordered in the crystal structure, with the catalytic Ca 2ϩ ion shown as a black sphere. B, a rigid-body ensemble model that better fits the experimentally determined SAXS data was determined using SASREF ( 2 of 1.67 (61)). In this model, the PH domain (residues 11-146), EF1/2 (residues 147-221), and the remaining domains of PLC␤3-⌬892 (residues 222-892) were treated as independent, rigid bodies. Additional restraints included restricting the PH domain C terminus to within 5 Å of the N terminus of EF1/2 and the C terminus of EF1/2 within 5 Å of the N terminus of EF3 and the rest of the enzyme. In the resulting model, the PH domain is translated by ϳ140°and translated ϳ70 Å from its position in the crystal structure. This large motion is facilitated by EF1/2, which is rotated by ϳ155°and translated ϳ20 Å from its position in the crystal structure. EF1/2 hands are weakly conserved compared with the rest of core domains across PLC enzymes and variably ordered in crystal structures of PLC␤. Finally, the residues linking EF2 and EF3 may serve as a hinge region, facilitating the rotation of EF1/2 and allowing the PH domain to sample multiple conformations.
PLC⑀ variants were purified as described (9), with some modifications for proteins used in single-particle EM and liposomepelleting assays. After lysis and ultracentrifugation, the supernatant was applied to a 5-ml HisTrap FF column (GE Healthcare) equilibrated with Buffer A (20 mM HEPES, pH 8, 300 or 500 mM NaCl, 20 mM imidazole, 10 mM ␤-mercaptoethanol, 0.1 mM EDTA, and 0.1 mM EGTA) and eluted with a gradient of 0 -100% Buffer A supplemented with 500 mM imidazole pH 8. Fractions containing protein were pooled, bufferexchanged with an equal volume of Buffer E (20 mM HEPES, pH 8, 50 mM NaCl, 2 mM DTT, 0.1 mM EDTA, and 0.1 mM EGTA), and loaded onto a 1-ml MonoQ 5/50 GL column (GE Healthcare) pre-equilibrated with Buffer E. Protein was eluted with a gradient of 0 -100% MonoQ-S Buffer E containing 500 mM NaCl. Fractions containing protein were again pooled, bufferexchanged, and concentrated as described above before loading on a 1-ml MonoS 5/50 GL column (GE Healthcare) pre-equilibrated with Buffer E. The protein was eluted, pooled, and concentrated as described above before final purification over tandem Superdex 200 Increase columns (GE Healthcare).

Differential scanning fluorimetry
DSF assays were carried out as described previously (40,54), with minor modifications. Purified PLC⑀ (0.2-0.6 mg/ml) and PLC␤ variants (0.5 mg/ml) were incubated with 5ϫ SYPRO orange dye in the presence of 5 mM CaCl 2 . All assays were performed in duplicate or triplicate with protein from at least two independent purifications.

Basal activity assays
PLC⑀ and PLC␤ activity was measured as described previously (24,38,54). The PLC⑀ PH-COOH variant was assayed at a final concentration of 0.075 ng/l, PH-C2 at 0.1-1 ng/l, and EF-COOH at 0.5 ng/l. PLC␤3-⌬892 was assayed at a final concentration of 10 ng/l, and PLC␤3-⌬892 E60C/V164C (oxidized or reduced) was assayed at 12 ng/l. Oxidized PLC␤3-⌬892 E60C/V164C was assayed in the absence of reducing agents. All assays were performed at least in duplicate with protein from at least two independent purifications.

SEC-SAXS/MALS-DLS data collection and analysis
SEC-SAXS was performed at BioCAT (Sector 18) Advanced Photon Source equipped with an ÄKTA Pure FPLC and a Pila-tus3 1M detector (Dectris) with a beam size of 160 m ϫ 75 m (Table 1). Data were collected at room temperature, with a 12-KeV X-ray (1.033-Å wavelength) and ϳ3.5-m sample-todetector distance (q range ϭ 0.00365 to 0.383 Å Ϫ1 ). All purified PLC⑀ and PLC␤3-⌬892 variants were diluted to a final concentration of 2-3 mg/ml in 20 mM HEPES, pH 8, 200 mM NaCl, 2 mM DTT, 0.1 mM EDTA, and 0.1 mM EGTA. Samples were centrifuged at 16,000 ϫ g for 5 min at 4°C before data collection. Protein samples were eluted on a Superdex 200 Increase column (GE Healthcare; Figs. S1 and S2). The eluate was passed through the UV monitor and then the SAXS flow cell, which consists of a 1.5-mm quartz capillary with 10-m walls. Data were collected every 2 s with 1-s exposure times. For PLC⑀ PH-COOH, frames corresponding to the leading edge of the elution peak as opposed to the highest point of the peak were selected for SAXS analysis. The usual practice of averaging frames corresponding to the highest point in the peak manifested symptoms of aggregation in the Guinier analysis due to possible capillary fouling (Fig. S1A). For all other PLC⑀ and all PLC␤3-⌬892 variants, frames corresponding to the protein peak were used for SAXS analysis (Fig. S1, B-F). A buffer file was generated for all samples by averaging exposures flanking the elution peak and was used for background subtraction (Fig. S1).
Raw scattering data were reduced using BioXTAS RAW 1.4.0 (55) and ATSAS (56). The low variability in the distribution of R g corresponding to the individual frames as seen plotted with the "scattering chromatograms" (integrated intensity of individual exposures plotted against frame number) demonstrates the monodisperse nature of the samples (Fig. S3). For oxidized PLC␤3-⌬892 E60C/V164C, automated Guinier approximation for individual frames was not possible due to low signal to noise, and we therefore used averaged frames 188 -193 to enable reliable Guinier approximation and calculation of the R g . We used MALS data and the molecular mass distribution across the elution peak to confirm the monodisperse nature of PLC⑀ PH-COOH, oxidized PLC␤3-⌬892 E60C/V164C, and reduced PLC␤3-⌬892 E60C/V164C (Fig. S4). For the SEC-MALS-DLS (dynamic light scattering) experiments, the samples were loaded on the SEC column with an exclusion limit of 1.25 MDa (WTC-030S5, Wyatt Technologies), which was followed by a UV detector, a MALS detector, and a DLS detector (DAWN Helios II, Wyatt Technologies). This experiment enabled us to not only establish that the samples were monodisperse but also to acquire a more reliable estimate for the molecular mass of samples analyzed using the ASTRA software (Wyatt Technologies) that were ambiguous with SAXS. I(q) versus q curves generated after buffer subtraction and averaging selected exposures were used for more detailed analyses of the SAXS data. R g , I(0), and D max for each protein were obtained using PRIMUS (57). 10 -15 ab initio models were calculated using DAMMIF, and then aligned and averaged with DAMAVER (58). For all PLC⑀ and PLC␤3-⌬892 variants, the mean NSD for all calculated models ranged from 0.560 Ϯ 0.018 to 0.968 Ϯ 0.032. One ab initio model was excluded from both the PLC⑀ EF-COOH and oxidized PLC␤3-⌬892 E60C/V164C data sets (discard criteria NSD Ͼ mean ϩ 2 ϫ S.D.). Final ab initio envelope structures were generated using DAMMIN (59). Plots were generated from buffer-subtracted averaged data (raw scattering and Guinier plots) or pair distance distribution (P(r) plots) and plotted using GraphPad Prism version 7.0.
Theoretical SAXS data for the PLC␤3 core crystal structure was generated using its coordinate file (PDB entry 3OHM (12)). CRYSOL3.0 (56, 60) was used to compare the calculated SAXS data with the experimentally determined data for PLC␤3-⌬892. SASREF (61) was used to generate a computational model that fit the empirical SAXS data. For this model, the following interdomain distance restraints were placed. The C terminus of the PH domain (residues 10 -146) was within 5 Å of the N terminus of the EF1/2 hands (residues 147-221), and the C terminus of the EF1/2 hands was within 5 Å of EF3-proximal CTD (residues 222-892). The theoretical SAXS scattering curve was then compared with the empirically determined data. Plots were generated from buffer-subtracted averaged data (raw scattering and Guinier plots) or averaged envelope data (P(r) plots) and plotted using GraphPad Prism version 7.0.

Negative stain EM sample preparation and imaging
For each PLC⑀ variant, 3 l of freshly purified protein (0.015-0.018 mg/ml) was applied to glow-discharged carboncoated 400-mesh copper grids (Electron Microscopy Sciences) and stained with 0.75% uranyl formate (47). Samples were imaged with a CM200 field electron gun transmission electron microscope (Phillips) operated at 200 kV. Micrographs were collected with a US4000 4k ϫ 4k CCD camera (Gatan), using a 1-s exposure time, defocus range of 5 m, and at a nomal magnification of 50,000ϫ, resulting in a sampling of 2.31 Å/pixel at the specimen level.

Single-particle 3D EM reconstructions
Data were processed with a beta release of RELION 2.1 (62). Raw micrographs were CTF-corrected, and particles were autopicked before reference-free 2D class averaging. For PLC⑀ PH-COOH, 47,300 particles were extracted, and the 2D classes contained 34,155 particles. For PLC⑀ PH-C2, 124,747 particles were extracted, and the 2D classes contained 91,982 particles. For PLC⑀ EF-COOH, 58,549 particles were extracted, and the 2D classes contained 53,028 particles. To minimize any potential model bias, de novo initial models were generated for PH-COOH, PH-C2, and EF-COOH from their respective 2D class averages. From the respective selected 2D classes for each variant, we generated two particle subsets containing ϳ100 particles per 2D class to build de novo initial models with the RELION-2.1b1 stochastic gradient descent method. One initial model per subset was chosen at random and used as a reference for 3D class averaging of particles from the best 2D classes. Four 3D classes were generated per initial model. When possible, the dominant 3D class from each of the two sets of four 3D classes was chosen for 3D auto-refinement, resulting in two final, refined structures. The representative workflow is shown in Fig.  S5. Fourier shell correlation (FSC) curves corresponding to refined models were calculated using standard methods (63).

Liposome pelleting assays
200 M hen egg white phosphatidylethanolamine and 100 M porcine brain phosphatidylinositol 4,5-bisphosphate (Avanti Polar Lipids) were mixed and dried under N 2 . Lipids were resuspended in sonication buffer (50 mM HEPES, pH 7, 80 mM KCl, 3 mM EGTA, and 2 mM DTT) and sonicated in two 30-s bursts (14,38). For oxidized PLC␤3-⌬892 E60C/V164C, DTT was omitted from the sonication buffer. Liposomes (65 l) were combined with either 2 g of total PLC⑀ variant (0.15 nmol of PLC⑀ PH-COOH, 0.19 nmol of PLC⑀ PH-C2, 0.17 nmol of PLC⑀ EF-COOH) or 1 g of total PLC␤3-⌬892 variant (0.125 nmol) and sonication buffer to a final volume of 100 l. Control binding reactions contained protein and buffer only, and total protein controls were incubated with liposomes or buffer. All samples were incubated for 1 h on ice. All samples, excluding total protein control samples, were then centrifuged at 119,000 ϫ g for 1 h. Total protein control samples were incubated on ice for the duration of the ultracentrifugation. For centrifuged samples, the supernatant was removed, and the pellet was resuspended in 100 l of sonication buffer. The PLC⑀ pellets were also subjected to sonication to resuspend the pellet. For analysis by SDS-PAGE and ImageJ, 16 l of the supernatant, resuspended pellet, or total protein control was denatured with 4 l of 5ϫ SDS loading dye, and 5 l of the total sample was analyzed by SDS-PAGE. All gels were stained with Bio-Safe Coomassie Blue (Bio-Rad), and band density was quantified with ImageJ.

Statistical methods
All plots were generated using GraphPad Prism version 7.0. One-way ANOVA was performed with Prism 7.0 and followed by Tukey post hoc multiple comparisons as noted in the figure legends. All error bars represent S.D.

Figure creation
Single-particle EM 3D envelopes were rendered with UCSF Chimera (45). SAXS molecular envelopes were generated from dummy atom coordinate files with e2pdb2mrc.py (64) using a low-pass filter to 20 Å and rendered in Chimera. All ribbon diagrams and dummy atom models were generated using PyMOL (Schrödinger, LLC, New York).