End-Product Diacylglycerol Enhances the Activity of PI-PLC through Changes in Membrane Domain Structure

Abstract

Diacylglycerol (DAG)-induced activation of phosphatidylinositol-phospholipase C (PI-PLC) was studied with vesicles containing PI, either pure or in mixtures with dimyristoyl phosphatidylcholine, distearoyl phosphatidylcholine, sphingomyelin, or galactosylceramide, used as substrates. At 22°C, DAG at 33 mol % increased PI-PLC activity in all of the mixtures, but not in pure PI bilayers. DAG also caused an overall decrease in diphenylhexatriene fluorescence polarization (decreased molecular order) in all samples, and increased overall enzyme binding. Confocal fluorescence microscopy of giant unilamellar vesicles of all of the compositions under study, with or without DAG, and quantitative evaluation of the phase behavior using Laurdan generalized polarization, and of enzyme binding to the various domains, indicated that DAG activates PI-PLC whenever it can generate fluid domains to which the enzyme can bind with high affinity. In the specific case of PI/dimyristoyl phosphatidylcholine bilayers at 22°C, DAG induced/increased enzyme binding and activation, but no microscopic domain separation was observed. The presence of DAG-generated nanodomains, or of DAG-induced lipid packing defects, is proposed instead for this system. In PI/galactosylceramide mixtures, DAG may exert its activation role through the generation of small vesicles, which PI-PLC is known to degrade at higher rates. In general, our results indicate that global measurements obtained using fluorescent probes in vesicle suspensions in a cuvette are not sufficient to elucidate DAG effects that take place at the domain level. The above data reinforce the idea that DAG functions as an important physical agent in regulating membrane and cell properties.

Introduction

Phosphatidylinositol phospholipase C (PI-PLC) is a phosphodiesterase that hydrolyzes the phosphodiester bond between glycerol and phosphate in PI, yielding lipid-soluble diacylglycerol (DAG) and water-soluble phosphorylinositol (1). Like other lipases, but at variance with most enzymes, PI-PLC acts on a substrate that does not occur in solution, but rather is found in the solid (aggregated) state, i.e., in the cell membranes. This has a number of consequences that render the kinetics and regulation of these enzymes rather unique phenomena. The substrate may exhibit very different physical properties (e.g., diffusion rate and molecular order) according to the membrane composition, temperature, etc., and those properties may in turn influence the enzyme activity. Moreover, one of the reaction end-products, DAG, is also a lipid that exhibits structural and physical properties very different from those of PI, with the result that the enzyme activity modifies the physical state of the substrate and the latter influences any further enzyme activity.

In an early study, Titball (2) reported the presence of PI-PLC in Bacillus cereus and other bacteria. He also noted that sequence homology had been reported between B. cereus PI-PLC and several eukaryotic PI-PLCs. More recent studies (1,3) have revealed the crystal structures of the B. cereus (4) and other bacterial PI-PLCs. These enzymes exhibit a number of interesting properties. For example, interfacial activation, V_max/K_M, increases when the PI substrate is in an aggregated rather than a monomeric form (5–7), and surface dilution inhibition (i.e., the enzyme-specific activity) decreases when the surface concentration of the substrates is diluted with nonsubstrate lipids even if the bulk substrate concentration is kept constant (7,8). Bacterial PI-PLCs operate in the scooting mode, completing various rounds of substrate turnover at the substrate interface before dissociating from the cell or vesicle (7). These enzymes are activated by the nonsubstrate phosphatidylcholine (PC) in the form of vesicles or micelles via allosteric binding of PC to PI-PLC (8–10). Many other important aspects of the structure and mechanism of PI-PLC have been described, as reviewed in Goñi et al. (1).

Previous studies from our laboratory have described the role of lipids other than PI and PC in PI-PLC (11). Simultaneous measurements of enzyme activity, interfacial enzyme binding, and fluorescence of different probes revealed that both enzyme binding and activity decreased with increasing lipid molecular order and with increasing vesicle diameter (decreasing bilayer curvature). For the particular case of DAG, Villar et al. (12,13) found that this lipid was an efficient activator of PI-PLC. In fact, DAG has been shown to mix nonideally with phospholipids, giving rise to in-plane separations of DAG-rich and -poor domains, even in the fluid state. DAG molecules increase the negative curvature of the lipid monolayers, and DAG-phospholipid mixtures tend to convert into inverted nonlamellar phases. Moreover, DAG has been found to activate a number of membrane-related enzymes (see Goñi and Alonso (14) for a review of DAG properties). Here, we studied in parallel the effects of DAG on the physical properties of bilayers and on PI-PLC activity and membrane binding. We conclude that DAG decreases the bilayer average molecular order and induces lateral domain separation in vesicles. PI-PLC binds preferentially to DAG-induced fluid domains, which are rapidly hydrolyzed.

Materials and Methods

Wheat-germ PI and egg DAG were purchased from Lipid Products (South Nutfield, UK). Egg sphingomyelin (SM), pig brain galactosylceramide (GalCer), dimyristoylphosphatidylcholine (DMPC), and distearoylphosphatidylcholine (DSPC) were obtained from Avanti Polar Lipids (Alabaster, AL). Bovine serum albumin (essentially free from fatty acids) and DiI-C18 were obtained from Sigma (St. Louis, MO). PI-PLC (EC 3.1.4.10) from B. cereus was obtained from Molecular Probes (Grand Island, NY), and labeled PI-PLC, 1,1′-dioctadecyl-3,3,3′3′-tetramethylindocarbocyanine perchlorate (DiI), and 2-dimethylamino-6-lauroylnaphthalene (laurdan) were supplied by Invitrogen (Eugene, OR). Enzyme preparations contained on average 2800 U/mg of protein. Salts, organic solvents, and other reagents were of analytical grade and were supplied by Merck (Darmstadt, Germany). All buffers were prepared with Milli-Q water (Millipore, Billerica, MA).

Liposome preparation

Large unilamellar vesicles (LUVs) were prepared by the extrusion method using Nuclepore filters (Costar, Cambridge, MA; 0.1 μm pore diameter unless otherwise stated) at room temperature, as detailed previously (15). The vesicles were prepared in 10 mM Hepes, 50 mM NaCl (pH 7.5). Vesicle size was estimated by quasi-elastic light scattering using a Malvern Zeta-sizer instrument. Irrespective of lipid composition, vesicle diameters were in the 100–110 nm range. More details regarding the preparation of these vesicles can be found in Nieva et al. (15).

Enzyme activity assays

The PI concentration was 0.3 mM in all experiments unless otherwise stated. For optimal catalytic activity, the enzyme was assayed at 39°C in 10 mM Hepes, 50 mM NaCl (pH 7.5), in the presence of 0.1% bovine serum albumin with continuous stirring (39°C is the optimal enzyme temperature, and our previous studies with this enzyme were carried out at that temperature (11–13)). Enzyme was used at a final concentration of 0.16 U/mL, equivalent to 57 ng protein/mL. The total volume in the enzyme assays was 1.0 mL. Enzyme activity was assayed by determining water-soluble phosphorus. Aliquots (50 μL) were removed from the reaction mixture at regular intervals and extracted with 250 μL of a chloroform/methanol/hydrochloric acid mixture (66/33/1 by volume), and the aqueous phase was assayed for phosphorus. For this purpose, the samples were digested with perchloric acid and the resulting free phosphate was assayed colorimetrically in the form of a molybdate/malachite green complex.

Fluorescence polarization assays

The fluorescence polarization of 1,6-diphenyl-1,3,5-hexatriene (DPH) was measured using an SLM 8100 spectrofluorometer equipped with standard polarization accessories and a circulating water bath. The excitation and emission wavelengths were 360 and 430 nm, respectively. The fluorescence polarization was calculated as

$$P=\left(I\text{vv-}\mathit{GI}\text{vh}\right)/\left(I\text{vv}+GI\text{vh}\right),$$

where Ivv and Ivh represent the intensity of vertically and horizontally polarized fluorescent light, respectively, when excitation light is vertically polarized. The correction factor G = Ihv/Ihh. The fluorescent probe DPH was added to the phospholipid to obtain a probe/lipid molar ratio of 1:250.

Intrinsic fluorescence assay of interfacial enzyme binding

The intrinsic fluorescence spectra of PI-PLC either alone or in the presence of vesicles, but always in the absence of bovine serum albumin, were recorded in an SLM Aminco spectrofluorometer equipped with a thermostated cell holder and a magnetic stirrer. Small aliquots of a concentrated LUV suspension were added to a fixed concentration of enzyme solution (20 nM) in a cuvette with continuous stirring. Samples were excited at 295 nm and emission was collected between 300 and 420 nm. The slit width was 5 nm for both excitation and emission. In some experiments, a fixed lipid concentration was used (134 μM) and the change in fluorescence emission was measured after 1 min. In other cases, the intrinsic fluorescence of PI-PLC was titrated with increasing amounts of lipid added at 1.5 min intervals.

Spectra were corrected for the light scattered by LUVs of identical composition and concentration, but in the absence of protein. The signal was also corrected for dilution and inner filter effects using the soluble Trp analog NATA, which does not partition into membranes. The apparent mole fraction partition coefficients, K_x(app), were determined by fitting the experimental values to the hyperbolic function

$$\left(F/F_0\right)=\frac{1+\left(F_{\max }/F_0\text{-1}\right)\left[\mathrm{L}\right]}{K+\left[\mathrm{L}\right]},$$

where [L] is the lipid concentration and K is the lipid concentration at which the bound enzyme fraction is 0.5. Then, K_x(app) = [W]/K, where [W] is the molar concentration of water.

The fraction of PI-PLC bound to the membranes was estimated according to the equation

$$f_{\text{bound}}=K_{\mathrm{x}\left(\text{app}\right)}\left[\mathrm{L}\right]/\left(\left[\mathrm{W}\right]+K_{\mathrm{x}\left(\text{app}\right)}\left[\mathrm{L}\right]\right).$$

GUV preparation and fluorescence microscopy

Giant unilamellar vesicles (GUVs) were prepared according to the electroformation method developed by Angelova et al. (16), modified as described previously (17), in a special chamber supplied by L.A. Bagatolli (Odense, Denmark) that allows direct visualization under the microscope (18). Stock solutions of lipids (0.2 mg/ml total lipid containing 0.2 mol % DiI or 2 mol % laurdan) were prepared in a chloroform/diethylether/methanol (4:5:1, v/v) solution. Then, 3 μl of the appropriate stocks was added onto the surface of Pt electrodes and solvent traces were removed by evacuating the chamber under high vacuum for at least 2 h. The Pt electrodes were then covered with 400 μl of 10 mM Hepes, 150 mM NaCl, pH 7.4, previously equilibrated to 60°C. The Pt wires were connected to an electric wave generator (TG330 function generator; Thurlby Thandar Instruments, Huntington, UK) under AC field conditions as follows: 1) frequency 500 Hz, amplitude 30 mV for 5 min; 2) frequency 500 Hz, amplitude 300 mV for 20 min; and 3) frequency 500 Hz, amplitude 900 mV for 90 min.

Direct visualization of GUVs

After GUV formation, the chamber was placed on an inverted confocal fluorescence microscope (Nikon D-ECLIPSE C1; Nikon, Melville, NY). The excitation wavelength for DiI was 561 nm. The images were collected using band-pass filters of 593 ± 20 nm. When required, 1 μl of PI-PLC labeled with Alexa Fluor 633, at 200 μg/ml, was added to study its effect on the GUVs. In the latter case, an excitation light at 635 nm was used and images were collected using a long-pass filter of 650 nm. All of these experiments were performed at 25–28°C. Image treatment was performed using the software EZ-C1 3.20 (Nikon).

Generalized polarization of laurdan-labeled GUVs

Laurdan-labeled GUVs were visualized on an inverted confocal microscope (Leica TCS-SP5) using a 63×, 1.2 NA objective. Laurdan is a polarity probe; upon excitation, its emission undergoes a spectral shift due to the reorientation of water molecules present in the glycerol backbone region of the membrane. This shift can be correlated to the lipid phase (19). In the gel phase, when no water is present, laurdan emission peaks at 440 nm, whereas in the liquid crystalline phase the spectrum is red-shifted ≈50 nm. This shift is quantified using the generalized polarization (GP) function:

$$\text{GP}=\left({\mathrm{I}}_{\mathrm{r}}\mathit{-}{\mathrm{G}}_{\text{GP}}{\mathrm{I}}_{\mathrm{f}}\right)/\left({\mathrm{I}}_{\mathrm{r}}+{\mathrm{G}}_{\text{GP}}{\mathrm{I}}_{\mathrm{f}}\right),$$ (1)

where I_r and I_f are the emissions collected in the red and blue spectral windows, respectively. The G_GP factor is used in the GP value to compensate for the difference in the efficiency of collection in regions I_r and I_f. The G factor is calculated according to the equation

$${\mathrm{G}}_{\text{GP}}=\left(\mathrm{G}{\mathrm{P}}_{\text{ref}}+\mathrm{G}{\mathrm{P}}_{\text{ref}}\mathrm{G}{\mathrm{P}}_{\text{mes}}{\text{-GP}}_{\text{mes}}^{\text{-1}}\right)/\left(\mathrm{G}{\mathrm{P}}_{\text{mes}}+\mathrm{G}{\mathrm{P}}_{\text{ref}}\mathrm{G}{\mathrm{P}}_{\text{mes}}{\text{-GP}}_{\text{ref}}^{\text{-1}}\right),$$ (2)

where GP_mes is the GP value of laurdan in pure dimethylsulfoxide (DMSO) using the same setup and settings as those used for the real sample. GP_ref is the reference value of laurdan in DMSO (GP_ref = 0.207) (20).

Laurdan was excited at 780 nm (two-photon excitation, Mai-Tai HP DS; Spectra Physics) and its emission was collected in two spectral windows (I_r, 415–445 nm; I_f, 490–510 nm) using nondescanned photon-counting hybrid detectors. A variable retarder (Berek polarization compensator, model 5540; New Focus) placed in the excitation beam before the microscope scanning head ensured circularly polarized two-photon excitation at the sample plane, effectively compensating for any mispolarization due to the reflecting components inside the microscope. The variable retarder was adjusted so that no photoselection effects would be visible on the equatorial plane of spatially homogeneous (monodomain) GUVs, thereby avoiding artifacts in the quantitation of the GP images due to uneven excitation of the fluorochromes.

GP images were calculated by computing the GP function for each pixel in the image using in-house-developed software according to Eq. 1.

NMR measurements

LUVs were prepared in the presence of H₂O/D₂O. The final phospholipid concentration was 40 mM. NMR spectra were recorded in a Bruker AV500 spectrometer (Karlsruhe, Germany) operating at 500 MHz for protons, and 202.4 MHz for ³¹P and full proton decoupling. The instrument was equipped with an inverse broad-band probe of 5 mm and gradients in the z axis, and 24,000 scans were averaged for each measurement. The data were registered and processed with TOPSPIN 1.3 software from Bruker.

Results

DAG and PI-PLC activity

Under the conditions described in Materials and Methods, PI-PLC activity was linear for the first 5 min of the assay (see the case for pure PI vesicles in Fig. 1 A). Rates were usually measured within the first 3 min. The observed rate for PI-PLC acting on pure PI bilayers was 4.0 nmol/min, corresponding to a specific activity of 70 nmol/min × μg of protein. The presence of egg DAG in the bilayers, up to 40 mol %, did not cause any marked change in the enzyme rate (Fig. 1 B).

Figure 1.

PI-PLC assays on PI bilayers in the presence of other lipids. (A) Time course of PI-PLC activity on pure PI vesicles. (B) PI-PLC initial rates on PI/DAG vesicles as a function of DAG mol % concentration in the bilayers. (C–F) PI-PLC initial rates on bilayers containing PI plus various concentrations of DMPC, SM, DSPC, or GalCer in the absence (●) or presence (○) of DAG at an additional 33 mol %. PI concentration (0.3 mM) was kept constant. T = 39°C. Average values ± SD; n = 3. In some cases, the SD bars are smaller than the symbols. To see this figure in color, go online.

A series of experiments were performed at 22°C and 39°C in which the substrate consisted of mixtures of PI and other lipids (DMPC, DSPC, SM, and GalCer) at different proportions, and the fraction of nonsubstrate lipid varied from 0 mol % to 80 mol %. The PI concentration in the assays was kept constant. When required, DAG was added to the bilayers at 33 mol % of the total lipids. Fig. 1, C–F, show the data for the four series of compositions tested at 39°C. In the absence of DAG, PI-PLC behaves as expected from our previous studies (11), i.e., there is a tension between the activating effect of phosphorylcholine and the inhibition due to increased lipid order. For a lipid such as DSPC, the final outcome is that there is no net change in enzyme activity (Fig. 1 E).

In the case of PI-SM mixtures (Fig. 1 D), the activatory effect prevails up to 30 mol % SM, but at higher SM ratios, the ordering effect of the high-melting-point SM predominates. Since DMPC is fluid at the assay temperature of 39°C, only an activating effect is seen (Fig. 1 C). Correspondingly, for the high-melting-point GalCer, which does not possess a phosphorylcholine group, only an inhibition is observed (Fig. 1 F).

The effect of DAG at 33 mol % is to induce a clear enzyme activation in mixtures of PI with DSPC, SM, or GalCer, but not with DMPC (Fig. 1). Together with the observation that DAG did not activate PI-PLC when acting on pure PI, this finding leads us to conclude that DAG has an activating effect only when the bilayers contain lipids with a high melting temperature, i.e., the molecular order of the bilayer at the assay temperature is high. The above observations are summarized for the equimolar mixtures PI/X at 39°C in Fig. 2 A. When similar experiments are performed at 22°C (Fig. 2 B), the results are comparable except that DAG activates PI-PLC acting on PI/DMPC. This is probably due to the fact that 22°C is just below the DMPC melting point (23°C), meaning that the order of PI/DMPC bilayers will be relatively high. The requirement of a highly ordered bilayer in order to detect a clear activating effect of DAG is illustrated in more detail in Fig. 3, where enzyme activities with or without DAG are shown as a function of temperature.

Figure 2.

(A and B) PI-PLC rates measured at 39°C (A) and 22°C (B). Initial rates on bilayers containing equimolar mixtures of PI and other lipids in the absence (gray bars) or presence (open bars) of DAG at an additional 33 mol %. Other conditions as in Fig. 1. Average values ± SD; n = 3.

Figure 3.

PI-PLC rates as a function of temperature. (A) PI-PLC initial rates on pure PI or PI:DAG vesicles. (B–E) PI-PLC initial rates on bilayers containing PI + equimolar concentrations of DMPC, SM, DSPC, or GalCer, respectively. Assays were performed in the absence (●) or presence (○) of DAG at an additional 33 mol %. PI concentration (0.3 mM) was kept constant. Average values ± SD (n = 3). In some cases, the SD bars are smaller than the symbols. To see this figure in color, go online.

Some temperature effects are interesting irrespective of bilayer order, mainly the drop in enzyme activity observed above 40°C in most assays in the presence or absence of DAG (Fig. 3, A–E). The decreased activity presumably reflects the enzyme thermal denaturation. In cases in which DAG shows a clear activatory effect at 39°C, e.g., with PI/SM and PI/DSPC (Fig. 3, C and D), the lipid environment appears to protect the enzyme from denaturation even at 60°C.

Effect of DAG on the lipid molecular order

In view of the above observations, the molecular order of the different bilayer compositions, with and without DAG, was measured as DPH fluorescence polarization. The results are summarized in Fig. 4. As expected, DAG decreased polarization (i.e., decreased the lipid chain molecular order (21)) for all of the lipid mixtures under study. In all cases, a decrease of 15–25% was observed. Thus, there was no correlation between the rather similar decrease in order (Fig. 4) and the very different effects of DAG on enzyme activity (>5-fold activation for mixtures of PI with SM, DSPC, or GalCer versus little or no change for pure PI or PI/DMPC bilayers). Moreover, the PI/GalCer/DAG mixture, which allowed a high PI-PLC activity (Fig. 1 F), had a higher molecular order than PI/SM or PI/DSPC, which supported only a low activity (Fig. 1, D and E). Thus, the overall changes in membrane molecular order, as revealed by DPH polarization, do not explain the effects of DAG on PI-PLC activity.

Figure 4.

Effect of DAG on lipid molecular order. Bilayer molecular order was measured at 22°C as polarization of DPH fluorescence emission. When required, DAG was added as an additional 33 mol %. Average values ± SD; n = 3).

DAG and PI-PLC binding to lipid bilayers

Binding of PI-PLC to lipid bilayers was studied through changes in the intrinsic fluorescence of PI-PLC Trp residues upon titration of the protein with increasing amounts of lipid vesicles. Binding is accompanied by an increase in fluorescence, due to the less polar environment experienced by Trp when it is transferred from an aqueous to a nonpolar environment in the bilayer. The apparent binding constants K are estimated from fluorescence versus [L] hyperbolic plots (Fig. 5, A and B and Fig. S1, A, C, and E in the Supporting Material) obtained by titrating the intrinsic enzyme fluorescence with increasing amounts of lipid. Then, the fractions of bound lipid are computed for each lipid concentration using K_x(app) as described in Materials and Methods. K and f_bound values are shown in Table 1 and Figs. 5 and S1, respectively. The data show that 1) DAG clearly favors PI-PLC binding to the bilayer, 2) this effect is more marked for those bilayers that were originally more ordered, and 3) there is a good correlation between increased binding and increased enzyme activation. Since the enzyme is active during the binding experiment, a certain proportion of PI, estimated in the 0–10% range, is hydrolyzed by the enzyme. This probably alters somewhat the f_bound values shown in Figs. 5 and S1, but the main result of the experiment, i.e., that DAG facilitates enzyme binding to the bilayer, remains fully valid.

Figure 5.

PI-PLC binding to lipid bilayers. (A and B) Binding was measured at 22°C as changes in the Trp intrinsic fluorescence of PI-PLC to bilayers containing PI (A) or equimolar mixtures of PI and SM (B). Measurements were performed in the absence (●) or presence (○) of DAG at an additional 33 mol %. (C and D) Fraction of bound enzyme as a function of lipid concentration. Average values ± SD; n = 3. Sometimes the error bars are smaller than the symbols. Similar data for other lipid compositions are shown in Fig. S1. Some relevant results are compiled in Table 1.

Table 1.

PI-PLC binding to lipid bilayers in the absence or presence of DAG

	K (mM)	Binding ratio
PI	0.93 ± 0.010
PI/DAG (1:0.33)	0.44 ± 0.005	2.1 ± 0.2
PI/DMPC (1:1)	0.44 ± 0.004
PI/DMPC/DAG (1:1:0.33)	0.058 ± 0.001	3.2 ± 0.2
PI/SM (1:1)	2.38 ± 0.042
PI/SM/DAG (1:1:0.33)	0.080 ± 0.012	13.0 ± 0.2
PI/DSPC (1:1)	3.16 ± 0.013
PI/DSPC/DAG(1:1:0.33)	0.125 ± 0.004	18.5 ± 1.0
PI/Galcer (1:1)	1940 ± 130
PI/Galcer/DAG (1:1:0.33)	18.7 ± 0.57	971 ± 24

Binding constants K (mM) are the lipid concentrations at which the fluorescence increase is half-maximal for each bilayer composition. Data from experiments as shown in Figs. 5, A and B, and S1, A, C, and E. Binding ratios correspond to f_bound with DAG/ f_bound without DAG at 100 μM PI. Data were obtained from experiments as shown in Figs. 5, C and D, and S1, B, D, and F. Average values ± SD; n = 3.

LUV binding assays were carried out at 22°C, a temperature at which enzyme activity is relatively low (Figs. 2 and 3). Under our experimental conditions, during K_x(app) measurements (when LUV did not initially contain DAG), the DAG concentration increased by an amount between 0% and 10% of the original PI, due to the enzyme activity. This figure is difficult to establish with precision, but the aim of the experiment, i.e., to show that the presence of DAG increases enzyme binding, is clearly achieved as shown by the data in Fig. 5.

Bilayer lateral heterogeneity and the effect of DAG

Confocal microscopy of GUVs is a convenient method to explore lateral heterogeneity, or domain formation, in bilayers (22–24). We examined GUVs of each lipid mixture under study at 22°C by fluorescence confocal microscopy to explore any possible relationship between DAG-induced enzyme activation and domain formation. All mixtures were stained with laurdan and examined by two-photon confocal microscopy. The so-called laurdan GP is a useful parameter to provide information about molecular order at the lipid-water interface (19,25), i.e., the level at which PI-PLC interacts with the bilayer. Representative images are shown in Fig. 6. PI/DMPC GUVs (Fig. 6, A and B) exhibit hardly any lateral heterogeneity, and DAG does not have any clear effect on the morphology of these bilayers. The same is true for pure PI GUV (not shown). However, in GUVs made of PI/SM (Fig. 6, C and D), DAG induces lateral phase separation, giving rise to less ordered domains (blue-green) coexisting with more ordered (yellow) domains. With PI/GalCer GUVs, domains are clearly visible in both the absence and presence of DAG (Fig. 6, E and F).

Figure 6.

Confocal microscopy of GUV containing PI and other lipids. Bilayer compositions are (A) PI:DMPC (1:1, mol ratio), (B) PI:DMPC:DAG (1:1:0.66 mol ratio), (C) PI:SM (1:1 mol ratio), (D) PI:SM:DAG (1:1:0.66 mol ratio), (E) PI:GalCer (1:1 mol ratio), (F) PI:GalCer:DAG (1:1:0.66 mol ratio), and LAURDAN staining. (G) PI:DSPC:DAG (1:1:0.66 mol ratio), DiI staining. (H) Same vesicles as in (G), but the stain is Alexa 635-labeled PI-PLC. Bar: 5 μm.

Laurdan GP was separately measured for the less and more fluid domains whenever phase separation could be distinguished (Fig. 7 A). Laurdan staining showed lateral phase separation in certain lipid mixtures (PI/SM/DAG, PI/DSPC, PI/DSPC/DAG, PI/GalCer, and PI/GalCer/DAG), but not in others. Moreover, and in a somewhat counterintuitive way, DAG caused the formation of ordered phases when they did not exist (Fig. 7 A, PI/SM), or increased the order in preexisting ordered domains (Fig. 7 A, PI/DSPC). It appears that DAG facilitates lateral phase separation of the bilayer components. Laurdan fails to reveal any DAG effect on PI/GalCer bilayers, even if the glyceride modifies the properties of those bilayers in many ways (Figs. 3 E, 5 F, and 6, E and F). This is probably explained by the low insertion of laurdan into the rigid GalCer-containing bilayers.

Figure 7.

Quantitative data obtained by confocal microscopy from experiments as shown in Fig. 6. (A) GP of laurdan in GUV regions. Black bars: less ordered domains; gray bars: more ordered domains. (B) Fluorescence intensity of membrane-bound Alexa 635-labeled PI-PLC. Open bars: homogeneous vesicles. Black bars: vesicles containing domains in which the enzyme is only found in the less ordered regions. Average values ± SD. Ten vesicles were measured in three different preparations for each lipid composition.

Fluorescence confocal microscopy also allows one to image enzyme docking to the GUV bilayers when the enzyme is labeled with an appropriate probe (Alexa 635) that does not modify its catalytic activity (22). As expected from the binding data in Fig. 5, the normalized intensity of Alexa 635 increases clearly in the presence of DAG (quantitative data in Fig. 7 B). More specifically, in heterogeneous vesicles, the enzyme binds preferentially to the more fluid regions, which are also labeled by DiI. An example of the PI/DSPC/DAG mixture is shown in Fig. 6, G and H. Alexa 635 labeling of the more fluid domains was also found for PI/SM/DAG and PI/DMPC/DAG mixtures. PI-PLC follows the trend of other phospholipases that preferentially bind the most fluid domains (23).

We conclude from the above data that DAG acts by increasing the bilayer affinity for the enzyme (Fig. 5). Moreover, in a mechanism that may or may not be related to the above findings, DAG operates through the generation of separate more and less fluid (ordered) domains (Figs. 6 and 7) in such a way that the more fluid (less ordered) ones bind enzyme with high affinity (Fig. 7). Mixtures containing GalCer would again constitute an exception to this rule, since in such mixtures domain separation is clearly visible even in the absence of DAG (Fig. 6, E and F).

Lipid phase studies by ³¹P-NMR

Multilamellar vesicles (MLVs) composed of the various mixtures under study, in the presence and absence of DAG, were examined by ³¹P-NMR (26). The results are shown in Fig. S2. In the absence of DAG (left-hand column), all spectra reveal lineshapes indicative of the lamellar phase, with a main peak and a shoulder at the lower-field side of the spectrum. Two peaks are resolved when the mixture contains two different phospholipids. The main effect of DAG is the appearance of an isotropic signal (at 0 ppm) that is predominant in the PI/GalCer/DAG sample. An isotropic signal in this kind of spectra may be due to a cubic phase or to the presence of small vesicles (<60 nm) whose rapid tumbling averages out the chemical-shift anisotropies. To distinguish between these two phenomena, the PI/GalCer/DAG dispersions were centrifuged under conditions such that the MLV and cubic phases, but not the small vesicles, would sediment. The sediments were resuspended in buffer and reexamined by ³¹P-NMR. The relative height of the isotropic signal was clearly decreased in the sediment spectrum (Fig. S2, top right), indicating that DAG had given rise to small vesicles. In our previous study of PI-PLC (11), we showed that the enzyme activity increased rapidly with decreasing vesicle radius. Thus, the DAG-activating effect in the case of PI/GalCer bilayers is at least partly due to this phenomenon of small vesicle shedding.

Discussion

DAG is an interesting lipid in the context of PLCs because in addition to being very potent in modifying the bilayer properties (14), it is also an enzyme end-product. Previous studies from our laboratory showed that DAG is an efficient activator of PI-PLC (12,13). In those studies, vesicles composed of PI, PE, PC, and cholesterol were used, and the PI proportion varied between 5 mol % and 40 mol %. When egg DAG was incorporated into the liposome formulation at an additional 5 mol % or 10 mol %, enzyme activity increased in a dose-dependent way. DAG is known to activate a series of membrane-related enzymes (see review in Goñi and Alonso (14)). Villar et al. (13) explained the observed activation in terms of a DAG-induced increase in the bilayer’s propensity to form inverted phases, which is known to facilitate protein insertion into membranes (27,28). Also, DAG was suggested to act through a decrease in membrane order (14), which would in turn enhance PI-PLC docking to the membrane and thus increase enzyme activity.

The results presented here, which were obtained with a large variety of lipid compositions and temperatures, provide a rather different view of the effects of DAG. First, DAG does not always activate PI-PLC (for example, see Fig. 1, B and C). Second, although DAG does cause a decrease in membrane order (Fig. 4), this does not correlate with the changes in enzyme activity (Figs. 1, 2, and 3), i.e., DAG causes a similar decrease in order (decreases DPH polarization) in PI/SM and PI/DMPC mixtures, but only in the former is PI-PLC activation observed (Figs. 1, C and D, and 4). Third, under our experimental conditions, DAG failed to cause any marked formation of nonlamellar inverted phases (Fig. S2), and the observed isotropic signal was mainly due to the generation of small-diameter vesicles.

Observing GUVs by fluorescence confocal microscopy allowed us to gain a more thorough understanding of the effect of DAG on PI-PLC. This approach was pioneered by Holopainen et al. (29) and later applied to the study of PLC by Riske and Döbereiner (30) and Ibarguren et al. (22,23). The results in Figs. 6 and 7 show that DAG can markedly modify the bilayer lateral heterogeneity. Note that the GUV experiments were performed at 22°C and should be correlated to the enzyme activities shown in Fig. 2 B. The main effect of DAG is to increase the difference in lipid molecular order between the more and less ordered domains in the vesicles, or even to generate, or make visible (as in PI/SM), a phase separation that did not exist, or could not be seen, in its absence (Figs. 6, C and D, and 7 B). (GalCer-containing bilayers exhibit a different behavior; see below.) For pure PI vesicles in which DAG addition causes no increase in enzyme activity, no change in lateral segregation is seen either (Figs. 6, A and B, and 7 A), and the increase in bound enzyme is modest (Figs. 5 F and 7 B). PI/DMPC mixtures represent an intermediate case: DAG addition somewhat increases PI-PLC activity at 22°C (Fig. 2 B) and increases enzyme binding (Fig. 7 B) without causing visible (macroscopic) phase separation (Figs. 6, A and B, and 7 A). However, for the PI/SM and PI/DSPC bilayer compositions in which DAG increases PI/DSPC activity, DAG causes macroscopic domain separation (in the former) or increases this effect (in the latter) (Figs. 6, C and D, and 7 A). Specifically for PI/DSPC, the ordered domains are made even more ordered (Fig. 7 A) in the presence of DAG. The latter may occur because in the presence of DAG, PI moves to the more fluid domains, i.e., the possibility of separate fluid domains induced by DAG causes an overall rearrangement of the lipid distribution in the bilayer. Changes in domain distribution and fluidity upon addition of more or less rigid lipids are commonly observed (24,31,32). In general, GUV microscopy studies allow us to study heterogeneities that would go undetected using spectroscopic techniques with samples in cuvette. As an example, Figs. 5 and S1 show an overall increase in enzyme binding upon addition of DAG, but Fig. 7 C reveals the increased binding that takes place specifically in the fluid regions of the vesicles.

Taken together, the above results show that DAG activates PI-PLC whenever it can increase enzyme binding. This increase does not appear to be due to an overall fluidization such as that detected through changes in DPH polarization (Fig. 4). Rather, it appears to be the DAG-induced lateral separation of more fluid domains (Fig. 7) that facilitates enzyme binding to a region that allows a high enzyme activity. PI/DMPC mixtures at 22°C may be a hypothetical case of DAG-promoted microdomains that are not detected by confocal microscopy but favor enzyme binding and activity (Figs. 2 B, 6, A and B, and S1 B). GP of laurdan, which reports on molecular order at the lipid-water interface, was not useful for predicting changes in PI-PLC activity when used in cuvette studies (11), but provides meaningful information when GP data from specific domains can be retrieved.

Vesicles containing GalCer constitute a particular case. As mentioned above, DAG causes a large increase in enzyme activity (at >20 mol % GalCer; Figs. 1, 2, and 3) and increases overall enzyme binding (Fig. S1), and yet laurdan fails to detect any effect (Fig. 7). In the absence of specific proof, we suggest that because of the tight hydrogen-bonding net that occurs in ceramides and their derivatives (33,34), the affinity of these bilayers for the fluorescent probes is altered, and thus the information provided by these probes cannot be compared with that obtained in other bilayers. Alternatively, or complementarily, the activating effect of DAG on PI/GalCer bilayers may occur through the DAG-dependent shedding of small vesicles (Fig. S2), which PI-PLC is known to degrade at high rates (11).

An additional mechanism for DAG-dependent PI-PLC activation that may coexist with DAG-induced lateral domain generation is DAG-generated packing defects in the bilayer. Recent studies (35,36) have shown how conically shaped lipids, particularly dioleoylglycerol, can induce lipid packing defects and facilitate the adsorption of peripheral proteins that use hydrophobic residues as membrane anchors. This was already hinted at by Ahyayauch et al. (11) in the context of PI-PLC binding to bilayers. The packing-defects activation mechanism may have been operating in many or all of the mixtures studied here in addition to the domain-formation mechanism discussed above. In particular, lipid packing defects may explain the DAG activation in PI-DMPC mixtures in which no DAG-induced macroscopic domain formation was detected.

The biological implications of our study are important. In addition to being an end-product of PI-PLC, DAG regulates its activity through changes in the physical properties of the bilayer. Basáñez et al. (37) described the lag time of PC hydrolysis by bacterial PI-PLC, and explained the phenomenon as being due to a slow accumulation of DAG in the bilayer. Once the DAG concentration reached a certain level (∼10 mol %), enzyme binding to the bilayer was much increased and a burst of enzyme activity followed. A similar behavior was found for PI-PLC (12) and attributed to the same effect. These data are in agreement with our current observations of increased enzyme binding promoted by DAG. With respect to the nonsubstrate polar lipids used in our study, even if saturated phosphatidylcholines (e.g., DMPC and DSPC) are not common in cell membranes, the relevance of domains enriched in SM and/or glycosphingolipids has been put forward repeatedly (38,39). Moreover, microdomains in the gel phase may exist in eukaryotic membranes (40) that are susceptible to fluidization by DAG generated by a neighboring PI-PLC. It could be argued that the DAG concentrations used in this study (33 mol %) are too high to be of physiological significance. This could be true considering the average lipid composition of the whole membrane, but given that the lipids that influence the activity of an enzyme are mostly those in the close vicinity, it is perfectly feasible that a local DAG concentration such as used in our study would be achieved in that microenvironment. Enzyme activity usually drops above 40°C (Fig. 3), but even at high T the enzyme activity is usually much higher in the presence of DAG. This may be further proof that DAG facilitates enzyme binding to the bilayer, with protein stability often being enhanced in a hydrophobic (membrane or detergent) environment (41–43). Thus, the above data suggest that DAG is a physical (and not only chemical) regulator of cell properties.