Lipid composition is expected to play an important role in modulating membrane enzyme activity, in particular if the substrates are themselves lipid molecules. A paradigmatic case is FAAH (fatty acid amide hydrolase), an enzyme critical in terminating endocannabinoid signalling and an important therapeutic target. In the present study, using a combined experimental and computational approach, we show that membrane lipids modulate the structure, subcellular localization and activity of FAAH. We report that the FAAH dimer is stabilized by the lipid bilayer and shows a higher membrane-binding affinity and enzymatic activity within membranes containing both cholesterol and the natural FAAH substrate AEA (anandamide). Additionally, co-localization of cholesterol, AEA and FAAH in mouse neuroblastoma cells suggests a mechanism through which cholesterol increases the substrate accessibility of FAAH.
FAAH (fatty acid amide hydrolase) is a membrane-bound enzyme that is responsible for the intracellular hydrolysis of the bioactive lipid AEA (anandamide or N-arachidonoylethanolamine) and other congeners known as eCBs (endocannabinoids) . The discovery of FAAH  and the demonstration that it terminates the signalling and biological activity of AEA in vivo has inspired pharmacological strategies aimed at augmenting the eCB tone through FAAH inhibitors (for a review see ). The 3D structure of FAAH has been resolved at a high resolution for a truncated form of the rat enzyme , and for a humanized form of the same enzyme where specific residues were substituted to generate a human-like active site . The collected structural information has identified the catalytic mechanism of FAAH, but a role for the surrounding membrane lipids in controlling the enzyme's activity remains as yet unknown.
Membrane lipid composition affects eCB uptake and signalling, and cholesterol has been demonstrated to be a key determinant of this regulation . In many cell types, increasing or decreasing membrane cholesterol content enhances or inhibits AEA uptake respectively [6,7]. A contributing factor to this effect may be the affinity of cholesterol for AEA . In the same context, cholesterol has been shown to modulate AEA binding to the CB1 (type-1 cannabinoid receptor)  and the TRPV1 (type-1 vanilloid receptor) . Recently, it has been proposed that cholesterol modulates AEA transport across the membrane, leading to increased AEA hydrolysis by FAAH . However, a description of the possible molecular mechanism of direct FAAH modulation by cholesterol is still missing. In C6 glioma cells, both cholesterol enrichment and depletion in the PM (plasma membrane) did not modulate FAAH activity , but, due to a fine regulation of cholesterol homoeostasis within the cell, a variation of membrane cholesterol content is unlikely to promptly influence the concentration of cholesterol within the intracellular compartments (for a review see ).
With the aim of dissecting the requirements for catalytic activity and enzyme interaction with membranes, using SAXS and FRET we analysed the conformational changes induced by lipid bilayers on FAAH. Using both synthetic and reconstructed lipid vesicles from different cellular compartments [i.e. PM or ER (endoplasmic reticulum)] we demonstrated a key role for membrane lipids in stabilizing a dimeric form of FAAH with cholesterol and AEA, both of which modulate its enzymatic activity within the membrane. Furthermore, MD simulations supported a novel mechanism by which cholesterol may help to open the membrane port of FAAH  and thus increase its accessibility to the enzyme substrate AEA. Overall, the results of the present study further our understanding of the molecular events underpinning the modulation of membrane enzymes by the surrounding lipids. FAAH has been proposed as a paradigm of such membrane enzymes that bind lipophilic substrates (generally embedded within the membrane) and cleave them with water to release hydrophilic products .
MATERIALS AND METHODS
Reagents and antibodies
All chemicals were of the purest analytical grade. AEA, cholesterol, AA (arachidonic acid), ethanolamine, PEA (N-palmitoylethanolamine), SEA (N-stearoylethanolamine), DPPC (dipalmitoyl phosphatidylcholine) and the protease inhibitor cocktail were from Sigma Chemicals. POPC (1-palmitoyl-2-oleoyl phosphatidylcholine), DOPC (dioleoyl phosphatidylcholine), brain SM (sphingomyelin; containing almost exclusively C18:0) and cholesterol were purchased from Avanti Polar Lipids. IPTG was purchased from Promega, trehalose was from Cargill and the Talon resin was from Clontech. b-AEA (biotin–AEA) was synthesized as reported previously .
Cy2 (carbocyanine)–streptavidin and DyLight 405–streptavidin were purchased from Jackson ImmunoResearch Labs. Alexa Fluor® 555-conjugated CTB (cholera toxin B), donkey anti-rabbit Alexa Fluor® 488-conjugated secondary antibodies and Prolong Gold anti-fade kit were purchased from Invitrogen. Rat FAAH was expressed with a His6 tag in Escherichia coli BL21 (DE3) pLysS competent cells (Merck) using the pTrcHisAFAAH-ΔTM plasmid  and was purified as reported previously . Rabbit anti-FAAH polyclonal antibody was purchased from Primm.
SAXS measurements were performed using the SWING beamline at the SOLEIL synchrotron radiation facility (Saint-Aubin, France). The measuring cell under vacuum was kept at a constant temperature (10°C) during the measurements. A total of 25 frames of 1 s each were recorded and averaged after visual inspection for radiation damage (none was found). The scattering intensities were measured over the q range qmin=0.0005 Å−1 (1 Å=0.1 nm) to qmax=0.50 Å−1 [q=(4π/λ)sinθ, with 2θ being the scattering angle]. FAAH samples were measured at a concentration of ~1.0 mg/ml. The radius of gyration (Rg) and the molecular mass of the proteins were calculated using the Guinier approximation . The distance distribution function p(r) was determined using the indirect Fourier-transform method of the GNOM program package . Ab initio calculations of the shape of FAAH from the SAXS pattern were performed using the DAMMIN program . A total of ten independent calculations were performed, averaged and filtered after their consistency was checked with the SUPCOMB program , as described previously . All of the models obtained yielded very similar shapes, as shown by the low value (0.651±0.012) of the NSD (normalized spatial discrepancy) calculated using the DAMAVER program suite . As FAAH is a membrane enzyme prone to aggregation in aqueous solutions, we evaluated the best experimental conditions able to stabilize soluble forms of the protein to make them suitable for SAXS analysis. We found that, among several factors, concentrations of Tris/HCl buffer ≥ 0.5 M and a high ionic strength with salt (e.g. NaCl) concentrations ≥ 0.1 M are able to avoid the formation of insoluble aggregates due to the high protein concentration used for SAXS analysis. Thus the purification and the stabilization of the protein was carried out with 0.1 M NaCl dissolved in 0.5 M Tris/HCl buffer (pH 7.5). The SAXS data produced under these experimental conditions were used to obtain the overall conformation of FAAH in solution.
The possible effect of different membrane lipid compositions on the structure of FAAH was also investigated by SAXS using POPC, DPPC and the major FAAH substrate AEA at a submicellar concentration (<20 μM) . All lipids were diluted into the protein solution [0.1 M NaCl dissolved in 0.5 M Tris/HCl buffer (pH 7.5)] titrating a lipid stock solution (2 mM dissolved in ethanol). Under these experimental conditions no micelle formation was observed by SAXS. Lipids were analysed at dilutions at least 5-fold higher than the protein (FAAH) concentration. None of these experimental conditions affected the scattering curve of FAAH oligomers. Higher lipid concentrations were not used because they form micellar structures with a volume several fold larger than that of FAAH, thus completely masking the scattering signal of the protein.
Fluorescence spectra were recorded at 25°C using a PerkinElmer LSB50 fluorimeter and 10 mm×2 mm path length quartz fluorescence microcuvettes (Hellma). The pyrene-bound liposomes used in FRET studies contained 5% (w/w) Py-PE (1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-1–pyrenesulfonyl) purchased from Molecular Probes. FAAH was used at a final concentration of 0.2 μM, whereas the liposome concentration varied between 10 and 600 μM in a final volume of 100 μl. Using these diluted protein samples, both in the presence and absence of 0.1 M NaCl, aggregates and differences in the membrane-binding properties of FAAH were not observed (results not shown). Solutions containing freshly prepared FAAH and Py-PE liposomes were used in all experiments. Emission spectra (λex=292 nm and λem=300–420 nm) were recorded in lysis buffer without trehalose. Data were plotted as the fractional loss of tryptophan fluorescence (ΔF/ΔFmax) against the liposome concentration as described previously . Experimental data were analysed by non-linear regression through an hyperbolic binding isotherm using the Kaleidagraph program (Synergy Software). LUVs (large unilamellar vesicles) were prepared with POPC as described previously . LUVs containing lipid rafts (described herein as lipid rafts) were prepared with DOPC, brain SM and cholesterol in an equimolar ratio (1:1:1) as reported previously . Cholesterol- and AEA-enriched LUVs were prepared using POPC, AEA and cholesterol at molar ratios from 10:1:1 to 10:1:5. The same vesicles were also prepared with PEA, SEA, AA and ethanolamine instead of (and at the same molar ratio as) AEA.
Cell membrane isolation and characterization
The PM and ER membrane were isolated from rat livers as described previously  with some modifications. Briefly, the hepatic tissue was disrupted using a Potter glass and mixed with an equal volume of disruption buffer [HB buffer: 5 mM Tris/HCl (pH 7.8) containing 10 mM KCl and 5 mM EDTA]. After centrifugation at 2000 g for 5 min at 4 °C the supernatant was transferred to a 12 ml ultracentrifugation tube with the sequential addition of 9.6 ml of 60% sucrose buffer, 4.5 ml of 37% sucrose buffer and Top solution (1 ml of HB buffer) followed by centrifugation at 150000 g for 2 h at 4 °C. In the final step the PM and the membrane of the ER were recovered with a glass Pasteur pipette, washed with HB buffer and then centrifuged for 1 h at 150000 g at 4 °C. The final pellet was dried for 12 h in a desiccator and stored at −20°C until use. The analysis of PE (phosphatidylethanolamine), PS (phosphatidylserine), PC (phosphatidylcloline) and SM in the PM and ER extracted from liver was carried out by HPLC using a Luna Silica 3 mm, 150 mm×4.6 mm column (Phenomenex), using acetonitrile/methanol/85% phosphoric acid (100:10:0.8, by vol.) as the mobile phase (flow rate, 1 ml/min) and a UV detector set at 205 nm (PerkinElmer). Quantitative analysis was performed by the external standard method for each of the investigated lipids; the calibration linear range was 0.010–2 mg/ml. PM and ER extracts were dissolved in chloroform and injected in the HPLC system without further treatment. Identification of the compound was on the basis of the retention time. Cholesterol quantification was obtained with the same chromatographic column using hexane/propan-2-ol (99:1, v/v) as the mobile phase (flow rate, 0.6 ml/min) and a UV detector set at 202 nm. The calibration linear range was 0.005–1.5 mg/ml and sample extracts were dissolved in chloroform before injection into the HPLC system.
Imaging and co-localization analysis
All experiments were carried out using a Leica TCS SP5 DMI6000 confocal microscope. Details of cell culture, treatment and staining are reported in the Supplementary Materials and methods section (at http://www.biochemj.org/bj/457/bj4570463add.htm). For co-localization analysis, we determined the Pearson's correlation coefficient and the intensity correlation quotient using the ImageJ plugin JACoP that groups together the most important co-localization tools . The value of Pearson's correlation coefficient can range from +1 to −1; +1 represents a perfect correlation, −1 represents a perfect exclusion and 0 represents a random localization . In addition, apparent co-localization due to random staining or a very high intensity in one window will have values of intensity correlation quotient near to zero, whereas if the two signal intensities are interdependent (co-localized) these values will be positive with a maximum of 0.5 .
The results of the present study are means± S.D. for least four independent determinations each performed in triplicate. Statistical analysis was assessed by the non-parametric Mann–Whitney U test using GraphPad Prism 5.
All MD simulations were performed using ACEMD  locally and on the GPUGRID.net distributed computing network . All simulations were all-atom, explicit solvent simulations using TIP3P water. The CHARMM 27 forcefield  was used for the protein; CHARMM Lipid C36 parameters  were used for POPC and cholesterol; and AEA was parameterized with the CGenFF forcefield and the ParamChem web service . The protein structure was taken from PDB code 1MT5, but with the MAFP (methoxyarachidonoyl fluorophosphonate) inhibitor removed. CHARMM-GUI  was used to build the membrane components. Details about the set-up, equilibration and production runs of all simulations are reported in the Supplementary Materials and methods section.
FAAH structure is modulated by membranes
In the present study the truncated form of rat FAAH, resolved at 2.8 Å by X-ray crystallography , was analysed by SAXS. As calculated from the p(r) function (and in accordance with the Guinier analysis; Figures 1A and 1B), the SAXS analysis of FAAH yielded a radius of gyration of 129±13 Å and a value of the maximum dimension of the particle (Dmax) of 410±10 Å. Ab initio calculations of the shape of the molecule, performed from the SAXS pattern with the DAMMIN program , indicated that FAAH has an overall structure with three lobes. Each of them is well-fitted by the crystallographic unit of the protein, that is an octamer of dimers (Supplementary Figure S1 at http://www.biochemj.org/bj/457/bj4570463add.htm). Indeed, superimposing the crystallographic unit of FAAH (PDB code 1MT5) on the low-resolution DAM models of the oligomers, it appears that in the absence of detergent FAAH is composed of three octamers (Figure 1C). Thus the enzyme shows a peculiar topology in solution where the crystallographic unit seems to behave as a building block for an oligomeric structure elongating along a specific axis, leading to an unprecedented structural organization of the enzyme (Figure 1C). As FAAH is an enzyme stably bound to membranes in vivo, the protein oligomer described above may be catalytically active in solution, but does not represent the bioactive conformation. Thus, in order to mimic the membrane milieu, we used detergents with a small micellar volume . In the presence of 1% (w/v) Triton X-100 we obtained a scattering profile where a typical oscillation (at q values higher than 0.12 Å−1), due to the form factor of the detergent micellar structure, was evident (Figure 2).
Oligomerization state of FAAH in solution
Effect of detergents that mimic the membrane milieu on FAAH structure
The Guinier analysis of this scattering curve yielded a value of the radius of gyration of 35.6±0.2 Å (Figure 2), very similar to the theoretical value obtained from the dimeric biological unit of the crystallographic structure  (Supplementary Figure S1). Thus in the presence of the detergent FAAH is mainly organized as a dimer, suggesting that upon interaction with membranes the oligomeric form of FAAH dissociates. Interestingly, Triton X-100 did not affect the dependence of FAAH activity on substrate concentration. This followed typical Michaelis–Menten kinetics yielding Km, kcat and kcat/Km values of 16±3 μM, 5.6±0.6 s−1 and 3.5×105 μM−1·s−1 respectively, in the presence and absence of the detergent. Incidentally, these values are in agreement with those already reported for FAAH by others [2,15]. Since the dissociated FAAH dimers had the same kinetic properties as the oligomeric form of the enzyme, it can be concluded that Triton X-100 does not affect the catalytic activity of FAAH, and that the oligomerization state of the protein does not influence the enzyme activity.
To further understand the effect of membranes on the structure and function of FAAH, we analysed the intrinsic fluorescence of the enzyme in the presence of LUVs constituted by POPC. The fluorescence spectrum of FAAH is very broad due to the presence of more than one emitting species ; however, increasing the concentration of POPC vesicles led to a relevant red-shift of the wavelength at the emission maximum (λmax from 343.5 to 346.0 nm; Figure 3). Interestingly, this effect was not replicated by simply diluting the enzyme, which suggests an overall higher exposure of the aromatic residues of the protein to the water solvent in the presence of membranes (Figure 3). Instead, the observed red-shift seems to be attributable to a general increase in the solvation of the protein due to the dissociation of the octameric structure. These data indicate that the protein oligomer observed at the high protein concentration used for the SAXS measurements (i.e. ~8.0 μM and ~1.0 mg/ml) was also present (Figure 3) in more diluted solutions (up to ~0.2 μM, i.e. ~0.03 mg/ml).
FAAH dissociation in the presence of membranes
FAAH preferentially binds to membranes containing both AEA and cholesterol
The structural analysis was extended further by measuring the membrane-binding properties of FAAH using FRET. To ascertain whether FAAH had any preference for specific membranes we made LUVs of different lipid compositions. The membrane binding of FAAH was not affected by the physicochemical properties of the membrane as shown by the half-saturation binding values ([L]1/2) being similar for membranes in the liquid-disordered crystalline phase (POPC vesicles) and the solid-like gel phase (β-phase) (DPPC vesicles) (Table 1). Therefore in the subsequent FRET experiments only POPC LUVs were used. This had the added biological relevance of POPC being one of the most common lipids in animal cell membranes . A role for cholesterol in modulating AEA movement within the membrane has been reported previously in biophysical studies [8,39]. In the present study, we found that the co-presence of AEA and cholesterol within POPC membranes, at a stoichiometric ratio from 1:1 to 1:5, induced a pronounced decrease in [L]1/2 (15±7); this indicates a strong increase (~5-fold over the POPC control) in the affinity of FAAH for AEA/cholesterol-containing membranes (Figure 4 and Table 1).
|LUV membranes||[L]1/2 (μM)|
|POPC/AEA/cholesterol (10:1:1 and 10:1:5)||15±7|
|Lipid rafts/AEA (10:5)||107±17|
|LUV membranes||[L]1/2 (μM)|
|POPC/AEA/cholesterol (10:1:1 and 10:1:5)||15±7|
|Lipid rafts/AEA (10:5)||107±17|
FAAH preferentially binds to membranes containing both AEA and cholesterol
To ascertain whether this preferential interaction could be ascribed to a general modification of the physicochemical properties of the membrane, or was rather due to the presence of a specific class of polyunsaturated lipids or functional groups, we tested also the effect of AEA hydrolysis products (AA and ethanolamine), and of other fatty acid amides, such as SEA and PEA. In all cases higher [L]1/2 values were obtained (Table 1), demonstrating that the preferential binding of FAAH occurred only with membranes containing AEA. As expected for a membrane enzyme associated with the ER , FAAH shows a higher binding affinity for membranes in the liquid-disordered state (POPC) and a weaker interaction with lipid raft structures typical of the PM . Thus we tested further whether the preferential binding of FAAH was maintained in lipid-raft-containing membranes.
Embedding AEA within lipid-raft-containing LUVs  did not significantly affect the binding of FAAH (Figure 4 and Table 1), suggesting that this enzyme prefers membranes containing AEA within cholesterol-rich regions in the liquid-disordered phase, rather than organized in lipid rafts. We extended these experiments further by analysing FAAH binding to PM and ER membranes isolated from the rat liver, as described in the Supplementary Materials and methods section. These results showed that FAAH preferentially binds to ER membranes (containing ~3% mol/mol of cholesterol; Supplementary Table S1 at http://www.biochemj.org/bj/457/bj4570463add.htm), yielding [L]1/2 values very similar to those obtained with AEA/cholesterol-containing synthetic vesicles (Figure 4 and Table 1). It is worth mentioning again that the ER membrane is the primary site of accumulation of AEA (Supplementary Figure S1).
FAAH activity is modulated by cholesterol
To clarify whether the preferential interaction of FAAH with AEA/cholesterol-containing membranes could also influence the enzyme's activity, we measured AEA hydrolysis by FAAH in the presence of different synthetic membranes. FAAH showed a similar catalytic activity in solution and when bound to POPC vesicles (Figure 5). Externally adding FAAH to POPC vesicles that contained increasing AEA/cholesterol molar ratios (from 1:1 to 1:5), an increased enzyme activity of up to ~4-fold over the controls was observed (Figure 5), indicating that FAAH activity can be modulated by the co-presence of cholesterol and AEA within the membrane in a cholesterol-concentration-dependent manner (Figure 5). This effect can be explained by the increased affinity of FAAH for membranes containing both AEA and cholesterol (Table 1), leading to higher amounts of membrane-embedded enzyme with respect to the control. Incidentally, the cholesterol/POPC ratio did not affect the specific activity of FAAH.
FAAH activity in the presence of membranes with different lipid compositions
FAAH co-localizes with AEA and cholesterol in mouse neuroblastoma cells
Fluorescence microscopy analysis of mouse N18 neuroblastoma cells revealed that FAAH is distributed in several dotted structures widely diffused in the cytoplasm and is particularly prominent in the perinuclear zone (Figure 6). These findings extend previous studies showing that FAAH is primarily associated with the membranes of the ER and excluded from the PM . To compare the degree of association of FAAH with different lipids and/or membrane compartments, a series of co-staining experiments was performed with specific lipid markers: b-AEA, a stable analogue of AEA, was used to stain the ER and adiposomes ; filipin, a fluorescent macrolide that specifically binds to both raft and non-raft membrane cholesterol, was used to stain the free cholesterol ; and Alexa Fluor® 555-conjugated CTB was used to stain the ganglioside GM1, a lipid selectively confined to lipid rafts . The extent of spatial overlap among FAAH and these lipids was measured using Pearson's correlation coefficient, a widely accepted parameter to quantify the degree of co-localization between fluorophores . A strong association between b-AEA and FAAH was found in N18 cells, with a higher co-localization between filipin-stained cholesterol and AEA, and a smaller, yet significant, association between filipin and FAAH (Figure 6). In contrast, FAAH staining did not significantly overlap with that of CTB (Figure 6), suggesting a lower affinity of FAAH for authentic lipid rafts. Overall, FAAH appears to be mainly localized in the AEA- and cholesterol-containing regions of internal membranes, but not in the PM lipid rafts.
Intracellular distribution of FAAH, AEA and cholesterol in mouse N18 neuroblastoma cells
MD simulation of FAAH within the membrane
The change in the catalytic activity of FAAH in the presence of cholesterol suggests a plausible entry pathway for AEA from the lipid bilayer. Indeed, a 200-ns long MD simulation of FAAH embedded in the membrane showed that the protein relaxes to adopt a conformation similar to one described previously , with the accessible entry of the AEA-binding pocket at the interface between the polar head of the membrane and the solvent. Such a conformation is compatible with the entry pathway of AEA into the binding cavity from the membrane.
In order to better understand the effect of cholesterol on the binding of AEA to FAAH, we simulated the full event of binding of the substrate , thus ascertaining whether a direct interaction with cholesterol may explain the preferential membrane binding and the increased catalytic activity of FAAH. We performed more than 1500 parallel simulations of FAAH associated with a membrane containing AEA/cholesterol/POPC at a 1:5:10 molar ratio, each for an average time of 165 ns (totalling over 250 μs of sampling time). Of these, 1000 simulations were free-ligand-binding simulations , in which 26 binding events were seen. An additional 550 simulations resampled these 26 events in order to improve the binding analysis and to clarify how AEA interacts with FAAH. The data yielded an interesting picture of AEA binding to FAAH (Figure 7 and Supplementary Movie S1 at http://www.biochemj.org/bj/457/bj4570463add.htm), suggestive of a specific role for cholesterol in the modulation of the enzyme activity. AEA approached FAAH at the previously proposed membrane port , exploring the pocket formed by helices α17, α18 and α19 and the loop between sheet β9 and helix α22 (Figure 7C, panel 1). The primary barriers to entry were steric hindrance by side chains and a salt bridge formed between those secondary-structure elements. Once these barriers were overcome, the polar head group of AEA moved into the enzyme where it formed weak and transient hydrogen bonds with water and the backbone of helix α9 and helix α22, as well as with the loop preceding it (Figure 7C, panel 2). AEA then displaced more water from the enzyme and moved further up into the enzyme (Figure 7C, panel 3), at which point the primary barrier was the highly disordered internal structure of the enzyme (in particular the very large loop between sheet β3 and helix α8), as well as further AEA dehydration. Finally, AEA took a similar conformation to that shown for the MAFP inhibitor in the crystal structure (Figure 7C, panel 4).
MD simulations of AEA binding to FAAH
The most significant outcome of the MD simulations was the possible explanation of the role of cholesterol in increasing AEA binding. When the salt bridge is intact, it closes the enzyme gate and AEA is prevented from entering. The continued interaction with the cholesterol head group helps to keep the port open while AEA enters, and allows other side chains to take conformations that make AEA binding easier. Cholesterol remains close to the enzyme even after AEA has fully entered, continuing to interact with the residues that keep the port open and fitting snugly against the enzyme in the region where AEA was before entering. In membranes made of only AEA/POPC, the probability that another lipid (e.g. another AEA) donates a hydroxy group for the same purpose would be significantly lower. Overall, MD simulations seem to explain the role of cholesterol in increasing FAAH activity as has been measured experimentally (Figure 5).
The exit of the AEA cleavage reaction products (i.e. ethanolamine and AA) is unlikely to be strongly modulated by cholesterol, whereas a role could also be played by increased lateral diffusion of AEA in cholesterol-containing membrane domains. To address this point, diffusion coefficients were computed for AEA using MD simulations. For the AEA/POPC (9:75) and the AEA/cholesterol/POPC (9:45:75) systems these coefficients were found to be 16.9±5.9×10−8 and 13.6±5.0×10−8 cm2/s respectively, with no statistically significant difference. Incidentally, these values for the diffusion coefficients are consistent with previous data obtained experimentally . Next, we investigated whether the effect of cholesterol on FAAH activity could be also related to an increased interlayer (flip-flop) movement of AEA , measured as the cholesterol-dependent increase in the AEA flip-flop rate between bilayers. Using all-atom simulations (see the Supplementary Materials and methods section), we did observe a qualitative higher flip-flop rate for AEA with cholesterol.
The crystallographic unit of FAAH constitutes octamers of dimers and the derived biological unit is a dimeric protein  (Supplementary Figure S1). The SAXS measurements reported in the present paper show that in solution FAAH adopts a higher hierarchical organization of its quaternary structure, constituting three octamers of dimers (Figure 1). Combined SAXS and fluorescence analyses show that the binding of FAAH to membranes (or to detergents that mimic the membrane milieu) induces a dissociation of these oligomers and stabilizes the dimeric form of the enzyme. Remarkably, the dimer of FAAH is a functional unit with the same catalytic activity as the octamer.
Additional FRET measurements allowed assessment of the preferential interaction of FAAH with membranes containing both AEA and cholesterol, but not organized in lipid rafts. The ~5-fold increase in their affinity of FAAH for membranes in the presence of cholesterol and AEA appears to be specific for AEA and not for other eCBs, and is not observed with either lipid alone. Moreover, data obtained with the PM and ER membrane from the rat liver are superimposable on to those obtained with synthetic vesicles. In particular, considering that ER membranes contain ~3% cholesterol and that AEA is mainly localized in this compartment (Supplementary Figure S1), the results of the present study strongly suggest a physiological role of these lipids in maintaining the proper intracellular localization of FAAH.
By confocal analysis we revealed that FAAH is consistently confined to AEA/cholesterol-rich regions, with a very low level of association with the GM1-rich membrane domains on the cell surface (Figure 6). Such a preferential binding of FAAH could contribute in vivo to stably maintain the enzyme localization in a compartment with a more efficient hydrolysis of AEA, i.e. those within cell membranes containing non-raft cholesterol, such as the ER, or those of intracellular stores of AEA, such as the adiposomes .
FAAH mainly interacts with one leaflet of the membrane bilayer (Figure 7A), thus it is conceivable that AEA reaches the active site of the enzyme only from that leaflet via the so-called membrane port . Cholesterol could induce a higher catalytic activity of FAAH by directly interacting with the enzyme, thus increasing the accessibility to AEA of the membrane port. Consistently, our MD studies clearly indicate that the binding of AEA is directly aided by cholesterol, which is more concentrated around the membrane port of FAAH (Figure 7B). Through its interaction with key salt bridge residues that gate the entrance of AEA, cholesterol facilitates opening of the port so that AEA can enter and reach the active site of the enzyme.
These results support the concept of a substrate–enzyme interaction, whereby a third player comes into the game: the membrane lipids. In the case of the metabolic enzymes of eCBs, these observations might have therapeutic relevance. On a more general note, the present study sheds light on the key role of the lipid environment in determining biological activity. This concept emerged in previous years for membrane receptors , and in the present study we show that it holds true also for a membrane enzyme such as FAAH, that is at the heart of eCB signalling in many pathophysiological events .
cholera toxin B
fatty acid amide hydrolase
large unilamellar vesicle
Annalaura Sabatucci, Clotilde Beatrice Angelucci and Enrico Dainese performed the SAXS experiments and analysis; Sergio Oddi performed the confocal analysis; Michele Del Carlo, Chiara Di Pancrazio and Enrico Dainese performed the membrane-binding analyses; Toni Giorgino, Nathaniel Stanley and Gianni De Fabritiis performed the MD simulations; Enrico Dainese and Mauro Maccarrone designed the experimental strategies; and Enrico Dainese and Mauro Maccarrone wrote the paper with the help of Gianni De Fabritiis and Michele Del Carlo and the input from all co-authors.
We thank SOLEIL for provision of synchrotron radiation facilities (to E.D.) and Dr Javier Perez for assistance in using the SWING beamline, and the volunteers of GPUGRID who donated graphics processing unit computing time to the project.
This work was supported by the Ministero dell’Istruzione, dell’Università e della Ricerca [grant number PRIN 2010-2011], by the Fondazione Italiana Sclerosi Multipla (FISM) [grant number 2010], the Fondazione della Cassa di Risparmio di Teramo TERCAS [Contract 2009-2012 (to M.M.)], the Spanish Ministry of Science and Innovation [grant number BIO2011-27450 (to G.D.F.)], the European Union via the Biostruct-X project within the FP VII programme (to E.D. and M.M.), the Consiglio Nazionale delle Ricerche via the short-term mobility programme 2013 (to T.G.) and the National Institutes of Health [grant number DA017259 (to B.F.C.)].
These authors contributed equally to this work.