Structural Features of Membrane-bound Glucocerebrosidase and α-Synuclein Probed by Neutron Reflectometry and Fluorescence Spectroscopy*

Background: A specific interaction exists between α-synuclein and glucocerebrosidase on the lipid membrane, resulting in enzyme inhibition. Results: Binding glucocerebrosidase has a profound effect on α-synuclein, moving roughly half of its embedded helical region above the membrane plane. Conclusion: A model is proposed with structural insights into glucocerebrosidase inhibition by α-synuclein. Significance: α-Synuclein-glucocerebrosidase interaction provides a molecular connection between Parkinson and Gaucher diseases. Mutations in glucocerebrosidase (GCase), the enzyme deficient in Gaucher disease, are a common genetic risk factor for the development of Parkinson disease and related disorders, implicating the role of this lysosomal hydrolase in the disease etiology. A specific physical interaction exists between the Parkinson disease-related protein α-synuclein (α-syn) and GCase both in solution and on the lipid membrane, resulting in efficient enzyme inhibition. Here, neutron reflectometry was employed as a first direct structural characterization of GCase and α-syn·GCase complex on a sparsely-tethered lipid bilayer, revealing the orientation of the membrane-bound GCase. GCase binds to and partially inserts into the bilayer with its active site most likely lying just above the membrane-water interface. The interaction was further characterized by intrinsic Trp fluorescence, circular dichroism, and surface plasmon resonance spectroscopy. Both Trp fluorescence and neutron reflectometry results suggest a rearrangement of loops surrounding the catalytic site, where they extend into the hydrocarbon chain region of the outer leaflet. Taking advantage of contrasting neutron scattering length densities, the use of deuterated α-syn versus protiated GCase showed a large change in the membrane-bound structure of α-syn in the complex. We propose a model of α-syn·GCase on the membrane, providing structural insights into inhibition of GCase by α-syn. The interaction displaces GCase away from the membrane, possibly impeding substrate access and perturbing the active site. GCase greatly alters membrane-bound α-syn, moving helical residues away from the bilayer, which could impact the degradation of α-syn in the lysosome where these two proteins interact.

Mutations in SNCA, the gene encoding for ␣-synuclein (␣-syn), 3 are linked to familial, early onset Parkinson disease (PD) (1)(2)(3), implicating this neuronal membrane-binding protein (4) as a pathogenic agent (5). Intracellular inclusions of fibrillar ␣-syn known as Lewy bodies are classic hallmarks of PD, and recently, glucocerebrosidase (GCase), a lysosomal hydrolase, has also been observed in them (6). Emerging data from numerous studies indicate a correlation between GCase deficiency and increased levels of ␣-syn (7)(8)(9). Importantly, mutations in GBA1, the gene encoding for GCase, are currently recognized as the most common risk factor for PD development (10) and recently have also been substantiated for dementia with Lewy bodies (11). Mutations in GBA1 also cause the lysosomal storage disorder Gaucher disease; it remains an open question as to how these two diseases may be linked (12).
GCase is a 497-residue lysosomal glycoprotein that catalyzes the hydrolysis of glucosylceramide into glucose and ceramide. Both anionic phospholipids and the activator protein, saposin C (Sap C), are considered vital for maximal catalytic activity (13)(14)(15)(16)(17); however, the presence of either alone is sufficient to promote enzyme function in vitro.
In our work (18,19), we have demonstrated an intimate molecular relationship by identifying a specific interaction between ␣-syn and GCase under the acidic solution conditions found in lysosomes, a site of ␣-syn degradation (20,21) and where GCase functions. Using site-specific fluorescence and Förster energy transfer probes (19), we found that ␣-syn-GCase interaction on the membrane surface involves both the N-and C-terminal portions of ␣-syn in contrast to that found in solution, which is specific only for the last 23 residues. Intriguingly, ␣-syn acts as a potent mixed inhibitor of GCase in vitro (19), but in the presence of its cofactor Sap C, GCase activity is fully rescued (22). Although the specific role of GCase activity in modulating ␣-syn levels remains controversial, this apparent reciprocal relationship is paving the way toward new PD treatments. For example, viral vector-mediated increase in wild-type GCase levels was shown to reverse PD-related features in a Gaucher disease mouse model (23,24), evidence that enhancing GCase activity is a potential therapeutic strategy for synucleinopathies.
Here, we sought to gain structural information on the membrane-bound ␣-syn⅐GCase complex. Although there are extensive structural studies on the nature of ␣-syn membrane interaction (4), little is known on how GCase associates with the membrane, although x-ray crystal structures of soluble GCase are available (25). We have employed neutron reflectometry (NR) to investigate the structural features of the two proteins on a sparsely tethered bilayer lipid membrane (stBLM) (26,27). NR is unique among available techniques in that it allows the structural characterization of membrane-associated proteins bound to actual lipid bilayers in an aqueous environment (28,29), as opposed to proteins in solid state or detergent-solubilized forms currently prevalent in the field.
The reflection of neutrons from stBLMs simultaneously probes the protein and the bilayer and readily distinguishes the layers of lipid acyl chains and phospholipid headgroups as well as the membrane-bound proteins (30 -36). Both the extent of polypeptide insertion into the bilayer and extension into the aqueous surrounding can be measured. Although it is of lower resolution than some structural methods, such as x-ray crystallography and solidstate nuclear magnetic resonance spectroscopy, NR has been successfully used to characterize a number of peripheral (e.g. ␣-syn (34), HIV-1 matrix (33), colicin N (35), PTEN tumor suppressor (30), and the GRASP Golgi stacking protein (31)) and integral membrane proteins (e.g. hemolysin (32) and outer membrane protein F (35)) on physiologically relevant, fluid phospholipid bilayers. By exploiting the high sensitivity of neutron scattering length density differences for deuterated ␣-syn (d-␣-syn) versus protiated GCase, the location of ␣-syn within the complex in the phospholipid bilayer was determined. Based on the orientation of GCase obtained from NR and other complementary techniques, including intrinsic Trp fluorescence and circular dichroism (CD) spectroscopy, an interaction model is proposed.
Preparation of Lipid Vesicles for Bromine Quenching and CD Experiments-Lipid vesicles were prepared as described previously (38). For Trp penetration depth experiments, brominated phosphatidylcholine lipids (Br 6 -7 , Br 9 -10 , and Br [11][12] ) were added to a final concentration of 30% while maintaining an overall 1:1 molar ratio of phosphatidylserine and phosphatidylcholine. Vesicle size (hydrodynamic radius ϳ45 nm) was characterized by dynamic light scattering measurements at 25°C using 5ϫ diluted lipid sample (ϳ0.6 mM) on a Dynapro Nano-Star (Wyatt) equipped with a 661-nm laser and Dynamics version 7.0.2 software. Autoattenuation of laser power was used to obtain optimal intensity counts of 1 ϫ 10 6 . Ten acquisitions with 5-s acquisition time were collected. Correlation time and peak cut-off ranges are 1.5-6.0 ϫ 10 4 s and 0.5-10,000.0 nm, respectively. Cumulant fit was applied to the autocorrelation curve to obtain the average hydrodynamic radius using a sphere model. Vesicle size distribution was extracted from the regularization fit using the Rayleigh sphere model (full width at halfmaximum ϳ60 nm). All buffers were filtered through 0.22-m filters (Millipore).
Preparation of stBLMs for Neutron Reflectometry-Sparsely tethered bilayer membranes were prepared as described previously with minor modifications (37). Vesicles (25 mg/ml) of the desired lipid composition were prepared by resuspending dried lipid films in aqueous pH 5.5 buffer containing 50 mM MES and 500 mM NaCl. The suspension was sonicated for 1 h in a bath sonicator and extruded at least 30 times through a 100-nm polycarbonate membrane using an extrusion kit (Avanti Polar Lipids). stBLMs were then formed by incubating the SAM with vesicle solutions for 1 h at room temperature. Non-fused vesicles were rinsed off with aqueous pH 5.5 buffer containing 50 mM MES and 25 mM NaCl.
Protein Expression and Sample Preparation-To produce uniformly deuterated wild-type ␣-syn, Escherichia coli BL21-(DE3)pLysS cells were first grown overnight at 37°C in a starter culture (1 ml of Luria broth) inoculated with a freshly transformed single colony. The resulting cells were then used to inoculate 20 ml of H 2 O-based minimal medium (39) overnight at 30°C. Finally, 1 liter of D 2 O-based minimal medium supplemented with deuterated glucose (1 g/liter) and deuterated Celtone (3 g/liter) (Cambridge Isotope Laboratories, Inc.) was then inoculated and grown overnight at 30°C. Cells were then transferred (250 ml into 2.5 liters) and grown to an OD of ϳ2 and induced with isopropyl 1-thio-␤-D-galactopyranoside (1 mM) for 6 h. Protein was purified as previously described (40). A calculated deuteration yield of ϳ98.9% was obtained from mass spectrometric analysis (15,233 Da in H 2 O). Taliglucerase ␣ (recombinant GCase) was obtained from Protalix Biotherapeutics Corp. (Carmiel, Israel). Samples were prepared in appropriate buffer (50 mM MES, 25 mM NaCl, pH 5.5), buffer-exchanged, and concentrated to the desired concentration using Amicon Ultra YM3 and YM30 centrifugal filters (Millipore) for ␣-syn and GCase, respectively.
GCase Membrane Binding Characterization-stBLMs formed on gold-coated microscopy glass slides were mounted on a custom-made surface plasmon resonance (SPR) instrument (SPR Biosystems, Germantown, MD) in the Kretschmann configuration (30). Bilayer completeness was characterized by electrochemical impedance spectroscopy and ensured to be Ͼ99%. Before the protein addition, the SPR resonance angle of the neat bilayer was monitored over time to obtain a stable baseline. Shifts in the resonance were used to quantify protein binding to the stBLM after injection of increasing concentrations of protein into the sample chamber. The system was allowed to equilibrate for 30 min before each subsequent protein addition at 20°C.
Trp fluorescence was measured ( ex ϭ 295 nm, obs ϭ 300 -500 nm, 0.3-s integration time, 1-nm slits, 25°C) in a microquartz cuvette with 3-mm path length (Starna Cells, working volume ϭ 120 l) using a Fluorolog 3 spectrofluorometer (Horiba Jobin Yvon) in the absence and presence (30 mol %) of brominated POPC lipids in POPC/POPS vesicles. Experiments were performed for 300 nM and 1 M GCase in 50 mM MES, 25 mM NaCl buffer, pH 5.5, and at a lipid/protein molar ratio of 350. Although variations in the absolute amount of quenching are observed, the trend relative to the three bromine positions is consistent and reproducible.
Neutron Reflectometry-NR measurements were performed at the NG7 and NGD-MAGIK reflectometers at the NIST Center for Neutron Research (NCNR) (41). Reflectivity curves were recorded for momentum transfer values 0 Յ q z Յ 0.25 Å Ϫ1 . stBLMs were prepared on SAM-coated 3-inch diameter silicon wafers assembled in an NCNR reflectometry flow cell (42). Every subsequent condition of the stBLM was measured using at least two isotopically different bulk solvents (contrasts) (i.e. aqueous buffer prepared from D 2 O and H 2 O). For each measured contrast, adequate counting statistics were obtained after 5-7 h. The flow cell allows for in situ buffer exchange; therefore, subsequent measurements were performed on the same sample area. The flow cell was maintained at room temperature. In order to minimize nonspecific binding, measurements of the stBLM with GCase (300 nM) alone were made after a short incubation period (20 min) followed by a rinse step. For measurements with ␣-syn (300 nM), a rinse step was avoided because its membrane binding is highly dynamic, and thus stBLM was incubated with the protein solution during the entire measurement time.
Data Analysis-SPR equilibrium response (R eq ) data were modeled by a Langmuir adsorption model (30), where p is the concentration of GCase, K d is the equilibrium dissociation constant, and B max is the saturation value of the SPR response at p 3 ∞. Data fitting was performed using IGOR Pro 6.32A (Wavemetrics). Trp mean depth (h m ) was obtained using the following equation (43), where F 0 and F are the Trp fluorescence intensities measured in the absence and presence of bromine; h is the distance from the bilayer center; and the fitting parameters, and S, are the dispersion and area, respectively. Data fitting was performed using IGOR Pro 6.32A (Wavemetrics). The one-dimensional structural profile along the lipid bilayer normal was obtained from NR using a hybrid of a stratified slab model for the solid substrate (44), a continuous distribution model for the stBLM (45), and a monotonic Hermite spline for the model-free envelope of the protein (28). Individual slabs were implemented for the bulk silicon, the silicon oxide, the chromium, and the gold layers. Fit parameters are thickness and neutron scattering length density for each layer (Table 1), except for the bulk silicon. One global roughness parameter applies to all substrate interfaces. Individual submolecular groups implemented in the continuous distribution model are ␤-mercaptoethanol, tether PEG chains, tether glycerol groups, substrate-proximal and substrate-distal phosphatidylcholine and phosphatidylserine headgroups, substrateproximal and substrate-distal methylene, and methyl chains of lipid and tether molecules. Fit parameters are the bilayer hydrocarbon thickness for each bilayer leaflet, bilayer completeness, tether surface density, tether thickness, and ␤-mercaptoethanol surface density (Table 1). One roughness parameter is applied to all distributions.
The Hermite spline that describes the volume occupancy profile of the protein, VO(z), is defined by control points, which are on average 15 Å apart. The spatial extension of the protein along the bilayer normal determines the number of control points, and it is determined iteratively during model optimization. Fit parameters for each control point are the volume occupancy of the envelope and the deviation from a position defined by equidistant control points throughout the spline. A constant neutron scattering length density (nSLD) is applied to spline functions that describe the volume occupancy of either completely protiated or deuterated protein material.
In the case of the ␣-syn⅐GCase complex that combines protiated GCase and d-␣-syn, the nSLD is a function of z: nSLD(z). This variable nSLD is modeled using a monotonic Hermite spline itself, sharing the position of the control points that define the volume occupancy spline. The nSLD can vary between values for entirely protiated and deuterated material (i.e. GCase and d-␣-syn). Volume occupancy profiles for the individual components of the complex are extracted after the fit by a linear decomposition. First, the nSLD profile is decomposed into the two contributions (frac syn (z), frac GCase (z)) from the protiated component (GCase, nSLD GCase ϭ 1.67 ϫ 10 Ϫ6 Å Ϫ2 ) and the deuterated component (␣-syn, nSLD syn ϭ 6.12 ϫ 10 Ϫ6 Å Ϫ2 ) using the equation, nSLD͑z͒ ϭ frac syn ͑z͒ ϫ nSLD syn ϩ frac GCase ͑z͒ ϫ nSLD GCase In a second step, individual volume occupancies for the protiated and deuterated components are calculated using the following equations. Data modeling that determines the orientation of a protein at the lipid bilayer requires the use of a structural model that is assumed to be invariant (rigid body modeling) (31,33). For GCase, the Protein Data Bank structure with the identifier 1OGS (46) was augmented by adding hydrogen atoms using the MolProbity Web service (47). The protein possesses four glycosylation sites, and glycans were added to those sites choosing the most frequently occurring species (48), using the appropriately patched ␣-mannose, ␤-xylose, N-acetyl-D-glucosamine, and fucose subunits from the CHARMM force field (49). The structure of the attached glycans was energy-minimized using NAMD, and the glycoprotein was equilibrated via a 10-ns simulation using the generalized Born approximation for the solvent (50). The obtained structure file was used for modeling of the NR data with two continuous fit parameters, equivalent to the two Euler angles (␤, ␥), describing the orientation of the protein at the interface.
The orientation of Chain A in 1OGS (46) is defined as the (0°, 0°) orientation. Every other orientation (␤, ␥) can be obtained by extrinsic rotations of the protein around the axes of the bilayer coordinate system. First, the protein is rotated by ␥ about the membrane normal, or the z axis. Second, the protein is rotated by ␤ about the x axis, which is in plane with the membrane. The third Euler angle, ␣, would correspond to a third extrinsic rotation around the z axis. NR is invariant to this orientation. The fit allows for orientations within 0°Յ ␤ Յ 90°a nd 0°Յ ␥ Ͻ 360°.
Data modeling and optimization of model parameters was performed using the ga_refl and Refl1D software packages developed at the NCNR (42). Optimization of model parameters is achieved either by the combined use of a genetic algorithm, a simplex amoeba algorithm for efficient searching of parameter space, and a Levenberg-Marquardt non-linear least square algorithm to refine the fit, or by using a differential evolution Markov chain. All reflectivity curves of one data set were fit simultaneously to the same model, sharing fit parameters, for example, for the solid substrate. Either a Monte Carlo simulation (26) or the statistical analysis of the Monte Carlo Markov chain (42) was used to determine the median model parameter values and their 68% confidence limits. Both methods deliver a bias-free and objective estimate of parameter uncertainties and, in this way, avoid overparameterization of the model.
The model structures of the membrane-associated ␣-syn⅐ GCase complex were based on the secondary structure of micelle-bound ␣-syn (51) and the x-ray structure of GCase (46), combined with the previously generated model of the ␣-syn⅐GCase complex in solution (18). The structures were generated with the Maestro and Prime programs (Schrödinger, Inc.).

Effect of ␣-Synuclein on GCase Membrane Binding Probed by
Trp Fluorescence and Circular Dichroism Spectroscopy-GCase contains 12 intrinsic Trp residues located within its hydrophobic interior and in solvent-accessible loops (Fig. 1A, top). To determine whether membrane association of GCase is modulated by ␣-syn, we measured the steady-state Trp fluorescence of GCase upon binding to phospholipid vesicles containing the heavy atom, collisional quencher bromine, which allows the determination of Trp side chain penetration depth (Fig. 1A,  bottom). ␣-Syn does not contain any native Trp residues and thus does not contribute to the measured intensities. Three independent bromine positions (Br 6 -7 , Br 9 -10 , and Br 11-12 ) served as markers for 11, 8.3, and 6 Å from the bilayer center (52). Equimolar phosphatidylcholine and phosphatidylserine (POPC/POPS) vesicles contained either none or one of the three different brominated lipids (30 mol %). The logarithm of the ratio of Trp fluorescence intensities measured in the absence (F 0 ) and presence of bromine (F) is plotted versus depth (43). Trp fluorescence quenching for GCase alone is observed for all sites, with Br 11-12 exhibiting the greatest effect, indicating that at least one or more of the Trp residues is partitioning into the hydrocarbon acyl chains, with a mean depth (h m ) of 6.6 Ϯ 0.7 Å from bilayer center, consistent with prior work using nitroxide spin-labeled lipids as quenchers (53). A marked difference is observed upon the addition of ␣-syn. Trp quenching is increased by Br 6-7 with a concomitant decrease by Br [11][12] , suggesting that ␣-syn lifts GCase up toward (ϳ2.3 Å) the lipidwater interface. A membrane-binding, N-terminal ␣-syn peptide (residues 1-19) had no effect on the GCase penetration depth, showing that perturbations of GCase membrane insertion only occur with full-length protein (data not shown).
To gain insights into whether this observed Trp displacement event is coupled to a structural rearrangement of the enzyme, we employed CD spectroscopy to probe for the changes in secondary structure content. The measured CD spectra are indistinguishable for GCase alone in buffer and in the presence of POPC/POPS vesicles, indicating the preservation of its folded structure on the membrane (Fig. 1B, top). Thus, the CD results require that any conformational changes that involve the insertion of Trp residues upon membrane binding must involve no net change of enzyme secondary structure.
For the complex with ␣-syn, however, the CD spectrum reveals a small increase in helical content when compared with the sum of the spectra of the individual proteins (Fig. 1B, bottom). Although it is not discernible whether the increased CD signal is solely attributable to GCase or ␣-syn or both, we postulate that it is more likely to be due to ␣-syn because it is known to be conformationally labile (4) and undergoes folding from disordered to helical structure in the presence of phospholipids (Fig. 1B, middle). Under this assumption, the change in secondary structure would be a modest ϳ5% increase in ␣-helicity, only seven additional helical residues. Based on the data, we suggest that GCase interaction with ␣-syn results in a shallower penetration depth of the enzyme on the bilayer with minimal change to its folded structure.
GCase Membrane Interaction and Structural Orientation Determined by Neutron Reflectivity-To obtain structural information on how GCase interacts with ␣-syn on a membrane, we first needed to measure neutron reflection of GCase on a POPC/POPS stBLM and determine its binding orientation. Membrane binding and binding affinity of GCase with POPC/POPS stBLM were confirmed and assessed by SPR spec-troscopy (Fig. 2). As expected, the relative membrane binding affinities correlate with GCase activity, with a strong preference for negatively charged phosphatidylserine lipids, confirming that the enzyme-lipid interaction is vital for catalysis (Fig. 2,  inset).  Best fit neutron scattering length density profiles for all four measurements calculated from the composition-space model are shown in the inset. B, simplified molecular distributions for each organic interface layer of the POPC/POPS stBLM and GCase obtained from the best fit of reflectivity data to the spline fit model. The median protein envelope is shown with its associated 68% confidence intervals (black). Volume occupancy is indicated on the right axis. Comparison of the median orientation fit using a structural model of glycosylated GCase is shown with 68% confidence intervals (green). See Table 1 for parameters.
Two reflectivity curves were measured using either H 2 O-or D 2 O-based buffer for the stBLM alone and the stBLM after GCase incubation, respectively. Representative D 2 O data are shown in Fig. 3A (Table 1). All measurements were evaluated using a global fit to obtain a structural profile of the stBLM and of the enzyme at the membrane interface (28,31). The resulting bilayer structure and median protein envelope yield an excellent fit to the measured reflectivity (Fig. 3B).
The protein envelope from the free form spline model indicates that GCase is peripherally associated with a peak density at a distance of 50 Å from the bilayer center, with the majority of GCase density in the aqueous surrounding. In accord with the Trp penetration result (Fig. 1A), a portion of the enzyme interacts within the phospholipid headgroups and penetrates into the acyl chain region of the outer leaflet. Using previously established rigid body modeling methods (28,31) and a Monte Carlo Markov chain data refinement algorithm (42), the spatial association and orientation of glycosylated GCase on the stBLM is obtained based on the GCase x-ray crystal structure (Protein Data Bank code 1OGS) (46). Within the 68% probability bounds (Fig. 4A), the majority of orientations are quite similar (Fig. 4B), with only modest variation in the Euler rotation angles of ␤ and ␥. A comparison of the free form spline envelope (black curve) and the envelope from the orientation fit (green curve) for values of ␤ Յ 90°is shown in Fig. 3B. The profiles that are outside the lipid membrane are nearly indistinguishable, suggesting that the globular fold of soluble GCase is preserved upon membrane association, consistent with the CD data (Fig.  1B). An example orientation (20°, 90°) is shown in Fig. 4C. GCase has a (␣/␤) 8 TIM barrel (Fig. 4C, taupe), which contains the two catalytic glutamic acids (red) and two flanking ␤-sheet domains (Fig. 4C, cyan). Of the probable orientations, this orientation best balances interactions between the glycans and the lipid headgroups with accessibility of the catalytic active site to the membrane.
The spline fit for GCase density shows greater density penetrating more deeply into the outer lipid leaflet compared with the orientation fit based on the crystal structure (Fig. 3B). From the orientation fit, the putative membrane interface (Fig.  4D) suggests that flexible loops 2 and 3 could insert into the bilayer, resulting in the penetration of Trp-348 and Trp-393 into the acyl hydrocarbon chains, as observed by fluorescence measurements. It is plausible that in order to accommodate side chain/peptide penetration, lipid reorganization also occurs, which could contribute to differences in the density profiles. Consistent with this view, a cluster of bound sulfate ions is seen interacting with loop 2 in the x-ray crystal structure, hinting that loop 2 might interact with phosphate headgroups of the membrane (54). Additional experimental support was obtained using a synthetic peptide consisting of the loop 2 sequence (Ac-GSKFWEQSVR-NH 2 ) to show that it has an affinity for POPC/ POPS vesicles (data not shown). In contrast, the corresponding loop 3 peptide (Ac-EGGPNWVRNFVD-NH 2 ) did not interact with the membrane based on Trp fluorescence (data not shown), but this result does not preclude GCase-membrane interactions that include the loop 3 Trp.
␣-Syn-GCase Interaction on the Membrane Probed by Neutron Reflectivity-The use of deuterated ␣-syn and protiated GCase facilitated extraction of the density profiles for each protein in the membrane-bound complex (Fig. 5). From these fits, there is an approximate 1.0:1.2 molar ratio of d-␣-syn and GCase at the membrane, consistent with the presence of 1:1 complex as determined previously (19). The extracted profile for GCase from the complex is in agreement with that obtained for GCase alone (i.e. the majority of the density is peripheral to the membrane, and a portion penetrates into the outer leaflet) (Fig. 5B), suggesting that GCase maintains a similar orientation in the complex. Interestingly, the mean of the penetration of GCase in ␣-syn⅐GCase complex does appear somewhat smaller than for free GCase but remains within 68% confidence limits (Fig. 5B).
The overall volume occupancy with the complex is about twice as large compared with data collected for GCase alone, reaching a maximum of 25% at 40 Å from the bilayer center.

TABLE 1 Median model parameter values and 68% confidence intervals for POPC/POPS stBLMs in the presence of GCase and ␣-synuclein
Parameters for the neat bilayer that were determined simultaneously are omitted. This level of coverage implies relatively close packing, with on average 20 -30 Å of unoccupied membrane lying between a protein complex and its nearest neighbor; thus, nonspecific steric contacts are possible. An additional small amount of protein density (1-7% volume occupancy) extending beyond 80 Å is also observed, which we attribute to the presence of nonspecifically associated, oligomerized, or aggregated proteins due to the longer incubation time (ϳ5-7 h) in these experiments. Similar features were observed for NR data collected for GCase alone under comparable conditions. Striking differences are observed between the structural envelopes extracted for d-␣-syn in the complex (Fig. 5C, green) compared with that collected for d-␣-syn alone (Fig. 5C, red). Clearly, interaction of d-␣-syn with the bilayer is perturbed in the presence of GCase (Fig. 5, C and D). This new insight is not evident from any other data and is only revealed through NR measurements. Unlike the protein alone, the structural envelope of ␣-syn in the complex is redistributed, exhibiting a second peak at about 50 Å from the bilayer center, evidence for an interaction of ␣-syn and GCase beyond the membrane surface (Fig. 5C). The observed changes cannot simply be due to crowding at the membrane surface; ␣-syn in the absence of GCase has been measured at comparable membrane coverage, and no changes have been observed resembling those seen for the complex (34). Unfortunately, we are near the limit of detection by NR, so lowering protein concentrations of ␣-syn and GCase below 300 nM is not feasible at this time; therefore, it is uncertain to what extent steric contacts might be perturbing the NR results.

Fit parameters GCase a GCase b GCase and d-␣-syn d-␣-syn
Integration of these profiles shows that ϳ48 residues as compared with ϳ73 residues of d-␣-syn are embedded in the membrane in the presence and absence of GCase, respectively (Fig.  5D). The ␣-syn alone data are fully supported by other structural studies where the N-terminal region up to around residue 95 is membrane-associated (4) and consistent with previous NR work on protiated ␣-syn (34). Recall that the CD spectra of membrane-bound ␣-syn and GCase indicated that ␣-syn maintains or possibly slightly increases its helical content (Fig. 1B). The implication is that ␣-syn retains its membrane-associated helical structure in the presence of membrane-bound GCase, yet the NR data suggest that roughly half of the ␣-helical structure lies above the plane of the bilayer. Such a situation has never before been observed for membrane-bound ␣-syn (4); all previous studies are consistent with the entire helical region being membrane-associated. However, the NR data cannot rule out the possibility that some portion of ␣-syn sitting above the bilayer plane still retains some contact with lipid molecules.

DISCUSSION
The work here shows that GCase binds to and partially inserts into the POPC/POPS membrane, involving structural reorganization of GCase loops at the membrane interface that extend into the hydrocarbon tail region of the membrane. The  Fig. 4B, with Trp residues and glycosylation sites also shown. D, molecular surface for GCase. The (20°, 90°) orientation, the same as in Fig. 4C, is shown in the left panel (side view). In the right panel, the membrane-facing view of GCase (bottom view, rotated 90°from the left panel) is shown. The surface is colored blue for lysine and arginine residues, cyan for histidine, red for aspartate and glutamate, and green for side chains of hydrophobic residues (alanine, valine, leucine, isoleucine, methionine, phenylalanine, and tryptophan). The enzyme active site cavity is indicated as well as the tryptophan residues of loops 2 and 3.
entrance to the active site for the most probable orientations (Fig. 4, C and D) does not directly face the plane of the bilayer. Instead, the entrance lies just above the membrane-water interface, where it presumably remains lipid-accessible and, therefore, catalytically competent. Sap C, the enzymatic cofactor of GCase (13,14), has been proposed to interact with GCase at a site near the active site entrance (55), in a cleft that lies to the left of the entrance, as seen in Fig. 4C. This proposed interaction site is solvent-accessible and lies adjacent to the membrane. The positioning of this site helps to explain how Sap C can bind GCase both in the presence and absence of membrane (22) and suggests that its enhancement of GCase activity probably involves direct interaction with both GCase and the membrane, as proposed previously (56).
The addition of GCase has a profound effect on the NR profile for ␣-syn, causing most of its embedded helical region to move above the plane of the membrane. To visualize how the membrane-associated ␣-syn⅐GCase complex might look, a model structure was generated based on micelle-bound ␣-syn (51). In this structure, the N-terminal region is composed of two ␣-helices consisting of residues 3-37 and 45-92. The first helical region, known to be most important for membrane binding (34,40,(57)(58)(59), is embedded in the membrane in the model, whereas the second helix is positioned above the membrane and in contact with GCase (Fig. 6A, Exposed model). The remaining C-terminal tail region was modeled based on the Spline model fits to the data are shown as solid lines. Best fit neutron scattering length density profiles for all four measurements calculated from the composition-space model are shown in the inset. B, simplified molecular distributions for each organic interface layer of the POPC/POPS stBLM and the median protein envelope for protiated GCase obtained from the best fit of reflectivity data to the spline fit model. Dashed lines, 68% confidence intervals. Volume occupancy is indicated on the right axis. See Table 1 for parameters. C, median protein envelope attributable to d-␣-syn (green) is shown. For comparison, the median protein envelope obtained from fits to NR of d-␣-syn in the absence of GCase (300 nM; red) on POPC/POPS stBLM is also shown. Dashed lines represent 68% confidence intervals. Volume occupancy is indicated on the right axis. See Table 1 for parameters. D, the integrated profiles for d-␣-syn alone (red) and d-␣-syn extracted from ␣-syn⅐GCase complex (green) are shown. The integral is normalized to the total number of residues (140 residues). Dashed lines represent 68% confidence intervals. GCase is shown in gray and ␣-syn in blue and red for the exposed (A) and embedded model (B), respectively. The ␣-syn N and C termini are labeled. The viewpoint corresponds to a rotation of 60°about the membrane normal of Fig.  4C. C, comparison between the predicted cross-sectional profiles for ␣-syn in the two ␣-syn⅐GCase models (A and B) and the spline fit extracted from the ␣-syn⅐GCase NR data (reproduced from Fig. 5C). Dashed lines represent 68% confidence intervals. Volume occupancy is indicated on the right axis.
solution ␣-syn⅐GCase complex (18). The cross-sectional profile calculated based on this model shows good agreement with that generated from the NR experiment, in particular, reproducing two of the three peaks, positioned at 13 and 50 Å from the bilayer center (Fig. 6C). This model is in accord with prior Förster resonance energy transfer fluorescence data, where it was shown that residues 57, 100, and 136 directly interact with GCase, whereas residues 7 and 40 are in close proximity to GCase and in direct contact with the bilayer (19).
Of course, our earlier fluorescence data (19) and the current NR results do not reveal exactly where on the GCase surface the ␣-syn helices make contact; such details must await future studies.
For comparison, the profile for a model structure with all helical residues fully embedded (Fig. 6B, Embedded model) is also shown (Fig. 6C). The embedded model profile has a higher peak at 14 Å, similar to membrane-bound ␣-syn alone (Fig. 5C), and a much smaller one at 48 Å. Neither model produced the third peak in the profile at roughly 90 Å. This peak probably arises from ␣-syn interacting with a small population of GCase multimer or aggregate at the membrane, the presence of which is implicated in the NR profile for GCase (Fig. 5B). Interestingly, the ␣-syn NR profile peak magnitudes lie between those of the exposed and embedded model profiles, suggesting that the complex structure might be an intermediate between the two models or that they are in dynamic equilibrium. This movement probably exposes amyloidogenic core residues of ␣-syn (residues 30 -100), increasing their accessibility to lysosomal proteases, and thus could play some role in PD etiology. Because the dissociation constant for complex formation is in the micromolar range (19), we expect that there is also some contribution from free membrane-bound ␣-syn at the concentrations used in this study (0.3 M). Nevertheless, the data corroborate our previous suggestion that on the membrane surface, the ␣-syn-GCase interaction involves a larger ␣-syn region compared with that found in solution (19).
Although the addition of ␣-syn has no significant effect on GCase secondary structure or its orientation relative to the membrane, it does shift GCase away a small distance from the membrane. Previously, it was suggested that inhibition by ␣-syn might be due to hindered substrate entry into the active site as well as release of product. The results here suggest an additional possibility, that displacement of GCase makes the active site too far from the bilayer for optimal substrate access. However, other explanations cannot be excluded, such as subtle alterations in the active site configuration as well as orientation of nearby loops.
In conclusion, using neutron reflectometry, we have structurally characterized a challenging system, namely the formation of a protein complex on the membrane, revealing features not obtainable by other techniques. Future work on how lysosomal degradation of ␣-syn is affected by GCase interaction and how Sap C reverses GCase inhibition by ␣-syn should provide additional insight into the mechanistic link between ␣-syn and GCase in PD.