obtained for the simulations of the E2 conformation of SERCA in a pure POPC lipid bilayer (blue) and in a

Transcription

1 Supplementary Figure S1. Distribution of atoms along the bilayer normal. Normalized density profiles obtained for the simulations of the E2 conformation of SERCA in a pure POPC lipid bilayer (blue) and in a mixture of POPC lipids and C 12 E 8 detergents (red). C=O is the distribution of POPC glycerol linkage carbonyl oxygen atoms and O is the distribution of the ether oxygen atoms in the detergent. 1

2 Supplementary Figure S2. SERCA crystal structures with black lines indicating the rough estimates of hydrophobic thicknesses. The protein structures are shown in grey ribbons. Tryptophan residues used to estimate the thickness are shown in purple space filling, tryptophan residues not used (272 and 794) are shown in pink space filling, and basic residues used for the estimate are shown in cyan space filling. 2

3 Supplementary Figure S3. 2D structure of the lipids and C 12 E 8 detergent used for membrane setups in molecular dynamics simulations. 3

4 Supplementary Figure S4. Hydrophobic matching for SERCA in the E2 and E2P conformations. Average structures of SERCA are shown in grey ribbons. Tryptophan residues are shown in blue spheres, basic residues in the membrane region are shown in cyan spheres. Average hydrophobic/hydrophilic interfaces, shown as surfaces, are calculated from MD simulations of SERCA positioned in the different membrane environments. The membrane environments are: (di-c14:1)pc lipids, POPC lipids {C16:0, C18:1}, and SGPC lipids {C18:0,C20:1}, with hydrophobic thicknesses of 23 Å, 29 Å, and 32 Å, respectively. The membrane interface is defined by the carbonyl oxygen atoms in the lipid glycerol linkage and colored based on deviation from the average position of each leaflet. 4

5 Supplementary Figure S5. Side chain adaption. Detail showing the side chain rotation of Lys 972 (in the E2 conformation) to adapt to the thin (di-c14:1)pc bilayer (red) and the thick SGPC bilayer (blue). Average structures of SERCA are shown in grey ribbons and the Lys side chains in licorice. 5

6 (a) (b) Supplementary Figure S6. Average helix tilts including the standard deviations. (a) E2 conformation, (b) E2P conformation. The helices are for the tilt analysis defined as follows: TM1 (res ), TM2 (res ), TM3 (res ), TM4 (res ), TM5 (res ), TM6 (res ), TM7 (res ), TM8 (res ), TM9 (res ), TM10 (res ). 6

7 Supplementary Figure S7. Low resolution (Fobs-Fcalc) difference electron density map (green) for the SERCA E2 structure (pdb-id 2YFY) contoured at 2.0 σ. The importance of low resolution diffraction in bilayer imaging is evident. The protein is shown in blue cartoon representation. Symmetry related molecules are shown in cylindrical representation (grey). Resolution ranges for the calculated maps are given above the structures. 7

8 Supplementary Figure S8. Nomenclature of atoms and atom types (CHARMM) for the aliphatic part of a POPC lipid. 8

9 Supplementary Figure S9. Nomenclature of atom types (CHARMM) for the detergent. 9

10 Supplementary Figure S10. Transmembrane part of SERCA. SERCA (in the E2P conformation) shown in grey ribbons with the transmembrane selection used for RMSD measurements and alignment of protein structures shown in purple. Average membrane surface, as defined by the lipid phosphorus atoms in a POPC bilayer, is shown in grey. 10

11 (a) (b) Supplementary Figure S11. RMSD of the transmembrane part of the protein structure for the studied simulations. (a) E2 conformation, (b) E2P conformation. 11

12 Supplementary Table S1. Comparison of structures for the E2 conformation. RMSD in Å for the TM part of the average structures compared to each other and to the X-ray structure E2 POPC+C 12 E 8 (di-c14:1)pc POPC SGPC X-ray POPC+C 12 E (di-c14:1)pc POPC

13 Supplementary Table S2. Comparison of structures for the E2P conformation. RMSD in Å for the TM part of the average structures compared to each other and to the X-ray structure E2P POPC+C 12 E 8 (di-c14:1)pc POPC SGPC X-ray POPC+C 12 E (di-c14:1)pc POPC

14 Supplementary Table S3. Bond parameters introduced for the detergent C 12 E 8. Bond Force constant (kcal/(mol Å 2 )) Equilibrium length (Å) CC32A-OH CC32A-CTL

15 Supplementary Table S4. Angle parameters introduced for the detergent C 12 E 8. Angle Force constant (kcal/(mol rad 2 )) Equilibrium angle ( ) UB force constant (kcal/(mol Å 2 )) UB equilibrium length (Å) H-OH1-CC32A OH1-CC32A-CC32A OH1-CC32A-HCA OC30A-CC32A-CTL HCA2-CC32A-CTL CC32A-CTL2-CTL CC32A-CTL2-HAL

16 Supplementary Table S5. Dihedral parameters introduced for the detergent C 12 E 8. Dihedral angle Force constant n Delta (kcal/mol) ( ) H-OH1-CC32A-CC32A H-OH1-CC32A-HCA OH1-CC32A-CC32A-OC30A OH1-CC32A-CC32A-HCA CC32A-OC30A-CC32A-CTL OC30A-CC32A-CTL2-CTL OC30A-CC32A-CTL2-HAL CC32A-CTL2-CTL2-CTL CC32A-CTL2-CTL2-HAL CTL2-CTL2-CC32A-HCA HAL2-CTL2-CC32A-HCA

17 Supplementary Table S6. Average protein hydrophobic thickness. The values are based on the MD simulations and the standard deviations are included. (di-c14:1)pc POPC SGPC E ± ± ± 1.8 E2P 20.2 ± ± ±

18 Supplementary Note 1 Hydrophobic Thickness of SERCA The hydrophobic thickness of the studied crystal structures of SERCA has been evaluated very similar to Lee s approach 5. The position of the nitrogen atoms of tryptophan imidazole rings, the nitrogen atom of the lysine residues 262 and 972, and the nitrogen atoms (not Nε, though) of arginine 63 were used as markers that are not allowed inside the hydrophobic part of the bilayer. Trp residues 272 and 794 were omitted from the evaluation, as it was judged unrealistic that they could contribute to the docking to the membrane interface. Using the markers to evaluate the hydrophobic thickness, the protein structure was rotated until the largest hydrophobic thickness allowing these atoms to be at the hydrophobic-hydrophilic limit or outside was found (Supplementary Fig. S2). This resulted in a measured hydrophobic thickness of 23 Å for the E2 conformation and 24 Å for the E2P conformation. Using the same definition of protein hydrophobic thickness on the MD simulations, the average measures given in Supplementary Table S6 are found. There is no simple way to define the hydrophobic thickness of SERCA, and the definition applied above can only be used to express that SERCA has features that do not seem to be well-matched with membranes of the thickness it is located in natively. From the simulations it seems that the role of the tryptophan sidechains as anchors to the membrane interface is not crucial, and that the energy penalty for having a tryptophan in the hydrophobic part of the membrane is not too high. Therefore, it might be more relevant to examine the vertical distance between the charged side chains (residues 262, 972, and 63) only even though they are not found vertically opposite each other. The distances between the charged groups vary between Å depending on the thickness of the membrane. In all cases the membranes show points of minimum thickness which are less than the average distance between the charged groups for that simulation. 18

19 Supplementary Methods Protein-Bilayer Topology Analytical data have shown that crystallization of the protein under the conditions described in the Protein Purification section is accompanied by the formation of an envelope around the transmembrane segment consisting of approximately a 30:80:1 molar ratio of phospholipid, detergent and protein, respectively 18. The hydrocarbon content of this mixed detergent/phospholipid envelope roughly corresponds to the phospholipid content of the native membrane, and we expect the mixture to form a modified bilayer around the membrane protein sector, akin to the large ternary complexes of protein/lipid/detergent arising as an intermediate state during the solubilization of the native membrane with nonionic detergent 16, 17. Use of Low Resolution Data Analysis of electron density maps calculated at various resolution ranges reveals that inclusion of low resolution data improves electron density features for the head group region of the bilayer leaflets. These features are presumably dominated by the distribution of semi-ordered phosphate head groups at the bilayer interfaces. We conclude that lipid-detergent bilayers, forming a continuum of ordered and semiordered molecules between crystallographically packed protein molecules, contribute to the crystal diffraction at low resolution and contrasts the disordered molecules of a bulk solvent. For density maps calculated at a low resolution limit at or above app. 30 Å, the coherence of the observed density peaks in the difference maps was weak as regards to contributions from a bilayer. Electron density maps calculated at or below a low resolution limit of 40 Å did show clear peaks in the putative head group region with low noise in the water-containing bulk solvent region. This holds true for both 2F obs -F calc and F obs -F calc electron density maps. Depicted in Supplementary Fig. S7 is the F obs -F calc electron density map calculated for various resolution ranges. Assuming a crystal composition as determined for comparable SERCA crystals, the lipiddetergent content is approximately one-third of the total with crystallographically packed protein, thus a protein/bulk solvent model can be assumed to provide meaningful, although not perfect, phases even at 19

20 very low resolution where the relative contributions from semi-ordered lipid molecules increases. Indeed the R-factor determined for 226 reflections at a resolution range of Å is 32%. POPC Membrane as Model of Sarcoplasmic Reticulum To mimic the in vivo sarcoplasmic reticulum (SR) membrane, a palmitoyl oleyl phosphatidylcholine (POPC) membrane model was chosen. The SR membrane (from which SERCA is isolated) has a somewhat simpler composition than the plasma membrane, containing mostly phospholipids and only little cholesterol 54. It has been determined that approximately 70 % of the phospholipids in rabbit SR membrane consist of phosphatidylcholine (PC); approximately 23 % of phosphatidylethanolamine (PE); while the remaining 7 % mainly consists of phosphatidylserine and phosphatidylinositol 55. For both PC and PE, more than 80 % are derived from fatty acids with C 16 or C 18 chains of which about 50 % are unsaturated. Therefore, a POPC membrane with a C 16:0 and a C 18:1 tail seems to be a reasonable choice to mimic the native conditions in a simple manner. For this, a POPC membrane patch of 115 Å x 125 Å was generated using the Membrane Builder available in VMD 44. MD Setups with Different Bilayer Compositions Setups with different bilayer compositions (Supplementary Fig. S3) were constructed from the setups with POPC bilayer as described in the following. For the dimyristoleoyl phosphatidylcholine ((di-c14:1)pc) bilayer ({C14:19c, C14:1 9c}) two and four terminal carbon atoms of the palmitoyl and oleyl chains of POPC were removed, respectively. Additionally, a double bond was introduced in the former palmitoyl chain by removing a hydrogen atom from carbon atoms C39 and C310. To force the new double bond of the lipid tail into a cis conformation, an edited force field was constructed favoring dihedrals of the cis double bond (for HEL1 CEL1 CEL1 HEL1 - see Supplementary Fig. S8 - the force constant was changed to 15 kcal/mol and the multiplicity to 1, keeping a phase of 180º). This edited force field was used during the minimization and for the first 200 ps of the lipid equilibration. Thereafter, all double bonds were in the cis conformation and the 20

21 normally applied force field was utilized for the remaining 300 ps of lipid equilibration and for the following equilibration and production run. To model a thicker bilayer, two carbon atoms were added to each of the POPC acyl chains to give a C18:0 and a C20:1 chain, thereby constructing the stearoyl gadoleyl phosphatidylcholine (SGPC) lipid. It was necessary to make room for the extra atoms in the centre of the bilayer, and this was achieved by displacing the two leaflets of the bilayer away from one another by a distance corresponding to the length in the z-direction of the extra ethyl group added. Setups with C 12 E 8 detergent molecules (see chemical structure in Supplementary Fig. S3) were constructed to mimic the environment in the crystal, where measurements suggest a stoichiometry of about 30 lipid molecules and 80 detergent molecules per protein molecule 18. To obtain a lipid:detergent stoichiometry similar to the one in crystals, every other POPC molecule in the system setup was exchanged with two single-tailed detergent molecules. A pdb-file of C 12 E 8 from the HIC-Up 56 database was used as input structure. Two C 12 E 8 molecules were positioned in the bilayer by aligning four carbon atoms in the tail of the detergent with four carbon atoms in the tails of a POPC lipid and afterwards deleting the lipid from the setup. Initial test-simulations of this system showed that the presence of detergent molecules made the bilayer area increase and thereby made the simulation box decrease in the direction along the membrane normal (z-direction). To ensure that the protein was adequately solvated, an extra 20 Å slab of water was added to the simulation box, with the same ionic concentration as the rest of the system. The Validity of a Random Lipid-Detergent Mixture The simulation time of 50 ns is not enough to allow for significant reorganization of the detergent and lipid molecules in the plane of the bilayer. The initial setup is therefore build as a random mixture of detergents and lipids, also at the interface with SERCA, as we know from fluorescence quenching studies on 21

22 brominated lipid 57 and brominated detergent 58 that both components are present at the protein interface in phospholipid/detergent mixtures. Parameters Applied for the Detergent The parameters applied in the simulations of the detergent molecule, C 12 E 8, are based on the CHARMM27 lipid parameters and the CHARMM ether parameters 51, 52 and outlined in the following. Atom types and partial charges for the ether fragment were assigned from the PEG monomer (-CH 2 -O-CH 2 -) from the CHARMM ether force field. For the aliphatic tail, atom types and partial charges were assigned as for lipid tails. For the terminal hydroxymethyl group, atom types and charges for serine were applied. The shift between the ether fragment and the aliphatic chain was performed as for methyl propyl ether in CHARMM The majority of the necessary parameters were present in either CHARMM27 or the revised ether force field in CHARMM However, parameters for the linkage regions between the aliphatic chain and the ether part, and between the ether part and the hydroxymethyl group, needed to be added to the parameter files. The added parameters are based on comparisons between the corresponding parameters in CHARMM27 or CHARMM35 if available, e.g. the CC32A-CTL2 (see Supplementary Fig. S9 for atom types) bond parameters are based on CTL2-CTL2 parameters from CHARMM27 and CC32A-CC32A from CHARMM35 (which are identical). Parameters involving ether oxygen atoms were adapted from CHARMM35 (the atom type is not present in CHARMM27), e.g. OC30A-CC32A-CC32A-CTL2 dihedral parameters. Parameters for two dihedral angles for the terminal hydroxyl part were adapted from the revised CHARMM35 and in the shift between ether and aliphatic region, CHARMM27 parameters were used. Parameters not present in either CHARMM27 or CHARMM35 are listed in Supplementary Tables 3-5. The Simulation Protocol To avoid potential problems arising from CHARMM27 lipid parameters producing a membrane which grows in thickness over time 59, we apply the NP z AT protocol. Still, an equilibration of 2 ns of the bilayer area and 22

23 thickness (in the NPT ensemble) was carried out before locking the area. To be confident that the 2 ns equilibration was suitable to reach a thicknesses for the bulk membranes corresponding to an experimental situation, we compared well defined experimental thickness values for both POPC and (di-c14:0)pc (DMPC) 60 to those observed in our simulated membranes (we did simulations with SERCA in (di-c14:0)pc, which are not included in the paper). As (di-c14:0)pc represents a considerable change relative to the initial POPC bilayer, it is to be expected that the (di-c14:1)pc, POPC+detergent, and SGPC bilayers can equilibrate the thickness within the same time frame. Analysis of Molecular Dynamics Simulations In all analysis of the MD simulations, 1750 snapshots evenly distributed over the last 35 ns of the 50 ns MD simulations were used, unless otherwise specified. Protein Structure Stability To evaluate the change in the transmembrane (TM) part (see Supplementary Fig. S10) of the protein compared to the original crystal structure, the Cα RMSD for the TM part of the simulated structure was calculated relative to the crystal structure (Supplementary Fig. S11). Since the major focus here is proteinlipid interactions, the cytoplasmic and luminal loops and the cytoplasmic domains were omitted from the RMSD calculations. Including these structural parts would result in a larger RMSD, since the watercontacting elements are expected to be more mobile, and thereby this RMSD would not reflect the changes caused by the different lipid environments. Transmembrane Helix Tilt We define the tilt of a TM helix as the angle between the membrane normal (the z-axis) and the vector connecting the center-of-mass of the first and last four Cαs of the helix. To compare helix tilts observed for different bilayers and for the two conformations, the average helix tilt as well as the standard deviation for the tilt was evaluated for each TM helix over the last 35 ns of the simulations (Supplementary Fig. S6). The 23

24 time scale covered by the MD simulations (50 ns) is not enough to show major rearrangements of the transmembrane helices, e.g. corresponding to a conformational shift. However, we only wish to study how the conformations of interest (E2 or E2P) adapt to the environment, and the helix tilts are seen to easily change around within the 50 ns, and thus the simulation time is expected to be sufficient to study the adjustment of the protein structure to the membrane environment. Average Structure Since no large-scale conformational changes are observed within the studied timeframe (15-50 ns), it is possible to evaluate an average structure based on the simulation, which will represent the structure and thereby the structural changes imposed by the different bilayer environments. The structure for each analysed frame was aligned to the one at 15 ns using only rotation around the z-axis and translation to avoid an unrealistic tilt of the bilayer. Based on this, the average coordinates for all non-hydrogen atoms were computed. The average structures could then be compared to each other and to the original X-ray structures as seen in Supplementary Table S1 and Supplementary Table S2. The same selection of Cαs as used in Supplementary Fig. S11 for RMSD measurements was utilized. Model Based Structure Factors To derive the model based structure factors for the calculation of the difference electron density map shown in Fig. 1c, 10 snapshots evenly distributed over the last 35 ns of the 50 ns MD simulation of SERCA (E2 conformation) in the bilayer setup with a mixture of POPC lipids and detergent (C 12 E 8 ) were used. The coordinates of the POPC and detergent molecules from all 10 snapshots were combined with the protein coordinates from the first snapshot into one coordinate file. Atoms in the protein chain were given an occupancy of 100 % and a crystallographic B-factor of 80, and the lipid/detergent atoms where given an occupancy of 10 % and a relatively higher crystallographic B-factor of 140. This setup thereby represents a model with positions of lipids and detergents averaged over 10 unit cells, verified by inspection as to not be in overlapping positions. The derived model is thus thought to be a good representation of the bulk bilayer 24

25 of lipids and detergents present in crystals of SERCA, as used for the recording of experimental X-ray diffraction. On this basis structure factors were calculated representing the F calc+lipids. Structure factors of the protein in the same conformation, but excluding the lipids and detergents, were used to calculate the model based F calc-lipids. Structure factors and phases of both representations were calculated at 60-3 Å, and the Fourier synthesis used for calculating the maps used in the analysis were performed at 60-8 Å. Average Membrane To obtain a representation of the average bilayer surfaces, as defined for a certain set of lipid atoms, a trigonal x,y grid with a resolution of 5 Å was laid out for each leaflet. For each analysed frame of the trajectories, the z-coordinates of the selected atoms were registered at the grid point closest to their x,y position. Finally, the average z-coordinate for each grid point could be evaluated. Grid points which were hit in less than 5 % of the timeframes were considered insignificant. Points that fell below the significance limit were represented by a linear interpolation between the closest significant points to fill out the holes on the surface. Before collecting the (x,y) and z coordinates of the lipid selection for the current timeframe, the system was aligned such that the TM part of the structure (same as used for RMSD measures) was aligned with the average structure previously found; as before without tilting the bilayer. Where horizontal lines are used to show the average membrane position, they are positioned at the simple average of z-positions of the selected lipid atoms obtained for the trajectory, after aligning each frame as for computing the average membrane surface. When a single number is given for the estimated hydrophobic thickness of a bilayer, it is evaluated as the z distance between the horizontal lines described above, where the carbonyl atoms of the lipids are used to define the hydrophobic-hydrophilic interfaces of the bilayer. 25

26 Membrane-water Interface The nature of the membrane-water interface can be illustrated by density profiles along the membrane normal for the carbonyl groups of the lipid glycerol linkage. As seen from Fig. S1, the distribution of the carbonyl groups are broader for the detergent-containing bilayer, and it is also seen that the bilayer is effectively thinner than for the pure POPC lipid bilayer. This, together with the distribution of oxygen atoms of the detergent polyether head-group (C 12 E 8 (O)), which shows very broad peaks and a non-zero density in the middle of the bilayer, makes it clear how the hydrophobic-hydrophilic interface is very diffuse in the detergent-containing bilayer. The data are based on every second timeframe of the standard set of timeframes used for analysis. 26

27 Supplementary References 54. van Meer, G., Voelker, D. R. & Feigenson, G. W. Membrane lipids: Where they are and how they behave. Nature Rev. Mol. Cell. Biol. 9, (2008). 55. Gould, G. W., McWhirter, J. M., East, J. M. & Lee, A. G. Effects of diet on the function of sarcoplasmic reticulum. Biochem. J. 245, (1987). 56. Kleywegt, G. J. & Jones, T. A. Databases in protein crystallography. Acta Cryst. D54, (1998). 57. de Foresta, B. et al. Membrane solubilization by detergent - use of brominated phospholipids to evaluate the detergent-induced changes in Ca 2+ -ATPase lipid interaction. Biochemistry 28, (1989). 58. de Foresta, B., Legros, N., Plusquellec, D., le Maire, M. & Champeil, P. Brominated detergents as tools to study protein-detergent interactions. Eur. J. Biochem. 241, (1996). 59. Klauda, J. B. et al. Update of the CHARMM all-atom additive force field for lipids: Validation on six lipid types. J. Phys. Chem. B 114, (2010). 60. Kucerka, N., Tristram-Nagle, S. & Nagle, J. F. Structure of fully hydrated fluid phase lipid bilayers with monounsaturated chains. J. Membr. Biol. 208, (2005). 27