Data Files Accompanying:"Impacts of Lipase Enzyme on the Surface Properties of Marine Aerosols".
Authors:J, M, Schiffer.| M, Luo.| A, C, Dommer.| G, Thoron.| M, Pendergraft.| M, V, Santander.| D, Lucero.| E, Pecora de Barros.| K, A, Prather.| V, H, Grassian.| R, E, Amaro.
Journal:J. Phys. Chem. Lett. 9, 14, 3839-3849
Contact: Rommie, E, Amaro., ramaro@ucsd.edu, Department of Chemistry and Biochemistry, University of California, San Diego, La Jolla, CA 92093
Cite as:J, M, Schiffer.| M, Luo.| A, C, Dommer.| G, Thoron.| M, Pendergraft.| M, V, Santander.| D, Lucero.| E, Pecora de Barros.| K, A, Prather.| V, H, Grassian.| R, E, Amaro. (2018): Data from: "Impacts of Lipase Enzyme on the Surface Properties of Marine Aerosols". In Center for Aerosol Impacts on Chemistry of the Environment (CAICE) Collection. UC San Diego Library Digital Collections.

Method:Langmuir Surface Adsorption Isotherms. Palmitic acid (>99 %) was purchased from Sigma Aldrich. Dipalmitoyl phosphatidic acid (DPPA) (>99 %) was purchased from Avanti Polar Lipids. Chloroform (>99.9%) was purchased from Fisher Scientific. Palmitic acid was prepared in Chloroform with a concentration of 1 mg/mL and DPPA was prepared in 4:1 chloroform : methanol solution with a concentration of 0.42 mg/mL.60 These solutions were used to introduce the lipid to the interface. Milli-Q water with electric resistance of 18.2 MΩ was used for the aqueous subphase. NaCl salt were purchased from Fisher Scientific and were purified by baking at 200°C overnight to remove organic contaminants. We used one subphase for our experiments, NaCl solution (0.4 M), which was chosen to be near that of sea water concentrations. The pH of the subphase was controlled at 2 or 6. These pH values lie in the range of aerosol acidity.38,39 
A computer controlled film balance with a Langmuir trough (KSV NIMA LB, S/N AAA100505) was used for these experiments. The temperature of the solution in the trough was controlled by a water circulator (Beckman Geneline Cooler). For the experiments done in the Langmuir trough (Figure 1), the lipid solution was first spread onto the surface of 55 mL subphase in the trough to form a monolayer. Then the monolayer was compressed to 5 mN/m and was held at this surface pressure for 408 minutes before lipase injection as describe previously.61 The BCL solution was made in 0.4 M NaCl solution with a concentration of 4.55 mg/mL. We injected 0.2 mL lipase solution underneath the lipid monolayer and the final concentration of lipase should be 500 nM. The area of the monolayer was monitored throughout the experiment with the computer controlled film balance. We also perform experiments using the computer controlled film balance with a petri dish. For the experiments done in a petri dish (Figure S1), the lipid monolayer was first formed on the surface of 10 mL subphase in the petri dish. Then after 30 minutes, BCL solution was injected underneath the lipid monolayer using a microsyringe which go through the monolayer surface. The final concentration of BCL is 1 mg/mL. The surface pressure was monitored with the computer controlled film balance throughout the experiment.

Method:Computational. MD Simulation System Building. We used Schrödinger PrepWizard to prepare BCL structures for molecular dynamic simulation.  The Burkholderia Cepacia triacylglycerol lipase (BCL) crystal structure (PDB: 3LIP) was prepared for simulations, maintaining the four waters from the crystal structure around the calcium in the PDB structure. The full BCL-bilayer system was built with CHARMM-GUI, and the orientation of BCL in the membrane was determined using the orientation of protein in membrane (OPM) server. The system was protonated at approximately pH 6, with the pKa titratable residues determined with the Maestro integrated PROPKA,  as the pH within marine aerosols is known to decrease significantly from that of seawater (pH 8.4). Control simulations of bilayer-only systems were also built in the same manner as described above at pH 6. 
All residues were maintained at the protonation states determined with PROPKA for aqueous environments except for three: 1) the active site histidine (H286), since this residue is buried at the interface and thus is likely to maintain a higher pKa than 7.20 in the presence of phospholipids/fatty acids, and 2) the cysteines C180 and C270, for which the disulfide bond was broken. The rationale behind breaking the disulfide bond was based on the location of the disulfide bond proximal to the membrane surface, which in each system contains protonated carboxyl groups and protonated phosphate groups capable of becoming deprotonated upon changes in local environments. Thus the disulfide bond was broken to sample a greater portion of conformational space and simulate more realistic protein motions given the chemical environment of a lipid interface. The distance between cysteine sulfurs and structural calcium distances were tracked through the simulation to verify the role of local unfolding in BCL dynamics and these findings are discussed briefly in the results (Figure S12, S14, Table S17). Convergence of protein dynamics was also verified through RMSD analysis (Figure S15). 
The full systems were then built using CHARMMGUI and the orientations of proteins in membranes (OPM) data base to position BCL in lipid bilayers. OPM placed the orientation of BCL such that residues 24, 26, 122, 126-127 129-143, 145, 146, 149, 150, 218, 220-221, 223, 232-247, 249-250, 253 293, 294, 296 were embedded in the bilayer. Four lipid bilayers were constructed, and these lipid were chosen because they are known to be highly concentrated in sea spray aerosols. Four systems were built at pH 6: Two systems were built within lipid bilayers  of DPPA at 45 Å2 per lipid and 60 Å2 per lipid, and two more within lipid bilayers of palmitic acid at 24 Å2 per lipid and 30 Å2 per lipid. Another two systems were built at pH <4: lipid bilayers of palmitic acid at 24 Å2 per lipid and 30 Å2 per lipid to elucidate the role of surface charge and pH on BCL structure. We solvated each system in a TIP3P (33) cubic water buffer of 70 Å, and x and y axis values of 70 Å as well. 5 Na+ ions were added to balance the charge of BCL, the ion concentration was set at 0.4 M NaCl to mimic the lower range of the salt concentration in a sea spray aerosol particle. The total number of atoms in each system at pH 6 is 72991 for DPPA at 45 Å2 per lipid, 68924 for DPPA at 60 Å2 per lipid, 71363 atoms for palmitic acid at 30 Å2 per lipid, and 73210 for palmitic acid at 24 Å2 per lipid. The total number of atoms in each system at pH <4 is 66070 for palmitic acid at 24 Å2 per lipid and 62554 for palmitic acid at 30 Å2 per lipid. The only difference between the pH <4 and pH 6 simulations are the protonation states of the palmitic acid head groups and the number of sodium ions present in solution. All other protein side chains except for H286 have pKa’s determined by PROPKA above 8.4 or below 4 and so the BCL protonation states are identical in both simulations. 
MD system parameterization. System coordinates from preparation (above) were parameterized in CHARMMGUI using the charmm36 force field parameters for the proteins and lipids, and periodic boundary conditions were implemented. The calcium ion bound to the BCL was parameterized as has been previously discussed. This force field has been previously tested for membrane protein environment. Calcium remained bound throughout all of the simulations as seen in Figure S12. 
GPU-enabled MD simulations. Simulations were performed with NVIDIA GK110 (GeForce GTX Titan) GPUs using the CUDA version of PMEMD in AMBER16. The simulation systems above were minimized with the CPU version of AMBER16 and equilibrated using the GPU version of AMBER16. We minimized in one step with 2500 steps of steepest-descent algorithm and the remaining 2500 steps with conjugate-gradient algorithm. Positional restraints and dihedral restraints were used during minimization to keep the the lipids in all trans configurations and to restrain the motion of protein residues and lipid head groups. 
Following minimization, five steps of equilibration were performed, including a stage of heating: 1) 25 ps of hydrogen only minimization with a restraint weight of 250 kcal/mol on the protein and solvent, 2) 4 ps of water minimization with a restraint weight of 20 kcal/mol on the protein and salt atoms, 3) 20 ps of water heating with a restraint weight of 20 kcal/mol on the protein and salt atoms, and 4) 40 ps of full equilibration. We equilibrated the system using harmonic equilibration at 298.15 K over four sequential 500-ps runs, decreasing the restraint potential on the backbone on each step, starting at 4.0 kcal/mol and ending at 1.0 kcal/mol. 	GPU-enabled AMBER16 production runs were carried out as an NPT ensemble at 298.15  K and 1 bar with a 2-fs time step and particle mesh Ewald (PME) electrostatic approximation and non-bonding cutoff of 12.0Å. These conditions were used to mimic the conditions of (1) the mesocosm laboratory experiments of phytoplankton blooms in a waveflume, and (2) smaller phytoplankton growth and sea spray aerosol forming experiments performed in marine aerosol reference tanks (MARTs) to mimic the open ocean. These experiments are where the different lipids were identified in the aerosol phase and where lipase activity was measured. MD input files are provided as part of the data sharing files. For each of the six system simulated, five copies of the simulation were performed for 110 ns each. 
Simulation analysis. MD trajectories were processed within an ipython jupyter notebook environment using PYTRAJ, MDTRAJ and numpy, as well as with cpptraj for PCA analysis and VMD for visualization. All frames were aligned using backbone atoms. Pytraj was used to perform dihedral, distance, radial distribution function, tilt angle, RMSF, membrane curvature and hydrogen bond based measurements. Mdtraj was also used to measure membrane curvature. Curvature was calculated by fitting a quadratic surface to the C16- (in the case of PA) or C216- (in the case of DPPA) carbon coordinates using the least-squares method available in the NumPy computing package. Once the surface was fitted, the coordinates of the local extremum were calculated and the curves along the surface in each principle direction (x and y) from the critical point were determined. The radius of curvature in each direction was calculated by finding the osculating circle at the vertex of each curve. All images were taken with visual molecular dynamics (VMD). 
Principal Component Analysis (PCA). The trajectory was stripped of solvent and aligned against backbone residues. PCA was performed for all atoms for each frame of the trajectory using CPPTRAJ. Three principal components were specified for calculation. 
Pairwise Internal Distance Clustering. MD trajectories were processed within an ipython jupyter notebook environment using PyEmma. All internal distances between all nitrogen atoms within the BCL backbone were calculated. Internal distances that were measured to be more than 7 Å or less than 3 Å during the entirety of the simulations were removed. From the remaining non-redundant distances, all internal distances that did not vary by more than 0.5 Å were also removed. This resulted in a total of 121 non-redundant internal distances for the more concentrated DPPA simulation, 84 non-redundant internal distances for the less concentrated DPPA simulation, 83 non-redundant internal distances for the more concentrated palmitic acid simulation at pH 2, 69 non-redundant internal distances for the more concentrated palmitic acid simulation at pH 6, 71 non-redundant internal distances for the less concentrated palmitic acid simulation at pH 2, and 65 non-redundant internal distances for the less concentrated palmitic acid simulation at pH 6. Regular space clustering was performed on these non-redundant internal distances resulting in the number of clusters seen in Table 1. 
SWISS-MODEL: A homology model of the closed structure of BCL was built using the FASTA sequence from PDB 3LIP and building into the closed structure of lipase from Psuedomonas glumae, PDB 1TAH . The results from SWISS-MODEL are found in Figure S3. The QMEAN score was -0.51, Cβ was -1.53, the All Atom was -1.21, Solvation was -0.28, and Torsion -0.16, with a GMQE of 0.96. 
APBS Electrostatics The PDB2PQR automated pipeline was used to investigate the electrostatics of the BCL surface and active site pocket. A pH of 6.0 was chosen to align with the pH from the simulation and SSA. All proteins structures that underwent APBS electrostatics were prepared with maestro’s prep wizard, had side chain protonation states optimized with PROPKA, and using the PARSE force field for the calculations.