FULL PAPER Zhang et al. binding mode of SM was then optimized within the 1.0 MPa. After these preparative steps, a production of structural restraints of hSMSl(Figure 4) Minimization 10 ns long was conducted under the same temperature was conducted using the adopted basis New. and pressure. During this process, the temperature was ton-Raphson algorithm with the CHARMm force field controlled with the Hoover method whereas the pres- implemented in the discovery studio software sure was controlled by coupling to a pressure bath using extended system algorithm. The mass of the pressure piston was 1000 amu. Langevin piston collision fre et to 25.0 ps The dielectric constants of 332 protein and water molecules were set to 1.0 and 80.0 respectively. Distance cutoff in generating the list of pairs was 14. 0 A Switching function was used between 10. A and 12.0 A to treat non-bonding interactions RMSD values relative to the initial structure were moni- tored as an indication of equilibrium along the MD trajectory. The last snapshot on the resulting md tra- jectory was retrieved, minimized without restraint in couple with the implicit membrane model. This final 8 efined model of the hSMSI/SM complex is shown in Figure 5. In order to evaluate this structural model, the PROCHECK program(version 3. 5.4 was applied to check its stereochemical quality. The Ramachandran plot of this hSMS I structure produced by PROCHECK Figure 4 The binding mode of sphingomyelin to hSMS1 Three lipid complex structure nesis study on the hSMSl/ Computational muta residues, i.e. H285, H328 and D332, in the catalytic site are criti- cal for the catalytic mechanism of hSMSI We then performed computational mutagenesis on the hSMSl complex structure to evaluate the impor In order to refine the structural model of the tance of both the functional groups on the lipid substrate hSMSI/SM complex, it was further subjected to a and several amino acid residues on hSMSISuch results long-time molecular dynamics (MD)simulation will provide additional proofs on the catalytic mecha- which an implicit GB/SA model of bilayer membrane nism of hSMS1. Two types of mutations were pe was applied. Major parameters for setting the implicit formed accordingly: on the substrate side, SM was mu membrane GB/SA model include: Grid spacing for tated in turn into PC, PE, PA, PS, and PG(Figure 1) lookup table (DGP)=1.5 A: Half membrane switching whereas on the hSMsl side, several important amino length(MSW)=2.0 A; Half smoothing length(SW) 3 A; Non-polar surface tension coefficient turn into alanine. Both types of mutations were done (SGAMMA)=0. 418 kJ/(mol-A2). Number of angular using the Discovery Studio software by deleting or ntegration points 50 was used for volume integration adding some atoms while keeping original atoms as GB/SA calculation. To keep the catalytic site stable for much as possible. After mutation, two rounds of mini- SN2 nucleophilic substitution reaction, three pairs of mization were applied to the hSMSIAipid complex harmonic distance restraint were applied with harmonic structure with the same force field parameters described force constant 418(kJ/molA): atom pairs 1-2, 4-5. above. Finally, binding energies between the substrate and 6-7(Figure 4). Firstly, hSMS 1/SM was positioned molecule and hSMS I were calculated using in the center ot the implicit membrahe model. The pla- hSMSi/ipid complex (Table 3 and Table 4) Iting tered at ==0 with a membrane thickness of 30 A. To- Normal mode analysis and protein motion analy pology and coordinate files were generated with the of hSMS Discovery Studio software. Minimization was per formed using the program CHARMm(release c33b2) We conducted protein domain motion analysis for Atotalof5000stepsofsteepestdescentminimizationtheDyndomon-lineserver(http://fizz.cmp.uea.ac.uk/ and 5000 steps of conjugate gradient minimization were dyndomn 4 The initial and the last configuration of the performed subsequently As the first step of MD simulation, the entire system hSMS I/SM complex in MD simulation were submitted was heated froma to in100 ps. Then, the as inputs. The Dyn Dom server returned the information whole system was equilibrated for 400 ps under a con- of the " hinge axes"(Figure 7). In order to confirm the stant temperature of 310 K and a constant pressure of mes of protein domain motion ana 1570www.cjc.wiley-vch.deO2011SioC,Cas,Shanghai&WileY-VchVerlagGmbh&Co.Kgaa,WeinheimChin.j.chem.2011,29,1567-1575FULL PAPER Zhang et al. 1570 www.cjc.wiley-vch.de © 2011 SIOC, CAS, Shanghai, & WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chin. J. Chem. 2011, 29, 1567—1575 binding mode of SM was then optimized within the structural restraints of hSMS1 (Figure 4). Minimization was conducted using the adopted basis New￾ton-Raphson algorithm40 with the CHARMm force field implemented in the Discovery Studio software. P O O O O N N H H N N H N O O H285 H328 D332 1 2 3 4 5 6 7 8 Figure 4 The binding mode of sphingomyelin to hSMS1. Three residues, i.e. H285, H328 and D332, in the catalytic site are criti￾cal for the catalytic mechanism of hSMS1. In order to refine the structural model of the hSMS1/SM complex, it was further subjected to a long-time molecular dynamics (MD) simulation in which an implicit GB/SA model of bilayer membrane41 was applied. Major parameters for setting the implicit membrane GB/SA model include: Grid spacing for lookup table (DGP)＝1.5 Å; Half membrane switching length (MSW)＝2.0 Å; Half smoothing length (SW)＝ 0.3 Å; Non-polar surface tension coefficient (SGAMMA)＝0.418 kJ/(mol•Å2 ). Number of angular integration points 50 was used for volume integration in GB/SA calculation. To keep the catalytic site stable for SN2 nucleophilic substitution reaction, three pairs of harmonic distance restraint were applied with harmonic force constant 418 (kJ/mol•Å2 ): atom pairs 1—2, 4—5, and 6—7 (Figure 4). Firstly, hSMS1/SM was positioned in the center of the implicit membrane model. The pla￾nar membrane is perpendicular to the z axis and cen￾tered at z＝0 with a membrane thickness of 30 Å. To￾pology and coordinate files were generated with the Discovery Studio software. Minimization was per￾formed using the program CHARMm (release c33b2).42 A total of 5000 steps of steepest descent minimization and 5000 steps of conjugate gradient minimization were performed subsequently. As the first step of MD simulation, the entire system was heated from 0 K to 310 K in 100 ps. Then, the whole system was equilibrated for 400 ps under a con￾stant temperature of 310 K and a constant pressure of 1.0 MPa. After these preparative steps, a production of 10 ns long was conducted under the same temperature and pressure. During this process, the temperature was controlled with the Hoover method43 whereas the pres￾sure was controlled by coupling to a pressure bath using extended system algorithm.44 The mass of the pressure piston was 1000 amu. Langevin piston collision fre￾quency was set to 25.0 ps－1 . The dielectric constants of protein and water molecules were set to 1.0 and 80.0, respectively. Distance cutoff in generating the list of pairs was 14.0 Å. Switching function was used between 10.0 Å and 12.0 Å to treat non-bonding interactions. RMSD values relative to the initial structure were moni￾tored as an indication of equilibrium along the MD trajectory. The last snapshot on the resulting MD tra￾jectory was retrieved, minimized without restraint in couple with the implicit membrane model. This final refined model of the hSMS1/SM complex is shown in Figure 5. In order to evaluate this structural model, the PROCHECK program (version 3.5.4)45 was applied to check its stereochemical quality. The Ramachandran plot of this hSMS1 structure produced by PROCHECK is shown in Figure 6. Computational mutagenesis study on the hSMS1/ lipid complex structure We then performed computational mutagenesis on the hSMS1 complex structure to evaluate the impor￾tance of both the functional groups on the lipid substrate and several amino acid residues on hSMS1. Such results will provide additional proofs on the catalytic mecha￾nism of hSMS1. Two types of mutations were per￾formed accordingly: on the substrate side, SM was mu￾tated in turn into PC, PE, PA, PS, and PG (Figure 1); whereas on the hSMS1 side, several important amino acid residues near the binding pocket were mutated in turn into alanine. Both types of mutations were done using the Discovery Studio software by deleting or adding some atoms while keeping original atoms as much as possible. After mutation, two rounds of mini￾mization were applied to the hSMS1/lipid complex structure with the same force field parameters described above. Finally, binding energies between the substrate molecule and hSMS1 were calculated using an implicit GB/SA membrane model for each resulting hSMS1/lipid complex (Table 3 and Table 4). Normal mode analysis and protein motion analysis of hSMS1 We conducted protein domain motion analysis for hSMS1 to study its major conformational motions with the DynDom on-line server (http://fizz.cmp.uea.ac.uk/ dyndom/).46 The initial and the last configuration of the hSMS1/SM complex in MD simulation were submitted as inputs. The DynDom server returned the information of the "hinge axes" (Figure 7). In order to confirm the outcomes of protein domain motion analysis, normal