Molecular Modeling of the Three-Dimensional Structure of Human Sphingomyelin Synthase CHEMISTRY p),1.0% of residues are in the generally allowed regions (a, -b, -l, -P), and only 1.0% of residues are in the disallowed regions. Thus, the quality of the hSMSI structural model derived in our study is acceptable in this aspect According to our model ( Figure 5A), residues H285 H328, and D332 are fully involved in the catalytic mechanism. D332 can form hydrogen bond with H328 which may act as hydrogen donor during catalysis. The choline group on the Sm molecule stretches well into a hydrophobic patch formed by F173, F177, H274, and L281. Y338 and R342 can form hydrogen bond with SM. With these two kinds of interactions, the catalytic center on SM, i.e. the phosphorus atom, is well posi tioned inside the catalytic pocket of hSMSl(Figure 5B) Our model(Figure 4)is generally consistent with the possible catalytic mechanism of hSMSl, which is given in the Introduction section of this article Figure 9 The binding mode of D609 with hSMSI derived Mutagenesis study on hSMSl/ipid complexes through molecular docking. hSMSI is rendered in ribbons whereas D609 is rendered in stick model. The shaded region in- According to our computation results(Table 4),mu- dicates the lipid bilayer tation of SM into other possible lipid substrates, in- cluding PG, PE, PA, and PS(Figure 1), leads to a de similarity high enough with hSMSI, especially for the crease in binding energy of hSMSI/PG,hSMSI/PE transmembrane domain. To tackle this problem, we re SMSI/PA, and hSMS 1/PS complexes as compared to trieved all known membrane proteins with six TMs, 61 that of the hSMSl/SM complex. However, mutation of PDBTM34 Then, we examined them visually one by above. The binding energy between hSMSl and PC template. Firstly, those largely different from hSMSI in hSMS1/SM. Based on the computed binding energies, terms of biological function, such as aquaporin-like the choline group on SM is very important to binding proteins, were excluded because they usually form a between hSMsI and SM. These results are consistent channel-like shape encircled by transmembrane helices. with Huitema's experimental proofs: none of the In addition, qualified candidates must be an enzyme non-choline phospholipids Secondly, for obvious reason, the TMs on the qualified than PC can trigger the production of sM. On the other candidates should be similar to those of hSMSI in terms hand, PC was efficiently recognized as a substrate of of length. After careful evaluation of all 61 entries, we SMS and SM itself had a donor of the phosphorylcho- chose the crystal structure of Escherichia coli GlpG line group, which can also be transformed back to PC (PDB entry: 2IC8)as the template. Indeed, the length of Therefore our model can, at least in a qualitatively TMs in GlpG is comparable to the counterparts in manner, explain the high selectivity of hSMSI towards hSMSI (Table 2). In addition, there is an internal cavity PC and sm on GlpG that harbors the Ser-His catalytic site, which is To better estimate the contribution of key residues to similar to hSMSl that catalyzes PC into SM. An ex pinding affinity between hSMSI and SM, computa- bi tracellular loop on GlpGare is also very similar to the tional mutagenesis, i.e. alanine scanning, was conducted counterpart on hSMSl with two conserved amino acid in our study on some residues near the catalytic site on esidues(H328 and D332) hSMSI. For example, H285, H328, and D332 are Our initial model of the hSMSI/SM complex struc- known to be three highly conserved amino acids in SMS ture was constructed through homology modeling ar families, even in lipid phosphatases/phosphotransferases, molecular docking. It was then subjected to a long-time and are key components participating in the catalytic MD simulation for refinement. An implicit membrane reaction. Mutation of these three key residues on model, which is basically a GB/SA model, was em- hSMSI will lose catalytic activity completely. In Ta ployed in simulation to mimic the real environment ble 4 and Figure 5A, one can see that mutation of round hSMSl. The final model of the hSMSI/SM com- D332A has little influence on the binding affinity be- plex structure is shown in Figure 5. The Ramachandran tween hSMSI and SM. However, the carboxyl group on lot of this hSMSl structure produced by Procheck d332 functions as a hydrogen acceptor to stabilize the (Figure 6)indicates that 87.5% of residues on our model transition state during the catalysis process. As for H285 are in the most favored regions(A, B, L), 10.5% of and H328, they are very important in both the binding residues are in the additionally allowed regions(a, b, 1, process and the subsequent catalytic process. Mutatio Chin J. Chem. 2011, 29, 1567--1575 C2011 SIOC, CAS, Shanghai, WILEY-VCH Verlag gmbH Co KGaA, Weinheim wwcjc. wiley-vch. de 1573Molecular Modeling of the Three-Dimensional Structure of Human Sphingomyelin Synthase Chin. J. Chem. 2011, 29, 1567—1575 © 2011 SIOC, CAS, Shanghai, & WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cjc.wiley-vch.de 1573 Figure 9 The binding mode of D609 with hSMS1 derived through molecular docking. hSMS1 is rendered in ribbons whereas D609 is rendered in stick model. The shaded region in￾dicates the lipid bilayer. similarity high enough with hSMS1, especially for the transmembrane domain. To tackle this problem, we re￾trieved all known membrane proteins with six TMs, 61 in total, from the membrane structural proteins database PDBTM.34 Then, we examined them visually one by one with two criteria to select the most suitable one as template. Firstly, those largely different from hSMS1 in terms of biological function, such as aquaporin-like proteins, were excluded because they usually form a channel-like shape encircled by transmembrane helices. In addition, qualified candidates must be an enzyme. Secondly, for obvious reason, the TMs on the qualified candidates should be similar to those of hSMS1 in terms of length. After careful evaluation of all 61 entries, we chose the crystal structure of Escherichia coli GlpG (PDB entry: 2IC8) as the template. Indeed, the length of TMs in GlpG is comparable to the counterparts in hSMS1 (Table 2). In addition, there is an internal cavity on GlpG that harbors the Ser-His catalytic site, which is similar to hSMS1 that catalyzes PC into SM. An ex￾tracellular loop on GlpGare is also very similar to the counterpart on hSMS1 with two conserved amino acid residues (H328 and D332). Our initial model of the hSMS1/SM complex struc￾ture was constructed through homology modeling and molecular docking. It was then subjected to a long-time MD simulation for refinement. An implicit membrane model, which is basically a GB/SA model, was em￾ployed in simulation to mimic the real environment round hSMS1. The final model of the hSMS1/SM com￾plex structure is shown in Figure 5. The Ramachandran plot of this hSMS1 structure produced by PROCHECK (Figure 6) indicates that 87.5% of residues on our model are in the most favored regions (A, B, L), 10.5% of residues are in the additionally allowed regions (a, b, l, p), 1.0% of residues are in the generally allowed regions (~a, ~b, ~l, ~p), and only 1.0% of residues are in the disallowed regions. Thus, the quality of the hSMS1 structural model derived in our study is acceptable in this aspect. According to our model (Figure 5A), residues H285, H328, and D332 are fully involved in the catalytic mechanism. D332 can form hydrogen bond with H328, which may act as hydrogen donor during catalysis. The choline group on the SM molecule stretches well into a hydrophobic patch formed by F173, F177, H274, and L281. Y338 and R342 can form hydrogen bond with SM. With these two kinds of interactions, the catalytic center on SM, i.e. the phosphorus atom, is well posi￾tioned inside the catalytic pocket of hSMS1 (Figure 5B). Our model (Figure 4) is generally consistent with the possible catalytic mechanism of hSMS1, which is given in the Introduction section of this article. Mutagenesis study on hSMS1/lipid complexes According to our computation results (Table 4), mu￾tation of SM into other possible lipid substrates, in￾cluding PG, PE, PA, and PS (Figure 1), leads to a de￾crease in binding energy of hSMS1/PG, hSMS1/PE, hSMS1/PA, and hSMS1/PS complexes as compared to that of the hSMS1/SM complex. However, mutation of SM into PC is quite different from lipid without choline above. The binding energy between hSMS1 and PC is a bit lower, i.e. more favorable, if compared to hSMS1/SM. Based on the computed binding energies, the choline group on SM is very important to binding between hSMS1 and SM. These results are consistent with Huitema’s experimental proofs:10 none of the non-choline phospholipids (PG, PE, PA, and PS) other than PC can trigger the production of SM. On the other hand, PC was efficiently recognized as a substrate of SMS and SM itself had a donor of the phosphorylcho￾line group, which can also be transformed back to PC. Therefore our model can, at least in a qualitatively manner, explain the high selectivity of hSMS1 towards PC and SM. To better estimate the contribution of key residues to binding affinity between hSMS1 and SM, computa￾tional mutagenesis, i.e. alanine scanning, was conducted in our study on some residues near the catalytic site on hSMS1. For example, H285, H328, and D332 are known to be three highly conserved amino acids in SMS families, even in lipid phosphatases/phosphotransferases, and are key components participating in the catalytic reaction. Mutation of these three key residues on hSMS1 will lose catalytic activity completely.17 In Ta￾ble 4 and Figure 5A, one can see that mutation of D332A has little influence on the binding affinity be￾tween hSMS1 and SM. However, the carboxyl group on D332 functions as a hydrogen acceptor to stabilize the transition state during the catalysis process. As for H285 and H328, they are very important in both the binding process and the subsequent catalytic process. Mutation