The interest in monolayer-protected gold NPs is motivated by the relatively inert and thus biocompatible nature of Au, and by its electronic and optical properties.(1-3) Gold is also easily passivated via Au–S bonds(5) to organothiolates, whose terminal groups can be chemically designed to finely tune the NP degree of hydrophobicity.
Many open questions are still unanswered. What is the role played by electrostatics at determining the type of interaction with lipid membranes? Cationic NPs are generally reported to be more toxic than anionic NPs.(6-8) Recent neutron scattering data by Tatur et al.,(8) suggest that anionic Au NPs may not enter the hydrophobic core of zwitterionic lipid membranes at all, simply adhering to their surface in the fluid phase and leading to lipid dehydration. Van Lehn et al.(9) indicate a stable binding to zwitterionic bilayers and the possibility of passive membrane translocation. Recent centrifugation-based assays suggest that PEG-passivated Ag NPs, bearing a small negative charge, do interact with zwitterionic vesiscles affecting their precipitation behavior.(10) Finer details concerning the arrangement of the charged ligands on the NP surface might affect the NP–membrane interaction and possibly explain the broad range of behaviors that so far have been observed experimentally.(11-13)
In the last couple of years, computational modeling has contributed to sketch a possible mechanism of interaction of anionic NP with zwitterionic lipid membranes. The first phases of such interaction have been elucidated via both atomistic(14-16) and coarse-grained (CG)(17) molecular dynamics (MD) simulations. It is now clear that electrostatic attraction between the charged ligands and the polar heads of zwitterionic lipid in the fluid phase drives the adhesion of the NP to the membrane surface. At the other end of the pathway, thermodynamics-based, implicit solvent and implicit bilayer models indicate that the most stable NP transmembrane state may correspond to the so-called “snorkeling” configuration.(9) In this configuration, the center of mass of the NP is embedded in the membrane core, while the charged ligand terminals stably interact with the lipid head regions of both leaflets. The all-atom (AA) MD simulations performed by Heikkilä et al.(14) and the CG ones performed by Gkeka et al.(17) could not observe any spontaneous penetration of the NP into the membrane core, due to the limited sampling time. Van Lehn et al. observed via AA simulations the spontaneous insertion of the NPs only at the highly curved edge of a lipid bicelle,(15) where the process is mediated by the protrusion of a lipid tail out of the hydrophobic membrane core. When interacting with a flat membrane, the insertion process has been reproduced only via biased simulations, either favoring a lipid–ligand hydrophobic contact by imposing an external driving potential on one lipid tail,(16) or directly forcing the NP in the center of the bilayer by removal of the overlapping lipids.(17) So far, no unbiased simulation of the insertion process has been performed, and the kinetics of the process has not been described.
Concerning the influence of the NP surface pattern, no clear picture has emerged so far from either thermodynamics-based models or MD simulations. The implicit bilayer and implicit solvent model by Van Lehn et al. does not predict any substantial difference in the water–membrane free energy of transfer of NPs with random or striped ligand patterns.(9, 18) Gkeka et al., based on a rigid-sphere model of the NP, calculated the water–membrane free energy of transfer of NPs with homogeneous or random arrangement of hydrophobic and charged beads on the surface, and concluded that the NPs with a homogeneous pattern should passively translocate through the membrane more easily than those with a random arrangement.(19)
In this paper, we present CG unbiased MD simulation of the whole interaction process, and conclude that anionic gold NPs do insert in the membrane core, the final snorkeling configuration being energetically highly favorable. Our unbiased simulations show that the insertion process is indeed mediated by the spontaneous protrusion of a lipid tail that initiates the NP-membrane hydrophobic contact. This stage is followed by the dropping of a charged ligand (an “anchor”) to the opposite leaflet, thus stabilizing the NP-membrane complex. Eventually, more and more anchors are dropped leading to the final snorkeling configuration. Our free-energy calculations show that the anchoring process is almost irreversible.
We furthermore show that the kinetics of the process depends on the NPs’ ligand composition and surface arrangement. Our calculations show that the interaction free energy profile of the NPs with a random surface arrangement of anionic and hydrophobic ligands is characterized by two metastable minima, corresponding to the surface-adsorbed configuration and to the snorkeling configuration. NPs whose ligands form large hydrophobic or anionic surface patches, instead, go through three metastable configurations. The transition from the adsorbed state to the snorkeling state is indeed slowed down by a significant energy barrier, which stabilizes an intermediate metastable state, in which the NP is semiadsorbed.
CG Model. We considered a Au144(SR)60 NP, R being either a hydrophobic octanethiol ligand (OT) or an anionic 11-mercaptoundecanesulfonate (MUS). The diameter of the Au core is about 2 nm (Supporting Figure S1), while the monolayer-protected NP has an overall diameter of about 4 nm. We modeled the Au–Au and Au–S interactions using an elastic network, while we developed a CG model of the ligands based on the popular Martini force field(20) (Figure 1). At CG level, the OT ligands are modeled as a chain of two hydrophobic (type C1) beads. The MUS ligands are described by three hydrophobic beads (Martini type C1) and one negatively charged terminal bead (type Qda). The details of the parametrization can be found in the Supporting Information. We remark here that our CG description does not distinguish between, e.g., mercaptoundecanoic acid and mercaptoundecanesulfonate, allowing for the direct comparison with several independent previous computational (AA) and experimental works.
One possible reason for concern about the use of CG models to study anionic NP–membrane interactions is the treatment of electrostatic interactions. Contrary to AA models, which parametrization includes long-range electrostatics, the Martini force field sharply cuts off Coulomb interactions at short distances (1.2 nm). In order to validate our CG model in this respect, we compared the three-dimensional spatial distributions of the passivated Au NPs in water to previous atomistic simulations (Supporting Figures S2 and S3). We obtained a satisfactory overlapping of all the partial radial distribution functions (RDF) for hydrophobic moieties, charged ligands, counterions, and water. In previous simulations of the interaction between charged dendrimers and lipid membranes, Lee and Larson(21) proposed to include the long-range electrostatic contributions into the MARTINI force field by implementing the Particle-Mesh-Ewald method. We tested this approach as well, but observed no substantial changes of the RDFs (Supporting Figure S4), together with obvious computational disadvantages. We calculated the Debye length, which measures the screening effects of counterions in an electrolytic solution, for the Na+ counterions surrounding our anionic NP in water. The Debye length can be deduced from the fitting of the counterions RDF at long distances to the Debye–Hückel distribution f(r) = Ae–Br + C, where A, B, and C are positive parameters, and r is the distance from the center of the charged NP. The Debye length LD is the inverse of B. From the fitting procedure, applied to the ion RDFs as obtained with plain Coulomb cutoff, the Debye length is 0.18 nm (Supporting Figure S5), in good agreement with the value obtained by Heikkilä et al.(22) via AA simulations (LD = 0.20 nm). More details on the model validation can be found in the Supporting Information.
We looked at three different NPs: (a) MUS:OT 2:1 ligand composition, with random surface arrangement of the ligands, (b) MUS:OT 1:1, random, and (c) MUS:OT, 1:1 with a patched arrangement, made of a central hydrophobic (OT) stripe flanked by two charged (MUS) poles. All NPs are shown in Figure 1, and the protocol for their construction is described in the Supporting Information. Our model lipid membrane is a patch of 512 phosphatidylcholine (POPC) lipids.
The interest in monolayer-protected gold NPs is motivated by the relatively inert and thus biocompatible nature of Au, and by its electronic and optical properties.(1-3) Gold is also easily passivated via Au–S bonds(5) to organothiolates, whose terminal groups can be chemically designed to finely tune the NP degree of hydrophobicity.Many open questions are still unanswered. What is the role played by electrostatics at determining the type of interaction with lipid membranes? Cationic NPs are generally reported to be more toxic than anionic NPs.(6-8) Recent neutron scattering data by Tatur et al.,(8) suggest that anionic Au NPs may not enter the hydrophobic core of zwitterionic lipid membranes at all, simply adhering to their surface in the fluid phase and leading to lipid dehydration. Van Lehn et al.(9) indicate a stable binding to zwitterionic bilayers and the possibility of passive membrane translocation. Recent centrifugation-based assays suggest that PEG-passivated Ag NPs, bearing a small negative charge, do interact with zwitterionic vesiscles affecting their precipitation behavior.(10) Finer details concerning the arrangement of the charged ligands on the NP surface might affect the NP–membrane interaction and possibly explain the broad range of behaviors that so far have been observed experimentally.(11-13)In the last couple of years, computational modeling has contributed to sketch a possible mechanism of interaction of anionic NP with zwitterionic lipid membranes. The first phases of such interaction have been elucidated via both atomistic(14-16) and coarse-grained (CG)(17) molecular dynamics (MD) simulations. It is now clear that electrostatic attraction between the charged ligands and the polar heads of zwitterionic lipid in the fluid phase drives the adhesion of the NP to the membrane surface. At the other end of the pathway, thermodynamics-based, implicit solvent and implicit bilayer models indicate that the most stable NP transmembrane state may correspond to the so-called “snorkeling” configuration.(9) In this configuration, the center of mass of the NP is embedded in the membrane core, while the charged ligand terminals stably interact with the lipid head regions of both leaflets. The all-atom (AA) MD simulations performed by Heikkilä et al.(14) and the CG ones performed by Gkeka et al.(17) could not observe any spontaneous penetration of the NP into the membrane core, due to the limited sampling time. Van Lehn et al. observed via AA simulations the spontaneous insertion of the NPs only at the highly curved edge of a lipid bicelle,(15) where the process is mediated by the protrusion of a lipid tail out of the hydrophobic membrane core. When interacting with a flat membrane, the insertion process has been reproduced only via biased simulations, either favoring a lipid–ligand hydrophobic contact by imposing an external driving potential on one lipid tail,(16) or directly forcing the NP in the center of the bilayer by removal of the overlapping lipids.(17) So far, no unbiased simulation of the insertion process has been performed, and the kinetics of the process has not been described.Concerning the influence of the NP surface pattern, no clear picture has emerged so far from either thermodynamics-based models or MD simulations. The implicit bilayer and implicit solvent model by Van Lehn et al. does not predict any substantial difference in the water–membrane free energy of transfer of NPs with random or striped ligand patterns.(9, 18) Gkeka et al., based on a rigid-sphere model of the NP, calculated the water–membrane free energy of transfer of NPs with homogeneous or random arrangement of hydrophobic and charged beads on the surface, and concluded that the NPs with a homogeneous pattern should passively translocate through the membrane more easily than those with a random arrangement.(19)In this paper, we present CG unbiased MD simulation of the whole interaction process, and conclude that anionic gold NPs do insert in the membrane core, the final snorkeling configuration being energetically highly favorable. Our unbiased simulations show that the insertion process is indeed mediated by the spontaneous protrusion of a lipid tail that initiates the NP-membrane hydrophobic contact. This stage is followed by the dropping of a charged ligand (an “anchor”) to the opposite leaflet, thus stabilizing the NP-membrane complex. Eventually, more and more anchors are dropped leading to the final snorkeling configuration. Our free-energy calculations show that the anchoring process is almost irreversible.We furthermore show that the kinetics of the process depends on the NPs’ ligand composition and surface arrangement. Our calculations show that the interaction free energy profile of the NPs with a random surface arrangement of anionic and hydrophobic ligands is characterized by two metastable minima, corresponding to the surface-adsorbed configuration and to the snorkeling configuration. NPs whose ligands form large hydrophobic or anionic surface patches, instead, go through three metastable configurations. The transition from the adsorbed state to the snorkeling state is indeed slowed down by a significant energy barrier, which stabilizes an intermediate metastable state, in which the NP is semiadsorbed.
CG Model. We considered a Au144(SR)60 NP, R being either a hydrophobic octanethiol ligand (OT) or an anionic 11-mercaptoundecanesulfonate (MUS). The diameter of the Au core is about 2 nm (Supporting Figure S1), while the monolayer-protected NP has an overall diameter of about 4 nm. We modeled the Au–Au and Au–S interactions using an elastic network, while we developed a CG model of the ligands based on the popular Martini force field(20) (Figure 1). At CG level, the OT ligands are modeled as a chain of two hydrophobic (type C1) beads. The MUS ligands are described by three hydrophobic beads (Martini type C1) and one negatively charged terminal bead (type Qda). The details of the parametrization can be found in the Supporting Information. We remark here that our CG description does not distinguish between, e.g., mercaptoundecanoic acid and mercaptoundecanesulfonate, allowing for the direct comparison with several independent previous computational (AA) and experimental works.
One possible reason for concern about the use of CG models to study anionic NP–membrane interactions is the treatment of electrostatic interactions. Contrary to AA models, which parametrization includes long-range electrostatics, the Martini force field sharply cuts off Coulomb interactions at short distances (1.2 nm). In order to validate our CG model in this respect, we compared the three-dimensional spatial distributions of the passivated Au NPs in water to previous atomistic simulations (Supporting Figures S2 and S3). We obtained a satisfactory overlapping of all the partial radial distribution functions (RDF) for hydrophobic moieties, charged ligands, counterions, and water. In previous simulations of the interaction between charged dendrimers and lipid membranes, Lee and Larson(21) proposed to include the long-range electrostatic contributions into the MARTINI force field by implementing the Particle-Mesh-Ewald method. We tested this approach as well, but observed no substantial changes of the RDFs (Supporting Figure S4), together with obvious computational disadvantages. We calculated the Debye length, which measures the screening effects of counterions in an electrolytic solution, for the Na+ counterions surrounding our anionic NP in water. The Debye length can be deduced from the fitting of the counterions RDF at long distances to the Debye–Hückel distribution f(r) = Ae–Br + C, where A, B, and C are positive parameters, and r is the distance from the center of the charged NP. The Debye length LD is the inverse of B. From the fitting procedure, applied to the ion RDFs as obtained with plain Coulomb cutoff, the Debye length is 0.18 nm (Supporting Figure S5), in good agreement with the value obtained by Heikkilä et al.(22) via AA simulations (LD = 0.20 nm). More details on the model validation can be found in the Supporting Information.
We looked at three different NPs: (a) MUS:OT 2:1 ligand composition, with random surface arrangement of the ligands, (b) MUS:OT 1:1, random, and (c) MUS:OT, 1:1 with a patched arrangement, made of a central hydrophobic (OT) stripe flanked by two charged (MUS) poles. All NPs are shown in Figure 1, and the protocol for their construction is described in the Supporting Information. Our model lipid membrane is a patch of 512 phosphatidylcholine (POPC) lipids.
การแปล กรุณารอสักครู่..
