level of hydration is correctly reproduced from direct observation of the selfassembly of either bicontinuous membranes (Fig. 3e) or micelles (Fig. 3g).
In addition to the obvious dependence on the chemical structure, different preparation techniques at the same level of hydration can lead to different structures.
Although the exact sequence of steps cannot be replicated in MD simulations,
exploiting the periodic boundary conditions used in the simulations, plus a suitably
sized model system, allows one to easily reproduce the correct final outcome. Small
model systems (replicated due to periodic boundary conditions) give rise to vesicles
or bilayers (Fig. 3b, j), whereas larger model systems allow membranes to form with
significant curvature (Fig. 3e) or give rise to micelles (Fig. 3g). Therefore, environmental conditions can be taken into account by CG–MD simulations accurately
enough that the correct supramolecular structure is obtained by self-assembly, using
no other information other than thermodynamic properties of the building blocks of
the macromolecule.
5 Response of Biological Membranes to Addition
of Macromolecules
Once a supramolecular structure is formed, the next challenge is set for MD
simulations: whether they can predict effectively the response by the supramolecular
assembly to the addition of a new component. This is typically a tough challenge for
experiments, just as it is for simulations. Due to the large number of molecules
involved, it is not unlikely to observe hysteresis when modeling the changes upon
insertion of new macromolecules. The model system first reaches one equilibrium
phase as it self-assembles, and remains in that phase even after it becomes
destabilized after new macromolecules are added. This may be the desired behavior
when preparing a supercritical condition in the laboratory; however, for simulations
the time gap between the simulated time and the laboratory time may be long enough
to generate unwanted hysteresis and prevent an accurate evaluation of the system’s
response.
The accelerated characteristic times of MD simulations with CG models are
invaluable for minimizing this problem and achieving the highest predictive power.
As proof of concept, we here review a study of the simulated effect of adding one or
more fullerene macromolecules (C540) to a phospholipid bilayer (Fig. 4) and to a
multilamellar stack (Fig. 5). In all simulations performed with this model, the insertion
of fullerene macromolecules appears to follow unimpeded diffusion into the bilayer
structure. Explicit calculation of the potentials of mean force (PMFs) of insertion
confirms this fact, both at the atomistic and at the CG level. Following the long time
(microsecond) evolution of the fullerene-phospholipid system, large membrane deformations appear that are clearly correlated to increased local concentration of fullerene
macromolecules. Coupling between contiguous bilayers in a multilamellar stack
(Fig. 5) also suggests that the CG model could be able to produce a multilamellar
stack from direct self-assembly.
100
G. Fiorin et al.
In addition to the obvious dependence on the chemical structure, different preparation techniques at the same level of hydration can lead to different structures.
Although the exact sequence of steps cannot be replicated in MD simulations,
exploiting the periodic boundary conditions used in the simulations, plus a suitably
sized model system, allows one to easily reproduce the correct final outcome. Small
model systems (replicated due to periodic boundary conditions) give rise to vesicles
or bilayers (Fig. 3b, j), whereas larger model systems allow membranes to form with
significant curvature (Fig. 3e) or give rise to micelles (Fig. 3g). Therefore, environmental conditions can be taken into account by CG–MD simulations accurately
enough that the correct supramolecular structure is obtained by self-assembly, using
no other information other than thermodynamic properties of the building blocks of
the macromolecule.
5 Response of Biological Membranes to Addition
of Macromolecules
Once a supramolecular structure is formed, the next challenge is set for MD
simulations: whether they can predict effectively the response by the supramolecular
assembly to the addition of a new component. This is typically a tough challenge for
experiments, just as it is for simulations. Due to the large number of molecules
involved, it is not unlikely to observe hysteresis when modeling the changes upon
insertion of new macromolecules. The model system first reaches one equilibrium
phase as it self-assembles, and remains in that phase even after it becomes
destabilized after new macromolecules are added. This may be the desired behavior
when preparing a supercritical condition in the laboratory; however, for simulations
the time gap between the simulated time and the laboratory time may be long enough
to generate unwanted hysteresis and prevent an accurate evaluation of the system’s
response.
The accelerated characteristic times of MD simulations with CG models are
invaluable for minimizing this problem and achieving the highest predictive power.
As proof of concept, we here review a study of the simulated effect of adding one or
more fullerene macromolecules (C540) to a phospholipid bilayer (Fig. 4) and to a
multilamellar stack (Fig. 5). In all simulations performed with this model, the insertion
of fullerene macromolecules appears to follow unimpeded diffusion into the bilayer
structure. Explicit calculation of the potentials of mean force (PMFs) of insertion
confirms this fact, both at the atomistic and at the CG level. Following the long time
(microsecond) evolution of the fullerene-phospholipid system, large membrane deformations appear that are clearly correlated to increased local concentration of fullerene
macromolecules. Coupling between contiguous bilayers in a multilamellar stack
(Fig. 5) also suggests that the CG model could be able to produce a multilamellar
stack from direct self-assembly.
100
G. Fiorin et al.