nonbonded interaction potentials to fit the thermodynamics-based MARTINI force
field. The CG model for Chl molecules follows the degree of coarse graining of the
MARTINI force field. Somewhat in line with the general MARTINI parameterization philosophy, which focuses on partitioning properties, the nonbonded parameters were chosen such that the distribution of the CG Chl beads between hydrophilic
and hydrophobic regions in the bilayer is correctly represented, as compared to the
atomistic reference simulation. Here, particular attention was paid to the interactions of the polar center of the porphyrin ring with the lipid beads and to the polarity
of the aromatic ring, which needs to be carefully tuned to obtain the correct
distribution between the polar headgroup and the hydrophobic tail regions of the
lipid bilayer. The bonded interactions in the CG pigments were derived such that
the CG model reproduces the shape and the conformational behavior of the
atomistic Chl molecules. The overall shape of the porphyrin ring and the different
conformations of the phytol tail are well represented in this CG model. As a last
aspect of validation of the CG model, we have analyzed the propensity of the Chl
pigments to aggregate in the lipid bilayer. It was found that Chl molecules do
aggregate, with clusters that form and break multiple times in the course of the
simulation, i.e., the aggregation is not overly strong. Qualitatively, these data are
corroborated by fluorescence quenching experiments that show that chlorophylls in
lipid bilayers have a tendency to aggregate at low lipid to Chl ratios of less than
1,250 lipids/chlorophyll. Summarizing the structural behavior, the distribution of
the pigments in the bilayer (which are indicative of a correct balance of hydrophobicity and hydrophilicity) and the pigment association are very well represented in
the CG model compared to atomistic simulations and experimental data (Debnath et
al., unpublished data).
After driving the CG model parameters for the Chl–lipid system, this new model
was now combined with the MARTINI model for proteins to perform some first
simulations of the pigmented LHCII complex (in trimeric as well as monomeric
form). In addition, classical atomistic (explicit solvent) simulations of trimeric and
monomeric LHCII in a model membrane were performed to provide a reference for
validation of the CG simulations. The first CG simulations of the LHCII complex
have proven to be very promising. Unlike our initial attempts without the careful
parameterization of the pigments, the trimeric protein–pigment complex has been
structurally stable, most notably without the presence of any artificial elastic
network between the protein core and the pigments (see Fig. 3). The properties of
the complex from the CG model are in excellent agreement with those from the
atomistic model. In the future, this CG model will be used to study various aspects
of LHCII protein–protein interactions in the lipid bilayer that, on the one hand, go
beyond the time and length scales accessible to atomistic simulations alone and, on
the other hand, require a more chemically realistic description of the protein/
pigment/lipid system than typical generic CG models.
Computational Studies of Biomembrane Systems: Theoretical Considerations. . .
269
field. The CG model for Chl molecules follows the degree of coarse graining of the
MARTINI force field. Somewhat in line with the general MARTINI parameterization philosophy, which focuses on partitioning properties, the nonbonded parameters were chosen such that the distribution of the CG Chl beads between hydrophilic
and hydrophobic regions in the bilayer is correctly represented, as compared to the
atomistic reference simulation. Here, particular attention was paid to the interactions of the polar center of the porphyrin ring with the lipid beads and to the polarity
of the aromatic ring, which needs to be carefully tuned to obtain the correct
distribution between the polar headgroup and the hydrophobic tail regions of the
lipid bilayer. The bonded interactions in the CG pigments were derived such that
the CG model reproduces the shape and the conformational behavior of the
atomistic Chl molecules. The overall shape of the porphyrin ring and the different
conformations of the phytol tail are well represented in this CG model. As a last
aspect of validation of the CG model, we have analyzed the propensity of the Chl
pigments to aggregate in the lipid bilayer. It was found that Chl molecules do
aggregate, with clusters that form and break multiple times in the course of the
simulation, i.e., the aggregation is not overly strong. Qualitatively, these data are
corroborated by fluorescence quenching experiments that show that chlorophylls in
lipid bilayers have a tendency to aggregate at low lipid to Chl ratios of less than
1,250 lipids/chlorophyll. Summarizing the structural behavior, the distribution of
the pigments in the bilayer (which are indicative of a correct balance of hydrophobicity and hydrophilicity) and the pigment association are very well represented in
the CG model compared to atomistic simulations and experimental data (Debnath et
al., unpublished data).
After driving the CG model parameters for the Chl–lipid system, this new model
was now combined with the MARTINI model for proteins to perform some first
simulations of the pigmented LHCII complex (in trimeric as well as monomeric
form). In addition, classical atomistic (explicit solvent) simulations of trimeric and
monomeric LHCII in a model membrane were performed to provide a reference for
validation of the CG simulations. The first CG simulations of the LHCII complex
have proven to be very promising. Unlike our initial attempts without the careful
parameterization of the pigments, the trimeric protein–pigment complex has been
structurally stable, most notably without the presence of any artificial elastic
network between the protein core and the pigments (see Fig. 3). The properties of
the complex from the CG model are in excellent agreement with those from the
atomistic model. In the future, this CG model will be used to study various aspects
of LHCII protein–protein interactions in the lipid bilayer that, on the one hand, go
beyond the time and length scales accessible to atomistic simulations alone and, on
the other hand, require a more chemically realistic description of the protein/
pigment/lipid system than typical generic CG models.
Computational Studies of Biomembrane Systems: Theoretical Considerations. . .
269