Team:Aalto-Helsinki/Model Results
Results and Discussion
Based on molecular dynamics simulations of DCD-1L at different salt concentrations and temperatures, we evaluated occurring changes in protein secondary structure through secondary structure assignment using DSSP and observed Ramachandran plot representations. Additionally we characterized the structure of the simulated peptide by scrutinizing the radius of gyration, solvent accessible surface area and hydrogen bonding.
Characterization of the amino acid sequence
Despite the great diversity of antimicrobial peptides, they exhibit many commonly shared characteristics. Such characteristics include the presence of multiple basic amino acids and amphipathic structures consisting of clusters of hydrophobic and hydrophilic amino acids. Commonly, AMPs have a positive net charge which has been shown to contribute to their initial binding to the often negatively charged bacterial cell membrane. The amphipathic structures found in AMPs have been found to enhance interactions with the phospholipids of the bacterial cell membrane, especially at the water-lipid interface.
DCD-1L consists of 48 amino residues and is processed proteolytically from a 110 amino acid precursor peptide. Contrary to many common AMPs, DCD-1L has a net negative charge at physiological pH. [10] The expected pI of DCD-1L is 5.07 and the expected molecular weight of the peptide is 4818.50 g/mol. Using a web-based tool Composition Profiler[16] we attempted to identify differences in amino acid residue expression in the sequence of DCD-1L when compared to the distribution of the 20 natural amino acids commonly found in nature. A p-value of 0.05 was used as a threshold and Swissprot51 dataset[1], which closely mirrors the amino acid distribution found in nature, was used for background screening. Figure 1 shows the results of the amino acid expression screening with the amino acids being color coded according to how often they are commonly found in β-helical structures. We find that especially the amino acids Aspartate, Glycine, Valine and Lysine are over-expressed when compared to the frequency of these amino acids commonly found in nature. Of the afore mentioned amino acids, Aspartate exhibits a negatively charged side chain, while Glycine, Valine are considered hydrophobic and Lysine polar amino acids. According to figure 1 Lysine is an amino acid commonly found in alpha helical structures, which are common for many AMPs[10][2]. The over-expression of negatively charged Aspartate residues is expected considering the overall negative net charge of DCD-1L. Additionally, in figure 1 we have included the amino acid sequence of DCD-1L in color differentiating between polar and hydrophobic residues. Unsurprisingly, DCD-1L appears to have alternating clusters of polar and hydrophobic amino acids, with such amphipathic regions being common for AMPs as previously discussed.
Figure 1. Differences in expression of the 20 natural amino acids in DCD-1L sequence based on analysis by Composition profiler. The residues have been color coded according to the frequency at which they contribute to the α-helix secondary structure. Additionally the amino acid sequence of DCD-1L has been provided and colored according to the presence of polar and hydrophobic amino acid residues.
Radius of gyration and solvent accessible surface area
Protein interactions are often accompanied by significant changes in conformation. The radius of gyration describes the distribution of an objects mass around and axis. In polymer physics, the radius of gyration is one of the measures used to describe the dimensions of a polymer chain or, in this case, our polypeptide chain. The radius of gyration can be determined experimentally through static light scattering, small angle neutron- (SANS) and x-ray scattering (SAXS) experiments and thus allows a point of reference for checking a computational model’s accuracy. Based on our simulation trajectories, we calculated both the time evolution of the radius gyration over the course of our simulation and the average radius of gyration for the last 5 ns of simulation. The time evolution of the radius of gyration for DCD-1L simulated at different temperatures has been presented in figure 2. In all of our simulations we observe a clear decrease in the radius of gyration of our protein, which we link to the aggregation of our protein structure and the assumption of a random coil-like structure as the more constrained α-helical secondary structure is quickly decomposed due to protein-solvent interaction. The average radius of gyration as a function of both salt concentration and system temperature has been plotted in figure 3. We observe a rising trend in the radius of gyration as temperature is increased, based on which we can postulate that the protein assumes a looser random coil conformation as temperature is increased.
Figure 2. Left: Radius of gyration of DCD-1L as a function of simulation time. Right: Solvent accessible surface area as a function of simulation time. Data for DCD-1L simulated at 290 K, 300 K, 310 K and 320 K has been presented.
Figure 3. Radius of gyration averages over the last 5 ns of simulation time as a function of NaCl concentrations (left) and temperature (Right). The error bars denoting standard deviation.
Additionally, we probed the change in protein solvent accessible surface area[5] as a function of simulation time. The solvent accessible surface area was calculated using a probe radius of 0.14 nm which closely corresponds to that of a water molecule. Recently, it has been suggested that solvent accessible surface area can be used to help better predict assignment of protein secondary structure[9]. Generally, folding of soluble protein decreases the area in contact with the solvent, which is reflected as a decrease in solvent accessible surface area. The decrease in solvent accessible surface area is generally more pronounced for hydrophobic residues. Lins et al.[9] analyzed the solvent accessible surface area of 587 proteins using NMR and X-ray diffraction and found that folding of the protein tends to equilibrate the hydrophobic and hydrophilic solvent-accessible areas, with the ration of the two being close to 1. As such, the surface of a soluble protein is not only hydrophilic, as is often wrongly assumed. In our simulations, we observe a slight decrease in the solvent accessible area across all simulated systems. This hints that the resulting random coil structure is tight enough to restrict penetration of the structure by water molecules. The reduction of solvent accessible surface area can be observed in figure 2. We also calculated the average solvent accessible surface area over the last 5 ns of simulation time, which has been plotted in figure 4 as a function of both salt concentration and system temperature. While there appears no clear trend in the solvent accessible surface area as a function of salt concentration, an increasing trend is observed with respect to increased system temperature. This gives further evidence to the hypothesis that the protein assumes a looser conformation at higher temperatures, which is more permeable to water molecules.
Figure 4. Solvent accessible surface area averages over the last 5 ns of simulation time as a function of NaCl concentrations (left) and temperature (Right). The error bars denoting standard deviation.
Lastly, we examined the number of hydrogen bonds present in the protein structure as a function of simulation time. The importance of hydrogen bonding is paramount in the formation of protein secondary structure and folding. The average number of hydrogen bonds present in the protein structure as a function of both concentrations of NaCl and system temperature have been presented in figure 5. The averages were calculated over the last 5 ns of simulation time. While the average number of hydrogen bonds within the structure appears to remain constant within the margin of error as salt concentration in the system is varied, there appears a decreasing trend in the number of hydrogen bonds found in the protein backbone as as system temperature is increased. This corresponds with the assumption that the protein assumes a loosed random coil conformation at higher temperatures leading to increased distance between amino acid residues.
Figure 5. Averages of the number of hydrogen bonds in the protein structure over the last 5 ns of simulation time as a function of NaCl concentrations (left) and temperature (Right). The error bars denoting standard deviation.
Assignment of protein secondary structure elements
Secondary structure assignment using DSSP
The DSSP algorithm calculates the most likely secondary structure assignment element wise based on the protein three dimensional structure. Based on individual atom coordinate extracted from the simulation trajectory, the algorithm first calculates the H-bond energy between all atoms. The positions of hydrogen atoms are discarded from the original trajectory and instead replaced at 0.1 nm of the backbone N in the opposite direction from the backbone C=O bond. The DSSP algorithm uses the best two possible H-bonds to assign the most likely secondary structure element. The DSSP algorithm thus requires knowledge of the protein three dimensional structure for secondary structure assignment and cannot thus function with only the amino acid sequence.[15; 7] However, this makes the algorithm easy to implement on a series of simulation trajectories as we have done here. Here, we used the Gromacs v.4.6.7 interface for DSSP to calculate secondary structure assignment for our simulated peptides as a function of simulation time as well as extract a distribution of the different secondary structure elements for the equilibrated protein structures. The secondary structure according to the DDSP algorithm as a function of simulation time has been presented in figure 6 for DCD-1L in varying concentrations of NaCl and in figure 8 for DCD-1L with counter ions at varying temperatures.
In the beginning of the simulation, the starting structure of the DCD-1L peptide is that of a standard right-handed α-helix. In a standard α-helix, each amino acid residue corresponds to a 100° turn in the helix and a translation of 0.15 nm along the helical axis, with the standard pitch per turn of the helix being 0.54 nm. The standard hydrogen bonding pattern of an α-helix is as follows: the N-H group of an amino acid forms a hydrogen bond with the C=O group of the amino acid four residues earlier.[8]
Based on our simulations of DCD-1L in different NaCl concentrations, we see a rapid degradation of the α-helical structure of our peptide. However, in some cases of DCD-1L in different salt environments, some of the alpha helical structure is maintained, though limited to only small, local regions of the peptide, mostly in the region 22th to 27th amino acid in the sequence.
Figure 6. Secondary structure assignment according the DSSP algorithm as a function of simulation time. Data for DCD-1L peptide simulated with NaCl concentrations of 20 mM, 50 mM, 70 mM, 100 mM, 120 mM, 150 mM, 200 mM, 300 mM and 500 mM is presented.
Figure 7. a.) Presence of secondary structure elements in unfolded DCD-1L structures as assigned by the DSSP algorithm. b.) Presence of helical secondary structures in unfolded DCD-1L peptide as assigned by the DSSP algorithm. Data for DCD-1L peptide simulated with NaCl concentrations of 20 mM, 50 mM, 70 mM, 100 mM, 120 mM, 150 mM, 200 mM, 300 mM and 500 mM is presented
However, more notable is the shift of the protein secondary structure from the standard α-helix to a 5-helix, also referred to as the pi helix. The 5-helix is characterized by its differing hydrogen bonding pattern from that of a typical α-helix, in which the N-H group of an amino acid forms a hydrogen bond with the C=O group of the amino acid five residues earlier. The 5-helix is thought to have originated from a destabilizing insertion of a single amino acid into an α-helix and is thus generally selected against in evolution unless they contribute to a functional advantage of the protein.[8] The fact that DCD-1L assumes the 5-helical structure in aqueous environments could in part explain its ability to remain anticlimactically active, as has been commonly reported[13; 14]. Based on our simulations, we can additionally discern that the 5-helicity appears to be curiously confined to the first half of the peptide sequence. Interestingly, it has been previously hypothesized that this mainly cationic first half of the DCD-1L peptide is responsible for initial interaction with the negatively charged bacterial cell membrane[4].
We observe somewhat similar results when we examine the effect of temperature on the DCD-1L secondary structure, with the amount of helicity that is retained varying wildly between different temperatures without any clear trend. However, it is notable that a significant amount of helicity is retained in simulations carried out at 310 K, a temperature that closely matches that of standard body temperature.
Figure 8. Secondary structure assignment according the DSSP algorithm as a function of simulation time. Data for DCD-1L peptide simulated with Na+ counterions at temperatures of 290 K, 300 K, 310 K and 320 K.
References
[1] Writers, YEAR. Name of article / book. Publication. Accessible at: [url here].
[2] Writers, YEAR. Name of article / book. Publication. Accessible at: [url here].
[3] Writers, YEAR. Name of article / book. Publication. Accessible at: [url here].
[4] Writers, YEAR. Name of article / book. Publication. Accessible at: [url here].
[5] Writers, YEAR. Name of article / book. Publication. Accessible at: [url here].
