A model of lipid-free apolipoprotein A-I revealed by iterative molecular dynamics simulation

Abstract

Apolipoprotein A-I (apo A-I), the major protein component of high-density lipoprotein, has been proven inversely correlated to cardiovascular risk in past decades. The lipid-free state of apo A-I is the initial stage which binds to lipids forming high-density lipoprotein. Molecular models of lipid-free apo A-I have been reported by methods like X-ray crystallography and chemical cross-linking/mass spectrometry (CCL/MS). Through structural analysis we found that those current models had limited consistency with other experimental results, such as those from hydrogen exchange with mass spectrometry. Through molecular dynamics simulations, we also found those models could not reach a stable equilibrium state. Therefore, by integrating various experimental results, we proposed a new structural model for lipid-free apo A-I, which contains a bundled four-helix N-terminal domain (1-192) that forms a variable hydrophobic groove and a mobile short hairpin C-terminal domain (193-243). This model exhibits an equilibrium state through molecular dynamics simulation and is consistent with most of the experimental results known from CCL/MS on lysine pairs, fluorescence resonance energy transfer and hydrogen exchange. This solution-state lipid-free apo A-I model may elucidate the possible conformational transitions of apo A-I binding with lipids in high-density lipoprotein formation.

Fig 1. Structural comparison of the four lipid-free models.

A) The X-ray model. B) The CCL/MS model. C) The MD model. D) The CMD model. The left panels show the initial structure of each protein while the right panels show structures after simulation. All proteins are represented in ribbons using Chimera. The region of residues 195 to 217 in the CMD model is indicated by purple.

Fig 2. Structural stability of the four models.

A) RMSDs of all protein atoms (without H atom) were measured over the entire trajectory. The protein structure in each frame was aligned to the initial structure, respectively. RMSDs were measured every 2 ps during the 15 ns all-atom simulation. B) The average solvent-accessible surface area of every hydrophobic residue exposed to water, measured every 20 ps in the all-atom simulations. SASAs were calculated with a 1.4 Å probe radius and a grid size of 0.25 Å. 8 types of hydrophobic residues were taken into account (Ala Leu Val Ile Pro Phe Met Trp). There are 92 total hydrophobic residues in lipid-free apo A-I models (68 in X-ray model). C) Potential energy comparisons of the four apo A-I models. Potential energy of the protein was measured every 20 ps during the simulations (from 0.4 ns to 15 ns). Potential energy includes bond, angle and dihedral angle energy, electrostatic interactions and Van der Waals interactions. The 0.4ns Cα atoms restrained simulations are not included.

Fig 3. Secondary structure comparisons of experimental lipid-free models and the MD results.

The central line represents the residue index. Rectangles along the line represent α-helices. Rectangles with half width represent β-strands. Secondary structures of all models are measured by VMD using a uniform standard. Experimental results are showing on top and structures after MD simulations are show below).

Fig 4. The distribution of neighboring Lys pairs in different lipid-free apo A-I models and their consistency to the CCL/MS experiment.

The relative distance of Cβ in Lys residues in the output structure of the simulation is measured by VMD. Hollow circles indicate experimental CCL/MS data (Davison et al [40]). The cut-off distance between two Cβ in Lys residues chosen for the calculation was 20 Å. Gray circles present the experimental data from CCL/MS method. The X and Y axes of the plot indicate the residue number of apo A-I (1–243). The cross linking data is shown through Lys pairs with a CCL link distance of 20 Å (grey circles), and compared to the CCL/MS, MD, and CMD PDB models (red plus signs). Circles with a red plus sign mean the generated model Lys distance agrees with the CCL experimental data, consequently a circle without a plus sign means the CCL data did not match the model.

Fig 5. Salt bridge distribution of lipid-free apo A-I MD model.

The backbone of apo A-I is presented by ribbons and colored by residue index from red (N-terminus) to blue (C-terminus). Inter-helix salt bridges (listed on the right) are presented with atomic bond. Acidic amino acids are colored by yellow and basic amino acids are colored by green.

Fig 6. Alternative conformation of local structure in the MD model.

A) The variable position of H5 and H6. The backbone of apo A-I C-terminus is presented in ribbons. Structures were aligned by backbone of residue 170 to 185. B) The changeable center angle in H2. The backbone of apo A-I is presented in ribbons. Structures were aligned by the backbone of residues 51 to 62. The ribbon is colored by index from red to blue. C) Surface hydrophobicity representation of MD model (only helix 1 and 2 are displayed). Hydrophobic surface is represented in orange, while hydrophilic surface is represented in blue. D) RMSDs of multiple simulations with the MD model. RMSDs of all protein atoms were measured over the entire trajectory of over seven simulations of the MD model (MD1 to MD7 indicate seven simulations, respectively).