1. Introduction
One of the key molecular interactions in living organisms is the interaction between amino acids and lipid bilayers of cell membranes. Lipid bilayers consist of amphipathic phospholipids that are self-assembled with long hydrophobic tails buried in the center and hydrophilic heads facing the intracellular (cytosolic) and extracellular water environment. About 25% of the membrane surface is occupied by integral membrane proteins [
1] that serve as gatekeepers, transporters, or signal transducers [
2]. The interactions between amino acids and lipids govern the structure folding and functions of integral membrane proteins.
Cell membranes prevent the entrance of undesirable extracellular molecules so as to protect the interior of cells. However, cell membranes are permeable to some peptides. These cell-permeating peptides (CPP), which have highly different efficiency [
3,
4], are classified into three classes: positively charged amino acid residues, alternative pattern of polar and non-polar amino acid residues, and hydrophobic peptides with low net charge [
4]. Cell-permeating peptides have been actively investigated as antibiotics [
5] and cell-specific drug delivery agents [
6]. However, the mechanism of cell permeation is poorly understood and such understanding requires an improved knowledge of how amino acids interact with the membrane.
One experimental strategy to probe energetics of amino acids is to measure partitioning between water and a lipid environment [
7]. For example, Wimley and White measured the transfer free energy of short peptides (Ace-WLxLL) from water and a palmitoyloleoylphosphatidylcholine (POPC) membrane [
8] and showed preference of aromatic but not charged residues at the membrane hydrophobic center. Hessa et al. measured a “biological hydrophobicity scale” by observing how peptides with different residues were selected for inserting into or transporting out of the membrane by translocation machinery Sec translocon [
9]. More recently, McDonald and Fleming performed mutations of alanine to aromatic residues in the outer membrane protein phospholipase A (OmpLA) and demonstrated the dependence of transfer free energy of aromatic side chains on the depth inside the membrane [
10]. Consistent with their partitioning coefficients, direct measurement of permeability of membranes to amino acids indicated that hydrophobic amino acids are much more permeable than hydrophilic ones [
11].
Computationally, partitioning of amino acids in hydrophilic and hydrophobic environments can be estimated from protein structures according to percentages of buried residues and protein stability [
12], and were found useful in locating trans-membrane helices [
13]. Distributions of amino acid residues in transmembrane helical proteins revealed vastly different statistical depth-dependent profiles for different residues [
14]. Molecular dynamics simulations have also played a significant role in understanding the interaction between amino acid residues and the cell membrane at atomic details. For example, experimental partitioning of 10 short peptides (Ace-WLxLL) at water/cyclohexane and water/phospholipid interfaces can be interpreted at molecular level by molecular dynamics simulations [
15,
16]. Distribution of 17 amino-acid analogs (without main-chain backbone atoms) across a dioleoylphosphatidylcholine (DOPC) bilayer [
17] was obtained by umbrella sampling and revealed the importance of water defects in permeation of polar and charged residues. Other studies focused on the permeation mechanism of specific charged residues such as charged tryptophan [
18] and arginine [
19,
20].
In this study, we performed bias exchange metadynamics simulations of the permeation of 20 standard amino acids from water to the center of a large Dipalmitoylphosphatidylcholine (DPPC) membrane system (consisting of 256 molecules). The large membrane system was chosen because the energy profile of a permeating amino acid is quantitatively affected by the size of the lipid bilayer [
20]. DPPC was chosen because it is a commonly used model for biological membranes of mammalian cells [
21]. Metadynamics molecular dynamics simulations explore free energy landscapes along several pre-chosen collective variables by using positive Gaussian potentials to bias against previously visited regions [
22,
23,
24]. It has been successfully applied to the water-to-membrane transfer of small molecules and compared to experimental measurement [
25,
26,
27,
28]. In particular, it has successfully found that the conformational transition of aspirin [
25] plays an important role for accurate estimate of transfer free energy.
To our knowledge, this is the first systematic study for membrane permeation of all 20 amino acids. Unlike the previous study of 17 side chain analogues [
17], both backbone and side chain atoms of amino acids are included in our simulations to better mimic the permeation of the whole amino acid. We found that hydrophobic and positive charged amino acids are easier to penetrate into the cell than polar and negatively charged amino acids. Correlation analysis between free energy cost of permeation and physico-chemical properties of amino acids suggest hydrophobicity as the main driving force of cell permeation.
3. Discussion
In this study, we have obtained the free energy profiles of all 20 natural amino acids. In general, polar amino acids have larger free energy barriers than nonpolar amino acids, and negatively charged amino acids are the most difficult to enter into the membrane. The results are consistent with previous experimental and computational studies. The obtained highly reproducible free energy profiles benefit from the use of both directional and torsion angles for conformational sampling. We observed conformational transitions for many amino acids from helical to sheet or sheet to helical regions in backbone torsion angles during cell permeating (
Figure 3 and
Figure S7).
The free energy cost of the ACE-Arg+-NH2 translocation from water to the center of 256 DPPC bilayer is 23 kJ/mol. This result is lower than 60.5 kJ/mol, the free energy of transfer of cationic arginine side chain into 64 DOPC bilayer [
17] and 74.4 kJ/mol into 72 DPPC bilayer [
21]. Our result is much closer to 29 kJ/mol, the free energy of transfer of ACE-Arg+-NH2 into 288 DMPC bilayer [
20]. However, the FES profiles are different. The FESs show a minimum of −37 kJ/mol for 256 DPPC bilayer at the interface between region II and III and −3.5 kJ/mol for 288 DMPC bilayer in the region of lipid headgroup. We also obtained the free energy cost of the ACE-Trp-NH2 translocation at about 18 kJ/mol, close to 20.9 kJ/mol, the free energy transfer of uncharged Trp into 40 DOPC bilayer [
18] with similar FES curves.
Experimentally, Chakrabarti [
11] summarized the order of amino acid permeability by different experiments to establish the order of amino acid permeability as Phe > Met > Leu > Ile; Leu > Ala; Gly > His. This is largely consistent with Phe > (Ile > Leu) > Met; Leu > Ala; Gly > His from our calculation on the free energy changes from water to the center. Naoi et al. [
39] obtained amino acid permeability according to an order by Leu > Phe > Trp > Met > Tyr and Val > Thr >Ser > Ala > Gly. This is largely consistent with Phe > Trp > Leu > Met > Tyr and Val > Ser~Thr > Gly > Ala from our calculation. In general, the order of Leu, Phe and Met is different in different experiments [
40]. The observed minor differences are likely due to the approximate nature of the membrane model system and its large difference from experimental membrane systems and experimental conditions.
To understand the mechanism of cell permeating, we performed the correlation analysis between calculated free energy profiles and physio-chemical properties of amino acids. We found that hydrophobicity plays the most important role in cell permeating. This indicates that for an non-natural amino acid, its cell permeability can be estimated from its measurable hydrophobicity. Another correlated scale is the STERIMOL length of the side chain. This suggests the intrinsic role of side chains in cell permutation in addition to hydrophobility.