Chiral Recognition of Dansyl Derivatives with an Amino Acid-Based Molecular Micelle: A Molecular Dynamics Investigation

In this study, the chiral separation mechanisms of Dansyl amino acids, including Dansyl-Leucine (Dans-Leu), Dansyl-Norleucine (Dans-Nor), Dansyl-Tryptophan (Dans-Trp) and Dansyl-Phenylalanine (Dans-Phe) binding to poly-sodium N-undecanoyl-(L)-Leucylvalinate, poly (SULV), were investigated using molecular dynamics simulations. Micellar electrokinetic chromatography (MEKC) has previously shown that when separating the enantiomers of these aforementioned Dansyl amino acids, the L-enantiomers bind stronger to poly (SULV) than the D-enantiomers. This study aims to investigate the molecular interactions that govern chiral recognition in these systems using computational methods. This study reveals that the computationally-calculated binding free energy values for Dansyl enantiomers binding to poly (SULV) are in agreement with the enantiomeric order produced in experimental MEKC studies. The L-enantiomers of Dans-Leu, Dans-Nor, Dans-Trp, and Dans-Phe binding to their preferred binding pockets in poly (SULV) yielded binding free energy values of −21.8938, −22.1763, −21.3329 and −13.3349 kJ∙mol, respectively. The D-enantiomers of Dans-Leu, Dans-Nor, Dans-Trp, and Dans-Phe binding to their preferred binding pockets in poly (SULV) yielded binding free energy values of −14.5811, −15.9457, −13.6408, and −12.0959 kJ∙mol, respectively. Furthermore, hydrogen bonding analyses were used to investigate and elucidate the molecular interactions that govern chiral recognition in these molecular systems.


Introduction
More than half of all pharmaceutical drugs currently in use are chiral, and the synthesis of these drugs often yields a racemic mixture, containing both enantiomers of the compound [1]. While the physical and chemical properties of the enantiomers are similar, they often produce very different pharmacological effects [1] [2] [3]. In many cases, one enantiomer may produce the desired medicinal effect, whereas the other may cause adverse health effects. For example, (+)-Ethambutol is primarily used to treat Tuberculosis, whereas its enantiomer, (−)-Ethambutol, causes blindness [4]. Due to the enantiomers of chiral drugs often producing vastly different effects, the United States Food and Drug Administration has mandated that each enantiomer of a chiral drug be tested for enantiomeric purity prior to being marketed [5]. Due to the enantiomers sharing nearly identical physical properties such as boiling point, density, mass and solubility, this often makes it difficult to separate enantiomers based on those aforementioned properties [6]. However, the enantiomers differ in stereoconfiguration, which allows them to interact differently with chiral separation mediums, thus allowing for enantioseparation [5].
Some commonly used chiral separation techniques are high-performance liquid chromatography (HPLC) and capillary electrophoresis (CE), which utilize chiral stationary and pseudostationary phases, respectively [7] [8]. In using chiral chromatographic techniques, if the two enantiomers of a chiral compound, the R-and S-enantiomer, have different binding affinities to the chiral separation medium, then enantiomeric separations can occur. While both aforementioned techniques are able to effectively separate enantiomers on the basis of their binding affinity to the chiral separation medium, CE has several advantages over HPLC. A much smaller sample size is required for CE as compared to HPLC and often, CE yields a higher number of theoretical plates, ultimately producing better enantiomeric resolution in a shorter period of time [9] [10]. In HPLC, the chiral recognition medium is covalently linked to a type of solid support material such as silicon beads [11]. Therefore, the user would need to purchase individual columns containing the desired stationary chiral recognition medium.
Consequently, if the user wanted to switch out the chiral separation medium, it would entail switching out the column(s). However, in CE, the chiral recognition medium can be exchanged quite easily with another chiral medium since it is part of the mobile phase, in which the chiral medium acts as the pseudostationary phase.
Chiral selectors such as cyclodextrins, polysaccharides, crown ethers and chiral micelles have proven to be effective chiral selectors in CE [7] [12] [13]. This research will focus on the latter, chiral micelles, more specifically, a class of chir-Open Journal of Physical Chemistry al micelles known as amino acid-based micelles. When micelles are used as the pseudostationary phase in CE, the technique is commonly known as Micellar Electrokinetic Chromatography (MEKC) [8] [10] [14] [15].
Amino acid-based micelles are composed of surfactant units, each containing a hydrocarbon chain; one end of the chain contains a terminal alkane, and the opposite side of the chain is connected to an amino acid head group. Amino acid-based surfactants have advantages over other chiral selectors in that amino acids are ubiquitous therefore deeming it cheaper than other options; these surfactants are also biodegradable and environmentally-friendly [16]. These surfactants have been extensively studied using single amino acid and dipeptide head groups, in which the dipeptide micellar systems have shown significant advantages in achieving enhanced enantiomeric resolution in MEKC compared to single amino acid based micellar systems [10]. This is due in part to the increased number of chiral centers on the head group, thus allowing for enhanced chiral selectivity. In addition, previous research reveals that the dipeptide amino acid-based micelles act as better chiral selectors in their polymerized forms [15] [17]. The polymerized surfactants are formed via free radical polymerization when the surfactant solution is subjected to gamma radiation at concentrations significantly above their critical micelle concentration (CMC), usually 5 -10 times the CMC. This gamma radiation causes the terminal alkenes to undergo free radical polymerization to form covalent bonds with neighboring surfactant units (see Figure 1(a)). These polymerized micelles are commonly referred to as amino acid-based molecular micelles (AABMMs) [17]. AABMMs have significant advantages over the traditional, non-polymerized micelle, including: AABMMs do not have a critical micelle concentration (CMC), reduced joule heating and most notably enhancing the enantiomeric resolution [15].
This manuscript focuses on one particular AABMM, poly-sodium N-undecanoyl-(L)-Leucylvalinate, or poly (SULV). Previous literature reports that when studying a wide-range of chiral compounds, enantiomeric separation was achieved for  than 75% of the compounds when using poly (SULV) as the pseudostationary phase in MEKC [14]. Chiral recognition with poly (SULV) has been examined using a variety of techniques. One of those techniques is Nuclear Overhauser Effect Spectroscopy Nuclear Magnetic Resonance (NOESY NMR), which has been used to study the primary site of interaction (s) of chiral analytes to molecular micelles [18]. Molecular dynamics (MD) simulations have also previously been used to investigate AABMMs, such as poly (SULV) to further investigate the factors that contribute to chiral recognition [5] [17]. Some of the chiral analytes previously studied with poly (SULV) include, beta blockers and binaphthyl compounds [5] [19]. This manuscript will focus on examining the binding of various Dansyl amino acids to the AABMM poly SULV.
Due to the intractable number of testing conditions and combinations that exist between: AABMMs, chiral analytes, counterions, surfactant concentration, and pH levels, it is important to develop a predictive model that is ultimately more cost-effective, efficient and requires less time than experimental methods. This investigation is part of a long-term project aimed at developing a Quantitative Structure-Enantioselective Retentions Relationship (QSERR) model [20] [21]. More specifically, the QSERR model will be developed to study and predict the best MEKC conditions for chiral selectivity using various AABMMs that contain at least one chiral center using various dipeptide combinations of the L-form of alanine, valine and leucine as well as the achiral amino acid glycine.
To build this predictive model, experimental data from Nuclear Magnetic Resonance (NMR) studies and MEKC will be combined with the insights gained from the MD simulation studies.
Here, we report an MD simulation study of Dansyl amino acids, including Dansyl-Leucine (Dans-Leu), Dansyl-Norleucine (Dans-Nor), Dansyl-Tryptophan (Dans-Trp) and Dansyl-Phenylalanine (Dans-Phe) binding to poly (SULV) to further the knowledge of a previous study that examined the same aforementioned systems with experimental MEKC and NOESY NMR [22]. The chemical structures of these compounds are shown in Figures 1(b)-(e). The Dansyl amino acid analytes examined in this study can be divided into two main groups: aliphatic (Dans-Leu and Dans-Nor), and aromatic (Dans-Phe and Dans-Trp).
This data is important to incorporate into the QSERR model in the future because it expands the depth and diversity of the analyte training set by introducing chiral analytes with varying characteristics; this will ultimately improve the predictions of the QSERR model. Combining experimental MEKC and NOESY NMR data with computational MD simulation studies will allow for the elucidation of these chiral separation mechanisms and helps to further the knowledge base that is necessary to develop a predictive QSERR database.

Experimental Details
Building the poly (SULV) micelle computationally Molecular modeling and MD simulation methods employed in this research Open Journal of Physical Chemistry project have been reported in detail in previous work from our group [5] [17] [18] [19] [22]. The methods for this research project are summarized as follows.
First, a monomer unit of N-undecanoyl-(L)-Leucylvalinate (ULV) molecule was built with an overall net charge of -1 within the graphical interface, Xleap, within the AMBER18 software package (see Figure 1(a)) [23]. The average structure of poly (SULV) was then calculated using the following procedure.
Step 1. The root mean squared deviation (RMSD) of the 480 ns MD simulation was determined. The RMSD is used to monitor the coordinate changes of a system over time in comparison to a reference set of coordinates. In this case, the reference structure was the initial frame. The RMSD values were calculated with Equation (1).
The number of atoms in the system, the mass of each atom, the coordinate vector for the target atom, the coordinate vector for the reference atom, and the total mass of the system are denoted as N, i m , i X , i Y and M, respectively [25] [26]. The RMSD protocol has previously been used in other studies from our group and is summarized as follows [5] [17] [18] [19] [27]. Throughout the 480.0 ns simulation, the poly (SULV) micelle structure will begin to change over time as it tries to settle into its energetically-favorable structure. The RMSD plot is composed of a Y-axis representing the RMSD value, and an X-axis representing Open Journal of Physical Chemistry the simulation time. The RMSD value will initially increase at a rapid rate at the beginning of the simulation then proceed to equilibrate and level off. The leveled-off region in the RMSD plot denotes which frames in the trajectory file contain representative structures of the equilibrated micelle.
Step 2. After determining which frames were associated with the equilibrated region, the average structure was determined from those frames, ultimately yielding an average, equilibrated structure of poly (SULV). This average structure was then used as the reference structure for the subsequent RMSD analysis on the 480 ns trajectory file.
Step 3. To obtain the final equilibrated poly (SULV) molecular micelle structure that was used for this study, the RMSD was calculated on all frames of the 480 ns trajectory file with the reference frame being the average theoretical structure obtained from Step 2. From here, the frame containing the lowest RMSD value is the representative poly (SULV) molecular micelle structure. The poly (SULV) molecular micelle structure was then stripped of its aqueous solvent consisting of water and sodium ions. The non-solvated, equilibrated structure of poly (SULV) was then used for the ligand docking analyses.

Ligand Docking
The Molecular Operating Environment (MOE) software package was used to identify the binding pockets on the representative poly (SULV) and dock ligand enantiomers into each of the identified pockets [28]. The MOE software contains a Site Finder module, which was used to identify the binding pockets of poly (SULV); Site Finder utilizes the alpha sphere method to identify molecular cavities and potential binding sites [19] [29]. Alpha spheres are placed within cavities where four receptor atoms sit at its spherical boundaries [30]. Non-polar regions with poor hydrogen bonding capabilities are represented by white spheres.
Regions where hydrogen bonding interactions are likely to occur are represented by red spheres [19]. Therefore, the hydrophobicity and hydrophilicity of a binding pocket can be quantified. Site Finder identified three binding pockets within poly (SULV). MD simulations were conducted on each enantiomer in all three binding pockets of poly (SULV) using AMBER18. The gaff and ff19SB force fields were employed in these simulations. The gaff force field was used to describe the chemical parameters of the surfactant tail constituent and each of the Dansyl enantiomers using the antechamber package within AMBER18 [32]. The ff19SB force field defines the chemical parameters of amino acids and was employed to provide proper parametrization of the amino acid head group of the surfactant [33]. Each MD simulation of the enantiomer to poly (SULV) complex was solvated with 20 sodium ions and solvated with TIP3P water molecules within a volume of a 10 Angstrom truncated octahedron. The MD simulation is summarized as follows. The first step in the MD simulation is initiated with a 50 ps minimization step. This was followed by a 20 ps warm-up step, which allowed the system time to increase to 300 K. The system was then allowed to equilibrate to a pressure of 1 atm with a time of 20 ps. The final production run for each MD simulation was conducted for 60 ns. Each simulation was monitored for anomalies with the Visual Molecular Dynamics (VMD) software to ensure that the simulation ran properly. This validation method was used to ensure that the enantiomer being investigated did not leave the pocket during the simulation or do anything else unexpected.

Binding Free Energy Analyses
The binding free energy values (kJ•mol −1 ) for each enantiomer: poly (SULV) molecular complex were calculated using post-trajectory analysis.
The binding free energy value of each enantiomer in the three pockets of poly (SULV) were used to calculate the percent occupancy, P i , which calculates where a given enantiomer is most likely to occupy based on its ΔG binding values in each of the binding pockets. The percent population was calculated with Equation (3) [5].

Hydrogen Bond Analyses
The CPPTRAJ utility in AMBER18 was used to perform hydrogen bond analyses on each of the 24 poly (SULV):enantiomer complexes [25]. Hydrogen bonding interactions occur when a hydrogen atom is covalently linked to heavy atoms, such as fluorine, oxygen or nitrogen, and that same hydrogen also shares intermolecular interactions with other nearby heavy atoms (usually fluorine, oxygen or nitrogen). The heavy atom that the hydrogen atom is covalently linked to is known as the donor atom, whereas the acceptor atom is the nearby atom that the hydrogen atom is having intermolecular interactions with. The CPPTRAJ utility is able to analyze each trajectory file and track the hydrogen bonds that are being broken and formed throughout the 60 ns simulation. It does so by following two main geometric criteria that are characteristic for hydrogen bonds: 1) the distance between the donor-to-acceptor heavy atoms must be within 3 Å of each other, and 2) the donor-hydrogen-acceptor angle cutoff must be ±30˚ [5].

Discussion
Previously reported data shows that when using poly (SULV) as the pseudosta- were not fully determined [22]. By investigating the interactions between the Dansyl amino acids and poly (SULV) via MD simulation studies, molecular-level insight will be able to provide information about the molecular interactions that govern chiral recognition in these molecular systems.
As mentioned previously, the MOE software package identified poly (SULV) as having three binding pockets, which can be found in Figure 2. Additionally, each pocket in relation to each other in poly (SULV) is shown in Figure 3. The binding free energy values and the occupancy percentages for each pocket with each of the Dansyl amino acid enantiomers are summarized in Table 1. The binding free energy data for each Dansyl amino acid will be summarized in the following sections.
Furthermore, previous studies also showed that experimental NOESY NMR data was inconclusive in determining which interactions governed chiral selectivity due to spectral overlap between the NOE intensities of the Dansyl amino acids and poly (SULV) [22]. Therefore, hydrogen bond analyses will be used to rationalize why the L-enantiomers of the Dansyl amino acids bind stronger to poly (SULV) than the D-enantiomers, while providing molecular-level insight  into the specific bonding interactions that govern chiral selectivity in these systems. The trajectory files for each molecular system will be analyzed for its intermolecular hydrogen bonds that are formed throughout the simulation. The hydrogen bonding analyses will provide detailed information about the acceptor atoms and donor atoms that are participating in hydrogen bonding [5] [18] [19].
Additionally, the highest number of consecutive frames that a hydrogen bond is maintained will be denoted as the 'max lifetime' for that specific hydrogen bond; this provides insight into the frequency of the hydrogen bond being formed [36].
The hydrogen bond occupancy represents the percentage of frames that a hydrogen bond is maintained throughout an entire simulation. Table 2 displays bonds that had greater than a 10% hydrogen bond occupancy throughout the simulations. All hydrogen bond occupancies less than 10% were considered negligible and therefore did not contribute greatly to chiral selectivity. In the following sections, the binding free energy values for each Dansyl amino acid will be rationalized using the data collected from hydrogen bonding analyses.   poly (SULV) as the pseudostationary phase.
As confirmed by the binding free energy values in Table 1  As confirmed by the binding free energy values in Table 1  As confirmed by the binding free energy values in Table 1   As confirmed by the binding free energy values in Table 1

Conclusion
The binding free energy values for the Dansyl amino acids examined in this study to poly (SULV) were all in agreement with the enantiomeric order determined by MEKC, in which the L-enantiomer of each Dansyl amino acid interacted stronger with poly (SULV) than the D-enantiomer. Spectral overlap in NOESY NMR spectra ultimately led to inconclusive results on the exact interactions that occurred between the Dansyl amino acids binding to poly (SULV). Open Journal of Physical Chemistry Therefore, MD simulations were able to provide molecular-level insight into the hydrogen bonding interactions that govern the chiral separation in these systems. In conclusion, the binding free energy and hydrogen bond analyses studies on poly (SULV) in conjunction with the enantiomers of Dans-Leu, Dans-Nor, Dans-Trp and Dans-Phe were able to provide insight into the molecular interactions that govern chiral selectivity. Furthermore, the computational hydrogen bond analyses supported the computational binding free energy values, which helped to rationalize the reasoning for why the L-enantiomers of each Dansyl amino acid had stronger binding interactions with the amino acid-based molecular micelle, poly (SULV), as opposed to its D-enantiomer. In conclusion, this computation investigation was in agreement with experimental MEKC enantiomer order data, and able to provide insight into the chiral separation mechanisms of these systems. This study also shows that when experimental NMR studies provide inconclusive results of binding interactions due to spectral overlap, MD simulations may be able to supplement information about the underlying chemical interactions.