Probing Fast Enantio-Recognition of Drugs with Multiple Chiral Centers by Electrospray-Tandem Mass Spectrometry and Its Mechanism

Abstract

Chiral drugs are very complex substances since individual enantiomers may differ in pharmacological and toxic effects, making it necessary to analyze enantiomers separately. In this study, we investigated the chiral differentiation of two ezetimibe enantiomers (i.e., SRS-EZM and RSR-EZM) and their mechanisms in complex with β-cyclodextrins (CDs) and metal ions as the auxiliary ligands. For this purpose, two complementary approaches have been employed: electrospray-tandem mass spectrometry (ESI-MS/MS) with collision induced dissociation (CID) and molecular modeling methods, including density functional theory (DFT) calculations and molecular dynamics (MD) simulations. The results showed a good agreement between experimental and theoretical data. It was demonstrated that SRS-EZM can be easily distinguished from RSR-EZM by applying CID in ESI-MS/MS. SRS-EZM is likely to form a more stable complex with β-CD and metal ions, and thus the [SRS-EZM]-Cu-[β-CD] cluster is more energetically difficult to separate from the SRS-EZM molecule compared with RSR-EZM. Such a difference may be attributed to the interactions between the drug molecule and the metal ion, as well as the cavity shape changes of the β-CDs upon complexation with molecular guests. Therefore, enantiomers in chiral drug can be recognized as ternary complexes of metal-analyte-β-CD by ESI-MS/MS with CID.

1. Introduction

Over the past few years, chiral analysis has attracted global attention because of the uses of chiral molecules in drug safety. Chiral molecules are essential in life processes, consisting of most biologically active compounds of living organisms [1]. Enantiomers of racemic drugs often show different behaviors in their biological activity, mechanisms, and toxicity. It is common that one enantiomer is active while other isomer is toxic in biological systems. Therefore, great emphasis should be put on the enantiomeric analysis for safe medication [2]. It was recommended that the study of enantioselective identity and stability ought to be performed before undertaking other evaluation works for drugs [2,3].

Along with the rising analytical requirements, tremendous efforts have been devoted to searching for a high-efficiency and low-expense approach. In recent years, various methods have been applied for chiral analysis, involving high-performance liquid chromatography (HPLC), gas chromatography (GC), supercritical fluid chromatography (SFC), thin-layer chromatography (TLC), and capillary electrophoresis (CE) [3,4]. However, these approaches share several common disadvantages, e.g., consuming too much time and needing plenty of samples. Mass spectrometry (MS) is a powerful tool for analysis owing to its strong sensitivity, high speed, and good molecular specificity [5,6,7]. Electrospray ionization ion trap multistage MS can provide abundant structural information for compound identification and useful stereochemical information derived from sterically controlled fragments. Thus, MS has gradually played an important role in chiral analysis. Several pure MS-based approaches have emerged by creating a chiral environment, among which the collision-induced dissociation (CID) of the diastereomeric complex method is widely used [8,9]. In this process, normally, metal-bound trimeric clusters [M^II(A)(ref)₂-H]⁺ that consist of metal ions (M^II), an analyte (A), and two chiral references (ref) are formed [9].When performing a CID experiment, the relative intensities of some characteristic fragment peaks in the mass spectra facilitate differentiation between stereoisomers, and thus the chiral discrimination of two enantiomers can be achieved by measuring the chiral recognition ratio (CR method) [10,11].

There are quantities of materials able to act as the chiral reference. During the past decades, supramolecular architectures and materials have attracted immense attention, which open the possibility of obtaining a large variety of aesthetically interesting structures and extend the applications in gas storage, sensors, separation, catalysis and others, mainly owing to their molecular recognition features. Cyclodextrins (CDs), composed of 6, 7, and 8 α-(1-4)-linked D-glucopyranose units and shaped like a torus, represent one of the most popular supramolecules, enjoying a rather low toxicity, high aqueous solubility, and binding capacity, which exhibit wide range of utilities at reasonable price [12]. β-CD is the most emblematic member of this family and has been widely used in industrial and research fields, such as drug delivery and chiral selectors. In the chiral analysis field, β-CD and its derivatives usually act as a chiral stationary phase in various chromatographic columns, as well as mobile phase additives [13,14]. However, it is reported many times that there is low enantioselective binding and efficiency when using β-CDs alone as the chiral selectors [15,16]. Considering the low cost of the native β-CDs and the premises for the proven formation of specific inclusion complexes with enantiomers, their use as chiral auxiliaries in the presence of metal ions was initiated. Recently, some β-CDs and metal ions complex systems have been reported to enhance enantiomers separations [17]. Although the metal ion can change the binding mode between analytes and β-CDs by affecting the size and shape of the cavity, the detailed mechanisms about a big ternary cluster (including β-CDs, metal ions and analytes) underlying the separation remain unknown. This results in the screening of the type and/or size of the cavities and ions, as well as their combinations and other relevant conditions still being based on trials instead of rational design, making the separation process tedious and expensive. For example, the influence of hydrogen bonding in the chiral recognition process is especially difficult to analyze due to the limitations of the experimental method [18]. Therefore, finding the binding mode and chiral recognition mechanism between the analyte and selector is of great significance for both fundamental studies and potential applications, like designing the chiral reference individually.

To understand its selective mechanism in molecular details, there are several available methods, including nuclear magnetic resonance (NMR) [19] and computational methods [11], applied to predict ligand selectivity in more studies. Compared with NMR, the theoretical calculation methods, including quantum mechanics (QM), molecular dynamics (MD), molecular docking, and quantitative structure activity relationships (QSARs), have been used to optimize all the systems in gas phase efficiently. These methods are acknowledged to be valuable tools for the molecular recognition mechanism correlation and owns different features. Density functional theory calculations (DFT) were used to perform a full geometry optimization from quantum mechanics aspects, while MD can provide the time-dependent trajectory of the host and guest molecules, revealing their binding mode from molecular mechanics aspects. The combination of the two methods is often used in studies. However, most computational calculations about β-CD are limited to small systems, and it is necessary to extend their application for more complicated clusters.

Our intention for this study is to propose a new chiral separation method by using β-CD/metal complex as the selector in MS and to find a “modular, robust molecular-modelling workflow” for how to analyze the formation of β-CD/metal and enantiomer drugs in general as well as revealing the molecular recognition mechanisms, which is a fundamental study for choosing and designing chiral selector. Ezetimibe (EZM), a potent cholesterol absorption inhibitor, has been applied to cure diseases related to the overabundance of cholesterol in individual or combined therapies. There are three asymmetric carbon atoms in EZM, thus producing eight different enantiomers. The desired one is designated as 3(S), 4(R)-1-(4- fluorophenyl)-[3-(4-fluorophenyl)-3(S)-hydroxypropyl]-(4-hydroxypenyl)-2-azetidinone (SRS-EZM), as shown in Figure 1, while the others are presented as a chiral impurity without pharmacological effects. HPLC is the common method to separate them with obvious disadvantages of low efficiency and high expense. It would therefore be valuable to develop a new technique capable of higher efficiency. We have reported that the entecavir enantiomers were discriminated using the tandem MS of metal coordination trimeric complexes and the MS method was compared with the HPLC method in the USP convention. The latter requires up to 11 min in gradient elution, while the MS method requires no more than 2 min [9]. In this study, we take EZM as an example for the effort of finding an effective way to analyze the enantiomers. CR method in MS was performed to achieve chiral separation and Cu is chosen to act with β-CD by trial. DFT and MD simulations were performed to understand their interaction details.

2. Materials and Methods

2.1. Materials and Reagents

SRS-EZM and its isomeric reference standard (RSR-EZM) were obtained from Zhejiang Huahai Pharmaceutical (Taizhou, China). Cupric sulfate (CuSO₄), Zinc chloride (ZnCl₂), magnesium chloride (MgCl₂), lead(II) sulfate (PbSO₄), Mercuric nitrate (Hg(NO₃)₂), Cobaltous chloride (CoCl₂), and β-CD were purchased from commercial source (analytical grade reagent). Methanol (MeOH) and Dimethylsulfoxide (DMSO) used for sample preparation were of HPLC grade and were purchased from Merck KGaA (Darmstadt, Germany).

2.2. Solution Preparation

Metal ions solution: Precisely weigh 1 mg of CuSO₄, ZnCl₂, MgCl₂, PbSO₄, Hg(NO₃)₂, CoCl₂ and dissolved in MeOH, and then diluted to 50 µg/mL with same solvent.

β-CD solution: Precisely weigh 10 mg. Then, dissolve in a small amount of DMSO (200 µL), before diluting with MeOH to 50 µg/mL.

SRS-EZM and RSR-EZM solution: Precisely weigh 1 mg of isomers separately. Then, dissolve in a small amount of DMSO (200 µL). Then, dilute with MeOH to 50 µg/mL.

Sample solution: Dilute metal ion solution to 12.5 µg/mL, β-CD solution and SRS-EZM or RSR-EZM solution 25 µg/mL separately. Then, mix 25 µg/mL β-CD and 12.5 µg/mL metal ion solution with 25 µg/mL SRS-EZM or RSR-EZM in ep tube with vortex for 1 min to form a ternary complex.

All above solutions were mixed well separately and stored at 4 °C before use.

2.3. ESI-Q-Trap Mass Spectrometry Analysis

Electronic spray ionization mass spectra were recorded on a LC/MS (amazon ETD Q-Trap) (Bruker 2011) and operated in positive ionization mode, with the spray voltage of −4 kV and auxiliary sheath gas flow speed is 5 L/min. All the samples were continuously infused at 5 µL/min via a 250 µL syringe. Applied voltages were −500 and −4500 V for the ion transfer capillary and the tube lens, respectively. The ion transfer capillary was held at 250 °C. The m/z ranges were set to 300–2000 m/z in profile mode and in the normal mass range during full scan experiments. Spectra were analyzed using the acquisition software XCalibur 2.0.7 (Thermo Scientific, San Jose, CA, USA) without smoothing and background subtraction. During MS/MS scans, collision-induced dissociation (CID) was performed with an activation Q of 0.25 and an activation time of 20 ms occurred in the linear ion trap analyzer (LTQ) and detection in the Orbitrap in centroid mode. The experiments were applied on Agilient Q-ToF 6550 later to ensure its accuracy.

2.4. Computational Details

2.4.1. DFT Study

Molecular modeling was carried out to further elaborate the complexation mechanism of EZM, Cu and β-CD according to the MS results. Given such a large complex, fractions of the system are calculated separately, namely [Cu^II(β-CD)]²⁺, [Cu^II(β-CD)(EZM)]²⁺, and [Cu^II(β-CD)(EZM)]²⁺. The structure of β-CD was obtained from the crystal structures (PDB code:1Z0N). It was optimized by DFT calculations employing the hybrid B3LYP functional and the 6-31G* basis set. Cu was then added, and the Los Alamos effective core potential basis set (LANL2DZ) was used. The geometries of complex SRS-Cu-[β-CD], RSR-Cu-[β-CD], 2SRS-Cu-[β-CD], and 2RSR-Cu-[β-CD] were optimized by the same method with [Cu^II(β-CD)]²⁺. Frequency calculations were performed to verify the true nature of minima. The minimal energies were estimated for each cluster. Basis superposition set errors (BSSEs) were also evaluated using the counterpoise corrections and added to the total energies.

2.4.2. MD Simulation

The stable binding model from DFT calculation was used as the starting structure of MD simulation. The universal force field (UFF) was employed to model the metal complex system in the subsequent MD simulations. Here, the equilibrium bond lengths and bond angles as well as the force constants for dihedral potentials were further redefined so that the molecular mechanical force field could well reproduce the quantum chemically predicted geometries. MD simulations were carried out with the Materials Studio package. For each complex, all molecules were put in vacuum and simulated under constant-NPT conditions (T = 500 K, which is equal with the temperature in the MS). Snapshots were extracted from each simulation trajectory with an equal interval of 600 ps. After two stages of minimization, 1 ns MD simulations were performed.

3. Results and Discussion

3.1. ESI-MS Analysis

To achieve chiral recognition, metal-analyte-reference interactions play important roles in achieving the selectivity of the enantiomers due to their different stereochemical effects. In the study, β-CD and metal ions act as the reference respectively. As we know, single charged and double charged cluster ions are both possible to investigate and it has been shown to be a feasible alternative that the trimetric cluster ion [M^II(ref)(A)₂-H]⁺ can take the place of [M^II(A)₂(ref)-H]⁺. However, the single charged metal–analyte–reference cluster [M^II(β-CD)(EZM)₂-H]⁺ as well as the [M^II(β-CD)₂(EZM)-H]²⁺ were not detected in MS¹spectra. Therefore, [M^II(β-CD)(EZM)₂-H]²⁺ was selected as the parent ion.

3.2. The Effect of Cu²⁺, Zn²⁺, Mg²⁺, Pb²⁺, Hg²⁺, Co²⁺ on Recognizing EZM Enantiomers

The effect of Cu²⁺, Zn²⁺, Mg²⁺, Pb²⁺, Hg²⁺, and Co²⁺ on distinguishing EZM enantiomers is different based on their interaction with ligands. With the addition of Pb²⁺ and Hg²⁺, there is no difference in MS/MS spectrogram and no corresponding [Pb(EZM)₂(β-CD)-H]²⁺ and [Hg(EZM)₂(β-CD)-H]²⁺ ternary complexes found. This demonstrates that Pb²⁺ and Hg²⁺ have no benefits with β-CD for increasing the distinguishing effect on EZM enantiomers.

In addition, Zn²⁺, Mg²⁺, and Co²⁺ have shown similar distinguishing effect to Cu²⁺. They all show one abundance difference of fragment ion in corresponding [M^II(EZM)(β-CD)-H]²⁺ (M = Cu²⁺, Zn²⁺, Mg²⁺, Co²⁺) ternary complexes (SRS- EZM and RSR- EZM) MS/MS spectrogram. By comparing the abundance of this special fragment ion, we can discriminate SRS- EZM and RSR- EZM. Among that, Cu²⁺ corresponds to fragment ion peak [EZM-H₂O+H]⁺ (m/z = 392.14) and Zn²⁺, Mg²⁺, Pb²⁺ correspond to fragment ion peak [M^II (EZM)₂(β-CD)-H]²⁺ (M = Cu, Co, Zn, Mg, A. [Cu(EZM)₂(β-CD)]²⁺: m/z = 1007.7987, B. [Co(EZM)₂(β-CD)]²⁺: m/z = 1005.8000, C. [Zn(EZM)₂(β-CD)]²⁺: m/z = 1008.2984, D. [Mg(EZM)₂(β-CD)]²⁺: m/z = 988.3264). By comparation of the difference of fragment ion ratio of abundance between corresponding SRS-EZM and RSR-EZM ternary complexes, Cu²⁺ shows the best distinguishing effect (Figure 2).

Taken together, the effects of Mg²⁺, Zn²⁺, Pb²⁺, Hg²⁺, and Co²⁺ combined with β-CD as ligands on differentiating SRS-EZM and RSR-EZM were studied. It turned out that different metal ions produced different fragments from the ternary complexes, which can be used to recognize SRS-EZM and RSR-EZM. Besides, the efficiency of recognition is also different. After analyzing the characteristics of metal ions, we finally present a possible conjecture that the loss of electron capability and the size of the radius are important factors that affect the discrimination function of metal ions as ligands.

3.3. Cu²⁺ Distinguishing EZM Enantiomers

3.3.1. The Effect of CID Collision Energy to Ternary Complexes

To observe the difference of fragment ions, we increased the CID collision energy progressively. As shown in Figure 3, when collision energy increased from 10 V to 20 V, the fragment ions abundance of ternary complexes was gradually increasing. When collision energy was set at 16 V, the fragment ion [EZM-H₂O+H]⁺ (m/z = 392.14) of RSR-EZM reached maximum and the corresponding abundance fragment ion of SRS-EZM only about 50%.

3.3.2. The Binding Form of Ternary Complexes

CID of [Cu^II(β-CD)(EZM)₂-H]²⁺ cluster ion typically yields several different fragment ions such as [Cu^II(β-CD)(EZM)-H]⁺, [Cu^II(β-CD)-H]⁺ and the neutral EZM, and there was obvious difference between EZM enantiomers in MS² spectra. As shown in Figure 3, the m/z of precursor ion is 1007.79, which yields four major fragment ions: two dimers corresponding to the loss of only one neutral EZM and two EZM (single charged at m/z 1605.44 and 1196.29, respectively); one protonated ion of EZM at m/z 410.16 and one product ion at m/z 392.15 via the loss of H₂O from EZM. It can be seen that the EZM product ion’s intensity of RSR is higher than SRS, which indicates that RSR tends to lose one neutral EZM more easily than SRS. In other words, [Cu^II(β-CD)(SRS-EZM)₂-H]²⁺ should have a stronger affinity than [Cu^II(β-CD)(RSR-EZM)₂-H]²⁺. Therefore, when the same collision energy is performed on the two clusters, [Cu^II(β-CD)(RSR-EZM)₂-H]²⁺ will produce a higher abundance of EZM fragment and the two EZM enantiomers can be differentiated.

In addition, a ternary complex was detected as [Cu(EZM)₂(β-CD)-H]²⁺ whether in ion trap MS or in Q-TOF MS. After collision in the MS/MS, the [Cu(EZM)₂(β-CD)-H]²⁺ produced several fragments, among which the [EZM-H₂O+H]⁺ (m/z = 392.15) shows a distinct difference between SRS-EZM and RSR-EZM (Figure 3,

Table 1. The difference of the fragment ions abundance of [Cu(EZM)₂(β-CD)]²⁺ ternary complexes at different collision energy.

	[Cu(RSR-EZM)2(β-CD)]2+	[Cu(SRS-EZM)2(β-CD)]2+	Difference Value
CE10	12.4%	8.2%	4.2%
CE12	21.3%	13.1%	8.2%
CE14	43.2%	23.6%	19.6%
CE16	97.3%	46.7%	50.6%
CE18	99.4%	99.2%	0.2%
CE20	99.7%	99.6%	0.1%

). The recognition of SRS-EZM and RSR-EZM was based on the collision-induced dissociation of the diastereomeric complex method.

3.4. Computational Calculations

The DFT and MD simulations were performed to investigate the differentiation mechanisms of EZM enantiomers in the gas phase of MS. Molecular structures of the complex were built up stepwisely to illustrate the formation of the clusters.

3.4.1. The [Cu^II(β-CD)]²⁺ Complex

For geometric optimization of the [Cu^II(β-CD)]²⁺ complex through DFT calculation, the copper ion was put in four different initial places, namely the center of the large rim edge, the edge of the large rim, the center of the small rim, and the edge of the small rim. The optimization can only proceed when the copper ion lies in the small rim. Figure 4 shows the top and side views of the initial and optimized structures when the ion placed at the center or the edge of the small rim. DFT calculation indicates that the copper ion was stabilized by three electron-donating oxygens of the 6-hydroxyl group localized on the primary rim edge while keeping the ion inside the cavity structure. The optimized Cu-O bond lengths range from 2.04 to 2.40 Å. Furthermore, it is also shown that the structure of β-CD can be affected in various degrees by adding the copper ion. When the copper ion is in the center, the conformation of β-CD become more distorted after optimization (Figure 4A,B), induced by the strong attraction from the ion. In contrast, when the copper ion is initially placed close to the edge of the rim (Figure 4C,D), the β-CD maintains a more “natural” state with less compensation of the conformational change induced by the Cu coordination. DFT calculations suggest the optimized structure shown in Figure 4D as a more energy-favorable state, with the energy 13.3 kcal/mol lower than the optimized structure shown in Figure 4B. The more negative the energy is, the more stably the complex forms. This is reasonable because the distorted β-CD is energy-unfavorable, which could be compensated by the coordination of the ion to a certain extent for the energy-favorable state of the whole system. To satisfy the β-CD conformation requirement, copper ions can hardly be stabilized when the distance between them is too far, and that is the reason why the optimization failed when placing the ion at the large-rim side. In brief, the copper ion plays a bridge role to connect the β-CD and EZM in the complex because of its coordination ability.

In previous studies, it is generally recognized that the large rim of β-CD containing secondary hydroxyl groups is the main binding site for guest inclusion [9,20,21], but one should also pay attention to the contribution from the small rim. It has been demonstrated that benzoic acid could insert into the carboxylic group at the small rim [22]. Another work also reveals that cations are likely to form a coordination bond at the small rim with β-CD when considering the distance between opposite oxygen atoms within the small rim is two-fold shorter when compared with the large rim. The author studied the complexes involving β-CD with Na⁺, Mg²⁺, Al³⁺, Cu⁺, and Zn²⁺ using the DFT computational method and showed clearly that all these metal ions are located in the small rim in the optimized structure [23]. It is also reported many times that the size and shape of the entrance and cavity of β-CD will make proper changes for guest binding [24,25]. Li et al. have determined that the methylation of the β-CD would affect the size of the cavity and the shape of the entrance [26].

Similarly, the copper ion also exhibits a significant effect on the rim parameters. The rim diameters of the β-CD in the optimized complex are measured according to the scheme shown in Figure 5A. For the atoms on the β-CD, the positions of the O₂/O₃/C₆ atoms are determined by the sugar ring while others are more flexible due to the rotations around C₂−O₂, C₃−O₃, and C₅−C₆ bonds (Figure 5B). Therefore, the former three atoms can be used to reveal the rim size. It is known that the entrance size of the small rim of β-CD is defined by the atom-to-atom distances (R values) between C₆ on opposite rings. The secondary hydroxyl groups (O₂ and O₃ oxygens) form the bigger rim. The maximum and minimum values are labeled Rs (max), Rs (min), R_L(max) and R_L(min)respectively. For the small rim, seven C₆···C₆′ distances (C₆ in glucose unit 1 between C₆ in glucose unit 4,5,C₆ in glucose unit 2 between C₆ in glucose unit 5,6, C₆ in glucose unit 3 between C₆ in glucose unit 6,7 and C₆ in glucose unit 4 between C₆ in glucose unit 7) are measured. For the large rim, a set of three cross-ring distances are measured. For O₃ in glucose unit 1, the three measured distances are O₃ in glucose unit 1 between O₃ in glucose unit 4 and 5, together with the O₂ in glucose unit 4; for O₂ in glucose 1, they are O₂ in glucose unit 1 between O₂ in glucose unit 4 and 5, together with the O₃ in glucose unit 5. The maximum and minimum of all cross-ring distances calculated are recorded respectively. They reflect how much the upper/lower rim is opened up and their difference value(ΔR) give us indication to the shape of the entrance. A bigger ΔR refers to a more elliptical entrance while a smaller one corresponds to a rounder shape [27,28].

The large rim and the small rim of CD are both considered to be a possible binding site.

Table 2. Various Structural Parameters Calculated from Optimized Structures (Å).

	RL (Max)	RL (Min)	△RL	Rs (Max)	Rs (Min)	△Rs
β-CD	11.11	10.25	0.86	13.03	8.62	4.41
β-CD+Cu	10.09	9.82	0.27	11.40	10.22	1.18
β-CD+Cu+SRS	13.91	12.44	1.47	10.94	8.78	2.61
β-CD+Cu+RSR	13.02	12.27	0.75	14.18	5.83	8.35
β-CD+Cu+2SRS	12.41	12.27	0.14	12.50	8.79	3.71
β-CD+Cu+2RSR	13.32	12.90	0.42	14.23	6.27	7.96

shows that, when the copper ion is added, the large rim diameters of β-CD change from 11.11 Å, 10.25 Å to 10.09 Å, 9.82 Å respectively, which is in accordance with the formation of Cu-O bond. The Rs(min) increases correspondingly, indicating that the small rim becomes bigger. Furthermore, the ΔR of the β-CD/Cu complex become smaller in the existence of copper ion, which indicates that the β-CD entrance becomes rounder to accommodate small-molecule drugs more easily.

In conclusion, we can see that the copper ion can form coordination bonds with oxygen atoms, changing the rim size of the β-CD. It has been reported that metal cations can assist host–guest interactions [27]. Cai introduced Fe²⁺ and Mg²⁺ to enhance the detection of β-CD/toluene complexes. Using electrospray mass spectroscopy, the existence of ternary complexes (dication/β-CD/toluene) has been proved [28]. Therefore, it can be indicated that all these changes seem to make preparation for the binding other molecules.

3.4.2. The Complex of [Cu^II(β-CD)(EZM)]²⁺

With a basic understanding of how the copper ion acts with the β-CD, we explore the combination mode of the drug molecule and [Cu^II(β-CD)]²⁺ system. One EZM molecule was randomly placed in the vicinity of the above-mentioned optimized [Cu^II(β-CD)]²⁺ structure to form [Cu^II(β-CD)(EZM)]²⁺ complex. The structures of the optimized [Cu^II(β-CD)(EZM)]²⁺ complex are shown in Figure 6 and the corresponding binding free energy is calculated according to the following reactions.

$$\mathrm{EZM} + {\left[{\mathrm{Cu}}^{\mathrm{II}}\right(\beta -CD\left)\right]}^{2+} \stackrel{\Delta G_1}{⟶} {\left[{\mathrm{Cu}}^{\mathrm{II}}\right(\beta -CD\left)\right(\mathrm{EZM}\left)\right]}^{2+}$$

(R1)

$$\mathrm{EZM} + {\left[{\mathrm{Cu}}^{\mathrm{II}}\right(\beta -CD\left)\right(\mathrm{EZM}\left)\right]}^{2+} \stackrel{\Delta G_2}{⟶} {\left[{\mathrm{Cu}}^{\mathrm{II}}\right(\beta -CD\left){\left(\mathrm{EZM}\right)}_2\right]}^{2+}$$

(R2)

For the RSR-EZM case, as shown in Figure 6A, the complex exhibits the lowest energy when the drug molecule was mainly stabilized by the edge of the large rim of β-CD. The Cu-O distances were 1.92 Å, 1.92 Å, and 2.74 Å, all of which were derived from the interaction between the oxygens and the small rim of β-CD, leaving no space for the RSR-EZM molecule to enter from the small rim and coordinate with Cu. Meanwhile, the R_L (max) changed from 13.03 Å to 14.17 Å, providing other binding sites for the drug molecule where a H-bond forms between the oxygen on the drug and a hydrogen on the rim with a length of 1.75 Å.

Interestingly, there is another stabilizing pattern which contains two kinds of binding mode: the coordinate bond and the H-bond (Figure 6B). The RSR-EZM partially enters the inner cavity of the β-CD and forms a Cu-O coordinate bond with a length of 2.05 Å. The rest of the Cu-O bond lengths were 2.29 Å and 2.12 Å, respectively. The H-bond length between the hydrogen atom nearby and the oxygen atom on the β-CD was 1.66 Å. Another H-bond forms at the large rim side, similar to the situation in Figure 6A with the same bond length. The binding energy for this pattern (Figure 6B) is 7.42 kcal/mol higher than the former optimized cluster (Figure 6A).

In the SRS-EZM case, we observed a similar binding mode with that shown in Figure 6B; that is, at least one part of the drug molecule entered the cavity and coordinated with the copper ion. The optimized structure of SRS-EZM allows the drug molecule completely penetrating to the cavity of β-CD (Figure 6C). The phenyl ring moved to a position where it can act with the copper ion, possibly through π-π conjugation. In this complex, there are two oxygen atoms on the β-CD which interacting with the Cu(II) ion. With 2.10 and 2.04 Å Cu-O bond lengths, making the whole complex more stable. Besides, the OH group near this phenyl also form two H-bonds with the β-CD. The bond lengths are 1.67 Å and 1.65 Å, respectively. In this optimized conformation, the coordinate bond between the copper ion and the drug molecule contributes significantly.

We also observe another energy-favorable form (Figure 6D) that is same as the one shown in Figure 6B, where the EZM partially enters the inner cavity of the β-CD and forms a Cu-O coordinate bond with a length of 2.12 Å. The other part which possesses a hydroxyl group acts with the H atom on the rim formed a H-bond with a length of 1.76 Å. The Cu-O distances are 2.12 Å, 2.13 Å, and 2.17 Å respectively. The energy is 2.02 kcal/mol higher than the optimized structure (Figure 6C), but 2.33 kcal/mol lower than the similar form for RSR-EZM (Figure 6B).

As it is described previously, both the RSR-EZM and SRS-EZM are observed to be stabilized with one OH group coordinating with the Cu and the other forming H-bond with the oxygen on the β-CD (Figure 6B,D). However, the SRS-EZM tend to pass through the β-CD cavity and stretch to other side (Figure 6C) while the RSR-EZM is more likely to drop down from the cavity (Figure 6A). This may result from the difference of their conformation. Taking Figure 6C as an example, assuming that the SRS-EZM are replaced with RSR-EZM, the hydroxyphenyl and the 3-(4-fluorophenyl) in the drug molecule would both tilt to the β-CD and even insert into its skeleton (Figure 6E). Compared with the SRS-EZM, the RSR-EZM cannot form coordinate bonds with the copper ion. Besides, the angle between the two 4-fluorophenyls enlarges to almost 180°, causing the steric geometry conformation of the drug molecule to become greatly distorted. During our optimization, when the RSR-EZM is put completely penetrating the whole β-CD cavity manually for the initial structure, the optimized procedure always turns out to be a failure, which is corresponding to the assumption.

3.4.3. The Complex of [Cu^II(β-CD)(EZM)₂]²⁺

Due to their stereo-isomeric difference, RSR-EZM is likely to bind at the entrance of the β-CD while SRS-EZM tends to penetrate the cavity and coordinate with Cu. However, when the second EZM molecule enters the system, it remains challenging to clarify what the most stable configurations are and how they interact with each other. Based on the result of the MS/MS fragments, we can know that the second EZM molecule more easily broken up than the first one, suggesting that the interaction force between EZM and [Cu^II(β-CD)(EZM)]²⁺ is weaker compared to coordination. Normally, the H-bond is weaker compared with coordination bond. Therefore, the second EZM molecule has a strong possibility to bind with the first EZM molecule or the β-CD instead of Cu.

The optimized structures of [Cu^II(β-CD)(EZM)₂]²⁺ are shown in Figure 7. Unsurprisingly, H-bond becomes the main binding form for the RSR-EZM(Figure 7A). Both of the two drug molecules connected the β-CD with H-bond, whose lengths are 1.89 Å and 1.86 Å, respectively. To get a more stable configuration, there is another H-bond forming to link the two RSR-EZMs with a distance of 1.75 Å. Copper ion coordinated with the oxygens on the β-CD, the bond lengths of which are 1.92 Å, 1.92 Å and 2.72 Å, respectively, similar to the [Cu^II(β-CD)(EZM)]²⁺(Figure 6A). However, compared to the complex containing only one drug molecule, the entrance size and shape has been enlarged. It is shown in Table 2 that the corresponding distances changed from 13.02 Å, 12.27 Å, 14.18 Å, 5.38 Å to 13.32 Å, 12.9 Å, 14.23 Å, 6.27 Å respectively. A larger and rounder entrance is more suitable for the two drug molecules to bind with.

Interestingly, SRS-EZM can stabilize in two forms (Figure 7B,C). The optimized structures illustrate that one of the drug molecules would coordinate with the Cu(II) ion and the other would connect the β-CD with the H-bond. In Figure 7B, the Cu(II) ion form a coordinate covalent bond with one oxygen atom on the SRS-EZM with a bond length of 2.03 Å. The remaining coordinate bonds involved the oxygens on the β-CD with a longer bond length of 2.15 and 2.16 Å. Except that, the two SRS-EZMs both interacted with the β-CD through four H-bonds. The SRS-EZM not only coordinated with copper ion, but also formed two H-bonds with the length of 1.74 Å and 1.76 Å, respectively. The other drug molecule binds at the other side with 1.76 Å and 1.77 Å H-bonds. The two drugs have no interactions between each other. In general, the SRS-EZM ternary complex owns a lower binding energy than RSR-EZM (~14.46 kcal/mol lower), indicating that the former is more stable, which is consistent with the result of the MS spectra.

The MD simulation was also performed to investigate the relative binding strength of the two EZM with the [Cu^II(β-CD)]²⁺ complex. The most stable structures of [Cu^II(β-CD)(EZM)₂]²⁺ derived from the DFT results were taken as the initial model. The snapshots taken at different simulation time from the trajectory are shown in Figure 8. For both RSR and SRS complex, the drug molecules can stay close to the β-CD when the temperature gradually increased up to 500 K within 500 ps, indicating a reasonable initial structure suggested by the DFT calculation. Afterwards, the RSR complex started to break up at 570 ps and one of the drug molecules moved away from the β-CD, while the other still stayed inside. In contrast, the SRS complex shows obvious stability within the 1000 ps run, which is consistent with the previous results, revealing that the [Cu^II(β-CD)(SRS-EZM)²]²⁺ is more stable than [Cu^II(β-CD)(RSR-EZM)²]²⁺.

MD simulations can provide auxiliary information for clarifying the enantiomers separation mechanism. As is shown in Figure 8, the Cu(II) ion in RSR complex stayed at the center of the small rim, while in SRS complex, the copper ion coordinated with the oxygen atom on the outside drug molecule all the time. Different configurations lead to different binding mode, which are related to their different stability. Compared with SRS-EZM, RSR-EZM cannot penetrate the cavity to coordinate with Cu(II) ion. H-bond is the only connection between RSR-EZM and β-CD. Thus, [Cu^II(β-CD)(RSR-EZM)²]²⁺ squint towards to lose one RSR-EZM molecule. To further prove such a difference between the two enantiomers, three parallel simulations were performed for each drug molecule by placing them at different initial places. The distances between the EZM and [Cu^II(β-CD)]²⁺ are calculated from these MD trajectories, and the normalized distributions of the distances calculated from the three runs are shown in Figure 9. It is obvious that compared with RSR-EZM, SRS-EZM always keeps closer around the β-CD, starting from various initial structures. It further proves the better stability of the RSR-complex.

To deeply understand the role of the drug enantiomers in the ternary complex and the application scope of our method, we studied the recognition of captopril and ibuprofen enantiomers by MS using Cu²⁺ and β-CD as ligand but failed. Captopril, β-CD and metal ions cannot form stable ternary complex because the sulfhydryl group of captopril is oxidated easily in the presence of metal ions. Ibuprofen, β-CD and metal ions cannot form ternary complex but can be recognized by D-glucose and metal ion complex, possibly due to the large opening size of β-CD relative to the ibuprofen molecule. Therefore, there are some limitations for MS with β-CD and the metal ions complex method for the recognition of enantiomers, such as suitable ligand, enantiomer stability, and molecular size.

4. Conclusions

In conclusion, electrospray-tandem mass spectrometry and molecular modeling were used in our study to probe the chiral analysis mechanism of RSR- and SRS-EZM. Collision-induced dissociation of the ternary complexes produces stable charge reduced intermediates, mainly the losses of the EZM from the ternary complexes. Their branching ratios depend on the relative bond strength between the EZM and [Cu^II(β-CD)]²⁺. There is a distinct difference in the EZM peak between SRS and RSR enantiomers, indicating their binding mode is different. By applying DFT calculation, it is shown that copper ion plays an essential part in this separation process. The smaller rim hydroxyl oxygen atoms formed bonds with copper ion in all structures. Furthermore, copper ion can not only coordinate with drug molecule, but also change the rim size and shape to provide proper binding sites for the drug, showing the so-called induced-fit binding effect. DFT theoretical study and MD stimulations revealed that SRS-EZM is able to enter the cavity of β-CD and forms a strong coordinate covalent bond with the copper ion while RSR-EZM achieves its stability only depending on the H-bonds. Therefore, the [Cu^II(β-CD)(RSR-EZM)²]²⁺ complex owns a higher free energy and is more easily broken up when subject to an external energy. In this study, we successfully illustrated the chiral analysis mechanism of such a complicated ternary complex by combining practical experiment and theoretical modeling, which can be an alternative analytical method for other chiral molecules.