Update on chiral recognition mechanisms in separation science

Abstract

The stereospecific analysis of chiral molecules is an important issue in many scientific fields. In separation sciences, this is achieved via the formation of transient diastereomeric complexes between a chiral selector and the selectand enantiomers driven by molecular interactions including electrostatic, ion‐dipole, dipole‐dipole, van der Waals or π‐π interactions as well as hydrogen or halogen bonds depending on the nature of selector and selectand. Nuclear magnetic resonance spectroscopy and molecular modeling methods are currently the most frequently applied techniques to understand the selector‐selectand interactions at a molecular level and to draw conclusions on the chiral separation mechanism. The present short review summarizes some of the recent achievements for the understanding of the chiral recognition of the most important chiral selectors combining separation techniques with molecular modeling and/or spectroscopic techniques dating between 2020 and early 2024. The selectors include polysaccharide derivatives, cyclodextrins, macrocyclic glycopeptides, proteins, donor‐acceptor type selectors, ion‐exchangers, crown ethers, and molecular micelles. The application of chiral ionic liquids and chiral deep eutectic solvents, as well as further selectors, are also briefly addressed. A compilation of all published literature on chiral selectors has not been attempted.

List of Abbreviations

2,3‐DM‐β‐CD

heptakis(2,3‐di‐O‐methyl)‐β‐CD

2,6‐DM‐β‐CD

heptakis(2,6‐di‐O‐methyl)‐β‐CD

2‐M‐β‐CD

heptakis(2‐O‐methyl)‐β‐CD

3‐M‐β‐CD

heptakis(3‐O‐methyl)‐β‐CD

6‐M‐β‐CD

heptakis(6‐O‐methyl)‐β‐CD

AGP

α₁‐acid glycoprotein

BGE

background electrolyte

CCS

collisional cross‐section

CD

cyclodextrin

CE

capillary electrophoresis

CEC

capillary electrochromatography

CIL

chiral ionic liquid

COF

covalent organic framework

CSP

chiral stationary phase

DCV

daclatasvir

DES

deep eutectic solvent

DFT

density functional theory

Dns

dansyl

ECD

electronic circular dichroism

HDA‐β‐CD

heptakis(2,3‐di‐O‐acetyl)‐β‐CD

HPLC

high‐performance liquid chromatography

IL

ionic liquid

IM

ion mobility

MD

molecular dynamics

MIP

molecularly imprinted polymer

MOC‐Val

methoxycarbonyl‐Val

MOF

metal‐organic framework

NMR

nuclear magnetic resonance

NOE

nuclear Overhauser effect

poly‐SULV

poly‐sodium N‐undecanoyl‐(l)‐leucylvalinate

ROESY

rotating‐frame Overhauser enhancement spectroscopy

S‐DABN

(S)‐(–)−1,10‐binaphthyl‐2,20‐diamine

SFC

sub/supercritical fluid chromatography

TM‐β‐CD

heptakis(2,3,6‐tri‐O‐methyl)‐β‐CD

1. INTRODUCTION

“What can more resemble my hand or my ear, and be more equal in all points, than its image in a mirror? And yet, I cannot put such a hand as is seen in the mirror in the place of its original.” As described by the philosopher Immanuel Kant, in 1783, in his discourse on metaphysics [1], a chiral object and its mirror image, although looking alike, are nonetheless incongruent. The epistemological analysis of Kant also applies to nature with regard to the interaction of chiral substances with chiral targets which makes the differentiation of enantiomers a fundamental natural phenomenon. Consequently, chiral molecules play important parts in many aspects of life sciences, medicine, synthetic chemistry, food chemistry, as well as many other fields [2]. In separation sciences, enantioseparations are primarily achieved by chromatographic and electrophoretic techniques. Enantioseparations occur in a chiral environment with a chiral selector either attached to a solid support (typically in chromatography) or added to the background electrolyte (BGE) (in capillary electrophoresis [CE]). Such enantioseparations are based on the formation of transient diastereomeric complexes between selectors and analytes in thermodynamic equilibria. Despite many efforts, the mechanisms underlying these reactions between selector and selectand are not completely understood to date. As a consequence, additional techniques including nuclear magnetic resonance (NMR) spectroscopy and molecular modeling and molecular dynamics (MD) simulation approaches have frequently been applied in combination with separation techniques. Although limited to soluble selectors, NMR, especially techniques such as nuclear Overhauser effect (NOE) spectroscopy and rotating‐frame Overhauser enhancement spectroscopy (ROESY) provide conclusions about the spatial proximity of atoms or functional groups of selector and guest molecules [3, 4, 5, 6]. Much less applied spectroscopic techniques that have been utilized include ultraviolet spectroscopy, fluorimetry, Fourier transform and attenuated total reflectance IR spectroscopy, or electronic and vibrational circular dichroism. Furthermore, molecular modeling methods have been applied to obtain information about binding thermodynamics and the selector‐selectand complex structures [6, 7, 8, 9, 10, 11].

The present short review summarizes recent data on the understanding of chiral recognition mechanisms at the molecular level in analytical separation sciences based on studies combining chromatographic or electrophoretic enantioseparations with molecular modeling and/or NMR spectroscopy with a focus on the literature published between 2020 and early 2024. Because of the large number of compounds that have been evaluated as chiral selectors or selectands, respectively, a review on the topic cannot be comprehensive. Therefore, only frequently used types of selectors are addressed including polysaccharides, cyclodextrins (CDs), macrocyclic antibiotics, proteins, crown ethers, donor‐acceptor type selectors, ligand‐exchangers, ion‐exchanger selectors and micelles, but some data on seldom used selectors such as metal‐organic frameworks (MOFs) and covalent organic frameworks (COFs) will be presented. Furthermore, chiral ionic liquids (CILs) and deep eutectic solvents (DESs) will be briefly addressed. Previous studies on the elucidation of chiral recognition mechanisms in separation sciences were compiled in a monograph [12], book chapters [13, 14, 15, 16] or review papers such as [17, 18, 19, 20, 21, 22, 23, 24, 25]. The application of NMR spectroscopy for this purpose has been summarized in [3, 4, 5, 6] and reviews on molecular modeling approaches for separation sciences can be found in [6, 7, 8, 9, 10, 11]. Within the reviewed period of time, the progress in the development of chiral stationary phases (CSPs) for high‐performance liquid chromatography (HPLC) [26, 27, 28] and gas chromatography [29] was provided and an overview of chiral mobile phase additives in chromatography was published [30].

2. INTERMOLECULAR INTERACTIONS

The International Union of Pure and Applied Chemistry defines the term “chiral recognition” as “attraction between molecules through noncovalent interactions that exhibit complementarity only between partners with specific chirality” and the term “molecular recognition” as “attraction between specific molecules through noncovalent interactions that often exhibit electrostatic and stereochemical complementarity between the partners” [31]. In separation science, discrimination between stereoisomers is accomplished via the formation of transient complexes between a chiral selector and the analyte stereoisomers. This is driven by diverse intermolecular interactions including ionic interactions, ion‐dipole or dipole‐dipole interactions, π‐π interactions, van der Waals interactions, and hydrogen bonds [6, 32, 33]. Strong, long‐range interactions such as ionic (Coulombic) interactions may be primarily involved in the initial, non‐stereoselective binding of analyte enantiomers to a selector because they are formed in the case of both enantiomers, whereas short‐range directional interactions such as hydrogen bonds or π‐π interactions contribute to stereoselective binding [21]. In addition, steric factors resulting from the spatial arrangement of the binding cavity or cleft of the selector may support chiral recognition. In recent years it has been shown especially for polysaccharide chiral selectors that σ‐holes and π‐holes assisted in the recognition of analytes. A σ‐hole bond is a noncovalent interaction between a covalently‐bound atom such as a halogen (or another atom of groups IV–VII) and a negative site, that is, a lone electron pair of a Lewis base or an anion. It involves a region of positive electrostatic potential, termed a σ‐hole, on the extension of one of the covalent bonds to the atom having electrophilic properties. π‐Holes are electron‐deficient regions often observed on polarized double bonds and π‐acidic aromatics and are able to interact with nucleophiles [34]. In general, the strength of σ‐hole bonds increases as the electronegativity and the polarizability of the electrophilic atom decreases and increases, respectively. Thus, the strength of this type of bond increases in the direction of lighter to heavier atoms in a certain group of the periodic table [35] Furthermore, in both σ‐ and π‐hole bonds the strength of the interaction may be enhanced by increasing the electron‐withdrawing ability of the substituents directly bonded to or in the neighboring of the hole region. Positive σ‐ and π‐holes interact in a highly directional manner with negative sites such as the lone electron pairs of Lewis bases. The contribution of σ‐holes and π‐holes to the interaction between polysaccharide chiral selectors and halogenated solutes has been studied in detail by Peluso et al. [36, 37, 38]. For a detailed discussion of non‐covalent interactions and their role in enantioseparations as well as multidisciplinary approaches to their understanding see the comprehensive review by Peluso and Chankvetadze [6].

In the case of barrel‐ or cup‐shaped selectors such as CDs or calixarenes, the expulsion of so‐called high‐energy water molecules from the cavities plays a role in complex formation [39, 40]. For example, a CD cavity is occupied by a number of water molecules, which depends on the size of the given cavity. Due to the fact that these water molecules can form only a limited number of hydrogen bonds and, consequently, not a stable hydrogen‐bond network they are “unorganized”. Thus, there is a strong enthalpic driving force for the inclusion of complex formation with guest molecules, which will expel the water molecules into the bulk phase where they can form more hydrogen bonds. For a detailed discussion of this effect see [39, 40].

Thermodynamic parameters are often applied in chromatography in an attempt to understand the retention mechanism. However, as correctly pointed out by the group of Felinger, conclusions drawn from van ’t Hoff plots should be considered with care in order to avoid misinterpretations [41, 42]. This fact was also pointed out earlier by other groups, for example, Asnin and Stepanova [43], and is primarily due to the fact that two different adsorption sites are present on the surface of the CSP, an enantioselective binding site (the chiral selector) and a non‐enantioselective binding site (the achiral solid support). Moreover, further parameters like the length of a column, the particle size, or the flow rate can affect van ’t Hoff plots and, consequently, the derived enthalpy and entropy values. Nonetheless, with an awareness that the numbers obtained from analytical chromatographic data are not “true” but rather “estimated” values, they may be useful for the explanation of some phenomena observed for a given CSP and experimental conditions such as, for example, the temperature‐dependent reversal of the enantiomer elution order observed for ketoprofen, naproxen or ibuprofen on polysaccharide CSPs [44, 45, 46] or the elution profiles of halogenated 4,4′‐bipyridine derivatives [47] or planar chiral ferrocenes bearing iodoethynyl substituents [48]. Thermodynamic data derived from chromatography for the analysis of entropic or enthalpic contributions in the case of enantioseparations has also been applied in the case of other types of selectors, for example, macrocyclic antibiotics [49, 50] or ion‐exchange type selectors [51, 52, 53].

3. CHIRAL SELECTORS

In principle, conclusions on the chiral recognition mechanism of a selector can be obtained by two approaches. The first utilizes a (large) number of analytes with structural modifications in order to determine the structural requirements of a certain group of solutes for a successful enantioseparation. The second way includes spectroscopic and/or molecular modeling techniques to rationalize an experimentally observed enantioseparation. From the many spectroscopic techniques, NMR including various NOE techniques such as ROESY are most often applied because they yield structural information of the proximity of atoms of (soluble) selectors and selectands and can be often collected under conditions similar to the experimental separation conditions [3, 4]. Unfortunately, NMR does not necessarily provide differences in the signals of the enantiomers in the diastereomeric selector‐selectand complexes so that the origin of the chiral recognition is not obvious. Probably more often, molecular docking and MD simulations of selector selectand interactions have been applied in recent years [6, 7, 11] because of the relative ease of obtaining such data due to the progress of computing technologies. However, such data should be critically evaluated as addressed below and also in references [54, 55]. Molecular modeling is often performed with the selector molecule not considering the binding to the solid support in HPLC. This can affect the accessibility of solutes to the binding sites of the selectors. In the case of CDs, especially the modeling of structures complexes involving randomly substituted CDs may be misleading. Randomly substituted CDs are a mixture of a large number of differently substituted positional isomers, which all contribute to a chiral separation. Consequently, modeling of the structure of a single given isomer with a defined substitution pattern does not adequately represent the separation system so conclusions on the complex structures are not meaningful and have to be considered critically.

A successful enantioseparation is still associated with trial and error in selecting the appropriate chiral selector out of a group of selectors and also largely depends on the experience of the researcher. Reliable prediction of the separation of a solute by a selector by computational techniques has always been the “dream” of separation scientists and attempts are ongoing to provide such information. Thormann evaluated the impact of the complexation constant and the mobility of the diastereomeric complexes on the separation of enantiomers by a neutral selector in capillary electrokinetic chromatography and capillary isotachophoresis to provide in‐sight into electrophoretic processes [56]. Recently, a chemical structure‐based machine learning approach based on the 3D molecular conformation of solutes has been developed to predict the HPLC enantioseparation on 18 chiral columns including polysaccharide, macrocyclic antibiotic, CD, donor‐acceptor, and protein CSPs [57]. The algorithm, named 3DMolCSP, also correctly predicted the enantiomer elution order on four selected columns. Thus, the newly developed program could serve as a valuable tool for selecting suitable CSP significantly facilitating successful enantioseparations in chiral chromatography in the future.

3.1. Polysaccharides

Amylose and cellulose derivatives have been established as the most frequently used commercial chiral selectors in LC. The polysaccharides are linear helical polymers composed of d‐glucopyranose units linked via β(1,4) (cellulose) or α(1,4) (amylose) glycosidic bonds. The hydroxy (OH) groups of the glucose molecules are derivatized with benzoate or, more frequently, with phenylcarbamate moieties which feature methyl and/or chlorine substituents in various positions on the aromatic ring. The benzoate or phenylcarbamate residues are oriented in such a way that chiral helical grooves are formed with the polar groups positioned inside the grooves near the carbohydrate backbone, while the hydrophobic phenyl moieties are located outside. Selector‐selectand complexes between analytes and polysaccharide selectors appear to be primarily mediated via hydrogen bonds as well as π‐π interactions and van der Waals forces. In addition, π‐hole as well as σ‐hole interactions including halogen and chalcogen bonds were identified to contribute to enantioselective interactions between solute and polysaccharide‐carbamate‐based selectors [36, 37, 38]. Especially polarizable iodine substituents showed a remarkable ability to interact as halogen bond (electrophilic) donors with the carbonyl group of cellulose tris(3,5‐dimethylphenylcarbamate) and amylose tris(3,5‐dimethylphenylcarbamate) as acceptor. Furthermore, the carbamate linkage allows some flexibility of the aromatic rings for enhancing π‐π interactions and van der Waals forces upon binding of the solute. Finally, the mobile phase composition may modulate the recognition process. For summaries of general features of polysaccharide‐based chiral selectors see, for example [14–21, 23]. Molecular modeling approaches have been summarized recently [58, 59, 60, 61] as have applications of polysaccharide CSPs to the enantioseparations [62, 63] or the preparation of polysaccharide selectors [63, 64].

Depending on the derivatization, the cellulose or amylose CSPs differ in size and shape of the chiral grooves [19, 65]. This has recently been confirmed by Sechi et al. [66], who observed different chiral cavities in amylose carbamate CSPs (Figure 1). While amylose tris(3,5‐dimethylphenylcarbamate) featured relatively large and “flat”, cup‐shaped chiral cavities, amylose tris(4‐methylphenylcarbamate) displayed smaller open‐shaped cavities and amylose tris(2,5‐dimethylphenylcarbamate) hindered cavities as derived from MD simulations of 3/4 left‐handed helical phenylcarbamate‐based nonamers of amylose with n‐hexane/2‐propanol 90:10 (v/v) as virtual solvent. Apart from the form and size of the cavities, the pendant groups also resulted in different electron density surfaces of the chiral selectors (Figure 1). The modeling data also explained the high versatility of 3,5‐dimethylphenylcarbamate‐substituted polysaccharides as chiral selectors in chromatographic enantioseparations due to the cup‐shaped cavity.

FIGURE 1.

Typical shapes of the chiral cavities of amylose‐based carbamate chiral selectors shown as line models (A–C) and electron density surfaces (D‐F) with the methyl groups colored in orange. 3,5‐dimethylphenylcarbamate (A, D), 4‐methylphenylcarbamate (B, E), and 2,5‐dimethylphenylcarbamate (C, F). (Reproduced with permission [66], © by Elsevier 2023).

Peluso and Chankvetadze proposed a computational approach based on conformational and electrostatic potential analysis for analyte and selector for the elucidation of the enantiorecognition mechanism using the enantiomers of 2‐(benzylsulfinyl)benzamide and cellulose‐based chiral selectors [57]. High enantioselectivity had been observed for the compound on a cellulose tris(3,5‐dimethylphenylcarbamate) CSP. First, the low‐energy conformers of the target compound as well as structural analogs and their electron density isosurfaces were calculated and subsequently correlated to the chromatographic data. A significant correlation between chromatographic k values and the isosurfaces regarding the sulfoxide and/or carbonyl group of the analyte was observed. The exceptional selectivity of 2‐(benzylsulfinyl)benzamide on the cellulose tris(3,5‐dimethylphenylcarbamate) CSP using 2‐propanol as eluent was explained by two distinct conformational forms of the enantiomers. The first eluted (R)‐enantiomer showed less conformational freedom because of constriction caused by an intramolecular hydrogen bond, which reduced the ability to form intramolecular hydrogen bonds with the selector. In contrast, the (S)‐enantiomer did not exhibit conformational constraints enabling interactions with the selector. The authors also studied the effect of chlorine substituents on the phenylcarbamate ring, which significantly affected the capability of the carbamate moiety to act as a hydrogen bond donor and acceptor. Furthermore, the enantioseparation capability of 17 methylated and/or chlorinated cellulose phenylcarbamates toward 2‐(benzylsulfinyl)benzamide was estimated from the electron density isosurfaces of the phenylcarbamate side chains of the selector [67]. An exclusively theoretical study used density functional theory (DFT) calculations to elucidate the mechanisms underlying the chromatographic enantioseparation of the β‐blocker drugs atenolol and carvedilol [68]. Hydrogen bonding and π‐π stacking interactions were identified as the main driving forces.

Wang and colleagues modeled enantiomeric separations by MD simulations as a dynamic interfacial process to illustrate the interaction processes between the selector and selectand occurring during the chromatographic separation [69, 70]. The chiral selector amylose tris(3,5‐dimethylphenylcarbamate) was docked onto a silica rod, simulating a CSP with the selector coated onto silica gel solid support. Four helical strands of 18mers of the polysaccharide were required to cover the considered silica rod surface area. This arrangement allowed to consider also the role of the solid support and allowed to study the interaction of enantiomers with more than one polysaccharide helix. Different parallel and antiparallel arrangements of the four strands were considered in the computational evaluations. Considering different metrics for hydrogen bonds, it was concluded that the HPLC enantiomer elution sequence under normal‐phase and polar organic mode conditions was best represented by the S/R ratio of the computed values of the averaged maximum lifetime or the overall average lifetime of the hydrogen bonds formed between the enantiomers and the selector. The model racemates interacted in different ways during the movement along the “column” modeled from four amylose selectors on a silica rod. Within a single strand of the amylose tris(3,5‐dimethylphenylcarbamate) selector, interactions with solute occurred in various ways via π‐π interactions and hydrogen bonds. For some molecules, for example, naringenin the authors also noted interactions of the molecule with two adjacent amylose strands [62]. In this study, dimerization of naringenin was also observed, but the authors stressed that they used a high concentration of the analyte during their modeling procedure to speed up the computational process so that a dimerization appears to be unlikely to take place under dilution conditions such as the chromatographic process.

Recent studies combining HPLC enantioseparations on polysaccharide CSPs with computational modeling techniques have been summarized in Table 1. In contrast to an earlier summary [17], an increasing number of publications studied cellulose‐based selectors [66, 71–76] compared to amylose‐based selectors [69, 70, 77–79]. Most often the enantioseparations were achieved under normal‐phase conditions [70–74, 76–79], but reversed‐phase elution [75, 78] and polar organic mode [66, 69] were also applied. The group of Carotti combined circular dichroism with molecular modeling to establish the enantiomer elution order of two tetracyclic quinoline derivatives composed of four stereoisomers when pure enantiomeric standards are not available [71]. The stereoisomers of the quinolines were separated on a cellulose tris(3,5‐dimethylphenylcarbamate) CSP using n‐hexane/2‐propanol (99:1, v/v) as mobile phase. Fractions of the enantiomers were collected, and the electronic circular dichroism (ECD) spectra were measured and compared to the theoretical spectra obtained by ab initio time‐dependent DFT simulations. This enabled conclusions on the enantiomer elution order, which was further substantiated by molecular docking simulations as the docking score for the individual enantiomers was in perfect agreement with the elution order deduced from ECD.

TABLE 1.

Examples of combinations of high‐performance liquid chromatography (HPLC) enantioseparations and molecular modeling for the elucidation of the recognition mechanism of polysaccharide‐based chiral selectors.

Selector a	Analyte(s)	HPLC mode	Modeling technique	Ref.
Cellulose tris(3,5‐dimethylphenylcarbamate)	Quinolines	Normal phase	Molecular docking	[71]
Cellulose tris(3‐chloro‐4‐methylphenylcarbamate)	β‐Blocker drugs	Normal phase	MD simulations	[72]
Cellulose tris(3,5‐dimethylphenylcarbamate)	Flavanone glycosides	Normal phase	Molecular docking	[73]
Cellulose tris(3,5‐dimethylphenylcarbamate)	Ofloxacin and flumequine	Normal phase	Molecular docking	[74]
Cellulose tris(3,5‐dimethylphenylcarbamate)	Antihistamine drugs	Reversed‐phase	Molecular docking	[75]
Cellulose tris(4‐methylbenzoate)	Psychoactive drugs	Normal phase	Molecular docking	[76]
Amylose tris(3,5‐dimethylphenylcarbamate)	Decursinol and derivatives	Normal phase	Molecular docking	[77]
Amylose tris(3‐chloro‐5‐methylphenylcarbamate)	Ibuprofen, naproxen, and ketoprofen	Reversed‐phase	Molecular docking	[78]
Amylose tris(3‐chloro‐5‐methylphenylcarbamate)	Furanocoumarins and dihydroflavones	Normal phase	Molecular docking	[79]
Cellulose tris(3,5‐dimethylphenylcarbamate), cellulose tris(3‐chloro‐4‐methylphenylcarbamate), cellulose tris(4‐chloro‐3‐methylphenylcarbamate), cellulose tris(3,5‐dichlorophenylcarbamate), amylose tris(3,5‐dimethylphenylcarbamate), amylose tris(3‐chloro‐5‐methylphenylcarbamate), and amylose tris(5‐chloro‐3‐methylphenylcarbamate)	Bipyridines	Normal phase	MD simulations and DFT calculations	[37, 47, 80, 81]
Cellulose tris (3,5‐dimethylphenylcarbamate), cellulose tris (3,5‐dichlorophenylcarbamate), amylose tris (3,5‐ dimethylphenylcarbamate), and amylose tris (3‐chloro‐5‐methylphenylcarbamate)	Ferrocenes		MD simulations and DFT calculations	[48, 82, 83]
Cellulose tris(3,5‐dimethylphenylcarbamate) and amylose tris(3,5‐dimethylphenylcarbamate)	3,3′‐Dibromo‐5,5′‐bis‐ferrocenylethynyl‐4,4′‐bipyridine, and 3,3′‐dibromo‐5,5′‐bis(2‐phenylethynyl)−4,4′‐bipyridine	Normal phase, reversed‐phase, and polar organic mode	DFT calculations	[66]

^a Only the selector included in modeling studies is listed here. In some cases, additional enantioseparations of the analytes on other polysaccharide CSPs are reported in the references.

Abbreviations: DFT, density functional theory; MD, molecular dynamics.

As stated above, a series of detailed studies for the enantioseparation of axially chiral halogenated bipyridines were performed by the groups of Peluso and Chankvetadze pointing out the involvement of σ‐ and π‐holes for chiral recognition via halogen and chalcogen bonds on coated or immobilized cellulose‐ and amylose‐based CSPs under primarily normal phase elution conditions [37, 47, 80, 81]. Van der Waals interactions provided the dominant contribution for the overall binding of the analytes to the chiral selectors, but hydrogen bonds and halogen or chalcogen bonds played a pivotal role in the enantiodiscrimination of the bipyridines depending on the CSP as well as the type of (halogen) substituent affecting the electrostatic potential calculated for the respective analytes. The derived interaction energies between the selector and selectand enantiomers corresponded to the enantiomer elution order observed in HPLC experiments. Similarly, the separation of the enantiomers of planar chiral ferrocenes featuring halogen and/or iodoethynyl substituents on cellulose and amylose CSPs was analyzed [48, 82, 83]. MD simulations of selected analytes and CSPs confirmed the penetration of the molecules into the cavities of the polysaccharide selectors. As in the case of the halogenated bipyridines, halogen bonds significantly contributed to the chiral recognition. A snapshot of the MD simulation of the (R)‐enantiomer of a ferrocenyl compound with amylose tris(3,5‐dimethylphenylcarbamate) is shown in Figure 2 confirming the penetration of the compound into the groove of the selector [83]. During the 100 ns simulation, the solute remained in the cavity, which was formed by six aromatic moieties of the selector. Interestingly, a hydrogen bond established between the π‐ethynyl electron cloud and the amide hydrogen of the selector played a role in the binding of the (R)‐enantiomer (Figure 2B).

FIGURE 2.

Structure of the chiral ferrocenyl compound and representative snapshot of molecular dynamics trajectories (100 ns) of the (R)‐enantiomer with amylose tris(3,5‐dimethylphenylcarbamate). (A) Electron density surface; green: aromatic ring, red: C = O, blue: N‐H, gray: phenyl ring (Ph), cyclopentadienyl rings (Cp) and C≡C group of the ferrocenyl compound. (B) Tube model of the complex; orange: Ph, Cp and C≡C group of the ferrocenyl compound, magenta: iodine, green: aromatic rings of amylose tris(3,5‐dimethylphenylcarbamate). (Reproduced with permission [83], © by Wiley 2022).

As a “combination” of both structures, the enantioseparation of 3,3‐dibromo‐5,5‐bis‐ferrocenylethynyl‐4,4‐bipyridine bearing two ferrocenylethynyl units linked to an axially chiral core was performed on amylose and cellulose chiral selectors and compared to the analogous structure with two phenyl groups in place of the ferrocenyl moieties [48]. Significantly larger separation factors were observed for the ferrocenyl compound compared to the phenyl compound. Despite its spacious structure, the P‐enantiomer of the bis‐ferrocenyl analyte showed a high affinity for the more compact amylose‐based selector. MD simulations indicated a higher ability of the ferrocene to penetrate the groove of the amylose‐based polymer compared to the analogous benzene compound. The process was mainly driven by van der Waals interactions.

Apart from HPLC, polysaccharide CSPs are also frequently applied for enantioseparations in sub/supercritical fluid chromatography (SFC). Accordingly, molecular modeling techniques were also applied to illustrate the selector‐selectand interactions in this chromatographic technique. The SFC enantioseparations of the β‐blocker drugs atenolol, metoprolol, and propranolol were studied on amylose tris(3‐chloro‐5‐methylphenylcarbamate) as chiral CSP [84]. Molecular docking revealed higher negative binding energies for the (R)‐enantiomers of the drugs compared to the (S)‐enantiomers, which correlated with the enantiomer elution order. Hydrogen bonds, van der Waals, π‐π, and dipole‐dipole interactions were identified as driving forces of the enantiodiscrimination. Interestingly, the interaction of the atenolol enantiomers with the polysaccharide selector almost exclusively occurred via hydrogen bonds, while the other compounds mainly established hydrophobic interactions in the complexation process. The four stereoisomers of luliconazole were separated by SFC using amylose tris(3‐chloro‐5‐methylphenylcarbamate) and amylose tris[(S)‐α‐methylbenzylcarbamate] as chiral selectors [85]. Only the latter separated all four stereoisomers under the applied experimental conditions. Differences in the binding energies and their values were in accordance with the stereoisomer elution order. Molecular docking revealed the participation of hydrogen bonds and π‐π interactions in the complexation process. However, the absolute number of interactions did not reflect the elution order per se, emphasizing that the strength and/or lifetime of interaction is more important for the enantioseparation process than the absolute number of interactions.

Chitosan is a linear polysaccharide composed of d‐glucosamine (2‐amino‐2‐deoxy‐d‐glucopyranose) moieties linked via β(1,4) glycosidic bonds. The amino and hydroxy groups can be derivatized by attaching substituted phenylcarbamates and phenylurea residues. Chen et al. prepared a series of chitosan‐based CSPs with a methylcyclohexyl urea substituent at the 2‐amino group and phenylcarbamate moieties with methyl and chloro substituents in various positions at the hydroxy groups [86]. Methyl substituents bearing chitosan phenylcarbamates separated more enantiomers as compared to the respective chlorinated derivatives, but the latter displayed higher α values in case of successful enantioresolutions. Hydrogen bonds were considered to be substantial in the separation process. Another study also evaluated the effect of regioselective modifications of 2, 3, and 6 positions of chitosan [87]. In this case, the 2‐amino and 3‐hydroxy groups contained 3,5‐dimethylphenyl substituents, while the 6‐hydroxy group displayed a 4‐chlorophenyl or a 3,5‐dichlorophenyl moiety. Compared to the analogous chitosan CSP containing 3,5‐dimethylphenyl groups in all three positions, the regioselective derivatives showed in some cases superior enantioseparations for the set of test racemates. Different interaction sites between the chiral selectors and model compounds were observed depending on the substitution of the 6‐hydroxy group substituent (4‐chlorophenyl vs. 3,5‐dichlorophenyl). The authors concluded that the combination of electron‐withdrawing and electron‐donating substituents in the regioselective chitosan derivatives was not generally beneficial for chiral separations except for selected cases.

3.2. Cyclodextrins

CDs are cyclic oligosaccharides formed by α(1,4) linked d‐glucopyranose units. The commercially used native CDs differ in the number of d‐glucopyranose molecules, α‐CD is built from six, β‐CD from seven, and γ‐CD from eight d‐glucopyranose units. CDs display a hollow toroidal structure with a lipophilic cavity and a hydrophilic outside surface. The secondary 2‐ and 3‐hydroxy groups are located at the wider rim, while the narrower rim features the primary 6‐hydroxy groups. The OH groups can be derivatized resulting in a large number of charged or uncharged derivatives. Suitable CD derivatives can be immobilized to solid supports for use in LC. Due to their complexation ability, CDs have found multiple applications including in analytical chromatographic and especially electrophoretic separation techniques [54, 88–90]. As CDs are typically water soluble, they can be studied in solution under comparable conditions by spectroscopic methods such as NMR spectroscopy as well as separation techniques such as CE, allowing direct correlations between complex structures and enantioseparations. Essentially the same applies to molecular modeling methods which can be combined with separation studies to rationalize the chiral recognition mechanism. Recent examples of the analysis of the structures of CD‐analyte complexes by NMR and/or molecular modeling combined with separation techniques have been compiled in Table 2. Further examples have been summarized, for example, in [17, 55, 60, 91–93]. Applications of MD simulations for CD complexes have also been reviewed [17, 54, 55, 94].

TABLE 2.

Examples of combinations of liquid phase separation techniques with molecular modeling and/or NMR spectroscopy for the elucidation of the chiral recognition mechanism by CDs.

CD a	Analyte(s)	Separation technique	Spectroscopic technique	Modeling technique	Ref.
Subetadex	N ‐Ethylbuphedrone	CE	NMR	–	[98]
α‐CD, 18‐crown‐6	Aromatic amino acids	CE	ESI‐MS	Molecular docking	[95]
β‐CD and methylated β‐CDs	Daclatasvir	CE	NMR and MS	–	[99]
TM‐β‐CD	Mexiletine	CE	–	Molecular docking	[100]
Sulfated β‐CD	Ofloxacin, gemifloxacin, lomefloxacin, and gatifloxacin	CE	–	Molecular docking	[101]
CM‐β‐CD	Acidic and basic drugs	CEC	NMR	Molecular docking	[102]
HP‐β‐CD	Mandelic acid and derivatives	HPLC	–	MD simulations	[103]
Phenylcarbamate‐β‐CD	Thalidomide	HPLC	–	Molecular docking	[104]
3,5‐Dichlorophenylcarbamate‐β‐CD	Proton pump inhibitors	HPLC	–	Molecular docking	[105]
β‐CD, HDAS‐β‐CD, and HMDS‐β‐CD	Tetrahydrozoline	CE	NMR	Molecular docking	[96]
α‐CD, β‐CD, γ‐CD, and HDA‐β‐CD	Terbutaline	CE	NMR	–	[97]
α‐CD and MDAI‐β‐CD	Kynurenine	CE	–	Molecular docking	[106]
CE‐β‐CD and CE‐γ‐CD	Licarbazepine	CE	NMR	MD simulations	[107]
CM‐β‐CD	Citalopram	CE	–	Molecular docking	[108]
β‐CD and HP‐β‐CD	Ketoconazole, miconazole	CE	NMR	Molecular docking	[109]
Cationic β‐CDs	Organic acids, pesticides, NSAIDs, and fluoroquinolones	CE	NMR	–	[110]
SBE‐β‐CD	5‐Hexyl‐2‐methyl‐3,4‐dihydro‐2H‐pyrrole	CE	–	Molecular docking	[111]
β‐CD, Succ‐β‐CD, SBE‐β‐CD, and subetadex	Mephedrone and butylone	CE	NMR	–	[112]

^a Only the CD for which complex structures have been concluded from spectroscopic and/or modeling data are indicated. In some cases, additional CDs have been evaluated for enantioseparations of the analytes in the references.

Abbreviations: CE, capillary electrophoresis; CEC, capillary electrochromatography; CE‐β‐CD, carboxyethyl‐β‐cyclodextrin; CE‐γ‐CD, carboxyethyl‐γ‐cyclodextrin; CM‐β‐CD, carboxymethyl‐β‐CD; ESI‐MS, electrospray ionization‐mass spectrometry; HDA‐β‐CD, heptakis(2,3‐di‐O‐acetyl)‐β‐CD; HDAS‐β‐CD, heptakis(2,3‐di‐O‐acetyl‐6‐O‐sulfo)‐β‐CD; HMDS‐β‐CD, heptakis(2‐O‐methyl‐3,6‐di‐O‐sulfo)‐β‐CD; HP‐β‐CD, hydroxypropyl‐β‐CD; MDAI‐β‐CD, mono‐6A‐deoxy‐6‐(1‐allylimidazolium)‐β‐CD; NMR, nuclear magnetic resonance; SBE‐β‐CD, sulfobutylether‐β‐CD; subetadex, heptakis[6‐S‐(2‐carboxyethyl)−6‐thio]‐β‐CD; Succ‐β‐CD, succinyl‐β‐CD; TM‐β‐CD, heptakis(2,3,6‐tri‐O‐methyl)‐β‐CD; NSAID, non‐steroidal anti‐inflammatory drug.

The complexation of solutes by CDs has been investigated extensively so that a fairly good understanding of the complex structures exists [17, 54, 55]. Because of the orientation of the d‐glucosyl moieties of CDs, the cavity has a hydrophobic surface formed by the glycosidic oxygen bonds and C‐H bonds. Therefore, the inclusion of a guest molecule is mediated via hydrophobic interactions within the cavity as well as hydrogen bonds with polar groups on the CD rims. Moreover, van der Waals interactions can contribute to complex formation. Inside the cavity, CDs display high electron density and Lewis base character as the non‐bonding electron pairs of the glycosidic oxygens are directed toward the interior of the cavity. Thus, dipole‐dipole interactions can be established between CDs and solutes as well as π‐π interactions in the case of aromatic analytes. Moreover, charged CDs can engage in ionic interactions, and, finally, a conformational change of the CD can occur upon solute complexation (“induced fit”) that can support the complexation process. In addition to the interactions between CDs and solutes, the displacement of the so‐called (“unorganized”) high‐energy water molecules from the interior of the cavity to the bulk liquid phase can be a driving force for solute complexation [39, 40] as briefly addressed above. As a consequence, the complexation of a solute by a CD is commonly enthalpy‐driven in contrast to the classical entropy‐driven complexation processes based on hydrophobic interactions.

Typically, guest molecules form inclusion complexes with CDs in a way that lipophilic moieties of the molecules enter the cavity from the wider secondary or the narrower primary side. However, “superficial” complexes at one of the CD rims as well as external complexes have also been observed in the case of successful enantioseparations. A recent study reported the CE enantioseparations of terbutaline using α‐CD, β‐CD, γ‐CD, and heptakis(2,3‐di‐O‐acetyl)‐β‐CD (HDA‐β‐CD) as chiral selectors [97]. In the case of α‐CD and HDA‐β‐CD, the (S)‐enantiomer of the drug migrated first, while the (R)‐enantiomer migrated first in the presence of β‐CD and γ‐CD. NMR studies of the analyte‐CD complexes revealed inclusion complexes in the case of β‐CD and HDA‐β‐CD although the exact structure of the terbutaline HDA‐β‐CD complex could not be unambiguously concluded from the NMR data. Based on NOE responses, either the aromatic ring of the drug entered the cavity from the wider secondary side or, alternatively, complexation occurred via the tert.‐butyl moiety of the analyte also enters from the wider secondary side. On the other hand, the solute could not penetrate the small cavity of α‐CD and interacted only with the outside of the CD. In the case of γ‐CD, terbutaline appeared to lie “on top” of the secondary side of the CD torus with hardly any penetration into the cavity [97].

Although not specifically addressed by the authors, external complexes seemed to be formed also between kynurenine and α‐CD as well as with the ternary complex with α‐CD and mono‐6A‐deoxy‐6‐(1‐allylimidazolium)‐β‐CD [106]. An external complex was also described for the interaction between thalidomide and per‐phenylcarbamate‐β‐CD [104]. An HPLC column with this selector covalently attached is commercially available. However, molecular modeling clearly indicated that the CD derivative does not form the toroidal structure typical for CDs because of the space‐filling phenylcarbamate substituents, which block the cavity. Consequently, the interaction between the enantiomers of thalidomide and the selector was exclusively mediated via the phenylcarbamate moieties. π‐π interactions between the phenyl moieties of the selector and thalidomide were established as were hydrogen bonds between the carbamate functional groups and the oxygen atoms of the glutarimide ring of the drug. (R)‐thalidomide formed two hydrogen bonds versus one in the case of (S)‐thalidomide in agreement with the enantiomer elution order observed in HPLC under polar organic mode elution conditions. Interactions with the phenylcarbamate substituents were also observed in the case of proton pump inhibitors and 2,5‐dichlorphenylcarbamate‐β‐CD [105]. Nonetheless, the majority of the studies listed in Table 2 used native CDs or CD derivatives with small substituents and, consequently, reported the formation of inclusion complexes stabilized by lipophilic interactions and hydrogen bonds.

Molecular modeling has also been applied to rationalize chiral separations in retrospect, which have been reported in earlier publications without specifically investigating the chiral recognition mechanism. Docking experiments and free energy of bonding calculations were performed to rationalize the separation of the enantiomers of noradrenaline (norepinephrine) by β‐CD [113]. The geometries of the complexes as well as the free binding energy revealed a greater stability of the complex of (S)‐noradrenaline compared to the (R)‐enantiomer. Both enantiomers formed 3 hydrogen bonds with the CD, but stronger (shorter) bonds were found for (S)‐noradrenaline. The higher stability of the complex of the (S)‐enantiomer was further corroborated by a higher HOMO‐LUMO band gap energy.

Dou et al. [114] used an approach combining MD simulations and quantum mechanics/continuous solvent model calculations to rationalize the superiority of HDA‐β‐CD over β‐CD in the CE enantioseparation of terbutaline studied by the group of Chankvetadze in combination with NMR spectroscopy earlier [97]. Whereas the terbutaline‒β‐CD complex featured the phenyl ring of the drug inside the cavity, the exact structure of the terbutaline‒HDA‐β‐CD complex could not be unequivocally derived from ROESY NMR data where two scenarios appeared plausible, one with the aromatic ring of the drug inserted into the cavity of HDA‐β‐CD and another one with the tert.‐butyl moiety within the CD cavity as also described above [97]. MD simulations always placed the phenyl ring inside the cavity in the case of both CDs from the wider secondary side with the alkyl side chain directed toward the solvent albeit to different extents [114]. HDA‐β‐CD showed higher deformation due to the increased flexibility because of the acetyl substituents. In both HDA‐β‐CD complexes with the terbutaline enantiomers the tert.‐butyl groups were close to the acetyl moieties in accordance with one of the NMR‐derived structures. Especially the alkyl group of (R)‐terbutaline was almost included within the secondary rim indicating strong binding of this enantiomer by the CD. In fact, during the inclusion process, (R)‐terbutaline underwent a conformational change in such a way that the tert.‐butyl group was further bent toward the secondary rim and rested almost parallel to it. The binding free energy of the (R)‐enantiomer compared to that of the (S)‐enantiomer also indicated the enantiomer migration order S > R as observed in the earlier CE experiments [97]. In the case of β‐CD, the modeled structures of the complexes indicated more interactions between the secondary rim and the side chain of (S)‐terbutaline compared to the (R)‐enantiomer and the bonding free energy data also supported the opposite CE enantiomer migration order R > S compared to HDA‐β‐CD. The somewhat contradictory conclusions drawn from the modeling data [114] compared to the CE/NMR [97] with regard to the complex structures may be due to the fact that the modeling may not match the conditions of the experimental study. On the other hand, a racemic compound was used for the determination of the complex structure by NMR, which did not allow for detecting differences in the structures of the complexes of the individual enantiomers. In a similar manner, molecular modeling including MD simulations was applied as a complementary technique to rationalize the complexation of oxazolidinones such as linezolid or tedizolid by heptakis(2,3‐di‐O‐acetyl‐6‐O‐sulfo)‐β‐CD [115]. The calculated differences in the strength of the binding of the enantiomers could predict the enantioseparation.

In most of the cases complexes with a 1:1 stoichiometric ratio between CD and solute have been reported, when a single CD is used for chiral discrimination. In the case of a dual system using simultaneously α‐CD and a positively charged CD, mono‐6A‐deoxy‐6‐(1‐allylimidazolium)‐β‐CD, for the enantioseparation of kynurenine, a ternary complex composed of both CDs and the analytes were modeled [106]. The same applied to the synergistic system of α‐CD and 18‐crown‐6 for the chiral separation of aromatic amino acids [95]. The ternary complex featured the aromatic moiety of the amino acids immersed into the CD cavity, while the side chain with the protonated amino groups protruded from the cavity and interacted with the crown ether. Nonetheless, higher‐order complexes formed between a single CD and analytes have been reported in the past [55]. Studying the enantioseparation of daclatasvir (DCV) by native and single isomer methylated β‐CDs it was noted that an enantioseparation was achieved in the presence of β‐CD, heptakis(2,6‐di‐O‐methyl)‐β‐CD (2,6‐DM‐β‐CD) and heptakis(2‐O‐methyl)‐β‐CD (2‐M‐β‐CD), while only a single peak was obtained in the presence of heptakis(3‐O‐methyl)‐β‐CD (3‐M‐β‐CD)and heptkais(6‐O‐methyl)‐β‐CD (6‐M‐β‐CD) and a two peaks with a plateau in between were observed in case of heptakis(2,3‐di‐O‐methyl)‐β‐CD (2,3‐DM‐β‐CD) and heptakis(2,3,6‐tri‐O‐methyl)‐β‐CD (TM‐β‐CD) (Figure 3) [99]. NMR data including rotating frame nuclear Overhauser enhancement spectroscopy (ROESY) indicated 1:1 complexes in the case of 3‐M‐β‐CD, 2,3‐DM‐β‐CD, and TM‐β‐CD, while higher order complexes with a stoichiometric ratio of up to 1:3 (DCV:CD) were concluded in case of β‐CD, 2‐M‐β‐CD, 6‐M‐β‐CD, and 2,6‐DM‐β‐CD. Schematic structures of the complexes of 2,6‐DM‐β‐CD, 6‐M‐β‐CD, and TM‐β‐CD are shown in Figure 4A–C as examples of the derived complex structures. DCV formed inclusion complexes with all CDs with the biphenyl structure situated in the cavity of the CDs and the methoxycarbonyl‐Val (MOC‐Val) pyrrolidine moieties protruding from the primary and secondary rims. In the case of the CDs forming higher order complexes, NOE enhancement between the signals of the Val methyl protons and the internal H‐3 and H‐5 of a 2,6‐DM‐β‐CD species were observed. Such NOE effects were not detected for the CDs forming 1:1 complexes. The exact stoichiometry (1:2 or 1:3) of the higher order complexes could not be derived by NMR spectroscopy, but 1:3 complexes for β‐CD and 2,6‐DM‐β‐CD were corroborated by electrospray ionization‐ time‐of‐flight‐mass spectrometry (ESI‐TOF‐MS) [99]. Interestingly, only a 1:1 complex between DCV and 2,6‐DM‐β‐CD was concluded from NMR experiments when only a low concentration relative to the concentration of DCV was applied in the sample solution so that complex stoichiometry appeared to depend on the CD concentration (unpublished data).

FIGURE 3.

Structure of daclatasvir (DCV) and electropherograms of the separation of DCV enantiomers in the presence of single isomer methylated β‐cyclodextrins (β‐CDs) and γ‐CD in 50 mM sodium phosphate buffer, pH 2.5. For details of the experimental conditions [99, 116].

FIGURE 4.

Chematic representations of complexes formed between daclatasvir (DCV) and (A) heptakis(2,3‐di‐O‐methyl)‐β‐cyclodextrin (2,3‐DM‐β‐CD), (B) heptakis(6‐O‐methyl)‐β‐CD (6‐M‐β‐CD), (C) heptakis(2,3,6‐tri‐O‐methyl)‐β‐CD (TM‐β‐CD), and (D) 1:1 and 2:1 complexes formed with γ‐CD as derived from nuclear magnetic resonance (NMR) spectroscopy. The nuclear Overhauser effects (NOEs) derived from rotating‐frame Overhauser enhancement spectroscopy (ROESY) experiments are indicated by arrows; blue: intermolecular NOEs between DCV and the CD, red; intramolecular NOEs, and green: intermolecular NOEs between DCV molecules in the 2:1 complex.

In the case of 2,3‐DM‐β‐CD and TM‐β‐CD, two peaks and a plateau were observed in CE experiments with DCV (Figure 3) [99]. The same phenomenon had previously been found in the presence of γ‐CD (Figure 3) [116]. However, in the latter case, the simultaneous presence of a 1:1 complex as well as a 2:1 complex (DCV:γ‐CD) were derived by NMR spectroscopy as schematically shown in Figure 4D and corroborated by MS. Two peaks with a plateau in between are typically observed in separation sciences such as HPLC and CE for equilibria such as enantiomerization, which are slow on the timescale of the respective separation technique. However, as concluded from subsequent molecular modeling and isothermal titration calorimetry studies, this CE phenomenon was interpreted as a slow dissociation of the DCV–CD complexes [117]. Molecular modeling as well as NMR studies clearly indicated a folded structure of the DCV molecule in the CD complexes with the MOC‐Val residues bending toward the rims of the CD. This was found for the 1:1 complex with TM‐β‐CD as well as for the DCV dimer included in the cavity of γ‐CD. As a consequence, the dissociation of the complexes was slowed down resulting in the described CE behavior. An important question in this context was the kinetics of the formation of the 2:1 complex, that is, the inclusion of a dimeric molecule or the successive inclusion of two molecules. Isothermal titration calorimetry indicated the formation of a DCV dimer in an aqueous solution, which could be substantiated by MD simulations [117]. The dimerization appeared to be initiated by intermolecular hydrogen bonds between the MOC‐Val‐pyrrolidine moieties resulting in a dimer with stacked biphenyl rings. Subsequently, this preformed dimer entered the cavity of γ‐CD from the wider side also driven by hydrogen bonds as shown in Figure 5.

FIGURE 5.

Molecular dynamics (MD) simulations of the daclatasvir (DCV) dimer entering the cavity of γ‐cyclodextrins (γ‐CD) from the wider rim at (A) 0 ns, (B) 0.04 ns, and (C) 3.5 ns, and distances of intermolecular hydrogen bonds (HBs) driving the inclusion. (Reproduced with permission [117], © by Elsevier 2023).

CDs have also been applied for the discrimination of enantiomers by ion mobility (IM) coupled to MS (IM‐MS). In IM, ionized analytes are separated in an inert gas under the influence of an electric field based on size, shape, and charge and MS detection measures the collisional cross‐section (CCS) values of the ions to provide information on their structures [118]. Various IM‐MS techniques have been described [118]. IM‐MS has been applied to differentiate between isomers including enantiomers as summarized in [119]. Enantiomers do not differ in their CCS values per se, but they can be analyzed as diastereomeric complexes with complexing agents such as CDs, typically in combination with a metal ion.

Some studies employed calculational methods to explain the chiral recognition principle observed in IM‐MS. For example, the chiral analysis of amino acids by IM‐MS was studied by Yang et al. [120]. Partial discrimination was observed for the complexes with native CDs, but this was greatly enhanced by the addition of monovalent or divalent metal ions. Especially Mg²⁺ proved to be highly effective. The authors also used DFT calculations to rationalize the experimental observations. Figure 6 shows the optimized structures of the complexes of Phe with β‐CD in the absence (Figure 6A) and in the presence of Mg²⁺ (Figure 6B) [120]. It is interesting to note that for the diastereomeric Phe‒β‐CD complexes insertion of the aliphatic side chain of Phe has been modeled (Figure 6A), which is in contrast to representations in aqueous media where the phenyl ring of aromatic amino acids was found within the cavity [121]. The ternary complex containing Mg²⁺ displayed differences in the interactions between the enantiomers and the metal ion (Figure 6B). The latter interacted with primary hydroxy groups of β‐CD in both complexes. Interaction between d‐Phe and Mg²⁺ occurred via the amino group, while side‐chain oxygens were involved in the case of l‐Phe. The same group also studied the discrimination of the enantiomers of mandelic acid and derivatives mediated by the native CDs, α‐CD, β‐CD, and γ‐CD as well as the divalent metal ions Mn²⁺, Fe²⁺, Co²⁺, Ni²⁺, Cu²⁺, and Zn²⁺ [122]. The chiral discrimination depended on the CD as well as the type of metal ion. For the ternary complex formed by mandelic acid, α‐CD and Cu²⁺, two favored structures were obtained which correlated to the fact that two mobility peaks for each mandelic acid enantiomer were also observed in IM‐MS. For the β‐CD and γ‐CD complexes only one orientation per enantiomer was found with the Cu²⁺ ion coordinating the carboxyl group of the drug inside the CD cavity.

FIGURE 6.

Optimized structures of (A) complex formed by β‐cyclodextrin (β‐CD) with d‐Phe (left) and l‐Phe (right) and (B) side view (top) and top view of the ternary complexes between β‐CD, Mg²⁺ ions and d‐Phe (left) and l‐Phe (right). (Reproduced with permission [120], © by Elsevier 2022).

The enantiodifferentiation of the enantiomers of penicillamine could be achieved using β‐CD and the monovalent ion Li⁺ [123]. Molecular modeling suggested that the Li⁺ ion formed hydrogen bonds with the primary OH groups of the CD as in the case of divalent ions. The penicillamine enantiomers were positioned inside the CD cavity with the carboxyl group entering from the secondary side. In the case of l‐penicillamine, Li⁺ interacted with the OH group of the drug, while the interaction of Li⁺ with d‐penicillamine occurred via the oxygen of the CO group. Furthermore, IM‐MS has also been used to analyze the enantiomers of thyroxine via the ternary complex with β‐CD and Ca²⁺ [124]. According to docking structures in the absence of Ca²⁺, the 3,5‐diiodo‐4‐hydroxyphenyl ring of thyroxine entered the CD cavity from the secondary wider rim. Adding Ca²⁺, d‐thyroxine displayed a stronger bent structure “wrapped” around the metal ion, while in the case of the ternary l‐thyroxine‒β‐CD‒Ca²⁺ complex Ca²⁺ interacted with the drug and the secondary hydroxy groups of the CD at the rim. β‐CD has also been applied as a complexing agent to increase the resolution of diastereomeric dipeptides in IM‐MS [125].

The popular combination of CDs with CIL or chiral DES or CD‐derived CILs will be addressed in section 3.9 below.

Judged by the number of publications, CD–analyte complexes are the most often modeled structures when it comes to rationalization of observations made in the process of (analytical) enantioseparations. This may be due to the fact that CDs are still relatively small molecules as compared to polysaccharide selectors, for example. Moreover, a large body of evidence has been accumulated over the years so that a fairly good understanding of the structures of CD‐analyte complexes exists and significant progress has also been made with regard to the computational methods. However, data should not be overinterpreted and should be evaluated critically and carefully as addressed by [54, 55]. In order to save computing time, assumptions and simplifications are often made and the computational model does not always correctly reflect the separation medium in which the complexes are formed during analysis as well as the fact that the complexes are mobile. Specifically, in the case of separations using randomly substituted CDs, modeling is often performed with a CD with a certain substitution pattern, for example, in references [101, 103, 107–109, 111]. The CD isomer used in such modeling does not adequately reflect the large variability of the multitude of CD isomers with different degrees of substitution and different substitution patterns in the mixture and the observed enantiomer migration order in CE is averaged over the entire range of CDs present. For example, the opposite migration order of the enantiomers of DCV has been observed in the presence of randomly methylated β‐CD as compared to 2,6‐DM‐β‐CD and also depended on the degree of substation of the randomly methylated CDs [126]. Moreover, different positional and substitution isomers of a CD may possess opposite chiral recognition for the analyte [126, 127]. Thus, rationalizing the enantiorecognition of randomly substituted CDs by modeling a certain single isomer CD out of the multitude of isomers present in the mixture may not lead to rational data, and agreement between experimental and theoretical data may be coincidental. In case of attachment of a CD to a solid support as in HPLC or capillary electrochromatography (CEC), the cavity of the CD may only be accessible for the solutes from one side, either the narrower or wider side. Summarizing, the combination of analytical enantioseparations with calculational techniques for the elucidation of the enantioseparation mechanism can be very helpful, but such studies have to be conducted in a rational manner and the data have to be critically interpreted in order to avoid misleading conclusions. These precautions co not apply only to CDs but to all chiral selectors which do not possess a uniform structure.

3.3. Further oligosaccharide selectors

Apart from CDs, linear and cyclic oligosaccharides have also been applied as chiral selectors. A group of cyclic oligosaccharides are cyclofructans, which are built from β(2,1) linked d‐fructofuranose units, and represent a relatively new type of chiral selector. Native cyclofructans containing six or seven d‐fructofuranose units as well as O‐alkyl and acyl derivatives, carbamoyl derivatives, or sulfated derivatives have been evaluated for chromatographic or CE enantioseparations [128]. In contrast to CDs, cyclofructans exhibit a disk‐like shape with an electronegative side composed of the hydroxy groups in the 3‐ and 4‐positions of the fructofuranose units and an electropositive side formed by the 1‐ and 6‐methylene moieties. The complexation of solutes is mediated via polar interactions such as dipole–dipole interactions or hydrogen bonds depending on the nature of the cyclofructan derivative, the solute, and the operation mode of the separation technique [15, 19]. The development of cyclofructans and derivatives as chiral selectors and their applications for analytical enantioseparations has been recently summarized [129, 130] but new studies on the chiral recognition mechanism of these selectors have not been published in the period of time covered by this review.

Maltodextrins are a complex mixture of oligo‐ and polysaccharides composed of α(1,4) linked d‐glucopyranose units, which are obtained by the hydrolysis of starch. They form helical structures with a hydrophobic interior and a hydrophilic outside and have been applied as chiral selectors in CE [131]. With regard to this selector, the synergism with the CILs tetramethylammonium d‐gluconic acid and tetramethylammonium shikimic acid in CE enantioseparations of basic and acidic drugs has been investigated [132]. The combinations led to improved enantioseparations as compared to the sole use of maltodextrin as a chiral selector. NMR studies of a sample containing citalopram and the mixture of maltodextrin and a CIL indicated low field shifts primarily of the aromatic signals of the drug indicating interactions between citalopram and the selector system. Molecular docking indicated more negative values of the total binding energies for the combination maltodextrin/CIL compared to maltodextrin alone, which was in accordance with the improved chiral separation. The modeled structure of the ternary complex formed by the citalopram enantiomers, a maltodextrin helix, and the d‐gluconic acid‐derived CIL showed increased interactions compared to a dimeric structure formed from the drug and maltodextrin. While the latter interaction was mediated via a single hydrogen bond, additional hydrogen bonds as well as hydrophobic π‐π interactions were observed in the ternary complex. In accordance with the enantiomer migration order, (S)‐citalopram featured more interactions with the selector system compared to the (R)‐enantiomer.

3.4. Macrocyclic antibiotics

Macrocyclic antibiotics have been frequently applied as chiral selectors in analytical liquid phase enantioseparations. In separation sciences, the most prominent representatives of this class are vancomycin, ristocetin A, teicoplanin, and teicoplanin aglycone. The compounds feature a macrocycle containing aromatic rings, which are formed from interconnected amino acids and substituted by carbohydrates. Vancomycin contains three macrocycles, while ristocetin and teicoplanin contain four. Macrocyclic antibiotics form a C‐shaped basket‐like structure with the carbohydrate moieties positioned at the surface. Because of the presence of different functional groups, chiral recognition of solutes may be based on various interactions including hydrogen bonds as well as ionic, dipole‐dipole, and π‐π interactions [14, 15, 18–21]. In addition, steric hindrance can also contribute to the process. Modeling approaches using macrocyclic antibiotics as selectors in chiral chromatography have been reviewed [59] and recent applications for enantioseparations of pharmaceutical drugs were summarized [133, 134].

A new chiral column containing vancomycin as a chiral selector was synthesized by attachment of the vancosamine moiety to aminopropyl silica via succinic acid [135]. Molecular modeling of the interaction with mandelic acid revealed higher binding energies for the (R)‐enantiomer, which was in agreement with the enantiomer elution order in HPLC experiments. π‐π Interactions and hydrogen bonds played a major role in the stereoselective binding of the mandelic enantiomers to vancomycin.

A detailed computational study for evaluating the binding modes of the enantiomers of the dipeptide carnosine to the selector teicoplanin A2‐2 was published by the group of Carotti to explain the chromatographic elution order S > R under reversed‐phase conditions [136]. (R)‐Carnosine engaged a stabilizing charge–charge interaction through its ionized imidazole ring with a carboxylate group of the chiral selector, while interactions between (S)‐carnosine and the selector limited conformational freedom of the selectand and impaired the association with the chiral selector. MD simulations revealed ionic interactions and hydrogen bonds as driving forces for the complexation event. The carnosine enantiomers were not bound inside the aglycone basket but mainly interacted via ionic interactions with the carboxylic acid function of the selector. Hydrogen bond contacts with functional groups in the proximity of the anionic site of teicoplanin A2‐2 resulted in the stereoselective recognition of the carnosine enantiomers. The chromatographic enantioseparation of a set of 4‐aryl dihydropyrimidinone derivatives on a teicoplanin aglycone containing CSP under polar organic mode elution conditions was reported by Bolognio et al. [137]. Hydrogen bonds as well as polar interactions and π‐π interactions contributed to the binding of the analytes to the chiral selector. MD simulations also indicated that repulsive steric effects supported the enantioseparation process. π‐Stacking as well as hydrogen bonds were also deduced from molecular modeling of the interaction between the pantoprazole enantiomers and the teicoplanin aglycone [138]. The chromatographic elution order S > R observed under reversed‐phase conditions was consistent with the formation of three π‐interactions in the case of the (R)‐enantiomer versus two for (S)‐pantoprazole. Furthermore, the binding positions to the selector differed for the enantiomers.

The group of Carotti also evaluated the chiral interaction between the enantiomers of the dipeptides Ala‐Ala, Gly‐Leu, Leu‐Gly, and Leu‐Leu and ristocetin A as chiral selectors simultaneously considering two alternative binding modes of the selector to a solid support [139]. Ristocetin A can be linked to silica gel via one of the two amino functions of the molecule, one of them present in the carbohydrate vancosamine (termed linkage A) and the other one in the 3,4‐dihydroxyphenylalanine moiety in the macrocycle (termed linkage B). This results in different orientations of the chiral selector as shown in Figure 7. In HPLC experiments under reversed‐phase conditions, the l‐ or ll‐stereoisomers eluted before the respective d‐ or dd‐enantiomers. Modeling studies were performed assuming a ratio of 22.5% linkage A and 77.5% linkage B, which were concluded from the reactivity of the respective amino groups under the coupling conditions based on their pK_a values. Depending on the dipeptide, different driving forces were identified by MD simulations to account for the enantiomer elution order. In accordance with the enthalpy contributions observed in the thermodynamic evaluations of the elution sequences of Ala‐Ala‐and Leu‐Leu, molecular modeling indicated that selector‐selectand complexation is mainly controlled by electrostatic interactions. The carboxylic acid group of d‐Leu‐d‐Leu was situated closer to the positive charges of the selector than the ll‐enantiomer. In addition, hydrogen bonds proved to be a discriminating factor between the enantiomers as was also observed for Ala‐Ala. This was especially true for B‐linked ristocetin A, which represented the more abundant linkage. Hydrogen bond frequency was slightly higher for d‐Leu‐Gly with ristocetin A bound via linkage B compared to the l‐enantiomer, but this was not found for d‐Leu‐Gly bound to the A‐linked selector nor for the enantiomers of Gly‐Leu toward both systems. In this case, the enantiorecognition mechanism for the enantiomers of Gly‐Leu was mainly driven by van der Waals interactions. Moreover, the interaction sites varied depending on the binding mode of the selector. The example of Gly‐Leu and Leu‐Gly highlighted the importance of the amino acid sequence of the dipeptides determining the binding modes between the dipeptides and the chiral selector [139].

FIGURE 7.

Chemical structures (left) and molecular modeling structures (right) of ristocetin A linked to silica gel via the amino groups in the vincosamine moiety (linkage A, top) or in the 3,4‐dihydroxyphenylalanine moiety (linkage B, bottom). (Reproduced with permission [139], © by Elsevier 2022).

Furthermore, the aminoglycoside antibiotic kasugamycin composed of two saccharide moieties was evaluated as a new chiral selector in CE [140]. A number of basic drugs including ephedrine/pseudoephedrine, promethazine, or amlodipine could be enantioseparated using weakly basic BGE (pH 7.5–8.5). One‐ and two‐dimensional NMR experiments with the enantiomeric pair ephedrine/pseudoephedrine indicated interactions between the aliphatic side chain of the enantiomers and kasugamycin. Docking simulations supported the presence of hydrogen bonds between the amino and hydroxy groups of the solutes and oxygen atoms of kasugamycin resulting in different orientations of ephedrine and pseudoephedrine relative to the selector.

3.5. Proteins

The stereoselective binding of drug enantiomers by serum proteins such as human and bovine serum albumin or α₁‐acid glycoprotein (AGP) has been well‐documented since the 1950s. Thus, proteins and glycoproteins have also been used as chiral selectors in separation sciences [141]. Due to the complexity of proteins, a multitude of non‐covalent interactions between the protein and analyte enantiomers contribute to the complexation including hydrogen bonds, π‐π, dipole‐dipole, and hydrophobic as well as ionic interactions [12, 18, 21].

The stereoselective binding of benzoin, chlorpheniramine, and propranolol by AGP was revisited employing the native protein as well as a derivative, where the Trp residue in position 26 had been modified by reaction with 2‐nitrophenylsulfenyl chloride [142]. It had previously been observed that the modified AGP lost the chiral recognition ability toward β‐blockers such as propranolol, while enantioseparation of benzoin and chlorpheniramine was retained although resolution was lower compared to the use of native AGP as chiral selector [142, 143]. Using molecular modeling including MD simulations it was concluded that all compounds were bound to a cavity of AGP lined by His25, Trp26, Tyr47, Arg128, Thr129, Asp161, and Glu168. The enantiomers of all three studied compounds bound in a similar manner and position in this cavity but differed from each other in the number of established ionic interactions, hydrophobic interactions, and/or hydrogen bonds. This suggested that (R)‐benzoin, (S)‐chlorpheniramine, and (R)‐propranolol formed the tighter complexes compared to the respective enantiomers, which correlated with the HPLC elution order in the case of benzoin and chlorpheniramine. The major difference between the binding positions of the compounds was that benzoin and chlorpheniramine did not bind in the proximity of Trp26, whereas the propranolol enantiomers docked in a position slightly shifted toward His25 and Trp26. Modeling of the structure of the modified AGP was less reliable but the authors concluded that modification by Trp26 by a 2‐nitrophenylsulfenyl group would result in a more rigid structure as well as small shifts in the position of the amino acids lining the binding pocket. This might explain the loss of enantioselectivity of the Trp26‐modified AGP toward propranolol in HPLC experiments because of its close binding to the modified amino acid [142].

Fagerström et al. revisited the chiral recognition of an analog of the β‐blocker drug propranolol containing a phenanthrene moiety (Figure 8) by cellobiohydrolase Cel7A using X‐ray crystallography, molecular modeling, thermodynamic and CE data [144]. According to X‐ray analysis, the enantiomers of the analog were bound in a comparable manner and the position of the enantiomers of propranolol and the analog were also almost identical [145]. However, as concluded from the complex structure of the (R)‐enantiomer conformational changes in the protein appeared to be induced. In molecular docking experiments, both enantiomers of the analog were bound in a similar way with regard to the position of hydroxymethyl groups of the secondary nitrogen but differed in the position of the phenanthryl ring and the established interactions. Specifically, water‐mediated hydrogen bonds were important for the binding of the (R)‐enantiomer to the cavity. According to crystallographic structure, water molecules were positioned differently for both enantiomers and appeared to impose a limitation on the flexibility of the ligand. The intra‐molecular interactions reduced the free movement of water molecules instead of expelling them from the binding site. By sequential elimination of water molecules and repeated docking, 14 water molecules in the vicinity of the ligand were identified, which influenced the docking of the (R)‐enantiomer. The authors emphasized that the importance of water molecules in the binding site provided further significant information on the complexation of compounds by cellobiohydrolase Cel7A and may serve as a model system for computational studies of the effect of water molecules on binding and selectivity of protein selectors [144].

FIGURE 8.

Structure of 2‐[2‐hydroxy‐3‐(phenanthren‐9‐yloxy)propylamino]propane‐1,3‐diol and stereoview of the molecular docking structure of the (S)‐enantiomer (violet) and the (R)‐enantiomer (green) into the active site of cellobiohydrolase Cel7A. Selected amino acid side chains, the main chain, and water molecules as well as interactions are shown. (Reproduced with permission [144], © by Wiley 2023).

3.6. Chiral crown ethers

Based on the coordination of cations in the macrocyclic ring via hydrogen bonds to the oxygen atoms [14, 15, 19–21, 130] chiral crown ethers have specifically been applied to the enantioseparation of primary amines by HPLC and CE although the separation of secondary amines and even amides have been reported [146].

The enantioseparation of quinolone drugs featuring primary, secondary, and ternary amino functions on a column to which (+)‐(18‐crown‐6)−2,3,11,12‐tetracarboxylic acid had been immobilized was studied [147]. Separation of the enantiomers was achieved in the reversed‐phase mode using acidic eluents to ensure the protonation of the amino groups of the analytes. Despite the fact that the authors described interactions between the protonated amino compounds with oxygen atoms of the crown ether ring, the structures shown in the publication only displayed ionic interactions between the protonated amino groups with the anionic carboxylate groups of the crown ether, which somewhat contradicts the established separation mechanism of chiral crown ethers mentioned above. In a subsequent study, the separation of quinolones with two stereogenic centers by the (+)‐(18‐crown‐6)−2,3,11,12‐tetracarboxylic acid selector was evaluated [148]. As described in the previous study [147], ionic interactions between the protonated amino functions of the analytes and the carboxylate groups of the selector were modeled [148].

Upmanis et al. reported the enantioseparation of the llll‐stereoisomer of the tetrapeptide Tyr‐Arg‐Phe‐Lys‐NH₂ and its dddd‐enantiomer on chiral columns containing either immobilized (R)‐ or (S)‐(3,3ʹ‐diphenyl‐1,1ʹ‐binaphthyl)−20‐crown‐6 as selector in the polar organic elution mode [149]. NMR and MS studies were carried out with the soluble selectors. The llll‐stereoisomer was retained stronger on the column containing immobilized (S)‐(3,3ʹ‐diphenyl‐1,1ʹ‐binaphthyl)−20‐crown‐6, while the dddd‐tetrapeptide eluted second from the (R)‐(3,3ʹ‐diphenyl‐1,1ʹ‐binaphthyl)−20‐crown‐6 containing stationary phase. MS data for the soluble selector (not attached to the solid support) indicated the formation of complexes with 1:1, 1:2, and 1:3 stoichiometry (tetrapeptide:crown ether). As the ratio between the intensity of the signals for the complexes was not affected by the amount of crown ether added to the sample, the authors concluded that a non‐stereospecific binding pattern might occur in the gas phase as compared to the situation in LC. Complexation‐induced shifts of NMR signals of the tetrapeptide corroborated the formation of complexes. Shifts were more pronounced for the llll‐tetrapeptide compared to the dddd‐enantiomer in the complex with (S)‐(3,3ʹ‐diphenyl‐1,1ʹ‐binaphthyl)−20‐crown‐6, while the opposite was observed for (R)‐(3,3ʹ‐diphenyl‐1,1ʹ‐binaphthyl)−20‐crown‐6. This is in agreement with the HPLC elution sequences. From 2D NMR data, the authors concluded that the interactions between the crown ethers and the peptide enantiomers occurred via non‐enantioselective hydrogen bonds between the protonated ε‐amino group of the Lys residue and the oxygens of the crown ether and a stereoselective hydrogen bond between the α‐amino group of Tyr and oxygens of the crown ether. In addition, π‐interactions involving the Phe residue and aromatic rings of the selectors might contribute to chiral recognition [149].

3.7. Donor‐acceptor type selectors (Pirkle‐type selectors)

Donor‐acceptor‐type selectors are also called brush‐type selectors or Pirkle‐type selectors after William H. Pirkle, the pioneer of this selector type [150]. These small molecule selectors typically contain an electron‐deficient aromatic moiety such as a dinitrophenyl ring acting as a π‐acceptor and/or an electron‐rich moiety, for example, a naphthalene of phenanthrene ring, which can function as a π‐donor. Thus, chiral recognition is mediated via face‐to‐face or face‐to‐edge π‐π interactions or dipole‐dipole stacking. Furthermore, hydrogen bonds as well as a steric hindrance in the case of bulky residues can contribute [14, 15, 18–21]. Applications of the commercial Whelk‐O1 phase in supercritical fluid chromatography and normal‐phase chromatography have been compared recently [151].

The chiral recognition of the conformational enantiomers of nevirapine and oxcarbazepine by the Whelk‐O1 phase was studied by Franzini et al. [152]. The Whelk‐O1 selector features a dinitrophenyl ring connected to 1,2,3,4‐tetrahydrophenanthrene moiety via an amide bond forming a “cleft” between the dinitrophenyl and the naphthalene partial structure of the tetrahydrophenanthrene ring. The analytes contain a dibenzoazepine ring with a non‐planar seven‐membered ring, which is known to be chiral resulting in a special case of planar chirality. However, the enantiomers can interconvert due to a flip‐mechanism of the ring. Thus, at 25°C only a single peak could be observed, but enantioseparation was achieved at −50°C. The enantiomerization barrier was about 15.9 kcal/mol for nevirapine and 16.9 kcal/mol in the case of oxcarbazepine. Conformational analysis of the (pR)‐enantiomers by semiempirical calculations in vacuo revealed one conformation for nevirapine but two in the case of oxcarbazepine so that two conformations were considered for the latter compound in docking studies using the (S,S)‐configured chiral selector present in the commercial column. For nevirapine, the data indicated that the homochiral diastereomeric (pS)‐(S,S) complex was more stable than the heterochiral (pR)‐(S,S) complex in the case of nevirapine, while it was opposite in the case of oxcarbazepine. This was true for both conformations of the dug used in the calculation process. Regardless of the stereochemistry of the complexes, a hydrogen bond was always observed between the ring oxygen of the analytes and the amide bond of the selector. Shorter hydrogen bond distances were always found in the more stable host–guest complexes. In the case of (pR)‐nevirapine, one aromatic ring was directed over the naphthyl moiety of the selector, while the second ring was directed face‐to‐edge toward the host dinitrobenzene ring. (pS)‐Nevirapine was situated within the cleft of the host, with the aromatic rings aligned over the host naphthyl and dinitrobenzene rings, respectively. In the complex with conformer 1 of (pR)‐oxazepine, the guest adopted an outer complexation geometry, with one aromatic ring stacked parallel to the host dinitrophenyl ring and the other ring facing the ethylene bridge of the host. Conformer 1 of (pS)‐oxcarbazepine and both enantiomers of conformation 2 of oxcarbazepine were placed inside the host cleft, with one ring stacked on the dinitrophenyl host ring and the other ring aligned over the host naphthyl ring [152].

The interactions of the (S,S)‐Whelk‐O1 selector with a set of three partially saturated 2H‐indazole compounds differed depending on their substitution pattern of the aromatic ring [153]. Interestingly, the less retained (3S,3aR)‐enantiomer of the unsubstituted compound (Figure 9) interacted with two selector molecules via one π‐cation and one edge‐to‐face π‐π stacking interactions with the nitro substituent of one selector molecule and the tetraphydrophenanthrene moiety of the second selector molecule, while the stronger retained (3R,3aS)‐enantiomer interacted with a single selector molecule via a hydrogen bond and a π‐cation interaction between the amide nitrogen of the selector and the nitro groups, respectively (Figure 9) [153]. In the case of derivatives with a methoxy or a fluorine substituent at the aromatic ring, both enantiomers interacted with a single selector molecule primarily via π‐π interactions. The first eluting (3R,3aS)‐enantiomers adopted constrained conformations in the complexes.

FIGURE 9.

Structures of the (S,S)‐Whelk‐O1 selector (left) and the partially saturated 2H‐indazole model compound (right) and interacting frames of the (3S,3aR)‐stereoisomer (A, cyan) and the (3R,3aS)‐enantiomer (B, orange). The Whelk‐O1 selector is represented in green balls and sticks. π‐π stacking and π‐cation interactions are shown as green and cyan dashed lines, respectively, while hydrogen bonds are represented by black dashed lines. Charge‐charge interactions are shown as dashed magenta dashed lines. (Reproduced with permission [153], © by Elsevier 2020).

Finally, the enantioseparation of a set of chiral phenylhydantoins on the (S,S)‐Whelk‐O1 phase has been studied under normal‐phase elution [154]. Docking of selected compounds into the selector indicated complex formation via π‐π interactions and hydrogen bonding. The complexes were different in the presence of a methyl group at the hydantoin nitrogen N1, which caused a steric effect with the naphthalene moiety in the case of the (R)‐enantiomer. While the hydrogen bond and π‐π interaction remained, a different orientation compared to the (R)‐enantiomers of non‐methylated solutes was observed. The structures of the diastereomeric complexes between the selector and the (S)‐enantiomers of the hydantoins were not affected by N1 methylation.

3.8. Chiral ion‐exchanger selectors

Chiral ion‐exchange selectors are sometimes considered a subgroup of donor‐acceptor selectors because π‐π interactions and hydrogen bonds contribute to chiral recognition apart from the eponymous ionic interactions [14, 15, 18–21]. The most often used anion‐exchange selectors are based on basic Cinchona alkaloids, for example, quinine or quinidine, while cation‐exchangers contain sulfonic acid or carboxylic acid groups as reviewed in [155]. Zwitterionic ion‐exchange selectors also exist [156]. The chromatographic characteristics of the zwitterionic selectors including their orthogonality for LCxLC application have been studied [157]. Furthermore, the so‐called fragmentation approach has been applied to understand the contribution of the functional groups of ion‐exchange selectors to the enantioseparation mechanism in chromatographic elution modes [158] and SFC [159].

Varfaj et al. achieved the enantioseparation of mandelic acid, 3‐phenyllactic acid, and 3‐(4‐hydroxyphenyl)lactic acid on zwitterionic CSP containing quinidine and trans‐2‐aminocyclohexanesulfonic acid [CHIRALPAK ZWIX(‐)] using acidic polar organic elution mode conditions [160]. Assignment of the stereoisomers of the analytes to the chromatographic peaks was achieved by ECD studies in combination with ab initio time‐dependent DFT simulations. The (S)‐enantiomers of the compounds eluted before the (R)‐enantiomers. Applying MD simulations revealed an essential role of the p‐hydroxy group in the retention mechanism. MD analysis of 3‐(4‐hydroxyphenyl)lactic acid and the chiral selector indicated a preferential interaction of the selector with the (R)‐enantiomer as compared to the (S)‐enantiomer, which was in agreement with the observed elution order in HPLC. Interestingly, the p‐hydroxy group of the (R)‐stereoisomer interacted frequently with the sulfonic acid group of the ZWIX(‐) selector, while in the case of the (S)‐enantiomer a hydrogen bond was established only a few times with the methoxy group of the quinoline moiety of the selector during the simulation timeframe.

3.9. Chiral ILs and DESs

ILs and, more recently, DES have been applied for enantioseparations by CE as summarized, for example, in references [161, 162, 163, 164, 165]. From a principal point of view, ILs and DES share similarities but also differences [166]. ILs as well as DES are composed of anions and cations and are liquid at or close to room temperature (up to approx. 100°C), but they differ in their starting materials and their synthesis. ILs are composed of organic cations and organic or inorganic anions, while DES is a combination if hydrogen bond acceptors (Lewis bases) and hydrogen bond donors (Lewis acids). Furthermore, the synthesis of ILs may involve several steps, reagents, and organic solvents and requires long reaction times. DES is prepared by (short‐term) heating or mechanical grinding of the two components until a homogenous liquid is formed. For further general information on ILs and DES see reference [166]. In the case of the application to analytical enantioseparations often either a single component of ILs or both components are chiral molecules.

Typically, CILs are used in combination with another chiral selector such as a CD, enhancing peak resolution compared to the use of the chiral selector alone. Most of the so far accumulated evidence in CE enantioseparations indicated a modification of the electroosmotic flow as the cause of the improvement although the effect is often more pronounced when the IL contains chiral components compared to analogous non‐chiral ILs. Structural evidence for the participation of a component of the CIL in the formation of the complex between the chiral selector and the selectand is scarce so far, and only a few studies provided evidence of the participation of the CIL in the interaction between the selector and the selectand. For example, the presence of a ternary complex between homocysteine derivative, γ‐CD and a CIL was concluded from NMR spectroscopy [167], while molecular modeling was applied to derive the structure of the complex formed by mandelic acid, teicoplanin, and the CIL [168]. In a recent study, glucosyl‐β‐CD was combined with CILs derived from l‐hydroxyproline or (1S,2R)−1‐carboxy‐2‐hydroxypropan‐1‐amine for the CE enantioseparation of the basic drugs amlodipine, nefopam and econazole [169]. The presence of the CILs in the BGE significantly improved the enantioseparations compared to the sole use of glucosyl‐β‐CD. Molecular docking structures showed a difference in the complexation of the amlodipine enantiomers by the CD in the presence and absence of the l‐hydroxyproline/trifluoroacetic acid CIL although none of the CIL components directly interacted with amlodipine. For example, (R)‐amlodipine established two hydrogen bonds with the CD, while in the presence of the CIL, the orientation of the solute in the CD cavity differed and a σ‐π stacking interaction and a hydrogen bond was established. In the case of the (S)‐enantiomer two hydrogen bonds were derived in the presence and absence of the CIL but the orientation of the compound relative to the CD differed in both cases [169]. Interactions between carboxymethyl‐β‐CD, the solute mirtazapine, and a CIL derived from (1S)‐(+)‐camphor‐10‐sulphonic acid were observed by NMR although the exact structure of the complex could not be concluded [170]. Another example is the synergism between maltodextrin as the primary selector and the tetramethylammonium salts of d‐gluconic acid and shikimic acid as CILs to improve the maltodextrin‐mediated CE enantioseparations [132] discussed above in section 3.3.

Another approach has been the synthesis of CILs using an established chiral selector such as a CD as one of the components. This has been realized for sulfobutylether‐β‐CD [171] and carboxymethyl‐β‐CD [172]. In the case of sulfobutylether‐β‐CD, the counterion was the (S)‐(3‐chloro‐2‐hydroxypropyl)trimethylammonium ion. Molecular modeling was performed with a CD substituted with two sulfobutylether residues at positions 2 and 6 of a single d‐glucopyranose unit, which does not reflect the true nature of the CD as discussed above, so that conclusion should be considered with care. Different orientations and interactions between the enantiomers of amlodipine and the CD were noted in the presence of the chiral counterion [171]. The CIL tetramethylammonium carboxymethyl‐β‐CD provided high chiral separation ability of several basic drugs in CE [172]. Again, only a defined CD isomer was modeled not representing the chiral selector in a correct way, but additional interactions were noted for the CIL compared to the sole presence of carboxymethyl‐β‐CD.

CD‐based CILs composed of mono‐6‐deoxy‐6‐(3‐benzylimidazolium)‐β‐CD and the amino acids l‐Cys, l‐Asp, or l‐Lys were evaluated in chiral composite membrane separation experiments using mandelic acid as a model compound [173]. Better performance of the membranes was noted for dissociated mandelic acid, which could be rationalized by the presence of additional electrostatic interactions between the mandelic acid anions and the positively charged selector. Molecular docking also provided an explanation for the higher enantiorecognition of the CIL containing l‐Cys or l‐Lys compared to the CIL containing l‐Asp due to the presence of additional hydrogen bonds formed with the thiol or ε‐amino function in contrast to the carboxylate group of l‐Asp.

Because of the relatively short period of time since their introduction in separation sciences, chiral DES has not yet been extensively studied especially with regard to the separation mechanism. Typically, DES was used in combination with established chiral selectors such as CDs, see for example [174, 175, 176, 177, 178]. Very recently, several DES based on lactobionic acid as sole chiral selectors have been introduced, but the interactions between the analytes and the selector components were not elucidated by spectroscopic or modeling approaches [179]. The synergism between the selector clindamycin and chiral DES composed of choline and l‐malic acid, l‐tartaric acid, and l‐lactic acid indicated somewhat improved enantioseparations of basic analytes in the presence of the DES in a borate BGE, pH 8.0 [180]. The optimal DES contained a combination of l‐lactic acid and l‐malic acid. In the NMR spectrum of citalopram as a model compound with the system clindamycin combined with the “mixed” DES downfield shifts of aromatic and aliphatic protons of citalopram were noted indicating interactions of the analyte with the selector system. Moreover, cross peaks in the 2D NMR spectrum showed interactions between citalopram and clindamycin as well as interactions between the drug and lactic acid and malic acid. The structures of the complexes of the drug enantiomers obtained by molecular docking displayed hydrogen bonds and π‐interactions. In the presence of the DES, a tighter fit between clindamycin and citalopram was noted compared to the structure in the absence of the DES.

Although not combined with analytical enantioseparations, Pietro et al. demonstrated by ¹H‐ and ¹³C‐NMR spectroscopy including ROESY experiments that amitryptiline and cyclobenzaprine formed a 1:1 inclusion complex with β‐CD in a DES composed of choline chloride and urea clearly indicating that chiral discrimination takes place in DES [181]. An MD simulation study on DES derived from menthol as a hydrogen bond acceptor and acetic acid, lauric acid, or pyruvic acid as a hydrogen bond donor was conducted in order to acquire a better understanding of their molecular structure [182]. Interactions between the hydroxy group of menthol and the carbonyl oxygens were found. Furthermore, several physico‐chemical properties including interaction energies between the components, viscosity, density, or dipole moments were calculated. Interestingly, (S)‐menthol showed a lower self‐diffusion coefficient in the DES compared to (R)‐menthol based on interaction energies with other species as well as hydrogen bond numbers. The authors pointed out, that the consequences of the theoretical data with regard to the applicability of the DES to enantioseparations remain to be studied.

3.10. Chiral micelles

Chiral micelles have been employed in the micellar electrokinetic chromatography mode for the separation of neutral and charged analytes by CE. Because of the dynamic equilibrium of the surfactant molecules between the solution and the micelles, suitable conditions for NMR spectroscopic or modeling studies have not been established in order to adequately represent this specific situation. Instead, polymeric micelles have been studied and the interaction between analytes and the micelles could be concluded from modeling data as summarized in [18, 19].

Garcia and colleagues investigated the chiral separation mechanisms of the dansyl (Dns) amino acids Dns‐Leu, Dns‐Nle, Dns‐Trp, and Dns‐Phe by the polymeric micelle system poly‐sodium N‐undecanoyl‐(l)‐leucylvalinate (poly‐SULV) using MD simulations [183]. The calculated binding free energy values for the interaction of the Dns amino acids with poly‐SULV indicated a stronger interaction with the l‐enantiomers, which was in agreement with the enantiomer migration order d‐ > l‐amino acids. For a poly‐SULV micelle composed of 20 surfactant monomers, three binding pockets were identified by MD simulations as shown in Figure 10A, while Figure 10B summarizes the highest‐scoring docking positions of the Dns‐Leu enantiomers in the three pockets. Binding free energy values were used to calculate the percent occupancy of each binding pocket for a given enantiomer and hydrogen bond analysis was used to rationalize the stronger binding of the l‐enantiomers to poly‐SULV compared to the d‐amino acid derivatives. Both enantiomers of Dns‐Leu and Dns‐Nle bound preferentially to pocket 1. However, while the l‐enantiomers displayed an occupancy of 98.84% (Dns‐l‐Leu) and 95.17% (Dns‐l‐Nle), respectively, the percentage of the d‐enantiomers in pocket 1 was lower (87.84% and 64.57%) with a significant percentage of binding to pocket 3 as well (10.89% and 32.93%). For the aromatic amino acids Dns‐Trp and Dns‐Phe, a diverse picture was obtained. Dns‐l‐Trp favored pocket 3 (72.29%), while Dns‐d‐Trp preferentially occupied pocket 2 (62%). In the case of Dns‐Phe, both enantiomers favored pocket 3 followed by pocket 1. Typically, the l‐enantiomers established more hydrogen bonds in the favored pockets compared to the d‐enantiomers. Thus, they interacted strongly with the poly‐SULV micelles in accordance with the migration order observed in micellar electrokinetic capillary chromatography. An exception was Dns‐l‐Phe, which established a single major hydrogen bond in pocket 3 as compared to three in the case of the d‐enantiomer. However, the lifetime of this hydrogen bond as well as its occupancy was much longer than the three hydrogen bonds formed by the d‐enantiomer indicating also preferential binding of Dns‐l‐Phe by poly‐SULV [183].

FIGURE 10.

(A) Spatial orientation of the three binding pockets in the poly‐sodium N‐undecanoyl‐(L)‐leucylvalinate (poly‐SULV) micelle and (B) highest scoring docked positions of Dns‐l‐Leu (blue, top) and Dns‐d‐Leu (red, bottom) in binding pockets 1–3. (Reproduced with permission [183], © Scientific Research Publishing 2021).

3.11. Miscellaneous selectors

Despite the fact that essentially the enantiomers of the vast majority of analytes could be separated using one of the established chiral selectors, the search for new selectors is still an active research topic. As it is impossible to address each novel selector within the frame of this short review only a few recent approaches will be discussed here. A relatively large group of selectors comprises MOFs and COFs. Both are highly ordered 3D functional materials with uniform pores. MOFs are built from metal ions (nodes) connected via di‐ or multidentate organic linkers. In contrast, COFs are obtained by the reaction of organic precursors forming covalent bonds. Thus, depending on the used organic and (in the case of the MOFs) inorganic starting materials a large structural variety is possible for both classes of compounds. Various applications of the materials have been reported including their application to enantioseparations [184, 185, 186, 187, 188].

A MOF constructed from a D3‐symmetry helical chiral Ru(phen)₃‐derived tricarboxylate ligand connected via Eu nodes was utilized for the enantioseparation of 1,1′‐bi‐2‐naphthol [189]. As revealed by DFT calculations, chiral recognition was achieved by N···H−O hydrogen bonds and π−π stacking between the host and guest. The (S)‐enantiomer formed a more stable complex, which was consistent with HPLC data. Jiang et al. prepared chiral porous Zr(IV)‐MOFs containing enantiopure 1,1′‐biphenyl‐20‐crown‐6 [190]. The chiral crown ether moieties were periodically aligned within the framework channels. The selector was applied to the enantioseparation of amino acids and their derivatives as well as the primary amino group‐containing drugs. Chiral recognition occurred via the crown ether moiety as derived from DFT calculations. As shown for Phg‐OMe, both enantiomers bind via 3 hydrogen bonds established between the protonated amino group and oxygens of the crown ether, but the binding energy was higher for the l‐enantiomer than that for d‐Phg‐OMe, which was in agreement with the enantiomer elution order in HPLC experiments. A chiral MOF was synthesized from the chiral organic ligand l‐His, the non‐chiral ligand 2‐methylimidazole and copper(II) and evaluated as CSP for the enantioseparation of basic drugs by open‐tubular CEC [191]. Molecular docking of (R)‐salbutamol to the selector resulted in three hydrogen bonds, while no hydrogen bond was detected for the (S)‐enantiomer. This was consistent with the longer migration time of (R)‐salbutamol in CEC and further supported by Gibbs free energy calculations.

Several COFs containing a CD derivative were constructed and applied to the discrimination of drug enantiomers via chromatography or membrane separations by the group of Ji [192, 193, 194]. Molecular modeling for deriving the molecular basis of the chiral recognition was conducted using the monomeric CDs. The formation of inclusion complexes stabilized by hydrogen bonds and π‐σ interactions were observed. Interestingly, for a β‐CD COF thin film nanocomposite membrane molecular docking simulations revealed two types of interactions with β‐CD in a COF cell unit derived from crystallographic data [194]. In the first one, the enantiomers of propranolol entered the CD cavity with the naphthalene ring from the wider side forming an inclusion complex as expected. However, in a second energy‐minimized configuration propranolol interacted with the outer side of β‐CD stabilized by hydrogen bonds. The authors speculated as the reason for this phenomenon that the inherently larger pore size of the COF backbone compared to β‐CD resulted in a rapid passing of a proportion of the propranolol molecules through the channels failing to interact with the hydrophobic cavity of the CD. The preferential binding of (S)‐propranolol as concluded from the average binding energy was in agreement with the experimental results.

A “dual” selector for HPLC enantioseparations was constructed by combining the ion‐exchange type selector quinine and a lipophilic CD derivative [195]. In this selector, quinine was incorporated in the spacer, which attached a methyl or chlorine substituted per‐(phenylcarbamoyl)‐β‐CD to the solid support. The enantioseparation of neutral, acidic, and basic analytes was evaluated in the reversed‐phase and polar organic elution modes. The functionality of the phenyl substituents of the CD was found to affect the separations. For example, flavonoids were resolved better on selectors containing a chloride‐substituted CD, which may be due to the electron‐withdrawing groups strengthening π‐π interactions. In contrast, basic enantiomers could not be resolved on these CSPs indicating that chloro‐phenyl substituents are unfavorable for the resolution of analytes with amine moieties. Molecular docking indicated that neutral analytes were preferentially included in the CD cavity while organic acids interacted with the CD moiety or the quinine selector depending on the phenylcarbamate substituent of the CD. For example, Dns‐Phe and 2‐(2,4‐dichlorophenoxy)propanoic acid were associated with the quinine part of the selector in the case of 3,5‐dimethylphenylcarbamate substituents on the CD, while the analytes entered the CD cavity when substituted with 3,5‐dichlorophenylcarbamate. The authors concluded that more work is required to understand the chiral separation mechanism of such dual selectors.

Chiral molecularly imprinted polymers (MIPs) are polymers synthesized by polymerization of functional monomers in the presence of a typically stereochemically pure template. The chiral recognition is based on the spatial arrangement of the interaction groups of the template and the polymer. Due to the fabrication process, the application of MIPs is limited to the template itself or structurally (closely) solutes. MIPs possess a variety of applications as recently summarized [196]. For the development of an extraction material specific for (S)‐(–)−1,10‐binaphthyl‐2,20‐diamine (S‐DABN) the electrostatic interaction potentials of S‐DABN as template and the monomers acrylamide, methacrylic acid and 2‐vinylpyridine were calculated and a 1:1 complex between the template and the individual monomers were subsequently constructed [197]. Because the strongest interaction was found for methacrylic acid, the MIP was subsequently constructed with this monomer and applied to the specific extraction of S‐DABN from aqueous solutions. Although no structure of the complex between S‐DABN and the imprinted polymer was modeled the study showed that interactions between a solute and the monomers used in the polymerization process can be used for a rational design of MIPs although this does not represent the actual structure of the MIP.

4. CONCLUSIONS AND OUTLOOK

As can be concluded from the examples compiled in the present short review, the desire for an understanding of the mechanism underlying chiral separations is still thriving. While a fairly good body of evidence for the interaction between a solute and a chiral selector has been accumulated for some selectors such as the most frequently applied polysaccharide‐derived selectors or CDs, understanding of other groups of selectors is still scarce. New analytes are constantly tested so that also further studies of well‐established selectors are ongoing. Despite the large number of known selectors, new types are investigated including their chiral recognition mechanism.

For the understanding of selector‐selectand interactions at the molecular level, mainly NMR spectroscopic techniques and molecular modeling methods are employed. Especially when combined with CE, NMR data can be very useful because the measurements can be performed under conditions comparable to the separation system. Of course, this is not possible for macromolecule selectors. With regard to molecular modeling, computational methods become more and more sophisticated as well as widespread and easy to use. However, especially the ease of application appears problematic as it is “seductive” to calculate the structure of a complex. In order to be meaningful, the separation conditions (mobile phase, BGE, etc.) and the correct structure of the selector must be considered. Otherwise, conclusions from such data may be misleading as the separation system is not adequately represented. For example, this has not been solved for randomly substituted CDs. Furthermore, binding of a selector to a solid support can also affect the accessibility of the chiral recognition sites of the selector. Especially challenging are also ternary systems such as the addition of CILs or chiral DES to separation systems containing another chiral selector. Consequently, modeling data should be carefully interpreted with regard to conclusions of the enantioseparation mechanism at the molecular level. The “dream” of an analytical chemist is the prediction of enantioseparations by a given chiral selector or using software to select the appropriate technique and conditions for an analyte. Despite the progress made in understanding chiral separation mechanisms, this is still in the far future.

CONFLICT OF INTEREST STATEMENT

The author declares no conflict of interest.

Scriba GKE. Update on chiral recognition mechanisms in separation science. J Sep Sci. 2024;47:2400148. 10.1002/jssc.202400148

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no new data were created or analyzed in this study.