**Colistin Sulfate Chiral Stationary Phase for the Enantioselective Separation of Pharmaceuticals Using Organic Polymer Monolithic Capillary Chromatography †**

#### **Ali Fouad 1,2, Montaser Sh. A. Shaykoon <sup>2</sup> , Samy M. Ibrahim 3, Sobhy M. El-Adl <sup>3</sup> and Ashraf Ghanem 1,\***


Academic Editors: Maria Elizabeth Tiritan, Madalena Pinto and Carla Sofia Garcia Fernandes Received: 16 January 2019; Accepted: 21 February 2019; Published: 26 February 2019

**Abstract:** A new functionalized polymer monolithic capillary with a macrocyclic antibiotic, namely colistin sulfate, as chiral selector was prepared via the copolymerization of binary monomer mixtures consisting of glycidyl methacrylate (GMA) and ethylene glycol dimethacrylate (EGDMA) in porogenic solvents namely 1-propanol and 1,4-butanediol, in the presence of azobisiso-butyronitrile (AIBN) as initiator and colistin sulfate. The prepared capillaries were investigated for the enantioselective nano-LC separation of a group of racemic pharmaceuticals, namely, α- and β-blockers, anti-inflammatory drugs, antifungal drugs, norepinephrine-dopamine reuptake inhibitors, catecholamines, sedative hypnotics, antihistaminics, anticancer drugs, and antiarrhythmic drugs. Acceptable separation was achieved for many drugs using reversed phase chromatographic conditions with no separation achieved under normal phase conditions. Colistin sulfate appears to be useful addition to the available macrocyclic antibiotic chiral phases used in liquid chromatography.

**Keywords:** colistin sulfate; enantioselective; encapsulation; capillary chromatography; monolith; organic polymer

#### **1. Introduction**

Most of the drugs currently in use are worldwide marketed as racemates. Enantiomers can exhibit different activities in biological systems, in particular, their pharmacology, toxicology, pharmacokinetics and metabolism. Therefore, it is important to separate single enantiomers to limit side effects that may arise from unwanted enantiomers [1–3]. To access enantiomerically pure compounds, enantioselective chromatographic techniques have been considered as the most feasible method compared to other more expensive and time-consuming approaches [4–10]. Among these techniques, High Performance Liquid Chromatography (HPLC) is the most widely used technique in enantiomer separation [11–13]. In HPLC, a chiral selector is required to form a Chiral Stationary Phase (CSP), the main driver for the chiral separation. The CSP is normally bound, immobilized adsorbed or otherwise attached to an appropriate support. The enantiomers are then resolved by the formation of temporary diastereomeric

complexes between the analyte and the CSP. The stationary phase support plays a very important role in any research investigation in this field [14–16].

Because of their advantages, the use of monoliths as stationary phases for HPLC represents a promising alternative to particle packed columns for CEC, conventional HPLC columns and nano-HPLC capillaries [17–21]. The preparation of organic polymer monolithic stationary phases via surface modification with a suitable precursor followed by the polymerization process results in increased stability of the monolith and affords greater adherence to the confining wall [22–28]. Many CSPs have been previously reported attached to monolithic support for chiral separation, especially in enantioselective capillary chromatography [29–31]. However, only a few were macrocyclic antibiotics [32].

Natural macrocyclic antibiotic materials play a very important role in chiral separation as useful CSPs. In general, the macrocyclic antibiotics most widely used as chiral selectors are vancomycin, vancomycin aglycon, norvancomycin, teicoplanin, and teicoplanin aglycon, ristocetin A, thiostrepton, rifamycin, kanamycin, streptomycin, fradiomycin, eremomycin and avoparcin [33]. The unique features of these chiral selectors include different chiral centers, inclusion cavities, phenyl rings, several hydrogen donor and acceptor sites, sugar moieties, and other groups which are the main drivers for their good chiral recognition abilities in different chromatographic modes. The chiral recognition mechanism in most of these antibiotics chiral selectors relies heavily on complexation, hydrogen bonding, inclusion complex formation, dipole interactions, steric interactions, and anionic and cationic binding. These chiral selectors have been employed for the enantiomeric resolution of a variety of racemates in HPLC, CEC, and CE [34–37]. Furthermore, a few were previously used in preparation of chiral monolithic columns for the enantioselective separation of racemic pharmaceuticals [38]. Colistin sulfate represents a new addition to the macrocyclic antibiotic family enabling its multi-chirality sites and functional groups to provide chromatographic interactions with racemic analytes [38]. Furthermore, the encapsulation of a macrocyclic antibiotic in an organic polymer monolith in capillary HPLC hasn't been previously reported. The ease of in situ preparation in capillaries or narrow channels of microfluidic devices render these ideal stationary phases for microscale separation formats [39].

Here we introduce a new chiral macrocyclic antibiotic, namely colistin sulfate, encapsulated in organic polymer monolithic capillary for the enantioselective nano-liquid chromatographic separation of a set of racemic pharmaceuticals.

#### **2. Results**

#### *2.1. Preparation and Characterization of Polymer Monoliths*

The use of macrocyclic antibiotics in chiral separation was previously reported in both conventional HPLC and CEC. In particular, macrocyclic antibiotic-based silica monolithic columns were previously studied [32,38,40]. However, the macrocyclic antibiotics were immobilized on the activated monoliths by a tedious reductive amination process [32]. No work was previously reported on the polymer monolithic antibiotic-based CSP in capillary liquid chromatography. Here we report the first use of colistin sulfate as a macrocyclic antibiotic chiral selector entrapped in organic polymer monolith for enantioselective capillary LC (Figure 1). The miscibility and solubility of colistin sulfate was tested in porogenic solvents used in monolith preparation, namely 1,4-butanediol, ethanol and *n*-propanol. When 1,4-butanediol was used in the polymerization mixture, a highly homogeneous solution occurred, however, better solubility was achieved when used in combination with 1-propanol as porogenic solvent.

Colistin sulfate-based polymer monolithic column (C1) was prepared via in situ copolymerization of colistin sulfate with monomers (40%) (GMA (20%) as a functional monomer and EGDMA (20%) as a cross linker) in the presence of a ternary porogenic system composed of 1-propanol (48%), 1,4-butanediol (6%) and chiral selector (6%). The ratio of the monomers to the porogens was fixed at

40:60 *w*/*w*, respectively; this was selected to provide columns with a good balance of permeability, surface area and mechanical stability.

**Figure 1.** Chemical structure of colistin sulfate (**A**) and teicoplanin (**B**).

Scanning Electron Microscopy (SEM) and Surface Properties of the Monoliths

Scanning electron microscopy (SEM) photos were taken to study the morphology of the prepared monolith. Column C1 showed a porous structure with interconnecting channels allowing the flow of mobile phase with reduced column back-pressure (Figure 2). The textural surface properties of the monolithic columns, including the specific surface area and the pore structure, were previously calculated by our group. The pore size distribution was determined from the adsorption isotherms using the Barrett–Joyner–Halenda (BJH) method. Specific surface area (SBET) was calculated using multi-point adsorption data from a linear segment of the N2 adsorption isotherms using the Brunauer–Emmett–Teller (BET) theory [41]. The monolithic column previously prepared using similar procedure and demonstrated good enantioseparation exhibited surface area of 28.67 m2/g, pore size of 169.2 nm and total pore volume of 0.12 cm3/g.

**Figure 2.** SEM of column C1 at 1200× and 25,000× (left and right, respectively) shows small micro-globules with rough surface.

Elemental analysis was used to determine the nitrogen (1.9 and 2.68 % *w*/*w*) and sulfur (0 and 0.4% *w*/*w*) content in the C1 column and blank column (G column), respectively. The blank column (G column) was prepared using the same polymerization mixture with addition of water instead of colistin sulfate. The measured the nitrogen contents were 1.9 and 2.68 % *w*/*w* and the measured sulfur contents were 0 and 0.4% *w*/*w* in the C1 column and blank column (G column), respectively. Elemental analysis was conducted to ensure the relevance of the presence of colistin sulfate in the prepared C1 column. These results confirm the presence of the chiral selector in the prepared C1 column.

#### *2.2. Enantioseparation of Different Classes of Pharmaceutical Racemates*

The colistin sulfate-based polymer monolithic capillary column was prepared as described above and investigated for the nano-LC enantioseparation of a set of different classes of racemic pharmaceuticals, namely: β-blockers, α-blockers, anti-inflammatory drugs, antifungal drugs, norepinephrine-dopamine reuptake inhibitors, catecholamines, sedative hypnotics, antihistamines, antibacterial drugs, anticancer drugs and antiarrhythmic drugs. Although reversed phase enantio-selective LC examples are limited, macrocyclic antibiotics were previously used in enantioseparation chromatography under reversed phase chromatographic mode [34,36–38,42–45]. The initial mobile phase selected for the enantioseparation separation of racemates **1**–**37** (Figure 3) was a binary mixture of methanol/water screened from 95:5 to 5:95 *v*/*v* at 1 mL/min flow rate at fixed UV detection 219 nm with eleven compounds separated (Rs ≥ 1) (Table 1). For examples, in MeOH/H2O 80:20 *v*/*v*, only ibuprofen (**7**) was separated, while in MeOH/H2O 40:60, indoprofen (**10**), hexaconazole (**15)** and miconazole (**16**) were separated. In MeOH/H2O 10:90 *v*/*v*, aminoglutethimide (**22),** tyrosine (**29**) and *O*-methoxymandelic acid (**34**) were also separated. The addition of an additive, namely triethylamine (TEA) 1% *v*/*v* in 10:90, resulted in the separation of acebutolol (**4**) normetanephrine (**21**), propafenone (**26**), tyrosine (**29**) and 4-hydroxy-3-methoxymandelic acid (**35**) (Figure 4), while non-acceptable separations were achieved by addition of the acidic additive namely trifluoroacetic acid (TFA). In an attempt to use normal phase namely *n*-hexane/2-propanol mixture ranging from 10–90% (*v*/*v*) resulted in resolution less than 1. All chromatographic data are summarized in Table 1.


**Table 1.** Chromatographic data, separation and resolution factors for the significantly resolved compounds, using reversed mobile phases, flow rate: 1 μL/min.

**Figure 3.** Chemical structures of the investigated racemates.

**Figure 4.** Enantioselective nano-LC separation of; (**a**) racemic 4-hydroxy-3-methoxymandelic acid (**35**); (**b**) phenylalanine (**30**) (mobile phase: methanol/water 40:60 *v*/*v*,); (**c**) tyrosine (**29**) and (**d**) *O*-methoxymandelic acid (**34**) on a C1 capillary column (150 μm ID, 25 cm length). UV: 219 nm, flow rate: 1 μL/min.

Because of the novelty associated with using colistin sulfate as a chiral selector, confirmatory tests were done by injecting the separated enantiomeric drugs using capillary monolithic column without chiral selector (blank column, cf. Figure 1). The injected drugs included tyrosine (**29**), phenylalanine (**30**), *O*-methoxymandelic acid (**34**) and 4-hydroxy-3-methoxymandelic acid (**35**). Only single peaks were obtained under chromatographic conditions similar to those previously used when using capillary columns with colistin sulfate as CSP (C1 column). Furthermore, the single *S*-enantiomer of acebutolol (*S*-acebutolol) was injected on the C1 column (Figure 2). Only a single peak was obtained when used alone and mixed with its racemic mixture, which resulted in a peak with higher intensity, but unfortunately with low resolution. Also it was observed that *S*-acebutolol eluted first in the same retention time as the eluted single peak of single isomer *S*-acebutolol. The results achieved from the injection of the enantiomers on both the blank and C1 column, confirm the presence of the chiral selector in situ the capillary and that it was not washed out or dissolved in the mobile phase. The investigated repeatability of the used C1 column is considered as a proof of stability of the chiral selector contained in the capillary. It was also observed that the chiral separation was mostly achieved at high water content in the mobile phase; although, the chiral selector itself can be dissolved in water. This does not contradict what has been previously reported in literature where similar solvent used for dissolving the chiral selector can be used as mobile phase in the same column [32].

The combination of thin-hair capillary format in capillary HPLC is also beneficial as swapping from existing conventional liquid chromatography LC (mL flow, more solvent) to micro/nano flow LC (less solvent) is beneficial. The expected outcome will be environmentally responsible, cost effective and efficient analytical sample preparation and separation technologies for analytical laboratories throughout the world. Some featured benefits include but not limited to: (1) up to 4× increase in sensitivity; (2) improved turn-around-time with up to 5× faster separations; (3) up to a 95% reduction in mobile phase consumption and (4) improved robustness–less sample introduced into the MS when used in LC/MS and, ultimately, less instrument downtime.

For example, the chiral analysis for one run in conventional HPLC consumes at least 20–30 mL of environmentally unfriendly solvents for 30 min separation. On the other hand, in nano-HPLC, running a similar analysis under reversed phase conditions consumes less than 100 μL of water-based mobile phase. The capillary monolithic column is 10,000 less in internal diameter and operates with one million times less solvent volume than a conventional column. Consequently, this approach is economically efficient where only a single sorbent, namely colistin antibiotic, was used as CSP in capillary HPLC reducing materials/solvents consumption. Taking ibuprofen as an example, it was efficiently separated on the prepared monolithic column (Rs = 1.02) while it was recently enantio-separated (Rs = 1.05) on a mixed sorbents containing eremomycin and bovine serum albumin BSA-based CSP under reversed phase conditions using mobile phase: MeOH:KH2PO4 (0.1 M, pH 4.5); 50:50 (*v*/*v*); flow rate: 0.5 mL/min; and fixed UV 220 nm [34]. Another example is the recent use of mixed chiral sorbents based on silica with immobilized macrocyclic antibiotics eremomycin and vancomycin for the enantioselective of β-blockers such as atenolol and amino acids like phenylalanine by conventional HPLC using a mobile phase of MeOH:ACN–TEAA (0.1%, pH 4.5) (95:5, *v*/*v*), and a flow rate of 1 mL/min [37]. Nano-HPLC chromatograms for some of the separated compounds under different ratios of methanol and water are given in Figure 4.

#### *2.3. Column Repeatability*

The repeatability is the ability to prepare equally performing capillaries at different time (run to run) and locations (batch to batch). To determine the repeatability of the prepared capillaries, two capillaries were prepared on the same day using the same polymerization mixture to test column-to-column repeatability. Moreover, batch-to-batch repeatability was tested by preparing three different batches at different days using the same polymer mixtures. 4-Hydroxy-3-methoxymandelic acid (**35**) was selected to test the capillaries' performance in terms of repeatability as it was nearly baseline resolved on both columns. Reproducibility of the retention times of both 4-hydroxy-3-methoxymandelic acid (**35**) peaks was satisfactory. In the run-to-run repeatability using one column, the average retention times for the two peaks were 23.5 min (RSD = 1.7%) and 30.6 min (RSD = 1.27%); respectively. In column-to-column repeatability, the average retention times for the two peaks are 23.5 min (RSD = 2.2%) and 30.6 min (RSD = 1.9%); respectively. In batch-to-batch repeatability, the average retention times for peak one and peak two are 22.5 min (RSD = 3.9%) and 31.4 min (RSD = 5.3%); respectively. The retention times and relative standard deviations (RSD) of the retention times ranged between 1.2% and 5.3%. These results suggest that the monolithic capillary columns can be used for reproducible routine analysis. It is worth mentioning that the acceptable %RSD values for intra-batch and inter-batch are 2.5% and 15%; respectively. Furthermore, the column loadability was tested by injecting more than 300 runs on the same column; 4-hydroxy-3-methoxymandelic acid (**35**) was injected in different orders started at run number 160 and ended by run number 307. The same separation was achieved (Figure 5).

**Figure 5.** The loadability of the monolithic columns of 4-hydroxy-3-methoxymandelic acid (**35**) started at run no. 160 up to run no. 307, on C1 capillary column (150 μm ID, 25 cm length). mobile phase: methanol/water 40:60 *v*/*v*, UV: 219 nm, flow rate: 1 μL/min.

#### *2.4. Effect of the Concentration of Chiral Selector*

The optimum concentration of the colistin sulfate in the polymerization mixture was selected after the preparation of three different capillaries with different concentrations of colistin sulfate (10, 20 and 30 mg/mL). The results revealed that 10 mg/mL afforded better separation and resolution while increasing the concentration to 30 mg/mL or more resulted in poor separation and resolution.

#### **3. Discussion**

Various macrocyclic antibiotics have been previously synthesized and applied on silica or polymer surfaces as a stationary phase either by immobilization, coating or by covalent bonding [24,28,32–36]. Whilst coating or physical adsorption is considered an suitable method to prepare CSPs, covalent bonding increases the chances for using diverse mobile phases and creates a more robust CSP [46]. It is worth pointing out that most of the CSPs have been prepared via immobilization to bond the chiral selectors to the solid supports. This has resulted in robust and more stable CSP, however, time consuming process offering less coverage of the CS compared to the one pot technique [47]. The way the CSP has been prepared (coating vs. bonding) may affect the chiral recognition mechanism. Thus, bonded-type phase showed a lower chiral recognition power than the coated-type phase.

Schmid et al. have been reporting since 2006 the development of dynamically-coated chiral stationary phases [48] using a macrocyclic antibiotic, namely vancomycin. Few macrocyclic antibiotics were previously used in preparation of chiral monolithic columns for the enantioselective separation of racemic pharmaceuticals [38]. Of interest, in 2010 Schmid et al. [32] published an article describing the preparation of a new chiral stationary phase by dynamic coating of a reversed-phase HPLC monolithic column with vancomycin-derivatives as chiral selector. They then investigated the separation of amino acids using reversed phase chromatographic conditions, namely triethyl-ammonium acetate (TEAA) buffer and methanol. As the underivatized vancomycin is hydrophilic, it could not be adsorbed on the commercial hydrophobic chromolith monolith. Consequently, vancomycin was derivatized to *N*-(2-hydroxydodecyl)-derivative before immobilization on the chromolith. Vancomycin is reversibly adsorbed via a hydrophobic side chain to the reversed-phase material. Similarly, Haroun et al. [49] dynamically coated the macrocyclic antibiotic teicoplanin on RP18 and RP8 columns. Teicoplanin has a hydrophobic C10 side chain which is attached to the glucopyranosyl group (Figure 1). This side chain was used to immobilize the chiral selector on the hydrophobic reversed phase material. This dynamically coated phase was used for the separation of aromatic amino acids. Similary, in this manuscript, colistine possesses a C9 hydrophobic side chain that can be used for the immobilization on the hydrophobic monolith prepared in this manuscript. It is worth to note that (1) continuous polymers formed from hydrophobic monomers can be used as stationary phase in reversed phase chromatography (RPC) and (2). Solvents used for dissolving the chiral sector can be used as mobile phase (not in excess) on the same column [32].

The chiral recognition of macrocyclic antibiotics used as chiral selectors for the enantio-separation of different compounds is due to the presentation of ionisable acidic or basic functional groups contributing to stereoselectivity, the presence of multiple stereogenic centers, and the presence of both hydrophobic and hydrophilic groups. Therefore, the transient non-covalent diastereomeric complexes with macrocyclic antibiotic are based on both electrostatic interactions and secondary interactions such as hydrophobic, hydrogen bonds, dipole-dipole, π–π interactions, and steric repulsion. Macrocyclic antibiotics have been successfully applied to HPLC and also to CEC for chiral separation of pharmaceutical drugs using stationary phases in the reversed-phase and the normal-phase modes [50–55].

Colistin sulfate has never been used as chiral selector although it possesses many points of interaction which significantly increase its enantiorecognition ability. It is well established that under reversed phase conditions, the formation of inclusion complexes within the cavity is the most predominant mechanism of retention and enantioselectivity. Moreover, the presence of different functional groups creates more points of interaction between the enantiomers and the CSP via π–π bonding, hydrogen bonding, dipole–dipole stacking, etc. which can increase the selectivity towards some analytes. For example, in miconazole (**16**), hydrophobic interactions are the prevailing CSP-analyte interactions, whereas hydrogen bonding seems to be more important in the enantiointeractions between the other analytes and CSPs [32]. Initial testing with mixture of methanol-based mobile phase, enantioselective separation was observed for many analytes with polar groups including acebutolol (**4**), indoprofen (**10**), hexaconazole (**15**), normetanephrine (**21**), aminoglutethimide (**22**), propafenone (**26**), tyrosine (**29**), *O*-methoxymandelic acid (**34**) and 4-hydroxy-3-methoxymandelic acid (**35**). This confirms the importance of solvent polarity in the chiral separation mechanism in terms of the inclusion complex stability. The large retention times observed is due to the very low flow rate used. Higher flow rate may result in high backpressure. Peak tailing of the more retained isomers was overcome by the use of mobile phase additives such as triethanolamine (TEA), which resulted in improved resolution, although, their negative effect on the lifetime of the capillary columns as well as its potential problems with nano-LC systems (e.g. precipitation in the pumps and valves) [56] can be dominant. No remarkably peak tailing of acebutolol (**4**), atenolol (**5**) and tyrosine (**29**) racemates was observed, ascribed to the existence of the amino groups on the surface of the monolithic matrices. It was also observed that the chiral separation was mostly obtained at high water content of the mobile phase; this indicates that water facilitates the interaction between the CSP and the racemates. We postulate that chiral separation in this study was mainly achieved via the formation of inclusion complexes as discussed previously. The use of normal organic phase resulted in high back pressure and very short life time of the prepared column. Nevertheless, the use of *n*-hexane/2-propanol mobile phase mixture ranging from 10–90% (*v*/*v*) resulted in resolution less than 1.

#### **4. Experimental**

#### *4.1. Reagents and Materials*

Colistin sulfate (99%), ethylene glycol dimethacrylate (EGDMA, 98%), glycidyl methacrylate (GMA, 98%), 1-propanol (99%), 1,4-butanediol (99%), trifluoroacetic acid (TFA, ≥99.5%), sodium hydroxide and hydrochloric acid were purchased from Sigma Aldrich (Milwaukee, WI, USA). Acetone (AR grade) and ethanol (HPLC grade) were purchased from BDH (Kilsyth, Vic., Australia). Methanol (HPLC) grade was purchased from Scharlau (Sentmenat, Spain). All other reagents were of the highest available grade and used as received. The fused-silica capillaries (150 μm internal diameter) were purchased from Polymicro Technologies (Phoenix, AZ, USA). 2,2-Azobis(isobutyronitrile) (AIBN) was obtained from Wako (Osaka, Japan). Water used for dilutions and experiments was purified by a Nano-pure Infinity water system (NJ, USA). The racemic analytes were mostly purchased from Sigma Aldrich.

#### *4.2. Preparation and Characterization of the Monolithic Columns*

#### 4.2.1. Activation of the Fused Silica Capillaries

Briefly, the fused silica capillaries were rinsed using a Harvard syringe pump (Harvard Apparatus, Holliston, MA, USA) and a 250 μL gas-tight syringe (Hamilton Company, Reno, NE, USA) with acetone and water 2–3 times each, activated with 0.2 mol/L sodium hydroxide (NaOH) for 6 h confirming the absence of any air bubbles, washed with water 3–4 times till neutral (pH 7), then washed with 0.2 mol/L hydrochloride (HCl) for 12 h, rinsed with water and ethanol 2–3 times each. A 20% (*w*/*w*) solution of 3-(trimethoxysilyl)propyl methacrylate in 95% ethanol adjusted to pH 5 using acetic acid was pumped through the capillaries at a flow rate of 0.25 μL/min for 6 h. The capillary was then washed with acetone one time and dried with a stream of nitrogen for 2 min. then left at room temperature for 24 h.

#### 4.2.2. Preparation of Colistin Sulfate Functionalized Monomer

The short (∼25 cm in length) surface modified capillary was filled by Harvard syringe pump with the degassed polymerization mixture at 0.25 μL/min using the syringe pump. Colistin sulfate polymer-based monolithic capillary column was prepared via in situ copolymerization of binary monomer mixtures consisted of GMA (20%) as a monomer and EGDMA (20%) as across linker along with different porogens namely; 1-propanol (48%), 1,4-butanediol (6%), in the presence of 1 wt% AIBN (with respect to monomers) and colistin sulfate (6%) as chiral selector. The blank column (G column) was prepared using the same procedure by addition of water (6%) instead of water. The filled capillaries were then sealed with a septum, placed in 70 ◦C water bath for 18 h for the polymerization reaction to take place. The unreacted monomers were removed from the monolithic columns by pumping methanol at a flow rate of 100 μL/h for 24 h before being investigated under light microscope to ensure its consistency and homogeneity of the polymerization mixture inside the capillary. The filled capillaries were conditioned with mobile phase for 1–3 days at μL/min (Figure 6). The ratios of the monomers to the porogens were kept 40% and 60%, respectively. The ratios of the porogens were fixed as 48% 1-propanol, 6% 1,4-butanediol and 6% chiral selector, all percentages are *w*/*w*.


**Figure 6.** Steps showing the preparation of polymer monolithic capillary columns.

#### 4.2.3. SEM of the Prepared Monoliths

SEM was performed to study the morphology of the prepared capillaries. The capillaries were cut into ~1 cm sections and put perpendicularly on 12.7 mm pin-type aluminum stub using double face epoxy resin tape. SEM was carried out and high resolution images were collected by sputter coating the capillary sections with gold Using ZEISS SIGMA FE-SEMs for High Quality Imaging and Advanced Analytical Microscopy (ZEISS Sigma, Jena, Germany).

#### *4.3. Instrumentation*

A nano-liquid chromatographic system consisting of an LC-10AD VP pump (Shimadzu, Kyoto, Japan), injector model 7725i-049 (Rheodyne, Park Court, CA, USA), a UV-Vis detector model MU 701 UV-VIS (GL Science, Tokyo, Japan) and a Shimadzu CDM-20A communications bus module was used. The system flow was split after direct injection. The data was processed by the Shimadzu Lab-Solutions software version 5.54 SP2 (Shimadzu, Kyoto, Japan).

#### *4.4. Standard Solutions and Sample Preparation*

Stock solutions of the racemic analytes at concentrations of 1 mg/mL in filtered HPLC grade methanol were prepared. Prior to injection, the stock solutions were further diluted 10× by mobile phase and filtered through Minisart RC 15 0.2 μm pore size filters (Sartorius, Goettingen, Germany). Tested compounds: β-blockers: alprenolol (**1**), metoprolol (**2**), propranolol (**3**), acebutolol (**4**), atenolol (**5**); α-blockers: naftopidil (**6**); anti-inflammatory drugs: ibuprofen (**7**), naproxen (**8**), flurbiprofen (**9**), indoprofen (**10**), cizolirtine (**11**), cizolirtine citrate (**12**), carprofen (**13**), glafenine (**14**); antifungal drugs: hexaconazole (**15**), miconazole (**16**), diniconazole (**17**) sulconazole (**18**); norepinephrine-dopamine reuptake inhibitor: nomifensine (**19**); catecholamines: arterenol (**20**), normetanephrine (**21**); sedative

hypnotics: aminoglutethimide (**22**); anti-histamines: chlorpheneramine (**23**); anticancer drugs: ifosfamide (**24**); antiarrhythmic drugs: tocainide (**25**), propafenone (**26**); flavonoids: flavanone (**27**); amino acids: glutamic acid monohydrate (**28**), tyrosine (**29**), phenylalanine (**30**); anti-platelet agents: clopidogrel (**31**); immunomodulatory drugs: thalidomide (**32**); miscellaneous: 1-acenaphthenol (**33**), *O*-methoxymandelic acid (**34**) 4-hydroxy-3-methoxymandelic acid (**35**), 1-indanol (**36**) and ampicillin (**37**). The chemical structures of the investigated racemates are shown in Figure 3.

#### *4.5. HPLC Conditions*

The mobile phase consisted of water/methanol (*v*/*v*) for the reversed phase HPLC and *n*-hexane/2-propanol for normal phase HPLC. For all samples, the injected volume was 0.2 μL at room temperature with flow rate 1 μL/min on C1 capillary column (150 μm ID, 25 cm length). Preliminary UV analyses were performed at a wavelength of 219 nm.

#### **5. Conclusions**

The macrocyclic antibiotic colistin sulphate has been used for the first time as a chiral selector entrapped in a polymer monolith for enantioselective capillary chromatography. The new capillary column was investigated for the enantioselective separation of a set of racemic drugs. Acceptable separation was achieved for many drugs, including acebutolol (**4**), ibuprofen (**7**), indoprofen (**10**), hexaconazole (**15**), miconazole (**16**), normetanephrine (**21**), aminoglutethimide (**22**), propafenone (**26**), tyrosine (**29**), *O*-methoxymandelic acid (**34**) and 4-hydroxy-3-methoxymandelic acid (**35**) under reversed phase chromatographic conditions, while normal phase conditions did not achieve any acceptable separations. The method provides more economical analysis under environmentally benign reversed phase conditions.

**Author Contributions:** Conceptualization, M.S.A.S., S.M.I. and S.M.E.-A.; Methodology, A.F.; Software, A.F. and A.G.; Validation, A.F. and A.G.; Formal analysis, A.F. and A.G.; Investigation, A.F. and A.G.; Resources, A.F. and A.G.; Data curation, A.F.; Writing—original draft preparation, A.F.; Writing—review and editing, A.F.; Visualization, A.G.; Supervision, M.S.A.S., S.M.I. and S.M.E.-A.; Project administration, S.M.E.-A. and A.G.; Funding acquisition, M.S.A.S.

**Funding:** This research was supported, in part, by Al-Azhar University, Egypt and University of Canberra, Australia. Funding was supported by the Egyptian Cultural and Educational Bureau, Minister of Higher Education Egypt (Cultural Affairs and Missions Sector) as Ph.D. joint mission stipend offered to Ali Fouad.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Samples of the compounds used in this research are available from the authors.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **New Brush-Type Chiral Stationary Phases for Enantioseparation of Pharmaceutical Drugs**

**Anamarija Kneževi´c 1,\* , Jurica Novak 2,3 and Vladimir Vinkovi´c <sup>1</sup>**


Academic Editors: Maria Elizabeth Tiritan, Madalena Pinto and Carla Sofia Garcia Fernandes Received: 25 January 2019; Accepted: 21 February 2019; Published: 25 February 2019

**Abstract:** The importance of chirality in drug development is unquestionable, with chiral liquid chromatography (LC) being the most adequate technique for its analysis. Among the various types of chiral stationary phases (CSPs) for LC, brush-type CSPs provide the base for interaction analysis of CSPs and enantiomers, which provide valuable results that can be applied to interaction studies of other CSP types. In order to analyze the influence of aromatic interactions in chiral recognition, we designed a set of ten new brush-type CSPs based on (*S*)-*N*-(1-aryl-propyl)-3,5-dinitrobenzamides which differ in the aromatic unit directly linked to the chiral center. Thirty diverse racemates, including several nonsteroidal anti-inflammatory drugs and 3-hydroxybenzodiazepine drugs, were used to evaluate the prepared CSPs. Chromatographic analysis showed that the three new CSPs separate enantiomers of a wide range of compounds and their chromatographic behavior is comparable to the most versatile brush-type CSP—Whelk-O1. The critical role of the nonbonding interactions in positioning of the analyte (naproxen) in the cleft of **CSP-6**, as well as the analysis of interactions that make enantioseparation possible, were elucidated using computational methods. Furthermore, the influence of acetic acid as a mobile phase additive, on this enantiorecognition process was corroborated by calculations.

**Keywords:** chiral chromatography; chiral recognition; intermolecular interactions; chiral drugs; Whelk-O1 column; mobile phase additives

#### **1. Introduction**

Chirality is an essential property in the development of pharmaceutical drugs, as well as in agrochemistry, food science, etc. It was put into the foreground in 1992 when the FDA issued its policy statement concerning the development of stereoisomeric drugs [1]. Even though the development of enantiopure compounds was not mandatory, it became the de facto standard in the pharmaceutical industry, also enabling the new possibilities in drug development, i.e., the chiral switch concept [2]. Accordingly, the need for fast, simple and reliable methods for separation of enantiomers, determination of enantiomeric excess (*ee*) and absolute configuration (AC) has increased. Although chiral liquid chromatography (chiral LC) has a great influence on the determination of absolute configuration [3], its main role is being the most powerful method for the separation of enantiomers [4] and for measuring the enantiopurity of organic compounds. Since the 1990s, when the availability of chiral stationary phases (CSPs) became widespread, chiral LC using CSPs has been the most widely used technique for chiral separations [5]. Besides being simple, versatile and reliable

method, chiral LC has also became an ultrafast technique with enantioseparations in the sub-minute time frame [6–12].

Over time, various types of CSPs have been developed, including more than a hundred which have been commercialized. Their properties and mechanism of enantioseparation depend on the nature of the chiral selector [5,13]. Polysaccharide-based CSPs are the most broadly applied for chiral LC separations [14]. However, there are also several Pirkle-type (or brush-type) CSPs, e.g., Whelk-O1 and ULMO, whose versatility has enabled their wide application [13,15]. These CSPs, which consist of small organic molecules covalently bound to the support, are also the most widely investigated CSPs regarding the chiral recognition mechanism [13,16]. The understanding of chiral separation mechanism is essential for estimating elution order, predicting the types of analytes which can be separated on a certain CSP, and improving the design of new selectors. Clearly, the enantioseparation process is easier to elucidate by analyzing interactions of the analyte and a small selector compared to polysaccharide CSPs, whether using experimental or computational methods [5]. Chiral recognition of brush-type CSPs is based on well-documented interactions between the selector and enantiomers of the analyte—hydrogen bonds, dipole-dipole interactions, van der Waals interactions and, in particular, aromatic interactions [16].

The goal of this research was threefold. First, ten new brush-type CSPs were prepared, with a molecular design based on previous studies of chiral recognition in CSP [13,17–19]. The objective was to obtain a CSP with a versatility as similar as possible to the Whelk-O1 column, the brush-type CSP with the broadest application in industry and academia. We opted for a structure with one amide bond, analogous to the Whelk-O1 column, and two aromatic groups—a 3,5-DNB aromatic unit and a substituted aromatic moiety (Figure 1). We tested the prepared CSPs using thirty diverse racemates and determined the CSPs with the best performance. Second, the variability of the aromatic moiety enabled us to elucidate the influence of aromatic substituent on the enantiorecognition process. Keeping that in mind, we investigated the role of nonbonding interactions relevant for chiral recognition on CSPs using computational methods, with special attention to hydrogen bond networks and aromatic interactions. Third, we explored the influence of acetic acid as an additive in the mobile phase using extensive ab initio methods. To the best of our knowledge, this is the first study that examines the possibility of positioning a minor additive within the chiral binding site and its influence on enantiospecific interactions.

**Figure 1.** Structures of new brush-type CSPs and Whelk-O1.

#### **2. Results and Discussion**

#### *2.1. Preparation of New Brush-Type Chiral Stationary Phases (CSPs)*

We previously described the synthesis and determination of an absolute configuration of (*S*)-*N*-(1-aryl-allyl)-3,5-dinitrobenzamides (**DNB-1**–**DNB-10**) [20]. These compounds are excellent candidates for the preparation of a new Pirkle-type CSPs for HPLC since they possess all properties of a good selector—they are rigid and contain a strong π-acceptor acid (DNB) as well as a π-donor base aromatic group. Furthermore, allyl group in the structure of DNBs enables their binding to silica gel using a simple and straightforward procedure.

For the binding of the chiral DNB selectors onto silica gel we chose a method that does not introduce additional polar moieties into the CSP structure (Scheme 1) [21]. Hydrosilylation of DNBs was conducted using chlorodimethylsilane and Speier's catalyst, followed by the replacement of a chloride with an ethoxy group using trimethylamine and ethanol. Obtained compounds were passed through short silica column and attached to 5 μm silica gel in boiling toluene. Before the end-capping procedure, we analyzed prepared CSPs using elemental analysis and infrared spectroscopy in order to determine the amount of selector attached onto the silica gel. We performed end-capping in the last step to protect free silanol groups with hexamethyldisilazane [22]. Ten new CSPs were thus prepared from the corresponding DNBs (Figure 1) with the optimal 0.2 mmol selector per 1 g of CSP loading [21]. The as-prepared CSPs were finally packed in steel columns using the slurry packing technique.

**Scheme 1.** Preparation of new brush-type chiral stationary phases (CSPs) from corresponding (*S*)-*N*-(1-aryl-allyl)-3,5-dinitrobenzamides (DNBs).

#### *2.2. Evaluation of Prepared CSPs*

In order to test the performance of the prepared CSPs we used a diverse set of analytes (Figure 2) which is usually used by our group to evaluate new CSPs [18,19]. The evaluation of prepared CSPs was performed in normal phase chromatography mode using the mixtures of hexane and 2-propanol as the mobile phase at room temperature and with UV detection at 254 nm. Compounds **1**–**7** and **9** were analyzed in the mobile phase containing 10% of 2-propanol in hexane, while the mobile phase for the analysis of compounds **8** and **10**–**21** contained 20% of 2-propanol due to their longer retention times. Our goal was to evaluate CSPs with respect to their difference in molecular structure and consequently to neglect the influence of quality of the column packing, column dimensions, etc. Therefore the relevant value for consideration is the separation factor (selectivity, α), which depends the most on the structure of the CSP and has the greatest influence on overall resolution.

Tested set of compounds can be subdivided into four subsets. Compounds **1**–**4** are small molecules with one or none carbonyl groups which could be the acceptor but not the donor of hydrogen bonds. Because of these characteristics, the satisfactory enantioseparation of compounds **1**–**4** is usually hard to achieve on brush-type CSPs. The second subset consists of compounds **5**–**9** which possess bigger aromatic groups and hydroxyl or amino groups capable of hydrogen bond formation. These compounds are usually well separated on brush-type CSPs. Finally, there are two subsets consisting of

aromatic amides, **10**–**21**, with the last subset (**15**–**21**) also bearing an ester group. These compounds can form strong hydrogen bonds and aromatic interactions which makes them easier to separate using brush-type CSPs.

**Figure 2.** Structures of the racemic analytes used for the evaluation of prepared CSPs.

The results indicate that CSPs which contain substituted phenyl aromatic groups, in most cases do not resolve compounds **1**–**4** (Figure 3). Enantioseparation of these compounds is better using the naphthyl series of CSPs, although the separation of compounds **3** and **4** is achieved on just a few CSPs. In general, enantioseparation results obtained for compounds **5**–**9** are better. Overall, the lowest separation factor is observed with CSPs like **CSP-3**, **CSP-4** and **CSP-5**, which have more than one methyl substituent or methyl at the *ortho* position of the phenyl core.

**Figure 3.** Separation factors of compounds **1**–**9** on prepared **CSP-1**–**CSP-10**, mobile phase hexane-2-propanol = 90:10. For compound **8**, the mobile phase ratio was 80:20.

Compounds **10**–**14** are resolved well using all CSPs, with the only exceptions being **CSP-3** and **CSP-5** (Figure 4a). These CSPs demonstrate the lowest separation factors due to their phenyl aromatic group in the combination with two or three methyl substituents near the chiral carbon atom, which disturb the enantiorecognition process due to steric hindrance. Also, the retention times of these compounds are shorter for the phenyl series of CSPs than the naphthyl series (Table S2). This means that interactions are weaker, as expected for smaller aromatic systems.

**Figure 4.** Separation factors of compounds **10**–**14** (**a**) and **15**–**21** (**b**) on prepared **CSP-1**–**CSP-10**, mobile phase hexane-2-propanol = 80:20.

The initial assumption was that 3,5-dinitrobenzoyl derivatives of amino acids with an isopropyl protecting group **15**–**21** would be resolved well on the prepared CSPs since they possess aromatic, ester and amide groups potentially capable of forming interactions with CSPs which may result in chiral recognition. However, the difference between the phenyl series of CSPs and larger aromatic substituents is even more pronounced in these examples (Figure 4b). Here, it is demonstrated that only CSPs with larger aromatic groups (naphthyl and phenantryl) show very good enantio-separations of these compounds. The lowest results in the naphthyl series were obtained for **CSP-8** which possesses a methyl group at position 2 of the naphthyl ring.

It is interesting to point out that the enantioseparation of racemate **19** on the prepared CSPs is quite unexpected (Figure 5). On the CSPs which show the overall lowest enantioseparation capabilities, its separation factor is the highest in the **15**–**21** subset. On the contrary, on the CSPs that demonstrate the best performance, this separation factor is the lowest of the subset. This result exemplifies how chiral recognition is difficult to predict and often depends on subtle details.

The above presented results demonstrate that both size and substitution of aromatic group play an important role in the enantioseparation capabilities of prepared CSP. Larger aromatic groups (naphthyl and phenanthryl) overall display higher separation factors for the tested set of racemates. This is expected since intermolecular aromatic interactions, which positively influence the chiral recognition, are stronger for larger aromatic groups. Substitution of more than one methyl group on the phenyl ring decreases the separation factor for all the tested racemates. On the other hand, the influence of the monosubstitution of the aromatic group shows two opposite trends. When the substitution is at the 4-position, there is no noticeable difference compared to the aromatic group without substituents (**CSP-1** compared to **CSP-2** and **CSP-6** to **CSP-7**). Contrarily, substitution at the 2-position greatly influences the enantioseparation capabilities of the prepared CSPs (**CSP-2** compared to **CSP-4** and **CSP-7** to **CSP-8**). The methyl substituent in this case is near the chiral center and can influence the chiral recognition in two ways—it sterically hinders the chiral center which decreases enantiorecognition and it increases the rigidity of the CSPs by reducing the rotational freedom around C\*-CAr bond (Figure 6). The influence of the rigidity of CSP on enantioseparation capability is substantial. **CSP-9** compared to **CSP-6** demonstrates higher flexibility and somewhat lower enantioseparation characteristics (see Figures 3 and 4). On the other hand, **CSP-8** compared to **CSP-6** has higher rigidity and considerably

α α

decreased enantioseparation results (which are partially due to steric hindrance of the chiral center). This indicates that a good CSP should be sufficiently rigid to strongly interact with one sterically compatible enantiomer, but also flexible enough to accommodate the analyte and achieve a maximum number of interactions for a wide range of compounds.

**Figure 5.** Chromatogram of the separation of racemate **19** on **CSP-10** (**a**) and **CSP-8** (**b**); mobile phase hexane-2-propanol = 80:20, flow 1mL/min, UV detection at 254 nm.

**Figure 6.** Relaxed potential energy surface scan around C\*-C<sup>α</sup> bond of **DNB-6** (red), **DNB-8** (yellow) and **DNB-9** (blue). Dihedral angle Φ is defined by atoms Cβ-Cα-C\*-N. Calculations were performed on M06-2X/aug-cc-pVDZ level of theory.

#### *2.3. Enantioseparation of Pharmaceutical Drugs on Prepared CSPs*

Since CSPs bearing naphthyl and phenanthryl aromatic groups were demonstrated as more versatile than phenyl CSPs, **CSP-6**–**CSP-10** were chosen for testing the enantioseparation of several pharmaceutical drugs including non-steroidal anti-inflammatory drugs (NSAIDs) and 3-hydroxy-benzodiazepine drugs (Figure 7).

**Figure 7.** Structures of the tested non-steroidal anti-inflammatory drugs (NSAIDs) and 3-hydroxybenzodiazepine drugs.

α α Given that NSAIDs are organic acids, acetic acid (0.1%) was selected as the mobile phase additive. Acidic additives are frequently used in the mobile phases for the analysis of acidic analytes because of their positive influence on the peak shape and retention time [23]. Results of the enantioseparation of NSAIDs on the prepared **CSP-6**–**CSP-10** (Figure 8a) showed that only naproxen is resolved well on prepared CSPs, while some of the drugs from our test set are only partially resolved on **CSP-7**.

**Figure 8.** (**a**) Separation factor of NSAIDs on prepared **CSP-6**–**CSP-10** using a mobile phase hexane-2-propanol-CH3COOH = 90:10:0.1 (hollow symbols) and hexane-2-propanol = 80:20 + 1 g dm−<sup>3</sup> NH4OAc (filled symbols); (**b**) Separation factor of compounds **8**, **9** and 3-hydroxy-benzodiazepine drugs on prepared **CSP-6**–**CSP-10** using a mobile phase hexane-2-propanol = 80:20 + 1 g dm−<sup>3</sup> NH4OAc.

It was demonstrated that in some cases neutral salts, such as ammonium acetate or ammonium formate, have a positive influence on the enantioseparation of racemates [23]. For example, NSAIDs showed excellent enantioseparation results on Whelk-O1 columns using a mobile phase with ammonium acetate as an additive [24]. Therefore, we investigated the enantioseparation performance of the newly prepared **CSP-6**–**CSP-10** on NSAIDs also using ammonium acetate as a mobile phase additive (Figure 8a). Although **CSP-8** and **CSP-9** still only separate the enantiomers of naproxen, the performance of the remaining CSPs was substantially improved. This is especially evident in the case of **CSP-10** (Figure 9) which can separate all of the tested NSAIDs using the abovementioned conditions.

Compounds **8** and **9**, which are structural analogs of 3-hydroxybenzodiazepines with an ester group at the 3-position, were resolved well using the prepared CSPs. Therefore, we decided to test the naphthyl series of CSPs for the enantioseparation of 3-hydroxybenzodiazepine drugs (Figure 8b). Since these compounds possess acidic and basic groups, ammonium acetate was also used as a mobile phase additive. 3-Hydroxybenzodiazepine drugs show very good enantioseparation results on the prepared CSPs, with the best results obtained on **CSP-6**, **CSP-7** and **CSP-10**.

**Figure 9.** Chromatograms of the separation of flurbiprofen on **CSP-10** using mobile phase: hexane-2-propanol-CH3COOH = 90:10:0.1 (**a**) and hexane-2-propanol = 80:20 + 1 g dm−<sup>3</sup> NH4OAc (**b**).

Overall, the best separation factors for the enantioseparation of the tested drugs were obtained on **CSP-6**, **CSP-7** and **CSP-10**. Given the importance of these compounds in the pharmaceutical industry, their enantioseparation was previously analyzed on several commercial brush-type CSP, including Whelk-O1 [24]. In order to compare the results obtained on our CSP with the results obtained for Whelk-O1, we analyzed these pharmaceutical drugs using the same hexane—2-propanol = 80:20+1g dm−<sup>3</sup> NH4OAc mobile phase (Table 1).


**Table 1.** Comparison of the separation factors (α) of tested pharmaceutical drugs obtained on **CSP-6**, **CSP-7**, **CSP-10** and Whelk-O1 1.

<sup>1</sup> For the analyses (*S*,*S*)-Whelk-O1 column (5 <sup>μ</sup>m particle size, 250 × 4.6 mm I.D.) was used.

From the obtained results it is evident that these three new CSPs have promising enatioseparation capabilities which are comparable to those of Whelk-O1. Furthermore, it must be emphasized that the column Whelk-O1, being a commercial one, is technologically optimized regarding the choice of silica gel used as the support, the binding procedure, as well as the packing procedure. It is known that all of these conditions influence the enantioseparation capabilities of CSPs [25,26]. The prepared CSPs haven't been optimized at this point, leaving plenty of room for improvement of these CSPs. These columns have one more advantage over Whelk-O1. In order to prepare Whelk-O1 enantiopure selector, preparative chiral chromatography must be used which has low productivity due to the low solubility of the compound [27,28]. On the other hand, the enantiopure selectors of our CSPs (**DNB-6**, **DNB-7** and **DNB-10**) can be obtained by enzyme resolution using a cheap commercial lipase—*Candida antarctica* lipase B (CAL-B) [20,29].

#### *2.4. Analysis of Interactions Responsible for Chiral Recognition between Naproxen and* **CSP-6**

To explore the nature of chiral binding site, a set of molecular dynamics simulations in vacuum of **CSP-6** with naproxen was performed. Since we are interested in distinguishing the details that enable enantiomeric separation, simulations were run separately for (*S*)- and (*R*)-enantiomers of naproxen. According to a 1H-NMR study of chiral recognition of (*S*)- and (*R*)-naproxen in the presence of Whelk-O1 selector analog [30] and to the molecular dynamics simulations of the interface of modified Whelk-O1 selector and naproxen in *n*-hexane [31], both enantiomers of naproxen dock inside the cleft, most probably by an M1 mechanism [32,33]. The main feature of the M1 mechanism is a hydrogen bond between the drug and the amide hydrogen, while aromatic interactions with the dinitrophenyl moiety introduce additional stability to the complex. For simplicity in our computational models we have neglected the non-polar solvent and a linker, the interface to silica gel, substituting it with a hydrogen atom.

The complex of **CSP-6** and (*S*)-naproxen ((*S*)-**A**) has lower energy (by 1.8 kcal mol−1) than the lowest energy complex of **CSP-6** and (*R*)-naproxen ((*R*)-**A**) (Figure 10). This result is in accordance with the experimental results which demonstrate that (*R*)-naproxen is the first eluting enantiomer on **CSP-6** (Figure S1). Ten additional complexes within 2.9 kcal mol−<sup>1</sup> were examined (Figure S2). In (*S*)-A complex the naphthyl and DNB subunits are almost perpendicular to each other. Naproxen is hydrogen bonded to the amide hydrogen (2.00 Å), with the carbonyl oxygen being a H-bond acceptor and in the vicinity of the β-hydrogen of DNB, being only 2.44 Å apart. The second most important H-bridge is between an oxygen of the DNB nitro group and the hydrogen from the naproxen carboxyl group. The weakest interaction of this type is responsible for anchoring the opposite side of naproxen, the methoxy moiety, to the second nitro group of the DNB subunit. The parallel face-centered stacking arrangement between the two aromatic rings with the distance between the centroids of the naproxen and DNB rings of 4.11 Å, represents a good example of aromatic interaction between strongly electron-deficient (DNB) and neutral aromatic rings (Figure 10b). The next intriguing space oriented interaction is the H-aromatic interaction [34], where three hydrogens of naproxen, including the hydrogen on the chiral carbon atom, are oriented toward the naphthyl ring, forming a recognizable T-shaped motif whose rings' planes form an angle of 75◦. Alternative modes of complexation are also found, like the one where naproxen is outside the cleft ((*S*)-**A**-4) or a complex where naproxen is hydrogen bonded to the carbonyl oxygen of **CSP-6** ((*S*)-**A**-5). Those complexes are 2.5 kcal mol−<sup>1</sup> higher in energy than (*S*)-**A**, and will not be discussed further.

Just by reversing groups at the chiral carbon on naproxen in the (*S*)-**A** complex and running optimization, a new complex is identified, (*R*)-**A**-2, 2.8 kcal mol−<sup>1</sup> higher in energy (see Supplementary Materials). The O-H···O bond is shorter by 0.10 Å, while O···H-N is 0.03 Å longer than the analogous bonds in (*S*)-**A**, but the orientation of the methyl group toward the naphthyl part of **CSP-6** perturbs the parallel face-centered stacking arrangement almost perfectly possible in the (*S*)-enantiomeric complex. The H-aromatic interaction network is also perturbed, which can be seen from the fact that only hydrogens from the α ring are pointing toward naphthyl moiety.

The (*R*)-**A** complex, more stable than (*R*)-**A**-2 by 0.9 kcal mol<sup>−</sup>1, is characterized by even stronger hydrogen bonds between the naproxen carboxyl group and DNB (1.86 Å) and the amide hydrogen (1.99 Å). The anchoring interaction between the methoxy hydrogen and the nitro group is missing, which is reflected in the increased distance between the centroids of an aromatic ring of naproxen and

the DNB subunit to 5.29 Å. By comparing the (*R*)-**A** and (*R*)-**A**-2 structures it is possible to conclude that the main contribution to the stabilization of the complexes comes from hydrogen bonding with aromatic and H-aromatic interactions playing a minor, but significant role in the enantiorecognition process.

Our findings are in agreement with similar theoretical studies on the chiral recognition of Whelk-O1. For a series of complexes of naproxen with Whelk-O1 and modified CSPs, the M1 mechanism is identified as the most probable docking mechanism [31]. Furthermore, the formation of the hydrogen bond was found to be critical for the successful separation of enantiomers [35].

**Figure 10.** The lowest energy structures of (*S*)-**A** (**a**), (*S*)-**A**, top view (**b**) and (*R*)-**A** (**c**). Selected bond lengths are indicated in Å. B3LYP/def2-TZVPP//M06-2X/def2-TZVPP (COSMO, ε = 2.78).

#### *2.5. Influence of Acetic Acid as the Mobile Phase Additive on The Enantioseparation of Naproxen on* **CSP-6**

Although it is empirically known that minor additives (organic acids, bases or neutral salts) may have positive effect on the shape of peaks, retention times and selectivity [23], experimental studies on the mechanism of how additives influence enantioseparation are scarce. Several systematic experimental studies were performed, mostly on polysaccharide CSPs under normal phase conditions [36–38], as well as using polar organic mobile phases [38–40]. Generally, the consensus is that additives in the mobile phase improve the solubility of acidic analytes in nonpolar solvents and suppress non-chiral interactions of analyte with the CSP surface (masking effect), and therefore reduce tailing and can increase selectivity. However, an explanation for the way in which additives could affect the chiral recognition process between the analyte and CSP remains elusive. Some studies [36,39] hypothesize that additives have an ability to displace the analyte or to influence the selector-analyte complex by interfering with the interactions that stabilize the complex, however, these speculative explanations of experimental results have not been substantiated so far.

Our goal was to investigate the possibility of influence of additives on intermolecular interactions between analyte and CSP during enantiorecognition. We opted for acetic acid, a simple acidic additive, and a system we had already investigated—naproxen and **CSP-6**. Acetic acid is known to form dimers in the gas phase and in nonpolar solvents characterized by a strong double hydrogen bond [41,42]. Since naproxen primarily interacts via its carboxylic group with CSPs, we explored the possibility of an interplay of naproxen, acetic acid and **CSP-6**.

The most stabile complex of naproxen, acetic acid and **CSP-6** is (*S*)-**B**, which is 3.9 kcal mol−<sup>1</sup> lower in energy than (*R*)-**B** (Figure 11). A common motif of both structures is an 8-membered ring formed by a double hydrogen bridge connecting acetic acid and the carboxylic acid group of naproxen. In the (*S*)-**B** complex, the N-H···O bond length is 2.90 Å, while naproxen's carbonyl oxygen serves as a double H-bond acceptor, associating with acetic acid as well. The analogous bridge to (*R*)-naproxen in (*R*)-**B** is longer (3.14 Å) and of a different nature – the H-acceptor is a carboxylic acid oxygen, which is simultaneously an H-donor to acetic acid. Although acetic acid is not connected to naproxen via enantioselective interactions, it modifies H-bond network compared to a complex without the acid ((*S*)-**A** and (*R*)-**A**). Firstly, the interaction between the carboxylate hydrogen and the nitro group is

missing. Secondly, the methyl group of acetic acid is a weak H-donor to the nitro subunit. For (*R*)-**B**, the presence of additive makes the contact between naproxen and **CSP-6** along the carboxyl oxygen more favorable.

**Figure 11.** The lowest energy structures of (*S*)-**B** (**a**) and (*R*)-**B** (**b**). Selected bond lengths are indicated in Å. B3LYP/def2-TZVPP//M06-2X/def2-TZVPP (COSMO, ε = 2.78).

Competitive interaction patterns are also found (Figure S3, Supplementary Materials). Only one structure is missing an 8-membered double H-bond ring ((*R*)-**B**-2) where a more complex H-bond network can be seen. An amide is connected to acetic acid, which is a H-donor to naproxen, whose carboxylic acid group is an H-donor to the carbonyl oxygen of **CSP-6**. Again, naproxen is positioned above the DNB subunit but outside the cleft.

Computational insights into the influence of acetic acid on the enantioseparation of NSAIDs, like naproxen, reveal that additives can interact with both CSP and carboxylic acid analytes. Even non-enantiospecific interactions with the carboxylate moiety of the drug change the interaction arrangements. In the system we studied, the entire binding scheme is shuffled, which is reflected in the increase of N-H···O distance and weakening of both the hydrogen bond and aromatic interactions. The acetic acid also influences the energetics of complexes, increasing the energy difference between the most stable (*S*)- and (*R*)-complex, enabling better separation. Our theoretical consideration of complexes of naproxen and CSPs, with and without acetic acid, indicate that higher selectivity is possible due to interactions of acetic additive with the analyte and CSP.

#### **3. Materials and Methods**

#### *3.1. General Information*

All the solvents were puriss. p.a. or HPLC grade and were used directly as supplied by Sigma-Aldrich Chemie GmbH (Munich, Germany), Alfa Aesar (Karlsruhe, Germany) or Acros (part of Thermo Fisher Scientific, Geel, Belgium). For the preparation of chiral stationary phases HPLC silica gel Separon SGX particle size 5 μm and pore size 80 Å from Tessek Ltd. (Prague, Czech Republic) was used. IR spectra were recorded on an MB102 instrument (ABB Bomem, Zurich, Switzerland). Elemental analyses were done on a 2400 CHNS Elemental Analyzer (Perkin-Elmer, Waltham, MA, USA). Chiral HPLC analysis were performed using a Prominence System (Pump LC-20AT, DGU-20A5 Degasser, UV detector SPD-20A, Shimadzu, Kyoto, Japan) or a Knauer system (Pump Knauer 64, 4-Port Knauer Degasser, UV detector Knauer Variable Wavelength Monitor, Interface Knauer, Knauer, Berlin, Germany, and a CD-2095 detector, Jasco, Easton, MD, USA). The packing of prepared CSP into stainless steel columns (150 × 4.6 mm I.D.) was achieved using Knauer Pneumatic HPLC Pump. The (*S*,*S*)-Whelk-O1 column, Regis Technologies, Inc. (Morton Grove, IL, USA), was 5 μm particle size and had dimensions of 250 × 4.6 mm I.D.

Racemates **8**–**21** used for the evaluation of prepared columns were previously synthesized in the Laboratory for Stereoselective Catalysis and Biocatalysis at the Ruđer Boškovi´c Institute [18,19]. All other compounds were purchased from commercial sources and used without further purification. Unless otherwise noted, the analyses of abovementioned racemates were performed at room temperature using the flow of 1 mL/min and UV detection of the compounds at 254 nm.

#### *3.2. General Procedure for the Preparation of CSP-OHs*

The corresponding DNB amide (0.75 mmol) was suspended in dry DCM (10 mL) under an inert atmosphere. A solution of hexachloroplatinic(IV) acid hydrate (20 mg) in isopropanol (0.5 mL) was added to the reaction mixture, followed by chlorodimethylsilane (10 mL). The mixture was refluxed for 5 h, cooled to room temperature and concentrated under reduced pressure. Dark residue was once again dissolved in DCM (5 mL), concentrated under reduced pressure and used without further purification. To the solution of the crude product in dry DCM (8 mL) under inert atmosphere, the 1:1 mixture of triethylamine and absolute ethanol (10 mL) was added dropwise and stirred at room temperature for half an hour. The solvent was evaporated to give a dark residue which was filtered through a short column of silica gel (eluent DCM-MeOH = 100:1). The resulting yellow residue was dissolved in dry toluene (5 mL) and added to a suspension of 5 μm HPLC silica gel (1.50 g) in dry toluene (50 mL). The silica gel was dried prior use for 24 h in Dean-Stark apparatus. The suspension was refluxed for 24 h, then filtered using a G4 sinter and washed with toluene (30 mL) and methanol (2 × 30 mL). Prepared CSP-OHs were dried for 4 h at 60 ◦C and elemental analysis results and IR spectra were recorded.

**CSP-1-OH** CHN analysis: C 5.05; H 0.63; N 1.11 (0.23 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3442 (SiO2), 1642, 1540, 1343, 1220–1031 (SiO2), 807 (SiO2), 465 (SiO2).

**CSP-2-OH** CHN analysis: C 4.20; H 0.85; N 0.62 (0.18 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3450 (SiO2), 1641, 1544, 1346, 1219–1034 (SiO2), 807 (SiO2), 731, 466 (SiO2).

**CSP-3-OH** CHN analysis: C 4.91; H 0.93; N 0.75 (0.20 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3434 (SiO2), 1640, 1539, 1219–1034 (SiO2), 808 (SiO2), 464 (SiO2).

**CSP-4-OH** CHN analysis: C 5.39; H 0.82; N 0.51 (0.24 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3439 (SiO2), 1640, 1542, 1349, 1227–1030 (SiO2), 808 (SiO2), 731, 465 (SiO2).

**CSP-5-OH** CHN analysis: C 5.43; H 0.80; N 0.70 (0.22 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3445 (SiO2), 1638, 1540, 1214–1029 (SiO2), 806 (SiO2), 467 (SiO2).

**CSP-6-OH** CHN analysis: C 5.45; H 0.55; N 0.77 (0.21 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3448 (SiO2), 1645, 1543, 1347, 1217–1034 (SiO2), 805 (SiO2), 729, 465 (SiO2).

**CSP-7-OH** CHN analysis: C 6.13; H 1.15; N 1.10 (0.22 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3452 (SiO2), 1643, 1545, 1346, 1235–1023 (SiO2), 805 (SiO2), 731, 460 (SiO2).

**CSP-8-OH** CHN analysis: C 6.73; H 1.17; N 0.41 (0.24 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3469 (SiO2), 1638, 1545, 1348, 1230–1034 (SiO2), 807 (SiO2), 731, 463 (SiO2).

**CSP-9-OH** CHN analysis: C 5.36; H 0.63; N 0.96 (0.20 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3451 (SiO2), 1636, 1544, 1350, 1228–1036 (SiO2), 807 (SiO2), 730, 468 (SiO2).

CSP-10-OH CHN analysis: C 6.41; H 1.07; N 1.06 (0.21 mmol/1 g); IR (ν/cm<sup>−</sup>1): 3439 (SiO2), 1630, 1542, 1347, 1223–1029 (SiO2), 806 (SiO2), 728, 464 (SiO2).

#### *3.3. End-Capping and Packing Procedure*

To a yellow suspension of prepared CSP-OHs in dry toluene (20 mL) in an inert atmosphere, hexamethyldisilazane (2 mL) was added and resulting mixture was refluxed for 20 h. The suspension was cooled, filtered using a G4 sinter and washed with toluene (30 mL) and methanol (2 × 30 mL). After drying of prepared CSP overnight at 60 ◦C, they were packed into stainless steel columns (150 × 4.6 mm I.D.) using the slurry packing technique (1.5 g of CSP was suspended in 25 mL of solvent hexane-2-propanol = 2:8).

#### *3.4. Computational Methods*

Minimum energy structures of **DNB-6**, **DNB-8** and **DNB-9** are taken from our previous publication [20]. Relaxed potential energy surface scan was performed in Gaussian 16 [43] with Minnesota global hybrid functional M06-2X [44] with 54% of the exact exchange and aug-cc-pVDZ functional. The ethenyl moiety of (*S*)-**DNB-6** from our previous work [20] was replaced by hydrogen. Ground state force field parameters of **CSP-6**, naproxen and acetic acid were evaluated using the *antechamber* module of the Amber 16 package and the generalized Amber force field (GAFF) [45]. After geometry optimization of the complex of (*S*)-**DNB-6** and naproxen (**A**) and (*S*)-**DNB-6**, naproxen and acetic acid (**B**), molecular dynamics simulations at 300 K in vacuum were run, with a time step of 1 fs and a time simulation of 10 ns, for both the (*S*)- and (*R*)-enantiomers of naproxen. Geometries were saved every 5 ps. The 20 lowest energy structures from the trajectories were optimized at the B3LYP/def-SVP level of theory. The 10 most stable complexes were then re-optimized using the same functional and larger basis set (def2-TZVPP), while final energies were evaluated on M06-2X/def2-TZVPP level with solvation effects incorporated via COSMO model [46] and dielectric constant that equals to 2.78. Geometry optimizations and single point energy calculations were run in Turbomole [47].

#### **4. Conclusions**

In conclusion, we prepared ten new Pirkle-type chiral stationary phases based on (*S*)-N-(1-aryl-propyl)-3,5-dinitrobenzamide selectors. We evaluated the prepared CSPs using thirty diverse racemates, including several nonsteroidal anti-inflammatory and 3-hydroxybenzodiazepine drugs. Our aim was to design CSPs with similar versatility to Whelk-O1 and three of our CSPs (**CSP-6**, **CSP-7** and **CSP-10**) indeed displayed chromatographic behavior comparable to this widely used CSP. The prepared CSPs differ in the aromatic unit directly linked to the chiral center, which enabled us to elucidate the influence of the size and substitution of aromatic moiety on the enantiorecognition process. In order to substantiate experimental results, we investigated the role of nonbonding interactions relevant for chiral recognition on CSPs using computational methods. The model system was enantioseparation of naproxen on **CSP-6** where we elucidated the influence of hydrogen bond network and aromatic interactions on enantiorecognition process. Furthermore, the stability of the complexes is in accordance with the experimentally determined elution order. Finally, we investigated the influence of acetic acid as an additive in the mobile phase in the abovementioned system. To the best of our knowledge, this study is the first one to investigate the possibility of positioning a minor additive within the chiral binding site. We showed that non-enantiospecific interactions with the carboxylic moiety of the analyte can change the interaction arrangements and influence the energetics of the complexes responsible for chiral recognition.

**Supplementary Materials:** The Supplementary Materials are available online. Table S1: Enantioseparation results of racemates **1**–**7** and **9** on **CSP-1**–**CSP-10**, Table S2: Enantioseparation results of racemates **8** and **10**–**14** on **CSP-1**–**CSP-10**, Table S3: Enantioseparation results of racemates **15**–**21** on **CSP-1**–**CSP-10**, Table S4: Enantioseparation results of NSAIDs on **CSP-6**–**CSP-10**, Table S5: Enantioseparation results of 3-hydroxy-benzodiazepine drugs on **CSP-6**–**CSP-10**, Figure S1: Enantioseparation of naproxen on **CSP-6**, Figure S2: Higher energy structures of (*S*)-**A** and (*R*)-**A**, Figure S3: Higher energy structures of (*S*)-**B** and (*R*)-**B**.

**Author Contributions:** Conceptualization, A.K. and V.V.; Methodology, A.K. and V.V.; Formal analysis, A.K. and J.N.; Investigation, A.K and J.N..; Resources, V.V.; Writing—original draft preparation, A.K and J.N..; Writing—review and editing, V.V.; Visualization, A.K. and J.N.; Supervision, V.V.; funding acquisition, V.V.

**Funding:** This research was funded by Croatian Science Foundation, grant number IP-2016-06-1142 (LightMol).

**Acknowledgments:** The authors would like to acknowledge Goran Landek for preliminary results on CSPs, Darko Kontrec for performing the packing of columns, Irena Dokli and Andreja Lesac for technical support and Vladimir A. Potemkin for performing conformational analysis of naproxen. The authors acknowledge generous computer time provided by the Croatian National Grid Infrastructure (CRONGI).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Samples of the compounds **DNB-1**–**DNB-10** and **CSP-1**–**CSP-10** are available from the authors.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Chiral Separation of the Phenylglycinol Enantiomers by Stripping Crystallization**

#### **Lie-Ding Shiau 1,2**


Academic Editors: Maria Elizabeth Tiritan, Madalena Pinto and Carla Sofia Garcia Fernandes Received: 8 October 2018; Accepted: 5 November 2018; Published: 7 November 2018

**Abstract:** Stripping crystallization (SC) is introduced in this work for chiral purification of *R*-phenylglycinol from the enantiomer mixture with an initial concentration ranging from 0.90 to 0.97. As opposed to the solid–liquid transformation in melt crystallization, the three-phase transformation occurs in SC at low pressures during the cooling process. SC combines melt crystallization and vaporization to produce a crystalline product and mixture vapor from a mixture melt due to the three-phase transformation. Thermodynamic calculations were applied to determine the operating pressure for the three-phase transformation during the cooling process in the SC experiments. To consider the possible deviations between the calculated and the actual three-phase transformation conditions, the product purity and the recovery ratio of *R*-phenylglycinol were investigated within a range of operating pressures during the cooling process.

**Keywords:** crystallization; vaporization; purification; phenylglycinol

#### **1. Introduction**

Pure enantiomer is often needed for the desired therapeutic effect due to different pharmacological and pharmacokinetic processes for the enantiomers of drugs. However, separation of the enantiomers has long been a challenging task as the enantiomers have nearly identical physical and chemical properties [1]. Various enantioselective separation techniques, including enantioselective synthesis, chromatographic separation, and preferential crystallization, have been developed for chemical and pharmaceutical industries [2]. Preferential crystallization generally has been used as an attractive means to separate the conglomerate-forming enantiomers from racemate [3–6]. Although chromatographic separation has been investigated extensively [7–10], the synthesis of efficient chiral stationary phases in chromatographic methods is usually deemed a robust technology. Recently, Didaskalou et al. [11] reported the membrane-grafted asymmetric organocatalyst used as an integrated synthesis–enantioseparation platform. Rukhlenko et al. [12] explored the capabilities of the related enantioseparation method by analytically solving the problem of the force-induced diffusion of chiral nanoparticles in a confined region.

Phenylglycinol, also called 2-Amino-2-phenylethanol, is an important example of a chiral compound. Only *R*-phenylglycinol can be used as an important precursor of HIV-1 protease inhibitor [13]. Enantioseparation, using extractant impregnated resins [14] or liquid–liquid extraction [15,16] has been proposed to separate *R*-Phenylglycinol from the racemic mixture. Fundamentally, an enantioselective solvent is chosen and used as extractant for the enantioseparation of phenylglycinol.

Stripping crystallization (SC) is a new separation technology, which combines melt crystallization and vaporization to produce a crystalline product due to the three-phase transformation. SC has been successfully developed to separate the mixtures with close boiling temperatures, including mixed xylenes [17–19], the styrene/ethylbenzene mixture [20], and the *S*-ibuprofen/*R*-ibuprofen mixture [21]. As opposed to extractant impregnated resins or liquid–liquid extraction, no solvent is added in SC. Thus, no removal of solvent is required at the end of SC.

The objective of this research is to study the feasibility of SC in purification of *R*-phenylglycinol from a phenylglycinol mixture. The thermodynamic calculations are adopted to determine the three-phase transformation conditions for the SC experiments. The effects of various operating conditions on the enantiomeric purity and recovery ratio of *R*-phenylglycinol crystalline product are investigated

#### **2. SC Model**

As opposed to the solid–liquid transformation in melt crystallization operated at normal pressure during the cooling process [22–27], the three-phase transformation occurs in SC at low pressures during the cooling process. Thus, SC combines melt crystallization and vaporization to produce a crystalline product and mixture vapor from a mixture liquid or melt [17–21]. The SC process is simulated in a series of N stage operations shown in Figure 1, where each stage is operated at a three-phase transformation state. The SC process starts with a mixture liquid or melt. The vapor formed in each stage is removed, while the crystalline product and the remaining liquid or melt in each stage enter the next stage. Thus, only the crystalline product remains at the end of SC when the liquid or melt is nearly eliminated.

**Figure 1.** Schematic diagram of the stripping crystallization (SC) operation where each stage is operated at a three-phase transformation state.

When SC is applied to purify *R*-phenylglycinol (B-component) from the mixture of *S*-phenylglycinol (A-component) and *R*-phenylglycinol, the SC process starts with a mixture melt of phenylglycinol. The corresponding three-phase transformation condition in each stage is determined based on the following assumptions: (a) The ideal gas law is assumed for the vapor due to low pressures; (b) The ideal solution for the melt is assumed due to the structure similarity between *S*-phenylglycinol and *R*-phenylglycinol; (c) The Clausius–Clapeyron equation [28,29] is adopted to describe the temperature dependence of the saturated pressure for each component in the melt; (d) The sublimation based on the solid–vapor equilibrium is not considered here as the mixture melt is used in the beginning of the experiments. Some physical properties of *S*-phenylglycinol and *R*-phenylglycinol are listed in Table 1. For simplicity, ΔHV = 2ΔHm is assumed in the thermodynamic calculations.

As SC is applied to produce *R*-phenylglycinol crystalline product from a mixture melt due to the three-phase transformation, both the solid–liquid equilibrium and the vapor–liquid equilibrium need to be simultaneously satisfied. The solid–liquid equilibrium is described by the Schroder–Van Laar equation [1,28,29], while the vapor–liquid equilibrium is described by Raoult's law [28,29]. Consequently, as similar to a previous work reported by Shiau [21], the three-phase equilibrium equations can be derived in each stage. If Tn is specified in each stage, these equations can be simultaneously solved for Pn, (XA)n, (XB)n, (YA)n and (YB)n for n = 1, 2, . . . , N.

**Table 1.** Some physical properties for phenylglycinol.

a: The Merck Index [30]; b: Estimated by Clausius–Clapeyron equation [28,29]; c: Measured by Differential scanning calorimetry in this work.

Figure 2 displays the thermodynamic calculations of P(T), XB(T), and YB(T) during the cooling process. Thus, the corresponding pressure, P(T), and the corresponding melt composition of *R*-phenylglycinol, XB(T), decreases during the cooling process for SC. In other words, Figure 2 reveals that, as XB(T) in a melt decreases, the corresponding temperature and pressure for the three-phase transformation conditions decreases.

**Figure 2.** P(T), XB(T) and YB(T) based on the thermodynamic calculations for the threephase transformation.

As shown in Figure 1, the three-phase transformation occurs in the melt in each stage, leading to the formation of *R*-phenylglycinol crystalline product and mixture vapor. Sn and Ln represent the amount of *R*-phenylglycinol crystalline product and the melt, respectively, remaining in stage n, while Vn represents the amount of the mixture vapor formed and removed in stage n. The entire material balance in stage n can be described by

$$\mathbf{S\_{n-1}} + \mathbf{L\_{n-1}} = \mathbf{S\_n} + \mathbf{L\_n} + \mathbf{V\_n} \tag{1}$$

where Sn−<sup>1</sup> + Ln−<sup>1</sup> is the total amount of crystalline product and melt entering stage n. As Vn−<sup>1</sup> represents the amount of vapor formed in stage n − 1 that is subsequently removed, it is not part of the equation for stage n. Thus, the amount of melt decreases and the amount of crystalline product increases during the stage operation.

Although both the melt and the vapor consist of *S*-phenylglycinol and *R*-phenylglycinol, only *R*-phenylglycinol crystalline product is formed in each stage based on the solid–liquid equilibrium described by the Schroder–Van Laar equation [1,28,29]. It is assumed that no impurity trapping occurs in the formation of *R*-phenylglycinol crystalline product based on the thermodynamic calculations. The material balance of *R*-phenylglycinol in stage n can be described by

$$\rm{L}\rm{S}\_{\rm{n}-1} + \rm{L}\_{\rm{n}-1}(\rm{X}\_{\rm{B}})\_{\rm{n}-1} = \rm{S}\_{\rm{n}} + \rm{L}\_{\rm{n}}(\rm{X}\_{\rm{B}})\_{\rm{n}} + \rm{V}\_{\rm{n}}(\rm{Y}\_{\rm{B}})\_{\rm{n}} \tag{2}$$

It is observed during the experiments that the three-phase transformation occurs in the melt very quickly in each stage, leading to the formation of *R*-phenylglycinol crystalline product and the mixture vapor. Therefore, it is assumed in each stage that the heat released in forming *R*-phenylglycinol crystalline product is quickly removed by vaporizing some portion of the melt. Thus, the energy balance in stage n can be described by

$$(\mathbf{S}\_{\rm n} - \mathbf{S}\_{\rm n-1})\Delta \mathbf{H}\_{\rm m,B} = \mathbf{V}\_{\rm n} \Delta \mathbf{H}\_{\rm V,B} \tag{3}$$

where Sn – Sn−<sup>1</sup> represents the amount of crystalline product formed in stage n while Vn represents the amount of melt vaporized in stage n. Note that the heat of vaporization is assumed as ΔHV,B for a mixture melt due to ΔHV,A = ΔHV,B.

As only the mixture melt L0 with a known (XB)0 is injected into the sample container, one obtains S0 = 0. Equations (1) to (3) can be solved simultaneously for three unknown variables- Sn, Ln and Vn. Note that SN and LN represents the crystalline product and the melt, respectively, remaining at the end while the total amount vapor formed and removed at the end is given by ∑<sup>N</sup> <sup>n</sup> <sup>=</sup> <sup>1</sup> Vn.

#### **3. Experimental Section**

The experimental assembly consisted of a 5-mL sample container in a 1.5-L chamber as shown in Figure 3. The stainless chamber was immersed in a constant temperature bath. A mechanical vacuum pump was used to lower the pressure in the chamber. A temperature probe was positioned in the center of the mixture melt and a pressure gauge was connected to the big chamber. Thus, the operating temperature and pressure could be adjusted mid-experiment. Crystallization and vaporization of the mixture melt during the three-phase transformation could be observed in the chamber via a transparent cover.

*R*-phenylglycinol (purity >98%) and *S*-phenylglycinol (purity >98%) were purchased from ACROS. In the beginning of the experiment, 1 g mixture melt with a known concentration was injected into the sample container stirred by a magnetic bar at 70 rpm. Then, the temperature was lowered gradually from the melting point (77 ◦C). The cooling rate generally started at 0.5 ◦C/min in the beginning and then increased gradually to 1 ◦C/min in the later stage. As the temperature decreased, pressure was adjusted downward based on Figure 2. Thus, a series of three-phase transformations occurred in the melt, leading to the formation of *R*-phenylglycinol crystalline product and mixture vapor. The experiments were generally ended at around 55 ◦C and 58 Pa within 25 min when vaporization was no longer observed in the chamber. Figure 4 illustrates the schematic diagram of the batch experiments, in which the melt was simultaneously vaporized and crystallized due to the three-phase transformation. Upon completion, the final product, including the crystals and melt, in the sample container were weighed.

**Figure 3.** Schematic diagram of the experimental apparatus for the SC operation with the features: (1) Mechanical pump, (2) constant temperature bath, (3) thermocouple, (4) pressure gauge, (5) magnetic stirrer operated at 70 rpm, (6) 5-mL sample container in a 1.5-L chamber.

**Figure 4.** Schematic diagram of a batch SC experiment, where each stage corresponds to a three-phase transformation state at a given time: At t = 0, a mixture melt in the sample container; at 0 < t < tf, formation of *R*-phenylglycinol crystalline product and mixture vapor from a mixture melt due to the three-phase transformation; at tf, only *R*-phenylglycinol crystalline product and the remaining melt left in the sample container. Note that the vapor was condensed and collected outside the sample container in the chamber.

The enantiomeric purity of the final product was analyzed by Polarimeter (Horiba, model: SEPA-300). The polarimetry was measured by dissolving 0.1 g final product in 20 mL 1 M HCl solution. First, a plot of the measured specific optical rotation versus the known enantiomeric purity within the range XB,0 = 0.9 to 1.0 was fitted with a linear regression line. Then, by measuring the specific optical rotation of the final sample the enantiomeric purity could be determined. Note that [*α*] 20 *<sup>D</sup>* = −29.9◦ for *R*-phenylglycinol and [*α*] 20 *<sup>D</sup>* = 29.9◦ for *S*-phenylglycinol. It should be noted that, as only crystallization and vaporization occurred during SC, polarimetry could be used to determine the enantiomeric purity of the final product.

From a practical point of view, some solvent might remain in the mixture melt before SC. To elucidate the effects of residual solvent on the final product purity and recovery ratio of *R*-phenylglycinol, 0.1 g ethanol was added into 1 g mixture melt in the beginning of the SC experiments. It was found that the final product purity and recovery ratio for 1 g mixture melt with 0.1 g ethanol were nearly the same as those for 1 g mixture melt without ethanol. Thus, all ethanol was vaporized when SC was operated at low pressures during the cooling process. Solvent inclusion in the formation of *R*-phenylglycinol crystalline product was nearly negligible.

#### **4. Results and Discussion**

SC was applied to purify *R*-phenylglycinol for various 1 g feeds: Feed 1 with (XB)0 = 0.90, feed 2 with (XB)0 = 0.95, and feed 3 with (XB)0 = 0.97. Table 2 lists the thermodynamic calculations for 1 g feed 1, where T0 = 72.7 ◦C is the initial three-phase transformation temperature for the mixture melt. As vaporization was no longer observed in the experiments at around 55 ◦C, TN = 54.6 ◦C was chosen for N = 15 with ΔT = 1.2 ◦C. Thus, Tn was specified in each stage for n = 1, 2, . . . , N using Tn−<sup>1</sup> − Tn = ΔT. Pn, (XA)n, (XB)n, (YA)n, and (YB)n were determined in each stage by solving the thermodynamic equations while Sn, Ln, and Vn were determined in each stage by solving Equations (1) to (3) for L0 = 1 g and S0 = 0. Note that Pn, (XB)n and (YB)n in Table 2 were consistent with the results shown in Figure 2. Table 2 also indicates that, as n increased during the cooling process, Sn increased and Ln decreased. As SC was operated from 73 ◦C and 160 Pa (n = 1) to 55 ◦C and 58 Pa, (n = 15), only *R*-phenylglycinol crystalline product remain in the last stage (SN = 0.606 g) while the melt was nearly eliminated in the last stage LN = 0.098 g. Similar calculated results were obtained for feed 2 and feed 3.

**Table 2.** The thermodynamic calculations for 1 g feed with XB,0 = 0.90 (ΔT = 1.2 ◦C).


The calculated purity of *R*-phenylglycinol in the final product, including the final crystalline product and the remaining melt, is defined as

$$\chi\_{\rm B,C} = \frac{\rm S\_N + L\_N(\chi\_{\rm B})\_N}{\rm S\_N + L\_N} \tag{4}$$

where SN, LN and (XB)N are determined in the last stage based on the thermodynamic calculations. The calculated recovery ratio of *R*-phenylglycinol is defined as

$$\mathcal{R}\_{\mathbb{C}} = \frac{\mathbb{S}\_{\text{N}} + \mathcal{L}\_{\text{N}}(\mathbb{X}\_{\text{B}})\_{\text{N}}}{\mathcal{L}\_{\text{0}}\mathcal{X}\_{\text{B},0}} \tag{5}$$

where L0 is the initial weight of the mixture melt and XB,0 denotes the initial purity of *R*-phenylglycinol in the mixture melt. For example, as shown in Table 2, feed 1 yields SN = 0.606 g and LN = 0.098 g with (XB)N = 0.549 in the last stage (N = 15), leading to XB,C = 0.937 and RC = 73% using Equations (4) to (5).

The experimental recovery ratio of *R*-phenylglycinol is defined as

$$\mathbf{R}\_{\mathbf{f}} = \frac{\mathbf{W}\_{\mathbf{f}} \boldsymbol{\chi}\_{\mathbf{B},\mathbf{f}}}{\mathbf{L}\_{\mathbf{0}} \boldsymbol{\chi}\_{\mathbf{B},\mathbf{0}}} \tag{6}$$

where Wf refers to the final weight of the product including the crystalline product and the remaining melt obtained at the end of the experiment, and XB,f represents the experimental purity of *R*-phenylglycinol in the final product.

Figure 5 shows XB,f of the final product plotted against XB,0 of the initial feed for various feeds. The Solid circles represent the calculated XB,C and the number in parenthesis represents the calculated RC. Thus, the thermodynamic calculations predict that feed 1 can be purified from XB,0 = 0.90 to XB,C = 0.937 with RC = 73%, feed 2 can be purified from XB,0 = 0.95 to XB,C = 0.979 with RC = 70%, and feed 3 can be purified from XB,0 = 0.97 to XB,C = 0.995 with RC = 69%.

**Figure 5.** XB,f of the final product plotted against XB,0 of the initial feed for various 1 g feeds. The solid line represents XB,f = XB,0 indicating no further purification for the initial feed during SC. Solid circle represents the calculated XB,C and the number in parenthesis represents the calculated RC. Other symbols, including open circle, open triangle, cross sign, and open square, represent the average of the experimental XB,f for four repetitive experiments operated at the specified pressure and the error bar represents the 95% confidence interval for the experimental XB,f. The number in parenthesis represents the average of the experimental Rf with the 95% confidence interval for the experimental Rf. Note that no error bar is added for solid circle of the calculated XB,C.

Other symbols in Figure 5, including open circle, open triangle, cross sign, and open square, represent the average of the experimental XB,f for four repetitive experiments operated at the specified pressure and the error bar represents the 95% confidence interval for the experimental XB,f. The number in parenthesis represents the average of the experimental Rf with the 95% confidence interval for the experimental Rf. For example, cross sign represents the average XB,f when the operating pressures was controlled at P(T) during the cooling process. As a lower XB,f with a higher Rf was observed for each feed compared to the calculated XB,C and RC, it was speculated that the calculated pressure P(T) in Figure 3 might be higher than the actual three-phase transformation pressure, leading to less impurity (*S*-phenylglycinol) vaporized and more crystalline product formed from the melt during the cooling process.

To consider the possible deviations between the calculated and the actual three-phase transformation pressure, various operating pressures are compared during the cooling process. The open circle in Figure 5 represents the average XB,f when the operating pressures were controlled at 0.1 × P(T) during the cooling process. Similarly, the open triangle represents the average XB,f when the operating pressures were controlled at 0.5 × P(T). The open square represents the average XB,f when the operating pressures were controlled at 10 × P(T).

Figure 5 shows for each feed that XB,f increased with decreasing pressure while Rf decreased with decreasing pressure. For example, when SC was applied for XB,0 = 0.90, XB,f increased from 0.914 to 0.935 and Rf decreased from 86% to 46% as the operating pressure was decreased from P(T) to 0.1 × P(T). As shown in the figure, when 0.1 × P(T) was adopted for each XB,0, XB,f was close to XB,C with Rf (46% to 55%) < RC (69% to 73%). Consequently, compared to the calculated P(T) in Figure 3, 0.1 × P(T) should be closer to the actual the three-phase transformation pressure. On the other hand, when 10 × P(T) was adopted for each XB,0, XB,f was close to XB,0 with Rf = 93% to 99%, indicating that the feed was not further purified in the SC experiments.

Discrepancies between the thermodynamic calculations and the experimental results are attributed to (a) the assumption that each stage was operated at the three-phase transformation. However, experimentally, these might not always be achieved; (b) although pure *S*-phenylglycinol crystal should be formed based on the thermodynamic equilibrium, impurity trapping can occur under actual kinetic conditions. The scope of this work was to investigate the feasibility of SC in the purification of *R*-phenylglycinol from a phenylglycinol mixture. In future kinetic studies, the effects of process conditions (e.g., cooling rate) on the crystal growth kinetics and impurity inclusion will be explored based on the impurity trapping correlation proposed by Myerson and Kirwan [31,32].

#### **5. Conclusions**

SC was successfully applied for chiral purification of *R*-phenylglycinol from the phenylglycinol enantiomers. A lower pressure during the cooling process generally led to a higher experimental product purity with a lower experimental recovery ratio. When SC was operated under the optimal pressure, which was one-tenth of the pressure based on the thermodynamic calculations, the experimental product purity was close to the calculated product purity while the experimental recovery ratio was slightly lower than the calculated recovery ratio. In other words, when temperature and pressure was lowered from 72.7 ◦C and 15 Pa to 55 ◦C and 6 Pa during SC, the purity of *R*-phenylglycinol increased from 0.90 to 0.937, from 0.94 to 0.985, and from 0.97 to 0.995 respectively with the recovery ratio ranging between 46% to 55%.

As no solvent is added into the melt, SC is a clean separation technology. Compared to melt crystallization, neither solid/liquid separation nor crystal washing is required because no mother liquor adheres to the crystal surfaces upon completion. Although a portion of the phenylglycinol enantiomers is lost through the vapor stream of each stage, the vaporized mixture can be recycled for continuous operation or mixed with the feed in the next batch for batch operation. The major difficulty in application of SC lies in the required low pressures during the cooling process. Furthermore, the crystal growth kinetics and impurity trapping during SC need to be elucidated in order to design an apparatus for industrial application.

**Funding:** This research is funded by Ministry of Science and Technology of Taiwan (MOST106-2221-E-182-053) and Chang Gung Memorial Hospital (CMRPD2G0241).

**Acknowledgments:** The author would like to thank Ministry of Science and Technology of Taiwan (MOST106-2221-E-182-053) and Chang Gung Memorial Hospital (CMRPD2G0241) for financial support for this research. The author also expresses his gratitude to Yu-Chen Chen and Keng-Fu Liu for their experimental work.

**Conflicts of Interest:** The author declares no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

#### **Notation**


#### **Subscript**


#### **References**


**Sample Availability:** Samples of the compounds are not available from the authors

© 2018 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Improved Enantioselectivity for Atenolol Employing Pivot Based Molecular Imprinting**

**Andreea Elena Bodoki 1,†, Bogdan-Cezar Iacob 2,† ID , Laura Elena Gliga 2, Simona Luminita Oprean 1, David A. Spivak 3, Nicholas A. Gariano <sup>3</sup> and Ede Bodoki 2,\* ID**


Academic Editors: Maria Elizabeth Tiritan, Madalena Pinto and Carla Sofia Garcia Fernandes

Received: 3 June 2018; Accepted: 26 July 2018; Published: 27 July 2018

**Abstract:** In the last few decades, molecular imprinting technology went through a spectacular evolution becoming a well-established tool for the synthesis of highly selective biomimetic molecular recognition platforms. Nevertheless, there is still room for advancement in the molecular imprinting of highly polar chiral compounds. The aim of the present work was to investigate the favorable kosmotropic effect of a ternary complex involving a polar chiral template (eutomer of atenolol) and a functional monomer, bridged by a central metal ion through well-defined, spatially directional coordinate bonds. The efficiency of the chiral molecular recognition was systematically assessed on polymers obtained both by non-covalent and metal-mediated molecular imprinting. The influence on the chromatographic retention and enantioselectivity of different experimental variables (functional monomers, cross-linkers, chaotropic agents, metal ions, porogenic systems, etc.) were studied on both slurry packed and monolithic HPLC columns. Deliberate changes in the imprinting and rebinding (chromatographic) processes, along with additional thermodynamic studies shed light on the particularities of the molecular recognition mechanism. The best performing polymer in terms of enantioselectivity (α = 1.60) was achieved using 4-vinyl pyridine as functional monomer and secondary ligand for the Co(II)-mediated imprinting of S-atenolol in the presence of EDMA as cross-linker in a porogenic mixture of [BMIM][BF4]:DMF:DMSO = 10:1:5, *v*/*v*/*v*.

**Keywords:** metal-mediated molecular imprinting; hydrophilic template; atenolol; chiral separation; β-blockers; molecularly imprinted polymers; molecular recognition

#### **1. Introduction**

Biological or synthetic receptors selectively recognize their target chemicals based on a combination of weak, short-ranged intermolecular interactions, such as hydrogen bonding, π-π interactions and van der Waals forces; and their selectivity being further refined by additional repulsive steric confinements. Molecularly imprinted polymers (MIPs) able to mimic natural receptors, offer tailored selectivity towards target molecules and better chemical and thermal stability in a simple and cost-effective manner [1]. MIPs have been widely employed for the concentration, separation and analysis of various bioactives, either as stationary phases in chromatography and capillary electrophoresis [2–6], or as recognition elements in chemo- and biosensing [7–9]. Within the extensive body of literature, a considerable part focuses on the use of these polymers for chiral analysis [2,7,10–13]

in various pharmaceutical, biomedical or environmental applications. Differentiation between the chiral forms of a molecule (chiral discrimination) is considered the supreme form of molecular recognition. Most often, the needed stereospecific features of MIPs are acquired through the process of non-covalent molecular imprinting using a variety of functional monomers, cross-linkers and porogenic solvents. Even though over the last few decades MIPs have become a well-established analytical tool for the selective recognition and analysis of small molecules, there is still some room for advancement in the molecular imprinting of highly polar compounds and biomacromolecules. Recent attempts at imprinting polar compounds such as polyphenols, e.g., oleuropein [14], polar organic micropollutants, e.g., benzotriazole [15] have been described, especially for sample enrichment applications (SPE adsorbents). The addition of certain additives (hydrophilic functional polymersoligo- and polyethylene glycol methacrylate [16]; molecular crowding agents—polyethylene glycol; room temperature ionic liquids [17,18]) in the pre-polymerization mixture could improve the overall imprinting efficiency, hydrophilicity, flexibility, morphology and porosity of the resulting polymer. Furthermore, alternative imprinting protocols, such as metal ion-mediated molecular imprinting may further correct some of the observed shortcomings in the molecular imprinting and recognition of polar templates [16].

In the pivot-based or metal ion-mediated molecular imprinting (MMMI) process the metal ions act as a bridge between the functional monomer and the template. The monomers are thus regularly positioned around the template via coordinate bonds restraining the free motion of the species. Consequently, the number of non-specific binding sites decreases, and improved imprinting factors are achieved [19]. Since chiral molecular recognition by MIPs relies on very small energetic differences between the forming transient selector/select and complexes at the polymer's interaction sites, any improvement in the degree of order throughout the polymerization step is beneficial in conveying the template's (i.e., target enantiomer) molecular information to the emerging MIP recognition sites with the highest possible fidelity.

Metal pivot-based imprinted polymers designed for water soluble templates were previously reported by the enhancement of column permeability and affinity towards the polar template [16,18,20,21]. However, the number of metal-mediated imprinted polymers intended for chiral separation is rather scarce [16,17,22]. Various representatives of β-blockers, most often propranolol [23–26], have been employed as model compounds in demonstrating the enantioselectivity of various chiral selectors. Nevertheless, for atenolol (ATNL), as one of the most polar representatives of this class of drugs, usually the poorest enantioselectivity has been reported under optimized binding conditions [26–28].

The aim of the present work was to investigate the favorable kosmotropic effect of a ternary complex involving both the polar chiral template (eutomer of ATNL) and the functional monomer, bridged by the central metal ions through well-defined, spatially directional coordinate bonds. Various aspects of the formation of the ternary metal complex monitored by UV-Vis spectroscopy, as well as particularities regarding the molecular imprinting process (composition of pre-polymerization mixture, initiation of free-radical polymerization, etc.) are also discussed.

#### **2. Results**

#### *2.1. Ternary Metal Complexes of ATNL*

In order to engage a more rational approach in selecting the appropriate metal ion, functional monomer, molar ratio and porogenic solvent to be tested for subsequent molecular imprinting, UV-Vis spectroscopy provided a simple, fast, cost-effective and relatively straightforward instrumental method which is adjustable to small sample volumes for assessing the formation of the ternary complex. The electronic spectra (350–1100 nm) of binary and ternary complexes of several transition metal ions (Co(II), Cu(II) and Ni(II)) with the ATNL as primary ligand, and various functional monomers (4-vinyl pyridine (4-VPy), 1-vinyl imidazole (VIM), methacrylic acid (MAA), acrylamide (AM), *N*,*N* -methylenebis(acrylamide) (BAM), *N*,*O*-bismethacryloyl ethanolamine (NOBE), vinyl ferrocene

(VFC), 4-vinyl phenylboronic acid (4-VPBA), trans-2-chloromethylvinylboronic acid (CVPBA) and poly(ethylene glycol) methyl ether methacylate (Mn = 300 g mol<sup>−</sup>1, PEGMA) prepared in a mixture of DMF/DMSO 1:5 (*v*/*v*) were recorded.

Bathochromic and hypsochromic shifts or hyperchromic and hypochromic effects occurring in the electronic spectra of ternary metal-template-monomer species with respect to the spectra of binary metal-template analogues were interpreted as evidence of ternary complex formation. Co(II) ions were selected as the pivot for the MIPs using 4-Vpy as a monomer (Figure 1a,b), while Cu(II) ions were preferred for the imprinting process using an acidic monomer, MAA (Figure 1c).

**Figure 1.** Electronic spectra of Co(II) complexes in DMF/DMSO 1:5 (*v*/*v*) with (**a**) binary and ternary Co(II) complexes with ATNL (1:1 molar ratio) and 4-VPy (1:5 molar ratio); (**b**) binary and ternary Co(II) complexes with ATNL (1:1 molar ratio) and MAA (1:5 molar ratio); (**c**) binary and ternary Cu(II) complexes with ATNL (1:1 molar ratio) and MAA (1:5 molar ratio).

#### *2.2. Preparation of MIPs*

#### 2.2.1. Non-Covalent Molecular Imprinting

Based on the ability of ATNL to interact in a concerted manner with hydrogen-bonding monomers, choices of host functional monomers ranged from the most commonly used donor-acceptor type monomers, such as MAA, AM or VIM to the ones bearing a single hydrogen donor or acceptor motif, such as 4-VPy (Tables 1–3). Choices among the conventionally employed aprotic porogenic solvents for non-covalent imprinting (i.e., toluene, ACN) were often limited by the poor solubility of the highly polar template. Thus, the influence of different aprotic (ACN, DMF, DMSO) and protic (methanol (MeOH), butanol) porogenic solvents on the overall features of the resulting imprinted polymer were

also studied. Concomitantly, the influence of different additives (i.e., cross-linkers, ionic liquids) meant to balance the polymeric framework's flexibility and porosity, as well as the selectivity of the imprinted binding sites, were investigated. Both bulk (S4) and monolithic (M6, M16-20) MIPs were obtained using this imprinting approach. Modest to no enantioselectivity (α = 1.0–1.32) was recorded, amongst which the monolithic polymer with 4-VPy as functional monomer (M6, α = 1.30), reticulated with EDMA in the presence of an ionic liquid ([BMIM]BF4) in DMF/DMSO 1:5 (*v*/*v*) as porogenic media was considered the most promising candidate for further investigations by the MMMI approach. Comparable results were obtained employing MAA as a functional monomer in ACN as a porogenic solvent (M18, α = 1.32). The presence of a protic solvent such as MeOH or water, even in trace amounts, added to the solvation of the template in ACN compromised chiral recognition (M20, α = 1.00).

#### 2.2.2. One-Monomer Molecularly Imprinted Polymer (OMNiMIP)

In an effort of reducing to a minimum the number of variables involved in the traditional non-covalent molecular imprinting, the OMNiMIP approach introduced by Sibrian-Vasquez and Spivak [29] was also tested by using a single crosslinking monomer, NOBE, in addition to the chiral template, solvent (DMF) and initiator (Table 1). Under these conditions, the obtained enantioselectivity was somewhat below expectations (S1, α = 1.05). However, when using ACN as porogen, in which case the presence of a small amount of MAA for the solubilization of ATNL was necessary, a certain improvement in enantioselectivity of the slurry packed column (S4, α = 1.17) was observed.

#### 2.2.3. Metal Ion Mediated Molecular Imprinting

The protocols of molecular imprinting were adapted in compliance with the requirements of the metal ion-mediated approach. Three transition metal ions (Co(II), Cu(II) and Ni(II)) were employed as coordination centers (Tables 1–3). Initially, the use of metal ions alongside or together with the single crosslinking monomer (NOBE) did not provide high enough enantioselectivities (S2, α = 1.05; S3, α = 1.01). Furthermore, the use of different molar ratios of 4-VPy as secondary ligand alongside the cross-linker TRIM and porogenic mixture (DMF/DMSO 1:1, *v*/*v*) provided low chiral discrimination of the slurry-based columns (S6–10, α = 1.03–1.07). However, changing the crosslinker to EDMA, and adjusting the polymerization media by adding an ionic liquid ([BMIM]BF4 in DMF/DMSO 1:5, *v*/*v*) resulted in a considerable enhancement in chiral selectivity (S11, α = 1.32). Eventually, switching from slurry-based to monolithic columns (with expected gain in column efficiency), alongside the concerted benefits of using MMMI and ionic liquid, led to the fabrication of the best performing MIP monolith (M2, α = 1.60). The optimal composition of the pre-polymerization mixture (Table 2, M2) turned out to be S-ATNL:Co(II):4-VPy (1:1:6 molar ratio), EDMA as cross-linker (C:M ratio = 1:4), [BMIM]BF4 as ionic liquid in DMF/DMSO 1:5 (*v*/*v*) as porogenic system.

#### 2.2.4. Bulk Imprinting vs. MIP Monolith

Initially, bulk imprinting with photochemical initialization at lower temperatures was considered (S1–S11, Table 1). After the removal of excess reagents and template (Figure 2a,b), the polymers were ground and sieved, followed by the slurry packing into HPLC columns. Studies continued on monoliths polymerized in situ in the chromatographic column (M1–24, Tables 2 and 3), but with thermal radical initiation. To further improve the molecular imprinting efficiency and future chromatographic performances of the monoliths, a hydrophilic ionic liquid was also added to the polymerization mixture [18]. Significant differences in the morphology and porosity of the monolith obtained in the presence and absence of the ionic liquid were observed (Figure 2a–d).

**Figure 2.** Scanning electronmicrographs of monolithic imprinted polymers of M2 (**a**) before and (**b**) after MeOH:AcOH 9:1(*v*/*v*) washing, (**c**) M8 and (**d**) M5 before MeOH:AcOH 9:1(*v*/*v*) washing. Elemental mapping spectrum of M2 (**e**) before and (**f**) after template removal indicating the washout of the metal pivot ion (Co(II)).

#### *2.3. Chromatographic Evaluation of the MIPs*

The enantioselectivity of MIPs investigated as HPLC stationary phases usually exhibits good interassay reproducibility and adequate efficiency with a high sensitivity of detection. Therefore, the present study used standard HPLC columns (100 × 2.1 mm), either slurry packed with the imprinted polymer samples or as a monolith, to study their distinctive chromatographic behavior and recognition mechanism. The assessment and comparison of imprinting factors (IF = k'MIP/k'NIP) was avoided because the observed binding differences are not exclusively due to the existence of specific imprinted cavities, but also due to the significant differences in the morphology (shape, texture, rigidity, porosity, surface area) of the imprinted (MIP) and non-imprinted polymer (NIP). Thus, the efficacy of molecular recognition was assessed based on the highest enantioselectivity achieved for ATNL's enantiomers under different chromatographic conditions (mobile phase composition). The results are synthetically presented in Tables 1–3, along with some of the most representative chromatograms (Figure 3).




Mobile phase composition: Replaced corresponding DMF/DMSO 1:5, *v*/*v*. typical (60 ◦C) photo-induced polymerization (24 h), the pre-polymerization mixture also includes 15 mg of AIBN as free radical initiator. T—template, M—monomer, Me—metal ion, C—cross-linker, A—additive, P—porogen, IL—[BMIM]BF4.



Mobile phase composition: a ACN; b ACN:50 mM formate buffer, pH 3 = 95:5, *v*/*v*; c ACN:50 mM formate buffer, pH 3 = 90:10, *v*/*v*; d ACN:50 mM formate buffer, pH 3 = 80:20, *v*/*v*; e ACN:50 mM acetate buffer, pH 5 = 80:20, *v*/*v*. In a typical thermal (60 ◦C) or photo-induced polymerization (24 h), the pre-polymerization mixture also includes 15 mg of AIBN as free radicalinitiator.T—template,M—monomer,Me—metalion,C—cross-linker,A—additive,P—porogen,IL—[BMIM]BF4.

**Figure 3.** Changes in chromatographic selectivity of metal ion-mediated imprinted (M2), non-covalent imprinted (M6) and non-imprinted (M8) monolithic polymer columns. R(+)—R enantiomer of ATNL, S(−)—S enantiomer of ATNL. For chromatographic conditions see footnote of Table 2.

To gain better insight into the prevalent chromatographic retention mechanism, the influence of the imprinting approach and polymerization mixture constituents, in addition to the mobile phase composition (organic solvent; ratio, nature and pH of the aqueous buffer) on the polymers' retention properties were investigated. The nature of the metal ion used in the pivot-based molecular imprinting approach had an important effect on the monolith enantioselectivity (Figure 4a). In 4-VPy-based polymers selectivity is mainly controlled by hydrogen bonding interactions, which are disrupted even by minute amounts of protic solvent (i.e., MeOH, isopropanol, water) added to the mobile phase. In MAA-based polymers, the partition equilibrium is controlled both by ion-exchange and hydrogen bonding interactions, their contribution being dependent on the ratio and pH of the aqueous buffer. Thermodynamic retention studies (Figure 4b) performed on Co(II)-mediated 4-VPy-based imprinted (M2) and non-imprinted (M8) polymers indicate that the binding of ATNL enantiomers involves an important component of the enthalpic (e.g., hydrogen bonding) term.

**Figure 4.** (**a**) Enantioselectivity (α) and chromatographic retention (k'S/k'R) of metal ion-mediated imprinted monoliths (M2—Co(II), M3—Ni(II), M4—Cu(II)). (**b**) Thermodynamic retention study of ATNL's enantiomers on metal ion-mediated molecularly imprinted (M2, —S-ATNL, -—R-ATNL) and non-imprinted (M8, -—S-ATNL, —R-ATNL) polymers.

#### **3. Discussion**

#### *3.1. Ternary Metal Complexes of ATNL*

ATNL, (R,S)2-(4-[2-Hydroxy-3-(isopropylamino)propoxy]phenyl)acetamide (Figure 5), is a beta-adrenergic antagonist, a cardioselective drug with a prolonged effect. Its S(-) stereoisomer exhibits a significantly higher affinity for the β1-adrenergic receptors [30]. ATNL is a good chelating agent that may act as a bidentate ligand through the secondary alcohol and amine as electron pair donor moieties and allows the formation of five membered rings that include the central metal ion [31–33]. It is also noteworthy that one of the functional groups typically involved in metal ion coordination, the secondary alcohol moiety, is bound to the ATNL's chiral center. Mononuclear or binuclear binary complexes where the metal to ligand (ATNL) ratio is 1:1, 1:2 or 1:4 have been reported for several first-row transition metal ions (Me = Co(II), Ni(II), Cu(II), Zn(II)) [31,32].

**Figure 5.** Potential interaction sites (coordinate and/or hydrogen bonding) of ATNL during non-covalent and MMMI.

Mononuclear tetrahedral complexes of [Me(ATNL)2] 2+ type with ATNL acting as bidentate, mononuclear octahedral complexes of [Me(ATNL)4] 2+ type [31,34] where two ATNL molecules act as bidentate and the ATNL molecules in the axial position act as monodentate, or an O-bridged binuclear complex, [Cu2(ATNL)2Cl2], where ATNL acts as (O, NH) bridging ligand [33], have been reported for first row transition metal ions (Me = Co(II), Ni(II), Cu(II), Zn(II)).

Ternary complexes of ATNL and ligands with N and O atom donor sites from amine and carboxylic moieties have also been reported for first-row transition metal ions [31,35]. The stability of mixed ligand complexes depends on the characteristics of the approaching secondary ligand (e.g., chelating properties, size and spatial configuration of chelate ring etc.), but also on the possible interactions outside the coordination sphere (hydrogen bonding between coordinated ligands, charge neutralization, chelate effect, and electrostatic interaction between non-coordinated charged moieties of ligands). Ternary complexes of metal ion—primary ligand—secondary ligand in the ratio of 1:1:1, 1:1:2 and 1:2:1, where ATNL was either the primary ligand or the secondary ligand, were obtained. Results indicated the preferential formation of ternary complexes in the 1:1:2 ratio over binary complexes for Co(II) as coordination center [31,35].

Thus, one of the hypotheses of the present study relied on the potential engagement of the hydroxyl group linked to ATNL's chiroptic center (most likely in its non-deprotonated form [31,33]) in a specific, spatially well-oriented interaction with an appropriate functional monomer mediated by a central metal ion through coordinate bonds. Such a metal ion-mediated molecular self-assembly of the template-monomer in the polymerization mixture should promote beneficial effects in the chiral molecular imprinting process of this hydrophilic enantiomer.

A favorable outcome using MMMI for the synthesis of enantioselective polymers can only occur if in the pre-polymerization step a stable, well-defined and soluble ternary metal complex exists. If in noncovalent imprinting a nonpolar, aprotic porogen, such as toluene or chloroform, is the ideal choice for promoting template-functional monomer associations [19], then in MMMI polar solvents (i.e., DMF, DMSO, MeOH) are required. Polar media may concomitantly offer the prerequisites of a favorable preorganized state (soluble ternary coordination complex): (i) Solvation of the metal ion and polar template and (ii)deprotonation of ligands (both the chiral template and monomer [36]) essential for metal coordination and (iii)dissolution of the resulting ternary metal complex. The stronger and more-defined interactions within the ternary complex are anticipated to lead to more specific recognition sites upon molecular imprinting. Contingent on the nature of the metal ion, its affinity towards the template and the selected monomer and molar ratio of ligands, different ternary metal complexes may also arise. Therefore, the success or failure of MMMI depends on the state of equilibrium established between these coordination complexes.

In the electronic spectra of transition metal complexes d-d transitions, charge transfer transitions, internal ligand transitions, combination and overtone vibrations of the ligands, and intervalence charge transfer transitions, materialize as bands in the region that spans the near infrared, visible and UV region (4000–30,000 cm−1). Bathochromic and hypsochromic shifts or hyperchromic and hypochromic effects occurring in the electronic spectra of ternary metal-template-monomer species with respect to the spectra of binary metal-template analogues is an indication of a change in the ligand field environment around the coordination center [37–39] for the co-existing species in equilibrium. Based on the data provided by the electronic spectra, Co(II) ions (λmax = 555 nm) were selected as pivot for the MIPs using 4-VPy as a monomer (Figure 1a), while Cu(II) ions were preferred for the imprinting process using an acidic monomer, MAA (Figure 1c).

The electronic spectra of the binary mixtures of ATNL and Co(II), Ni(II) and Cu(II), respectively, recorded in a DMF:DMSO (1:5, *v*/*v*) exhibited relatively weak, low-energy bands in the 13,330–19,230 cm−<sup>1</sup> (750–520 nm) range which could be assigned to the d-d electronic transitions in a distorted octahedral or tetrahedral coordination environment. The electronic spectra also exhibited higher intensity bands at higher energy regions of the spectra that may be assigned to internal ligand transitions or d–π\*, L→M or M→L charge transfer bands. The collected data are in agreement with the bands observed in the electronic spectra of previously reported Co(II), Ni(II) or Co(II) complexes of ATNL [31–33].

The d-d electronic transitions translate into two low intensity bands in the electronic spectrum of the binary Co(II)–ATNL mixture, one centered at 17,605 cm−<sup>1</sup> (568 nm) with a shoulder at 18,868 cm−<sup>1</sup> (530 nm) and the other at a lower energy region 9090 cm−<sup>1</sup> (around 1100 nm) (Figure 1a). Upon addition of 4-VPy to the binary mixture, the local coordination environment around the Co(II) ion changes, and this translates into a broader multiple structured band centered at 18,348 cm−<sup>1</sup> (545 nm) that appears in the spectra of the ternary mixture. The hypsochromic shift of the bands is associated to a hyperchromic effect for the shoulder initially at 18,868 cm−<sup>1</sup> (530 nm) and to a hypochromic effect for the band initially at 17,605 cm−<sup>1</sup> (568 nm) (Figure 1a).

Nevertheless, the spectra of binary mixtures of Cu(II)–ATNL and Ni(II)–ATNL changes to a much lesser extent upon the addition of 4-VPy (data not shown). In these cases, it appears that the primary ligand, ATNL, gives rise to a more stable binary complex and that the local environment around Cu(II) and Ni(II), respectively, is significantly less influenced by the presence of the secondary ligand (data not shown).

Acidic monomers such as MAA interact with the Brönsted-basic template ATNL, and the protonation of the amine moiety may alter the beta-blocker's chelating properties. Such a phenomenon was observed in case of the binary Co(II)-ATNL mixture. The local environment around the Co(II) ion changes significantly upon addition of MAA. As shown by the electronic spectra, ATNL is apparently displaced from the coordination sphere of Co(II) (Figure 1b). In contrast, the electronic spectra indicates no alteration of the local environment around Cu(II) when the acidic monomer is added to the binary Cu(II)–ATNL binary mixture (data not shown).

#### *3.2. Preparation of MIPs*

#### 3.2.1. Non-Covalent Molecular Imprinting

ATNL has several hydrogen donor and acceptor atoms (Figure 5) which during the non-covalent molecular imprinting could interact in a concerted manner with hydrogen-bonding monomers. Choices of host functional monomers ranged from the most commonly used donor-acceptor type monomers, such as MAA, AM or VIM to the ones bearing a single hydrogen donor or acceptor motif, such as 4-VPy. Furthermore, being a Brönsted-basic template, during the ATNL interaction with MAA a partial or full proton transfer is expected to occur. Based on the nature of the employed porogen, contact hydrogen bonded assemblies may be formed in aprotic solvents such as ACN. Nevertheless, polar protic solvents tend to disrupt such electrostatic interactions having a negative impact on the imprinting factors, but also on the rebinding mechanism of the resulting imprinted material [40]. Different additives (i.e., cross-linkers, ionic liquids [36]) meant to balance the polymeric framework's flexibility and porosity, as well as the selectivity of the imprinted binding sites, were used in the pre-polymerization mixture (Tables 1–3). Modest to no enantioselectivity (α = 1.0–1.32) was recorded, both for the slurry-packed and monolithic columns. Monoliths with different functional monomers (i.e., 4-VPy, MAA) and significantly different porogenic media (DMF/DMSO, ACN) were able to provide similar enantioselectivities (M6, α = 1.30; M18, α = 1.32), as long as the cross-linker and ionic liquid were identical in the polymerization mixture. The ratio of monomer to cross-linker seems to be of less importance in this imprinting approach; however, the presence of a protic solvent such as MeOH or water, even in trace amounts, can compromise the polymer's enantioselectivity for the hydrophilic template. The non-covalent imprinting of ATNL's enantiomer by the OMNiMIP approach, at least under the tested experimental conditions, failed to provide noteworthy results in chiral chromatographic selectivity. One of the possible reasons for the low enantioselectivity recorded in the case of the NOBE-based columns may be the significant swelling effect observed during the sieving of the crushed polymer that may affect the structural integrity of the imprinted cavities.

#### 3.2.2. Metal Ion-Mediated Molecular Imprinting

Particular requirements are to be met for MMMI in terms of components of the pre-polymerization mixtures: polar porogenic solvent (DMF, DMSO, DMF/DMSO) that must provide the solubilization of the metal ion's salt (anhydrous acetates) while keeping the ternary complex in solution, and metal ions that should not compromise the efficiency of the employed free radical initiator (AIBN). Another concern when imprinting polar templates is the chemical and mechanical structure of the polymeric network. A good balance between the polymer's hydrophilicity and its flexibility determined by the nature of the functional monomer and cross-linker must be optimized. Various additives (i.e., cross-linkers, co-monomer, ionic liquids, chaotropic agents) affecting the prototropic forms or the basicity of participating ligands (ATNL enantiomer and functional monomer) may also be decisive in the formation of the ternary metal complex; therefore, any change in the pre-polymerization mixture had to be carefully considered (Tables 1–3).

#### Metal Ion and Functional Monomer (Secondary Ligand)

Following the effect of the nature of metal ion on the efficiency of pivot-based imprinting using 4-VPy as a secondary ligand; results show a decrease in the enantioselectivity of the resulting monoliths in the order Co(II) Ni(II) ≥ Cu(II) (α = 1.60, 1.38 and 1.30, respectively for M2, M3 and M4 respectively). These findings are correlated with the electronic spectra of the metal complexes recorded in the screening step (Section 2.1) that suggest a higher stability of the Co(II) ternary complex (Figure 1a). Since most of the metal ions from the polymeric framework are eliminated during the template removal process (Figure 2e,f), they are no longer involved in the molecular recognition during rebinding. Nevertheless, the chromatographic retention (k') is inversely correlated with the recorded enantioselectivities (Figure 4a); this is most probably due to the higher number of non-specific binding sites emerging when Ni(II) and Cu(II) are employed as mediators. Obviously, changing the nature of the secondary ligand is also critical for the chromatographic performance of the resulting chiral stationary phases (CSP). Keeping Co(II) as a pivot and using 1-VIM as a basic secondary ligand bearing the same electron donor moiety as 4-VPy; selectivity decreases to 1.43 (M14). Evidently, in principle, other metal ion–secondary ligand combinations could equal and possibly surpass the selectivity recorded for the S-ATNL:Co(II):4-VPy (1:1:1)-based monolith. Selecting the "right" metal ion-ligand pair is a matter of a rational choice where the number of potential combinations can be significantly narrowed down by a preliminary spectroscopic screening. Therefore, other potentially promising combinations of metal-secondary ligands (Cu(II)–MAA, Cu(II)–1-VIM, Ni(II)–1-VIM) using various polymerization mixture constituents (cross-linkers, porogenic solvent, with and without ionic liquid) were also tested for the imprinting of S-ATNL, but in all cases suboptimal enantioselectivities were recorded (M10, M15, M21–24, α = 1.00–1.14) as compared to the best performing M2 column (α = 1.60). Unfortunately, even if UV-Vis spectroscopy is able to indicate some promising metal-secondary ligand combinations as starting points, assessing the optimal MMMI conditions is far from being a straightforward process. In addition to the formation of the best stable and soluble ternary metal ion-mediated complex, all the other constituents of the polymerization mixture will collectively play a critical part in the final chromatographic outcome, namely enantioselectivity.

It must be stressed that in the absence of the metal ion (i.e., Co(II)), molecular imprinting is achieved by the conventional non-covalent approach, with a poorer performance in terms of molecular recognition (M6, α = 1.30).

#### Functional Co-Monomers

Other co-monomers (i.e., AM, BAM, 4-PBA, CVPBA) added alongside the secondary ligand (4-VPy) had a negative effect on enantioselectivity (S9–10 and M11–13, α = 1.00–1.29). The interaction between co-monomers may reduce to a large extent the binding interactions with the template [41] and may also increase the heterogeneity of the resulting binding sites. Adding hydrophilic macromonomers to the polymerization mixture was reported as another convenient strategy to enhance the imprinting factor of polymers in case of water soluble templates [16]. In our case, instead of further boosting the hydrophilicity of the polymeric network, the addition of poly(ethyleneglycol) methyl ether methacrylate (PEGMA) fully compromised enantioselectivity (M7, α = 1.00).

#### Cross-Linker

Using EDMA instead of TRIM as a cross-linker turned out to be decisive for a more favorable polymer morphology and improved chiral selectivity, regardless of the type of column used (slurry packed–S10, α = 1.07 and S11, α = 1.32; monolith, M1, α = 1.00 and M2, α = 1.60). The beneficial effect of the ionic liquid on the imprinting efficiency was also unequivocally demonstrated by the selectivity value of column M2, α = 1.60 vs. column M5, α = 1.08. Moreover, chromatographic retention is inversely correlated with the recorded selectivity factors in the presence (M2, k'S = 3.7) and absence M5, k'S = 6.3) of [BMIM]BF4 most probably due to the increased stability of the ternary complex.

#### Ionic Liquid

The presence of the ionic liquid translates into a spectacular change in the morphology of the imprinted material. A highly porous, globular polymeric framework (Figure 2b, M2) is obtained in the presence of the ionic liquid, while a much denser structure is observed in its absence (Figure 2d, M5). Before template removal, a film-like structure of excess reagents covered the outer surface of the polymer (Figure 2a, M2); while upon washing with MeOH:AcOH 9:1 (*v*/*v*) a highly indented surface was revealed (Figure 2b, M2).

#### 3.2.3. Bulk Imprinting vs. MIP Monolith

MIPs are frequently prepared by bulk polymerization, in which case the obtained polymer block is ground and sieved before use. Bulk imprinting is more suitable for photochemical initiation at lower temperatures, which in theory should provide more homogenous and better defined imprinted sites. However, obtaining CSPs using this approach is cumbersome, wastes large amounts of material, and is often accompanied by the physical damage of some of the binding cavities. Furthermore, the resulting highly irregular MIP particles with polydisperse granulometry dramatically reduce the efficiency of the HPLC columns packed with such materials [42]. Imprinted monoliths polymerized directly in the chromatographic column offer a much simpler and long known alternative [43], at the cost of being only compatible with thermal radical initiation and only amenable to a few porogens. Nevertheless, such a continuous, but highly porous polymeric bed should in principle provide lower backpressure and greatly reduced mass transfer resistance, and thus higher chromatographic efficiency in comparison with the packed MIP-based columns. Room temperature ionic liquids (RTILs), demonstrating excellent solvation features, represent a more environment-friendly alternative as solvents for the preparation of MIPs [44]. As a result of their particular structure, they tend to promote self-assembly and improve specific template–monomer interactions, thus limiting non-specific binding [44,45]. In addition to offering higher imprinting factors and faster polymerization rates, they seem to reduce polymer swelling and boost the permeability of imprinted polymers, including monoliths [46]. Thus, in the search for the optimal experimental conditions that provide the highest enantioselectivity for ATNL's enantiomers, in our study both imprinting approaches were tested, in the presence or absence of RTIL.

Initially anticipating better molecular imprinting results, bulk imprinting with photochemical initialization at lower temperatures was investigated. Upon polymer grinding and sieving (particle size 25–38 μm) it was expected that all slurry-packed particulate columns demonstrate similar properties in terms of flow, back-pressure and sample load. Given the simplicity and earlier success of the OMNiMIP approach in chiral separation [47], N,O-bismethacryloyl ethanolamine (NOBE) was tested as a single cross-linking monomer for the imprinting of ATNL's enantiomers. At first (S1-3, Table 1), pure DMF was selected as porogenic solvent which enabled the solvation of the template enantiomer (S-ATNL). Unfortunately, neither the non-covalent (only NOBE), nor the metal ion-mediated (NOBE/Cu(II); NOBE/Co(II) imprinting approach gave notable enantioselectivity (α < 1.04) in any of the tested mobile phases. Nevertheless, adding a small amount of MAA to the NOBE-based polymerization mixture (S4, Table 1) enabled the solubilization of the template in acetonitrile (porogen), and endowed a certain degree of enantioselectivity (α < 1.17) of the resulting MIP using acetonitrile/formate buffer, pH 3.0 (85:15, *v*/*v*) as mobile phase. Seeking for improvements in the chiral recognition of the synthesized polymers, other functional monomers able to form ternary metal complexes with ATNL's enantiomer were screened by UV-Vis spectroscopy. One promising monomer candidate, forming soluble ternary metal (Cu(II), Co(II), Ni(II)) complexes with the target enantiomer, is 4-VPy (Figure 1). Therefore, different polymerization mixtures (S6–10, Table 1) combining various molar ratios of the template (T), monomer (M), metal ion (Me) and TRIM as cross-linker (C) were screened by metal ion-mediated bulk imprinting using photochemical initiation in a mixture of DMF:DMSO (1:1, *v*/*v*) as porogenic solvent. Unfortunately, none of the slurry-packed columns filled with S6-10 imprinted polymers exhibited noteworthy chiral discrimination (α = 1.00–1.04).

Seeking higher chromatographic efficiencies potentially able to distinguish between the enantiomers of ATNL, studies continued on monoliths polymerized in the chromatographic column using the same molecular imprinting approaches (M1–24, Tables 2 and 3), but with thermal radical initiation. The polymerization mixture was further adapted, exploiting the favorable influence of RTILs on the imprinting process and on the future chromatographic performances of the resulting continuous polymeric bed. Therefore, a hydrophilic ionic liquid, namely 1-butyl-3-methylimidazolium tetrafluoroborate ([BMIM][BF4]), was added to the porogenic solvent system. A higher ratio of IL/porogenic solvent system seem to favor the imprinting efficiency; thus finally a mixture of [BMIM][BF4]:DMF:DMSO = 10:1:5, *v*/*v*/*v* was selected [16,17]. Moreover, TRIM was replaced by a more polar cross-linker bearing numerous hydrogen bond acceptor motifs, namely ethylene glycol dimethacrylate (EDMA). Eventually, derived from the tested bulk imprinting polymerization mixtures, keeping a molar ratio of S-ATNL:Co(II):4Vp = 1:1:5 and EDMA as cross-linker (molar ratio M:C = 1:4) resulted a MIP monolithic column (M2) offering the best recorded selectivity factor (α = 1.60) in pure acetonitrile as mobile phase. Interestingly, in the absence of [BMIM][BF4] the resulting MIP monolith, (M5) is fully deprived of enantioselectivity (α = 1.00). Yet again, without the Co(II) ion mediating the interaction between the template and monomer during the radical polymerization, the enantioselectivity of the imprinted monolith (M6) drops to 1.30. Nevertheless, since a certain degree of chiral selectivity is still preserved by the M6 monolith, this suggests that in the current experimental conditions a mixed mechanism of imprinting (both non-chiral and metal ion mediated) is most likely to occur in case of M2 monolith. Keeping the same molar ratio of T:Me:M = 1:1:5 as for M2, the enantioselectivity of the resulting MIP monoliths for Ni(II) and Cu(II) (M3 and M4) were also investigated. Although in these cases a certain degree of selectivity for the template enantiomer has been observed (α = 1.38 for M3 and α = 1.30 for M4), it only matched the selectivity factor registered for the monolith obtained in the absence of the metal pivot (M6). The recorded changes in the electronic spectra of the binary and ternary mixtures of S-ATNL-Me(II)-4-VPy display a tendency in the formation of a stable ternary complex in the order Co(II) > Ni(II) > Cu(II) (data not shown). The latter trend correlates well also with the recorded efficiency of chiral recognition for monoliths M2–4. For the control experiments, the reference, non-imprinted monolith (M8) prepared in the absence of the template, but in the presence of Co(II), did not demonstrate any noticeable enantioselectivity (α = 1.04). Evidently, no enantioselectivity is observed for the non-imprinted monolith in the absence of the metal mediator (M9, α = 1.00).

#### *3.3. Chromatographic Retention Mechanism*

The influence of the mobile phase composition on the polymers' retention properties, regardless of the employed functional monomer (MAA, 4-VPy) or imprinting approach, indicated a severe decrease of the column capacity factor for both enantiomers of ATNL upon the addition of protic solvents, such as MeOH, acetic acid, or aqueous buffers (data not shown). This would suggest that hydrogen bonding plays a major role in molecular recognition of ATNL's enantiomers on the tested MIPs.

When using monomers with ionizable functional groups, such as MAA (M17–18), a relatively good efficiency with modest selectivity (α = 1.15–1.32) may be achieved in mixed aqueous(pH ≤ 5)-organic mobile phases (i.e., ACN: 50mM formate buffer (pH 3.0) = 85:15, *v*/*v*) where the partition equilibrium is most probably controlled by ion-exchange interactions (both specific and non-specific) between the amine of the template and the carboxylic groups of the imprinted polymeric structure. Nevertheless, using pure ACN (aprotic, weak solvent with intermediate polarity) as mobile phase, ATNL is totally retained by the polymer due to the synergistic effect of both the electrostatic and the additionally emerging hydrogen bonding interactions.

In the case of the 4-VPy-based MIPs (S6–11; M1–13, Tables 1–3), the ion-exchange mechanism is absent, thus retention and selectivity are mainly controlled by hydrogen bonding interactions. As already mentioned, using 4-VPy as functional monomer, enantioselectivity (where applicable) is only recorded in pure ACN. Chromatographic retention of ATNL drops dramatically (k'~1.05) and is accompanied by a complete loss of selectivity upon the addition in the mobile phase of minute amounts (0.1%, *v*/*v*) of protic solvents able to compete with hydrogen bonding.

Unfortunately, upon addition of the porogenic system to the mobile phase, the "memory effect" of the tested polymers could not be improved in the current experimental setup. Due to its strong molar absorptivity, even 1% (*v*/*v*) of DMF:DMSO 1:5 mixture added to ACN hindered the UV signal of the eluting enantiomers (data not shown).

Furthermore, the thermodynamics of the molecular recognition were assessed on the best performing 4-VPy-based monolithic polymer (M2). The retention and sorption selectivities were studied in the temperature range 20–50 ◦C at a flow rate of 0.2 mL min−<sup>1</sup> ACN. The influence of temperature on the retention (k') of enantiomers is shown in Figure 4.

As expected, the experimental van't Hoff plots recorded for the MMMI polymer (M2) for both enantiomers of ATNL are linear (r2 > 0.91), showing decreasing retention with the increase of temperature. The average value of the thermodynamic terms (−ΔHi/RT and ΔSi/R + lnϕ) [48] for the two enantiomers were also calculated for the studied temperature range (Figure 4 inset). For the imprinted polymer, the percent contribution of the entropic term (ΔSi/R + lnϕ) for both enantiomers lays around 40%. The binding of ATNL enantiomers implies more selective energetic interactions with the M2 monolith, rather than the entropically controlled steric interactions with the imprinted memory sites [49,50]. The average percent contribution of the steric term (ΔSi/R + lnϕ) for the binding of the template is slightly higher (40.6%) in comparison with the binding of its antipode (R-ATNL, ~39.5%). In case of the non-imprinted polymer (M8), identical van't Hoff plots (r<sup>2</sup> > 0.86) were obtained for both enantiomers. For this polymer, the average steric term contribution for the template and its antipode are somewhat smaller, 34.6% and 35%, respectively. These results indicate that energetic interactions (hydrogen bonding) are mainly responsible for recording the chromatographic retention, but the steric complementary (shape and size) of the emerging imprinted cavities also brings their contribution to the selective binding of the ATNL enantiomers in comparison with the reference, the non-imprinted polymer.

#### **4. Materials and Methods**

#### *4.1. Reagents*

Analytical grade standard S(-)-ATNL 98% was purchased from Toronto Research Chemicals (Toronto, Canada). R(+)-ATNL 99% was provided from Santa Cruz Biotechnology, Inc. (Dallas, TX, USA). 2-(trifluoromethyl)acrylic acid 98% (TFMAA) and vinylferrocene 97% (VFC) were purchased from Tokyo Chemical Industry (Tokyo, Japan). Methacrylic acid (MAA) 99%, pentaerythritol triacrylate (PETRA), pentaerythritol tetraacrylate (PETEA), trimethylolpropane trimethacrylate (TRIM), 2,2 -azobis(2-methylpropionitrile) 98% (AIBN), 4,4 -azobis(4-cyanovaleric acid) 98% (ACVA), 4-vinylpyridine 95% (4-Vpy) and 1-vinylimidazole 99% (1-VIM) were purchased from Aldrich (Steinheim, Germany). Tetrabutylammonium hexafluorophosphate 98% (4BA6FPh) was provided from Fluka (Steinheim, Germany). Ortho-phosphoric acid 85% (*w*/*w*), glacial acetic acid 100%, hydrochloric acid 37% (*w*/*w*), dimethylformamide 99% (DMF) and ammonium hydroxide 25% (*w*/*w*) pro analysi were purchased from Merck (Darmstadt, Germany). Formic acid 95% (*w*/*w*), 1-butyl-3-methylimidazolium tetrafluoroborate 98% ([BMIM][BF4]), acrylamide 99% (AM), *N*,*N* -Methylenebis(acrylamide) 99% (BAM) and boric acid were obtained from Sigma-Aldrich (Steinheim, Germany) and sodium hydroxide 99.3% from Lach-Ner (Neratovice, Czech Republic). HPLC grade solvents (acetonitrile (ACN), methanol (MeOH), butanol (BuOH), were obtained from Sigma-Aldrich (Steinheim, Germany) and used without further purification. Anhydrous Co(II) acetate 98%, Cu(II) acetate 98% and Ni(II) acetate 99% were purchased from Alfa Aesar (Kandel, Germany) and dimethyl sulfoxide 99.5% (DMSO) was from Carl Roth (Karlsruhe, Germany).

NOBE was synthesized by a previously published method [29].

All other chemicals were analytical reagent grade and were used as received.

Ultrapure water (18.2 MΩ, Barnstead EASYPure ROdi) was used for the preparation of all samples, buffers and related aqueous solutions. Phosphate buffer at various pHs was prepared by dissolving phosphoric acid in ultrapure water and adjusted accordingly with NaOH (1 M).

Stock solutions of 2 mg mL−<sup>1</sup> ATNL enantiomer were prepared in 2 mL volumetric flasks using ACN as solvent and were stored in the refrigerator at +4–6 ◦C.

#### *4.2. Apparatus*

Chromatographic experiments were performed with an Agilent 1200 HPLC system (Agilent Technologies, Waldbronn, Germany), equipped with a degassing unit, quaternary pump, autosampler, column oven and a diode-array detector. Signal acquisition and data processing were performed in the Chemstation B03.01 (Agilent Technologies, Waldbronn, Germany) software. The detection was performed at 200 nm and the flow-rate was 0.2 mL min<sup>−</sup>1. All the mobile phases were filtered through a 0.22 μm membrane filter from Millipore before use. Standard samples of atenolol enantiomers (30 μg mL<sup>−</sup>1) dissolved in HPLC grade ACN were injected in a volume of 10 μL.

The UV-Vis spectra were recorded in conventional (1 cm optical path) quartz cuvettes by a double-beam SPECORD® 250 PLUS (Analytik Jena, Jena, Germany) spectrophotometer in the range of 350–1100 nm. The electronic spectra of each ligand and the colored metal complexes were measured in the same solvent as the one used for the MIP preparation.

#### *4.3. MIP Preparation*

The molecularly imprinted and the non-imprinted polymers were synthesized under the conditions illustrated in Tables 1–3. The pre-polymerization mixtures intended for non-covalent imprinting were prepared by weighing the solid components, followed by the addition of liquid monomers and the adequate solvent(s). In the case of metal ion-mediated imprinting, the anhydrous metal ion salts (acetates) were dissolved in the adequate dry solvent or solvent mixture and ATNL was added to the solution. Subsequently, functional monomers, cross-linkers, ionic liquid and the free radical initiator were admixed. In a typical radical polymerization, 15 mg of AIBN has been employed, whereas in the OMNiMIP approach 24 mg of the same radical initiator has been added. The pre-polymerization mixture was sonicated for 15 min and further degassed by purging a gentle flow of nitrogen for 5 min. Depending on the desired form of the polymer (bulk or monolith), the vials or stainless-steel HPLC columns filled with the pre-polymerization mixtures were sealed and caped. In the case of bulk polymers, photopolymerization was performed both at room temperature (22 ± 2 ◦C) under a UV lamp for 24 h, in 5 mL sealed glass vials, whereas thermal polymerization was carried out in the BMT Ecocell convection oven at 60 ◦C for 24 h in 5 mL vials or stainless-steel HPLC columns. Removal of the template was achieved by Soxhlet extraction with MeOH/acetic acid (9:1, *v*/*v*), for 24 h. The polymers were then ground using a laboratory mortar and pestle and then sieved using standard testing sieves (ϕ = 200 mm, CISA sieves, aperture 38 and 25 μm, respectively), and the fraction between 25 and 38 μm was collected. The particles were slurry packed into stainless-steel columns (length 100 mm, internal diameter 2.1 mm) using an LC-10AT VP pump (Shimadzu, Japan) to full volume for HPLC analysis.

The imprinted monoliths were prepared by in situ thermopolymerization, in stainless-steel HPLC columns (length 100 mm, internal diameter 2.1 mm) sealed with Teflon tape and screwable caps, and kept at 60 ◦C for 24 h in a BMT Ecocell oven. The removal of the template was achieved by washing with MeOH/acetic acid (9:1, *v*/*v*). Control polymers were synthesized under the same conditions in the absence of the template.

#### *4.4. Microscopic Characterization of MIPs*

For scanning electron microscopy (SEM) characterization, the polymers (MIP and NIP) were metalized with gold in a Polaron E–5100 plasma-magnetron sputter coater (Polaron Equipment Ltd., Watford, UK) in the presence of argon (45 s at 2 kV and 20 mA). Ultrastructural images were obtained in a FEI Quanta 3D FEG scanning electron microscope (FEI, Hillsboro, ON, USA)) at 30kV and different magnification powers.

#### *4.5. Chromatographic Evaluation of The Imprinted Polymers*

#### 4.5.1. Bulk MIP

The template was removed from the imprinted polymers by Soxhlet extraction with MeOH:acetic acid (90:10, *v*/*v*) for 24 h. After grinding and sieving (25 and 38 μm), the polymer particles were slurry packed using a Shimadzu LC-10AT HPLC pump into steel columns (100 × 2.1 mm) to full volume for chromatographic experiments. As mobile phase various solvent systems were tested isocratically at 20 ◦C, at a flow rate of 0.2 mL min<sup>−</sup>1, starting with pure ACN and gradually switching to mixtures of ACN with increasing ratio of aqueous buffers (50 mM formate buffer (pH 3.0); 50 mM acetate buffer (pH 5.0) and 100 mM borate buffer (pH 9.3)).

#### 4.5.2. MIP Monolith

The template and excess reagents were removed from the imprinted polymer monolith by pumping through the column MeOH:acetic acid (90:10, *v*/*v*) at a flow rate of 0.2 mL min−<sup>1</sup> for 24 h. The columns were equilibrated with the corresponding mobile phase for 12 h at a flow rate of 0.20 mL min−<sup>1</sup> to remove any remaining template. If not otherwise stated, HPLC analyses were performed isocratically at 20 ◦C, at a flow rate of 0.2 mL min−<sup>1</sup> using pure ACN or a mixture of ACN/aqueous buffer in variable proportions as the mobile phase, monitoring the eluted analytes at a wavelength of 200 nm.

For all imprinted polymers the separation factor, α, was measured as a ratio of capacity factors k'S enantiomer/k'R enantiomer, with k' determined by the following relation: k' = (tr − t0)/t0, where tr is the retention time of the analyte and t0 is the retention time of the void volume measured using acetone as marker.

#### **5. Conclusions**

Improved chiral selectivity of the important β-blocker atenolol was achieved by the addition of a metal pivot which gave an imprinting factor of 1.60 versus the traditional molecular imprinted polymer formulation without the pivot which gave an imprinting factor of 1.32. Atenolol is a hydrophilic drug, and molecular imprinting in polar and/or aqueous phases is difficult for traditional molecular imprinting methods based on non-covalent hydrogen bonding or electrostatic complexes between monomers and template, due to disruption of the complex by the polar/aqueous porogenic solvent. The use of a metal to overcome this complex disruption in polar solvents and to coordinate the template and functional monomer led to approximately 23% improvement in chiral selectivity when using Co(II), and a 16% improvement when using Ni(II). This has an important impact due to the large demand for imprinting templates that are only soluble in highly polar and/or aqueous-based solvents. Furthermore, a 25% enhancement in enantioselectivity was found when using monolithic materials versus ground and sieved bulk imprinted polymers. Investigation of binding parameters showed that better selectivity was not a result of increased binding affinity (i.e., larger k' values) versus non-metal systems but was due to increase in differential enthalpic contributions of binding between the imprinted polymer and each enantiomer of atenolol. Thus, it can be concluded that the underlying mechanism of improvement of enantioselectivity of the imprinted polymer is due to the metal pivot approach maintaining fidelity of the imprinted site during polymerization. The choice of functional monomer was shown to be important based on the affinity of the functional monomer for the metal-pivot; in particular, 4-vinylpyridine or vinylimidazole did not disrupt important template-metal interactions whereas methacrylic acid displaced at least one of the template-to-metal interactions. This may be general for metal-pivot systems that require close proximity of the template to the metal for creating a selective binding site. In addition, the choice of crosslinker was important for optimum performance, for example, entry M2 shows that an EDMA crosslinked molecularly imprinted material provided 60% enhancement in enantioselectivity versus a nearly identical formulation using TRIM. Other molecularly imprinted materials using the crosslinkers PETRA and PETEA also showed little to no enantioselective performance, supporting the conclusion that the difunctional crosslinker EDMA is required for these systems versus any trifunctional crosslinkers.

**Author Contributions:** Conceptualization, A.E.B., B.-C.I., D.A.S. and E.B.; Data curation, A.E.B., B.-C.I., S.L.O.O. and E.B.; Investigation, A.E.B., B.-C.I., L.E.G. and N.A.G.; Methodology, A.E.B., B.-C.I., D.A.S. and E.B.; Resources, S.L.O., D.A.S. and E.B.; Writing—original draft, A.E.B., B.-C.I., D.A.S. and E.B.

**Acknowledgments:** This work was supported under the contract funded by the University of Medicine and Pharmacy "Iuliu Hatieganu" Cluj-Napoca, internal grant No. 4944/11/08.03.2016 and partly by a grant of Ministry of Research and Innovation, CNCS-UEFISCDI, project number PN-III-P1-1.1-TE-2016-0628, within PNCDI III. David Spivak and Nicholas Gariano were supported by the National Science Foundation under grant CHE-1411547. The ultrastructural characterization of the polymers by Lucian Barbu-Tudoran, as well as the help in the spectroscopic studies offered by Tudor Rusan is greatly acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Not available.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Chiral Stationary Phases for Liquid Chromatography: Recent Developments**

**Joana Teixeira 1, Maria Elizabeth Tiritan 1,2,3 , Madalena M. M. Pinto 1,2 and Carla Fernandes 1,2,\***


Received: 31 January 2019; Accepted: 26 February 2019; Published: 28 February 2019

**Abstract:** The planning and development of new chiral stationary phases (CSPs) for liquid chromatography (LC) are considered as continuous and evolutionary issues since the introduction of the first CSP in 1938. The main objectives of the development strategies were to attempt the improvement of the chromatographic enantioresolution performance of the CSPs as well as enlarge their versatility and range of applications. Additionally, the transition to ultra-high-performance LC were underscored. The most recent strategies have comprised the introduction of new chiral selectors, the use of new materials as chromatographic supports or the reduction of its particle size, and the application of different synthetic approaches for preparation of CSPs. This review gathered the most recent developments associated to the different types of CSPs providing an overview of the relevant advances that are arising on LC.

**Keywords:** liquid chromatography; enantioseparation; chiral stationary phase; chiral selector; chromatographic support

#### **1. Introduction**

Now more than ever, analytical and preparative enantiomeric separations play a crucial role in industry and academic research [1]. There are a wide variety of methods to achieve and analyze enantiomerically pure compounds, including liquid chromatography (LC) [2,3], supercritical fluid chromatography [4–6], diastereomeric crystallization [7,8], membranes [9,10], asymmetric catalysis [11], simulated moving bed [12,13], dynamic and enzyme-mediated kinetic resolution [14,15], among others.

LC using chiral stationary phases (CSPs) proved to be an essential tool with a wide range of applications, including preparative separation of enantiomers of diverse analytes [16,17], determination of enantiomeric composition [18,19], monitorization of asymmetric reactions [20,21], analysis of the stereochemistry of natural compounds [22,23], pharmacokinetic [24,25], forensic [26–28], environmental [29–31], and enantioselective studies [32,33], among others.

The development of CSPs for LC combined with the improvement of chromatography instrumentation revolutionized the enantioseparation approaches. LC using CSPs has demonstrated to be extremely useful, accurate, versatile, and it has been a widely used technique in diverse fields and applications, emphasizing, for example, the enantioseparation of underivatized amino acids [34,35], diverse classes of pharmaceuticals [36–40], atropisomers [41], as well as the study of intermolecular interactions between biomolecules and drugs [42], among others.

Over the last decades, several types of CSPs have been developed [43–46] and, among them, more than a hundred are currently commercially available [39]. These comprise Pirkle-type, ligand-exchange-type, molecularly-imprinted, and based on macrocyclic antibiotics, proteins, polysaccharides, cyclodextrins, crown ethers, cyclofructans, synthetic polymers, among others [43–46]. Nevertheless, although many different types of CSPs are described, the development of new CSPs continues to be a field of research with great importance.

#### **2. Chiral Stationary Phases: Recent Developments**

Since the first description of a CSP, in 1938, by Henderson and Rule [47], and to follow the constant challenges on different areas as well as the advances in chromatographic instrumentation, the development of new CSPs for LC has been a continuous and evolutionary subject.

In this review, for each type of CSPs, the most recent CSPs were presented describing the strategies used for their development (Figure 1).

**Figure 1.** Summary of recent strategies for development of new chiral stationary phases (CSPs) for liquid chromatography (LC).

Herein, only the CSPs that were not reported in previous fundamental reviews will be presented, highlighting the main advantages of the strategy used for their preparation, the best chromatographic results and the objectives to be achieved.

#### *2.1. Polysaccharide-Based CSPs*

The first application of a polysaccharide as a chromatographic chiral selector was described by Hessen and Hagel, in 1976 [48]. Since then, different polysaccharides were extensively used as CSPs due to their high enantioselectivity properties after derivatization [49]. Nevertheless, amylose and cellulose are the main polysaccharides used to obtain CSPs [50], followed by chitosan and chitin [51]. The chiral recognition ability of polysaccharide derivatives is dependent on diverse structural features, including sugar units, stereogenic centres of the glucopyranose units, type of the linkage and its position, as well as the adjacent polymer chains [49]. The helical twist of the polymer backbone is also essential for enantioselectivity [43].

Polysaccharide derivatives as efficient chiral selectors can include phenyl, alkyl or benzylcarbamates, esters, benzoate, or aryl or cycloalkyl groups [52]. Benzoate or phenylcarbamate moieties may comprise methyl, methoxy, among other groups, and/or chlorine substituents in the aromatic ring [52], affording different solubility and chiral recognition ability [52]. Moreover, the position of the substituents in the aromatic ring influences the enantioseparation performance of the chiral selector [49].

Polysaccharide derivatives can be coated onto a chromatographic support, as silica or derivatives, by an adsorption process [53,54] allowing a larger surface area [55] and high efficiency [43]. CSPs comprising coated polysaccharide derivatives can operate in normal phase, polar organic and reversed-phase elution mode; however, they have restrictions due to the non-compatibility with "non-standard" solvents, such as dichloromethane, chloroform, toluene, or acetone [43]. The use of those solvents in the mobile phase may cause the dissolution of the adsorbed polymer and, consequently, removal of the selector from the chromatographic column [43]. Immobilized polysaccharides emerged as a reliable alternative allowing the use of a broader selection of solvents as mobile phases [56–60]. Different procedures can be used for covalently bonded the polysaccharide derivatives to the chromatographic support, such as a polymerization reaction and photoinduced and enzymatic polymerization [56]. Nevertheless, despite the solvent versatility, in general, the potential of chiral recognition of immobilized polysaccharide-based CSPs is lower than the coated due to modification of stereospecific conformation that can occur during the immobilization process [56,57].

This type of CSPs is recognized as being the most successful and widely applied for both analytical [61–69] and preparative enantioseparations [17,70–76], being responsible for about 99% of reported chiral separations [50]. Among the developed polysaccharide-based CSPs, the 3,5-dimethylphenyl *tris*-phenylcarbamates of amylose and cellulose proved to have the best enantiorecognition performance [77–80]. In our group, this type of CSPs has proved to be effective for analytical as well as preparative applications [26,81–84].

The chiral recognition mechanisms concerning these CSPs are not yet completely understood. In an attempt to improve the knowledge related with structural features associated with the chiral recognition mechanisms and their chromatographic behavior at a molecular level, several studies concerning to docking, spectroscopy, molecular modelling, and quantum chemical calculations were recently performed and compiled by Scriba et al. [45].

Several reviews have assembled the advances on preparation and evaluation of this type of CSPs over the years [43,45,49–52,57–59,80,85–93]. Nevertheless, this research field is always evolving being the most recent polysaccharide-based CSPs presented on Table S1 (supplementary material).

Recent developments on polysaccharide-based CSPs comprise different approaches, with the general objective being the improvement of the enantioseparation performance. The strategies include the introduction of new polysaccharide derivatives (mainly new chitin and chitosan derivatives but also cellulose derivatives), hybrid selectors, and different chromatographic supports (monoliths, core-shell, microspheres), as well as the application of different methodologies for coating or immobilization procedures. Generally, for the new derivatives, the effect of different substituents on chiral recognition has been discussed [94–97]. Additionally, the effect of mobile phase composition on enantioseparation was also explored [94–96,98,99].

Recently, Han et al. [94] developed two CSPs, using a derivative of cellulose *tris* (3,5-dimethylphenylcarbamate) (CSP1) and the same derivative functionalized with carboxylic acid (CSP2) (Figure 2). They concluded that a large variety of substituents could avoid the chiral recognition properties of the cellulose derivatives, reducing the performance of the CSP. The best chromatographic results were obtained for *trans-*stilbene oxide, with α and Rs values of 1.84 and 9.59, respectively.

**Figure 2.** Chemical structures of polysaccharide-based CSP1–56 and CSP61.

Shen et al. [100] synthetized cellulose derivatives with different combination of carbamate substituents and prepared 25 new CSPs (CSP3–27) (Figure 2). The effect of the carbamate substituents at 2,3-positions and 6-position of the glucose moiety were the main focus of the study. It was found that the chiral recognition properties of the CSPs comprising derivatives with two different phenylcarbamates were higher than if CSPs only had one substituent. The resolution was improved by the presence of different carbamate substituents, suggesting that the chiral recognition was dependent on the electronic properties, position and number of substituents of the glucose unit [100]. The highest separation factor obtained by using these recent CSPs was 2.87, for Pirkle alcohol.

Chitin and chitosan-based CSPs have received particular attention in the last few years [51]. Through continuous efforts to develop effective CSPs, other recent reports describing the use of chitin [101,102] and chitosan [95–97] derivatives are rising, with the carbamates as one of the most studied [49]. The growing interest in these polysaccharides comes from the fact that they have low solubility, which allows the use of a wide variety of mobile phases [52]. The influence of substituents on chitin and chitosan derivatives was also investigated. For some analytes, these CSPs possessed an enhanced chiral recognition when compared to cellulose and amylose derivatives, which may be attributed to the variety of solvents that can be used [103].

Tang et al. [97] developed eight CSPs (CSP28–35) comprising chitosan 3,6-*bis*(arylcarbamate)-2 -(*p*-methylbenzylurea) with diverse substituents in the aromatic rings of the carbamates as well as in the amide group (Figure 2). Selectors with electron-donating substituents demonstrated a higher ability of enantioseparation. Previous reports emphasized that an electron-donating substituent at the 4-position of the aromatic ring was beneficial for the chiral separation [44]. Despite the selectors with 4-methyl substituent and 3-chloro-4-methyl portion presented a superior enantioseparation, the highest resolution (Rs = 18.1) and separation factor (α = 6.72) were obtained by the CSP with 3,5-dimethyl substituent [97].

In another study, Zhang et al. [95] prepared seven CSPs (CSP36–42) comprising derivatives of chitosan *bis*(phenylcarbamate)-(*N*-cyclobutylformamide) (Figure 2). The same substituent on different positions resulted on modifications on the suprastructure of the selector leading to different size of cavities, for example, due to different electronic effects. The obtained CSPs proved to have considerable stability on different solvents and a good enantiorecognition, allowing the researchers to obtain a separation factor of 8.64 for voriconazole [95].

Other new chitosan-based CSPs were developed, in this case, comprised of derivatives of chitosan (*bis*(methylphenylcarbamate)-(isobutyrylamide)) (CSP43–48) (Figure 2) [96]. The introduction of some substituents on specific positions of the aromatic ring linked to the carbamate were favorable for enantioseparation, such as methyl substituents. Additionally, the low solubility of chitosan was proved to be an advantage for the solvent tolerance and good enantioresolution performance achieved. As an example of its performance, high enantioselectivity and resolution were obtained for voricomazole, with α and Rs values 4.32 and 11.9, respectively [96].

Zhang et al. [102] synthetized derivatives of chitin using three different phenyl isocyanates (4-trifluoromethoxy, 3-chloro-4-methyl, 4-chloro-3-trifluoromethylphenylcarbamate) to develop three CSPs (CSP49–51) (Figure 2). All CSPs were applied for enantioseparation of tadalafil and its intermediate, demonstrating great enantiorecognition potential, with resolution and separation factor values of 4.72 and of 2.15, respectively [102].

Mei et al. [101] derivatized natural and regenerated chitins with 3,5-dimethyphenyl isocyanate, to prepare CSP52 and CSP53–55, respectively (Figure 2), with the difference between them only related to the raw material. The regenerated chitins were obtained from natural chitins, after a treatment with acetic anhydride, showing a more promising performance. They pointed out that CSP prepared from selectors with lower molecular weight provided an improved resolution [101]. The best chromatographic results were obtained for voricomazole, with Rs = 11.7 and α = 3.06.

Another strategy was the development of hybrid selectors [98]. Hybrid selectors or biselectors comprise two different polysaccharide derivatives coated on a chromatographic support [104]. Zhang et al. prepared the CSP56, which comprised a biselector based on derivatives of amylose and chitin (Figure 2), combining the low solubility of chitin derivatives with the excellent chiral recognition properties of amylose derivatives [98]. The obtained CSP presented an improved resistance against organic solvents with high enantioselectivity, with a Rs value of 8.49 for mephobarbital, and an α value of 4.32 for 2-(5-chloro-2-((4-methoxybenzyl)-amido)phenyl)-4-cyclopropyl-1,1,1-trifluorobut-3-yn-2-ol.

Regarding the use and preparation of chromatographic supports, a new technique of encapsulation in organic polymer monolith was reported, by Fouad et al. [103], as an alternative preparation of the chromatographic support. They functionalized an organic polymer monolith with carbamylated amylose as selector to obtain the CSP57. The synthesis of the amylose derivative was described previously [105]. The encapsulation of amylose was described as an economic methodology and it allowed the conjugation of amylose with reversed-phase elution mode for several analytes [103]. This promising technique allowed good results with a maximum resolution of 2.80 and a separation factor of 3.90 for the testes analytes.

Another different approach for preparation of chromatographic support was reported by Li et al. [99], who functionalized core-shell silver particles with cellulose derivatives through coating and intermolecular polycondensation and developed CSP58. A synergetic effect between silver and cellulose was observed considering the high values of resolution (Rs = 2.61) and enantioseparation (α = 8.42). This CSP demonstrated a particular selectivity toward analytes having the functional group ketone [99].

In another study, Bezhitashvili et al. [106] reported the covalent immobilization of a cellulose derivative (cellulose-(3,5-dichlorophenylcarbamate)) onto the surface of core-shell particles to obtain CSP59. The synthesis of the cellulose derivative was described previously [77]. The click chemistry was the synthetic methodology applied for the immobilization of the cellulose derivative to the chromatographic support [107]. The authors emphasized the short time of analysis achieved with baseline separations [106]. The highest separation factor obtained was 15.3 with a resolution of 11.0 for 2-(4-methylbenzylsulfinyl)-benzamide.

Huang et al. [108] developed a new methodology for coating cellulose *tris*(3,5-dimethylphenyl carbamate) derivative on silica microspheres, without any surface pre-treatment since no aggregation occurred, and prepared CSP60 (range of pore size 10–150 nm). The synthesis of the cellulose derivative was previously described [109]. The silica microspheres with reduced size have functionalized polymeric beads being highly crosslinked [108]. This technique allowed a high-loading of the chiral selector and the obtained CSP provided a good performance, being the best separation factor of 2.41 for 2,2,2-trifluoro-1-(9-anthryl) ethanol.

The CSP61 (Figure 2) was reported by Vieira et al. [110] using a technique of thermal immobilization of cellulose dodecanoate on silica particles without the use of chemical reagents. Despite of the absence of a chemical reagent during the procedure of immobilization, the selector was strongly linked to the chromatographic support allowing an exceptional selectivity. Some advantages of the immobilization technique were highlighted, including its low cost and eco-friendly feature [110]. The separation factor obtained was 3.10.

Besides the recent developments on polysaccharide-based CSPs, it is important to emphasize that this type of CSPs also cover a wide range of recent applications [111–113]. For example, Padró et al. [89] recently reviewed applications of polysaccharide-based CSPs in different fields. Studies comparing the enantioresolution performance of coated and covalently immobilized CSPs based on polysaccharides were also found [114].

Additionally, the influence of mobile phase is a common focus in several studies [81,115].

#### *2.2. Protein-Based CSPs*

The intrinsic ability of chiral recognition by enzymes, plasma proteins and receptors inspired the application of proteins in enantioseparation techniques [43]. Proteins are complex structures with a large surface area comprising a variety of stereogenic centers and different binding sites, which allow multiple possibilities of intermolecular interactions with small molecules [55]. The first application of a protein as CSP was reported in 1973, describing the separation of tryptophan enantiomers using a bovin serum albumin (BSA)-sepharose CSP [116]. After this first report, many CSPs based on proteins have been developed, with the most used proteins the human serum albumin (HSA), α1-acid glycoprotein (AGP), crude ovomucoid (OVM), and cellobiohydrolase I (CBH I) [86]. All these proteins as chiral selectors have been well documented for chromatographic enantioseparation for a wide range of chiral compounds and for binding affinity studies [117].

Proteins as CSPs are applied on affinity and pharmacokinetic studies since they can mimic the in vivo systems [118], being this feature very important in drug discovery [43]. The possibility of using aqueous or aqueous-organic as mobile phases was pointed out as other advantage of protein-based CSPs considering its compatibility with mass spectrometric detection. The disadvantages of this type of CSPs are the low capacity and efficiency. Moreover, the possibility of denaturation of protein limit the ranges of pH, ionic strength, temperature, and organic modifier composition of mobile phase [88], which is a result of its reduced chemical and biochemical stabilities.

HSA is the most applied protein-based CSP, and it is used predominantly on studies of drug-protein binding [43]. In separation techniques, it is applied for weakly acidic and neutral compounds [119] as well as zwitterionic molecules [120]. For preparation of this type of CSPs, the protein can be physically adsorbed onto the packing material or it can be covalently bound [121].

A number of reviews on CSPs based on proteins have appeared over the years, focusing on their developments and applications [43,45,50,85,87,88,118,120–122]. Recently, Bocian et al. [117] briefly reported several studies related to protein-based CSPs, including the most common (HSA and AGP) as well as the more unusual, namely, avidin and fatty acid binding proteins. The developed strategies presented on that review were mainly related with the chromatographic support as the introduction of monoliths. The introduction of new selectors was also described [117]. Bertucci and Tedesco [42] recently reviewed the advances concerning the HSA as chiral selector highlighting the application of competitors for a particular binding site of HSA as the greatest advance. Scriba et al. [45] compiled some studies, focusing on the understanding of the binding sites and the main interactions between protein-based CSPs with diverse analytes, by molecular modeling.

The most recent protein-based CSPs was not reported in those reviews; their chromatographic performance are presented on Table S2 (supplementary material). It was found that different proteins as new selectors were not introduced. Nevertheless, diverse techniques of preparation of chromatographic support, namely new techniques of immobilization [42,117,123], entrapment [124], or the application of monoliths [125] have been reported attempting to overcome the stability problems of the proteins.

Proteins are usually immobilized on silica and its pore size can be defined in order to optimize the separation. Matsunaga and Haginaka [126] immobilized AGP on silica particles with different sizes, 5, 3 and 2.1 μm (CSP62–64). The use of this protein as chiral selector was reported, in 1985, by Hellerstein et al. [127]. In another study, Bi et al. [124] entrapped the same protein on a silica support with 100 Å and 300 Å (CSP65–66). Relatively to the first study mentioned, the resolution and efficiency of CSPs with lower particle size were superior. As an example, an excellent resolution was obtained for benzoin (Rs = 14.2) [126]. In the latter study, the described methodology using CSP65–66 proved to be an alternative to high-throughput screening and analysis of biological interactions due to the good affinity results, a maximum of 2.10 × 106 <sup>M</sup>−<sup>1</sup> [124].

Matsunaga and Haginaka [128] also studied the effect of particle size, on the efficiency of the column, with cellulase as chiral selector (CSP67). The first application of cellulase as chiral selector was reported by Vandenbosch et al. in 1992 [129]. As in the previous studies, the column with the lowest particle size provided the greatest efficiency and enantioselectivity [128], with a resolution value of 10.7 for propranolol, for example.

Zheng et al. combined a covalent immobilization process with a cross-linking/modification methodology, using HSA as chiral selector, to achieve CSP68. The aim of this new immobilization strategy was to enhance the protein retention [130]. The CSP obtained presented a high binding affinity for warfarin, with a maximum affinity constant of 2.60 × <sup>10</sup><sup>5</sup> <sup>M</sup><sup>−</sup>1.

A polyclonal antibody CSP (CSP69) was developed by Bi et al. [131] through an alternative methodology, which consisted of the isolation and immobilization of the selector presented on a serum sample (on-line immunoextraction). This methodology was previously described by Matsuda et al. [132]. Once again, it was found that it could be an alternative to the traditional immobilization methodologies since it is not necessary extra steps of protein purification and immobilization [131]. The stability and the robustness of the CSP were also highlighted [131]. This methodology allowed the preparation of a CSP with considerable affinity. For example, a binding affinity value of 90.0 × 106 <sup>M</sup>−<sup>1</sup> was obtained for disopyramide.

A new protein-based CSP was reported by Fedorova et al. [133] using a different adsorption methodology, which consisted of BSA adsorbed on eremomycin and grafted on silica (CSP70). An improved resolution, with a good resolution (Rs = 2.14) was provided, in comparison with a CSP comprising only eremomycin.

One of the most recent developments concerning this type of CSPs was the functionalization of monoliths with proteins. Monoliths are based on silica and present the advantage of optimizing the proportion of monomers and cross-linkers. This optimization enables the control of the average size of the throughput channels and the porous [134]. Monoliths can be prepared with different materials and techniques; the advantages comprise a superior flow as well as an enhanced mass transfer resulting on a more efficient separation [121]. Pfaunmiller et al. [125] immobilized HSA on monoliths to obtain CSP71–72. The main objective was to optimize the amount of protein that could be immobilized. As a consequence, the prepared monoliths allowed an improvement on all chromatographic parameters [125].

The recent applications described for this type of CSPs are more diversified than the developments and some of them are related with the optimization of the chromatographic conditions [133–135]. Nevertheless, the number of publications describing the use of protein-based CSPs has been decreasing over the years [117,136]. Binding affinity studies between drugs and proteins and drug-protein interactions were also found [137–139].

#### *2.3. Cyclodextrin-Based CSPs*

The first application of cyclodextrins as CSPs was described by Armstrong and DeMond in 1984 [140]. Since then, several cyclodextrin-based CSPs have been reported [137,140–142]. Cyclodextrins consist on cyclic oligosaccharides [88]; this type of macrocycles can be divided into three classes, α, β, and γ [43]. The structure of a cyclodextrin consists on a truncated cone [43] with an interior non-polar cavity and free hydroxyl groups located on larger and tiny edges [143]. The hydroxyl groups can be derivatized with diverse polar or apolar substituents [55], which can influence the conformational flexibility of a given cyclodextrin, modifying the size of its cavity and creating additional binding sites [43].

The chiral recognition mechanism is typically based on the formation of an inclusion complex between the analytes and the internal cavity of the cyclodextrin [43]. Additionally, the analytes can establish different types of interactions with the exterior side, including dipole-dipole, hydrogen-bond, ionic, π-π, or London interactions [120]. Cyclodextrins present a considerable number of stereogenic centres, which also contributes to enantiorecognition [55].

Cyclodextrin derivatives can be prepared through physical coating or covalently bonding to a chromatographic support [144]. Covalent bonding of cyclodextrin derivatives is the most applied methodology, since it provides a powerful and resistant linkage to the chromatographic support. The most common linkers are ether, amino, and urea. Recently, a triazole linker was also described [144].

The high stability of this type of CSPs allows the use of an extensive variety of solvents as components of mobile phases, with a wide range of polarities, affording an efficient enantioseparation for different analytes [55]. This type of CSPs can be applied in multimodal elution conditions [142].

Several reviews have been devoted to the developments and applications of cyclodextrin-based CSPs [43,45,85,87,88,140,142,144]. Additionally, Guo et al. [145] reviewed the most recent developments concerning on cyclodextrin functionalized monolithic columns.

Studies related to chiral recognition mechanisms of this type of CSPs using diverse methodologies, such as nuclear magnetic resonance, docking, or molecular modeling, were also addressed to understand the molecular interactions as well as the effect of some chromatographic conditions, such as pH, temperature, or organic modifier, in the enantioseparation [146,147].

The recent cyclodextrin-based CSPs and the evaluation of their chromatographic performance are described on Table S3 (supplementary material). The most recent developments are comprised mainly of the introduction of new derivatives and application of different methodologies of immobilization to the chromatographic support. The preparation of hybrid CSPs to enhance the interactions between the analyte and the stationary phase was also emphasized. It was found that the majority of the new cyclodextrin-based CSPs were prepared based on the most widely used cyclodextrin as CSP, i.e., the β-cyclodextrin [148–151]. Moreover, the immobilization strategy of the chiral selector on the chromatographic support was, mainly, by click chemistry [149–152]. The main advantages of this approach are the mild reaction conditions and the enhanced tolerance of the CSPs to solvents and the range of pH values [153]. The introduction of new methodologies to prepare the chromatographic support was also focused, including the preparation of hybrid supports [154,155], the introduction of monoliths [156] and new chromatographic supports [157], or using a different technique to prepare the support [158].

Zhou et al. [152] reported the linkage of a C6-disubstituted cationic β-cyclodextrin onto an alkynylated β-cyclodextrin bonded to a silica support to afford the CSP73 (Figure 3). The obtained bilayer cationic β-cyclodextrin presented a remarkable enantioselectivity for the tested analytes. As an example of its excellent enantioseparation and resolution, α and Rs values of 2.39 and 4.40, respectively, were obtained for 4-(chlorophenyl) propyl ester [152].

Tang et al. [151] resorted to thiol-ene click chemistry to prepare a sulfoether-bridged cationic per(3,5-dimethyl) phenylcarbamoylated-β-cyclodextrin-based CSP (CSP74) (Figure 3) being able to establish π-π interactions and hydrogen bonding interactions with the tested analytes. Its enantiorecognition ability was demonstrated by a separation factor of 1.70 and resolution of 6.03 for 3-(chlorophenyl) propyl ester.

Zhou et al. prepared a perphenylcarbamate β-cyclodextrin chloride linked by click chemistry to an alkynyl silica support to obtain the CSP75 (Figure 3) [159]. After evaluation of its enantioseparation performance using diverse analytes, they concluded that the introduction of the 3-methoxypropylammonium substituent promoted favorable intermolecular interactions with the analytes. In addition, it was suggested that the mobile phase could cause steric hindrance which prevented the establishment of interactions that were crucial for enantiorecognition [159]. The performance of the CSP was promising with a maximum resolution value of 9.84 and a separation factor of 2.76 for 7-methoxyflavanone and 6-methoxyflavanone, respectively.

A new *N-*benzyl-phenethylamino-β-cyclodextrin was synthetized and bonded to mesoporous nanoparticles of silica obtaining the CSP76 (Figure 3) [149]. The new CSP demonstrated to have a superior flexibility and stability, in comparison with the native β-cyclodextrin-based CSP, being obtained through a more economic process [149]. Relatively to its performance, the higher separation factor and resolution values were 1.30 and 1.97, respectively, for carvedilol.

Four new cyclodextrin-based CSPs (CSP77–80) (Figure 3) were developed by chemical bonding of carboxymethyl-β-cyclodextrin derivatives to silica gel by an amidation reaction on aqueous solution [160]. The carboxymethyl moiety provided additional interactions with the tested analytes, in comparison with the native β-cyclodextrin demonstrating a superior enantioselectivity and resolution [160]. For example, an excellent separation factor value (α = 6.08) was achieved for methyl 2-amino-3-(3-(methylsulfonyl)phenyl)-propanoate hydrochloride. Moreover, for 1-((benzyloxy)carbonyl)-4-hydroxypyrrolidine-2-carboxylic acid, the resolution value was 9.56 [160].

**Figure 3.** Chemical structure of cyclodextrin-based CSP73–100.

The effect of different substituents on CSPs has also been investigated. Chen et al. [148] synthetized β-cyclodextrin derivatives with a phenylcarbamate moiety with different patterns of substituents, which were subsequently immobilized onto the silica gel through intermolecular polycondensation of the triethoxysilyl groups (CSP81–85) (Figure 3). They reported that the presence of an aromatic ring with electron-withdrawing groups on the β-cyclodextrin improved the chiral recognition for analytes with electron-donor groups since the number of possible π-π interactions was superior [148]. The hydrogen-bond interactions between the carbonyl group or nitrogen of the analytes and the amino group of phenylcarbamate of cyclodextrin-based CSP were also improved [148]. Relatively to the performance of the CSPs, the highest separation factor value achieved was 2.87 for Pirkle alcohol.

Li et al. [150] arrived at similar conclusions after preparing oxazolinyl-functionalized β-cyclodextrins covalently bonded to silica support (CSP86–88) (Figure 3). They described that analytes with electron-donating or hydrogen-bonding groups were easily enantioseparated due to a higher number of π-π and hydrogen-bonding interactions [150]. CSP88 was more suitable for enantioseparation of polar compounds since it promoted electrostatic interactions due to the presence of an ionic group [150]. Additional factors that could influence the enantioseparation performance of the CSPs, such as the spacer length, selector concentration, and rim functionalities, were also investigated [150]. A reduced surface concentration and a superior flexibility of the spacer decreased the enantioselectivity, since it weakened the interactions between the selector and the analyte [150]. Additionally, a superior selector concentration could be beneficial for enantioseparation of some racemates. The performance of the CSP was promising with an excellent resolution for ketoprofen (Rs = 22.0) and a separation factor value of 15.5 for loxoprofen.

The same group developed four thioether bridged cationic cyclodextrin-based CSPs (CSP89–92) (Figure 3) and the influence of the spacer length, selector concentration, and rim functionalities on the performance of the CSP were studied [161]. In this case, it was found that CSPs comprising a spacer with a superior length could compromise their ability of enantiorecognition; however, a higher concentration of the selector was positive [161]. The higher resolution value achieved was 12.7 for 4-nitrophenyl propyl oxide, and the separation factor value was 3.30 for styrene oxide.

Regarding the development of hybrid CSPs, a spherical β-cyclodextrin-silica hybrid CSP (CSP93) (Figure 3) was reported by Wang et al. [154] highlighted by the presence of multiple functional groups, which expanded the spectrum of possible interactions with analytes. The β-cyclodextrin derivative was introduced into the pore channels and pore wall framework, and the linker was attached just into the pore channels [154]. Both the interior and exterior of the CSP participate in the process of the enantiorecognition. For example, separation factor and resolution values of 1.63 and 4.65 were obtained for diclofop and mandelonitrile, respectively.

Regarding the use of new materials as chromatographic supports, a modification of the most common chromatographic support was performed by Zhao et al. [155] to obtain new CSPs. The modified silica gel was named hydride silica, and its surface was covered by silica-hydrogen bond instead of silica-hydroxyl. The hydride silica presents a superior resistance to water, a reduced polarity, and an improved separation rate and stability and it can be used with a wide variety of solvents [155,162]. Zhao et al. [155] prepared polar group derivatives of β-cyclodextrin bonded to hydride silica to obtain CSP94–97 (Figure 3). The higher resolution value achieved was 9.31 for methyl (2*R*,3*R*,4*S*,5*R*)-5-(4-fluorophenyl)-4-nitro-3-phenyl-3-(trifluoromethyl)-pyrrolidine-2-carboxylate; the best separation factor value was 3.65 for methyl (2*R*,3*S*,4*S*,5*R*)-5-(4-fluorophenyl)-4-nitro-3-(*p*-tolyl) pyrrolidine-2-carboxylate.

Ghanem et al. [156] used a different strategy to prepared new cyclodextrin-based CSPs. They encapsulated the trimethylated-β-cyclodextrin to a polymeric monolithic, to obtain a superior surface area, reduced pore size, and enhanced total pore volume, and developed the CSP98. They also studied the physical characteristics of the CSP to established relationships with the potential of enantiorecognition and concluded that a superior concentration of the selector improved the enantioseparation [156]. The CSP98 demonstrated a suitable mechanical and thermal stability as well

as reproducibility [156] with a maximum resolution value of 2.51 for flavanone, and a separation factor of 1.42 for carprofen.

A different chromatographic support was also proposed by Qiang et al. [157], who described a β-cyclodextrin CSP based on graphene oxide (CSP99) (Figure 3), which was covalently linked to amino silica gel by an amide bond. The graphene oxide and cyclodextrin presented a synergetic effect for enantiorecognition being the hydrogen bonding and π-π interactions the main interactions between the CSP and analyte. The CSP99 was also applied for hydrophilic interaction chromatography. Regarding the chromatographic results, a separation factor of 38.8 was achieved for equol, and a resolution value of 2.17 for 1-phenylethanol [157].

A light-assisted preparation of carboxyl methyl β-cyclodextrin-based CSP (CSP100) (Figure 3) was described by Tang et al. [158] who used ultra-violet light to link the chiral selector to silica, which promoted the modification of ionic bonds into covalent bonds. This technique proved to be eco-friendly and efficient [158]. The morphology and chemical composition of CSP100 was characterized. Moreover, it was concluded that its enantiorecognition ability was dependent of hydrogen bonding and dipole-dipole interactions [158]. The maximum resolution value achieved was 8.04 for chlortrimetron.

#### *2.4. Macrocyclic Antibiotic-Based CSPs*

Macrocyclic antibiotics are the second most versatile group of CSPs, after polysaccharides; their planning was inspired by cyclodextrins. The first report of macrocyclic antibiotics as CSP was in 1994, describing the application of vancomycin, thiostrepton, and rifamycin B as CSPs [163].

Macrocyclic antibiotics are divided into four groups: ansamycins, polypeptides, glycopeptides, and aminoglycosides [164]. Ansamycins comprise an aromatic unit linked to an aliphatic chain, and their classification is based on the aromatic moiety. If the aromatic unit is a naphthalene or naphthoquinone, it is denominated naphthalenic ansamycin, while if it is a benzene or benzoquinone, it is a benzenic ansamycin [164]. The most common ansamycins used as CSPs are rifamycins B and SV; the first one is enantioselective for cationic compounds and the second for neutral and anionic [165]. Polypeptides have few aromatic ring units while aminoglycosides do not have this type of structural feature [165]. Only one polypeptide is used as CSP, thiostrepton, whereas aminoglycoside class comprises more CSPs, such as fradiomycin, kanamycin, and streptomycin [166].

Glycopeptides are the most promising class of macrocyclic antibiotic-based CSPs, including avoparcin, ristocetin A, teicoplanin, vancomycin, and derivatized analogues from vancomycin, among others [165]. The chemical structure of glycopeptides consists on a glycosylated cyclic or polycyclic peptide. The central framework is a heptapeptide, in which five of the seven amino acid residues are common to all glycopeptides [164]. Glycopeptides have some flexibility due to the possibility of rotation of sugar groups [165].

The structure of macrocyclic antibiotics allows a variety of interactions with the analytes, such as hydrophobic, π-π, dipole-dipole, hydrogen-bond, electrostatic, ionic, and Van der Waals interactions [166], being either attractive or repulsive. It is possible for the formation of inclusion complexes to occur [167]. The high number of stereogenic centers in their structures also contributes for its high capacity of chiral recognition [55]. Nevertheless, the chiral recognition mechanism of this type of CSPs it is not currently quite understood [43].

The chromatographic support of this type of CSPs is, predominantly, silica gel [164]. Macrocyclic antibiotic-based CSPs are able to operate in all chromatographic elution modes [88]. Besides that, the macrocyclic antibiotic-based CSPs provide a complementary enantioselective profile [167].

Over the years, the developments carried out to obtain diverse macrocyclic antibiotic-based CSPs as well as their applications have been compiled [43,45,85,87,88,164,165,167,168]. Additionally, some authors focused their studies on the mechanism of chiral recognition [169,170].

The most recent reports related to new macrocyclic antibiotic-based CSPs as well as their chromatographic parameters are presented in Table S4 (supplementary material). The developments did not comprise the introduction of new antibiotics as chiral selectors but rather the use of new chromatographic supports, specifically the use of silica particles with sub-2-μm size. In fact, they are mainly based on the preparation of new teicoplanin and vancomycin-based CSPs by reducing the size of the packaging material [171–173]. The main objective was the improvement on chromatographic performance by reduction on analysis time and enhance of resolution and enantioselectivity.

Min et al. [171] described the preparation of a teicoplanin-based CSP bonded to sub-2 μm superficially porous particles (CSP101). The main focus was to avoid aggregation and to uniformize the size distribution, enhancing the surface area. The high resolution and enantioselectivity obtained in a short time of analysis were highlighted. The maximum resolution value was 5.60 for methionine, and the separation factor was 9.40 for norvaline.

Ismail et al. [172] developed a teicoplanin-based CSP with a sub-2 μm chromatographic support; however, in this case, they used totally porous silica particles (CSP102). They pointed out the flexibility of the CSP to operate on different elution modes. The selectivity, efficiency, and the short analysis time on ultra-high-performance liquid chromatographic (UHPLC) were also emphasized [172]. Its performance was promising achieving a resolution value of 10.7 for alanine, and a separation factor of 3.45 for mandelic acid.

Vancomycin was bonded to sub 2-μm diol hydride-based silica particles by Rocchi et al. [173]. Four new CSPs were developed (CSP103–106) with the same main objective: reduction of analysis time. It was inferred that this technique could be applied to other chiral selectors due to the promising results [173]. The maximum resolution value was 3.36 and the separation factor was 2.69 for haloxyfop.

Despite the reduction of particle size of the support, new materials were introduced as chromatographic support. Recently, Xu et al. [174] described the preparation of a vancomycin-based CSP through the combination of monoliths and polymeric cross-linking (CSP107). The new CSP possessed a good mechanical stability, permeability, and enantioselectivity [174]. The influence of some chromatographic conditions was also investigated. The performance of the CSP was satisfactory, with a resolution value of 1.47 for salbutamol, and a separation factor of 1.23 for carteolol. Hellinghausen et al. [175] prepared the CSP108 through the prior synthesis of vancomycin, by Edman degradation, and further binding it to superficially porous particles through a primary amine group of vancomycin, which resulted from the removal of an *N*-terminus leucine residue. The CSP108 presented promising results with a good resolution and separation factor values for 2-amino-2-phenylbutyric acid (Rs = 2.70 and α = 1.57).

A vancomycin-based CSP was recently prepared (CSP109) by a photochemistry-based method [176]. Additionally, the influence of flow rate, elution mode, buffer, and the mass of analyte were also investigated. The addition of 2-propanol, buffer and an increase on analyte mass improved its enantioresolution performance, since π-π interactions were superior. The chromatographic performance was good, with a maximum resolution of 3.08 and a separation factor of 4.23.

It is also important to highlight that the complementary behavior of the different macrocyclic antibiotic-based CSPs continues to be a subject of great relevance. In fact, several recent studies can be found in literature [166,167,177]. Most of them compared the enantioresolution performance of teicoplanin and teicoplanin aglycone CSPs [166,167] or of vancomycin and teicoplanin CSPs [177].

#### *2.5. Brush-Type or Pirkle-Type CSPs*

Brush-type or Pirkle-type CSPs were introduced in 1979, when Pirkle and House described the development and application of a chiral fluoro alcoholic CSP able to enantioseparate diverse classes of analytes [178].

Neutral synthetic chiral low-molecular mass molecules are the base of this type of CSPs [43]. These molecules should promote donor-acceptor interactions as a hydrogen-bond, π-π, or dipole-dipole, in addition to attractive and/or repulsive Van der Waals interactions [45]. As they comprise small molecules as chiral selectors, the mechanism of chiral recognition is, frequently, based on the "three-points" model, which refers that the establishment of at least three interactions between one of the enantiomers to be resolved and the CSP are essential for chiral recognition [179].

The chiral selectors are usually covalently linked to a silica support, which can have monosubstituted or trisubstituted silane groups, through a spacer [55]. Over several years, Pirkle's group has developed successive generations of CSPs [180], based on the principle of reciprocity [181] and on chromatographic [182,183] and spectroscopic [184,185] methods to understand the chiral recognition mechanisms. Among them, Whelk-O1 CSP, created by a rational approach, is the most applied and versatile CSP in both academic and industrial fields [186].

Initially, the preferred elution mode was the normal phase since it provides a favorable environment for the interactions needed to enantioseparate the analytes [88]. Nevertheless, this type of CSPs can also be used in polar organic and reversed-phase elution modes [187–189].

The advantages inherent to this type of CSPs are the compatibility with a wide range of solvents used as mobile phase, the stability to temperature and pressure, the considerable loading capacity and the possibility to be easily scaled up to preparative chromatography [190]. Another key advantage is the possibility of switching the configuration of the chiral selector and to use the inverted configuration column approach [43]. Its kinetic performance is reasonable and the fact that the structure of the chiral selector is relatively "simple" allows an easier knowledge of the chiral recognition mechanisms as well as a consecutive optimization of the selector [191]. Pirkle-type CSPs are characterized by their diversity and versatility since it is possible to use a variety of different small molecules as chiral selectors as well as introduce different substituents that can improve enantioselectivity. In addition, they can be highly specific for certain types of chiral compounds [120].

Pirkle-type CSPs have evolved over the years. Certain types of CSPs have more reported progresses, mainly due to the possibility of use of a wide variety of small molecules as chiral selectors. Several Pirkle-type CSPs can be found in literature comprising chiral selectors closely related to the original Pirkle's group CSPs and others structurally different [43,45,85,87,88,180,186,190,191]. Recently, a literature survey made by our group covering the report on Pirkle-type CSPs developed during the last 17 years was published [190]. We described 226 new CSPs, including a wide diversity of small molecules as chiral selectors, including amine, amino alcohol and amino acids derivatives, peptides, drugs, selectors based on natural products, and xanthone derivatives, among others [190].

The recent developments of this type of CSPs also include the use of new chromatographic supports, such as monolith supports, core-shell particles, or particles with a reduced size (sub 2-μm). The reduction of particle size enables the adaptation of Pirkle-type CSPs to UHPLC, the first ones to be converted, which are associated with the inherent advantages such as the reduction of analysis time and quantity of the solvent, improved efficiency, and enantioresolution [192,193]. Regarding the use of core-shell particles, it was found that the packaging with this type of material or, alternatively, with fully porous particles led to differences on chromatographic performance. The diffusion on core-shell particles is lower than in fully porous particles, which is especially beneficial for large analytes, since it prevents a decrease on efficiency due to an inefficient mass transfer. The distribution of particle size of core-shell particles is nearly unimodal, which increases efficiency on enantioseparation of small analytes [194,195].

Different synthetic methodologies to obtain the chiral selectors as well as for its immobilization on the chromatographic support were also introduced. The synthesis of biselectors was another approach [190]. The most recent Pirkle-type CSPs and the chromatographic parameters obtained after evaluation of their enantioresolution performance are presented on Table S5 (supplementary material).

Qiao et al. [196] developed a CSP based on *N*-ferrocenyl benzoyl-(1*S*,2*R*)-1,2-diphenyl ethanol as a chiral selector (CSP110) (Figure 4). The conjugation of a cyclopentadienyl carbon ring with an aromatic ring demonstrated to improve enantioselectivity. The chiral recognition mechanisms were also explored, revealing that hydrophobic, hydrogen-bond, π-π, and dipole-dipole interactions between the chiral ferrocene CSP and acidic and basic groups of the analytes were crucial. The performance of the CSP was promising achieving a maximum resolution value of 4.13 and a separation factor of 2.43 for 3-nitrophenol.

**Figure 4.** Chemical structures of Pirkle-type CSP110–120.

Çakmak et al. [197] synthetized an aromatic amine derivative of (*R*)-2-amino-1-butanol for the application as chiral selector of a new CSP (CSP111) (Figure 4). In the same study, they used docking, molecular dynamics simulation, and quantum mechanical computation methods to characterize the mechanisms of chiral recognition. The performance of the new CSP was good with a high resolution value of 3.85 for 2-phenylpropionic acid, and a separation factor of 2.75 for mandelic acid.

Four new pseudopeptide-based CSPs were developed (CSP112–115) (Figure 4) inspired by the possibility of enantiorecognition ability of an organocatalyst [198]. It was found that the enantioselectivity of the CSPs was dependent of the degree of derivatization of diproline portion and of the length of polymeric chain. The chromatographic results were promising, achieving, for example, separation factor and resolution values of 9.80 and 2.89, respectively, for 1-phenylethan-1-amine and *N*-(1-(naphthalen-2-yl) ethyl)-3,5-dinitrobenzamide.

Additionally, in another recent work, derivatives of amino acids and amino alcohols as CSPs were prepared by Yu et al. [199], based on C3-symmetric CSPs (CSP116–119) (Figure 4). It was found that a phenyl group linked to amide was crucial for chiral recognition and, despite the chiral selectors did not possess a π-acidic or π-basic group, their performance was promising. For example, a separation factor value of 2.58 was achieved for 2-phenyl-2-pentanol.

Wang et al. [200] synthetized the (*R*)-6-acrylic-binaphtol as chiral selector through addition of the acrylic group to the (*R*)-binaphtol and developed the CSP120 (Figure 4). The mechanisms of chiral recognition, the effect of the temperature and mobile phase composition were also discussed. It was found that the flexibility of the CSP and the π-π stacking event allowed the retention of the analytes without compromise the enantioseparation [200]. Regarding the chromatographic results, a separation factor value of 1.12 was achieved for 3,5-dinitro-*N*-(1-phenylethyl) benzamide.

Along with the continuous developments of this type of CSPs, it is important to emphasize that a broad range of recent applications have also been reported [201–205].

#### *2.6. Ion-Exchange-Type CSPs*

Ion-exchange-type CSPs were introduced by Salvadori et al. in 1985, who described the application of cinchona alkaloids as CSPs [206]. Nevertheless, Lindner group developed the majority of this type of CSPs [207]. Ion-exchanger selectors can be subdivided into three groups: anionic, cationic, or zwitterionic [208].

The most common anion-exchangers as chiral selectors are cinchona alkaloids [45] and terguride [43]. Anion-exchanger selectors are appropriate for enantioseparation of acidic compounds; their enantioselectivity are attributed to the five stereogenic centers of the basic nucleus common to quinine and quinidine [209]. Cation-exchanger selectors are useful for enantioseparation of basic analytes, which are structurally based on chiral sulfonic or carboxylic acid compounds as selectors [45]. Zwitterionic selectors were introduced, more recently, by Lindner et al. [210] by merging key cationand anion-exchange moieties in one single chiral selector [45]. Those CSPs can been applied for the enantioresolution of acid, basic, and amphoteric compounds [209]. Zwitterionic CSPs have overcome the main disadvantage of anion and cation-exchanger CSPs, since these two groups only separate enantiomers with opposite charge [210].

The chiral mechanism of recognition is mainly based on ionic interactions between the charged analytes and the opposite charged groups of the CSPs [208]. Hydrogen bonds as well as π-π interactions are also important for complex formation [45]. The ion-pairing of solvent controls the adsorption and retention of the analytes [43]. Polar-organic and reversed-phase elution modes are the preferential elution modes for this type of CSPs [43]. The retention and enantioselectivity are affected by the pH and the nature and concentration of acid or base added to the mobile phase [43].

The progresses resorting to this type of CSPs have been reviewed over the years [43,45,85,88,207, 209,211–213]. Recently, Ilisz et al. [213] compiled the most recent developments concerning to anionic and zwitterionic-exchange-based CSPs, which are related to the application of different techniques of preparation of chromatographic support to attempt the optimization of chromatographic parameters. The most recent ion-exchanger-type CSPs and their chromatographic parameters are presented on Table S6 (supplementary material). The majority of the recent developments focused on quinine and quinidine derivatives as chiral selectors [214–217], the key structural moiety representative of anion-exchanger selectors.

Todoroki et al. [214] developed a new technique to prepare new ion-exchange CSPs, specifically cinchona alkaloid-based quinine and quinidine-type fluorous-tagged-CSPs (CSP121–125) (Figure 5). The main objective was to improve the enantioseparation properties enabling a sensitive, selective, robust, and reproducible analysis methodology. The versatility of the new CSPs was another advantage, as it was capable to enantioseparate bulky, aromatic compounds, in addition to amino acids, such as threonine with a resolution value of 11.8, and asparaginine, with a separation factor of 4.56.

**Figure 5.** Chemical structures of Ion-exchange-type CSP132–140.

The complementarity profile between anion-exchange-type CSPs was the focus of a recent study reported by Lämmerhofer et al. [218], who prepared several cinchona carbamate selectors with distinct carbamate residues to obtain CSP126–131 (Figure 5). Different structural moieties were introduced to enhance the possibility of complementary; for example, the introduction of bulky groups to create steric hindrance or aromatic rings to provide π-π interactions. The complementary accomplished with the new CSPs allowed the expansion of the enantioselectivity range. The performance of the CSPs was promising, achieving, for example, a separation factor of 17.0 for leucine.

De Martino et al. [216] reported the synthesis of an anion exchange hybrid selector, the 3,5-dinitrobenzoyl-9-amino-9-deoxy-9-epiquinine, to develop CSP132 (Figure 5). The strategy applied was the association of typical moieties of Pirkle-type selectors with key moieties of anionic-exchange-based selectors, enlarging the possibility of multiple interactions with the analytes. The performance of this hybrid CSP was promising, with separation factor and resolution values of 2.06 and 11.0, respectively, for diazepam *N*-oxide.

A new immobilization technique based on click chemistry, was described by Lämmerhofer et al. [219], who prepared a cross-linked *tert*-butylcarbamoyl quinine-based CSP (CSP133). The technique allowed achieving a CSP with reduced resistance to mass transfer and retention times, as well as an improved stability. During the optimization of the procedure, some features were discussed, such as the amount of polysiloxane, chiral selector, radical initiator, and reaction solvent, as well as reaction time and size of the chromatographic support particles [219]. The performance of the CSP was promising; for example, with separation factor and resolution values of 1.54 and 5.20, respectively, for *N*-[(9*H*-fluoren-9-ylmethoxy) carbonyl]-phenylalanine.

The same group also resorted to click chemistry to prepare other CSPs based on *tert*-butylcarbamoyl quinine (CSP134–137) (Figure 5) [215]. The optimization of the selector's structure allowed the avoidance of non-specific interactions that could reduce chiral recognition [215]. The introduction of a sulfonic group afforded a reduction on the retention times and an improvement, in some cases, of separation factors since its negative charge provided electrostatic interactions, promoting an effect similar to the counterion effect [215]. The performance of the CSP was satisfactory achieving a maximum resolution value of 6.20 for *N*-(9-fluorenylmethoxycarbonyl)-phenylalanine and a separation factor of 1.66 for *N*-acetyl-phenylalanine.

The application of core-shell particles was another strategy. A new CSP based on *tert*-butylcarbamoylquinine selector (CSP138) (Figure 5) was described by the same group, to promote the enantioseparation of several proteinogenic amino acids [217]. Core-shell particles were introduced in order to improve the analysis time, which was a promising methodology for the bioanalytical area, since it could be combined with sensitive fluorescence detection or highly sensitive and selective mass spectrometric detection. The column presented a reasonable performance with good enantioselectivity and resolution. For example, α and Rs values of 1.55 and 4.08, respectively, were achieved for threonine [217].

Armstrong et al. [220] also resorted to core-shell particles to develop two new quinine-based CSPs (CSP139–140) (Figure 5) for ultrafast liquid and supercritical fluid chromatography. The new CSPs allowed fast analysis with high enantioselectivity and efficiency for the tested analytes. The performance of the new CSPs was promising, affording, for example, a maximum resolution of 25.5 and a separation factor of 14.5 for *N*-(3,5-dinitrobenzoyl)-leucine.

It is important to highlight that recent applications of this type of CSPs were diversified, with the focus, mainly, on the application of anion and zwitterionic-type CSPs with amino acids [221–225].

#### *2.7. Crown-Ether-Based CSPs*

Crown-ethers as CSPs were firstly described by Sousa et al. [226] for the enantioseparation of primary amine salts. Crown-ethers consist on macrocyclic polyethers, with a cavity of a specific size, being able to form complexes with analytes [227]. CSPs based on crown-ethers can be divided into two major groups: the crown- ethers comprising a 1,1 -binaphthyl group and those containing two tartaric acid groups [228]. The enantiomers of α-amino acids and primary amines may be separated by the first type of crown-ethers CSPs [229]; the second group can be applied for enantioseparation of primary and secondary amino compounds and non-amino compounds [229].

The mechanisms of chiral recognition are typically driven by triple hydrogen bonds established by an ionized ammonium group of the analytes and three oxygen of the CSP, leading to the formation of an inclusion complex [43]. The electron-donor oxygen particles are distributed on inside of the cavity of the crown-ether [88]. Steric hindrance from the substituents of the analyte ions and the functional groups of the crown-ethers can influence the enantioseparation [43]. Additional interactions are essential to complement the formation of the complex, including π-π, hydrogen-bond, and dipole-dipole interactions [230]. Mobile phases should be strong acidic aqueous solutions to achieve the total protonation of the amino group of the analytes [43]. Crown-ether-based CSPs can be obtained by a

coating process or by immobilization [231]. To avoid the leaching of the CSP from the column and to allow the analysis of hydrophobic compounds the use of covalently bonded CSPs it can be preferable than the coated [232].

The developments of crown-ether-based CSPs have been revised over the years [43,45,85,87,88,228, 229]. Hyun et al. [228] reviewed the most recent developments, related to both classes of crown-ethers as CSPs, highlighting techniques for the preparation of the chromatographic support or the protection of unreacted residues.

Crown-ether-based CSPs have some restrictions related to the target analytes; however, their preparation and the chromatographic conditions can be modifier to improve the chromatographic results. The majority of the recent developments encompassed the use of different strategies for immobilization of the chiral selectors to the chromatographic support (click chemistry) and the introduction of different functional groups on previous described selectors. The most recent crown-ether-based CSPs as well as their chromatographic performance presented on Table S7 (supplementary material). Some of the recent CSPs are based on calix[4]arene derivatives as chiral selectors [233–235]. The new derivatives are mainly prepared using click chemistry [234,235].

Accordingly, the CSP141 (Figure 6) comprising the aza-15-crown-5-capped methylcalix[4] resorcinarene derivative was developed by Ma et al. [233]. Structurally, the CSP possesses two key recognition sites, enhancing the possibility of interactions and, consequently, enantioseparation of analytes. The robustness of the CSP was highlighted since it could operate on different elution modes with a short analysis time (for example, k1 = 0.08 for *m*-nitrophenol).

Yaghoubnejad et al. [234] prepared a calix[4]arene functionalized with two L-alanine units to develop CSP142 (Figure 6), through covalent binding between the allyl groups at the lower rim of the chiral selector and the chromatographic support, using click chemistry. The CSP142 was able to enantioseparate both π-acidic and π-basic analytes. It was suggested that the used technique could easily be adapted to other derivatives to obtain improved CSPs. The maximum resolution value achieved was 1.43 and the separation factor was 2.00 for mandelic acid.

Click chemistry has also been explored, by Li et al. [235], for the preparation of a click-dibenzo-18-crown-6-ether-based CSP (CSP143) (Figure 6). The effect of pH and concentration of salt in the mobile phase on chromatographic parameters was evaluated [235]. It was found that the retention of strong acids decreased with the increment on salt concentration. Regarding the pH values, the retention of both acidic and basic analytes decreased with its reduction. The retention factors were good with a minimum of 0.10 for uracil.

CSPs comprising carboxyl derivatives of crown ethers as chiral selectors were also prepared. Németh et al. [236] synthetized derivatives of acridino-crown ethers containing a carboxyl group to obtain eleven CSPs (CSP144–154) (Figure 6). The CSPs were developed taking into account some structural features that could favor the interactions crucial for chiral recognition mechanisms. Due to the rigidity of the tricyclic ring system, the enantioselectivity was improved [236]. The obtained performance was reasonable with a maximum resolution value of 1.20 for 1-(1-naphthyl)-ethylamine, and a separation factor of 2.05 for 1-(4-nitrophenyl)-ethylamine hydrogen chloride.

The development of hybrid crown-ether-based CSPs was also reported. Deoxycholic-calix[4]arene hybrid-type selectors were synthetized, by Yaghoubnejad et al. [237], aiming to enhance the interactions of the obtained CSPs (CSP155–156) (Figure 6). The calix[4]arene unit was fundamental for the mechanisms of chiral recognition being responsible for the establishment of hydrophobic and π-π interactions, important for inclusion complexes formation. The presence of an acidic or basic modifier in mobile phase was beneficial for enantioresolution of acidic or basic analytes [237]. Relatively to its performance, a maximum resolution value of 3.93 and a separation factor of 4.30 were obtained for mandelic acid.

**Figure 6.** Chemical structures of crown-ether-based CSP141–156.

#### *2.8. Cyclofructan-Based CSPs*

Cyclofructans are the most recent type of CSPs being introduced, in 2009, by Armstrong et al. [238]. Moreover, they demonstrated that suitable derivatized of cyclofructans presented a superior enantioselectivity in comparison with native cyclofructans [238].

Cyclofructans are cyclic oligosaccharides formed by units of D-fructofuranose β(2→1) linked together [45]. They are also described as a crown-ether nucleus rounded by fructofuranose units, with its number between 6 and 8 [239]. Each unit has four stereogenic centres [88]. In opposition to cyclodextrins, the interior of the nucleus is hydrophilic [45]. The mechanism of chiral recognition is based on the formation of a complex, which is driven by polar interactions, including dipole-dipole and hydrogen-bond interactions [45]. Therefore, the analytes to enantioseparated should not be hydrophobic and may have hydrogen-acceptor and polarizable groups next to stereogenic center [240]. The acidic hydrogen-bond play an important role on chiral recognition, thus, the presence of a polarizable group that causes steric hindrance to the basic portion of the cyclofructan is favorable [240]. The main advantages of this type of CSPs are their high loadability and versatility, as it is able to enantioseparate basic, acidic, and neutral analytes [239]. Moreover, they can be used in different elution modes [239]. Although cyclofructan-based CSPs are recent, some reviews can be found related to its developments and applications [45,55,85,87,88,90]. The elucidation of their chiral recognition mechanisms has been the focus of some studies to clarify the interactions between the CSP and the analytes [45]. The most recent cyclofructan-based CSPs and its chromatographic behavior are presented in Table S8 (supplementary material). The latest developments include the synthesis of new derivatives of cyclofructan as chiral selectors, the preparation of new chromatographic supports, and the application of different immobilization strategies.

In order to evaluate the effect of electron-donating and electron-withdrawing groups on enantioselectivity, Khan et al. [239] synthetized chlorinated aromatic derivatives of cyclofructan 6 and developed CSP157–166 (Figure 7). The presence of a chlorine proved to be beneficial for enantioselectivity, in opposition to nitro group, especially in the ortho position of the aromatic ring, which negatively affected the chiral recognition [239]. A maximum resolution value of 6.90 for 2-2 -binaphthol, and a separation factor of 2.05 for Tröger's base were obtained.

The influence of the degree of substitution, as well as the size of the substituents, was researched by Padivitage et al. [241] by preparing CSP167–171 (Figure 7), with basic derivatives cyclofructan 6 as a selector. It was concluded that bulky groups caused steric hindrance and that a high degree of substitution (up to six substituents) negatively affected the enantioselectivity. Moreover, it was found that charged cyclofructans did not possess a superior ability of enantiorecognition [241]. Relatively to the performance of the CSPs, as an example, separation factor and resolution values of 1.43 and 3.10 were obtained for warfarin.

An alternative technique for preparation of a CSP was presented by Qiu et al. [242] (CSP172) (Figure 7), using click chemistry to immobilize the chiral selector, cyclofructan 6, to a resin. The resin was chosen as chromatographic support due the advantages inherent to this material, such as high adsorption capacity, high mechanical strength, lower cost, and reduced sensitivity to pH. Although the chromatographic results were only reasonable, with a resolution of 1.40 for *trans*-1-amino-2-indanol, and a separation factor of 1.41 for *N*-*p*-tosyl-1,2-diphenylethylenediamine, the stability and reproducibility of the CSP were emphasized [242].

Similarly to other types of CSPs, the applications of cyclofructan-based CSPs are becoming more prominent [243,244].

**Figure 7.** Chemical structures of cyclofructan-based CSP157–172.

#### *2.9. Molecularly-Imprinted CSPs*

A different approach to chiral separation has been applied by using molecularly-imprinted CSPs. The synthesis of artificial selectors that are specific for a selected target (template) [245] is the principle of this type of CSPs. Each molecular imprinted CSP can only be applied for a specific type of analytes, as they are frequently applied on preparative enantioseparation and extraction of the desired compounds [246].

Several reviews have been devoted to molecularly-imprinted CSPs, mainly focusing on their developments and the different fields of application [45,88,246–253]. Although the developments concerning this type of CSPs are becoming more usual, their enantioresolution performance is currently not competitive in comparison to the existing CSPs. The most recent developments comprised the adaptation of this type of CSP to monoliths, nanoparticles, and predominantly to polymers. The introduction of different supports such as alginate microspheres [254] or polymer functionalized with quantum dots [255] were also reported as well as the description of different functional monomers and crosslinking agents [256–259].

Recently, Gutierrez-Climente et al. [245] prepared a new CSP by molecularly-imprinted nanoparticles on silica beads to reduce the tailing effect, commonly observed with this type of CSPs, through the reduction of particle size. The influence of some factors, such as the buffer percentage and concentration, pH, temperature, and column length, on chromatographic performance was evaluated. Regarding the chromatographic results, a maximum resolution value of 1.44 and a separation factor of 1.45 were obtained for citalopram.

In another study, Yang et al. [260] prepared a molecularly-imprinted polymer on porous silica gel microspheres to improve the chromatographic performance of a previous developed CSP and to reduce the analysis time. The new CSP demonstrated a higher affinity than the nonimprinted polymer with the silica gel, and selectivity for the target analyte, oseltamivir, with a retention factor of 13.5.

The optimization of the capacity of enantioseparation of a molecularly-imprinted monolith using a molecular crowding agent was recently reported by Wang et al. [261]. The main aim was to enhance the interactions between the CSP and the target analyte (*S*)-amlodipine. The composition of mobile phase, ionic strength, pH, and content of organic modifier were also taken into account when attempting to improve chromatographic performance [261].

#### *2.10. Other CSPs*

Despite the CSPs already mentioned, there are other types of CSPs, such as ligand-exchange, based on synthetic polymers, among others. The ligand-exchange CSPs do not present significant recent developments. Regarding polymers, several synthetic polymers can be used as selectors of CSPs [262]. Nevertheless, despite the interest in this type of material, synthetic polymer-based CSPs are not yet commercialized.

Regarding synthetic polymers, their classification can be based on the type of polymerization, as addition or condensation polymers and cross-linked gels, which are prepared resorting to molecular imprinted technique [263]. A synthetic and optically active polymer can be used for preparation of CSPs if it possesses a helical conformation, which contributes to the wide range of applications and effective separations [264]. The chiral recognition mechanism is based on hydrogen-bond, π-π interactions, and steric factors [45].

As for the other types of CSPs, the CSPs comprising synthetic polymers were the focus of several reviews [45,55,262–267]. The most recent developments are related with the introduction of monoliths [267], specifically of nanoparticles and hybrid monoliths. Ding et al. [264] also reported the use of smart polymers.

Recently, Maeda et al. [268] synthetized derivatives of optically active poly(diphenylacetylene) with chiral and achiral substituents. The helical structure of the polymer demonstrated to influence the enantioselectivity. The performance of the CSP was promising, affording excellent enantioselectivity and resolution, with α = 19.3 and Rs = 15.7 for ruthenium (III) acetylacetonate.

In another recent study, optically active π-conjugated polymers formed by alternated units of thieno[3–b]cthiophene and glucose-linked biphenyl were prepared, with its backbone conformation important for the enantioseparation of the obtained CSP [269]. Its performance was satisfactory with a maximum separation factor of 1.56 for cobalt (II) acetylacetonate.

A stable, porous, and crystalline organic polymer was introduced by Zhang et al. [270] highlighting its enhanced stability and resolution. The enantioresolution performance of the obtained CSP was reasonable with a separation factor value of 1.21 for *trans*-metoconazole, and a resolution of 2.56 value for *p-*nitrochlorobenzene.

Additionally, it is important to highlight the introduction of chip-based columns, motivated by the same factors than the transition for UHPLC, i.e., reduction of analysis time and improve the efficiency; the adaptation of LC remains challenging due to technical aspects [271–273]. An advantage focused for this type of columns was the simplicity of the process of its production as referred in the first report by Manz et al. [274]. More recently, they have been applied to extraction methodologies [275,276]. Despite the incorporation of micro and nanoparticles, this remains a challenging issue [273]; the introduction of monolith chip-based columns has already been reported [277].

#### **3. Conclusions**

The development of new CSPs for LC is a continuous and challenger issue covering various types of CSPs. This review gathered the most recent developments associated to different types of CSPs providing an overview of the advances that are occurring on this research area.

The most recent strategies, summarized in Figure 1, comprised the introduction of new chiral selectors or new chromatographic supports, and the application of different immobilization or coating methodologies for preparation of the CSPs. Regarding the chiral selectors, novel structures or analogues related to previously reported selectors were described as well as the use of hybrid selectors. The focus in chromatographic supports with lower particle size, the innovation related to the application of new materials such as monoliths and core-shell particles, as well as the use of hybrid supports, were also reported. In addition, several non-conventional approaches for immobilization or coating the chiral selectors to the chromatographic support were included, with particular emphasis to click chemistry as well as new encapsulation techniques or thermal immobilization without the use of chemical reagents, among others. Regardless of the type of CSPs, the main objectives of the development strategies were similar, concerning the improvement of the enantioresolution performance of the CSPs, as well as the increase of versatility and range of applications. Additionally, the transition to UHPLC and the possibility of the new CSPs to be used in all elution modes or using mobile phases compatible with mass spectrometric detection has also been underscored.

Even though several innovative strategies have been applied and several new CSPs were recently developed, they do not yet go beyond the exploratory stage. Nevertheless, the various strategies presented constitute an important trigger aiming to achieve new CSPs, as commercially viable products with high versatility and broad range of analytical and preparative applications or, on the contrary, as exceptionally efficient for the enantioresolution of specific target analytes.

In our opinion, the development of efficient chromatographic tools for LC enantioresolution is a subject that should continue to receive special attention, since it has constructive repercussions in several other research areas, such as biomedical, toxicology and forensic sciences, environment, food and fragrances, industry, among others. Moreover, the goal of developing a universal CSP remains a dream to be reached by those working in this research area.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1420-3049/24/5/ 865/s1, Table S1: Recent developments of polysaccharide-based CSPs, Table S2: Recent developments of protein-based CSPs, Table S3: Recent developments of cyclodextrin-based CSPs, Table S4: Recent developments of macrocyclic-based CSPs, Table S5: Recent developments of donor-acceptor or Pirkle-type CSPs, Table S6: Recent developments of ion-exchange-based CSPs, Table S7: Recent developments of crown-ether-based CSPs, Table S8: Recent developments of cyclofructan-based CSPs.

**Author Contributions:** J.T. collected the primary data and contributed in writing of the manuscript. C.F., M.E.T., and M.M.M.P. supervised the development of the manuscript, and assisted in data interpretation, manuscript evaluation, and editing.

**Funding:** This work was supported by the Strategic Funding UID/Multi/04423/2013 through national funds provided by FCT—Foundation for Science and Technology and European Regional Development Fund (ERDF), in the framework of the programme PT2020, the project PTDC/MAR-BIO/4694/2014 (reference POCI-01-0145-FEDER-016790; Project 3599—Promover a Produção Científica e Desenvolvimento Tecnológico e a Constituição de Redes Temáticas (3599-PPCDT)) as well as by Project No. POCI-01-0145-FEDER-028736, co-financed by COMPETE 2020, under the PORTUGAL 2020 Partnership Agreement, through the European Regional Development Fund (ERDF), and CHIRALXANT-CESPU-2018.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
