Enantiopurification by Co-Crystallization within Cyclodextrin Metal–Organic Framework

Abstract

Tröger’s base analogs (TBAs) and their derivatives are versatile, Λ-shaped, tetracyclic chiral building blocks utilized in numerous fields of research. Although various methods for the enantiopurification of TBAs have been demonstrated in the literature, none has achieved it with the use of metal–organic frameworks (MOFs). This investigation introduces a convenient and scalable method to obtain enantiopure TBAs with the formation and digestion of a chiral MOF composed of fully recoverable and non-hazardous starting materials, namely, cyclodextrin-based metal–organic framework (CD-MOF).

1. Introduction

The enantiopurification of chiral compounds plays a fundamental role in modern chemistry and pharmacology on both laboratory and industrial scales. The rising demand for enantiopurity and accurate screening of chiral compounds has promoted notable advances in liquid and supercritical fluid chromatographic technologies. In recent decades, chiral resolution techniques have become more prevalent in various fields, including but not limited to the processing of pharmaceuticals, agrochemicals, and natural products. As a result, the efficient design and preparation of chiral stationary phases (CSPs) continue to gain more importance and grow [1,2,3,4,5,6]. Although CSPs with diverse compositions have been invented for the direct enantioseparation of various chiral mixtures, their use remains constrained by the production costs and scalability limits. In addition, recent trends toward the mitigation of environmental damage caused by chemical industries have encouraged the development of greener alternatives to common CSPs, such as porous materials. Porous materials, e.g., zeolites, have been utilized in the industrial-scale processing of petrochemicals for decades because of their availability and impressive durability. Zeolites are particularly efficient in catalytic transformations of hydrocarbons, including disproportionation, (trans)alkylation, isomerization, and arylation [7]. Despite the impressive efficacy of zeolites in the mentioned processes, inorganic zeolites lack the properties needed for chiral separations. Therefore, enantiopure building blocks are incorporated in microporous materials resulting in hybrid organic–inorganic zeolite analogs (zeotypes) capable of enantioseparation of chiral alcohols [8]. The advent of porous coordination polymers has also contributed to the growing importance of porous materials, placing them above and beyond many others, including zeolites and activated carbon [9,10].

Linking organic ligands with metallic cations results in the formation of a rather newer class of porous materials with exceptional and tunable properties. These materials, more often defined as metal−organic frameworks (MOFs) [11], are found to be efficient in the storage of gases and information [12,13], carbon capture [14,15,16,17], catalysis [18], and separation sciences [19]. Chiral MOFs are composed of chiral linkers and serve as separation media for enantiomeric resolution or catalyzing reactions asymmetrically [20]. Chiral MOFs have been employed for a wide variety of applications, especially in photonics [21] and discriminating sorbents [22]. Lately, a green and renewable MOF composed of γ-cyclodextrin (γ-CD) and alkali metal salts, namely CD-MOF [23,24,25], was found efficient in separating components of challenging mixtures in an energy-efficient and bio-compatible manner [20,26]. For instance, CD-MOF can separate ethylbenzene from styrene, haloaromatics, terpinenes, pinenes, and chiral compounds from one another [20]. Hartlieb and coauthors demonstrated the efficiency of CD-MOFs as separation media by utilizing them as CSPs in HPLC, and they succeeded in the challenging separation of xylene regioisomers from one another [20]. Their pioneering investigation indicated that CD-MOFs retain saturated species to a greater extent than unsaturated ones. They also correlated structural factors with retention time, e.g., the double bond location in the case of pinene and terpinene isomers. They specifically indicated that exocyclic double bonds increase retention more than endocyclic ones. In addition, they highlighted analytes carrying halogen substituents can interact with CD-MOFs through non-covalent bonding interactions (NCIs) [27] that contribute to the enhanced separation efficiency and supported their observations with ABD molecular modeling simulations. They also noted that CD-MOFs, as homochiral frameworks, can discriminate the enantiomers of chiral analytes and separate them from one another. These findings resulted in the introduction of an inexpensive and easy-to-prepare stationary phase [20] for HPLC applications with the potential for replacing older CSPs such as CD-bonded silica particles [28]. Their innovative approach has inspired us to utilize CD-MOFs as green and scalable separation media in this work to separate the enantiomers of a dicarboxylic acid-carrying ethano-bridged TBA by co-crystallization [29]. In this investigation, we test a co-crystallization strategy for exploiting CD-MOF chirality for the enantiopurification of an ethano-bridged TBA.

TBAs are a class of exceptionally chiral diamines whose structural rigidity prevents nitrogen inversion. TBAs have been employed in the design of phosphorescent organic light-emitting diodes (OLEDs) [30,31], molecular switches [32], supramolecular scaffolding [33], liquid crystal dopants [34], artificial muscles [35], metalloporphyrins [36], photosensitizers [37], fluorosolvatochromic, fluorescent and NMR active probes [38,39,40,41,42], macrocycles [43,44], perovskite solar cells [45], chiral catenanes [46], ultra-microporous, and gas separating membranes [47,48,49].

Methano-bridged TBAs, e.g., TBA 1 shown in Scheme 1, stereogenic-tertiary amine groups cannot undergo the inversion of configuration on their own and hence are normally stable. However, they racemize upon protonation and through the formation of iminium intermediates [34]. TBA racemization can be prevented by replacing the methanol-bridge with an ethano-bridge (Scheme 1), which makes their stereogenic-tertiary amines acid-resistant [34]. Therefore, we have used ethano-bridged TBA 3 in this investigation, which does not racemize and carries two carboxylic acid groups. These carboxylic acid groups make TBA 3 water soluble and provide interactive sites for establishing hydrogen bond interactions with the chiral alcohol groups of cyclodextrin in both solution and solid state, i.e., CD-MOF.

2. Materials and Methods

2.1. Materials

Sigma and Fisher Scientific supplied chromatography-grade solvents, including acetone (CH₃COCH₃), chloroform (CHCl₃), dichloromethane (CH₂Cl₂), dimethylformamide (DMF), ethyl acetate (EtOAc), and methanol (MeOH), ethanol (EtOH) and hexanes. Reagent-grade starting materials supplied by Ambeed, Tokyo Chemical Industry (TCI), and Combi-Blocks were used without further purification. Rigaku Cu-Synergy X-ray diffractometers were employed to collect crystallographic data. These data were processed and refined with Olex V2-1.3 software before depositing the solved structure in the Cambridge Crystallographic Data Centre (CCDC). Chiral HPLC chromatograms were recorded by multi-channel optical detection at 235 and 245 nm using Agilent’s 1260 Infinity HPLC instrument equipped with a Phenomenex chiral analytical column (Chirex 3126 (D)-penicillamine, 150 × 4.6 mm, 5 µm). The applied mobile phases included HPLC-Plus-grade H₂O, and MeCN purchased from Sigma Aldrich. Copper (II) sulfate (2 mM) was added to the aqueous mobile phase (10% v/v) used for the analytical chiral separations only. NMR spectra were recorded at 300 K using Bruker Avance III 600 MHz, Bruker Neo 600 MHz, and Bruker Avance III 500 MHz instruments running Topspin (version 4.0.8) for the analysis and plotting of the acquired spectra. Deuterated chloroform (CDCl₃), dimethyl sulfoxide (CD₃SOCD₃), methanol (CD₃OD), and deuterium oxide (D₂O) were purchased from Cambridge Isotope Laboratories (CIL) and used for NMR spectroscopic analysis.

2.2. Synthesis of Tröger’s Base Analog (±)-1

Ethyl 4-aminobenzoate (12.5 g) was crushed with a mortar and pestle and gradually added to trifluoroacetic acid (TFA, 200 mL) while mixing them with a magnetic stir bar until fully homogenous. Paraformaldehyde (PFA, 4.0 g, excess) was added before sealing the flask and allowing the mixture to stir at rt for 10 days. The reaction mixture was concentrated to about one-fourth of its initial volume using a vacuum pump (Edwards RV12, Hampton, NH, USA) equipped with two traps filled with dry ice/acetone and sodium hydroxide pellets. This concentrate was then poured on 200 g of crushed ice and basified to a pH of 8 with the addition of saturated sodium carbonate solution before extraction with EtOAc (2 × 50 mL). These organic layers were then combined, rinsed with saturated sodium hydrogen carbonate solution (3 × 30 mL), dried over magnesium sulfate, and paper-filtered. The filtrate was then dried using a rotavap, and the remaining crude was subjected to normal-phase flash chromatography, giving 9.5 g (77% yield) of TBA (±)-1 as a white solid. R_f 0.35 (silica gel; EtOAc–hexanes–DCM, 2:3:5 v/v). ¹H NMR (600 MHz, CDCl₃): δ 7.85 (dd, J = 8.8, 2.8 Hz, 2H), 7.64 (d, J = 2.8 Hz, 2H), 7.19 (d, J = 8.8 Hz, 2H), 4.76 (d, J = 16.5 Hz, 2H), 4.35 (s, 2H), 4.27 (d, J = 16.5 Hz, 2H), 4.30 (q, J = 6.1 Hz, 4H), 1.33 (t, J = 6.5 Hz, 6H).

2.3. Synthesis of Tröger’s Base Analog (±)-2

Tröger’s base analog (±)-1 (2.2 g), 1,2-dibromoethane (12.0 mL), and lithium carbonate (10.0 g) were mixed at 130 °C in DMF (40.0 mL) for 48 h. The mixture was then reduced to about one-fifth of its initial volume using a rotavap with a hot bath at 70 °C. The remaining residue was divided between water and EtOAc (100 mL, 1:1). The organic layer was separated, rinsed with saturated sodium carbonate solution (3 × 30 mL), dried over magnesium sulfate, and paper-filtered. The filtrate was then dried by a rotavap, and the remaining residue was subjected to normal-phase flash chromatography, giving 780 mg of the purified product TBA (±)-2 (35% yield). R_f 0.25 (silica gel; EtOAc–hexane, 30% v/v). ¹H NMR (600 MHz, CDCl₃): δ 7.70 (dd, J = 8.8, 2.8 Hz, 2H), 7.60 (d, J = 2.8 Hz, 2H), 7.11 (d, J = 8.8 Hz, 2H), 4.63 (d, J = 17.5 Hz, 2H), 4.53 (d, J = 17.5 Hz, 2H), 4.26 (q, J = 6.9 Hz, 4H), 3.55–3.67 (m, 4H), 1.31 (t, J = 7.1 Hz, 6H). ¹³C NMR (150 MHz, CDCl₃): 166.2, 154.9, 136.1, 130.6, 128.7, 127.9, 126.8, 60.8, 58.8, 54.3, 14.4.

2.4. Synthesis of Tröger’s Base Analog (±)-3

Tröger’s base analog (±)-2 (500 mg), sodium hydroxide (200 mg), and DI water (20 mL) were mixed and refluxed under nitrogen gas for 48 h. The mixture was allowed to cool to rt, filtered (glass microfiber), acidified with HCl solution, and concentrated with a rotavap before being injected into a C18 reversed-phase chromatography column. Purified hydrolyzed TBA (±)-3 (390 mg, 91% yield) eluted at 50% H₂O–MeOH. Collected and combined column fractions were reduced to one-tenth of their initial volume, chilled in a fridge, and then filtered. The obtained solid was rinsed with ice-cold water and further dried under a high vacuum overnight. ¹H NMR (600 MHz, CD₃OD): δ 7.68 (dd, J = 8.2, 2.3 Hz, 2H), 7.61 (d, J = 2.3 Hz, 2H), 7.16 (d, J = 8.2 Hz, 2H), 4.73 (d, J = 17.3 Hz, 2H), 4.53 (d, J = 17.3 Hz, 2H), 3.72 (m, 2H), 3.56 (m, 2H). ¹³C NMR (150 MHz, CD₃OD): 169.4, 156.3, 137.7, 131.8, 129.7, 128.8, 128.1, 59.5, 55.1. MS (ESI +): m/z [M + H]⁺ calcd for [C₁₈H₁₇N₂O₄]⁺: 325.34; found: 325.3.

2.5. Preparation of Potassium Salt of Tröger’s Base Analog (±)-3

Tröger’s base analog (±)-3 (973 mg, 3 mmol) and KOH (366 mg, 6 mmol) were sonicated in HPLC-grade MeOH until fully dissolved. This solution was reduced by rotavap and then dried under high vacuum at room temperature, resulting in 1242 mg of potassium salt of (±)-3.

2.6. Enantiopurification of Tröger’s Base Analog (–)-(S,S)-3

Stock solution of γ-CD (2075 mg, 1.6 mmol) and KOH (718 mg, 12.8 mmol) in HPLC-grade H₂O (32 mL) was prepared and gradually added to the potassium salt of TBA (±)-3 and sonicated to obtain a saturated solution of (±)-3. The resulting solution was passed through a 0.2 µm microfilter and poured into a nitrogen-flushed beaker. EtOH (40 mL) was allowed to diffuse slowly into the solution at rt over 10 days. Pale yellow cubic crystals of enantioenriched 3⸦CD-MOF were isolated, placed on a sintered filter, and rinsed with a 1:3 blend of H₂O–EtOH (3 × 5 mL), and then pure EtOH (3 × 5 mL). The first crop of enantioenriched 3⸦CD-MOF crystals were digested in H₂O, with sonification at room temperature. The pH of this solution was adjusted at 5 with the addition of HCl before isolating enantioenriched TBA 3 from the solution with a C18 reversed-phase chromatography column using H₂O–MeOH gradient. The isolated fractions containing enantioenriched 3 were combined, rotavaped, and dried under a high vacuum to afford the first crop of enantioenriched TBA 3. Enantioenriched 3 was dissolved in a minimal volume of fresh γ-CD stock solution and was subjected to the diffusion of EtOH in order to obtain the second crop of enantioenriched 3⸦CD-MOF crystals and repeating the steps over and over again until obtaining the third crop of crystals, i.e., (–)-(S,S)-3⸦CD-MOF, which was then collected, rinsed, digested in H₂O and chromatographed to obtain enantiopure (–)-(S,S)-3 as mentioned.

2.7. Enantiopurification of Tröger’s Base Analog (+)-(R,R)-3

The mother liquor was rotavaped to eliminate its EtOH and extra water content by reaching ~30% of its initial volume. The concentrate was then subjected to the diffusion of EtOH in order to eliminate (–)-(S,S)-3 enriched CD-MOF crystals from it. This step was repeated one more time to eliminate any remaining (–)-(S,S)-3 enriched CD-MOF crystals. The final mother liquor was then acidified with the addition of HCl before isolating (+)-(R,R)-3 from it with a C18 reversed-phase chromatography column and H₂O–MeOH gradient. The isolated fractions containing (+)-(R,R)-3 were combined, rotavaped, and dried under a high vacuum to afford an off-white solid. The remaining solutions of TBA 3 from the crystallization steps were then combined, acidified with the addition of HCl and chromatographed with a C18 column, rotavaped, and dried under a high vacuum to recover all remaining amounts of TBA 3 for storage.

3. Results and Discussion

In order to achieve enantiopurification of ethano-bridged TBA 3 through co-crystallization, we added the potassium salt of its racemate (±)-3 to the solution in which CD-MOF crystals grow. The NCIs exist between 3 and γ-CD, e.g., hydrophobic repulsion effect and hydrogen bond interactions, help the incorporation of 3 in CD-MOF, abbreviated as 3⸦CD-MOF in this work. This selective inclusion is in agreement with superior enantioselectivity and chiral recognition of TBAs by γ-CD-based CSPs in the reversed-phase mode [28]. CD-MOF, as a homochiral framework [27], interacts with (–)-(S,S)-3 more than (+)-(R,R)-3 and allows its incorporation in its chiral pores and cavities. Although NMR analysis of the digested MOF crystals indicated the inclusion of TBA 3 in CD-MOF, SCXD could not display the guest molecule, perhaps due to their random positioning within the MOF [27,50]. Moreover, crystallization of TBA 3 was found to be exhausting and futile in various solvent systems, buffers, and at various pH values and in the presence of different counter ions. These attempts led to the formation of amorphous chalk-looking solids; except for one TBA 3 sample stored in methanol at low pH that resulted in a partial methyl-esterification of TBA 3. Serendipitously, this partially esterified sample contributed to the crystallization and determining the structure of TBA 3 (Figure 1 and Figure 2) by single crystal X-ray diffractometry (SCXD). As expected, SCXD analysis of TBA mixture containing TBA 3 revealed that its ethano-bridge tightly ratchets down the dihedral angle (81°, Figure 1) existing between the aromatic residues, making it smaller than values (82–110°) reported for most methano- and some ethano-bridged TBAs [51,52,53,54,55].

Our SCXD analysis also revealed that although TBA 3 molecules are densely functionalized, they do not form any intermolecular hydrogen bond interactions among each other in solid state and instead trap water molecules in between their tertiary amine and carboxylic acid groups by the formation of well-defined hydrogen bonds (Figure 2). These hydrogen bonding interactions were accurately located and measured within the crystal lattice after visualizing water molecules trapped in between two TBAs (Figure 2). This demonstrated the interactive sites of TBA 3 are available for interacting with γ-CD. Our ¹H NMR spectroscopic titrations also indicated the existence of NCIs and the binding constant between (–)-(S,S)-3 and γ-CD.

SCXD analysis of the obtained CD-MOF crystals revealed the commonly known cubic structure assuming I432 cubic unit cell with cell dimensions of 31.1 × 31.1 × 31.1 Å (α = β = γ = 90°). As shown in Figure 3, the CD-MOF porous structure includes interconnected voids and tunnels, which are 9 and 17 Å in diameter, respectively. Based on our earlier findings [27] and reports of other researchers [55], these tunnels are narrower than the central spherical pores and, hence, can be more selective for TBA 3. These tunnels are just large enough to accommodate TBA 3 of specific handedness in their chiral cavity. In contrast, the spherical voids are much larger and, hence, may impose weaker enantioselectivity. This assumption explains why the first crop of three co-crystallized with CD-MOF could not exceed 26–30% ee despite multiple optimization attempts. This moderate level of enantioenrichment was more or less reproduced for the second and third crops, which in the end resulted in enantiopure 3. The first crop of CD-MOF crystals indicated enantioenrichment with an average ee value of 28% for (–)-(S,S)-3 in CD-MOF and (+)-(R,R)-3 in the mother liquor. Therefore, these enantioenriched fractions were recovered and separately subjected to co-crystallization with CD-MOF in freshly prepared alkaline solutions of γ-CD. Three to four repeated cycles of co-crystallization of 3 with CD-MOF resulted in enantiopure TBAs (–)-(S,S)-3 and (+)-(R,R)-3 whose optical purity was confirmed by reversed-phase chiral HPLC analysis and CD spectra (Figure 4).

4. Conclusions

A key feature of CD-MOFs is the large number of intraframework voids functionalized with free homochiral stereocenters (24 stereogenic hydroxyl groups per γ-CD torus), which enables it to enantioselectivity host specific types of chiral compounds. CD-MOF preparation [23] only requires inexpensive, fully recoverable, nonhazardous materials at ambient temperature that facilitate the scale-ups to compensate for the low overall yield. Although the ee value obtained at each cycle of co-crystallization is not particularly impressive, as it leads to 3–5% of enantiopure products by the end of the process, the entire process costs less and is more productive than chiral HPLC, as it leaves no hazardous wastes and requires little to no effort as the molecules self-assemble. The chiral analytical HPLC separations took an hour, and each run involved about 1 microgram of the analytes, which were detectable by the detector but not worth collecting. Preparative chiral HPLC could have been more productive; however, the columns may not have been affordable for all users. Moreover, the method may not successfully transfer because higher loads of TBA may precipitate. In addition, this specific type of column cannot be washed with high organic mobile phases, making them susceptible to fouling after several runs.

A major disadvantage of co-crystallization is the difficulty of controlling nucleation. CD-MOF nucleation can be influenced by the presence of insoluble particles that can be introduced from the air, laboratory reagents, and equipment. Filtration of the crystallization media and flushing the glassware with purified nitrogen, instead of wiping or drying on racks or in ovens, can minimize problems associated with dust, static electricity, and lint from laboratory wipes. The anti-solvent diffusion rate can also influence nucleation and crystallization rates, which highly depend on temperature, solvent purity, glassware shape, size, and design. TBA 3 and γ-CD are quite easily separatable by regular C-18 reversed-phase flash chromatography and can be reused or stored after recovery. In addition, CD-MOF’s mother liquor can be rotavaped in order to remove its ethanol content, brought back to the volume, and subject to co-crystallization again. TBA 3 is stable in aqueous solutions and can be retrieved from γ-CD solution due to their retention difference on regular C-18 columns. γ-CD has little to no affinity to the C-18 phase and exists in the column at a low organic gradient where TBA 3 requires at least 40% more organic solvent to be washed off the column. Since regular C-18 columns are sensitive to high pH values, we adjusted the pH to around 5 with an HCl solution, which was required to retrieve TBA 3 entirely. These pH changes could only racemize methano-bridged TBAs and not ethano-bridged ones such as TBA 3, as explained in the mechanism earlier [34]. All in all, this fundamental research has opened new possibilities by conceptualizing chiral separations by co-crystallizations within homochiral MOFs.

5. Patents

There are no patents nor competing financial interests to declare.