Preparation of chiral 3-oxocycloalkanecarbonitrile and its derivatives by crystallization-induced diastereomer transformation of ketals with chiral 1,2-diphenylethane-1,2-diol

Chiral 3-oxocycloalkanecarbonitriles were prepared by fractional crystallization and crystallization-induced diastereomer transformation (CIDT) of diastereomeric ketals with (1R,2R)-1,2-diphenylethane-1,2-diol. Investigation of the crystal structures by X-ray diffraction analysis revealed that the difference in hydrogen bonds caused the discrepancy of the solubilities between (R) and (S) diastereomers. Furthermore, CIDT to afford the (R)-diastereomer in good yield (95% yield) and with high diastereoselectivity (97% de) was accomplished, which is the first example of CIDT of neutral compounds via formation of the diastereomeric ketal with (1R,2R)-1,2-diphenylethane-1,2-diol.


Introduction
In recent years, the structure of active pharmaceutical ingredients and ne chemicals has become more complicated. To meet the demands for synthesizing organic molecules with a sophisticated design, building blocks containing chiral carbons as a component are widely used. Developing a new chiral building block will enable the synthesis of a new compound and provide benets to both the chemical and pharmaceutical industries.
As new candidate building blocks, we focus on 3-oxocycloalkanecarbonitriles, 1a 1a,b and 1b 1c (Fig. 1). In fact, active pharmaceutical ingredients or intermediates containing 3-oxocycloalkanecarbonitriles or its derivatives have been reported, 2a,b which implies that 1 has potential as a building block. In spite of its structural simplicity, the preparation of enantiomerically pure 1 remains a challenging task.
Although 3-cyanoketones 1 are easily prepared from Michael addition of cyanide ions to a,b-unsaturated ketones, 3a,b there are a few articles reporting the preparation of chiral 3-cyanoketones by catalytic enantioselective conjugate addition of cyanide to enones. Specically, 1b was obtained in 81% ee and 90% yield, 4 and nucleophilic addition of formaldehyde dialkylhydrazones to conjugated enones has been reported. 5a,b However, no optical resolution of these neutral compounds has been reported, as these compounds are not applicable for diastereomeric salt separation, which is the most popular method to resolve racemic compounds. In fact, 3-oxocyclopentanecarboxylic acid as an acidic compound was resolved by diastereomeric salt formation with (À)-brucine, but four sequential crystallizations were required to obtain (R)-enantiomer in 98% ee. 6 In order to introduce chiral moiety onto 1 and apply diastereomeric separation, we use chiral ketals as not only a protecting group but also chiral resolving auxiliary. Among several 1,2-diols, commercially available (1R,2R)-or (1S,2S)-1,2diphenylethane-1, 2-diol (dihydrobenzoin) shows great promise as a chiral auxiliary. 7 In fact, several articles reported the separation of two isomers via formation of the diastereomeric ketals, and subsequent isolation either by column chromatography or crystallization. 8a-c Preliminarily, we synthesized diastereomeric mixtures of 2, but unfortunately, it was oily substance, which indicated that diastereomer separation by recrystallization was not applicable. Therefore, we transformed nitriles 2 into amides 3, which are generally expected to solidify due to the formation of hydrogen bonds. First, we examined the preparation of chiral 3-oxocycloalkanecarbonitrile and their derivatives via ketalization with (1R,2R)-1,2-diphenylethane-1,2-diol through fractional crystallization ( Fig. 1). Even if separation of diastereomer is achieved through fractional crystallization, half of the diastereomeric mixture would remain as an undesired diastereomer. To our delight, 3 turned out to be racemized under basic conditions. Therefore, we performed crystallization-induced diastereomer transformation (CIDT) 9a-f on 3 and demonstrated the successful transformation of these compounds while keeping stereochemistry (Fig. 2).
For the purpose of applying the diastereomer separation by fractional crystallization, we hydrated nitriles 2 with a combination of hydrogen peroxide and potassium carbonate 12 to obtain crystalline amides 3.

Investigation of crystal structures by single-crystal X-ray diffraction analysis
To determine the stereochemistry of these diastereomers and to clarify why (R)-diastereomer crystallized preferably, we investigated the crystal structures of their diastereomers by singlecrystal X-ray diffraction (SXRD) analysis. To obtain both diastereomers of 3, we rst tried to separate the diastereomeric mixture of 3a by recycling preparative HPLC. However, 3a was inseparable due to the similar retention time of the diastereomers. In contrast, the corresponding nitriles 2a could be satisfactorily separated by recycling preparative HPLC. Here, we conrmed that both diastereomers of 2a were denitely oil. Then, (R)-2a and (S)-2a were transformed to crystalline amides (R)-3a and (S)-3a under the basic conditions mentioned above without epimerization. 12 In sharp contrast with 2a, even ten cycles of recycling preparative HPLC could not separate 2b. Aer transformation to the corresponding amide 3b, diastereomeric  mixtures of 3b were satisfactorily separated into (R)-3b and (S)-3b by recycling preparative HPLC. All amides 3 were crystallized to obtain crystals suitable for SXRD analysis and we were able to determine the stereochemistry ( Fig. 3 and 4). As shown in Fig. 3, both (R)-3a and (S)-3a had the same space group (P2 1 ) and a similar molecular arrangement to construct hydrogen bonding networks. Their amide groups act as hydrogen donor and acceptor to construct the same number of hydrogen bonds, namely, two amide protons bound to two oxygen atoms of the amide and the ketal. While the cis proton against the carbonyl oxygen in (R)-3a constructed hydrogen bonds with the amide functional group, the trans one did in (S)-3a. The parameters of the hydrogen bonds in (R)-3a and (S)-3a are summarized in Table 2. The hydrogen bonding distances in (R)-3a crystals were shorter than those of (S)-3a. The crystals of (R)-3a (mp 139-140 C, 1.26 g cm À3 ) had a higher melting point and larger calculated density than those of (S)-3a (mp 135-136 C, 1.23 g cm À3 ). We performed solubility tests on each diastereomer in toluene at 25 C and found that the values of (R)-3a and (S)-3a were 16.0 g L À1 and 21.5 g L À1 , respectively, as anticipated. These results suggest that the difference of strength of the hydrogen bonds caused the discrepancy of the solubility and consequently enabled fractional crystallization providing (R)-3a.
As shown in Fig. 4, the crystals of (R)-3b and (S)-3b had space groups of P2 1 and P2 1 2 1 2 1 , respectively. Similar molecular arrangements of constructing hydrogen bonding networks were observed in both diastereomers. While the cis proton against the carbonyl oxygen in (R)-3b constructed hydrogen bonds with amide functional groups, the cis one did not in (S)-3b. In other words, a cis proton of (S)-3b did not bind to an oxygen atom of the ketal. The crystals of (R)-3b (mp 164-165 C) had a higher melting point than those of (S)-3b (mp 152-154 C), but both had the same calculated density (1.24 g cm À3 ). Solubility tests on each diastereomer in toluene at 25 C revealed that the values of (R)-3b and (S)-3b were 35.3 g L À1 and 49.0 g L À1 , respectively. As discussed above, we conclude that the different hydrogen bonding networks caused the difference in both melting point and solubility and enabled fractional crystallization providing (R)-3b.    Table 1 shows that the separation capacity of 3a is superior to that of 3b in the aspect of fractional crystallization. To elucidate the reason for this difference, both diastereomeric mixtures (3a and 3b) were crystallized simply from the solution (toluene/ CHCl 3 ¼ 2/3) and the precipitated solid was analyzed using powder X-ray diffraction (PXRD) analysis. Fig. 5 shows the PXRD patterns of the crystallized diastereomeric mixture of 3 along with those of each single diastereomer as well as their simulated patterns calculated from SXRD. In the case of 3a, the PXRD pattern showed a superposition pattern of each diastereomer ((R)-3a and (S)-3a), which means the two diastereomers deposited separately. In contrast, the PXRD pattern of the solid precipitation of 3b shows a broad pattern with a partial component of (R)-3b. The pattern of (S)-3b was particularly hard to identify. These results suggest that the mixture of (R)-3b and (S)-3b might exist as an amorphous material and be what caused the broad PXRD pattern. If so, it would explain why the separation capacity of 3a is superior to that of 3b.

Epimerization and crystallization-induced diastereomer transformation (CIDT)
Even if a racemic compound is optically resolved by the diastereomer method, half of the undesired diastereomer remains in the ltrate. Therefore, epimerization of the remaining diastereomer (S)-3 into (R)-3 is desirable from the viewpoint of yield. The screening of the epimerization conditions using (S)-3a and (S)-3b is summarized in Table 3. DBU in toluene and potassium tbutoxide in dioxane or THF were not effective (entries 1-3). (S)-3a (>99% de) was smoothly epimerized with potassium t-butoxide (2 equiv.) at 50 C for 3.5 h in t-butanol, and the opposite diastereomer (R)-3a was slightly enriched with 7% de (entry 4). Under the same conditions, epimerization of (S)-3b (>99% de) proceeded more slowly than (S)-3a, where the % de of (S)-3b was 20% even aer 7 h (entry 5). Other bases such as NaH and KOH were not effective (entries 6 and 7). The combination of potassium tbutoxide and t-butanol for CIDT has been reported, 14 so we consider strong basic and protic conditions is appropriate for this epimerization. This positive result encouraged us to apply CIDT ( Table 4). The treatment of 3a (0.25 mmol, 2% de) with potassium t-butoxide (0.5 equiv.) in t-butanol (0.20 mL) at room temperature precipitated (R)-3a with 80% de (87% yield, entry 1). An increased amount of t-butanol (0.40 mL) with an extended stirring time (96 h) improved % de to 97% (95% yield, entry 2). Meanwhile, a similar procedure to entry 2 using 3b precipitated (R)-3b with 14% de (85% yield, entry 3). Although the attempt to increase the reaction temperature to 80 C in order to accelerate CIDT with addition of i-octane as a poor solvent provided better % de (44% de and 51% de, entries 4 and 5), these gures were not as high as those of 3a. As with the fractional crystallization of 3, the CIDT of 3a was superior to 3b. Anyway, we are convinced that the ketal moiety acted as not only a protecting group but also a chiral resolving auxiliary which is a useful tool for CIDT.

Conclusions
We have synthesized chiral 3-oxocycloalkanecarbonitrile and 3oxocycloalkanecarboxamide by fractional crystallization of ketal derivatives with (1R,2R)-1,2-diphenylethane-1,2-diol. Investigation of the crystal structures by X-ray diffraction analysis revealed that the difference in hydrogen bonds caused the discrepancy of the solubilities between the (R) and (S) diastereomers. Furthermore, CIDT to obtain (R)-3a in good yield (95% yield) and with high diastereoselectivity (97% de) was accomplished. Finally, successful derivatization of functional group and deprotection of ketals were performed without epimerization. To the best of our knowledge, we demonstrated the rst example of CIDT of neutral compounds via formation of the diastereomeric ketal with (1R,2R)-1,2-diphenylethane-1,2-diol. These ndings can be applied for synthesizing chiral and neutral building blocks containing carbonyl group.

General information
Starting materials, reagents, and solvents were obtained from commercial suppliers and used without further purication. Optical rotations were measured with a JASCO DIP-140 digital polarimeter at 20 C using the sodium D line, and optical rotation data were reported as follows: [a] 20 D (concentration c ¼ g/100 mL, solvent). 1 H and 13 C NMR spectra were acquired with a Varian Gemini 2000 NMR spectrometer at 300 MHz and 75 MHz, respectively. Chemical shis (d) of 1 H NMR were expressed in parts per million (ppm) relative to tetramethylsilane (d ¼ 0) as an internal standard. Multiplicities are indicated as br (broadened), s (singlet), d (doublet) and m (multiplet), and coupling constants (J) are reported in Hz unit. Chemical shis (d) of 13 C NMR were expressed in ppm downeld or upeld from CDCl 3 as an internal standard (d ¼ 77.0). Infrared spectra were acquired using in KBr disk with a JASCO FT/IR-460 plus spectrometer. Mass spectra were acquired with a Thermo Fisher Scientic Exactive spectrometer. Powder X-ray diffraction were acquired with a Bruker D8 ADVANCE. Single-crystal X-ray diffraction were acquired with Bruker APEX II and Bruker APEXII Ultra CCD diffractometers. Recycling preparative HPLC was performed with a JAI LC-908. Enantiomeric excess (ee) and diastereomeric excess (de) were determined by chiral HPLC analysis with a JASCO LC-2000Plus system and a SHIMAZU LC-2010 system.