The D-Leu used in this study was a special grade (purity > 98.0%) and purchased from WAKO Pure Chemicals. L-Leu and L-Val used in this study were pharmaceutical grade and purchased from AJINOMOTO Co., Inc.

The experimental flow is shown in Figure 1. First, a 30 g/L L-Val aqueous solution was prepared. L-Leu or D-Leu was added to this solution and the resulting solution was stirred. The solution was then evaporated until the concentration of L-Val in the solution (C_L-Val) was 20 wt%. Crystals formed in the solution. The resulting slurry was subsequently placed in a 50˚C water bath, cooled from 50˚C to 10˚C at a cooling rate of 10˚C/h, and

then maintained at 10˚C overnight. The slurry was filtrated using 0.45 μm membrane filters. The wet crystals were washed using MeOH. The concentrations of L-Leu, D-Leu, and L-Val were analyzed by high-performance liquid chromatography (HPLC; column: MCIGEL CRS10W; column temperature: 30˚C; detection wavelength: 254 nm; measuring time: 15 min; mobile phase: 1 mM CuSO₄ aq./MeOH [v/v] = 85/15; flow rate: 1 mL/min). The crystal structure was analyzed using powder X-ray diffraction (PXRD: Bruker D2 PHASER, CuKα, 30 kV) and the crystal shape was analyzed by scanning electron microscopy (SEM; JEOL JSM-6510).

The initial experimental system is represented as the molar ratio between the guest amino acid and the host amino acid. For example, L-Leu crystallization in L-Val is represented as “L-Leu/L-Val”. A comparison of the incorporation behaviors of L-Leu/L-Val and D-Leu/L-Valis shown in Figure 2, where the ratios of guest amino acids to L-Val in the crystals are plotted against those in the mother liquor at equilibrium. In the case of L-Leu used as aguest amino acid (blue solid circles), L-Leu was incorporated into L-Val crystals. The impurity content in the crystals was linearly plotted against tha tin the mother liquor. By contrast, in the case where D-Leu was used as aguest amino acid (red open circles), all D-Leu remained in the mother liquor. When the molar ratio of D-Leu to L-Val (represented as x_D-Leu) in the mother liquor was less than 0.12, D-Leu was not incorporated into L-Val crystals. However, when x_D-Leu was greater than 0.12, D-Leu crystallized with L-Val crystals. Thus, the solubility of D-Leu in the mother liquor was x_D-Leu = 0.12.

In summary, L-Leu was incorporated in a certain ratio of x_L-Leu in the mother liquor. D-Leu was not incorporated when its concentration in the mother liquor was less than the solubility limit; however, D-Leu was crystallized when its concentration in the mother liquor was greater than the solubility limit. In this report, we mainly discuss cases where x_L-Leu < 0.12 and x_D-Leu < 0.12 in the mother liquor.

The structure of crystals obtained by crystallization was analyzed by PXRD (Figure 3). The molar ratio between L-Val and L-Leu in the crystal is denoted as L-Leu/L-Val in this figure. The diffraction angles (2θ) of the (001) and (002) peaks in the pattern of L-Val were detected at 2θ = 7.3˚ and 14.6˚, respectively. The diffraction angles of the (001) and (002) peaks in the pattern of L-Leu were detected as 2θ = 6.0˚ and 12.1˚, respectively. In the case of L-Leu incorporated into L-Val crystals, these peaks were shifted to lower angles than those of single- crystalline L-Val and the extent of their shift was dependent on x_L-Leu in the crystals. The relationship between the diffraction angle and lattice spacing, known as Bragg’s law, is shown in Equation (1):

where d is the lattice spacing, n is an integer (n = 1, 2, 3, ∙∙∙) representing the order of diffraction, and λ is the X-ray wavelength (CuKα: 1.5418 Å). As θ decreases, d increases. Our PXRD results show that L-Leu was

incorporated into the unit cell of an L-Val crystal and that the c-axis expanded. Moreover, the incorporation of greater amounts of L-Leu into L-Val crystals resulted in broader peaks. It was speculated that the regularity of the crystal array decreased because L-Leu was randomly incorporated into the L-Val crystal lattice. We therefore considered these crystals to be a solid solution. L-Val and L-Leu peaks were not detected in any of the PXRD patterns of L-Leu/L-Val. Therefore, L-Val and L-Leu single crystalsdid not form; only the L-Leu/L-Val solid solution crystallized.

The crystal shape differed markedly between the L-Leu/L-Val crystal and the D-Leu/L-Val crystal (Figure 4). When x_D-Leu was less than 0.12 in the mother liquor, the molar ratio of the obtained crystal was D-Leu/L-Val = 0.03/99.97. D-Leu was hardly incorporated into L-Val by crystallization. The crystal clearly exhibited a hexagonal-plate morphology, which was identical to that of the L-Val single crystal. However, vaguely outlined fine crystals were obtained in the case of L-Leu/L-Val crystallization; this morphology was a result of the formation of a solid solution. In the L-Leu/L-Val solid solution, L-Leu was randomly incorporated into the L-Val crystal lattice. Because the side chain of L-Leu is larger than that of L-Val, L-Val crystal growth was stopped at the point of L-Leu incorporation, disturbing the normal growth of L-Val crystals. Therefore, the L-Leu/L-Val solid solution crystallized as vaguely outlined and fine crystals.

L-Leu was incorporated into L-Val crystals, whereas D-Leu was not. We speculate that this difference in incorporation behavior was due to the magnitude of the interaction energy between a molecule of the guest amino acid (L-Leu or D-Leu) and an L-Val crystal. The interaction energy is correlated with the stability of the crystal structure: when the interaction energy is lower, the crystal structure is more stable.

We calculated the interaction energy on the basis of this assumption. The interaction energy is difficult to determine experimentally; we therefore calculated it using a molecular dynamics package (Materials Studio, from Dassault Systemes Biovia K.K.). The L-Val 4 × 4 × 2 crystal structure was drawn and optimized geometrically using the cvff function, which is a generalized valence potential function used to model small organic crystals and gas-phase structures such as amides, carboxylic acids, proteins, and a wide range of organic compounds [12] [13] . A guest amino acid was used to replace one of the existing L-Val (100) surfaces (structure A). The crystal lattice in which the guest compound was removed from structure A was constructed (structure B). The crystal lattice in which all L-Val except the guest compound was removed from structure A was also constructed (structure C). Structures A-C are shown in Figure 5. The interaction energy (E_I) was estimated from Equation (2):

“Guest amino acid” means substituted amino acid. Materials Studio was used to estimate the interaction energy (module: forcite, potential function: cvff).

where E_A, E_B, and E_C were calculated as the total energy of structure A, structure B, and structure C, respectively.

The estimated interaction energy is shown in Table 1. In the case where L-Val was used as a guest amino acid, E_I represents the interaction energy between a molecule of L-Val and an L-Val crystal. A comparison of the interaction energies reveals that the energy of D-Leu was remarkably higher than those of L-Leu and L-Val. The interaction energy is correlated with the likelihood of incorporation. In the case of a high interaction energy, the crystal lattice became unstable and difficult to construct; therefore, D-Leu was difficult to be incorporated.