(T1430-01-06) Systematic Screening of Pharmaceutical Co-crystals Using Experimental and Computational Tools

Purpose: Pharmaceutical co-crystals are multi-component molecular solids wherein a drug and coformer are crystallized into a single, new solid phase. A rich library of co-crystals collectively demonstrates the potential of co-crystal engineering to improve drug product performance; however, the broad majority of these cases are discovered rather than designed, and thus the promise of tailoring drug product performance through co-crystallization remains quite limited. Co-crystal screens mainly focus on complementary hydrogen-bonding interactions and thus, the role of other molecular properties of drug and coformer on co-crystal formation is limited. To narrow these knowledge gaps, this study was designed to investigate and assess three methods for predicting co-crystal formation with naringenin (NAR), a model flavonoid with poor aqueous solubility and potential treatment for Alzheimer’s disease. Our test bed was composed of 39 co-formers which we found to produce 9 co-crystals with NAR and 30 physical mixtures.

Methods: A series of pharmaceutical and pharmacological selection criteria were developed to identify a pool of NAR-specific coformers that (1) lead to probable co-crystal formation, (2) likely improve naringenin solubility at physiological pH and (3) are safe for human use. Starting with 250 molecules collected from safe-for-human-use databases, an initial pool of 24 NAR-compatible co-formers (Group A) was selected and underwent a thorough solid-form screening procedure. Methods of co-crystal synthesis included liquid-assisted grinding (LAG), slurry at equimolar and non-equimolar stoichiometric ratios, reaction crystallization, and solvent evaporation. Critical method parameters e.g., solvent choice, volume of solvents, and time were also varied to maximize opportunity for co-crystal formation. However, based on the limited co-crystal yield from Group A, a second group of 15 co-formers (Group B) was chosen. Selection criteria were consistent with Group A, however, in addition a statistical analysis using ConQuest software of the Cambridge Structural Database (CSD) provided a series of frequently recurring hydrogen-bonding synthons to target. Powder X-ray Diffraction (PXRD) and Differential Scanning Calorimetry (DSC) were used to determine co-crystal formation. Our screening process generated a test bed to evaluate the accuracy of common models for predicting co-crystal formation with NAR. Cocrystallization outcomes for NAR and all thirty-nine coformers were determined by (1) calculating frequency of occurrence of synthon using ConQuest software to understand the role of hydrogen bonding and (2) using the co-crystal design tool, which is based on a variety of molecular descriptors, provided in Mercury 3.0. Lastly, the impact of drug-coformer molecular miscibility on co-crystal formation was examined by analyzing the endotherm patterns of physical mixtures of drug and coformer in DSC [1].

Results: Five co-crystals, with nicotinamide, isonicotinamide, proline, betaine, and picolinic acid were produced with LAG from Group A and confirmed by PXRD and DSC. The slurry method resulted in a NAR co-crystal with gallic acid and solvent evaporation produced two polymorphs of NAR: caffeine co-crystals. A different polymorph of the NAR: nicotinamide co-crystal was obtained using slurry method with non-equimolar stoichiometric ratio. Co-crystals of NAR with nicotinamide, isonicotinamide, proline, betaine picolinic acid, and caffeine have been previously reported [2, 3]. Two new co-crystals, with succinimide and tetramethyl pyrazine, were generated from Group B using both LAG and slurry methods. The higher frequency of synthon formation, calculated by ConQuest, for hydroxyl group (present in NAR) was found to be with carbonyl, aromatic nitrogen and carboxylate ion; however, our observed success rate for co-crystal formation for Group B is around 13% despite leveraging these hydrogen-bonding interactions. The prediction accuracy of the co-crystal design tool in Mercury 3.0 was better at 25%, but several false positives were observed with more than 50% of coformers and thus, molecular properties such as shape, size, dipole moment of drug and coformer need to be refined to better predict co-crystal formation. Lastly, the molecular miscibility screen using DSC showed the best correlation among the three methods.

Conclusion: Nine pharmaceutical co-crystals of NAR, six previously reported and three new co-crystals, were generated in this work. NAR is a weakly ionizable compound with poor aqueous solubility and low bioavailability, and thus cocrystallization is a reasonable option for potentially improving its performance. Our co-crystals have potential as novel drug products because both physicochemical and pharmacological selection criteria were established at the outset; however, the success rate observed for new co-crystal formation (4 out of 34) was around 10%. We found that prediction tools that utilized hydrogen-bonding and molecular properties to anticipate co-crystal formation performed modestly, while DSC-based miscibility screening showed better success. Overall, new tools/descriptors to predict and/or screen probable co-crystal formation are necessary to advance co-crystal engineering in pharmaceutical solid-form discovery.

References: 1. Manin AN, Voronin AP, Drozd KV, Manin NG, Bauer-Brandl A, Perlovich GL. Cocrystal screening of hydroxybenzamides with benzoic acid derivatives: a comparative study of thermal and solution-based methods. Eur J Pharm Sci. 2014;65:56-64.
2. Cui W, He Z, Zhang Y, Fan Q, Feng N. Naringenin cocrystals prepared by solution crystallization method for improving bioavailability and anti-hyperlipidemia effects. AAPS PharmSciTech. 2019;20:1-12.
3. Luo C, Liang W, Chen X, Wang J, Deng Z, Zhang H. Pharmaceutical cocrystals of naringenin with improved dissolution performance. CrystEngComm. 2018;20(22):3025-33.

Figure 1. Schematic representation of systematic screening process of pharmaceutical co-crystals of naringenin (NAR).

Figure 2. Assessment of predictive and screening tools in comparison with experimental results.