Absolute Configuration Determination of Chiral API Molecules by MicroED Analysis of Cocrystal Powders Formed based on Cocrystal Propensity Prediction Calculations

Abstract

Establishing the absolute configuration of chiral active pharmaceutical ingredients (APIs) is of great importance. Single crystal X-ray diffraction (scXRD) has traditionally been the method of choice for such analysis, but scXRD requires the growth of large crystals, which can be challenging. Here we present a method for determining absolute configuration that does not rely on the growth of large crystals. By examining microcrystals formed with chiral probes (small chiral compounds such as amino acids), absolute configuration can be unambiguously determined by microcrystal electron diffraction (MicroED). Our streamlined method employs three steps: (1) virtual screening to identify promising chiral probes, (2) experimental cocrystal screening and (3) structure determination by MicroED and absolute configuration assignment. We successfully applied this method to analyze two chiral API molecules currently on the market for which scXRD was not used to determine absolute configuration.

Graphical Abstract

Depicted is a new and robust pipeline for absolute configuration determination. First, virtual screening selects chiral probes compatible with the active pharmaceutical ingredient (API). Second, crystallization trials produce microcrystals suitable for analysis without further optimization. Finally, MicroED is used to determine the crystal structure. The absolute configuration of the API can be confidently assigned relative to the chiral probe.

Introduction

Of the 32 small molecule drugs approved by U.S. Food and Drug Administration in 2021, >70% were chiral compounds formulated as enantiopure products.^[1] For most chiral small molecule drugs, only one enantiomer is biologically active, while the other enantiomer can potentially lead to unwanted side-effects and even toxicity. D-penicillamine is a prime example of a drug molecule that must be formulated as enantiopure because its enantiomer, L-penicillamine, is highly toxic.^[2] Therefore, it is critical that researchers establish ways to isolate stereoisomers, characterize all stereocenters (absolute configuration) with a high degree of certainty and determine the biological activity of each stereoisomer/enantiomer. Several methods exist for determining the absolute configuration of chiral molecules, including circular dichroism,^[3] optical measurements^[4] and relative configuration assignment via nuclear magnetic resonance (NMR).^[5] Historically, single crystal X-ray diffraction (scXRD) has been the gold standard for establishing the absolute configuration of active pharmaceutical ingredients (APIs), either directly from the measured intensities using the Flack parameter^[6] or indirectly by the incorporation of a chiral probe or other internal reference.^[7]

Unfortunately, scXRD is limited by the need to grow suitably sized crystals (≥ 5μm x 5μm x 5μm), which can be challenging for many APIs. Recently, a new method has emerged for determining structures from microcrystalline powders termed microcrystal electron diffraction (MicroED), a 3D electron diffraction (3D ED) method.^[8–10] This method has recently been applied to a growing number of API molecules.^[11,12] Unfortunately, absolute configuration determination by MicroED is not as straightforward as for scXRD. As electrons are thought to produce minimal anomalous scattering, the Flack parameter cannot be used for MicroED data. For inorganic materials, combining electron diffraction imaging with atomic resolution crystal imaging has allowed for absolute configuration determination.^[13,14] Unfortunately, this method has not yet been applied to organic molecules, likely due to their radiation sensitivity. For organic molecules, accounting for dynamical scattering has been used to determine absolute configuration.^[15,16] Dynamical refinement requires processing MicroED data in programs not typically used by pharmaceutical crystallographers that are highly computationally intensive.

Here we present an alternative method for absolute configuration determination from MicroED data by incorporating a chiral probe. Crystallography produces exactly two potential solutions for chiral crystals. These two potential solutions are related by inversion symmetry and only one solution can possibly be correct, regardless of the number of chiral centers in the molecule. Because of this property, knowing the configuration of one chiral center in the crystal lattice with certainty allows for the unambiguous assignment of all other stereocenters in the crystal and therefore the absolute configuration of the molecule(s) in the crystal. Because of this, incorporation of a chiral probe into the crystal form easily allows for absolute configuration determination. This method has been widely applied in scXRD^[7] and has recently been applied to MicroED data via covalent modification^[17] and chiral salt formation.^[18] Covalent modification is not possible for all APIs and salt formation requires that the API contains an ionizable group, making these methods unsuitable for many API molecules. We sought to develop a pipeline for determining absolute configuration via MicroED based on cocrystallization with chiral probes in order to expand the reach of this method, especially to molecules lacking ionizable groups (Figure 1).

Figure 1.

A streamlined method for confidently assigning the absolute configuration of an active pharmaceutical ingredient (API). The chiral coformer acts as a probe, allowing for the unambiguous assignment of all stereocenters in the API molecule by MicroED.

We proposed a streamlined approach to growing cocrystal powders suitable for MicroED data collection by first employing virtual chiral cocrystal screening to narrow the list of potential chiral probes to those most likely to succeed. The top coformer hits can then be employed in cocrystal screens and crystal hits can be identified by standard characterization methods. Suitable hits can then be characterized by MicroED and absolute configuration can be assigned relative to the chiral probe.

For the purposes of this study, two chiral API molecules currently on the market were selected for absolute configuration determination using our pipeline: ipragliflozin and finafloxacin. Ipragliflozin is a sodium-glucose transporter 2 (SGLT2) inhibitor that was approved in Japan in 2014 for the treatment of type 2 diabetes mellitus.^[19] Finafloxacin is a fluoroquinolone antibiotic that was approved in the United States in 2014 for the treatment of otitis externa (swimmer’s ear).^[20] Ipragliflozin and finafloxacin were selected because they lacked scXRD structures in the Cambridge Structural Database (CSD) and we speculated that the absence of such structures indicated that growing suitably sized crystals for scXRD may have been challenging for these compounds. Without scXRD structures, researchers likely had to rely on other, less reliable techniques to establish the absolute configuration of these APIs. These compounds were also selected because they were easy to procure and exhibited properties amenable to virtual screening such as low molecular weight.

Cocrystalline solid formulations have grown in popularity for their ability to improve poor physicochemical properties of APIs and intermediates.^[21,22] Two primary challenges need to be addressed when pursuing a cocrystal: (1) selection of potential coformers and (2) finding suitable crystallization conditions. Selecting appropriate coformers is challenging because the chemical space of potential coformers is vast. Additionally, many crystallization techniques exist for obtaining cocrystals, including solution-mediated reactions, ball milling, evaporation, cooling and melt quench.^[23] Since there is no way to know a priori which experimental technique will yield a cocrystal for a given API-coformer pair, experimentalists often try multiple methods for each pair. This approach requires at least several grams of material and is very time-consuming. Virtual cocrystal screening directly addresses the first challenge, improving the chances of success and reducing the number of experiments to be performed. Virtual cocrystal screening encompasses a series of knowledge-based,^[24–29] physics-based^[30–38] and machine learning-based^[39,40] computational techniques that were developed to predict the likeliest compound-coformer pairs to form a cocrystal. In this study, we employed a fast and accurate virtual cocrystal screening method that combines a conductor-like model for realistic solvents (COSMO-RS),^[41] a machine learning (ML) model^[42] and pKa considerations to screen coformers.^[42] The hybrid physics-based and ML-based model provides better propensity rankings than either model used independently. This screening method is fast, generating results in approximately three days, making it conducive to the rapid timelines seen in the pharmaceutical industry. Virtual screening techniques reduce the list of potential coformers by ranking and prioritizing those most likely to form a cocrystal with the target compound.

Because our goal was absolute configuration determination and not generating a solid formulation suitable for consumption, our list of potential coformers was not limited to the Generally Recognized as Safe (GRAS) list. Instead, we selected chiral probes based on these criteria: (1) contains a chiral center, (2) easy procurement and (3) low molecular weight. Using these criteria, we produced a list of 43 chiral probes suitable for absolute configuration studies (Table S2). Ipragliflozin was already known to form a cocrystal with L-proline,^[43] so L-proline was also included in the virtual screen for ipragliflozin.

Results and Discussion

Virtual screening produced a ranked list of chiral probes for each API molecule. Based on the virtual screen rankings, cocrystallization experiments were carried out with 11 chiral probes for ipragliflozin and 10 chiral probes for finafloxacin. Experimental screening for ipragliflozin produced one crystal hit with L-proline (ranked 13^th). Experimental screening for finafloxacin produced eight hits observed by PLM, two of which passed further characterization criteria (high crystallinity by XRPD, stoichiometry confirmation by ¹H-NMR and melting point shift by DSC). The confirmed hits were formed with (1S)-(+)-menthyl chloroformate (ranked 3^rd) and (R)-(+)-α-methylbenzyl isocyanate (ranked 8^th).

The structure of ipragliflozin + L-proline was determined by MicroED (Figure 2A). With the chiral center in L-proline as a guide, the stereochemistry of all five stereocenters in ipragliflozin could be unambiguously assigned (Figure 2B). This configuration is in line with the expected structure.^[19]

Figure 2.

(A) The MicroED structure of an L-proline and ipragliflozin cocrystal (ranked 13^th of 44 in the virtual screen). (B) The absolute configuration of ipragliflozin was determined relative to the L-proline in this cocrystal.

The structure of finafloxacin and (1S)-(+)-menthyl chloroformate was also determined by MicroED (Figure 3A). To our surprise, this was not a cocrystal and instead was an adduct product formed by a reaction between finafloxacin and (1S)-(+)-menthyl chloroformate (Figure 3B). Thankfully this substitution reaction was unlikely to have affected the stereochemistry of the two chiral centers in finafloxacin. While this crystal form is unsuitable for formulation purposes, it can still be used for absolute configuration determination. Again, the chiral centers in the probe could be used to unambiguously infer arrangement of the two chiral centers in finafloxacin, which were in line with the available literature for finafloxacin.^[20] We speculate that the our cocrystallization screens produced an adduct in the case of finafloxacin because many of the chiral probes used in this study were not on the GRAS list and the list therefore contained some reactive compounds such as (1S)-(+)-menthyl chloroformate. As this substitution reaction did not impact our ability to determine the absolute configuration of finafloxacin, this was considered a successful result. We expect that the same virtual screening procedure with a list of coformers from the GRAS list would be much less likely to result in adduct formation, as would be desirable for solid formulation cocrystal screening.

Figure 3.

(A) The MicroED structure of the (1S)-(+)-menthyl chloroformate and finafloxacin crystal hit (ranked 3^rd of 43 in the virtual screen) revealed an adduct, not a cocrystal. (B) The putative reaction responsible is shown. (C) Despite this, the stereochemistry of the chiral groups in finafloxacin could still be determined relative to the (+)-menthyl carbamate.

Conclusion

This study outlines a novel pipeline for determining the absolute configuration of an API molecule when single crystal X-ray crystallography is too challenging, would require too much time and/or material, or is simply not possible despite considerable effort. We successfully applied this pipeline to two APIs currently on the market for which scXRD was not used for absolute configuration determination: ipragliflozin and finafloxacin. This effort was accomplished in ~6 weeks, required 15–20mg of API material per screen (800–880mg total) and <1 mg crystalline powder for MicroED. Virtual chiral cocrystal screening took approximately one week and cut the experimental screening time in half by reducing the number of potential coformers to a tractable number. Experimental cocrystal screening involved 40–44 experiments per API and spanned 2–4 weeks. Importantly, in both cases at least one cocrystal hit was produced from the virtual screen ranked list of coformers. These crystal hits could be directly used for MicroED measurements without the need to optimize the crystallization condition. MicroED structure determination was accomplished in about two weeks per sample, including the time required for structure refinement. We note that structure determination by MicroED can be limited by the resolution of the crystals, which is generally an inherent property of the packing interactions building each individual crystal lattice. Because of this, it can be helpful to have multiple crystal hits to choose from, with X-ray powder diffraction being an excellent guide to help prioritize samples for MicroED structure determination. We expect that this procedure can be applied to many API molecules; and we hope it is used to establish absolute configuration earlier in the drug development process with minimal effort and a high degree of confidence.

Experimental Section

The COSMO-RS + ML^[41,42] model was applied to the two investigated API molecules, ipragliflozin and finafloxacin, with the chiral probe list (Table S2). The API-coformer pairs were ranked by the ML probability, and pairs with ΔpKa > 3 were excluded (Tables S2 and S3). The top 11 and 10 coformers for ipragliflozin and finafloxacin respectively were subjected to experimental cocrystallization screening.

For each cocrystallization experiment, 15–20 mg of API was mixed with the coformer at a molar ratio of 1:1 in 0.2 – 1.0 mL solvent. Four solvents were tested for each API-coformer pair (Tables S5 and S7). Slurry experiments were magnetically stirred for 7 days at 25 °C, whereas reactive crystallization experiments were magnetically stirred for 3 days at 25 °C. If no solids were observed, the clear solutions were subjected to stirring at 5 °C for 2 days followed by slow evaporation. Samples were taken from each vial for XRPD analysis, and subsequently DSC, PLM and NMR if the XRPD pattern was different from those of the starting materials.

Three crystal hits were examined by MicroED: ipragliflozin L-proline, finafloxacin (1S)-(+)-menthyl chloroformate Type A (finafloxacin (+)-menthyl carbamate), and finafloxacin (1R)-(+)- α-methylbenzyl isocyanate Type A (Table S1). MicroED was carried out as previously described.^[12] Structures were determined of ipragliflozin L-proline and finafloxacin (+)-menthyl carbamate.

Further experimental details, tables, analysis and spectra can be found in the Supporting Information.

Data Availability

The raw data that support the findings of this study are openly available in Zenodo at https://doi.org/10.5281/zenodo.7308613, https://doi.org/10.5281/zenodo.7308682 and https://doi.org/10.5281/zenodo.7308788. Deposition Number(s) <url href=“https://www.ccdc.cam.ac.uk/services/structures?id=doi:10.1002/###.20220XXX”> 2168647 (for finafloxacin (+)-menthyl carbamate), 2174023 (for ipragliflozin L-proline)</url> contain(s) the supplementary crystallographic data for this paper. These data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe <url href=“ http://www.ccdc.cam.ac.uk/structures ”> Access Structures service</url>.