I. INTRODUCTION
The number of drugs based on active pharmaceutical ingredients (API) in the form of salts has increased in the past decades (Wermuth and Stahl, Reference Wermuth and Stahl2001) due to the potential of multi-component solids to tune stability (Dhondale et al., Reference Dhondale, Thakor, Nambiar, Singh, Agrawal, Shastri and Kumar2023), solubility, or tabletability (Paulekuhn et al., Reference Paulekuhn, Dressman and Saal2007; Braga et al., Reference Braga, Grepioni, Lampronti, Maini and Turrina2011) of medications. However, the development of new drugs based on multicomponent solids can be hindered by the lack of high-quality reference powder patterns and/or crystal structures required for phase identification and quantitative analysis by the pharmaceutical industry (Hilfiker et al., Reference Hilfiker, Blatter, Raumer and Hilfiker2006). Thus, recent progress in algorithms for structure solutions from powder XRD data and an increase in computational resources resulted in increases of crystal structures of multi-component API solids obtained from powder X-ray diffraction (XRD). The following examples are besifloxacin hydrochloride (Kaduk et al., Reference Kaduk, Gates-Rector and Blanton2023a), butenafine hydrochloride (Kaduk et al., Reference Kaduk, Gates-Rector and Blanton2023b), carbamazepine cocrystals (Guerain et al., Reference Guerain, Derollez, Roca-Paixão, Dejoie, Correia and Affouard2020; Surov et al., Reference Surov, Drozd, Ramazanova, Churakov, Vologzhanina, Kulikova and Perlovich2023), zonisamide/ε-caprolactam cocrystal (Barbas et al., Reference Barbas, Portell, Prohens and Frontera2022).
Rilpivirine (DrugBank No. DB08864) is a non-nucleoside reverse transcriptase inhibitor used for the treatment of HIV-1 in treatment-naive adults (Putcharoen et al., Reference Putcharoen, Kerr and Ruxrungtham2013). The crystal structure of free base rilpivirine has been solved by means of powder X-ray diffraction by Kaduk et al. (Reference Kaduk, Zhong and Blanton2015). However, the commercially available medication Edurant contains rilpivirine in the form of hydrochloride (Scheme 1). Patents WO2012125993A1 (Rendell et al., Reference Rendell, Mittelman, Erlich and Ratkaj2012), WO2013153161A2 (Hotter et al., Reference Hotter, Pichler, Adamer and Griesser2013), and EP2628732A1 (Sandoz, Reference Sandoz2013) and paper by Kommavarapu et al. (Reference Kommavarapu, Maruthapillai, Palanisamy and Sunkara2015) contain information about polymorphs and solvates of rilpivirine hydrochloride; however, data about crystal structures of any rilpivirine salts are still missing. Within our research devoted to XRD characterization of solid forms of known APIs (Goloveshkin et al., Reference Goloveshkin, Korlyukov and Vologzhanina2021; Korlyukov et al., Reference Korlyukov, Dorovatovskii and Vologzhanina2022, Reference Korlyukov, Buikin, Dorovatovskii and Vologzhanina2023; Buikin et al., Reference Buikin, Vologzhanina, Novikov, Dorovatovskii and Korlyukov2023), the crystal structure of the title compound has been obtained by means of synchrotron powder X-ray diffraction. The solid was also characterized by means of solid-state 13C, 15N, and 35Cl NMR spectroscopy.
Scheme 1. The molecular structure of rilpivirine hydrochloride.
II. EXPERIMENTAL
Rilpivirine substance was purchased from Clearsynth (CAS No. 700361-47-3) and used without any purification. The infra-red (IR) spectrum of the substance was recorded on an IR spectrometer with a Fourier transformer Shimadzu IRTracer100 (Kyoto, Japan) in the range of 4000–600 cm−1 at a resolution of 1 cm−1 (Nujol mull, KBr pellets) and coincided with that published before (Kommavarapu et al., Reference Kommavarapu, Maruthapillai, Palanisamy and Sunkara2015). 15N and 13C NMR solid-state NMR experiments were recorded on a Bruker AVANCE III WB 400 MHz spectrometer equipped with a 4.0 mm magic angle sample spinning (MAS) BB/HF probe (15 kHz) and a 2.5 mm MAS BB/H probe (35 kHz) (1H – 400.1 MHz, 13C – 100.6 MHz, 15N – 40.6 MHz, and 35Cl – 39.2 MHz). Samples were spun at 9.5–14 kHz (4.0 mm probe) at the magic angle (MAS) using ZrO2 rotors. 13C cross-polarization / magnetic angle spinning (CP/MAS) spectra were recorded with a recycle delay of 8 s and contact times of 3 ms at 13 kHz. 15N-CP/MAS spectra were recorded with a recycle delay of 14–18 s and contact times of 6–15 ms at 9.5 kHz. 1H MAS spectra were recorded using single-pulse sequences at 14 kHz (4.0 mm probe) and 34 kHz (2.5 mm probe) with a recycle delay of 15 s. 35Cl MAS spectra were recorded using Hahn-Echo pulse sequences at 14 kHz (4.0 mm probe) in a rotor-synchronized mode using 63.2 μs echo-time (1 rotor cycle) with a recycle delay of 0.5 s. 2D 13C–1H CP-FSLG-HETCOR spectra were recorded at 13 kHz MAS (4.0 mm probe) in a rotor-synchronized mode using 300–2100 μs contact time under CP conditions to detect direct/long-range C–H correlations. All the 13C, 15N, and 2D HETCOR spectra were recorded under high-power proton decoupling conditions using “spinal64”. All chemical shifts for 1H and 13C are relative to the external adamantane sample; 15N chemical shifts were calculated to this scale and were checked using the 15N-labeled-glycine sample; 35Cl chemical shifts are relative to the external NaCl sample.
Elemental analysis was calculated for C22H19ClN6 (%): C 65.5; H 4.7; Cl 8.9; N 20.9. Found (%): C 65.46(12); H 4.86(4); Cl 9.04(8); N 20.64(10).
A. Powder X-ray diffraction
The synchrotron PXRD data were recorded at the X-ray structural analysis beamline (Belok/XSA) of the Kurchatov Synchrotron Radiation Source (Svetogorov et al., Reference Svetogorov, Dorovatovskii and Lazarenko2020). Monochromatic radiation was used to measure the pattern and then θ angles were recalculated to a wavelength of 0.7500 Å. The sample was placed in a cryoloop of 200 μm in size and rotated around the horizontal axis during the measurement, which made it possible to average the diffraction patterns according to the orientations of the sample. The diffraction patterns were collected by the 2D Rayonix SX165 detector, which was located at a distance of 150 and 250 mm with 29.8 and 18° tilt angles, respectively, Debye–Scherrer (transmission) geometry was used with a 400 μm beam size. The 2θ range was 0.99–59.8° with a step size of 0.00994°. The exposure time was 10 min. The two-dimensional diffraction patterns obtained on the detector were further integrated to the standard form of the dependence of the intensity on the scattering angle I(2θ) using the Dionis software (Svetogorov, Reference Svetogorov2018). To calibrate the sample–detector distance, the polycrystalline LaB6 (NIST SRM 660a) was used as a standard with the known position of the diffraction peaks. The diffraction peaks were approximated by fundamental parameters with Gaussian 1/cos(θ) convolution.
The indexing of the powder pattern and the subsequent structure solution were performed with the Topas 5.0 software (Coelho, Reference Coelho2003; Bruker TOPAS 5 User Manual, 2014). All patterns were indexed in the C-centered monoclinic unit cell. The systematic absences suggested the space group C2/c, which was confirmed by the successful solution and refinement of the structure. A molecular model of rilpivirine was taken from a publication by Kaduk et al. (Reference Kaduk, Zhong and Blanton2015) and converted into a Fenske–Hall Z-matrix file using OpenBabel (O'Boyle et al., Reference O'Boyle, Banck, James, Morley, Vandermeersch and Hutchison2011). A simulated annealing algorithm of the Topas 5.0 software was applied to find the position of non-hydrogen atoms of rilpivirine and an independent chlorine atom in an asymmetric unit. The solution result was used as a starting geometry for the periodic density functional theory (DFT) calculations at PBE (Perdew–Baron–Erzenhopf) exchange–correlation functional level with the fixed unit cell using VASP 5.4.1 (Kresse and Hafner, Reference Kresse and Hafner1993, Reference Kresse and Hafner1994; Kresse and Furthmüller, Reference Kresse and Furthmüller1996a, Reference Kresse and Furthmüller1996b). Atomic cores were described using projected augmented wave (PAW) potentials (Blöchl, Reference Blöchl1994; Kresse and Joubert, Reference Kresse and Joubert1999). Valence electrons were described in terms of a plane-wave basis set.
Optimization result with the fixed unit cell was used as the starting geometry, and the sources of bond and angle restraints in the Rietveld refinement (Rietveld, Reference Rietveld1967) were used for synchrotron powder XRD data. Atomic coordinates were taken from the PBE-PAW optimized model and refined with the Topas 5.0 software. Anisotropic displacement parameters were refined equally for three types of carbon atoms, all chlorine, and all nitrogen. The positions of the hydrogen atoms were calculated geometrically and refined in the riding model with U iso(H) = 1.5U iso(C) for methyl groups and 1.2U iso(X) for other atoms. The final R-values are listed in Table I along with the corresponding values of Rietveld refinement results. Average/maximal atomic deviations of models I, II, III, and IV from the optimized model were equal to 0.06/0.11, 0.08/0.15, 0.06/0.12, and 0.08/0.15 Å. The comparison of the optimized model and model I (150 mm, 100 K) indicates that the majority of deviations corresponds to positions of methyl groups (Supplementary Figure S1). The Rietveld plot recorded at room temperature (RT) at 250 mm distance is given in Figure 1, and the other plots are given in Supplementary Material (Supplementary Figure S2).
TABLE I. Crystal data and structure refinement for rilpivirine hydrochloride.
Figure 1. XRD pattern IV. Red and blue lines correspond to the calculated profile and experimental patterns, respectively. The bottom trace shows the difference curve.
B. Computational methods
The plane-wave calculations were carried out in the VASP 5.4.1 program package (Kresse and Hafner, Reference Kresse and Hafner1993, Reference Kresse and Hafner1994; Kresse and Furthmüller, Reference Kresse and Furthmüller1996a, Reference Kresse and Furthmüller1996b). The PBE method (Perdew et al., Reference Perdew, Kurth, Zupan and Blaha1999) and PAW (Kresse and Joubert, Reference Kresse and Joubert1999; Kresse and Hafner, Reference Kresse and Hafner2000) were used. Valence electrons (2s and 2p for O, N, and C atoms; 1s for H) were described in terms of a plane-wave basis set. The kinetic energy cutoff for the wave functions was set to 800 eV. Automatic k-point sampling was used. The total energy and force convergence thresholds were set to 10−6 and 10−4 eV, respectively. The crystal structures from the powder X-ray diffraction experiment were used as the starting geometry for the calculations. The geometry optimizations were carried out with the lattice parameters fixed at their crystallographic values. The output files with the relaxed geometries were converted to a res-file using VESTA (Momma and Izumi, Reference Momma and Izumi2011).
III. RESULTS AND DISCUSSION
A. Details of structure refinement
Previously reported polymorphs of rilpivirine hydrochloride were characterized using laboratory diffractometers with CuKα radiation at RT. None of the reported patterns was indexed (Rendell et al., Reference Rendell, Mittelman, Erlich and Ratkaj2012; Hotter et al., Reference Hotter, Pichler, Adamer and Griesser2013; Sandoz, Reference Sandoz2013; Kommavarapu et al., Reference Kommavarapu, Maruthapillai, Palanisamy and Sunkara2015). For our structure obtained from RT data (sample IV), theoretically calculated pattern for CuKα radiation comprises characteristic peaks at 2θ = 9.1, 9.6, 11.0, 12.0, 13.7, 14.7, and 15.1° (Supplementary Figure S3), which correspond to the anhydrous rilpivirine hydrochloride described by Kommavarapu et al. (Reference Kommavarapu, Maruthapillai, Palanisamy and Sunkara2015). IR spectrum for this sample contains weak bands, which can be associated with secondary amine stretching vibrations observed at 3477 and 3404 cm−1. Besides, two strong bands at 1599 and 1516 cm−1 appear in the spectrum, which correspond to NH+ bending vibrations.
Based on the powder XRD pattern described at US2010189796 patent (Stokbroekx, Reference Stokbroekx2010), Kaduk et al. (Reference Kaduk, Zhong and Blanton2015) reported the crystal structure of Form II of free base rilpivirine, which is characterized by characteristic bands in IR spectra at 3316 (ν(R2NH) group), 2223 (ν(C≡N)), 2215 (ν(C≡N)), 1483 (aromatic ν(C–C)), and 1325 (δ(=CH)) cm−1. For our salt, bathochromic shift of the ν(R2NH) band denotes weakening of the N–H⋯N hydrogen bond in rilpivirine as compared with the N–H⋯Cl bonds in rilpivirine hydrochloride.
As it follows from Table I, at a given detector distance, the quality of refinement and convergence factors are almost independent on the temperature of the experiment. Thus, ambient condition powder data are more valuable for phase identification, while low-temperature data are more helpful for structure solution and refinement. In the latter cases, thermal motion is lower that simplifies the determination of molecular conformation and packing. Isotropic thermal parameters of atoms are listed in Table II. Indeed, the temperature increase is accompanied by an increase in U iso values from 0.001 to 0.024 Å2. The distance between the sample and the detector in our case strongly affects convergence factors and standard deviations. For the 250 mm distance, deviations for cell parameters are lower due to more sharp peaks at low angles (Supplementary Figure S2). At the same time, all convergence factors (R int, R wp, R p, R-Bragg) at this distance are higher than those for data obtained at the 150 mm distance. This fact is related to a lower number of observed reflections (2624 vs. 956 reflections for 150 and 250 mm). However, for the 150 mm distance, the pattern becomes nearly linear at 2θ > 30°; and difference between peaks becomes negligible. Although this fact allows to decrease all R-factors, it, in fact, does not affect the overall quality of refinement. Finally, more Chebychev coefficients are required for the refinement of background (8th order for I vs. 5th order for III). Thus, despite higher convergence factors, structure solution and the refinement of small organic molecules from synchrotron radiation obtained at the NIC “Kurchatov Institute” should be carried out using the 250 mm distance between a sample and the detector. Note that, for large-volume unit cell samples, the sample-to-detector distance should likely be decreased.
TABLE II. Isotropic thermal parameters of atoms (Å 2).
B. Molecular and crystal structures of rilpivirine hydrochloride
The asymmetric unit of this salt contains one protonated rilpivirine molecule and a chloride anion (Figure 2). The protonated atom of the cation was revealed from DFT calculations, 15N NMR solid-state spectrum, and the most probable H-bonding system as well. In accord with these data, the pyrimidine ring is protonated, so that the H(N) atom is able to take part in N–H⋯Cl bonding, while the second nitrogen atom of the heterocycle is involved in intramolecular C–H⋯N bonding. This bonding supports the delocalization of the lone pair of the amine N23 nitrogen. This is indicated by the similarity in C–N23 distances (1.361 and 1.366 Å), and N6–C5–N23–C24 and C5–N23–C24–C25 torsion angles (12.1 and 11.6°). Instead, the lone pair of the N10 amine nitrogen is probably delocalized only with heterocycle. The delocalization manifests itself through asymmetry of bond lengths (1.283 Å for C9–N10 and 1.471 Å for C11–N10) and torsion angles (2.6 and 90.0° for C11–N10–C9–C8 and C9–N10–C11–C12). Overall molecular conformation is folded; it is close to that of free base rilpivirine (Kaduk et al., Reference Kaduk, Zhong and Blanton2015) (Figure 3), with the exception of positions of substituted phenyl rings and the cyanovinyl group due to rotation along single C–N amine and C–C bonds.
Figure 2. Asymmetric unit of rilpivirine hydrochloride.
Figure 3. Molecular conformations of rilpivirine in pure solid (green) and rilpivirine hydrochloride (red).
Each molecule takes part in H-bonding with two chloride anions by N–H⋯Cl interactions. The parameters of these bonds for the theoretical model are listed in Table III. The anions and cations are connected by these bonds into infinite chains parallel to the crystallographic c axis (Figure 4). Besides, the anions are involved in weak C–H⋯Cl bonding that connects the chains into a 3D framework. These are two C–H⋯Cl bonds with one molecule, which support the folded conformation of the cation (Figure 2) and a bond with a hydrogen atom of the heterocycle of another molecule.
TABLE III. Parameters of H-bonds in the theoretical model of rilpivirine hydrochloride.
Figure 4. Fragment of N–H⋯Cl-bonded chains in rilpivirine hydrochloride. H-bonds are depicted with dotted lines.
C. Solid-state 15N and 13C NMR spectroscopy
Full analysis of solid-state NMR 13C and 15N chemical shifts has been undertaken. 35Cl NMR has also been recorded (Supplementary Figure S4), which confirmed the presence of the chloride anion in the solid phase; however, full analysis of chemical shifts has not been carried out yet. Rilpivirine hydrochloride in the solid state is represented by one polymorph with the ordered structure as follows from sharp peaks in 13C and 15N NMR spectra (Figures 5 and 6). 2D solid-state 13C-1H HETCOR NMR experiments with varied contact times were applied to investigate 13C–1H interactions (Figure 7). Solid-state NMR spectra and NMR 13C and 15N spectra in solution were also compared. In general, solid-state spectra correspond well with those in solution, although some chemical signals are significantly shifted. Besides, symmetrical signals corresponding to benzene rings due to their free rotation in solution are affected by the asymmetrical environment in solids; the signals corresponding to ortho- and meta-positions become double. 15N SSNMR spectra are valuable for the analysis of N⋯H hydrogen bonds in the solid. It is clearly seen that the N6 atom of a pyrimidine ring was protonated: the corresponding signal is upfield shifted compared to the solution more than signals from other atoms. Besides, in solution, evidence of protonation is less pronounced probably due to a dynamic equilibrium.
Figure 5. 13C CP MAS (13 kHz) solid-state NMR spectrum of powder Rilpivirin-HCl with full signal assignment, and comparison with 13C spectrum in DMSO solution (top). Atomic numeration is given in Figure 2.
Figure 6. 15N CP MAS (9.5 kHz) solid-state NMR spectrum of powder Rilpivirin-HCl with full signal assignment, and comparison with 15N spectrum in DMSO solution (reconstructed from 2D HSQC and HMBC) (top). Atomic numeration is given in Figure 2.
Figure 7. 2D 13C-1H FSLG HETCOR solid-state NMR spectrum (13 kHz MAS) used for signal analysis and assignment. Atomic numeration is given in Figure 2.
Besides the analysis of experimental solid-state NMR spectra, the quantum chemical calculations of chemical shifts were carried out. To calculate the values of chemical shifts, we used optimized geometry of the rilpivirin hydrochloride (theoretical model). The ability to calculate chemical shifts is implemented in the VASP 5.4.1 code; however, the usage of PAW potentials to account for core electrons can be considered as a potential source of errors. To bring the calculated values of chemical shifts into the same scale as experimental ones, we used values of isotropic chemical shifts of 15N and 13C calculated for preliminary optimized crystal structures of ammonia and orthorhombic SiMe4. The results of chemical shift calculations are summarized in Supplementary Table S3 and Figure S4. The differences in experimental and calculated values do not exceed 18.0 and 12.1 ppm for 15N and 13C chemical shifts. The most pronounced disagreement between the experiment and quantum chemical calculations is observed in the case of cyano groups and carbon atoms surrounded by acceptor nitrogen atoms (C5 and C9). In general, the presence of a nitrogen atom bonded to C1 leads to an increase in difference between experiment and calculated values of chemical shifts. On the other hand, the chemical shifts for methyl groups are reproduced within 3 ppm. For the most of carbon atoms of phenyl groups, the differences are as low as 10 ppm. The observable disagreement for the most part can be corrected using simple linear regression (Supplementary Figure S5). After the differences between experimental and calculated chemical shifts became satisfactory. As a result, we can conclude that the solid-state NMR study of rilpivirine hydrochloride is in agreement with the structure derived from powder X-ray data.
IV. CONCLUSION
The molecular and crystal structure of rilpivirine hydrochloride, which is used as an active pharmaceutical ingredient in an anti-HIV medication, has been studied at both RT and low temperature using synchrotron radiation and powder XRD data. No phase transition was observed. The solution is in accord with solid-state 15N and 13C NMR data. The analysis of convergence factors and standard deviations for different sample–detector distances revealed that powder patterns of small organic molecules from synchrotron radiation obtained at the NIC “Kurchatov Institute” should be obtained using the 250 mm distance between a sample and the detector.
V. DEPOSITED DATA
Four powder patterns of the title compound from synchrotron data obtained at 100 and 298 K at two detector distances (150 and 250 mm) have been submitted to the ICDD for inclusion in the Powder Diffraction File. The CIF files (CCDC 2303885-2303888) can also be obtained at www.ccdc.cam.ac.uk/structures.
Crystallographic Data Files were deposited with ICDD. Files can be requested by contacting the Managing Editor at [email protected].
SUPPLEMENTARY MATERIAL
The supplementary material for this article can be found at https://doi.org/10.1017/S0885715624000228.
ACKNOWLEDGEMENTS
This research was funded by the Russian Science Foundation (grant no. 23-73-00027). Solid-state NMR experiments were performed using the solid-state NMR facility at the Department of Structural Studies of N. D. Zelinsky Institute of Organic Chemistry of the RAS, Moscow (R.N.).
CONFLICTS OF INTEREST
The authors declare no conflicts of interest.
