Assessing the risk of salt disproportionation using crystal structure and

crystal (Specac Ltd., Orpington, UK) was used. The sample was kept in contact with the diamond crystal of the ATR accessory, and 64 scans were accumul...
0 downloads 0 Views 12MB Size
Article Cite This: Cryst. Growth Des. XXXX, XXX, XXX−XXX

pubs.acs.org/crystal

Assessing the Risk of Salt Disproportionation Using Crystal Structure and Surface Topography Analysis Mitulkumar A. Patel,† Suman Luthra,‡ Sheri L. Shamblin,§ Kapildev K. Arora,§ Joseph F. Krzyzaniak,§ and Lynne S. Taylor*,† †

Crystal Growth & Design Downloaded from pubs.acs.org by UNIV OF SUNDERLAND on 10/22/18. For personal use only.

Department of Industrial and Physical Pharmacy, College of Pharmacy, Purdue University, West Lafayette, Indiana 47907, United States ‡ Pfizer Inc, Worldwide Research and Development, Cambridge, Massachusetts 02139, United States § Pfizer Inc, Worldwide Research and Development, Groton, Connecticut 06340, United States S Supporting Information *

ABSTRACT: Salt disproportionation is a major issue for pharmaceutical products containing a salt form of a weakly basic drug, because conversion to the free base during processing or storage in the presence of excipients may negatively impact stability and bioperformance of the drug product. Several factors influencing disproportionation tendency have been elucidated; however, a complete mechanistic understanding of this phenomenon is still lacking. Specifically, it is unclear if the crystal structure of the salt plays a role, beyond influencing the salt solubility. Herein, by utilizing model compounds with similar pKa and pHmax values, and hence similar thermodynamic driving forces for disproportionation, we demonstrate that salt crystal structure appears to play a major role in influencing disproportionation tendency. Some salts with low pHmax values were found to be resistant to disproportionation, while other systems transformed to the free base following contact with the basic excipient, magnesium stearate. For salts that converted to the free base, the new crystal phase was imaged using atomic force microscopy and scanning electron microscopy. Based on these images, it was concluded that conversion occurred via a nucleation and growth process; i.e., transformation involved considerable structural rearrangement. Extensive crystallographic analysis provided some insight, suggesting that the packing and intermolecular interactions around the salt bridge may influence the susceptibility of the salt to conversion to the free base when challenged with a basic microenvironment. Thus, evaluation of disproportionation risk requires consideration not only of pHmax, but also salt crystal structure, whereby different solid state forms of a given salt may present different risk levels for conversion to free base. Further studies of correlations between crystal structure and disproportionation tendency are clearly warranted to ensure the development of robust salt formulations.



INTRODUCTION

typically resulting from the presence of certain excipients in the formulations.6 Disproportionation is a key challenge for pharmaceutical development, in particular, for salts of weakly basic compounds which are more prevalent than acidic drugs, and a deeper understanding of important factors impacting this process is required. Disproportionation can result from processing operations, the presence of a particular formulation component, or some combination thereof. Unit operations such as high shear wet granulation and compression have been shown to lead to salt-to-free base conversion.7 For formulations, important factors include excipient basicity as well as the particle size and surface area of the active pharmaceutical ingredient (API).8,9 In addition, environmental factors such as

Formulations containing the salt form of a drug are common. According to the Food and Drug Administration’s (FDA) orange book database, about half of marketed pharmaceutical products are available as the salt form.1 The salt enhances the solubility and bioavailability relative to the parent free form.2 In addition, if the free form is amorphous in nature, salt formation may yield a crystalline form, providing improved purification and handling during pharmaceutical manufacturing.3 The salt form is also desirable when the parent free form is reactive or undergoes rapid degradation; the salt form often has superior solid-state properties relative to the free form, leading to enhanced stability.4,5 Two major challenges with salt formulations are deliquescence and disproportionation.6 The former occurs when the salt is highly hygroscopic and liquification occurs at or above a critical relative humidity. Disproportionation, where the salt converts to the free form, occurs when there is a change in microenvironmental pH, © XXXX American Chemical Society

Received: August 7, 2018 Revised: September 21, 2018 Published: October 5, 2018 A

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

temperature and relative humidity (RH) are known to impact disproportionation.8,9 Two physicochemical properties often used to evaluate the risk of disproportionation are the pKa and the pHmax of the compound. The ionized form is stable below the pHmax, whereas above this value, the neutral form is the stable form.10 Thus, for the salt of a weakly basic API, any excipient generating a microenvironmental pH greater than the pHmax can theoretically cause disproportionation. Further, the pHmax of the API is inversely proportional to the solubility of the corresponding salt.11,12 Hence, a salt with higher solubility enhances the risk of disproportionation, as compared to the salt form with lower solubility.10,11 The solid-state form of a given salt is not typically considered when evaluating disproportionation tendency. Recently, it has been suggested that the solid-state properties of the API salt may play a role in governing disproportionation. Amorphous miconazole mesylate was shown to be highly prone to disproportionation as compared to the crystalline anhydrous form, while the dihydrate salt did not show any disproportionation.13 It was suggested that the molecular packing in the dihydrate salt hindered disproportionation. This study thus indicated that there might be a correlation between crystal structure and susceptibility to disproportionation. However, alternative explanations are also possible. For example, crystalline salts that appear to be resistant to disproportionation could, in principle, have reacted to form a very thin product layer at the surface, which then creates a barrier hindering further reaction. If this product layer constitutes a mass fraction lower than the analytical detection limit, which is likely if bulk analysis techniques are employed, then an incorrect conclusion could be drawn. The formation of a barrier layer has been reported for other systems during phase transformations.14,15 Consequently, it is important to determine if a given salt crystal structure is truly resistant to disproportionation, or if the extent of reaction falls below detection limits. Moreover, the interface between the reacting solids should be carefully evaluated for signs of reaction. In this study, the disproportionation tendency of several chemically and structurally diverse salts of model weakly basic compounds was evaluated using bulk and surface analytical techniques. Molecular packing was evaluated using single crystal structures in order to provide insight into possible structural features that might impact disproportionation tendency. Because previous research indicated that miconazole mesylate dihydrate is resistant to disproportionation while the anhydrate and amorphous forms are susceptible, both hydrates and nonsolvated salts were evaluated.13 The following compounds were evaluated for disproportionation tendency in the presence of a basic excipient, magnesium stearate: pioglitazone HCl, ziprasidone HCl, ziprasidone mesylate, sorafenib tosylate, and atazanavir sulfate. Each of these salts has a low pHmax (Table 1) and hence, theoretically, a high propensity to undergo disproportionation. Powder blends were evaluated using Raman spectroscopy to determine bulk disproportionation. Atomic force microscopy (AFM) and scanning electron microscopy (SEM) were employed to study changes occurring at the interface between a salt crystal film and the excipient. A comparison was made between the salt and free base crystal structures to further identify any constraints which might impact the likelihood of disproportionation.

Table 1. pKa and pHmax Values for Various Salts Solid Form

pKa

pHmax

PIOH ZIPH ZIPM ATZS SORT

5.2a 6.5c 6.5c 4.7e 4.9g

3.00b 3.42d 2.44d 2.25f 3.27h

a

pKa of PIO was obtained from literature.26 bObtained from previous literature.27 cpKa of ZIP was obtained from literature.28 dObtained using the experimentally determined solubility value of 0.000078 mg/ mL, 0.09431 mg/mL, and 0.899 mg/mL for ZIP free base, ZIPH, and ZIPM, respectively. epKa of ATZ was obtained from literature.29 f Obtained using the experimentally determined solubility value of 0.0057 and 1.60 mg/mL for ATZ free base and ATZS, respectively. g pKa of SOR was obtained from literature.26 hSolubility of SOR free base (0.0000359 mg/mL) and SORT (0.00152 mg/mL) were obtained from the literature.26



EXPERIMENTAL SECTION



METHODS

Materials. PIOH was obtained from the Tokyo Chemical Industry Co, Ltd. (Portland, OR). Magnesium stearate (MgSt) was supplied by Brand Nu Laboratories (Meriden, CT). Sodium hydroxide (NaOH) and hydrochloric acid (HCl) were purchased from Fisher Scientific (Pittsburgh, PA). Ziprasidone mesylate (ZIPM), ziprasidone hydrochloride (ZIPH), and ziprasidone (ZIP) free base were procured from Pfizer (Groton, CT). Atazanavir sulfate (ATZS) and sorafenib tosylate (SORT) were obtained from ChemShuttle (Hayward, CA), while atazanavir (ATZ) and sorafenib (SOR) free base were obtained from Attix Pharmaceutical (Toronto, Ontario, Canada). All chemicals were >99% pure. Ultrapure water was obtained using a Millipore ultrapure water system (Billerica, MA). The molecular structures of model API salts and the excipients are shown in Figure. 1. Amorphous ATZS was prepared by rapid solvent evaporation. Briefly, a concentrated methanolic solution of ATZS (150 mg/mL) was poured onto a glass dish and solvent was rapidly evaporated using dry nitrogen gas. The obtained glassy material was removed using a spatula and stored in a sealed vial. To obtain pioglitazone (PIO) free base, PIOH was slurried in 1 M NaOH solution with vigorous stirring for 3 h, followed by vacuum filtration, washing with water, and air drying. The resulting powder was collected and stored in an airtight container.

Preparation of Binary Mixtures for Bulk Disproportionation Study. Each crystalline salt form (PIOH, ZIPH, ZIPM, SORT, ATZS) was mixed at a 50% (w/w) ratio with the model excipient, MgSt. 50 mg of each salt form was geometrically mixed with 50 mg of MgSt using a spatula. The mixtures were prepared in triplicate. All samples were then stored in open vials at 75% RH at 40 °C, except ATZS samples, which were stored at 85% RH at 40 °C. Pure salts (without MgSt) were also stored at the same conditions. Raman Spectroscopy. Raman spectroscopy was used to monitor disproportionation. A RamanRxn2 Analyzer (Kaiser Optical Systems, Inc., Ann Arbor, MI) with a PhAT probe, as described previously,13 was used for analysis. For the light source, a 785 nm high power NIR diode laser (Invictus) was used. A laser spot size of 6 mm was used and back scattered radiation was collected and delivered to the base unit via the fiber bundle of the PhAT probe. The instrument was calibrated as described previously.13 A specially designed sample holder was used to obtain reproducible results. The CCD (charge coupled device) detector was cooled to −40 °C to reduce the offset in the baseline and all spectra were corrected by a dark subtraction. To generate each spectrum, 6 scans were accumulated at an exposure time of 10 s each with a laser power of 400 mW. Preparation of Salt Solutions for Spin Coating to Produce Crystal Films. Solutions of ZIPH (12 mg/mL) and ZIPM (25 mg/ mL) were prepared by dissolving the appropriate amount of salt in a B

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 1. Chemical structures of PIOH, ZIPH, ZIPM, ATZS, SORT, and MgSt. MeOH:DCM (1:1) mixture at 40 °C. The PIOH solution (150 mg/ mL) was prepared in MeOH at 40 °C, whereas the SORT solution (11 mg/mL) was prepared in a MeOH:H2O (5:1) mixture at 55 °C. A solution of ATZS (150 mg/mL) was prepared in MeOH at 25 °C. Preparation of MgSt Pellet. The MgSt pellets were prepared using a Specac IR press (Grasby Specac, Kent, UK). About 200 mg of MgSt powder was added to a die of 10 mm diameter followed by addition of about 50 mg of ethyl cellulose, which helped in removal of the MgSt pellet from the punch. The material was then compressed at 700 psi for 3 s to obtain the bilayer pellet. The MgSt surface was contacted with the crystal films for the interfacial disproportionation investigations. Film Preparation for Disproportionation Study. Films from different salt solutions were prepared using a KW-4A spin coater (Chemat Technology Inc., Northridge, CA). Silicon wafers, thallium bromoiodide (KRS-5), and glass substrates were used to prepare films for atomic force microscopy (AFM), Fourier transform infrared (FTIR) spectroscopy, and polarized light microscopy (PLM) analysis, respectively. Silicon wafers with a mirror finish were cut into pieces of approximately 1 cm2 in size and were taped on steel stubs for preparing films for AFM, scanning electron microscopy (SEM), and energy-dispersive X-ray spectroscopy (EDX) analysis. KRS-5 and glass slides were also cut into pieces of around 1 cm2, which were used for FTIR and PLM analysis, respectively. Twenty-five microliters of salt solution was placed on the substrate which was subsequently spun for 6 s at 50 rpm, followed by 30 s at 1500 rpm. The resultant films were stored at 40 °C/75% RH for 2 days to obtain crystalline salt films, with the exception of ATZS. To obtain an amorphous film of ATZS, the films were stored at 40 °C/85% RH, whereas to obtain a crystalline ATZS film, films were stored in enclosed chamber saturated with an ACT:IPA (1:1) solvent mixture at 40 °C for 2 days followed by storage at 40 °C/85% RH for 2 days. MgSt Pellet-Salt Film Disproportionation Study. To observe the disproportionation of different salt films, each film prepared on either silicon, KRS-5, or glass substrates, was kept in contact with a MgSt pellet at 40 °C/75% RH (with the exception of ATZS films which were stored at 40 °C/85% RH). Contact was achieved by simply placing the MgSt pellet onto the salt film. The films on silicon, KRS-5, and glass substrates that were each contacted with the MgSt pellet were then used for AFM, FTIR spectroscopy, and PLM analysis, respectively, with analysis taking place after removal of the MgSt pellet. Atomic Force Microscopy (AFM). Topographical images of film samples were obtained using a nanoIR2 AFM-IR instrument (Anasys Instruments, Inc., Santa Barbara, CA). Contact mode NIR2 probes (Model: PR-EX-NIR2, Anasys Instruments, Santa Barbara, CA) were used. To obtain AFM topographical images by contact mode, an x

and y resolution of 256 points and a scan rate of 0.5 kHz was used. To collect and analyze these images, the Analysis Studio software (v 3.10.5539, Anasys Instruments, Santa Barbara, CA) was used. The thickness of the films were determined by making a scratch on the film with a razor blade followed by an AFM scan across the scratch. Nanoscale IR (nano-IR) Spectroscopy. To obtain localized midIR spectra, a nanoIR2 AFM-IR instrument (Anasys Instruments, Inc., Santa Barbara, CA) was used in contact mode, with NIR2 probes (Model: PR-EX-NIR2, Anasys Instruments, Inc., Santa Barbara, CA). In this technique, the nanoscale resolution of an AFM is coupled with a tunable pulsed IR laser source to obtain site-specific chemical information. When the wavelength of the incident light corresponds to that of the absorption of the specific chemical entity in the sample, energy is absorbed. This process results in a localized thermal expansion, causing the AFM cantilever in contact with the sample to oscillate. The amplitude of oscillations has a characteristic decay pattern, which can be recorded as a function of time, known as a “ringdown”. To record this “ringdown” signal, a second harmonic mode of cantilever oscillation with a frequency center of 200 kHz and a frequency window of 30 kHz was selected. Fourier transformation techniques can then be applied to obtain the information about amplitudes and frequencies of oscillations. The “ringdown” is directly proportional to the absorption of the sample and an IR absorption spectrum is generated by measuring this “ringdown” over a wide range of wavenumbers.16 The IR optimization, background calibration, and data collection for AFM-IR spectra were performed over 1600−1800 cm−1 with a coaverage of 256 scans, at a data point spacing of 4 cm−1, as described previously.17 All the spectra were smoothed using a Savitzky-Golay function with a polynomial order of 3 and a side point of 5. Polarized Light Microscopy (PLM). Films prepared on a glass substrate were utilized for PLM analysis. Glass substrates with a salt film were mounted onto the stage of the Eclipse E600 POL microscope. The images were obtained by microscope equipped with a DS Fi1 camera (Nikon Corporation, Tokyo, Japan) using crosspolarized light in transmittance mode as described previously.18,19 Bulk Fourier Transform Infrared Spectroscopy (FTIR). FTIR spectra of the various films were collected using a Bruker Vertex 70 instrument (Bruker Co., Billerica, MA) equipped with a transmittance sample holder. OPUS software (Bruker Co., Billerica, MA) was used for data collection and analysis. FTIR spectra were generated by placing film samples prepared on KRS-5 substrates into the transmittance accessory, followed by accumulating 64 scans at a resolution of 4 cm−1. Similarly, to obtain the FTIR spectra of bulk powders, a Golden Gate attenuated total reflectance (ATR) sampling accessory with a diamond crystal (Specac Ltd., Orpington, UK) was used. The sample was kept in contact with the diamond crystal of the C

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 2. Raman spectra for samples corresponding to (A) PIOH, (B) ZIPH, (C) ZIPM, (D) ATZS, (E) ATZS amorphous, and (F) SORT. Where (a) pure salt, (b) salt + MgSt physical mixture, (c) salt + MgSt stored at accelerated condition (at 40 °C/75% RH, except for ATZS samples which were stored at 40 °C/85% RH) for 2 days, (d) salt + MgSt stored at accelerated condition (at 40 °C/75% RH, except for ATZS samples which were stored at 40 °C/85% RH) for 14 days, (e) respective pure free base (downward arrow indicates decrease in salt peak for PIOH, while upward arrow indicates appearance of free base peak in amorphous ATZS). Cressington sputter coater (208HR, Cressington Scientific Instruments, Watford, UK). The SEM images were acquired first and then the EDX analysis was performed at an accelerating voltage of 10 kV to collect elemental data. The resulting images were analyzed using an Oxford Inca system (Oxford Instruments Analytical, Buckinghamshire, UK). Each experiment was performed in triplicate to determine an average value for elements in the samples.

ATR accessory, and 64 scans were accumulated at a resolution of 4 cm−1. Energy Dispersive X-ray Spectroscopy (EDX) Coupled with Scanning Electron Microscopy (SEM). The surface texture and elemental distribution of contacted and uncontacted regions of the films were determined using a field emission SEM (Quanta SEM, FEI Co., Hillsboro, OR, USA) equipped with an EDX detector.20 The samples were first coated with platinum (∼2 nm thickness) using a D

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

powder pattern.24 This indicates successful formation of PIO free base (Figure S1). The XRPD patterns (Figure S1) indicated that PIOH exists as pure polymorphic Form I with no detectable free base. Similarly, XRPD analysis was carried out for all other samples (Figures S1 and S2). The Cambridge Structural Database (CSD) was used to identify the solid state form of the salts and free bases used in this study. ZIPH was found to be present as the monohydrate form (Figure S1), while ATZS and SORT were identified as anhydrous forms (Figure S2). The crystal structure of ZIPM and the free base forms of ATZ and PIO were not available in the CSD. Therefore, single crystals of these material were produced and analyzed by single crystal X-ray diffraction to obtain their structure (described in detail in the Supporting Information). The data indicated that ZIPM is a trihydrate while ATZ and PIO free base were anhydrous (Figures S1 and S2). XRPD patterns of all salts were then compared with their respective free bases. The results indicated that all salts were present in pure form, without any detectable free base (Figures S1 and S2). To observe the effect of amorphization on the disproportionation reaction, amorphous ATZS was prepared. XRPD analysis indicated an absence of any diffraction peaks in the amorphous ATZS material (Figure S3). The neat amorphous form of ATZS was found to remain stable at the experimental conditions employed (Figure S3). pHmax Determination. Previous research indicates that the pHmax is a critical factor which governs the rate and extent of disproportionation.10,25 Therefore, it was essential to evalute pHmax values for the salts used in this study. To calculate the pHmax values, literature pKa, and solubility values were used where available. For ZIP free base, ZIPH, ZIPM, ATZ free base, and ATZS, the solubility data was obtained experimentally. XRPD analysis of the residual solid after solubility analysis indicated no change in solid form, which suggests that the solubility of the desired solid-state form was obtained (Figure S4). The pKa and solubility values were utilized in eq 1 to calculate the pHmax values, which are reported in Table 1. Disproportionation in Binary Blends with MgSt. To study which model salts are prone to disproportionation, a binary blend of each salt with the basic excipient, MgSt, was prepared. These blends were stored at accelerated stability conditions and Raman spectroscopy was used to determine the event of disproportionation, i.e., the appearance of free base peaks. Specific spectral regions as described in Table S1 were used to identify the salt-to-free base conversion. Among the samples, PIOH showed disproportionation as noted by a decrease in the intensity of the characteristic salt peak at ∼1638 cm−1 (Figure 2A), which is consistent with previous observations.30 In addition, the amorphous form of ATZS was also found to disproportionate when stored with MgSt, as indicated by formation of ATZ free base peaks. On the other hand, ZIPH, ZIPM, ATZS crystalline, and SORT salts did not show any disproportionation (Figure 2), even after 14 days of storage at accelerated conditions with MgSt. XRPD analysis of these salt-MgSt blends after storage showed an absence of free base peaks (Figure S5), consistent with the Raman analysis. In contrast, PIOH and amorphous ATZS blends showed the presence of characteristic free base diffraction peaks (Figure S5). Furthermore, the control, neat API salt powders (without MgSt) did not show any disproportionation, with the exception of PIOH, which showed slight disproportionation after 14 days, consistent with previous observations.30

X-ray Powder Diffraction (XRPD). Diffraction patterns of various samples were collected using a Rigaku SmartLab (XRD 6000) diffractometer (The Woodlands, TX) in Bragg−Brentano mode. Cu Kα radiation (λ = 1.5405 Å) was used. About 40 mg of each sample was placed onto a glass sample holder. Diffraction data were recorded at room temperature between 10° and 40° 2θ with a step size of 0.02° with a scan time of about 8 min per sample. Solubility Determination. The slurry method was used to determine the solubility of salt and/or free base with samples measured in triplicate.21 An excess amount of solid was slurried in water for 48 h at 25 °C. An Optima L-100 XP ultracentrifuge (Beckman Coulter, Inc., Brea, CA) was then used at 35,000 rpm for 25 min to pellet the excess solid. An Infinity Series high performance liquid chromatography (HPLC) system (Agilent 1290, Agilent Technologies, Santa Clara, CA) was used to determine the concentration in the supernatant. An Inertsil ODS-3 C18 column with dimensions of 5 μm, 4.6 × 100 mm (GL Sciences Inc., Rolling Hills Estates, CA) was used. The mobile phase was prepared by mixing water and acetonitrile (1:1, v/v) along with 0.1% trifluoroacetic acid. A flow rate of 1 mL/min and an injection volume of 10 μL were used. An ultraviolet (UV) detector set at 220 nm was used to monitor the elution of the compound. A calibration curve (R2 value of 0.995) was applied to determine the concentration of each sample. Dilution with mobile phase was performed to obtain concentration values within the range of the calibration plot. Determination of pHmax. The pHmax values for different salt forms were obtained using the following equation.11,12

pH max = pK a + log

S0 K sp

(1)

where the pKa is related to the acid dissociation constant (Ka) of the protonated drug, S0 is the intrinsic solubility of the free base, and Ksp is the solubility product for the salt. Single Crystal X-ray Diffraction (SCXD). To obtain the single crystals, saturated solutions of ZM trihydrate in water, pioglitazone in ethanol, and atazanavir in IPA were prepared at 40 °C. The supernatant was filtered using a syringe filter (PTFE, 0.45 μm). The resulting solutions were allowed to slowly evaporate at room temperature until crystallization was observed. A single crystal of each compound, suitable for SCXD, was mounted on a Mitegen loop or micromesh followed by placement on the stage. The SCXD data were collected utilizing a Bruker Quest Instrument along with a Photon2 CMOS area detector. The instrument was equipped with a MicroMax002+ high-intensity copper X-ray source (Cu Kα radiation, λ = 1.54184 Å) in combination with confocal optics. The diffraction data were obtained at −173 °C. Detailed information regarding single crystal data processing is provided in the Supporting Information. Crystal Structure Analysis. The crystal structure of various salts and free bases were analyzed using Mercury software (v 3.8) from the Cambridge Crystallographic Data Centre (Cambridge, United Kingdom). A change in the molecular conformation of the API in their salt or free base form was evaluated by overlaying a portion of the molecule and comparing the difference in the rest of the molecular conformation, as described previously.22 The unit cells of various crystal structures were obtained from the respective single crystal structures. To determine the presence of any voids in the crystal structure, the single crystal structures of various salts were analyzed using the probe radius analysis function in the Mercury software, as described previously.18 As water molecules have a probe radius of 1.4 Å, the structures of various salts were searched for the presence of any voids with a probe radius of ≥1.4 Å.23



RESULTS Solid Form Identification and Single Crystal Structure Determination. To analyze the polymorphic form and phase purity of the various model salts, XRPD analysis was performed. The PIO free base obtained from the PIOH, showed an XRPD pattern well-matched to the reference E

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 3. FTIR spectra of samples corresponding to (A) PIOH, (B) ZIPH, (C) ZIPM, (D) ATZS, (E) ATZS amorphous, and (F) SORT. (a) pure free base, (b) pure salt, (c) uncontacted with MgSt, and (d) contacted with MgSt for 2 days. Arrows indicate appearance of characteristics peaks of free base formation.

The salt films were prepared by spin coating on various substrates and were then pretreated using various conditions to obtain the desired solid-state form of the salt. As the films obtained by spin coating were too thin to perform XRPD, thicker samples were also produced using similar experimental conditions, and the obtained material was then analyzed by XRPD. In addition, FTIR was used to compare the solid-state form of salt films obtained on the KRS-5 substrates. The FTIR spectra of each salt film were comparable to spectra obtained

Preparation and Characterization of Salt Films. Films of various API salts were prepared to study correlations between the change in salt crystalline surface topography/ chemistry and disproportionation. To identify the solid form of salt being generated in the film as well as to analyze the occurrence or absence of disproportionation, films were prepared on different substrates. Glass, KRS-5, and silicon wafers substrates were used to prepare films for characterization with PLM, FTIR, and AFM analysis, respectively. F

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 4. PLM images of samples for (A) PIOH, (B) ZIPH, (C) ZIPM, (D) ATZS, (E) ATZS amorphous, and (F) SORT in the uncontacted and MgSt-contacted regions.

Figure 5. AFM (deflection) images of samples for (A) PIOH, (B) ZIPH, (C) ZIPM, (D) ATZS, (E) ATZS amorphous, and (F) SORT in the uncontacted and contacted regions.

After 2 days of storage, the pioglitazone film spectrum is very similar to that of the pure base, indicating that this sample appears to have largely converted to the free base. The amorphous atazanavir film is not a perfect match with the reference free base, suggesting that disproportionation is not complete in this instance. Having established which films underwent disproportionation and which were apparently resistant based on FTIR data, changes in the surface morphology and topography of film following contact with MgSt were evaluated. First, to identify any gross visual change in the salt films, PLM was performed, taking images of noncontacted and MgSt-contacted regions after 2 days of contact. Except for the ATZS amorphous film, all other films showed clear evidence of birefringent crystals

from their respective reference powder samples. Both XRPD (Figure S6) and FTIR (Figure 3A−F) results indicated that desired forms of each salt were obtained in the films, and that the solid form was consistent with the form that was used for binary blend disproportionation studies. FTIR was also used to determine the occurrence of disproportionation in the salt film after contact with a MgSt pellet. Specific spectral regions as described in Table S1 were used to identify the salt-to-free base conversion. ZIPH, ZIPM, crystalline ATZS, and SORT did not show any changes after exposure to MgSt, indicating an absence of disproportionation (Figure 3). On the other hand, the spectra for PIOH and amorphous ATZS films changed after contact with MgSt, and showed peaks corresponds to the free base forms (Figure 3). G

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 6. SEM images of samples for (A) PIOH, (B) ZIPH, (C) ZIPM, (D) ATZS, (E) ATZS amorphous, and (F) SORT in the uncontacted and MgSt contacted regions.

Table 2. Elemental Distribution Obtained by EDX of Various Salt Films PIOH Element C N O S Cl Element C N O S Cl Element C N O S Cl

Uncontacted 74.5 6.4 8.1 6.9 4.1

± 0.3 ± 0.2 ± 0.1 ± 0.1 ± 0.1 ZIPH

Uncontacted 71.4 12.7 6.4 3.2 6.3

± 0.1 ± 0.1 ± 0.2 ± 0.2 ± 0.4 ZIPM

Uncontacted 62.7 12.1 16.9 5.2 3.1

± ± ± ± ±

0.4 0.4 0.2 0.2 0.1

ATZS Contacted

Element

± ± ± ± ±

C N O S

78.1 6.0 7.6 7.8 0.1

0.2 0.1 0.1 0.1 0.1

Uncontacted 70.1 11.7 16.2 2.0

± ± ± ±

Contacted

0.5 0.3 0.1 0.1

70.7 10.5 16.6 2.2

± ± ± ±

0.4 0.1 0.2 0.1

ATZS amorphous Contacted

Element

± ± ± ± ±

C N O S

71.2 12.9 6.3 3.5 6.1

0.4 0.2 0.1 0.1 0.2

Uncontacted 71.1 11.4 15.6 1.9

± ± ± ±

0.4 0.2 0.2 0.1

Contacted 70.9 10.9 16.7 1.5

± ± ± ±

0.3 0.2 0.2 0.1

SORT Contacted

Element

± ± ± ± ±

C N O Cl F S

62.3 11.8 17.5 5.5 2.9

0.4 0.3 0.3 0.1 0.2

(Figure 4) in the initial, uncontacted films. No change in morphology or birefringence for ZIPH, ZIPM, ATZS, and SORT was observed for regions of the film that had been in contact with MgSt. In contrast, large spherulites initially present in PIOH films were converted to smaller regions of crystalline material with altered birefringence, while contacted ATZS amorphous films showed the appearance of small birefringent regions.

Uncontacted 70.8 9.6 7.8 2.6 6.5 2.7

± ± ± ± ± ±

0.5 0.3 0.2 0.1 0.2 0.2

Contacted 70.4 9.5 7.6 2.8 6.9 2.8

± ± ± ± ± ±

0.4 0.2 0.3 0.2 0.2 0.1

AFM Analysis. To monitor the changes in surface topography of various API salt films after contact with MgSt, AFM analysis was carried out. The PIOH film shows a smooth morphology in the uncontacted region, while a dramatic change in the topography was observed for the MgSt contacted region (Figure 5) after 2 days. The films of ZIPH, ZIPM, crystalline ATZS, and SORT films did not show any difference in their topography for MgSt contacted or uncontacted regions (Figure 5). However, the ATZS amorphous film showed a H

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 7. Crystal structures of various salts showing interaction of API and their respective counterions for (A) PIOH, (B) ZIPH, (C) ZIPM, (D) ATZS, and (E) SORT.

EDX was performed to analyze any change in the chemical profile in MgSt-contacted and uncontacted regions (Table 2). The obtained elemental distribution data are similar to their respective theoretically predicted values based on the molecular formula (Table S2). For PIOH, the uncontacted region has a high chlorine amount, while the contacted region showed a negligible amount of chlorine (Table 2). In contrast, the other films do not show much difference in the elemental distribution between MgSt contacted and uncontacted regions. EDX analysis was also performed on the MgSt pellet. Only the MgSt pellet contacted to PIOH showed the presence of Cl in its elemental distribution (Table S3). All other samples of MgSt pellets that contacted different salt films had elemental distributions similar to that of the uncontacted MgSt pellet (Table S2). The elemental distribution of the uncontacted MgSt pellet showed slightly higher amounts of O and Mg, compared to the theoretically predicted values. This could be due to contamination with MgO. Single Crystal Structure Analysis. To identify the presence or absence of a direct salt bridge in the structure, the single crystal structure was obtained from the CSD or determined experimentally and analyzed using Mercury software (v 3.8). The data show that PIOH, ZIPH, ATZS, and SORT have a direct salt bridge in their crystal structure, while ZIPM does not have a direct salt bridge (Figure 7). Furthermore, in the structure of the ATZS, the salt bridge is shielded by two of the adjacent biphenyl rings, while in the structure of SORT, instead of one salt bridge linking the API and the counterion, there is a dual salt bridge. To evaluate the presence of channels in the crystal structure, which can allow easy access of water to the salt bridge, a void space analysis was carried out. The results of the crystal void space analysis showed that there are no voids greater than 1.4 Å (0% of unit cell volume) in the crystal structure of the various salts, except

change in topography in the MgSt-contacted region compared to the uncontacted region of the film. To identify disproportionation on the surface of the film, AFM coupled nano-IR was used. PIOH and ZIPM films were selected for this study, as in the case of PIOH there is change in the surface texture, while for ZIPM, there is no change. For PIOH, the nano-IR spectra of the uncontacted and the contacted regions correspond to salt and free base form, respectively. In the case of ZIPM, the nano-IR spectra of both the uncontacted and the contacted region correspond to the salt form (Figure S7). SEM/EDX Analysis. To further confirm the presence or absence of disproportionation in the films, SEM/EDX was used as an orthogonal technique. The SEM data showed surface features consistent with polycrystalline films for PIOH, ZIPH, ZIPM, ATZS, and SORT. As seen from the PLM study, the PIOH spherulites were very large and SEM images suggest that they have a smooth surface at high magnification (Figure 6A). The surface of the ZIPH showed rod-like crystals, ZIPM films showed feather-like crystals, and the surface of ATZS films showed needle-like morphology (Figure 6B−D). On the other hand, the surface of the amorphous ATZS film was found to be smooth (Figure 6E) while the SORT film showed discrete crystals (Figure 6F). The morphology of the PIOH and the amorphous ATZS films is different in the contacted versus the uncontacted region. In both cases, small crystals are formed in the portion of the film contacted with MgSt. In contrast, the surface of ZIPH, ZIPM, crystalline ATZS, and SORT did not show any differences in morphology between contacted and uncontacted regions. SEM analysis of the MgSt pellet was also performed. The surface of the pellet found to have large irregularly shaped plate-like structures, which is characteristic of MgSt (Figure S8). There is no morphology change for MgSt before or after contacting with the various salt films. I

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 8. Overlay of API conformation in salt (red) and the corresponding free base (blue) in the single crystal structure of (A) PIOH, (B) ZIPH, (C) ZIPM, (D) ATZS, and (E) SORT and their respective free base form.

Figure 9. Unit cells of salt (red) and free base (blue) for (A) PIOH, (B) ZIPH, (C) ZIPM, (D) ATZS, and (E) SORT and their respective free base forms.

rotation of molecules required for the process of conversion of salt to the free base are presented in Table 4. The data indicated that salt-to-free base conversion in the case of the PIOH required breaking 2 hydrogen bonds while other salts required breaking of 3 or more bonds, with considerable variations observed for hydrogen bond distances.

for ATZS, which showed the presence of 4.4% voids, which are isolated and do not form channels (Figure S9). The conformational differences between the API molecule in the salt versus the free base form were also evaluated from their crystal structures. A slightly larger extent of conformational difference was observed for PIO/PIOH and for ATZ/ ATZS compared to the other structures (Figure 8). To evaluate any differences in the unit cell parameters between salts and free base, the structures of unit cells are presented in Figure 9, while their parameters are presented in Table 3. In addition, information regarding bond breaking and



DISCUSSION Conversion of a salt to the free form is a type of phase transformation, albeit one that involves a reaction (proton transfer) as well as the nucleation and growth of the new J

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Table 3. Unit Cell Parameters for Various Salts and Free Bases Parameter Refcode Space group a b c α β γ Volume Z Z′ Parameter Refcode Space group a b c α β γ Volume Z Z′

PIOH FAYDUF Monoclinic, P21 10.0696 (17) Å 9.4318 (16) Å 10.1752 (19) Å 90° 95.178 (14)° 90° 962.44 (Å3) 2 1 ATZS LUQREW Triclinic, P1 9.861 (5) Å 29.245 (6) Å 8.327 (2) Å 93.56 (2)° 114.77 (3)° 80.49 (3)° 2150.41 (Å3) 2 2

PIO

ZIPH

FAYDUF03 Monoclinic, P21/c 4.6249 (3) Å 18.4928 (10) Å 22.4002 (12) Å 90° 92.016 (3)° 90° 1914.64 (Å3) 4 1

OMAKAQ01 Triclinic, P1 7.250103 (28) Å 10.98666 (8) Å 14.07187 (14) Å 83.4310 (4)° 80.5931 (6)° 87.1437 (6)° 1098 (Å3) 2 1 ATZ

LUQREW01 Monoclinic, P21 15.2650 (18) Å 5.8604 (6) Å 21.365 (2) Å 90° 95.966 (3)° 90° 1900.94 (Å3) 2 0

phase, in this case, the free base. Phase transformation of organic solids is typically considered to occur via one of three different routes, namely, solid-state, solution-mediated, or surface-facilitated phase transformations. Solid-state phase transformations, where true solid−solid transformations occur in the absence of liquid or vapor, have been described in detail by Paul and Curtin,31 as well as Byrn and coworkers,32 and are considered to involve four steps: molecular loosening in the parent compound, formation of an intermediate solid solution, nucleation, and growth of the new phase. Because pharmaceutical solids are typically exposed to water vapor, solution mediated phase transformation routes often have been invoked to describe transformations in powders, as well as in suspensions.33 Solution mediated phase transformations have been extensively discussed by Cardew and Davey and require the initial dissolution of the parent phase into solution, followed by nucleation and growth of the new phase with continued dissolution of the parent phase to replenish the solution concentration.34 More recently, there is emerging evidence that transformations can occur through surface facilitated mechanisms. Using molecular dynamics simulations as well as experimental observations, Gao and Olsen suggested that the metastable form II polymorph of acetaminophen transforms to the stable form I polymorph in aqueous suspension through surface disordering of the metastable form because of hydration, with subsequent ordering to the stable form.35 Warzecha et al. reported the formation of dense nanoclusters of drug on the surface of anhydrous olanzapine in water, which subsequently ordered yielding dihydrate crystals.36 Although both of the aforementioned examples were observed in aqueous solution, Gao and Olsen postulated that a similar surface facilitated mechanism was most likely responsible for transformation of acetaminophen form II to the stable form when exposed to high RH conditions.35

ZIPM OMAKAQ04 Monoclinic, P21/n 13.709 (6) Å 11.595 (4) Å 16.764 (8) Å 90° 103.496 (16)° 90° 2591.15 (Å3) 4 1 SORT AKENUA Monoclinic, P21/c 21.276 (4) Å 9.1160 (17) Å 16.077 (3) Å 90° 108.143 (3)° 90° 2963.14 (Å3) 4 1

ZIP VUJRUO Monoclinic, P21/a 7.377 (8) Å 8.038 (3) Å 33.627 (6) Å 90° 95.60 (4)° 90° 1984.44 (Å3) 4 1 SOR AKENOU Monoclinic, P21/c 8.1587 (16) Å 9.8055 (19) Å 27.758 (5) Å 90° 94.358 (3)° 90° 2214.22 (Å3) 4 1

Currently, salt-to-free base conversion in powders is modeled as a solution mediated phase transformation process that occurs at the surface whereby water associated with the solids is able to mobilize species.37 In this approach, some of the key parameters influencing the extent of phase transformation are the amount of the salt that is dissolved in a hydration layer, which in turn depends on the amount of water in the system as well as the salt solubility, the drug dissociation constant, the free base solubility, and the ability of excipients to increase the microenvironmental pH to values higher than the pHmax of the salt. pHmax defines the pH boundary below which the salt is the stable solid state form and above which the base is the stable solid state form. Therefore, it has generally been considered that pHmax is pivotal in dictating the risk for salt disproportionation.25,38 Somewhat surprisingly, our results are not supportive of this generally accepted model for disproportionation in powders. The salts evaluated herein have similar values of pKa and pHmax (Table 1) but show highly disparate disproportionation tendencies. Thus, only PIOH and amorphous ATZS undergo disproportionation, while all other salts appear resistant. Importantly, AFM and SEM surface analysis support that the salts are truly resistant to disproportionation with no evidence of transformation to a new form occurring at the crystal surface. The disparity in disproportionation extent between the various salts upon interaction with a common basic excipient suggests that structural factors may play a role and that a classical solution mediated phase transformation model may not always be appropriate to predict salt disproportionation tendency. For disproportionation to occur, the salt bridge between the drug and counterion must be disrupted, whereby a proton is lost from the drug. For the proton transfer to occur, the basic species must come into sufficiently close proximity with the reactive portion of the salt (the salt bridge). Proton transfer is facile for solvated molecules, but is likely to be K

DOI: 10.1021/acs.cgd.8b01188 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

between the API molecule and its counterion. Clearly, more examples of salts with different types of salt bridges need to be evaluated in order to draw any conclusions about this importance of this factor, if any, to the disproportionation reaction. Crystal void space analysis was carried out to identify if salts with facile disproportionation have voids or channels in the crystal structure which enable access of water molecule to the salt bridge. However, none of the salt crystals were found to have channels and voids, except ATZS, where isolated void space was observed. Therefore, there is no direct correlation between the void space and the susceptibility of the disproportionation. To understand other structural features which could have prevented or facilitated the salt to free base conversion of these salts, a comparison of the salt and their respective free base crystal structures was carried out. For solid-state phase transformations, researchers have historically suggested that transformations may occur more readily if the initial and final states show structural similarity.39 Given that the salt crystal lattice contains both the drug and the counterion, crystal structures might be expected to show considerable variations between salt and free form. ATZS and ZIPH have different space groups for the salt and free base crystals, while, in the case of PIOH, both salt and free base are present in same space group. However, for ZIPM and SORT, both salt and free base are present in the same space group and disproportionation was not observed, suggesting that similarity in space group between salt and free base crystals is not a key factor in facilitating disproportionation. This is in accordance with Mnyuk who has suggested that structural similarity is unlikely to dictate the likelihood or kinetics of a phase transformation.40 Conformational analysis indicated that among all structures, conversion of the salt to the free base, required greater conformational changes for ATZ and PIOH molecules. However, for other systems, molecular conformations were quite similar in salt and free base crystals. Molecular conformation differences might be important if there is a high energy barrier to be overcome in order to achieve the free base conformation, however such energy differences were not evaluated since this factor could not be qualitatively linked to disproportionation tendency. Based on the AFM and SEM data, disproportionation results in the nucleation and growth of new crystals of the free base form, with fragmentation of the original crystalline material. Thus, conversion involves a reconstructive transformation, which in turn requires breakage and reforming of hydrogen bonds and other interactions to convert from the salt to the free base. Therefore, the extent of hydrogen bond disruption, and the ease of reforming hydrogen bonding may play a role in determining disproportionation tendency. The chloride ion in PIOH forms two hydrogen bonds with two individual pioglitazone cations. In ZIPH and ZIPM, the hydrogen bonding network is complex as both anions and water are involved in the bonding network. Given that the free base is anhydrous, for these salts, conversion to the free form requires significant disruption of hydrogen bonding, between cation, water, and anion. For ATZS, the sulfate anions hydrogen bond not only to the atazanavir cations, but also to another sulfate anion. In SORT, three hydrogen bonds form between the anion and three sorafenib cations. PIOH thus appears to have the least complex H-bonding pattern and also fairly weak hydrogen bonds relative to the other salts. This may well

Table 4. Structrual Changes Occuring during Salt to Free Base Conversion Conversion PIOH to PIO

H Bonds in the salt form (distance (D···A)/Å) N−H···Cl (3.049) N−H···Cl (3.144)

ZIPH to ZIP

N−H···Cl (3.120) O−H···Cl (3.152) O−H···Cl (3.157)

ZIPM to ZIP

N−H···O (2.953) O−H···O (2.688) O−H···O (3.049) O−H···O (2.758) O−H···O (2.914) O−H···O (3.116)

ATZS to ATZ

SORT to SOR

O−H···N O−H···O O−H···N O−H···O N−H···O N−H···O O−H···O

(2.741) (2.734) (2.830) (2.917) (2.619) (2.639) (2.473)

N−H···O (3.021) N−H···O (2.886) N−H···O (2.760)

Bond between

Molecule orientation change

APIcounterion APIcounterion APIcounterion Watercounterion Watercounterion API-Water WaterMesylate WaterMesylate WaterMesylate WaterMesylate WaterMesylate Water-API Water-API Water-API Water-API API-Sulfate API-Sulfate Sulfate− Sulfate APITosylate APITosylate APITosylate

Rotate a molecule by 90°

Rotate a molecule by 180°

Change dihedral angle (>90°) + rotate a molecule by 90°

Slight change in dihedral angle (