The Use of Cooling Crystallization in an Ionic Liquid System for the Purification of Pharmaceuticals
Cameron C. Webera,b, Samir A. Kulkarnia, Andreas J. Kunov-Krusea,c, Robin D. Rogersd,e and Allan S.Myersona*
aNovartis-MIT Center for Continuous Manufacturing andDepartment of Chemical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA
b Current address: Department of Chemistry, Imperial College London, London SW7 2AZ, UK
c Current address: Department of Chemistry, Technical University of Denmark, 2800 Kgs. Lyngby, Denmark
d Center for Green Manufacturing and Department of Chemistry, The University of Alabama, Tuscaloosa, Alabama 35487, USA
e Current address: Department of Chemistry, McGill University, 801 Sherbrooke St West, Montreal QC H3A 0B8, Canada
Abstract
The application of ionic liquids (ILs) as solvents is frequently discussed in the context of their tunability, with the potential to tailor the solvent system uniquely to the process being investigated. Instead, here we study the potential for a singleIL, 1-ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide ([C2C1Im][NTf2]), to be used for the cooling crystallization of a wide range of active pharmaceutical ingredients (APIs). [C2C1Im][NTf2] was selected on the basis of its thermal stability, low reactivity and miscibility with solvents of moderate polarity which suggests that it is miscible with liquids of comparable polarities to many API molecules.The overwhelming majority of APIs tested were soluble at >50 wt%within [C2C1Im][NTf2] at elevated temperatures despite relatively poor solubility at room temperature. This dramatic effect was ascribed to the miscibility of most of the molten APIs with the IL.The solubility curves for nine APIs were measured which established the potential use of this IL as a crystallization solvent.Finally, cooling crystallizations were conducted using acetaminophen containing common impurities as models. The cooling crystallizations within [C2C1Im][NTf2] were found to produce acetaminophen in similar or greater purity with substantially improved yields relative to a number of control cooling and antisolvent crystallizations.
Introduction
Ionic liquids (ILs), commonly defined as salts that melt below 100 °C,1,2 are a unique class of solvents that have attracted considerable research interest in recent years.3 The source of this interest arises from the favorable properties someILs can exhibit, such as low flammability,4,5 low vapor pressures,6tunable physicochemical properties through appropriate ion selection,7,8large liquidus ranges and good thermal stability.ILs have been applied to a wide range of fields including synthesis and catalysis3,9 as well asmaterial10-13 and separation science.14,15Surprisingly, studies into the use of ILs as solvents for the purification bycrystallization of organic solids, particularly pharmaceuticals,remains relatively limited despite the industrial significance of this methodology.16,17
The tunability of ILs has been one of their most widely touted and exploited benefits as solvents, as the ability to manipulate solution interactions can lead to the design of ILs that are good solvents for chosen target compounds. This has been explored with regards to the solubility of pharmaceutical substances,18-21 and their crystallization.22-25These approaches typically involve the use of antisolvent crystallization23-25 or a combination of liquid-liquid extraction and antisolvent crystallization to recover the target,16although a recent publication has demonstrated the applicability of cooling crystallization of acetaminophen from hexafluorophosphate based ILs from the perspective of crystal engineering.22A complication of antisolventcrystallization methods is that the final solution contains the IL, soluble impurities and the antisolvent at the very least which complicates the recovery and recycling of the IL.Furthermore, the addition of a volatile antisolvent negates some of the advantageous physical properties of the IL, in particular their low vapor pressure, and the use of water as the antisolvent necessitates expensive and energy intensive drying processes. While another IL could be used as an antisolvent to retain these properties, as has been demonstrated by An et al.,23,24 this results in even more complicatedsubsequent separations.
In order to address these concerns we have considered the use of cooling crystallization in an IL as a purification methodology to reduce the complexity of the resultant solution. To achieve such a crystallization in good yields without degradation of the solvent or target compounds, the IL needs to be selected such that it possesses good thermal stability and relatively inert ions. Ideally, the IL will possess a relatively low viscosity to enable lower temperatures to be attained and the target compounds will be poorly soluble at low temperatures. This final requirement runs counterto the aims of much of the previous work conducted on organic crystallization in ILs, where designing or selecting ILs for high solubility has been the priority.16,20 As a result of these factors 1-ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide ([C2C1Im][NTf2]) was selected for initial study. In addition to possessing low viscosity, good thermal stability and relatively inert ions; [C2C1Im][NTf2] is miscible with a wide range of liquids with polarities from methanol to ethyl acetate.26 It was worth considering that in the extreme case of the target compound melting, it will either form an immiscible phase or be completely miscible with the IL. Based on the miscibility of [C2C1Im][NTf2] with all but the most and least polar liquids, it could be anticipated that this IL will be miscible with most APIs at least as the temperature approaches their melting point. This approach has been used previously for the crystallization of aromatic compounds such as catechol and acetophenone from [C8C1Im][NTf2], although to the best of our knowledge not for the crystallization or purification of APIs.27
To ascertain the generality of our approach and to probe the limits of miscibility and thermal stability, we have examined a wide range of different active pharmaceutical compounds (APIs) as model systems for the cooling crystallization approach (Figure 1). The melting point of each API is listed in Table 1. The potential for cooling crystallization to be applied has been examined by the conduct of variable temperature solubility experiments and investigating the miscibility of each of the APIs with the IL at their melting point. To demonstrate proof of principle and examine the efficacy of this purification approach, the cooling crystallization of acetaminophen (AAP) with various impurities has been conducted and compared with results obtained previously for antisolvent processes25 and with cooling crystallizations conducted in water.
Figure 1. Structures of all compounds, ILs and model impurities utilized in this study. (a) Acetaminophen (AAP), (b) fenofibrate (FF), (c) ibuprofen (IBU), (d) acetylsalicylic acid (ASA), (e) itraconazole (ITR), (f) griseofulvin (GSF), (g) salicylic acid (SA), (h) naproxen (NPX), (i) amoxicillin (AMOX), (j) etomidate (ETO), (k) rufinamide (RUF), (l) cyclosporine (CYC), (m) 4-aminophenol (4-AP), (n) 4-nitrophenol (4-NP), (o) 4’-chloroacetanilide (4’-CA) and(p) 1-ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide ([C2C1Im][NTf2]).
Experimental Section
Materials.Acetaminophen (AAP), 4-aminophenol (4-AP), 4-nitrophenol (4-NP), 4’-chloroacetanilide (4’-CA), fenofibrate (FF), ibuprofen (IBU), itraconazole (ITR), griseofulvin (GSF), acetylsalicylic acid (ASA), salicylic acid (SA), naproxen (NPX), amoxicillin trihydrate (AMOX)and etomidate (ETO) were purchased from Sigma Aldrich and used as received. Rufinamide (RUF) and cyclosporine (CYC) was obtained from Xian-Shunyi Bio-Chemical Technology Co., Ltd. and used as received.
1-Ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide ([C2C1Im][NTf2], Iolitec) was stored in a dry-box under a nitrogen atmosphere and dried under high vacuum prior to use.
The uncertainty in mass was ± 0.0001 g for all experiments. The uncertainty in temperature was ± 0.01 °C for solubility experiments conducted up to 100 °C and ±0.1 °C for solubility and crystallization experimentsconducted above 100 °C.
Solubility Measurements.In a typical solubility experiment, 1.00 g solvent was added to excess solid and the resultant suspension stirred for 24 h using a rare earth magnetic stirrer. The temperature was maintained by means of a Peltier heating system in an Avantium Crystal16 instrument. Temperatures above 100 °C were obtained using a heating block on a stirrer hotplate. The suspension was passed through a 0.20 μm syringe filter and the filtrate diluted in deuterated acetone. The mole fraction solubility was determined by1H NMR or HPLC with consistent results obtained between each method. For experiments above 25 °C, the syringe was preheated to prevent crystallization during transfer. All measurements were conducted in duplicate.
Crystallization of AAP from [C2C1Im][NTf2].In a typical crystallization experiment, AAP (0.550 g), impurity (4-AP, 4-NP or 4’-CA, 61 mg) and [C2C1Im][NTf2] (5.00 g) were combined in a scintillation vial. The vial was sealed then the suspension heated with stirring at 125°C until the solids dissolved. The solution was then cooled to 50 °C at 1°C min−1 and then the temperature maintained at 50°C for an additional 1.5 h before being filtered, washed with dichloromethane (3 × 5 mL) and dried at the pump.
Crystallization of AAP from water. In a typical crystallization experiments, an aqueous solution of AAP (0.032 g mL−1) containing either 4-NP or 4-AP (3.6 mg mL−1) was heated with stirring to 60 °C then cooled to 20 °C at 0.5 °C min−1. The temperature was then maintained at 20 °C for 1.5 h. The solid was collected by filtration and washed with ice water (5 mL) before being dried at the pump.
HPLC Analysis for AAP Crystallization.The HPLC instrument (Agilent 1100) was equipped with a UV diode array detector. The column used was a YMC-Pack ODS-A 150 4.6 mm i.d. column packed with 3 μm particles with 12 nm pore size (YMC America Inc.). The detection wavelength was set at 275 nm for AAP and 4-AP, 254 nm for 4’-CA and 230 nm for 4-NP. Samples were analyzed using an isocratic method with a 30:70 methanol:water mobile phase containing 0.3% trifluoroacetic acid with a 10 μL injection and 1 mL min−1 flow rate. Analysis run times were 5 min for 4-AP, 20 min for 4-NP and 35 min for 4’-CA.
Powder X-ray Diffraction. Powder XRD was performed using a PANalyticalX’Pert Pro diffractometer over the range 5–50° 2θ using a Cu Kαsource with a total collection time of 10 min 37 sec and sample stage rotation of 15 rpm. Solid samples were pressed flat onto a zero background sample holder without grinding to prevent polymorph transformation.
Results and Discussion
To ascertain whether a range of APIs were miscible with [C2C1Im][NTf2] near their melting points, 12 APIs were heated in the presence of [C2C1Im][NTf2] at 0.50 mass fraction. The miscibility of these APIs, as determined by visual observation, is noted in Table 1 alongside their melting points. From Table 1 it is evident that the overwhelming majority of APIs tested were miscible with [C2C1Im][NTf2] at their melting points despite generally beingpoorly soluble at 25°C. This observation largely accords with our initial hypothesis that since [C2C1Im][NTf2] is miscible with liquids of a wide range of different polarities, [C2C1Im][NTf2] should be a good solvent for solutes of intermediate polarity as their melting points are approached.It should be noted that although the viscosity of [C2C1Im][NTf2] is lower than many other ILs,28 it is higher than most conventional solvents. Consequently, many suspensions containing high solute concentrations formed slurries that were difficult to stir at room temperature although the use of rare earth magnetic stirrers made these suspensions tractable. This situation improved at elevated temperatures where most solutions could easily be stirred.
The only APIs not miscible with [C2C1Im][NTf2] upon heating to their melting point were AMOX and IBU. AMOX was too thermally unstable for its solubility to be measured so no substantial conclusions can be drawn in this case. IBU, however, did form two mutually immiscible liquid phases at 0.50 mass fraction. IBU possesses a relatively low melting point, a carboxylic acid functionality and a branched aliphatic side chain which collectively may account for this behavior. Other APIs containing carboxylic acid moieties i.e. SA, NPX and ASA were miscible at their melting point, although none of these compounds possessed the same aliphatic component as IBU. This indicates that the presence of a carboxylic acid group does not definitely lead to immiscibility with [C2C1Im][NTf2]. ASA,SA and NPX also possessed higher melting points than IBU and the higher temperatures may affect the structure and solvation properties of [C2C1Im][NTf2]. To investigate this further a sample of 0.50 mass fraction IBU in [C2C1Im][NTf2] was heated to 160 °C and remained immiscible, indicating that the higher melting temperatures of the other carboxylic acid containing compounds do not account for this behavior. Consequently it is apparent that the aliphatic component or the combination of this extremely nonpolar aliphatic component with the highly polar carboxylic acid moiety are responsible for the immiscibility of IBU.
Table 1. Miscibility of APIs tested at 0.50 mass fraction in [C2C1Im][NTf2] at their melting points.
API / Melting Point (°C) / Miscible?AAP / 169 / Y
FF / 81 / Y
IBU / 76 / N
ASA / 135 / Y
ITR / 166 / Y
GSF / 220 / Y
SA / 159 / Y
NPX / 152–155 / Y
AMOX / 194 / Na
ETO / 67 / Y
RUF / 232–234 / Y
CYC / 148–151 / Y
a Solid visibly decomposed rapidly upon heating
From the compounds examined for miscibility, solubility curves were measured for all except three compounds. The excluded APIs were AMOX due to issues with its thermal stability, RUF as its solubility even at 125 °C was below the limit of detection and CYC as it was structurally too complex to analyze using the NMR methodology employed.
The solubility of the compounds with melting points under 100 °C (FF, ETO and IBU) are depicted in Figure 2 and the values and errors in terms of both mass and molar solubilityare tabulated in the ESI. It is clear from this diagram that two very different behaviors are observed. The solubility of ETO and FF in [C2C1Im][NTf2] is significantly greater than IBU at all temperatures tested and notably increase exponentially as their melting point is approached. This behavior is qualitatively in agreement with the ideal solubility equation. The solubility of IBU, however, does not vary greatly and two immiscible phases form upon melting, as discussed previously.
Figure 2. Variation of solubility with temperature for FF, IBU and ETO. Solid vertical lines correspond with the melting points of each of the APIs and use the same color code. Lines between points are provided as a guide to the eye.Errors have been omitted as they are generally smaller than the data point.
For the compounds with higher melting points, namely AAP, GSF, ASA, NPX, SA and ITR, all were miscible with [C2C1Im][NTf2] at 0.50 mass fraction at their melting point as shown in Table 1. Despite this, there remains some distinct differences in the solubility profiles of these APIs (Figure 3 and ESI).It is important to note that in collecting these solubility data, a small amount of decomposition was observed for ASA at the higher temperature values (ESI) so some caution must be taken with the absolute solubility values obtained. Nonetheless, the solubility observed based on the amount of pure material in solution was reproducible and the extent of decomposition small. Given the solubility measurements were obtained by exposure to these temperatures for 24 h, far longer than would be required for a crystallization process, it is unlikely that this would preclude the crystallization of ASA under these conditions using [C2C1Im][NTf2].
Figure 3. Variation of solubility with temperature for AAP, GSF, ASA, NPX, SA and ITR. Solid vertical lines correspond to the melting point of each API and use the same color code. Line between points provided as a guide to the eye.Errors have been omitted as they are generally smaller than the data point.
In general and as would be expected, solubility is higher for compounds with lower melting points and an exponential increase in solubility observed as the melting point is approached, at least for APIs with melting points low enough so that these data could be obtained. Interestingly, the solubility profiles themselves differ even for compounds that possess similar melting points and this can be attributed to specific interactions between the APIs and the solvent. For example, SA and NPX are substantially less soluble than AAP in [C2C1Im][NTf2] at elevated temperatures on both a mole and weight basis despite possessing lower melting points. This behavior can be understood either due to unfavorable specific interactions between the carboxylic acid moiety and the [NTf2]− anion, although this would not account for the high ASA solubility, or due to the formation of aromatic clathrate type structures.10,29,30 The presence of a more electron rich aromatic ring for AAP due to the amide and phenolic functionalities may lead to stronger [C2C1Im]+–AAP interactions and its smaller size relative to NPX could lead to its more favorable inclusion. SA contains two strongly hydrogen bonding groups which would account for its lower solubility in the hydrophobic [C2C1Im][NTf2] IL at temperatures below its melting point. Attempts to fit these data to commonly used models such as the Van’t Hoff equation and the modified Apelblat model led to large average deviations. This is likely due to the very high solubilities and large temperature range over which the solubilities were measured which would lead to substantial changes to the bulk solvent, rendering many of the underlying assumptions of these models invalid.
From the solubility data obtained, it is clear that [C2C1Im][NTf2] has the potential to act as a medium for cooling crystallization for a wide range of different APIs. Only ETO and FF have significant solubility at room temperature while the other APIs vary from low solubility at room temperature to extremely high solubilities at elevated temperatures. This means that, with the exception of ETO and FF, a suitable temperature range above room temperature would exist for the cooling crystallization of each API. For FF, it appears that cooling below room temperature can facilitate sufficiently low solubilities to enable satisfactory crystallization yields. Given [C2C1Im][NTf2] is liquid at temperatures above −17 °C, this provides some scope for cooling below room temperature.31
To experimentally determine whether a cooling crystallization process could be successfully conducted within this IL and its effect on the purification of the API, the crystallization of AAP in the presence of some common impurities (4-AP, 4-NP and 4’-CA)was conducted as a model study.These have previously been explored as model systemsforantisolvent crystallization from IL mixtures.25The cooling crystallizations were performed by cooling a solution of AAP with 10 wt% impurity in [C2C1Im][NTf2]from 125°C–50°C at 1 °C min−1 followed by maintaining the temperature at 50°C for 1.5 h. The resultant crystals were then filtered, washedwith dichloromethane and dried.The cooling rate of 1 °C min−1 was arbitrarily chosen as an intermediate rate of cooling as the cooling rate has previously been shown to have a negligible effect on the metastable zone width and resultant properties of AAP crystallized from ILs.22