Article pubs.acs.org/molecularpharmaceutics
Impact of Polymers on Crystal Growth Rate of Structurally Diverse Compounds from Aqueous Solution Grace A. Ilevbare,† Haoyu Liu,‡ Kevin J. Edgar,‡ and Lynne S. Taylor*,† †
Department of Industrial and Physical Pharmacy, College of Pharmacy, Purdue University, West Lafayette, Indiana 47907, United States ‡ Department of Sustainable Biomaterials, College of Natural Resources and Environment, Virginia Tech, Blacksburg, Virginia 24061, United States S Supporting Information *
ABSTRACT: The presence of an effective crystal growth inhibitor in solution is desirable to prolong supersaturation since residual crystalline material in an amorphous formulation resulting from the manufacturing process or formed during storage or dissolution can potentially have a significant impact on the extent and duration of supersaturation. In this study, the effectiveness of a group of chemically diverse polymers, including several recently synthesized cellulose derivatives, on solution crystal growth of three structurally diverse compounds (celecoxib, efavirenz, and ritonavir) was quantified at different extents of supersaturation and compared. Despite the different chemical properties and structures of the model compounds, nonspecific hydrophobic drug−polymer interactions appeared to be important in determining the impact of a given polymer on crystal growth for of all these drug compounds. Specific intermolecular interactions were also found to be important for crystal growth inhibition of celecoxib and efavirenz by the hydrophilic polymer, PVPVA. These interactive forceshydrophobicity and specific intermolecular interactionsare likely to promote adsorption of the polymer onto the surface of the crystalline drugs, thus influencing crystal growth. The effectiveness of the polymers also depended on the rate of crystallization of the drug molecules. At a similar supersaturation ratio of ∼1.2, ritonavir and celecoxib had slower normalized crystal growth rates (0.20 and 0.91 mg min−1 m−2, respectively), while the normalized crystal growth rate of efavirenz was significantly higher (2.97 mg min−1 m−2), resulting in lower levels of crystal growth inhibition by the polymers for efavirenz. KEYWORDS: crystal growth, cellulose polymers, supersaturation, hydrophobicity, specific intermolecular interactions, amorphous formulation
■
permeability;13−16 cyclodextrins and surfactants can decrease the free fraction of drug which results in decreased intestinal membrane permeability of lipophilic drugs (BCS class II), while the permeability of drugs solubilized in cosolvents has been found to decrease with increasing cosolvent fraction.11,17−19 Solid-state modification, for example, delivery of an amorphous form of the drug, can increase aqueous solution concentration without reducing intestinal membrane permeability.20 The supersaturated solutions generated from the dissolution of amorphous solids may lead to an increase in absorption compared to that of a saturated solution if supersaturation can be maintained for a physiologically relevant type period. Using in silico modeling and simulation to predict drug absorption from the GI tract, the relationship between drug supersaturation and the improvement in oral bioavailability has been demonstrated.21,22 However, in order to
INTRODUCTION The rate and extent of drug absorption from the gastrointestinal (GI) tract depends on the solubility/dissolution characteristics and permeability of the drug through the GI membrane.1−3 If a drug candidate has reasonable membrane permeability, then often the rate-limiting process in absorption is the solubility of the drug dose in the GI tract;4 this is characteristic of liphophilic, class II compounds in the biopharmaceutical classification system (BCS). Low aqueous solubility compounds often suffer from limited bioavailability, and the formulation of these molecules into orally administered dosage forms with sufficient bioavailability is a drug delivery challenge. A number of solubility enhancing formulation strategies are routinely used to create elevated concentrations of the drug in the GI tract including: cosolvents,5 complexforming agents such as cyclodextrins,6,7 and surfactant-based formulations.8−10 Although significant increased apparent solubility may be achieved by these techniques, their impact on the overall fraction of the dose absorbed is erratic.11,12 It has been demonstrated that cosolvent, complexation, and surfactant-based solublization methods may lead to a lower effective © XXXX American Chemical Society
Received: January 16, 2013 Revised: April 9, 2013 Accepted: April 18, 2013
A
dx.doi.org/10.1021/mp400029v | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
presented in Table 1, while the abbreviations for the synthesized cellulose ester polymers (Figure 2) are presented in Table 2.
maintain supersaturation, crystallizationnucleation and/or crystal growthmust be prevented. Trace crystalline material in an amorphous formulation, either resulting from the manufacturing process or produced during storage, can thus potentially have a significant impact on the extent and duration of supersaturation; the presence of an effective crystal growth inhibitor in solution is therefore highly desirable to prolong supersaturation. It has been proposed that maintenance of supersaturation through the addition of polymeric additives is the result of intermolecular interactions in solution (hydrogen bonding) and/or steric hindrance of recrystallization.23−26 In a previous structure−property study, we found that the hydrophobicity of the polymer as well as the presence of ionized functional groups were important in determining the effectiveness of a given polymer as a crystal growth inhibitor for the model compound, ritonavir.27 The aim of this current study was to quantify and compare the effectiveness of a group of polymers, selected based their chemical and physical attributes, in inhibiting crystal growth from supersaturated solutions of three chemically diverse poorly water-soluble compounds: celecoxib, efavirenz, and ritonavir. A variety of polymers were investigated, including a number of recently synthesized cellulose polymers with greater chemical diversity than commercially available cellulose derivatives. The importance of hydrophobic interactions was investigated for selected polymers and crystallizing solute by varying the ionic strength of the medium.
Table 1. Commercially Available Polymers and Abbreviations Used in This Study polymer
abbreviation
poly(vinylpyrrolidone) (K 29/32) poly(vinylpyrrolidone vinyl acetate) (K 28) poly(acrylic acid) hydroxypropyl methyl cellulose (606 grade) hydroxypropyl methyl cellulose acetate succinate (AS-MF)
PVP PVPVA PAA HPMC HPMCAS
■
EXPERIMENTAL SECTION Materials. Efavirenz (Figure 1a) and ritonavir (Figure 1b) were purchased from Attix Corporation, Toronto, Ontario, Figure 2. Molecular structure of the novel synthesized cellulose derivatives and the substituent groups. These cellulose derivatives are not regioselectively substituted. The abbreviations are presented in Table 2.
Table 2. Synthesized Polymers and Abbreviations Used in This Study polymer cellulose cellulose cellulose cellulose cellulose cellulose cellulose cellulose cellulose cellulose cellulose
Figure 1. Molecular structure of (a) efavirenz, (b) ritonavir, and (c) celecoxib.
Canada. Celecoxib (Figure 1c) was provided by Pfizer Inc. (Groton, MA). Acetonitrile and methanol were purchased from Macron Chemicals (Phillipsburg, NJ). The commercially available polymers were obtained from various sources: poly(vinyl pyrrolidone) K29/32 and poly(acrylic acid) (Sigma-Aldrich Co., St. Louis, MO), poly(vinyl pyrrolidone vinyl acetate) K28 (BASF, Germany), and hydroxypropyl methyl cellulose 606 grade and hydroxypropyl methyl cellulose acetate succinate AS-MF grade (Shin-Etsu Chemical Co., Ltd. (Tokyo, Japan)). The synthesis of the cellulose ω-carboxyalkanoates is described in refs 28 and 29. The abbreviations used in this report for the commercially available polymers are
propionate adipate acetate 320S adipate acetate 320S suberate 0.63 acetate 320S suberate 0.90 acetate 320S sebacate acetate propionate 504-0.2 adipate 0.85 acetate butyrate 553-0.4 adipate 0.81 acetate propionate 504-0.2 adipate 0.33 acetate butyrate 553-0.4 adipate 0.25 acetate propionate sebacate 0.24 acetate butyrate sebacate
abbreviation CP Adp CA 320S Adp CA 320S Sub 0.63 CA 320S Sub 0.90 CA 320S Seb CAP Adp 0.85 CAB Adp 0.81 CAP Adp 0.33 CAB Adp 0.25 CAP Seb CAB Seb
Methods. Characterization of Seed Crystals. The morphology of celecoxib and efavirenz crystals was characterized using scanning electron microscopy (SEM). Seed crystals were prepared for SEM by dispersing them in water. Approximately 20 μL of the dispersion was placed on a glass slide and allowed to dry overnight in a vacuum oven at room temperature. Cover slides were mounted using double sticky copper tape and sputter-coated with Pt for 60 s prior to imaging. Subsequently, samples were imaged with a FEI NOVA nanoSEM field emission SEM using the Everhart−Thornley B
dx.doi.org/10.1021/mp400029v | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
of the compound. The group contributions at 25 °C are presented in ref 31. The SP values of the polymers are summarized in Table 3, while the SP values of the model compounds are presented in Table 4.
(ET) detector and through-the-lens detector (TLD). Parameters were 5 kV accelerating voltage, ∼4−5 mm working distance, and beam spot sizes of 3 and 30 μm aperture. Magnifications were 100−10 000×. Powder X-ray diffraction (PXRD) and differential scanning calorimeter (DSC) experiments were performed to further evaluate the seed crystals. Bulk seed crystals were analyzed using PXRD. PXRD patterns were obtained using a Shimadzu XRD 6000 diffractometer (Shimadzu Scientific Instruments, Columbia, MD) using Bragg−Brentano parafocusing geometry. The equipment was calibrated using a silicon standard which has a characteristic peak at 28.44° 2θ. The X-ray tube consists of a target material made of copper, which emits a Kα radiation. Experiments were performed using an accelerating potential of 40 kV and current of 30 mA. The divergence and scattering slits were set at 1.0°, and the receiving slit was set at 0.3 mm. The experiment was conducted with a scan range from 5 to 40° 2θ, while the scanning speed was 4 deg/min. Thermal transitions were measured using a TA Q2000 DSC (TA Instruments, New Castle, DE) attached to a refrigerated cooling accessory (RCS) (TA Instruments, New Castle, DE). Both the DSC and the RCS were purged with nitrogen gas. The thermogram was obtained by heating the sample (bulk drug compound, used as received from the manufacturer) at 5 °C/ min and 20 °C/min for melting point and glass transition temperature determinations, respectively. Melting point and enthalpy of fusion were determined from a first heat scan. The temperature range used for efavirenz and ritonavir was 25−150 °C, while the temperature range used for celecoxib was 25−200 °C. Thermal transitions were viewed and analyzed using the analysis software Universal Analysis 2000 for Windows 2000/ XP provided with the instrument. The surface area of the bulk seed crystals was determined using the Brunauer−Emmett−Teller (BET) method.30 The BET specific surface area was determined by the use of nitrogen adsorption/desorption isotherms at liquid nitrogen temperature (the BET bath temperature was −195.5 °C and relative pressures ranged from 0.06 to 1.0) using a TriStar 3000 V4.01 instrument (Micromeritics, Norcross, Georgia). The sample was prepared for analysis by degassing using a FlowPrep 060 instrument (Micromeritics, Norcross, Georgia) for 120 min at 100 °C, under constant flow of nitrogen gas. The data from the multipoint determination were analyzed and used to estimate surface area using the analysis software provided with the instrument. The linear portion of the adsorption isotherm with relative pressures ranging from 0.06 to 0.30 was used for analysis. Solubility Parameter. The solubility parameter (SP) was used to characterize the relative hydrophobicity of the polymers and the model compounds. SP values were estimated using the method proposed by Fedors.31 The solubility parameter can be evaluated using:31 δ=
∑i Δei ∑i Δvi
=
ΔEv V
Table 3. Solubility Parameter (SP) Values of the Polymers Investigated polymer abbreviations
SP (MPa1/2)
CAB Seb CAP Seb CAB Adp 0.25 CAP Adp 0.33 CAB Adp 0.81 CAP Adp 0.85 CA 320S Seb HPMCAS CA 320S Sub 0.63 CA 320S Sub 0.90 HPMC CA 320S Adp CP Adp PVPVA PAA PVP
19.62 19.94 20.05 20.56 20.86 21.27 22.36 22.44 22.62 22.66 22.68 22.95 23.28 25.77 25.92 28.39
Table 4. Physicochemical Properties of the Model Compounds property pKa molecular weight (g mol−1) Log P solubility parameter (MPa1/2) enthalpy of fusion (kJ mol−1) melting point (°C) equilibrium solubility (μg mL−1) theoretical amorphous “solubility” (μg mL−1) experimental amorphous “solubility” (μg mL−1)
ritonavir
efavirenz
celecoxib
1.8 and 2.6 720.9 5.6a 20.03 60.4 122.7 1.3 20.6
10.2 315.7 4.7a,b 21.89 14.5 139.0 8.2 19.8
11.1 381.4 3.5c 23.38 37.6 163.5 1.5 22.6
19.8
12.3
2.5
a
As reported in ChemBioDraw Ultra v11.0, CambridgeSoft Corp., Cambridge, MA. bReference 38. cReference 35.
Moisture Sorption Analysis. The moisture sorption profiles of amorphous celecoxib, efavirenz, and ritonavir were determined at 37 °C using a VTI-100 continuous vapor flow sorption instrument (TA Instruments, New Castle, DE). The amorphous samples were prepared using the melt-quench method. A mortar and pestle was used to reduce the size of the amorphous samples prior to moisture sorption analysis. Approximately 15.0 mg of the amorphous sample was placed into the sample pan and dried at 37 °C and low relative humidity until the weight change was less than 0.01 wt % over 5 min. Moisture sorption analysis at the equilibrium temperature (37 °C) was performed by equilibrating the sample under controlled relative humidity (RH) ranging from 5% RH to 95% RH, in 10% RH intervals. Equilibrium was assumed to be attained when the weight change was less than 0.01 wt % over 5 min at each RH step. Moisture sorption isotherms were obtained on three individually prepared samples of each compound analyzed. Determination of Amorphous “Solubility”Theoretical and Experimental. The theoretical amorphous solubility of the
(1)
where the Δei and Δvi are the additive atomic and group contributions to the energy of vaporization and molar volume, respectively, ΔEv is the energy of vaporization at a given temperature, and V is the corresponding molar volume which is calculated from the known values of molecular weight and density. The method is based on group additive constants; therefore, it requires only knowledge of the structural formula C
dx.doi.org/10.1021/mp400029v | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
model compounds was estimated using the experimentally determined crystalline solubility (Ceq), the estimated free energy difference between crystalline and amorphous forms (ΔGC) and the activity of the amorphous solute saturated with water [−I(a2)]:33,34 ⎡ ΔG ⎤ Camorphous = Ceq × exp[−I(a 2)] × exp⎢ C ⎥ ⎣ RT ⎦
S=
(2)
ΔHf (Tm − T )T Tm2
(4)
where C is the concentration of the supersaturated solution and C0 is the equilibrium solubility at a given temperature. Here it is assumed that concentration is a reasonable substitute for thermodynamic activity because of the extremely dilute solutions involved in these studies. All experiments were performed in triplicate. Stock solutions (solubilized drug) of the drug compounds were prepared by dissolving 200 mg of drug in 50 mL of methanol to make a final stock solution of 4 mg/mL. Supersaturated solutions were generated by adding a small volume of solublized drug in methanol (∼0.25 mL) to sodium phosphate buffer, pH 6.8 (50 mL); the amount of methanol added to the aqueous solution (1:200 ratio methanol to buffer) had a negligible impact on equilibrium solubility (based on solubility measurements performed using a similar ratio of organic solvent and buffer). The seeds (0.010 g) were added to the buffer and allowed to equilibrate at 37 °C prior to addition of solubilized drug. The test solution was stirred at a constant speed of 400 rpm using a Cole Parmer (model 5000020) overhead mixer attached to axial-flow impeller blade, with a digital mixer controller (Cole Parmer Instrument Co., Niles, IL). Desupersaturation profiles were measured using a SI Photonics UV spectrometer (Tuscon, Arizona) coupled to a fiber optic probe (path-length 10 mm) at a constant temperature of 37 °C. Wavelength scans (200−450 nm) were performed at 5 s time intervals. Data collection began immediately after generation of supersaturation. The slope of the concentration vs time curve over the first 2 min of the experiment was taken as the initial bulk crystal growth rate (Rg). The effect of the ionization state of the polymer on crystal growth rate was evaluated by performing experiments at pH 3.8 and 6.8, using 100 mM sodium acetate buffer and 100 mM sodium phosphate, respectively. Drug−polymer hydrophobic interactions were investigated by performing growth experiments in sodium phosphate buffer at two different ionic strengths, 50 mM and 100 mM. Turbidity (indicative of secondary nucleation) was detected by monitoring the extinction, at 280 nm for ritonavir and 350 nm for celecoxib and efavirenz, using a SI Photonics UV/vis spectrometer, fiber optically coupled with a dip probe (path-length 5 mm); the compounds have no absorbance at these wavelengths. Fourier Transform Mid-Infrared (FTIR) Spectroscopy. Interactions between PVPVA and the drug following adsorption onto celecoxib and efavirenz crystals were evaluated by adding 20 mL of PVPVA predissolved in buffer (100 μg/ mL) to approximately 10 mg of seed crystals, followed by overnight equilibration of the suspension. The seed crystals were separated from solution using ultracentrifugation. Ultracentrifugation was performed at 40 000 rpm (equivalent of 274 356g) in an Optima L-100 XP ultracentrifuge equipped with Swinging-Bucket Rotor SW 41 Ti (Beckman Coulter, Inc., Brea, CA) for 15 min. The seed crystals were dried in a vacuum oven at room temperature for ∼12 h prior to FTIR analysis. A Bio-Rad FTS 6000 spectrophotometer (Bio-Rad Laboratories, Cambridge, MA) was used to obtain IR spectra. The spectrophotometer was equipped with a global infrared source, KBr beamsplitter, and a DTGS detector. Spectra were obtained at ambient conditions using the ATR (attenuated total reflectance) sampling method. Spectra were recorded over the range of 4000−400 cm−1 with a resolution of 4 cm−1 and coaddition of 128 scans. A background spectrum was collected
Using the heat of fusion (ΔHf) and melting point (Tm) determined from DSC analysis as described above, the free energy change for transformation of an amorphous solid to a crystalline solid was estimated at 37 °C (T):32 ΔGC =
C C0
(3)
The activity of the hydrated amorphous solute in contact with water [−I(a2)] was estimated by applying the Gibbs−Duhem equation to the water sorption isotherm data for the amorphous solid. The detailed thermodynamic analysis is presented in ref 33. Amorphous samples were prepared by a solvent evaporation method, specifically spin-coating. Spin-coating was performed using a KW-4A spin-coater (Chemat Technology, Inc., Northridge, CA). Crystalline drug compound (30 mg) was dissolved in 0.5 mL of methanol. Two to three drops of solution were placed on a circular glass slide with an 18 mm diameter (VWR International, LLC) and spin-coated. Residual solvents were removed by drying the amorphous films under vacuum at room temperature for ∼24 h. The glass slides were weighed before and after spin coating (after drying) to determine the weight of the amorphous film. Prior to use, the samples were analyzed by cross-polarized light microscopy and powder X-ray diffraction analysis to verify their amorphous nature. A Nikon Eclipse E600 Pol microscope with a 20× objective was used with NIS-Elements software package (Version 2.3; Nikon Company, Tokyo, Japan). The experimental amorphous “solubility” of the model compounds was determined by dissolving the solid in 100 mM sodium phosphate buffer at pH 6.8, 37 °C. Solution concentration was measured using a SI Photonics (Tuscon, Arizona) UV spectrometer coupled to a fiber optic probe (pathlength 10 mm). Wavelength scans (200−450 nm) were performed at 45 s time intervals. Second derivatives (SIMCA P+ V.12 software (Umetrics Inc., Umea Sweden) of the spectra were taken for the calibration and sample data in order to mitigate particle scattering effects. Calibration solutions (1.0− 30.0 μg/mL) were prepared in methanol. The standard curve was linear over the aforementioned concentration range; r2 > 0.999 for the model compounds investigated. Crystal Growth Rate. The crystal growth rate was characterized by measuring the rate of desupersaturation in the presence of seed crystals. Celecoxib, efavirenz, and ritonavir were used as received from the manufacturer and prior to use, the seeds were sieved to a size below 250 μm. Crystal growth experiments were performed in a jacketed beaker connected to a digitally controlled temperature water bath. Crystal growth rate experiments were performed at initial solution concentrations ranging from 2.5 to 20.0 μg/mL in the absence and presence of predissolved polymer present at a concentration of 5 μg/mL. The supersaturation ratio (S) was expressed as: D
dx.doi.org/10.1021/mp400029v | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
Crystal Growth Rates of the Model Compounds as a Function of Supersaturation. Previously, we have observed that the effectiveness of inhibitory polymers is dependent on the initial solute supersaturation.41−43 Therefore, initial experiments focused on evaluating the growth rate behavior of the compounds in the absence of polymer as a function of supersaturation in order to understand how this differs between the various model compounds. For ritonavir and celecoxib, it was found that growth rate measurements could be obtained across a range of concentrations, up to a supersaturation equivalent to the predicted amorphous solubility. During these experiments, there was no evidence of significant secondary nucleation; that is, for these systems, it appears that desupersaturation is mainly due to the bulk crystal growth of the seed crystals present in solution. However, for efavirenz, at all concentrations above 10 μg/mL (S = 1.2) there appeared to be detectable secondary nucleation based on turbidity measurements; therefore, subsequent experiments with polymers were limited to this much lower supersaturation. For efavirenz, at concentrations above 10 μg/mL, the turbidity of the solution began to increase when macroscopic crystals began to form; an increase in turbidity was not observed for celecoxib and ritonavir within the concentration ranges (5−22 μg/mL) used in this study. Additional evidence for the absence of significant secondary nucleation for ritonavir (based on particle size distribution and mass balance calculations) is presented in ref 43. Crystal growth rates for the three compounds were compared at low supersaturation [S of 1.2 for efavirenz and ritonavir, and 1.3 for celecoxib (Table 6)]. The normalized crystal growth rates of the model compounds were then estimated by dividing the experimentally determined crystal growth rates by the product of the amount of seed crystals used (0.010 g) and the respective BET surface area values of the model compounds. Ritonavir had the slowest normalized crystal growth rate of 0.20 mg min−1 m−2, while the growth rate of celecoxib at a similar supersaturation was somewhat faster at 0.91 mg min−1 m−2. In contrast, the normalized crystal growth rate of efavirenz was much faster (2.97 mg min−1 m−2), resulting in a relatively faster desupersaturation in the presence of the seed crystals, even at this low S. For ritonavir and celecoxib, when the solution concentration was increased to 10 μg/mL (S of 7.6 and 6.6, respectively), the normalized growth rate of ritonavir increased considerably, as expected, and was very similar to that of celecoxib (Table 6). A similar trend was observed at the highest supersaturation employed (S of 14.7 and 15.4 for celecoxib and ritonavir, respectively). Impact of Polymers on Crystal Growth Inhibition of the Model Drug Compounds. The effectiveness of the polymers for the model compounds was investigated at an initial compound concentration of 10 μg/mL and this initial concentration corresponds to an initial supersaturation ratio (S) of 6.6 for celecoxib, 1.2 for efavirenz, and 7.6 for ritonavir. Higher levels of inhibition by the polymers were expected for efavirenz, since crystal growth rate experiments were performed at a relatively low supersaturation. The effectiveness (Eg) of the polymers in inhibiting crystal growth was estimated using the following equation:
with the same parameters prior to taking the spectra of the samples. Spectral interference from water vapor and CO2 was prevented by purging the spectrometer with conditioned air. Spectra were analyzed using GRAMS/AI V.7.02 software (Thermal Fisher Scientific, Inc., Waltham, MA). All spectra are plots of absorbance versus wavenumber.
■
RESULTS Properties of the Model Drug Compounds and Characterization of the Seed Crystals. Celecoxib, efavirenz, and ritonavir (Figure 1), were chosen as model compounds due to their poor aqueous solubility values. The equilibrium solubility values of crystalline celecoxib, efavirenz, and ritonavir at pH 6.8 and 37 °C, where all of the molecules exist in the unionized form, were determined to be 1.5, 8.2, and 1.3 μg/mL, respectively.34 At the concentration used in this study (5 μg/ mL), the polymers have an insignificant effect on the equilibrium solubility of the model compounds.34 Celecoxib and efavirenz are weak acids with pKa values of pKa 11.135 and pKa 10.2,36 respectively, while ritonavir is a very weak base with two thiazole functional groups with pKa values of 1.8 and 2.6,37 hence all compounds were predominantly un-ionized during growth rate measurements. In this study, Log P and solubility parameter (SP) values were used as descriptors of hydrophobicity. These values were predicted using group contribution methods and were obtained from the ChemBioDraw Ultra version 12.0 (CambridgeSoft, Cambridge, MA) and various literature sources.35,38 The model compounds have different positive Log P values, indicative of varying levels of hydrophobicity. SP values provide a numerical estimate of the cohesive interactions within a material and can also provide an indication of relative polarity.31 The lower the solubility parameter of a compound, the more hydrophobic it is. Based on Log P and SP values (Table 4), all of the model compounds can be characterized as being hydrophobic, whereby ritonavir is the most hydrophobic, while celecoxib is the least hydrophobic. The crystal forms of celecoxib and efavirenz seed crystals before crystal growth were confirmed to be the most stable Form III39 and Form I40 polymorph, respectively; these polymorphic forms were the intended starting materials. The PXRD pattern of celecoxib had distinctive peaks at 5.4, 14.9, 16.2, and 22.5 2θ (Supplementary Figure S1a) which are characteristic of the Form III polymorph, while the PXRD pattern of efavirenz had distinctive peaks at 6.0, 14.1, 21.1, and 24.8 2θ (Supplementary Figure S1b) which are characteristic of the Form I polymorph. These PXRD patterns are also consistent with the predicted powder pattern of the seed crystals (celecoxib CSD code: DIBBUL and efavirenz CSD code: AJEYAQ02), obtained from the Cambridge Structural Database using the powder pattern prediction tool in Mercury (CDD 3.1, Cambridge Crystallographic Data Center). In addition, the melting points of celecoxib (163.5 °C) and efavirenz (139.0 °C) seed crystals are consistent with reported literature values for their respective polymorphic forms.39,40 Representative SEM micrographs of celecoxib and efavirenz seed crystals are shown in Supplementary Figures S2a and S2b, respectively. Celecoxib seed crystals were plate-like, while efavirenz seed crystals were more rod-like. The BET surface areas of celecoxib, efavirenz, and ritonavir seed crystals were determined to be 1.145, 1.770, and 1.265 m2/g, respectively. A detailed description of ritonavir seed crystals is presented in refs 27 and 41.
Eg = E
Rg
o
Rg
p
(5) dx.doi.org/10.1021/mp400029v | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
where Rgo and Rgp are the initial bulk crystal growth rate in the absence and presence of polymer, respectively. Figure 3 shows
Figure 4. Crystal growth rate ratio of celecoxib at an initial concentration of 10 μg/mL. The data are arranged in order of hydrophobicity: least hydrophobic to most hydrophobic (left to right). Each column is an average (n = 3) of the effectiveness ratio of the polymers (5 μg/mL), and error bars indicate one standard deviation.
Figure 3. Seed-desupersaturation profiles of celecoxib in the absence (black squares) and presence (red circles) of an effective crystal growth inhibitor, CA 320S Sub 0.63, at an initial concentration of 10 μg/mL. The y-axis represents the amount of celecoxib in solution [solution concentration multiplied by volume of buffer (50 mL)]. The desupersaturation profiles (during the first two minutes) were fit to a line, which was used to estimate the initial bulk crystal growth rate of celecoxib. All experiments were performed in triplicate, and error bars indicate one standard deviation.
an example of the desupersaturation rate for seeded crystal growth in the presence and absence of a polymer at an initial celecoxib (CLB) concentration of 10 μg/mL. From these data, the effectiveness crystal growth rate ratio for the polymer shown, CA 320S Sub 0.63 (5 μg/mL), in inhibiting crystal growth of CLB at an initial concentration of 10 μg/mL was calculated to be ∼6.0; this polymer is quite an effective inhibitor for CLB at this supersaturation. Polymers with Eg < 1 were considered to be ineffective crystal growth inhibitors, while polymers with Eg > 1 were deemed as showing some growth inhibition. Figures 4, 5, and 6 show a comparison of the growth rate ratio of celecoxib, efavirenz, and ritonavir, respectively, at an initial concentration of 10 μg/mL in the absence of polymer (Rg0) to the growth rate in the presence of all the polymers investigated (Rgp). Although the impact of the majority of the polymers investigated in this study on crystal growth rate inhibition of ritonavir at a solution concentration of 10 μg/mL has been reported previously,27 the data are included in this study so that trends about effective crystallization inhibitors can be discerned. The polymers are arranged in order of increasing hydrophobicity (left to right based on decreasing SP values); the calculated SP values of the polymers tested in this study are summarized in Table 3. Some 16 polymers with different structures and properties were employed, 3 of which were synthetic polymers (white columns) while the other 13 were cellulose-based polymers (blue and red columns). A total of 11 of the 13 cellulose-based polymers were recently synthesized cellulose ω-carboxyalkanoate esters (red columns), designed to be effective amorphous solid dispersion polymers, spanning a wider range of hydrophobicities and chemical functionality than
Figure 5. Crystal growth rate ratio of efavirenz at an initial concentration of 10 μg/mL. The data are arranged in order of hydrophobicity: least hydrophobic to most hydrophobic (left to right). Each column is an average (n = 3) of the effectiveness ratio of the polymers (5 μg/mL), and error bars indicate one standard deviation.
commercially available cellulose derivatives.28 Out of the 16 investigated polymers, 13 were effective for efavirenz, 10 were effective for ritonavir, and 12 were effective for celecoxib ((Rg0/ Rgp) > 1), although to different extents. It should be noted that three of the in-house synthesized cellulose derivatives (CA 320S Sub 0.63, CA 320S Sub 0.90, and CA 320S Seb) were designed based on insights into the key polymer properties responsible for effective growth rate inhibition gained during an F
dx.doi.org/10.1021/mp400029v | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
Sub 0.63 > CA 320S Seb > CAP Adp 0.85 > CAB Adp 0.81. The second-generation polymer, CA 320S Sub 0.90, was the most effective cellulose-based polymer for both efavirenz and ritonavir, inhibiting crystal growth by a factor of 4.9 and 12.8, respectively. Interestingly, the hydrophilic and synthetic polymers, PVP and PVPVA, were effective crystal growth inhibitors for celecoxib and efavirenz but not for ritonavir, while PAA was ineffective for all of the model compounds. In summary, in general the most hydrophobic newly synthesized cellulose ω-carboxyalkanoate derivatives were ineffective in inhibiting crystal growth, while the moderately hydrophobic cellulose-based polymers (both commercially available and the new cellulose ω-carboxyalkanoate polymers) were effective crystal growth inhibitors for the model drug compounds. The very hydrophilic synthetic polymers were highly variable in effectiveness. Investigation of the Drug−Polymer Interactive ForcesImpact of Ionic Strength. To investigate the impact of nonspecific drug−polymer interactions on polymer effectiveness as crystal growth inhibitors, crystal growth experiments were performed at two different ionic strengths. If hydrophobic forces drive interaction of the polymer with the growing crystal, a higher ionic strength should promote interactions and hence polymer effectiveness due to the “salting out” effect.44 A comparison between the crystal growth effectiveness ratios of selected polymers at two different ionic strengths [50 mM (white columns) and 100 mM (red columns)] for celecoxib and efavirenz at an initial drug concentration of 10 μg/mL is shown in Figures 7 and 8. For
Figure 6. Crystal growth rate ratio of ritonavir at an initial concentration of 10 μg/mL. The data are arranged in order of hydrophobicity: least hydrophobic to most hydrophobic (left to right). Crystal growth rate experiments were performed in triplicate. Each column is an average of the effectiveness ratio of the polymers (5 μg/ mL), n = 3, and error bars indicate one standard deviation.27,29
initial study with ritonavir.28,41 These second generation novel cellulose-based polymers (see Table 5 for a summary of Table 5. Degree of Substitution (DS) of the SecondGeneration Polymers polymer abbreviation
DSa (CO2H)
DSb (Ac)
DS (total)
CA 320S Sub 0.90 CA 320S Sub 0.63 CA 320S Seb
0.90 0.63 0.57
1.82 1.82 1.82
2.72 2.45 2.39
a
DS (CO2H): DS of the carboxylic-bearing substituent; bAbbreviation: acetate (Ac).
properties) were much more effective at inhibiting the crystal growth of ritonavir compared to commercially available polymers as well as the best of the first generation cellulose derivatives that we investigated (CAP Adp 0.85 and CAB Adp 0.81) (Figure 6). For celecoxib, the moderately hydrophobic commercially available polymers were the most effective inhibitors (Figure 4): HPMC > HPMCAS > PVPVA; HPMC was a very effective crystal growth inhibitor of celecoxib, inhibiting crystal growth by a factor of 10.2, while the most effective of the newly synthesized cellulose derivatives was CA 320S Sub 0.90, which inhibited the growth by a factor of about 7. A set of chemically diverse polymers was effective in inhibiting crystal growth of efavirenz (Figure 5), whereby the two most effective polymers were PVPVA, a noncellulose derivative, and the newly synthesized cellulose derivative, CA 320S Sub 0.90. The best performing polymers were: PVPVA ≈ CA 320S Sub 0.90 > CAP Adp 0.85 ≈ HPMCAS, whereby PVPVA and CA 320S Sub 0.90 both inhibited growth by a factor of ∼5.0; it should be noted that, for efavirenz, the initial S was very low (1.2). In contrast, for ritonavir (Figure 6), the moderately hydrophobic newly synthesized cellulose-based polymers were most effective: CA 320S Sub 0.90 > CA 320S
Figure 7. Effectiveness crystal growth rate ratio of selected polymers (5 μg/mL) for celecoxib (10 μg/mL) at two ionic strength conditions (50 mM and 100 mM). Each column is an average of the effectiveness crystal growth rate ratio (n = 3), and error bars indicate one standard deviation.
both model compounds, the effectiveness of the cellulose based polymers was higher when the ionic strength was increased. In contrast, the effectiveness of PVPVA, a moderately hydrophilic and synthetic polymer that inhibited crystal growth for both compounds, was relatively insensitive to ionic strength suggesting that the drug polymer interaction is not driven by G
dx.doi.org/10.1021/mp400029v | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
Figure 8. Effectiveness crystal growth rate ratio of selected polymers (5 μg/mL) for efavirenz (10 μg/mL) at two ionic strength conditions (50 mM and 100 mM). Each column is an average of the effectiveness crystal growth rate ratio (n = 3), and error bars indicate one standard deviation.
hydrophobic forces for this polymer. The largest relative difference in the growth rate ratio at the two different ionic strengths was observed for HPMC, a 56% and 36% reduction in effectiveness for celecoxib and efavirenz at the lower ionic strength, respectively. Investigation of the Drug−Polymer Interactive ForcesFTIR Spectroscopy. The role of specific intermolecular interactions between celecoxib and efavirenz crystals and PVPVA was investigated using FTIR. Figure 9a and b shows the FTIR spectra (3600−2800 cm−1) of celecoxib and efavirenz seed crystals conditioned in the absence and presence of PVPVA. The IR stretching bands of the hydrogen bonded N− H units of celecoxib (sulfonamide functional group) and efavirenz (amide functional group) usually occur between 3500 and 3200 cm−1. The X−H stretching frequency in the aforementioned region is used to classify and characterize hydrogen bonds. 45,46 In general, hydrogen bonds are characterized by a decrease in wavenumber (red-shift) of the X−H peak due to lengthening of the bond.45,46 The N−H stretching peaks of celecoxib at 3331 and 3225 cm−1 shift to 3321 and 3217 cm−1, respectively, following the interaction with PVPVA (CLB-PVPVA spectrum), while other peaks remain essentially unchanged (Figure 9a). Similarly, the N−H stretching peak of efavirenz shifts from 3316 to 3308 cm−1 following the interaction with PVPVA (Figure 9b). Thus, the shifts in the N−H stretching peaks of celecoxib and efavirenz to lower wavenumbers suggest that specific intermolecular drug− polymer interactions occur between the drug molecules (celecoxib and efavirenz) and PVPVA, and these interactions appear to contribute to the overall effectiveness of PVPVA as a crystal growth inhibitor. Investigation of the Drug−Polymer Interactive ForcesEffect of Polymer Ionization. The effect of the ionization state of the synthesized cellulose polymers on their effectiveness as crystal growth inhibitors was investigated by comparing crystal growth rates for celecoxib at two pH
Figure 9. FTIR spectra of (a) celecoxib and (b) efavirenz in the absence and presence of PVPVA. The arrows indicate the N−H stretching band position before and after interaction with PVPVA.
conditions, 3.8 and 6.8. The ionization of the ω-carboxyalkanoate substituent groups will be nearly complete (∼100%) at pH 6.8,47 while they are less ionized (
