- Email: [email protected]
Mechanism of freeze-drying drug nanosuspensions
International Journal of Pharmaceutics 437 (2012) 42–50
Contents lists available at SciVerse ScienceDirect
International Journal of Pharmaceutics journal homepage: www.elsevier.com/locate/ijpharm
Pharmaceutical Nanotechnology
Mechanism of freeze-drying drug nanosuspensions Nae-Oh Chung, Min Kyung Lee, Jonghwi Lee ∗ Department of Chemical Engineering and Materials Science, Chung-Ang University, 221, Heukseok-dong, Dongjak-gu, Seoul, 156-756, Republic of Korea
a r t i c l e
i n f o
Article history: Received 11 April 2012 Received in revised form 25 June 2012 Accepted 20 July 2012 Available online 7 August 2012 Keywords: Nanoparticles Nanocrystals Freeze-drying Lyophilization Redispersibility Cryoprotectant Freezing rate Nanosuspension Nanoformulation PEG
a b s t r a c t Drug nanoparticles prepared in a liquid medium are commonly freeze-dried for the preparation of an oral dosage in solid dosage form. The freezing rate is known to be a critical parameter for redispersible nanoformulations. However, there has been controversy as to whether a fast or slow freezing rate prevents irreversible aggregation. A systematic investigation is presented herein regarding the effect of both the molecular weight of the cryoprotectant and the freezing rate in order to elucidate the mechanism underlying irreversible aggregation. It was found that irreversible aggregation occurred during drying rather than freezing, although a proper freezing rate is critical. A more homogeneous distribution of the cryoprotectant and drug nanoparticles led to more redispersible powders. Thus, keeping the local concentration distribution of the nanoparticles and cryoprotectant fixed during the freezing step plays a critical role in how the freezing rate affects the redispersibility. The kinetic approach of excluding the tendency of ice crystal growth permitted an explanation of the controversial results. This study will facilitate an in-depth understanding of the aggregation process of nanoparticles or proteins during freeze-drying. © 2012 Elsevier B.V. All rights reserved.
1. Introduction Freeze-drying, or lyophilization, is known to be the most effective unit operation to preserve perishable materials. The process was developed to preserve serum during WWII. Protein drugs are popular materials that require freeze-drying, and recently the extensive use of drug nanoparticles has increased the importance of drying technologies (Tang and Pikal, 2004; Haugh et al., 2010; O’Brien et al., 2004; Abdelwahed et al., 2006). The intrinsic energy penalty associated with nanosized dimensions requires special nanotechnologies for external energy and stabilization, which mostly produce liquid state products, i.e., nanodispersions or nanoemulsions (Abdelwahed et al., 2006; Jacobsand and Muller, 2002; Hill, 2001; Lee, 2004; Ploehn and Russel, 1990; Philip et al., 2003; Berglund et al., 2003a,b; Zhang et al., 2011; Wang et al., 2009; Zhou et al., 2010; Vyas et al., 2008). Therefore, a drying step is commonly needed to convert the liquid products into solid dosage forms. Unfortunately, phase and composition changes that occur in the conversion process often nullify the effects of external energy and stabilization (Jacobsand and Muller, 2002; Hill, 2001; Lee, 2004; Ploehn and Russel, 1990; Philip et al., 2003; Berglund et al., 2003a,b) in the preparation steps, resulting in the aggregation of nanoparticles. Irreversible aggregates of drug nanoparticles are
∗ Corresponding author. Tel.: +82 2 816 5269; fax: +82 2 824 3495. E-mail address: [email protected] (J. Lee). 0378-5173/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ijpharm.2012.07.068
unable to redisperse into nanoparticles upon dissolution and thus, the advantages of nanoformulation can be lost. Indeed, the development of suitable pharmaceutical unit operations for nanoparticles, such as drying, granulation, and compaction, is still a challenging task. The successful preparation of redispersible nanopowders that readily revert back to nanosuspensions upon reconstitution in water requires delicate process control and certain processing windows. Extensive research has been conducted on the relationship between processing variables and the redispersibility of final solid powders (Ho and Lee, 2011; Kesisoglou et al., 2007; Vergote et al., 2001; Kim and Lee, 2010). However, similar to protein drug lyophilization, the formulation of a detailed mechanism and an indepth understanding of this relationship has yet to be achieved. Most research has focused on the choice of cryoprotectant, temperature profile, concentration and vacuum pressure. In the freeze-drying process, a wide variety of cryoprotectants, such as glycerol, quaternary amines, carbohydrates, and synthetic polymers are used to protect nanoparticles from stresses and subsequent aggregation. It is generally known that a higher concentration of cryoprotectant and faster freezing result in better nanoparticles redispersibility. However, it has been reported that the use of a common cryoprotectant destabilized nanoparticles in some cases. Kamiya et al. investigated the effect of interactions between lipid nanoparticles and saccharides on freeze-drying (Kamiya et al., 2010) and found that the crystallization of a saccharide excluded drug-lipid nanoparticles, resulting in irreversible
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
aggregation. With an increase in the cryoprotectant concentration, the irreversible aggregation of nanoparticles may either increase or decrease depending on the system of nanoparticles and the freezedrying conditions. Controversy also exists regarding the freezing rate; a fast freezing rate (low freezing temperature) may allow for better redispersibility, but in other cases, a slow freezing rate yields better results. Thus far, the selection of cryoprotectants and related processing conditions has largely relied on empirical approaches. Previous reports focused on the freezing rate, which had mostly been neglected in earlier research. To facilitate an in-depth understanding, a control method was developed for the freezing rate by adjusting the temperature gradient. The freezing rate is closely related to the temperature profile of the actual industrial unit operation. By systematically controlling the freezing rate for nanocrystalline systems of water-insoluble drugs, it was found that, without any other changes, only variations in the freezing rate could result in significantly different redispersibilities. We defined a critical freezing rate above which nanoparticles can retain their redispersibility (Lee and Cheng, 2006). However, depending on the cryoprotectant and processing conditions, the successful processing of redispersible nanopowders requires faster or sometimes slower freezing rates (Lee et al., 2009). The choice of cryoprotectant is also critical. In the case of carrageenan, only 0.5 wt% could successfully protect naproxen nanocrystals (Kim and Lee, 2010). Although we revealed the importance of the freezing rate in the freeze-drying process, the mechanism behind the phenomenon is not yet well-understood. During freezing, ice crystals nucleate and grow while excluding most of the solutes into a cryoconcentrate phase. In the cryoconcentrates, further phase separation or crystallization of a component can occur. Therefore, by focusing on the parameters related to the phase separation and crystallization phenomena, a more in-depth mechanism may be elucidated. First of all, kinetic phenomena are most likely dependent on the diffusion characteristics of the nanoparticles and cryoprotectant. Therefore, in addition to the control of the freezing rate, the molecular weight of the cryoprotectant was systematically varied in this study to investigate its effect on diffusivity. The diffusivity issue was highlighted from the perspective of frozen structure development (internal spatial distribution) in order to reveal the mechanism of irreversible aggregation among nanoparticles. In particular, we hypothesized that morphological characteristics developed during freezing influence subsequent irreversible aggregation. 2. Materials and methods 2.1. Materials Naproxen, a relatively insoluble crystalline drug compound (API, Mw = 452.8 g/mol, purity >95%, Tokyo Chemical Industry, Japan) was used as a model drug, while hydroxypropyl cellulose (HPC, Mw = 60 kg/mol, surface energy = 45 mN/m, FMC, Philadelphia, PA, USA) was employed as a steric stabilizer for wet comminution. Ramda-carrageenan (non-gelling at 1% in a 0.2 M KCl solution), sucrose (purity >98%), and polyethylene glycol (PEG, Mn = 200–35,000 g/mol) obtained from Aldrich (St. Louis, MO, USA) were used as cryoprotectants. HPLC-grade water from Aldrich and yttria-stabilized zirconia beads (1 mm diameter, Performance Ceramics, OH, USA) were employed without further purification. 2.2. Preparation of nanocrystal dispersions Low energy comminution was used to produce nanocrystal dispersions with 1 mm zirconia beads (50 mL) as the grinding media. API was mixed with HPC previously dissolved in water (100 mL), and the beads and additional water were put into a glass bottle.
43
Fig. 1. Schematic illustration of a controlled freezing device with a temperature gradient, which decreases as a function of distance from the liquid nitrogen reservoir. Freezing starts at the region close to the reservoir, and the freezing rate decreases as the freezing front moves away from the reservoir.
The total weight of the slurry (API + stabilizer + water) was 8.4 g, and the concentration of API in water was 15 wt%. The weight ratio of HPC:drug was kept constant at 1:6. The comminution speed and time were 125 rpm and five days, respectively. After comminution at room temperature, the media beads were filtered out and the suspension was stored at 5 ◦ C. The cryoprotectants were then introduced into the nanocrystal suspension 24 h before the subsequent experiment. The solution concentration of cryoprotectant was varied from 5 to 7.5 wt%. 2.3. Controlled freeze-drying Nano-suspensions were frozen using a custom-made apparatus (Fig. 1) following a previously outlined procedure (Lee and Cheng, 2006; Lee et al., 2009). The freezing rate was varied as a function of the distance from the liquid nitrogen reservoir in the sample, thus reflecting changes in the temperature gradient, which was repeatedly checked as outlined in earlier studies (Lee and Cheng, 2006; Lee et al., 2009). As shown in Fig. 1, nano-suspensions were first put into a long tube, and the tube was subsequently placed in contact with a liquid nitrogen reservoir. After contact, freezing began from the region near the liquid nitrogen reservoir. Both the container and the reservoir were made of copper to enable rapid heat transfer. In addition to heat transfer through the copper, cold nitrogen gas flow aided in the propagation of the freezing front. Because the frozen portions of the suspension had different turbidities when compared to the liquid portions, the freezing interface between the frozen and liquid portions was distinct (Lee et al., 2009). Freezing rates could be calculated by monitoring the change in position of the freezing front with respect to time. To obtain reproducible results, all other experimental factors, such as the amount of liquid nitrogen, were kept constant. The frozen suspension was vacuumdried using the FD-1000 bench top freeze-dryer (EYELA, Tokyo, Japan, trap chilling temperature of −55 ◦ C, 6.1 Pa) for 3–24 h (unless otherwise indicated). 2.4. Characterization The particle size was measured with a Horiba LA 910 laser light scattering analyzer (632.8 nm He-Ne laser) after dispersing the particles in 150 mL of water and sonicating the mixture for 1 min. A relative refractive index of 1.2 ± 0.00i and a 1 min sonication of 40 W and 39 kHz were employed. After freeze-drying, disk-shaped powder samples (5.1 mm diameter) taken at different locations along the long tube were used for characterization (Fig. 1). The volume-averaged particle sizes of the powders were used in this study.
44
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
The morphology of the freeze-dried powders was characterized by optical microscopy (Olympus BX–51); quantitative data for the thickness of ice crystals were obtained at 30 different locations. The morphology of the freeze-dried powders was further examined with a Hitachi S-4300E scanning electron microscope (SEM, 3–4 kV, 3 nm platinum coating). The surface morphology of the powders was investigated by atomic force microscopy (AFM, XE100, Park Systems, South Korea) using a silicon cantilever (NSC15, MikroMasch, Spain) with a spring constant of 30.05 (±3.7) N/m. The distributions of PEG and naproxen were mapped with an IRT5000 FT-Raman microscope system (surface reflect-type, Jasco, Easton, MD, USA) capable of FT-IR (FT-IR 6100 – type A) and Raman (NRS-3000) measurements in order to elucidate their diffusion mechanism during freezing and drying. The FT-IR wavenumber ranged from 400 to 2500/cm (reflection-type microscope, accumulation: 200, resolution: 4/cm, scanning speed: 32 m/s), while the Raman wavenumber ranged from 1000 to 2000/cm (exposure time: 5 s, accumulation: 200, laser source: 532 nm ND:yak, resolution: 1.85/cm). The crystal structure of the dried powder was analyzed with an X-ray diffractometer (XDS 2000, Scintag, Santa Clara, CA, USA) at a scan rate of 1◦ /min (2: 5–40◦ ). Differential scanning calorimetry (DSC, Pyris 6, PerkinElmer, USA) data were also acquired. For the DSC measurements, 10 mg of sample was scanned twice in a N2 /air atmosphere at a heating rate of 10 ◦ C/min and a cooling rate of 20 ◦ C/min from −60 to 180 ◦ C. The viscosity of the cryoprotectant was measured with a rheometer (Brookfield RH10 capillary rheometer, Middleboro, MA, USA) at a fixed shear rate (10/s) and temperature (25 ◦ C); the average of 15 measurements was used. The zeta potential was measured with a dynamic electrophoretic light scattering spectrophotometer ELS-Z2 (Otsuka Electronics, Japan; average electric field: −16.53 V/cm, conversion equation: Smoulchowski) at 25 ◦ C. 3. Results 3.1. Effect of freezing rate and PEG molecular weight To gain a systematic understanding of the freezing mechanism of drug nanosuspensions, a controlled freezing apparatus with freezing rate control capability was designed. In conventional freeze-drying, freezing starts from the coldest region, typically the surface region of a vial, and depending on the imposed temperature gradient, freezing propagates into the rest of the liquid
phase at a poorly controlled freezing rate. Controlled freezing experiments were conducted utilizing the temperature gradient surrounding the liquid nitrogen reservoir (Fig. 1). In contrast to previous reports (Lee and Cheng, 2006; Lee et al., 2009), a flow duct was introduced to guide nitrogen gas flow and limit potential turbulent air flow from the outside environment. This improved the reproducibility of the freezing rate. The change in turbidity of the nanosuspensions was recorded as a function of time, thus providing information on the movement of the freezing front, which can be taken as the freezing rate. Following previous reports (Lee and Cheng, 2006; Lee et al., 2009), we confirmed the reproducibility of freezing rate control and the imposed temperature gradients as a function of distance from the liquid nitrogen reservoir. The effect of the freezing rate on the redispersibility of drug nanosuspensions in the presence of PEG with different molecular weights is shown in Fig. 2. Once particle aggregation occurs, the average particle size quickly reaches several microns from the original particle size of 148 nm; therefore our focus was on the transition between these two sizes, which corresponds to nanoparticle aggregation (Lee and Cheng, 2006; Lee et al., 2009). For 8 and 20 kg/mol PEG (Fig. 2(a)), faster freezing rates resulted in redispersed drug particles with a larger particle size. In other words, slower freezing was better than faster freezing for drug nanocrystal dispersions between 0 and 700 m/s. This is consistent with the results of previous reports on mannitol and PEG (Lee et al., 2009), but quite contrary to both the conventional understanding of freeze-drying and what is observed during the freeze-drying of drug nanosuspensions without cryoprotectant (Lee and Cheng, 2006). The molecular weight of polymer cryoprotectants is an important parameter that determines diffusion, entanglement density, viscosity and steric repulsion. One of the goals of this work was to investigate how an increase in the molecular weight affects the redispersibility of nanosuspensions. As shown in Fig. 2(a), it was found that PEG with a higher molecular weight prevented the aggregation of drug nanoparticles more effectively in a range of relatively low freezing rates. This was further confirmed in Fig. 2(b), where only the lowest molecular weight case (PEG 0.4k) showed significant aggregation, which manifested as a larger particle size. Although the particle size was not linearly dependent on the molecular weight, there seemed to be a transition in particle size caused by an increase in molecular weight. When comparing Fig. 2(a) and (b), it is evident that an increase in the PEG content
Fig. 2. Effect of the freezing rate on the volume-averaged particle size of redispersed drug nanoparticles after freeze-drying in the presence of (a) 5 wt% and (b) 7.5 wt% PEG (Mn = 0.4–20 kg/mol).
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
45
Fig. 3. Effect of the drying time on the volume-averaged particle size of redispersed drug nanoparticles after freeze-drying in the presence of 5 wt% PEG (Mn = 8 kg/mol).
from 5 to 7.5 wt% was followed by a significant improvement in redispersibility (except PEG 0.4 K and some low freeze rate cases). This effect of the cryoprotectant concentration is consistent with that observed in most previous studies (Lee et al., 2009). For comparison, 1 wt% carrageenan was used in place of PEG, and successful cryoprotection was observed. The resulting particle sizes were all below 0.2 m (0.15, 0.17, 0.17, 0.17, 0.18, 0.17, and 0.19 m for 588, 344, 222, 105, 86, 11, and 9.6 m/s, respectively). This result is also consistent with previous reports that carrageenan provides effective cryoprotection (Kim and Lee, 2010).
Fig. 4. Effect of the freezing rate on the thickness of ice crystals in frozen nanosuspensions in the presence of various cryoprotectants (5 wt%). The inlet OM images show the freeze-dried drug nanosuspensions (PEG 8k), with arrows denoting the direction of freezing propagation (scale bar = 100 m).
3.2. Occurrence of irreversible aggregation Determining the mechanism of irreversible aggregation and the effect of the freezing rate on this mechanism is not a simple task. However, Fig. 3 aids in providing an in-depth understanding of aggregation. The drug nanosuspensions with PEG 8k were frozen at different freezing rates and then freeze-dried for
Fig. 5. Typical SEM micrographs of naproxen/PEG freeze-dried powders (freezing rate = 150–200 m): (a) PEG 0.4k (5 wt%) and (b) PEG 20k (7.5 wt%).
46
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
different time periods. After subsequent reconstitution of the samples in water, the degree of irreversible aggregation was assessed by measuring the average particle sizes. No significant aggregation occurred during the initial three hours of drying. However, irreversible aggregation continually progressed up to 24 h, especially in the fast freezing rate ranges. At freezing rates between 200 and 700 m/s, the most significant difference in the degree of aggregation occurred between three and six hours, indicating that irreversible aggregation occurs primarily during the sublimation process due to internal changes, e.g., desiccation stress, heat of sublimation, etc. Although irreversible aggregation occurred during the drying step and not in the freezing step, aggregation still depended on the freezing rate. Therefore, while the frozen nanosuspensions did not aggregate before drying, they seemed to have different internal structures depending on the freezing rate. Certain internal structures are more susceptible to desiccation stress and/or heat of sublimation, which can cause irreversible aggregation. Thus the differences in the internal structure likely resulted in different degrees of aggregation.
3.3. Morphology of frozen nanosuspensions The internal structure of the freeze-dried powders from nanosuspensions was investigated via OM, SEM, and AFM. Typical OM micrographs of freeze-dried powders are shown in Fig. 4. Along the direction of freezing propagation, ice crystals grew as columnar or lamellar phases. This phase separation between ice and cryoconcentrates occurs through fingering of the freezing front (Colard et al., 2009; Deville et al., 2007; Uhlmann et al., 1964). The width of the columnar ice crystals distinctly depends on the freezing rate (Rohatgi and Adams, 1967; Lee et al., 2010a). As the freezing rate increases, the width inversely decreases (as shown in Fig. 4), regardless of the sample. Compared to the pure drug nanosuspension, the addition of a cryoprotectant increased the width of the ice crystals. Among the cryoprotectants used in Fig. 4, carrageenan yielded the largest increase in ice crystal width, followed by sucrose. A marginal difference was observed in samples containing PEG of different molecular weights. A lower molecular weight resulted in ice crystals with a larger width. The fingering phenomenon seemed to develop further in the samples containing lower molecular weight PEG. The ice crystal width does not directly support the tendency toward irreversible aggregation, since the results in Fig. 4 are not systematically correlated with the re-dispersion particle size results. However, phase separation indirectly determines the composition of the cryoconcentrate and the desiccation stress.
SEM images of well (Fig. 5(b)) and poorly dispersible (Fig. 5(a)) samples due to a difference in molecular weight show the typical surface features of dried cryoconcentrate regions facing ice crystals. In the left micrographs taken at low magnification, there is no significant difference in morphology related to redispersibility. At a higher magnification (right), a slight difference is apparent. Surface roughness was detected in the well redispersible samples at a slightly shorter wavelength (submicron) that more clearly showed the individual drug nanoparticles (Fig. 5(b)). Data regarding the difference in surface roughness and the identification of individual nanoparticles was obtained with AFM scans of PEG 8k. The AFM images revealed that the slower freezing rate generally produced surface roughness with a slightly shorter wavelength. The sample in Fig. 6(c), which is a readily redispersible powder, has the shortest wavelength corresponding to the size of the individual nanoparticles. In Fig. 6(a) (fast freezing, poorly dispersible), the wavelength of the surface roughness was larger than the size of the primary nanocrystals. When the PEG and nanoparticles were homogeneously mixed, the individual nanoparticles were more easily observed on the surface. In poorly redispersible samples (Figs. 5(a) and 6(a)), the surface of the cryoconcentrate seemed to have PEG that was more inhomogeneously dispersed. These findings suggest that the wavelength of the surface roughness could be an indirect sign of irreversible aggregation due to a lack of cryoprotectant between the nanoparticles.
3.4. Distribution of cryoprotectants and drug nanoparticles A morphological examination itself cannot reveal the local distribution of drug nanoparticles and cryoprotectants. To elucidate the detailed internal structure, both FT-IR and Raman mapping techniques were employed; typical results of FT-IR mapping are given in Fig. 7. The concentrations of PEG and nanoparticles varied perpendicular to the growth direction of the ice (downward) due to the fingering of the freezing front, while the wavelength of the concentration fluctuation seemed to be related to the size of the ice crystals. Interestingly, the position of the local maxima of the PEG was always different from that of the drug nanoparticles. The mismatch between the two maxima indicated that the rejection processes of the drug nanoparticles and PEG from the freezing front were different, which resulted in two regions within the cryoconcentrate phase: a nanoparticle-rich region and a PEG-rich region. The two regions were mainly formed kinetically, not thermodynamically, although there was no clear distinction between the two mechanisms. In the case of a slow freezing rate, the mismatch of the local maxima of the two components was not greater than the
Fig. 6. AFM micrographs showing the effect of the freezing rate on the morphology of dried powders (PEG 8k): (a) freezing rate = 590 m/s, (b) 100 m/s, and (c) 9 m/s.
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
47
rate cases. In other words, PEG cannot be homogeneously inserted between drug nanoparticles and thus it cannot prevent irreversible aggregation in this case. 3.5. Viscosity of nanosuspensions The results of viscosity measurements of the nanosuspensions with cryoprotectants and their controls (solutions without drug nanoparticles) are shown in Fig. 8. Shear thinning behavior was evident in all of the samples in Fig. 8(a). Carrageenan of only 1 wt%, which provided successful cryoprotection efficiency in a previous report (0.15–0.19 m particle sizes at 9.6–588 m/s freezing rates) (Kim and Lee, 2010), had the highest viscosity, followed by PEG and sucrose. Sucrose of 5 wt% was not effective for cryoprotection, which corroborates previous work (Lee et al., 2009). As can be inferred from the particle size data in Fig. 2, PEG 8k was more viscous than PEG 1k. Therefore, the viscosity results have a simple relationship with the redispersibility results: the higher the viscosity, the better the redispersibility results. The viscosity may not be the only determining factor, but it does have a major influence on cryoprotection. In particular, subsidiary phase separation into PEG-rich and nanoparticle-rich regions in cryoconcentrates appear to be dependent on the viscosity of the medium.
Fig. 7. FT-IR microscopy mapping images (30 × 20 m) showing the freezing rate dependence of the distribution of naproxen and PEG 8k (5 wt%) in freeze-dried powders. The characteristic peaks of naproxen (1733 cm−1 ) and PEG 8k (1355 cm−1 ) were used. The dotted lines indicate the local maxima of the concentration.
wavelength of the ice crystals (Fig. 7(c)). Still, the position of the local maxima seemed to follow an alternating turn between the PEG and drug nanoparticles. Due to the magnification limitation of FT-IR, the wavelength of the ice crystals is not fully covered in Fig. 7(c), but it was approximately 80 m (Fig. 4). The mismatch became more irregular as the freezing rate increased. In the case of a fast freezing rate, there were five and four local maxima for the nanoparticles and PEG, respectively (Fig. 7(a)). The wavelength of freezing front fingering could not be traced from the wavelength of the PEG or nanoparticle local maxima, which was about 5 m (Fig. 4). Therefore, it was observed that a fast freezing rate induced a more irregular accumulation of PEG and nanoparticles in the cryoconcentrates, resulting in a more inhomogeneous distribution. This explains the poor redispersibility in the fast freezing
3.6. Interactions between drug and PEG Several characterizations were conducted in order to determine the detailed aggregation mechanism. First, the molecular interactions between naproxen and PEG were examined using FT-IR. However, no significant evidence of the interactions was observed (data not shown), probably because the interactions were confined only to the surface of the naproxen nanoparticles. This result was also consistent with the XRD and DSC findings, where no meaningful differences were noted (data not shown). Similar to the results of a previous report (Lee et al., 2009), the crystalline phase of naproxen detected by XRD and DSC was not influenced by the freezing rate or the existence of PEG (Anon., 2012). Zeta potential values of the drug nanoparticles were also measured in order to detect differences in electrostatic repulsion and possible surface charge effects on redispersibility. No significant difference in zeta potential, which may be related to redispersibility, was identified. In addition, the values obtained before freeze-drying (naproxen = −16.77, naproxen/PEG 8k = −17.79, and
Fig. 8. Viscosity of (a) drug nanosuspensions containing various cryoprotectants and (b) pure cryoprotectant solutions.
48
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
crystal growth (6). Eventually, the liquid phase undergoes vitrification at a maximally concentrated state. The cryoprotectant molecules can form their own phase (crystalline or amorphous) during this step, although it is not preferable for good redispersibility. These phase developments can be restricted by kinetic factors. The key mechanisms underlying irreversible aggregation are chain entanglement in a steric stabilizer (HPC) and drug particle fusion. For better redispersibility, a highly water soluble cryoprotectant, present between the drug nanoparticles, should homogeneously cover and effectively protect the surface of the drug nanoparticles from possible entanglement and fusion. Therefore, the homogeneity of the distribution and the degree of protection are the main factors to be discussed. The distribution of the drug nanoparticles and cryoprotectant in cryo-concentrated phases was not found to be entirely homogeneous. First, differences in diffusivity can cause inhomogeneity. Although the solubility of a solute in ice is almost negligible, the nanoparticles or polymers (solutes) in solution can be engulfed by the ice phase as the freezing rate increases. If a polymer chain is considered to be a particle and there is no interaction among particles, which is obviously an oversimplification, then simple theoretical principles can be developed for flat interfaces. By considering the surface energy, mass balance, and diffusion equations, the critical freezing rates below which the cryoprotectant (PEG) or drug nanoparticles will be excluded, c,p and c,n , respectively, are Fig. 9. Schematic illustration of structure development during the freezing of drug nanosuspensions containing a cryoprotectant. Cryoprotectants and drug nanoparticles can be engulfed in an ice crystal phase depending on the freezing rate () and critical freezing rate (c,p and c,n ).
naproxen/sucrose = −15.62 mV) did not differ significantly from the values acquired after re-constitution in water (naproxen/PEG 8k = −14.34 and naproxen/sucrose = −16.88 mV). Regarding the stability of the suspensions due to steric and electrostatic repulsion, a rough guideline of 20 mV for the zeta potential has been suggested (Jacobsand and Muller, 2002). However, this does not take into account the complexity of steric and electrostatic repulsion. As the particle size decreases, steric repulsion plays a more important role than electrostatic repulsion, which relies on many variables such as molecular weight, adsorption and desorption kinetics and chain flexibility (Lee et al., 2010b). In our research, a zeta potential of −14 to −18 mV was sufficient to provide stable suspensions up to six months at room temperature and no significant size changes after six months. 4. Discussion 4.1. Distribution of drug nanoparticles and cryoprotectants The freeze-drying process is composed of two steps: freezing and drying at a low temperature. Freezing, the first step in freeze-drying, fixes the morphological and local compositional characteristics of the ice crystals and cryoconcentrates, which is critical in preventing irreversible aggregation. As ice crystals nucleate and grow following the direction of the temperature gradient, cryoprotectant molecules and drug nanoparticles are rejected from the ice phase into a cryo-concentrated liquid phase (Fig. 9). For a particle to be rejected by the ice phase, the surface free energy at a minimum separation, = sp − ( lp + sl ), should be positive, where sp , lp , and sl are the interfacial free energies of the solid-particle, liquid-particle, and solid–liquid interfaces, respectively (Deville et al., 2007; Uhlmann et al., 1964). The composition of the liquid phase follows the equilibrium phase boundary determined by thermodynamics in an ideal situation with ice
c,p =
c,n =
∇ p d ao 3rp
(1)
d
∇ n d ao 3rn
n
d
n (2)
where d is the distance between the particle (or polymer) and the interface (freezing front), is the viscosity of water, r is the radius of the particle (or polymer), ao is the intermolecular distance of water, and n is an exponent that is larger than 1 (Deville et al., 2007; Uhlmann et al., 1964). The critical freezing rate represents the excluding tendency of the solutes. Eq. (2) conveniently provides a basic understanding of the effects of influencing factors, such as the particle size (Deville et al., 2007; Uhlmann et al., 1964). If there is no further phase separation in the cryoconcentrated liquid phase, then the rejected solutes, namely the drug nanoparticles and cryoprotectants, at the moving front of the ice crystals will be homogeneously mixed in the cryoconcentrated regions. On the other hand, if one component is engulfed by ice crystals while others are not, then the formation of cryoconcentrate will entail an inhomogeneous distribution depending on the relative magnitude of the freezing rate (), c,p and c,n . Four different cases of freezing rates are schematically illustrated in Fig. 9. First, if the freezing rate, , is smaller than c,p and c,n , then both the drug nanoparticles and cryoprotectant will be rejected by the growing ice phase and form a homogeneous cryoconcentrate phase. Local fluctuations in the concentration will also be minimized. This could be the case for many instances of conventional freeze-drying, since the typical values of c,p and c,n are relatively large (Deville et al., 2007; Uhlmann et al., 1964). In this case, conventional knowledge of freeze-drying is appropriate. For example, the redispersibility will improve as the freezing rate increases, since smaller crystals form with reduced freezing and desiccation stress. A higher concentration of cryoprotectant would also be beneficial. The other extreme case occurs when the freezing rate is larger than both c,p and c,n (Fig. 9(d)). An extremely fast freezing rate appears to be rare. However, a rather homogeneous distribution can still result, preventing irreversible entanglement and particle fusion.
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
The size of the drug nanoparticles is at least one order of magnitude larger than that of PEG (radius of gyration of 20k PEG = 7 nm (Lee et al., 2008; Morar et al., 2006)). Thus, c,p may be larger than c,n , but the critical velocities of the drug nanoparticles (c,n ) and PEG (c,p ) are still difficult to evaluate since other surface free energy parameters are simply not available. Furthermore, if the interactions between the particles are considered (many particle systems), the viscosity term () should be a function of the concentration and the interactions. As the freezing rate increases, a solute of low diffusivity will first be engulfed inside the ice crystal phases, resulting in an inhomogeneous distribution of drug nanoparticles and cryoconcentrates. As the freezing rate increases above c,n , more drug nanoparticles will be trapped in ice crystals forming their own phases, which results in a poorly redispersible case (Fig. 9(b)). The accumulation of cryoprotectants outside the cryoconcentrate regions is illustrated in Fig. 9(c). In such a scenario, more drug nanoparticles come into direct contact with each other. These two cases explain our observation of the unique effects of freezing rate and molecular weight. If one of the cases occurs as the freezing rate increases, the effect of the freezing rate will be contrary to conventional knowledge. As the molecular weight of the PEG increases, r and increase, and c,p seems to get closer to c,n . As a result, more homogeneous distributions could form. This explanation may exclude the possibility of the mechanism illustrated in Fig. 9(c), as can be inferred by the particle size difference. It is intuitive that excluding a drug nanoparticle, which is far larger than a polymer chain, is usually quite easy. Previous reports on the unexpected effect of the freezing rate, which primarily found an increase in the redispersibility with a decrease in the freezing rate, have supported explanations regarding specific interactions between particles and cryoprotectants. The type of cryoprotectant and its interactions with drug nanoparticles are considered in the parameter in our treatment. The aforementioned kinetic approach is based on the simple view of competition between the critical velocity and freezing rate. However, if phase separation or a concentration fluctuation develops in the cryoconcentrated region, then other parameters such as the phase behavior of the concentrated solution, local temperature fluctuations, and entanglement fluctuations should also be considered.
4.2. Microstructure and local distribution The flat interface of the freezing front excludes solutes and develops a significant concentration gradient. The overall freezing rate is determined by the rate of heat extraction, while the local growth rate is limited by the low liquid diffusivity. To overcome the mismatch between the overall freezing rate and local growth rate, the flat freezing front takes on a dendritic shape (Kurz and Fisher, 1981). As ice crystals grow, heat flowing from the moving interface into the remaining unfrozen solution causes supersaturation and supercooling to occur. As heat transfer rates are normally much greater than mass transfer rates, nanoparticle or cryoprotectant diffusion tends to control the growth of the ice crystals. The accumulation of rejected nanoparticles and cryoprotectant will further decrease solute mobility, resulting in an engulfment of cryoconcentrated regions. Based on a materials balance and Fick’s law, theoretical equations to describe the inter-dendrite spacing in the case of dilute solutions have been proposed. Rohatgi and Adams (Rohatgi and Adams, 1967) suggested the following expression for the spacing between dendrites (ice crystals): L = (8DT )
0.5 −0.5
F
,
(3)
49
where D is the binary diffusion coefficient related to the concentration term, T is the supercooling between dendrites, and F is the freezing rate (Rohatgi and Adams, 1967). The researchers demonstrated that supersaturation and supercooling are independent of the freezing rate, as well as the relationship between L and D or F. Eq. (3) does not consider interactions between particles (polymers), the particle size distribution, the viscoelastic properties of the slurry, latent heat diffusion, or partial solidification in the cryo-concentrated regions. However, it quantitatively explains the results shown in Fig. 4. As the viscosity of a solution increases due to an increase in the molecular weight, D and L both decrease. The effect of the freezing rate also obeys Eq. (3) since an increase in F (freezing rate) is followed by a decrease in L. However, the scaling parameters of ‘−0.5’ were not precisely confirmed in our experiment. Note that −0.65, −0.68, −0.71, −0.35, −0.73, and −0.81 were obtained for cases without PEG, PEG 1 K, PEG 8 K, PEG 20 K, SUC, and CAR, respectively. In addition, the freezing rate, a purely experimental parameter in this study, is not strictly F in Eq. (3), but the two parameters are proportional to each other. As shown in Fig. 4, the redispersibility does not directly reflect changes in L. Therefore, the local distribution of the cryoprotectant and nanoparticles is more important than the pore morphology of a dried cake (ice crystal morphology), such as L. The freezing step determines the internal structures of the cryoconcentrated phases of a frozen cake where molecular mobility is restricted. In the cases of poor redispersibility, the subsequent drying process brings drug nanoparticles in close contact with each other, where actual irreversible aggregation occurs. This aggregation process depends on the local distribution of the cryoprotectant determined in the previous freezing step. The analysis and conclusions of this study should be generalized with caution. Although the analysis was based on a series of investigations regarding the effects of the freezing rate using drug nanoparticles with and without sucrose, mannitol, lactose, and PEG (Lee et al., 2009), generalization may require more extensive experiments using various materials and conditions. 5. Conclusions The redispersibility of dried nanoparticles is a critical factor, and poor redispersibility may arise due to irreversible aggregation. Unfortunately, redispersibility is an unpredictable factor; it decreases in some cases and increases in other cases with an increase in the freezing rate. For the development of redispersible solid formulations, the underlying mechanism of irreversible aggregation was investigated in detail in this study. The molecular weight of cryoprotectants and the freezing rate were systematically varied, and the redispersibility and distribution of the cryoprotectant were investigated. The redispersibility of nanoparticles increased with an increase in the molecular weight of the cryoprotectant. A slow freezing rate was better for the redispersibility of nanoparticles in the case of PEG. These results were explained by a kinetic approach based on the excluding tendency of the solutes (represented by the critical freezing rate) from growing ice crystals. Differences in the excluding tendencies of the cryoprotectant and drug nanoparticles could explain the dependence of the redispersibility on both the freezing rate and the molecular weight. The spatial distributions of PEG and naproxen, as mapped by FT-IR, allowed for the elucidation of their diffusion mechanism during freezing and drying. Acknowledgements This work was supported by a grant (10035574) from the “Platform Chemical Process Technology from Lignocellulosic Biomass Conversion” R&D Program funded by the Ministry of Knowledge
50
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
Economy, grants from the National Research Foundation of Korea (NRF) funded by the Korea government (MEST) (2012-014107), and the Korea Healthcare technology R&D Project, Ministry of Healthcare & Welfare (A103017). MKL would like to thank the Human Resource Training Project for Strategic Technology (MKE & KOTEF). References Abdelwahed, W., Degobert, G., Stainmesse, S., Fessi, H., 2006. Freeze-drying of nanoparticles: formation, process and storage considerations. Adv. Drug Deliv. Rev. 58, 1688–1713. The Tm of PEG 8k (5 wt%) in the first scan showed a slight increase in the cases of fast and slow freezing rates (55.02 ◦ C (8.75 J/g)) for 590 m/s and 54.95 ◦ C (8.80 J/g) for 9 m/s) when compared to the case of a middle freezing rate (52.29 ◦ C, 7.89 J/g, 100 m/s) while the Tm of naproxen (155.09, 155.11, and 155.95 ◦ C for 590, 100, and 9 m/s respectively) and PEG in the second scans (52.07, 52.18, and 52.50 ◦ C for 590, 100, and 9 m/s respectively) were not changed. Berglund, K.D., Przybycien, T.M., Tilton, R.D., 2003a. Coadsorption of sodium dodecyl sulfate with hydrophobically modified nonionic cellulose polymers. 1. Role of polymer hydrophobic modification. Langmuir 19, 2705–2713. Berglund, K.D., Przybycien, T.M., Tilton, R.D., 2003b. Coadsorption of sodium dodecyl sulfate with hydrophobically modified nonionic cellulose polymers. 2. Role of surface selectivity in adsorption hysteresis. Langmuir 19, 2714–2721. Colard, C.A.L., Cave, R.A., Grossiord, N., Covington, J.A., Bon, S.A.F., 2009. Conducting nanocomposite polymer foams from ice-crystal-templated assembly of mixtures of colloids. Adv. Mater. 21, 2894–2898. Deville, S., Saiz, E., Tomsia, A.P., 2007. Ice-templated porous alumina structures. Acta Mater. 55, 1965–1974. Haugh, M.G., Murphy, C.M., O’Brien, F.J., 2010. Novel-freeze-drying methods to produce a range of collagen-glycosaminoglycan scaffolds with tailore mean pore sizes. Tissue Eng. Part C: Methods 16, 887–894. Hill, T.L., 2001. A different approach to nanothermodynamics. Nano Lett. 1, 273–275. Ho, H., Lee, J., 2011. Redispersible drug nanoparticles prepared without dispersant by electro-spray drying. Drug Dev. Ind. Pharm., 1–8. Jacobsand, C., Muller, R., 2002. Production and characterization of a budesonide nanosuspension for pulmonary administration. Pharm. Res. 19, 189–194. Kamiya, S., Kurita, T., Miyagishima, A., Itai, S., Arakawa, M., 2010. Physical properties of griseofulvin-lipid nanoparticles in suspension and their novel interaction mechanism with saccharide during freeze-drying. Eur. J. Pharm. Biopharm. 74, 461–466. Kesisoglou, F., Panmai, S., Wu, Y., 2007. Nanosizing-oral formulation development and biopharmaceutical evaluation. Adv. Drug Deliv. Rev. 59, 631–644. Kim, S., Lee, J., 2010. Effective polymeric dispersants for vacuum, convection and freeze drying of drug nanosuspensions. Int. J. Pharm. 397, 218–224.
Kurz, W., Fisher, D.J., 1981. Dendrite growth at the limit of stability: tip radius and spacing. Acta Metall. 29, 11–20. Lee, J., 2004. Intrinsic adhesion force of lubricants to steel surface. J. Pharm. Sci. 93, 2310–2318. Lee, J., Cheng, Y., 2006. Critical freezing rate in freeze drying nanocrystal dispersions. J. Control. Release 111, 185–192. Lee, H., Venable, R., MacKerell Jr., A., Pastor, R., 2008. Molecular dynamics studies of polyethylene oxide and polyethylene glycol: hydrodynamic radius and shape anisotropy. Biophys. J. 95, 1590–1599. Lee, M.K., Kim, M.Y., Kim, S., Lee, J., 2009. Cryoprotectants for freeze drying of drug nano-suspensions: effect of freezing rate. J. Pharm. Sci. 98, 4808–4817. Lee, M.K., Chung, N.-O., Lee, J., 2010a. Membranes with through-thickness porosity prepared by unidirectional freezing. Polymer 51, 6258–6267. Lee, M.K., Kim, S., Ahn, C.-H., Lee, J., 2010b. Hydrophilic and hydrophobic amino acid copolymers for nano-comminution of poorly soluble drugs. Int. J. Pharm. 384, 173–180. Morar, A.S., Schrimsher, J.L., Chavez, M.D., 2006. PEGylation of proteins: a structural approach. Biopharm. Int. 19, 34–49. O’Brien, F.J., Harley, B.A., Yannas, I.V., Gibson, L., 2004. Influence of freezing rate on pore structure in freeze-dried collagen-GAG scaffolds. Biomaterials 25, 1077–1086. Philip, J., Gnanaprakash, G., Jayakumar, T., Kalyanasundaram, P., Raj, B., 2003. Three distinct scenarios under polymer, surfactant, and colloidal interaction. Macromolecules 36, 9230–9236. Ploehn, H.J., Russel, W.B., 1990. Interactions between colloidal particles and soluble polymers. Adv. Chem. Eng. 15, 137–228. Rohatgi, P.K., Adams, C.M., 1967. Effect of freezing rates on dendritic solidification of ice from aqueous solutions. Trans. Metall. Soc. AIME 239, 1729–1736. Tang, X., Pikal, M.J., 2004. Design of freeze-drying processes for pharmaceuticals: practical advice. Pharm. Res. 21, 191–200. Uhlmann, D.R., Chalmers, B., Jackson, K.A., 1964. Interaction between particles and a solid–liquid interface. J. Appl. Phys. 35, 2986–2993. Vergote, G.J., Vervaet, C., Van Driessche, I., Hoste, S., De Smedt, S., Demeester, J., 2001. An oral controlled release matrix pellet formulation containing nanocrystalline ketoprofen. Int. J. Pharm. 219, 81–87. Vyas, T.K., Shahiwala, A., Amiji, M.M., 2008. Improved oral bioavailability and brain transport of Saquinavir upon administration in novel nanoemulsion formulation. Int. J. Pharm. 347, 93–101. Wang, L.J., Dong, J.F., Chen, J., Eastoe, J., Xuefeng, L., 2009. Design and optimization of a new self-nanoemulsifying drug delivery system. J. Colloid Interface Sci. 330, 443–448. Zhang, Y., Gao, J., Zheng, H., Zhang, R., Han, Y., 2011. The preparation of 3,5-dihydroxy-4-isopropylstilbene nanoemulsion and in vitro release. Int. J. Nanomed. 6, 649–657. Zhou, H.F., Yang, Y., Liu, G.L., 2010. Preparation and characterization of a lecithin nanoemulsion as a topical delivery system. Nanoscale Res. Lett. 5, 224–230.
Contents lists available at SciVerse ScienceDirect
International Journal of Pharmaceutics journal homepage: www.elsevier.com/locate/ijpharm
Pharmaceutical Nanotechnology
Mechanism of freeze-drying drug nanosuspensions Nae-Oh Chung, Min Kyung Lee, Jonghwi Lee ∗ Department of Chemical Engineering and Materials Science, Chung-Ang University, 221, Heukseok-dong, Dongjak-gu, Seoul, 156-756, Republic of Korea
a r t i c l e
i n f o
Article history: Received 11 April 2012 Received in revised form 25 June 2012 Accepted 20 July 2012 Available online 7 August 2012 Keywords: Nanoparticles Nanocrystals Freeze-drying Lyophilization Redispersibility Cryoprotectant Freezing rate Nanosuspension Nanoformulation PEG
a b s t r a c t Drug nanoparticles prepared in a liquid medium are commonly freeze-dried for the preparation of an oral dosage in solid dosage form. The freezing rate is known to be a critical parameter for redispersible nanoformulations. However, there has been controversy as to whether a fast or slow freezing rate prevents irreversible aggregation. A systematic investigation is presented herein regarding the effect of both the molecular weight of the cryoprotectant and the freezing rate in order to elucidate the mechanism underlying irreversible aggregation. It was found that irreversible aggregation occurred during drying rather than freezing, although a proper freezing rate is critical. A more homogeneous distribution of the cryoprotectant and drug nanoparticles led to more redispersible powders. Thus, keeping the local concentration distribution of the nanoparticles and cryoprotectant fixed during the freezing step plays a critical role in how the freezing rate affects the redispersibility. The kinetic approach of excluding the tendency of ice crystal growth permitted an explanation of the controversial results. This study will facilitate an in-depth understanding of the aggregation process of nanoparticles or proteins during freeze-drying. © 2012 Elsevier B.V. All rights reserved.
1. Introduction Freeze-drying, or lyophilization, is known to be the most effective unit operation to preserve perishable materials. The process was developed to preserve serum during WWII. Protein drugs are popular materials that require freeze-drying, and recently the extensive use of drug nanoparticles has increased the importance of drying technologies (Tang and Pikal, 2004; Haugh et al., 2010; O’Brien et al., 2004; Abdelwahed et al., 2006). The intrinsic energy penalty associated with nanosized dimensions requires special nanotechnologies for external energy and stabilization, which mostly produce liquid state products, i.e., nanodispersions or nanoemulsions (Abdelwahed et al., 2006; Jacobsand and Muller, 2002; Hill, 2001; Lee, 2004; Ploehn and Russel, 1990; Philip et al., 2003; Berglund et al., 2003a,b; Zhang et al., 2011; Wang et al., 2009; Zhou et al., 2010; Vyas et al., 2008). Therefore, a drying step is commonly needed to convert the liquid products into solid dosage forms. Unfortunately, phase and composition changes that occur in the conversion process often nullify the effects of external energy and stabilization (Jacobsand and Muller, 2002; Hill, 2001; Lee, 2004; Ploehn and Russel, 1990; Philip et al., 2003; Berglund et al., 2003a,b) in the preparation steps, resulting in the aggregation of nanoparticles. Irreversible aggregates of drug nanoparticles are
∗ Corresponding author. Tel.: +82 2 816 5269; fax: +82 2 824 3495. E-mail address: [email protected] (J. Lee). 0378-5173/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ijpharm.2012.07.068
unable to redisperse into nanoparticles upon dissolution and thus, the advantages of nanoformulation can be lost. Indeed, the development of suitable pharmaceutical unit operations for nanoparticles, such as drying, granulation, and compaction, is still a challenging task. The successful preparation of redispersible nanopowders that readily revert back to nanosuspensions upon reconstitution in water requires delicate process control and certain processing windows. Extensive research has been conducted on the relationship between processing variables and the redispersibility of final solid powders (Ho and Lee, 2011; Kesisoglou et al., 2007; Vergote et al., 2001; Kim and Lee, 2010). However, similar to protein drug lyophilization, the formulation of a detailed mechanism and an indepth understanding of this relationship has yet to be achieved. Most research has focused on the choice of cryoprotectant, temperature profile, concentration and vacuum pressure. In the freeze-drying process, a wide variety of cryoprotectants, such as glycerol, quaternary amines, carbohydrates, and synthetic polymers are used to protect nanoparticles from stresses and subsequent aggregation. It is generally known that a higher concentration of cryoprotectant and faster freezing result in better nanoparticles redispersibility. However, it has been reported that the use of a common cryoprotectant destabilized nanoparticles in some cases. Kamiya et al. investigated the effect of interactions between lipid nanoparticles and saccharides on freeze-drying (Kamiya et al., 2010) and found that the crystallization of a saccharide excluded drug-lipid nanoparticles, resulting in irreversible
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
aggregation. With an increase in the cryoprotectant concentration, the irreversible aggregation of nanoparticles may either increase or decrease depending on the system of nanoparticles and the freezedrying conditions. Controversy also exists regarding the freezing rate; a fast freezing rate (low freezing temperature) may allow for better redispersibility, but in other cases, a slow freezing rate yields better results. Thus far, the selection of cryoprotectants and related processing conditions has largely relied on empirical approaches. Previous reports focused on the freezing rate, which had mostly been neglected in earlier research. To facilitate an in-depth understanding, a control method was developed for the freezing rate by adjusting the temperature gradient. The freezing rate is closely related to the temperature profile of the actual industrial unit operation. By systematically controlling the freezing rate for nanocrystalline systems of water-insoluble drugs, it was found that, without any other changes, only variations in the freezing rate could result in significantly different redispersibilities. We defined a critical freezing rate above which nanoparticles can retain their redispersibility (Lee and Cheng, 2006). However, depending on the cryoprotectant and processing conditions, the successful processing of redispersible nanopowders requires faster or sometimes slower freezing rates (Lee et al., 2009). The choice of cryoprotectant is also critical. In the case of carrageenan, only 0.5 wt% could successfully protect naproxen nanocrystals (Kim and Lee, 2010). Although we revealed the importance of the freezing rate in the freeze-drying process, the mechanism behind the phenomenon is not yet well-understood. During freezing, ice crystals nucleate and grow while excluding most of the solutes into a cryoconcentrate phase. In the cryoconcentrates, further phase separation or crystallization of a component can occur. Therefore, by focusing on the parameters related to the phase separation and crystallization phenomena, a more in-depth mechanism may be elucidated. First of all, kinetic phenomena are most likely dependent on the diffusion characteristics of the nanoparticles and cryoprotectant. Therefore, in addition to the control of the freezing rate, the molecular weight of the cryoprotectant was systematically varied in this study to investigate its effect on diffusivity. The diffusivity issue was highlighted from the perspective of frozen structure development (internal spatial distribution) in order to reveal the mechanism of irreversible aggregation among nanoparticles. In particular, we hypothesized that morphological characteristics developed during freezing influence subsequent irreversible aggregation. 2. Materials and methods 2.1. Materials Naproxen, a relatively insoluble crystalline drug compound (API, Mw = 452.8 g/mol, purity >95%, Tokyo Chemical Industry, Japan) was used as a model drug, while hydroxypropyl cellulose (HPC, Mw = 60 kg/mol, surface energy = 45 mN/m, FMC, Philadelphia, PA, USA) was employed as a steric stabilizer for wet comminution. Ramda-carrageenan (non-gelling at 1% in a 0.2 M KCl solution), sucrose (purity >98%), and polyethylene glycol (PEG, Mn = 200–35,000 g/mol) obtained from Aldrich (St. Louis, MO, USA) were used as cryoprotectants. HPLC-grade water from Aldrich and yttria-stabilized zirconia beads (1 mm diameter, Performance Ceramics, OH, USA) were employed without further purification. 2.2. Preparation of nanocrystal dispersions Low energy comminution was used to produce nanocrystal dispersions with 1 mm zirconia beads (50 mL) as the grinding media. API was mixed with HPC previously dissolved in water (100 mL), and the beads and additional water were put into a glass bottle.
43
Fig. 1. Schematic illustration of a controlled freezing device with a temperature gradient, which decreases as a function of distance from the liquid nitrogen reservoir. Freezing starts at the region close to the reservoir, and the freezing rate decreases as the freezing front moves away from the reservoir.
The total weight of the slurry (API + stabilizer + water) was 8.4 g, and the concentration of API in water was 15 wt%. The weight ratio of HPC:drug was kept constant at 1:6. The comminution speed and time were 125 rpm and five days, respectively. After comminution at room temperature, the media beads were filtered out and the suspension was stored at 5 ◦ C. The cryoprotectants were then introduced into the nanocrystal suspension 24 h before the subsequent experiment. The solution concentration of cryoprotectant was varied from 5 to 7.5 wt%. 2.3. Controlled freeze-drying Nano-suspensions were frozen using a custom-made apparatus (Fig. 1) following a previously outlined procedure (Lee and Cheng, 2006; Lee et al., 2009). The freezing rate was varied as a function of the distance from the liquid nitrogen reservoir in the sample, thus reflecting changes in the temperature gradient, which was repeatedly checked as outlined in earlier studies (Lee and Cheng, 2006; Lee et al., 2009). As shown in Fig. 1, nano-suspensions were first put into a long tube, and the tube was subsequently placed in contact with a liquid nitrogen reservoir. After contact, freezing began from the region near the liquid nitrogen reservoir. Both the container and the reservoir were made of copper to enable rapid heat transfer. In addition to heat transfer through the copper, cold nitrogen gas flow aided in the propagation of the freezing front. Because the frozen portions of the suspension had different turbidities when compared to the liquid portions, the freezing interface between the frozen and liquid portions was distinct (Lee et al., 2009). Freezing rates could be calculated by monitoring the change in position of the freezing front with respect to time. To obtain reproducible results, all other experimental factors, such as the amount of liquid nitrogen, were kept constant. The frozen suspension was vacuumdried using the FD-1000 bench top freeze-dryer (EYELA, Tokyo, Japan, trap chilling temperature of −55 ◦ C, 6.1 Pa) for 3–24 h (unless otherwise indicated). 2.4. Characterization The particle size was measured with a Horiba LA 910 laser light scattering analyzer (632.8 nm He-Ne laser) after dispersing the particles in 150 mL of water and sonicating the mixture for 1 min. A relative refractive index of 1.2 ± 0.00i and a 1 min sonication of 40 W and 39 kHz were employed. After freeze-drying, disk-shaped powder samples (5.1 mm diameter) taken at different locations along the long tube were used for characterization (Fig. 1). The volume-averaged particle sizes of the powders were used in this study.
44
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
The morphology of the freeze-dried powders was characterized by optical microscopy (Olympus BX–51); quantitative data for the thickness of ice crystals were obtained at 30 different locations. The morphology of the freeze-dried powders was further examined with a Hitachi S-4300E scanning electron microscope (SEM, 3–4 kV, 3 nm platinum coating). The surface morphology of the powders was investigated by atomic force microscopy (AFM, XE100, Park Systems, South Korea) using a silicon cantilever (NSC15, MikroMasch, Spain) with a spring constant of 30.05 (±3.7) N/m. The distributions of PEG and naproxen were mapped with an IRT5000 FT-Raman microscope system (surface reflect-type, Jasco, Easton, MD, USA) capable of FT-IR (FT-IR 6100 – type A) and Raman (NRS-3000) measurements in order to elucidate their diffusion mechanism during freezing and drying. The FT-IR wavenumber ranged from 400 to 2500/cm (reflection-type microscope, accumulation: 200, resolution: 4/cm, scanning speed: 32 m/s), while the Raman wavenumber ranged from 1000 to 2000/cm (exposure time: 5 s, accumulation: 200, laser source: 532 nm ND:yak, resolution: 1.85/cm). The crystal structure of the dried powder was analyzed with an X-ray diffractometer (XDS 2000, Scintag, Santa Clara, CA, USA) at a scan rate of 1◦ /min (2: 5–40◦ ). Differential scanning calorimetry (DSC, Pyris 6, PerkinElmer, USA) data were also acquired. For the DSC measurements, 10 mg of sample was scanned twice in a N2 /air atmosphere at a heating rate of 10 ◦ C/min and a cooling rate of 20 ◦ C/min from −60 to 180 ◦ C. The viscosity of the cryoprotectant was measured with a rheometer (Brookfield RH10 capillary rheometer, Middleboro, MA, USA) at a fixed shear rate (10/s) and temperature (25 ◦ C); the average of 15 measurements was used. The zeta potential was measured with a dynamic electrophoretic light scattering spectrophotometer ELS-Z2 (Otsuka Electronics, Japan; average electric field: −16.53 V/cm, conversion equation: Smoulchowski) at 25 ◦ C. 3. Results 3.1. Effect of freezing rate and PEG molecular weight To gain a systematic understanding of the freezing mechanism of drug nanosuspensions, a controlled freezing apparatus with freezing rate control capability was designed. In conventional freeze-drying, freezing starts from the coldest region, typically the surface region of a vial, and depending on the imposed temperature gradient, freezing propagates into the rest of the liquid
phase at a poorly controlled freezing rate. Controlled freezing experiments were conducted utilizing the temperature gradient surrounding the liquid nitrogen reservoir (Fig. 1). In contrast to previous reports (Lee and Cheng, 2006; Lee et al., 2009), a flow duct was introduced to guide nitrogen gas flow and limit potential turbulent air flow from the outside environment. This improved the reproducibility of the freezing rate. The change in turbidity of the nanosuspensions was recorded as a function of time, thus providing information on the movement of the freezing front, which can be taken as the freezing rate. Following previous reports (Lee and Cheng, 2006; Lee et al., 2009), we confirmed the reproducibility of freezing rate control and the imposed temperature gradients as a function of distance from the liquid nitrogen reservoir. The effect of the freezing rate on the redispersibility of drug nanosuspensions in the presence of PEG with different molecular weights is shown in Fig. 2. Once particle aggregation occurs, the average particle size quickly reaches several microns from the original particle size of 148 nm; therefore our focus was on the transition between these two sizes, which corresponds to nanoparticle aggregation (Lee and Cheng, 2006; Lee et al., 2009). For 8 and 20 kg/mol PEG (Fig. 2(a)), faster freezing rates resulted in redispersed drug particles with a larger particle size. In other words, slower freezing was better than faster freezing for drug nanocrystal dispersions between 0 and 700 m/s. This is consistent with the results of previous reports on mannitol and PEG (Lee et al., 2009), but quite contrary to both the conventional understanding of freeze-drying and what is observed during the freeze-drying of drug nanosuspensions without cryoprotectant (Lee and Cheng, 2006). The molecular weight of polymer cryoprotectants is an important parameter that determines diffusion, entanglement density, viscosity and steric repulsion. One of the goals of this work was to investigate how an increase in the molecular weight affects the redispersibility of nanosuspensions. As shown in Fig. 2(a), it was found that PEG with a higher molecular weight prevented the aggregation of drug nanoparticles more effectively in a range of relatively low freezing rates. This was further confirmed in Fig. 2(b), where only the lowest molecular weight case (PEG 0.4k) showed significant aggregation, which manifested as a larger particle size. Although the particle size was not linearly dependent on the molecular weight, there seemed to be a transition in particle size caused by an increase in molecular weight. When comparing Fig. 2(a) and (b), it is evident that an increase in the PEG content
Fig. 2. Effect of the freezing rate on the volume-averaged particle size of redispersed drug nanoparticles after freeze-drying in the presence of (a) 5 wt% and (b) 7.5 wt% PEG (Mn = 0.4–20 kg/mol).
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
45
Fig. 3. Effect of the drying time on the volume-averaged particle size of redispersed drug nanoparticles after freeze-drying in the presence of 5 wt% PEG (Mn = 8 kg/mol).
from 5 to 7.5 wt% was followed by a significant improvement in redispersibility (except PEG 0.4 K and some low freeze rate cases). This effect of the cryoprotectant concentration is consistent with that observed in most previous studies (Lee et al., 2009). For comparison, 1 wt% carrageenan was used in place of PEG, and successful cryoprotection was observed. The resulting particle sizes were all below 0.2 m (0.15, 0.17, 0.17, 0.17, 0.18, 0.17, and 0.19 m for 588, 344, 222, 105, 86, 11, and 9.6 m/s, respectively). This result is also consistent with previous reports that carrageenan provides effective cryoprotection (Kim and Lee, 2010).
Fig. 4. Effect of the freezing rate on the thickness of ice crystals in frozen nanosuspensions in the presence of various cryoprotectants (5 wt%). The inlet OM images show the freeze-dried drug nanosuspensions (PEG 8k), with arrows denoting the direction of freezing propagation (scale bar = 100 m).
3.2. Occurrence of irreversible aggregation Determining the mechanism of irreversible aggregation and the effect of the freezing rate on this mechanism is not a simple task. However, Fig. 3 aids in providing an in-depth understanding of aggregation. The drug nanosuspensions with PEG 8k were frozen at different freezing rates and then freeze-dried for
Fig. 5. Typical SEM micrographs of naproxen/PEG freeze-dried powders (freezing rate = 150–200 m): (a) PEG 0.4k (5 wt%) and (b) PEG 20k (7.5 wt%).
46
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
different time periods. After subsequent reconstitution of the samples in water, the degree of irreversible aggregation was assessed by measuring the average particle sizes. No significant aggregation occurred during the initial three hours of drying. However, irreversible aggregation continually progressed up to 24 h, especially in the fast freezing rate ranges. At freezing rates between 200 and 700 m/s, the most significant difference in the degree of aggregation occurred between three and six hours, indicating that irreversible aggregation occurs primarily during the sublimation process due to internal changes, e.g., desiccation stress, heat of sublimation, etc. Although irreversible aggregation occurred during the drying step and not in the freezing step, aggregation still depended on the freezing rate. Therefore, while the frozen nanosuspensions did not aggregate before drying, they seemed to have different internal structures depending on the freezing rate. Certain internal structures are more susceptible to desiccation stress and/or heat of sublimation, which can cause irreversible aggregation. Thus the differences in the internal structure likely resulted in different degrees of aggregation.
3.3. Morphology of frozen nanosuspensions The internal structure of the freeze-dried powders from nanosuspensions was investigated via OM, SEM, and AFM. Typical OM micrographs of freeze-dried powders are shown in Fig. 4. Along the direction of freezing propagation, ice crystals grew as columnar or lamellar phases. This phase separation between ice and cryoconcentrates occurs through fingering of the freezing front (Colard et al., 2009; Deville et al., 2007; Uhlmann et al., 1964). The width of the columnar ice crystals distinctly depends on the freezing rate (Rohatgi and Adams, 1967; Lee et al., 2010a). As the freezing rate increases, the width inversely decreases (as shown in Fig. 4), regardless of the sample. Compared to the pure drug nanosuspension, the addition of a cryoprotectant increased the width of the ice crystals. Among the cryoprotectants used in Fig. 4, carrageenan yielded the largest increase in ice crystal width, followed by sucrose. A marginal difference was observed in samples containing PEG of different molecular weights. A lower molecular weight resulted in ice crystals with a larger width. The fingering phenomenon seemed to develop further in the samples containing lower molecular weight PEG. The ice crystal width does not directly support the tendency toward irreversible aggregation, since the results in Fig. 4 are not systematically correlated with the re-dispersion particle size results. However, phase separation indirectly determines the composition of the cryoconcentrate and the desiccation stress.
SEM images of well (Fig. 5(b)) and poorly dispersible (Fig. 5(a)) samples due to a difference in molecular weight show the typical surface features of dried cryoconcentrate regions facing ice crystals. In the left micrographs taken at low magnification, there is no significant difference in morphology related to redispersibility. At a higher magnification (right), a slight difference is apparent. Surface roughness was detected in the well redispersible samples at a slightly shorter wavelength (submicron) that more clearly showed the individual drug nanoparticles (Fig. 5(b)). Data regarding the difference in surface roughness and the identification of individual nanoparticles was obtained with AFM scans of PEG 8k. The AFM images revealed that the slower freezing rate generally produced surface roughness with a slightly shorter wavelength. The sample in Fig. 6(c), which is a readily redispersible powder, has the shortest wavelength corresponding to the size of the individual nanoparticles. In Fig. 6(a) (fast freezing, poorly dispersible), the wavelength of the surface roughness was larger than the size of the primary nanocrystals. When the PEG and nanoparticles were homogeneously mixed, the individual nanoparticles were more easily observed on the surface. In poorly redispersible samples (Figs. 5(a) and 6(a)), the surface of the cryoconcentrate seemed to have PEG that was more inhomogeneously dispersed. These findings suggest that the wavelength of the surface roughness could be an indirect sign of irreversible aggregation due to a lack of cryoprotectant between the nanoparticles.
3.4. Distribution of cryoprotectants and drug nanoparticles A morphological examination itself cannot reveal the local distribution of drug nanoparticles and cryoprotectants. To elucidate the detailed internal structure, both FT-IR and Raman mapping techniques were employed; typical results of FT-IR mapping are given in Fig. 7. The concentrations of PEG and nanoparticles varied perpendicular to the growth direction of the ice (downward) due to the fingering of the freezing front, while the wavelength of the concentration fluctuation seemed to be related to the size of the ice crystals. Interestingly, the position of the local maxima of the PEG was always different from that of the drug nanoparticles. The mismatch between the two maxima indicated that the rejection processes of the drug nanoparticles and PEG from the freezing front were different, which resulted in two regions within the cryoconcentrate phase: a nanoparticle-rich region and a PEG-rich region. The two regions were mainly formed kinetically, not thermodynamically, although there was no clear distinction between the two mechanisms. In the case of a slow freezing rate, the mismatch of the local maxima of the two components was not greater than the
Fig. 6. AFM micrographs showing the effect of the freezing rate on the morphology of dried powders (PEG 8k): (a) freezing rate = 590 m/s, (b) 100 m/s, and (c) 9 m/s.
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
47
rate cases. In other words, PEG cannot be homogeneously inserted between drug nanoparticles and thus it cannot prevent irreversible aggregation in this case. 3.5. Viscosity of nanosuspensions The results of viscosity measurements of the nanosuspensions with cryoprotectants and their controls (solutions without drug nanoparticles) are shown in Fig. 8. Shear thinning behavior was evident in all of the samples in Fig. 8(a). Carrageenan of only 1 wt%, which provided successful cryoprotection efficiency in a previous report (0.15–0.19 m particle sizes at 9.6–588 m/s freezing rates) (Kim and Lee, 2010), had the highest viscosity, followed by PEG and sucrose. Sucrose of 5 wt% was not effective for cryoprotection, which corroborates previous work (Lee et al., 2009). As can be inferred from the particle size data in Fig. 2, PEG 8k was more viscous than PEG 1k. Therefore, the viscosity results have a simple relationship with the redispersibility results: the higher the viscosity, the better the redispersibility results. The viscosity may not be the only determining factor, but it does have a major influence on cryoprotection. In particular, subsidiary phase separation into PEG-rich and nanoparticle-rich regions in cryoconcentrates appear to be dependent on the viscosity of the medium.
Fig. 7. FT-IR microscopy mapping images (30 × 20 m) showing the freezing rate dependence of the distribution of naproxen and PEG 8k (5 wt%) in freeze-dried powders. The characteristic peaks of naproxen (1733 cm−1 ) and PEG 8k (1355 cm−1 ) were used. The dotted lines indicate the local maxima of the concentration.
wavelength of the ice crystals (Fig. 7(c)). Still, the position of the local maxima seemed to follow an alternating turn between the PEG and drug nanoparticles. Due to the magnification limitation of FT-IR, the wavelength of the ice crystals is not fully covered in Fig. 7(c), but it was approximately 80 m (Fig. 4). The mismatch became more irregular as the freezing rate increased. In the case of a fast freezing rate, there were five and four local maxima for the nanoparticles and PEG, respectively (Fig. 7(a)). The wavelength of freezing front fingering could not be traced from the wavelength of the PEG or nanoparticle local maxima, which was about 5 m (Fig. 4). Therefore, it was observed that a fast freezing rate induced a more irregular accumulation of PEG and nanoparticles in the cryoconcentrates, resulting in a more inhomogeneous distribution. This explains the poor redispersibility in the fast freezing
3.6. Interactions between drug and PEG Several characterizations were conducted in order to determine the detailed aggregation mechanism. First, the molecular interactions between naproxen and PEG were examined using FT-IR. However, no significant evidence of the interactions was observed (data not shown), probably because the interactions were confined only to the surface of the naproxen nanoparticles. This result was also consistent with the XRD and DSC findings, where no meaningful differences were noted (data not shown). Similar to the results of a previous report (Lee et al., 2009), the crystalline phase of naproxen detected by XRD and DSC was not influenced by the freezing rate or the existence of PEG (Anon., 2012). Zeta potential values of the drug nanoparticles were also measured in order to detect differences in electrostatic repulsion and possible surface charge effects on redispersibility. No significant difference in zeta potential, which may be related to redispersibility, was identified. In addition, the values obtained before freeze-drying (naproxen = −16.77, naproxen/PEG 8k = −17.79, and
Fig. 8. Viscosity of (a) drug nanosuspensions containing various cryoprotectants and (b) pure cryoprotectant solutions.
48
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
crystal growth (6). Eventually, the liquid phase undergoes vitrification at a maximally concentrated state. The cryoprotectant molecules can form their own phase (crystalline or amorphous) during this step, although it is not preferable for good redispersibility. These phase developments can be restricted by kinetic factors. The key mechanisms underlying irreversible aggregation are chain entanglement in a steric stabilizer (HPC) and drug particle fusion. For better redispersibility, a highly water soluble cryoprotectant, present between the drug nanoparticles, should homogeneously cover and effectively protect the surface of the drug nanoparticles from possible entanglement and fusion. Therefore, the homogeneity of the distribution and the degree of protection are the main factors to be discussed. The distribution of the drug nanoparticles and cryoprotectant in cryo-concentrated phases was not found to be entirely homogeneous. First, differences in diffusivity can cause inhomogeneity. Although the solubility of a solute in ice is almost negligible, the nanoparticles or polymers (solutes) in solution can be engulfed by the ice phase as the freezing rate increases. If a polymer chain is considered to be a particle and there is no interaction among particles, which is obviously an oversimplification, then simple theoretical principles can be developed for flat interfaces. By considering the surface energy, mass balance, and diffusion equations, the critical freezing rates below which the cryoprotectant (PEG) or drug nanoparticles will be excluded, c,p and c,n , respectively, are Fig. 9. Schematic illustration of structure development during the freezing of drug nanosuspensions containing a cryoprotectant. Cryoprotectants and drug nanoparticles can be engulfed in an ice crystal phase depending on the freezing rate () and critical freezing rate (c,p and c,n ).
naproxen/sucrose = −15.62 mV) did not differ significantly from the values acquired after re-constitution in water (naproxen/PEG 8k = −14.34 and naproxen/sucrose = −16.88 mV). Regarding the stability of the suspensions due to steric and electrostatic repulsion, a rough guideline of 20 mV for the zeta potential has been suggested (Jacobsand and Muller, 2002). However, this does not take into account the complexity of steric and electrostatic repulsion. As the particle size decreases, steric repulsion plays a more important role than electrostatic repulsion, which relies on many variables such as molecular weight, adsorption and desorption kinetics and chain flexibility (Lee et al., 2010b). In our research, a zeta potential of −14 to −18 mV was sufficient to provide stable suspensions up to six months at room temperature and no significant size changes after six months. 4. Discussion 4.1. Distribution of drug nanoparticles and cryoprotectants The freeze-drying process is composed of two steps: freezing and drying at a low temperature. Freezing, the first step in freeze-drying, fixes the morphological and local compositional characteristics of the ice crystals and cryoconcentrates, which is critical in preventing irreversible aggregation. As ice crystals nucleate and grow following the direction of the temperature gradient, cryoprotectant molecules and drug nanoparticles are rejected from the ice phase into a cryo-concentrated liquid phase (Fig. 9). For a particle to be rejected by the ice phase, the surface free energy at a minimum separation, = sp − ( lp + sl ), should be positive, where sp , lp , and sl are the interfacial free energies of the solid-particle, liquid-particle, and solid–liquid interfaces, respectively (Deville et al., 2007; Uhlmann et al., 1964). The composition of the liquid phase follows the equilibrium phase boundary determined by thermodynamics in an ideal situation with ice
c,p =
c,n =
∇ p d ao 3rp
(1)
d
∇ n d ao 3rn
n
d
n (2)
where d is the distance between the particle (or polymer) and the interface (freezing front), is the viscosity of water, r is the radius of the particle (or polymer), ao is the intermolecular distance of water, and n is an exponent that is larger than 1 (Deville et al., 2007; Uhlmann et al., 1964). The critical freezing rate represents the excluding tendency of the solutes. Eq. (2) conveniently provides a basic understanding of the effects of influencing factors, such as the particle size (Deville et al., 2007; Uhlmann et al., 1964). If there is no further phase separation in the cryoconcentrated liquid phase, then the rejected solutes, namely the drug nanoparticles and cryoprotectants, at the moving front of the ice crystals will be homogeneously mixed in the cryoconcentrated regions. On the other hand, if one component is engulfed by ice crystals while others are not, then the formation of cryoconcentrate will entail an inhomogeneous distribution depending on the relative magnitude of the freezing rate (), c,p and c,n . Four different cases of freezing rates are schematically illustrated in Fig. 9. First, if the freezing rate, , is smaller than c,p and c,n , then both the drug nanoparticles and cryoprotectant will be rejected by the growing ice phase and form a homogeneous cryoconcentrate phase. Local fluctuations in the concentration will also be minimized. This could be the case for many instances of conventional freeze-drying, since the typical values of c,p and c,n are relatively large (Deville et al., 2007; Uhlmann et al., 1964). In this case, conventional knowledge of freeze-drying is appropriate. For example, the redispersibility will improve as the freezing rate increases, since smaller crystals form with reduced freezing and desiccation stress. A higher concentration of cryoprotectant would also be beneficial. The other extreme case occurs when the freezing rate is larger than both c,p and c,n (Fig. 9(d)). An extremely fast freezing rate appears to be rare. However, a rather homogeneous distribution can still result, preventing irreversible entanglement and particle fusion.
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
The size of the drug nanoparticles is at least one order of magnitude larger than that of PEG (radius of gyration of 20k PEG = 7 nm (Lee et al., 2008; Morar et al., 2006)). Thus, c,p may be larger than c,n , but the critical velocities of the drug nanoparticles (c,n ) and PEG (c,p ) are still difficult to evaluate since other surface free energy parameters are simply not available. Furthermore, if the interactions between the particles are considered (many particle systems), the viscosity term () should be a function of the concentration and the interactions. As the freezing rate increases, a solute of low diffusivity will first be engulfed inside the ice crystal phases, resulting in an inhomogeneous distribution of drug nanoparticles and cryoconcentrates. As the freezing rate increases above c,n , more drug nanoparticles will be trapped in ice crystals forming their own phases, which results in a poorly redispersible case (Fig. 9(b)). The accumulation of cryoprotectants outside the cryoconcentrate regions is illustrated in Fig. 9(c). In such a scenario, more drug nanoparticles come into direct contact with each other. These two cases explain our observation of the unique effects of freezing rate and molecular weight. If one of the cases occurs as the freezing rate increases, the effect of the freezing rate will be contrary to conventional knowledge. As the molecular weight of the PEG increases, r and increase, and c,p seems to get closer to c,n . As a result, more homogeneous distributions could form. This explanation may exclude the possibility of the mechanism illustrated in Fig. 9(c), as can be inferred by the particle size difference. It is intuitive that excluding a drug nanoparticle, which is far larger than a polymer chain, is usually quite easy. Previous reports on the unexpected effect of the freezing rate, which primarily found an increase in the redispersibility with a decrease in the freezing rate, have supported explanations regarding specific interactions between particles and cryoprotectants. The type of cryoprotectant and its interactions with drug nanoparticles are considered in the parameter in our treatment. The aforementioned kinetic approach is based on the simple view of competition between the critical velocity and freezing rate. However, if phase separation or a concentration fluctuation develops in the cryoconcentrated region, then other parameters such as the phase behavior of the concentrated solution, local temperature fluctuations, and entanglement fluctuations should also be considered.
4.2. Microstructure and local distribution The flat interface of the freezing front excludes solutes and develops a significant concentration gradient. The overall freezing rate is determined by the rate of heat extraction, while the local growth rate is limited by the low liquid diffusivity. To overcome the mismatch between the overall freezing rate and local growth rate, the flat freezing front takes on a dendritic shape (Kurz and Fisher, 1981). As ice crystals grow, heat flowing from the moving interface into the remaining unfrozen solution causes supersaturation and supercooling to occur. As heat transfer rates are normally much greater than mass transfer rates, nanoparticle or cryoprotectant diffusion tends to control the growth of the ice crystals. The accumulation of rejected nanoparticles and cryoprotectant will further decrease solute mobility, resulting in an engulfment of cryoconcentrated regions. Based on a materials balance and Fick’s law, theoretical equations to describe the inter-dendrite spacing in the case of dilute solutions have been proposed. Rohatgi and Adams (Rohatgi and Adams, 1967) suggested the following expression for the spacing between dendrites (ice crystals): L = (8DT )
0.5 −0.5
F
,
(3)
49
where D is the binary diffusion coefficient related to the concentration term, T is the supercooling between dendrites, and F is the freezing rate (Rohatgi and Adams, 1967). The researchers demonstrated that supersaturation and supercooling are independent of the freezing rate, as well as the relationship between L and D or F. Eq. (3) does not consider interactions between particles (polymers), the particle size distribution, the viscoelastic properties of the slurry, latent heat diffusion, or partial solidification in the cryo-concentrated regions. However, it quantitatively explains the results shown in Fig. 4. As the viscosity of a solution increases due to an increase in the molecular weight, D and L both decrease. The effect of the freezing rate also obeys Eq. (3) since an increase in F (freezing rate) is followed by a decrease in L. However, the scaling parameters of ‘−0.5’ were not precisely confirmed in our experiment. Note that −0.65, −0.68, −0.71, −0.35, −0.73, and −0.81 were obtained for cases without PEG, PEG 1 K, PEG 8 K, PEG 20 K, SUC, and CAR, respectively. In addition, the freezing rate, a purely experimental parameter in this study, is not strictly F in Eq. (3), but the two parameters are proportional to each other. As shown in Fig. 4, the redispersibility does not directly reflect changes in L. Therefore, the local distribution of the cryoprotectant and nanoparticles is more important than the pore morphology of a dried cake (ice crystal morphology), such as L. The freezing step determines the internal structures of the cryoconcentrated phases of a frozen cake where molecular mobility is restricted. In the cases of poor redispersibility, the subsequent drying process brings drug nanoparticles in close contact with each other, where actual irreversible aggregation occurs. This aggregation process depends on the local distribution of the cryoprotectant determined in the previous freezing step. The analysis and conclusions of this study should be generalized with caution. Although the analysis was based on a series of investigations regarding the effects of the freezing rate using drug nanoparticles with and without sucrose, mannitol, lactose, and PEG (Lee et al., 2009), generalization may require more extensive experiments using various materials and conditions. 5. Conclusions The redispersibility of dried nanoparticles is a critical factor, and poor redispersibility may arise due to irreversible aggregation. Unfortunately, redispersibility is an unpredictable factor; it decreases in some cases and increases in other cases with an increase in the freezing rate. For the development of redispersible solid formulations, the underlying mechanism of irreversible aggregation was investigated in detail in this study. The molecular weight of cryoprotectants and the freezing rate were systematically varied, and the redispersibility and distribution of the cryoprotectant were investigated. The redispersibility of nanoparticles increased with an increase in the molecular weight of the cryoprotectant. A slow freezing rate was better for the redispersibility of nanoparticles in the case of PEG. These results were explained by a kinetic approach based on the excluding tendency of the solutes (represented by the critical freezing rate) from growing ice crystals. Differences in the excluding tendencies of the cryoprotectant and drug nanoparticles could explain the dependence of the redispersibility on both the freezing rate and the molecular weight. The spatial distributions of PEG and naproxen, as mapped by FT-IR, allowed for the elucidation of their diffusion mechanism during freezing and drying. Acknowledgements This work was supported by a grant (10035574) from the “Platform Chemical Process Technology from Lignocellulosic Biomass Conversion” R&D Program funded by the Ministry of Knowledge
50
N.-O. Chung et al. / International Journal of Pharmaceutics 437 (2012) 42–50
Economy, grants from the National Research Foundation of Korea (NRF) funded by the Korea government (MEST) (2012-014107), and the Korea Healthcare technology R&D Project, Ministry of Healthcare & Welfare (A103017). MKL would like to thank the Human Resource Training Project for Strategic Technology (MKE & KOTEF). References Abdelwahed, W., Degobert, G., Stainmesse, S., Fessi, H., 2006. Freeze-drying of nanoparticles: formation, process and storage considerations. Adv. Drug Deliv. Rev. 58, 1688–1713. The Tm of PEG 8k (5 wt%) in the first scan showed a slight increase in the cases of fast and slow freezing rates (55.02 ◦ C (8.75 J/g)) for 590 m/s and 54.95 ◦ C (8.80 J/g) for 9 m/s) when compared to the case of a middle freezing rate (52.29 ◦ C, 7.89 J/g, 100 m/s) while the Tm of naproxen (155.09, 155.11, and 155.95 ◦ C for 590, 100, and 9 m/s respectively) and PEG in the second scans (52.07, 52.18, and 52.50 ◦ C for 590, 100, and 9 m/s respectively) were not changed. Berglund, K.D., Przybycien, T.M., Tilton, R.D., 2003a. Coadsorption of sodium dodecyl sulfate with hydrophobically modified nonionic cellulose polymers. 1. Role of polymer hydrophobic modification. Langmuir 19, 2705–2713. Berglund, K.D., Przybycien, T.M., Tilton, R.D., 2003b. Coadsorption of sodium dodecyl sulfate with hydrophobically modified nonionic cellulose polymers. 2. Role of surface selectivity in adsorption hysteresis. Langmuir 19, 2714–2721. Colard, C.A.L., Cave, R.A., Grossiord, N., Covington, J.A., Bon, S.A.F., 2009. Conducting nanocomposite polymer foams from ice-crystal-templated assembly of mixtures of colloids. Adv. Mater. 21, 2894–2898. Deville, S., Saiz, E., Tomsia, A.P., 2007. Ice-templated porous alumina structures. Acta Mater. 55, 1965–1974. Haugh, M.G., Murphy, C.M., O’Brien, F.J., 2010. Novel-freeze-drying methods to produce a range of collagen-glycosaminoglycan scaffolds with tailore mean pore sizes. Tissue Eng. Part C: Methods 16, 887–894. Hill, T.L., 2001. A different approach to nanothermodynamics. Nano Lett. 1, 273–275. Ho, H., Lee, J., 2011. Redispersible drug nanoparticles prepared without dispersant by electro-spray drying. Drug Dev. Ind. Pharm., 1–8. Jacobsand, C., Muller, R., 2002. Production and characterization of a budesonide nanosuspension for pulmonary administration. Pharm. Res. 19, 189–194. Kamiya, S., Kurita, T., Miyagishima, A., Itai, S., Arakawa, M., 2010. Physical properties of griseofulvin-lipid nanoparticles in suspension and their novel interaction mechanism with saccharide during freeze-drying. Eur. J. Pharm. Biopharm. 74, 461–466. Kesisoglou, F., Panmai, S., Wu, Y., 2007. Nanosizing-oral formulation development and biopharmaceutical evaluation. Adv. Drug Deliv. Rev. 59, 631–644. Kim, S., Lee, J., 2010. Effective polymeric dispersants for vacuum, convection and freeze drying of drug nanosuspensions. Int. J. Pharm. 397, 218–224.
Kurz, W., Fisher, D.J., 1981. Dendrite growth at the limit of stability: tip radius and spacing. Acta Metall. 29, 11–20. Lee, J., 2004. Intrinsic adhesion force of lubricants to steel surface. J. Pharm. Sci. 93, 2310–2318. Lee, J., Cheng, Y., 2006. Critical freezing rate in freeze drying nanocrystal dispersions. J. Control. Release 111, 185–192. Lee, H., Venable, R., MacKerell Jr., A., Pastor, R., 2008. Molecular dynamics studies of polyethylene oxide and polyethylene glycol: hydrodynamic radius and shape anisotropy. Biophys. J. 95, 1590–1599. Lee, M.K., Kim, M.Y., Kim, S., Lee, J., 2009. Cryoprotectants for freeze drying of drug nano-suspensions: effect of freezing rate. J. Pharm. Sci. 98, 4808–4817. Lee, M.K., Chung, N.-O., Lee, J., 2010a. Membranes with through-thickness porosity prepared by unidirectional freezing. Polymer 51, 6258–6267. Lee, M.K., Kim, S., Ahn, C.-H., Lee, J., 2010b. Hydrophilic and hydrophobic amino acid copolymers for nano-comminution of poorly soluble drugs. Int. J. Pharm. 384, 173–180. Morar, A.S., Schrimsher, J.L., Chavez, M.D., 2006. PEGylation of proteins: a structural approach. Biopharm. Int. 19, 34–49. O’Brien, F.J., Harley, B.A., Yannas, I.V., Gibson, L., 2004. Influence of freezing rate on pore structure in freeze-dried collagen-GAG scaffolds. Biomaterials 25, 1077–1086. Philip, J., Gnanaprakash, G., Jayakumar, T., Kalyanasundaram, P., Raj, B., 2003. Three distinct scenarios under polymer, surfactant, and colloidal interaction. Macromolecules 36, 9230–9236. Ploehn, H.J., Russel, W.B., 1990. Interactions between colloidal particles and soluble polymers. Adv. Chem. Eng. 15, 137–228. Rohatgi, P.K., Adams, C.M., 1967. Effect of freezing rates on dendritic solidification of ice from aqueous solutions. Trans. Metall. Soc. AIME 239, 1729–1736. Tang, X., Pikal, M.J., 2004. Design of freeze-drying processes for pharmaceuticals: practical advice. Pharm. Res. 21, 191–200. Uhlmann, D.R., Chalmers, B., Jackson, K.A., 1964. Interaction between particles and a solid–liquid interface. J. Appl. Phys. 35, 2986–2993. Vergote, G.J., Vervaet, C., Van Driessche, I., Hoste, S., De Smedt, S., Demeester, J., 2001. An oral controlled release matrix pellet formulation containing nanocrystalline ketoprofen. Int. J. Pharm. 219, 81–87. Vyas, T.K., Shahiwala, A., Amiji, M.M., 2008. Improved oral bioavailability and brain transport of Saquinavir upon administration in novel nanoemulsion formulation. Int. J. Pharm. 347, 93–101. Wang, L.J., Dong, J.F., Chen, J., Eastoe, J., Xuefeng, L., 2009. Design and optimization of a new self-nanoemulsifying drug delivery system. J. Colloid Interface Sci. 330, 443–448. Zhang, Y., Gao, J., Zheng, H., Zhang, R., Han, Y., 2011. The preparation of 3,5-dihydroxy-4-isopropylstilbene nanoemulsion and in vitro release. Int. J. Nanomed. 6, 649–657. Zhou, H.F., Yang, Y., Liu, G.L., 2010. Preparation and characterization of a lecithin nanoemulsion as a topical delivery system. Nanoscale Res. Lett. 5, 224–230.