Determining drug release rates of hydrophobic compounds from nanocarriers

Abstract

Obtaining meaningful drug release profiles for drug formulations is essential prior to in vivo testing and for ensuring consistent quality. The release kinetics of hydrophobic drugs from nanocarriers (NCs) are not well understood because the standard protocols for maintaining sink conditions and sampling are not valid owing to mass transfer and solubility limitations. In this work, a new in vitroassay protocol based on ‘lipid sinks’ and magnetic separation produces release conditions that mimic the concentrations of lipid membranes and lipoproteins in vivo, facilitates separation, and thus allows determination of intrinsic release rates of drugs from NCs. The assay protocol is validated by (i) determining the magnetic separation efficiency, (ii) demonstrating that sink condition requirements are met, and (iii) accounting for drug by completing a mass balance. NCs of itraconazole and cyclosporine A (CsA) were prepared and the drug release profiles were determined. This release protocol has been used to compare the drug release from a polymer stabilized NC of CsA to a solid drug NP of CsA alone. These data have led to the finding that stabilizing block copolymer layers have a retarding effect on drug release from NCs, reducing the rate of CsA release fourfold compared with the nanoparticle without a polymer coating.

This article is part of the themed issue ‘Soft interfacial materials: from fundamentals to formulation’.

1. Introduction

Drug nanoparticles (NPs) and nanocarrier (NC) formulations of highly lipophilic drugs enable the delivery of compounds that previously could not be administered at therapeutic levels by conventional formulations [1,2]. Both pure drug NPs [1,2] and more complex NC constructs, such as liposomes, nanocapsules, polymeric NPs, micelles and polymersomes can improve the observed therapeutic effect of drug compounds by increasing solubility, improving pharmacokinetics or altering biodistribution. Drug dissolution or release data at simulated physiological conditions are an important component in preclinical development to evaluate drug formulations prior to in vivo testing, for quality control and also for regulatory approval [3]. The challenge of predicting in vivo drug dissolution or release through in vitro methods for nanoformulations is becoming more widely appreciated [4–6]. The in vitro assays for oral formulations of soluble drugs are often not appropriate as indicators for in vivoperformance of nanoscale formulations of lipophilic drugs. The variability of results obtained for the different methods has been investigated for both drug-loaded biodegradable microspheres [7] as well as pure drug NPs [8]. Further, poor correlation is often found between in vitro and in vivo release [9–11]. The problem with available assays centres on mass transfer limitations when dealing with highly insoluble compounds in NC or NP form. These limitations are associated with maintaining sink conditions during the assays, diffusion resistance associated with low solubility, long length scales in conventional test methods and sampling without inadvertently entraining NCs.

In this work, we present a novel in vitro method to overcome issues of low solubility, poor sensitivity, mass transfer limitations and separation to enable investigation of release kinetics of hydrophobic drugs from NCs. The limitation of previous assays, specifically the dominance of mass transfer limitations, was illustrated by the insightful experiments of Petersen et al. [12], that used a flow cytometry approach to quantify the release rate of lipophilic fluorophores from NCs to a lipid emulsion phase, which served as the hydrophobic sink. This method circumvents the problem of separating the released fraction from the NCs, as flow cytometry is used to quantify the fluorescence in the acceptor emulsion particles without interference from the NCs. Supporting experiments were performed to demonstrate assay conditions that lack kinetic hindrances, saturation limitations and separation steps. A 3000×faster rate constant for transfer was obtained for the transfer to lipid droplets (surface area approx. 500 cm²) compared with the non-emulsified lipid (surface area approx. 2 cm²) [12]. In the referenced method, transfer of the fluorescent dye Nile red occurred in less than 5 min for the lipid microparticle sink, but took several days even in the small geometry of the 24-well plate format. The intimate mixture of lipid microparticles with the NCs decreases diffusion length scales, and importantly mimics the interaction of the NCs with red blood cells, lipoproteins and cell walls that occurs in vivo. The experiment by Petersen et al. required the use of a fluorescent compound, because fluorescence was used as the readout in flow cytometry. Therefore, the technique is not applicable to non-fluorescent pharmaceutical actives, which includes almost all drugs.

Here, we present an extension of the earlier Petersen approach, to develop a drug release assay that is applicable to all hydrophobic drugs presented in NP or NC form, schematically depicted in figure 1d. Lipid microparticles are prepared as high surface area, hydrophobic, sink particles. Magnetic NPs are included in the lipid formulation to enable separation of lipid microparticles from drug NCs by the introduction of a magnetic field. After incubation with NC formulations, HPLC analysis of the drug in the lipid carriers enables time-resolved assessment of release kinetics. The separation protocol is validated against data obtained with the fluorescence flow cytometry assay according to Petersen et al. [12], and the release of the hydrophobic drugs itraconazole (ITZ) and cyclosporine A (CsA). Development experiments explore various parameters of the assay, including separation efficiency, solubility of the drug of interest in the liquid lipid phase and the relative ratios of the NC and lipid phases. We believe this assay provides a platform to assess drug release from NCs or NPs that can be widely adopted in the nanomedicine community.

Figure 1.

(a) 1 : 3 CTO lipid prepared with Miglyol after heating to 90°C and allowing to cool. (b) 1 : 3 CTO lipid (250 mg) which has been melted, mixed with 50 μl of Fe₃O₄ in hexane, and allowed to cool. (c) The mixture in (b) after homogenization in a PVA solution. (d) Schematic for the transfer process that occurs during dissolution. (e) Workflow for the separation process. Aliquots of the NC/microparticle mixture are separated, using the magnetized column at each time point. The ‘NCs’ fraction is eluted from the column via gravity, whereas the lipid droplets are retained on the magnetized beads. When the column is removed from the separator the lipid is eluted as the ‘released drug’ fraction. Each fraction is then assayed for drug content. (Online version in colour.)

2. Methods

(a). Materials

Perylene (greater than 99%), pyrene (greater than 99%), Nile red (99%), cholesterol (95%), vitamin E (97%) and tetrahydrofuran (THF, HPLC grade) were purchased from Sigma-Aldrich (St Louis, MO). Hostasol red (HosR) was from Clariant GmbH (Coventry, RI). The synthesis of 1.5 kDa hydroxyl terminated polystyrene (PS_1.5k-OH) and polystyrene-b-polyethylene glycol (PS_1.5k-b-PEG_5k-OH) has previously been reported [13]. ITZ was provided by Merck (West Point, PA), and CsA was purchased from LC Laboratories (Woburn, MA). Compritol 888 ATO and Miglyol 812 N were generously provided by Gattefoss Corporation (Paramus, NJ) and Sasol (Eatontown, NJ)/Condea Chemie (Witten, Germany), respectively. Vinol 528 (PVA; 93–100 kDa; and 87–89% hydrolysed) was from Air Products (Allentown, PA) but is no longer commercially available. Mowiol 3–83 was from Clariant (Frankfurt/Main, Germany). Superparamagnetic magnetite NPs in hexane were prepared by a thermal decomposition method as described previously (8.5 nm, 0.267 g Fe₃O₄ ml⁻¹ hexane) [14]. Prior to use, water was purified via 0.2 μm filtration and four-stage deionization to a resistivity of 17.8 MΩ or greater (NANOpure Diamond, Barnstead International, Dubuque, IA) or by passing it through a Milli-Q device (Millipore, Billerica, MA). Soya bean oil was manufactured by Crisco and was used as received.

MACS MS columns and a Mini MACS separator were obtained from Miltenyi Biotec (Auburn, CA). The MACS MS columns are prepared with a surfactant coating to promote wetting. Prior to the first use, the columns were rinsed as prescribed by the manufacturer. To maintain the columns for additional uses, after each experiment the columns were placed in a 0.1 wt.% boiling PVA solution to solubilize any residual material, flushed with 1 ml aliquots of boiling water twice, and then dried with pressurized air. Finally, 400 μl of 1% Brij97 in ethanol was added to the column and allowed to dry under vacuum to ensure wetting in subsequent experiments.

(b). Nanocarrier preparation

NCs were produced by flash nano-precipitation (FNP) in a multi-inlet vortex mixer [15]. One stream of THF (12 ml min⁻¹) was mixed against three streams of water (3×36 ml min⁻¹). The composition of the THF solution for each formulation is detailed in table 1. The resulting NC suspensions were dialysed in a Spectra Por regenerated cellulose dialysis membrane (Spectrum Laboratories, Inc., CA) with a molecular weight cut-off of 6–8 kDa. The 2 l external volume of water was stirred continuously and changed three times over 24 h. NC suspensions were diluted before dynamic light scattering (DLS) measurements were taken to determine the average particle size (Zetasizer, Malvern, Worcestershire, UK). Each measurement was taken two to three times.

Table 1.

Nanocarrier formulations.

formulation ID	core	concentration (mg ml−1)	block copolymer	concentration (mg ml−1)	size (nm)
NC1	perylene/PS-OHa	0.1/10	PS1.5k-b-PEG5k-OH	10	143
NC2	Nile red/VEb	0.02/10	PS1.5k-b-PEG5k-OH	10	110
NC3	ITZc	20	PS1.5k-b-PEG5k-OH	20	170
NC4	CsAd	20	n/a	0	401
NC5	CsA	20	PS1.5k-b-PEG5k-OH	20	435
NC6	CsA	20	PLA3.8k-b-PEG5k-OCH3	20	428

^aPS-OH, hydroxy terminated polystyrene.

^bVE, vitamin E.

^cITZ, itraconazole.

^dCsA, cyclosporin A.

(c). Fluorescence microscopy

Fluorescence microscopy was performed using a Nikon Eclipse TE300 Fluorescence microscope, and micrographs were captured with a Retiga EXi CCD digital camera. A Nikon high-pressure mercury lamp was used to illuminate samples. A red band-pass filter exciting samples between 510 and 560 nm and detecting emission above 600 nm was used to image HosR dye.

(d). Drug release assay protocols

(i). Preparation of bulk lipid mixtures and differential scanning calorimetry measurements

The bulk lipid mixtures were prepared by heating Compritol and either soya bean oil or Miglyol, in various Compritol-to-oil (CTO) weight ratios, above 90°C. The molten mixture was agitated to mix and cooled to room temperature. A trace fraction of HosR (electronic supplementary material) was dissolved in the molten lipid to facilitate lipid mass quantification in development experiments. The correlation between mass of lipid and HosR was determined by dissolving known masses of dyed lipid and measuring the fluorescence intensity of HosR. The resulting correlation between HosR and mass of the lipid was used in control experiments to determine the mass of lipid recovered in each sample (electronic supplementary material). Differential scanning calorimetry (DSC) measurements were performed on mixtures of Compritol and soya bean oil with CTO=1:0, 1 : 1, 1 : 2 and 1 : 3. Samples (5–9 mg) were heated at a rate of 4°C min⁻¹ to 100°C, held at 100°C for 2 min, then cooled at 4°C min⁻¹, in vented aluminium crimp seal pans with a nitrogen purge at 50 ml min⁻¹ (DSC Q2000, TA Instruments).

(ii). Assessment of optimal melt emulsification parameters

Several batches of lipid microparticles were prepared using a melt emulsification method [16], without a high-pressure homogenization step, in order to assess the optimal parameters to be incorporated in the Release Assay Protocol, below. The lipid mixture (250 mg) and 15 ml of a PVA solution (0, 0.1% or 1% w/w) were separately heated to greater than 90°C in a water bath. Magnetite NPs in hexane (8.5 nm, 0.267 g ml⁻¹ Fe₃O₄) were added to the fluorescent lipid mixture from step §2d(i) (0, 32 or 64 μl). The mixture was swirled by hand until homogeneous and heated to allow hexane evaporation. The hot PVA solution was added to the melted lipid, and the mixture was homogenized with an Ultra-Turrax homogenizer (T 25, Jahnke & Kunkel, Staufen, Germany) for 2, 5, 10 or 15 min at 3, 7, 13, 15 or 24 k r.p.m. The suspensions were allowed to cool and maintained at room temperature until used. Particle size distributions were obtained using a Microtrac S3500 particle size analyser (Microtrac. Montgomeryville, PA) by dispersing suspension aliquots in circulating water. Particle size distributions are reported as weighted by volume.

(iii). Release assay protocol

Based on the optimized parameters found in §2d(ii), the following final procedure for lipid suspension preparation was determined. First, one part of Compritol and three parts Miglyol (1 : 3 CTO) were heated in a water bath until the Compritol had melted and then allowed to cool (figure 1a). A 250 mg portion of 1 : 3 CTO lipid was heated to greater than 90°C in a 30 ml borosilicate vial (25×95 mm) in a water bath and Fe₃O₄ NPs in hexane (50 μl) were added to the melt, swirled to mix and allowed to sit in the water bath approximately 30 s (figure 1b). PVA in water (0.1 wt.%, 15 ml) was heated in a boiling water bath until greater than 90°C and then poured into the lipid mixture. The lipid was emulsified in 0.1% PVA at 13 k r.p.m for 10 min and then allowed to cool to room temperature (figure 1c).

To remove lipid droplets which could not be retained on the column, as well as excess PVA polymer in the suspension, 0.5 ml aliquots of the original suspension were passed through the MS Column in the MiniMACS separator. The retained fraction was washed twice with 0.5 ml of water. The particles retained in the column were recovered by removing the MS Column from the separator and gently flushing with 0.5 ml of water, using the plunger provided with the column. This was repeated until the desired volume of particles for an experiment was obtained, usually 7 ml. The resulting lipid concentration was approximately 10 mg ml⁻¹, confirmed by fluorescence spectroscopy as described in the electronic supplementary material. This process was also used to concentrate the lipid. For example, 1 ml of the original suspension was passed through the MS column with the separator, washed two times with 0.5 ml of water and eluted by the addition of 0.5 ml of water, to achieve a twofold higher concentration of approximately 25 mg ml⁻¹ lipid. A scintillation vial for carrying out each dissolution/release experiment was prepared by rinsing with a 1 wt.% PVA solution and allowing the residue to dry under house vacuum, in order to limit lipid particle adhesion to the vial walls for the duration of the experiment.

The separation step of the release assay protocol is depicted schematically in figure 1e. Each experiment was initiated by adding NCs in water to the 10 mg ml⁻¹ lipid suspension. After briefly shaking, a 0.5 ml aliquot was set aside to later determine the initial concentration of drug in the mixture. For each time point, a 0.5 ml aliquot of the lipid/NC suspension was added to the top of the MS column in the MiniMACS separator. The liquid fully drained by gravity (about 1 min) and was collected. Two 0.5 ml washes of water were sequentially added to the top of the MS column, and allowed to drain. The 1.5 ml of eluent was collected in a single tube. The MS column was then removed from the MiniMACS separator, and 0.5 ml of water was added and flushed through using the plunger provided with the columns. This flush was repeated twice more, and the combined 1.5 ml of liquid was collected in a tube. The separation process took 5–6 min total. Both tubes were freeze-dried and prepared for analysis, as described in §2d(iv). Time points are reported as the time at which the separation process was initiated.

(iv). Sample preparation for analysis

Separated samples of the release media in water were frozen on dry ice and lyophilized on a shelf at room temperature (VirTis Advantage, Gardiner, NY) with a condenser temperature set to −78°C and a pressure of less than 100 mTorr. Dry powders were obtained after 24 h and were dissolved with the addition of 0.4 ml of THF and sonication (Fisher Scientific FS6) for 10 min at room temperature. Finally, the tubes were centrifuged at 7 k r.p.m. for 15 min in a micro centrifuge (Eppendorf Centrifuge 5415 C) to settle insoluble material. The supernatant was analysed by HPLC or fluorescence spectroscopy as described below.

(v). Fluorescence spectroscopy

Fluorescence measurements were performed using a Hitachi F-7000 fluorescence spectrophotometer (scanned at 240 nm min⁻¹, 400 PMT voltage). Excitation/emission (EX/EM) slit widths used were either 5 nm/5 nm or increased to 10 nm/10 nm, to increase sensitivity. The excitation wavelength, λ_EX, and emission wavelength, λ_EM, are given for each fluorophore in the electronic supplementary material. The peak fluorescence emission intensity at λ_EM was correlated to the concentration of dye using standards of each dye dissolved in THF (electronic supplementary material).

(vi). High-performance liquid chromatography

HPLC analysis was performed with a Synergi Fusion C18 column at room temperature. For ITZ, a mobile phase of 50/50 acetonitrile/water at 1 ml min⁻¹ was used with a detector wavelength of 230 nm. The area under the curve (AUC) for the ITZ peak was determined by integration Surveyor PDA Plus acquisition software, and the AUC was correlated to drug concentration in THF by running 20 μl injections of ITZ standards with concentrations 1–100 μg ml⁻¹. The AUC versus concentration was fit by linear regression with R²>0.99, and the relationship predicted all experimentally measured values to within less than or equal to 5%. The derived relationship was used to calculate the concentration of unknown samples.

Separation of CsA peak resolution was achieved with a 40 min gradient mobile phase with a flow rate of 0.75 ml min⁻¹. The mobile phase composition was held at 70/30 acetonitrile/water for 5 min, ramped to 100/0 over 15 min, held for 5 min and then ramped back down to 70/30 over the last 15 min. CsA was detected at 210 nm. A standard curve was plotted by measuring the AUC for 20 μl injections of known concentrations of CsA from 2.3 to 230 μg ml⁻¹. The linear regression yielded a good fit of the data, with R²>0.99. The limit of quantitation was 11.5 μg ml⁻¹.

(e). Nile red release by flow cytometry protocol

The acceptor o/w emulsion consisted of 5 wt.% Miglyol 812 stabilized with 3 wt.% PVA (Mowiol 3–83) in an aqueous phase containing 2.25% glycerol and 0.01% thiomersal. The emulsion was prepared at room temperature using an Ultra-Turrax (T8, IKA Labortechnik, Staufen, Germany) for 15 min. The particle size was determined by laser diffraction without stirring using a Coulter LS 230 Particle Sizer (Beckman Coulter Inc., US-Fullerton). Eight measurements of 90 s were averaged and analysed using Mie theory with a refractive index of 1.332 for water and 1.45 for the sample. The mean particle size by volume was 7.5 μm.

The release of Nile red was determined by mixing 315 μl of NCs with 1 ml of the o/w emulsion in Eppendorf tubes. The ratio between the donor and acceptor was 1 : 100 w/w (referring to the whole weight of the NCs). The tubes were incubated and shaken at 37°C and 100 r.p.m. At different time points, 15 μl samples of the mixture were diluted with 1 ml water and measured in the flow cytometer (Epics XL MCL, Beckman Coulter Inc., US-Fullerton) at an approximate count rate of 250 events per second. Fluorescence excitation was performed at 488 nm, and the emitted fluorescence was detected at the photomultiplier tube with wavelengths between 665 and 685 nm. The measurement was stopped after 10 000 detection events. The amount of drug transferred into the acceptor o/w emulsion during the experiment was calculated from a calibration curve that was obtained by measuring acceptor emulsion samples loaded with different concentrations of Nile red.

3. Results and discussion

(a). Properties of the bulk lipid

Crystalline lipid phases have a limited capacity for loading drug compounds for drug delivery applications [17], and higher loading capacities are achieved by creating a mixture of oil and a high melting point fatty acid ester [18]. In this work, lipid-based particles were prepared to form a drug sink for the assay with a liquid oil phase to provide a high drug loading capacity and a crystalline lipid phase for particle mechanical integrity. It was hypothesized that the liquid lipid might inhibit crystallization of the solid lipid upon solidification of the molten lipid. Lipid mixtures were prepared with soya bean oil at various CTO ratios. The DSC cooling curve for each sample after melting is plotted in figure 2. The enthalpy of crystallization (AUC) for each sample decreased as the weight fraction of oil was increased. Furthermore, the onset of the phase transition was delayed as the weight fraction of oil was increased. Each peak was integrated to quantify the enthalpy of crystallization, and there was a linear correlation between enthalpy of the transition and the weight fraction of Compritol (figure 2 inset). When the enthalpy was normalized by the weight fraction of Compritol, a constant value of 125 J g⁻¹ of Compritol was obtained, which corresponded to the enthalpy of the transition for the pure material. Therefore, the lipid mixture undergoes phase separation during cooling and solidification, containing pure Compritol crystals and liquid oil, where the crystalline regions of the Compritol bind the soya bean oil. Many hydrophobic drugs have good solubility in liquid oils, and 75 vol% liquid oil in the lipid mixture was chosen to provide the lipid sink for dissolving drug.

Figure 2.

Characterization of mixtures of Compritol and soya bean oil after melting and solidification, at CTO=1:0, 1 : 1, 1 : 2 and 1 : 3. At a cooling rate of 4°C min⁻¹, there is a depression in the onset of solidification. Inset: the integrated values for the enthalpy of the transition are plotted as a function of the Compritol fraction in the mixture (filled squares). These values are normalized by actual weight fraction of Compritol, and plotted for each sample (open circles).

(b). Lipid particle size distribution

PVA-stabilized microparticles were formed from the lipid mixtures by emulsion-melt homogenization. The size distribution (PSD) was a critical parameter in developing the protocol; the largest particles had to be less than 30 μm to flow through the column packing, as per the manufacturers’ specifications. To examine the effect of homogenizer tip speed on PSD, 250 mg of 1 : 2 CTO lipid was homogenized for 5 min in 15 ml 1 wt.% PVA at speeds from 3 to 24 k r.p.m. The PSDs for the resulting suspensions are shown in figure 3a. At low speeds of 3 and 7 k r.p.m., the PSD is unimodal with average diameters of 78.9 and 18.9 μm, respectively. As the speed is increased to 13 k r.p.m., the average particle size decreases, but PSD becomes bimodal, with populations of 10 μm particles and 3 μm particles. At 18 k r.p.m., the relative fraction of 10 μm particles decreases and the population of smaller particles dominate. At the highest homogenization speed of 24 k r.p.m., the PSD shifts further to the left, resulting in a population of particles of average diameter 1.5 μm. Overall, higher homogenization speed leads to smaller size particles with wider relative size distribution.

Figure 3.

Particle size distributions of lipid suspensions and results of magnetic lipid retention experiments. (a) 1 : 2 CTO lipid emulsified in 1 wt.% PVA for 5 min at 3 k r.p.m. (filled squares), 7 k r.p.m. (open squares), 13 k r.p.m. (filled circles), 15 k r.p.m. (open circles) or 24 k r.p.m. (filled triangles). (b) The same 1 : 2 CTO lipid emulsified at 13 k r.p.m. in 1 wt.% PVA for 2 min (filled squares), 5 min (open squares) 10 min (filled circles) or 15 min (open circles). (c) 1 : 2 CTO lipid was homogenized for 15 min at 13 k r.p.m. in 0 wt.% PVA (filled squares), 0.1 wt.% PVA (open squares) or 1 wt.% PVA (filled circles). (d) 1 : 3 CTO lipid was homogenized with 0.1 wt.% PVA for 10 min at 13 k r.p.m. with the addition of 0 μg (filled squares), 32 μl (open squares) or 64 μl (filled circles) of Fe₃O₄, suspended in hexane, to the lipid melt. (e) The percentage of lipid mass retained in the column is as a function of magnetic nanoparticles added to the lipid mixture prior to homogenization, with formulations including 0, 25, 50, 100, 150 or 200 μl of the Fe₃O₄ suspension (f) lipid retention in the magnetized column, relative to the amount of lipid added to the column, where all preparations were prepared with 50 μl Fe₃O₄. (g–i) Fluorescence microscopy of diluted samples of (g) the original lipid suspension, (h) the fraction of particles retained on the magnetized column and (i) the lipid droplets washed off the magnetized column.

The effect of homogenization time on PSD was determined by homogenizing at a fixed speed of 13 k r.p.m. for times from 2 to 15 min. The resulting PSDs are plotted in figure 3b. After 2 min of homogenization, the resulting PSD was bimodal, with populations of particles at 15 and at 3 μm. As the time is increased from 2 to 5 min, the fraction of small particles increased relative to the fraction of large particles. After 10 min, the PSD had average diameter of 3.4 μm. Increasing the time further resulted the additional presence of 200 μm aggregates. This is evidence of over processing of particles, which results in aggregation or coalescence if newly formed particles are not adequately stabilized by surfactant or emulsifier [19].

Increasing time of homogenization ultimately resulted in a unimodal distribution. At high speeds, there are populations of large and small particles, whereas at long times, the population is roughly monomodal. At higher speeds, there is a larger distribution of the number of times each particle passes through the homogenizer tip, making the relative size differences between particles more pronounced. Increasing homogenization time gradually reduced particle size through repeated passes through the high-intensity shear region near the impeller tip. For homogenization times greater than 5 min, there was a greater likelihood that particles had been through the tip a similar number of times, and thus had a similar size. The homogenizer tip speed had a more direct impact on average particle size than the homogenization time did: an eightfold increase in tip speed (3–24 k r.p.m.) changed average particle size from 80 to 1.5 μm, whereas a similar change in homogenization time (2–5 min) only reduced particle size by roughly two-thirds.

The PVA was important for stabilizing droplets and preventing coalescence. For 1 : 2 CTO lipid, homogenized at 13 k r.p.m. for 10 min, 20–30 μm aggregates persisted when there was no PVA in solution, whereas this population disappeared when 0.1 or 1% PVA was used as a stabilizer (figure 3c). The addition of 32 or 64 μl of Fe₃O₄ NPs in hexane to 250 mg of molten 1 : 2 CTO lipid had essentially no effect on the PSD relative to a formulation which has no magnetic NPs (figure 3d). The calculated volume ratio of magnetic NPs to lipid, assuming full evaporation of hexane prior to homogenization, is 8.54 μl Fe₃O₄ per ml lipid, meaning that the magnetic NPs make up less than 1% of the volume of each lipid droplet.

(c). Magnetic separation control experiments

(i). Recovery of nanocarriers and lipid droplets

The complete recovery of (i) NCs in a solution of PVA and (ii) the lipid microparticles without magnetic inclusions with two 0.5 ml washes was first established and is described in the electronic supplementary material.

(ii). Magnetic lipid retention

Effective separation of the NCs from the magnetic lipid droplets required that the magnetic lipid droplets were retained on the column in the presence of the separation magnet. The fraction of lipid retained on the column was determined as a function of the amount of magnetic Fe₃O₄ that was added to the lipid formulations. Suspensions of lipid droplets, prepared with 1 : 3 CTO lipid, with soya bean oil and the HosR fluorescent dye, homogenized for 10 min at 13 k r.p.m. with 0.1% PVA were prepared with varied amounts of added magnetic NPs in hexane. The particles were passed over the MACS MS Column placed in the Mini MACS separator and washed twice with 0.5 ml of DI water, which was collected. The column was then removed from the separator, and the retained particles were eluted with three 0.5 ml washes of DI water, which was collected as the lipid retained fraction. The effectiveness of lipid particle retention was quantified by the relative amount of fluorescence in the two isolated samples after lyophilization. In figure 3e, these data are presented and show that a fivefold increase in the amount of magnetic NPs added to the lipid, resulted in a marginal increase in the fractional lipid retention on the column. The fraction retained was between 54% and 72% when Fe₃O₄ addition was increased from 25 to 200 μl.

The MiniMACS column is capable of capturing a maximum of 1×10⁷ cells, per the manufacturer information. Dilutions of lipid from 17 to 2 mg ml⁻¹, with 48 μg Fe₃O₄ per ml lipid, were passed over the column, washed and the retained particles were then recovered. It was consistently found that only approximately 60% of the administered concentration of lipid was retained on the column (figure 3f). Because the fraction of lipid retained on the column was independent of the number of particles passed over the column, surface saturation of the packing in the column was not occurring. Based on these two control experiments, 40% of particles that are not retained by the magnetized column should be removed prior to performing a release experiment, rather than using large fractions of the magnetic particles to increase retention, because the magnetic particles are the most expensive component of the assay. This ‘filtration’ step was performed with the magnetized column, in order to wash away the 40% of particles that could not be retained on the column, and was carried out as described in the final release assay protocol (§2d(iii)). The particles captured by the column were eluted and collected separately from the particles washed off the column, to be used for release studies. From the original suspension, 62% of the lipid mass is retained on the column and recovered in the ‘retained lipid’ fraction (table 2). This sample was passed over the magnetized column again, and 88% of the lipid mass was retained on the column again, with only 3% of the mass eluting in the ‘washed lipid’ fraction (table 2). Similarly, the first fraction washed off the magnetized column was passed over the column again, and 87% of the lipid mass was washed off again, with 13% of the mass being retained on the column (table 2). The particles retained on the column once are retained on the column to a high degree, justifying the use of pre-filtration to prepare particles for a release assay. More than 90% of the mass is recovered in each separation step, and any lost mass is presumed to be in the column.

Table 2.

Fractional recovery of lipid mass in magnetized column and in the wash.

suspension added to column	lipid mass washed off column	lipid mass retained on column
original lipid suspension	39.4±2.2%	61.8±1.2%
‘retained lipid’—61.8% of original lipid suspension
— fraction of original lipid suspension retained on column	3.9±1.1%	87.9±1.2%
‘washed lipid’—39.4% of original lipid suspension
— fraction of original lipid suspension washed off column	86.6±3.2%	12.8±4.4%

Fluorescence images of the original suspension, the ‘retained lipid’ and the ‘washed lipid’ were captured (figure 3g–i) and show that the original polydisperse suspension is divided into greater than 5 μm particles, which are the retained lipid, and smaller particles, which are not retained on the column. Assuming that the magnetite NPs are dispersed randomly throughout the lipid droplets formed, these images indicate that particles which were smaller do not have sufficient magnetic force to be retained on the column, and instead are washed away with the flow of the liquid. This washing step has several other benefits, such as washing away excess PVA, which might interact with the NCs in a release assay. Lastly, the magnetic lipid droplets may be resuspended in a medium other than water by eluting particles with 0.5 ml of the appropriate buffer.

(d). Release experiments

Having established the lipid formulation, the protocol for preparation and having demonstrated satisfactory retention in the column in the presence of the magnetic field and recovery once the field is removed, the functionality of the system as a tool for quantifying the release or dissolution of hydrophobic compounds without mass transfer limitations under sink conditions was validated. Because there are a number of assays which have emerged that will allow dissolution or release profile for NPs and NCs to be generated, the following experiments aim to better understand parameters that influence sink conditions and demonstrating the sink condition is maintained.

(i). Validation with flow cytometry assay

To validate this new protocol based on lipid droplets and magnetic separation, the transfer of Nile red out of vitamin E NCs (table 1, NC2) was determined using both assays. However, in the flow cytometry experiment, there was a 1 : 100 ratio of NCs to acceptor emulsion, whereas the ratio was 1 : 200 in the magnetic separation assay with lipid microparticles. The results for the two experiments are plotted in figure 4a. The rate of transfer of the Nile red is very rapid, as was previously observed [12], and the two methods yield the same transfer profiles. The flow cytometry method is capable of resolving time points less than 1 min apart, whereas the time required for the magnetic separation steps reduces the temporal resolution to approximately 5 min. However, for the intended release in medical applications, this resolution is adequate.

Figure 4.

(a) A comparison of Nile red dye transfer profiles from vitamin E nanocarriers (NCs; NC2) in a 1 : 200 mass ratio with the lipid microparticles, quantified by the protocol of magnetic separation of lipid droplets developed in this report (filled squares) and by fluorescence of emulsion droplets in a 1 : 100 ratio by flow cytometry as developed by Petersen et al. [12] (open circles). Representative datasets for the dissolution of (b) CsA (NC5) and (c) ITZ (NC3) NCs. Both formulations are stabilized by PS_1.5k-b-PEG_5k-OH. The fraction of drug is calculated based on the total amount of drug in a 0.5 ml sample of the mixture of NCs and lipid. The NC fraction (open squares) is the mass of drug quantified in the 1.5 ml of liquid eluted from the column in the MACS separator and the released drug fraction (filled squares) is the mass of drug recovered in lipid after flushing the lipid from the column. Each point represents the average of two samples separated at each time point. The sum of these two fractions is plotted as the mass balance (filled circles).

(ii). Drug solubility: Miglyol versus soya bean oil

Control experiments in §3(a)–3d(i) were carried out with lipid formulations using soya bean oil as the liquid oil. Drug solubility in the liquid oil becomes a critical issue for both maintaining sink conditions and for quantifying the amount of drug. Because fluorescence assays are very sensitive, the NC concentration can be very dilute relative to the concentration of lipid and saturation of the fluorophore in oil is not an issue. HPLC assays have limited resolution, requiring a much higher concentration of NCs for reliable quantitation, which raises issues of saturation.

A preliminary experiment was conducted to determine the solubility of ITZ in soya bean oil relative to Miglyol, which had been used by Petersen et al. [12] in the development of the flow cytometry assay. The solubility of ITZ was determined by adding excess bulk powder to Miglyol and soya bean oil, and equilibrating at room temperature. The suspension was 0.2 μm filtered, and the filtrate was analysed by HPLC to determine the drug concentration. It was found that there was more than a 10 times greater solubility for ITZ in Miglyol (6.4 μg ml⁻¹) than in soya bean oil (0.47 μg ml⁻¹). Then, ITZ NCs were prepared with the formulation detailed in entry NC3 in table 1 and lipid particles were prepared with 1 : 3 CTO lipid, where the oil was Miglyol or soya bean oil, with 50 μl of Fe₃O₄ NPs in hexane, homogenized for 10 min at 13 k r.p.m. in 0.1% PVA. The lipid particles were filtered over the magnetized column prior to the experiment to isolate the retained fraction and the fraction not retained on the magnetized column was discarded.

Dissolution profiles for NPs frequently follow a first-order rate equation of the form

3.1

and can be fit by the integrated equation

3.2

to obtain the rate constant, k. The initial drug concentration in suspension is c₀ in equation (3.2). Release data are often well-fit by this model, but these do not shed light on the underlying processes governing drug dissolution in the system. The rate constant for dissolution obtained by this method is convenient for the relative comparison of different formulations, and includes contributions from diffusion coefficients, surface area and boundary layer thickness.

The dissolution is plotted as a first-order process in figure 5a according to equation (3.2), for both experiments, and the rate constants are obtained by linear regression. When ITZ NCs are incubated with lipid droplets with Miglyol as the liquid oil phase, k=−2.2×10⁻²±2×10⁻³ (R²=0.97). In contrast, the dissolution of the same ITZ NCs is four times slower in the presence of lipid particles prepared with soya bean oil, with k=−5.5×10⁻³±5×10⁻⁴ (R²=0.94). It can be inferred from the slower release rate that the solubility of ITZ in the soya bean oil microparticle phase is lower, and the approach to saturation in the liquid media slows the rate of release, because the assumed sink condition is not valid. Based on these results, all following experiments are based on the formulation of magnetic lipid microparticles formulated with Miglyol, as outlined in §2d(iii).

Figure 5.

(a) Impact of drug solubility in oil on the dissolution rate of the drug, demonstrated by capturing the release profile in the presence of ITZ nanocarriers (NCs; NC3) using lipid droplets formulated with 1 : 3 CTO, using either soya bean oil (filled squares) or Miglyol (open circles) as the liquid oil phase. (b) The rate constant for dissolution of CsA is k=−3.8×10⁻²±3×10⁻³ when the NPs are not coated with an amphiphilic diblock copolymer (NC4, filled circles) but slows to k_{PLA3.8k-b-PEG5k-OCH3}=−8.4×10⁻³±7×10⁻⁴ when stabilized by PLA_3.8k-b-PEG_5k-OCH3 (NC6, open squares) and k_{PS1.5k-b-PEG5k-OH}=−9.1× 10⁻³±1×10⁻⁴ when stabilized by PS_1.5k-b-PEG_5k-OH (NC5, filled squares). These experiments were performed as described in §2d(iii) with Miglyol in the lipid phase. Inset: the particle size distribution of cyclosporine A particles when formulated without a stabilizer (NC4, filled circles) and with PS_1.5k-b-PEG_5k-OH (NC6, open squares) or PLA_3.8k-b-PEG_5k-OCH3 (NC5, filled squares) as a stabilizer.

(iii). Release profile dependence on lipid concentration

The next control experiment performed was to demonstrate that the dissolution assay quantifies the intrinsic drug release rate. To accomplish this, the dissolution profile of CsA NCs formulated as shown in table 1 (NC4), without any polymer stabilizer, was obtained by incubation with three different concentrations of magnetic lipid droplets. The standard concentration of lipid (1× lipid) was approximately 10 mg ml⁻¹ after homogenization and filtration to remove approximately 40% of the particles that are not retained in the column. A portion of these particles were diluted to 5 mg ml⁻¹ lipid with DI water (0.5× lipid), and another portion was concentrated to 20 mg ml⁻¹ lipid as described in the methods (2× lipid). Hence, the ratio of drug to lipid mass in these experiments was 1 : 12.5, 1 : 25 and 1 : 50.

The concentration of drug in the NC fraction, the released drug fraction and the sum of both fractions is plotted in figure 6. In figure 6a, the dissolution curves for the 1× lipid and 2× lipid samples are almost identical, whereas the 0.5× lipid sample has a slower dissolution profile, relative to the other two samples, though eventually all of the drug is transferred to the released drug fraction. Even though the lipid is not saturated in the 0.5× lipid sample, the approach to saturation of the reduced amount of lipid phase slows the rate of dissolution. Therefore, at the condition used in our experiments, the solubility of CsA in the lipid phase when incubated with 10 mg ml⁻¹ lipid (1× lipid) is sufficient to capture the intrinsic release rate.

Figure 6.

Demonstration that a sink condition has been achieved in the release assay. The concentrations of drug quantified (a) in the NC fraction, (b) in the released drug fraction and (c) the sum of the mass in the two fractions are plotted per ml of the mixture. The dissolution profiles correspond to the same CsA nanocarriers (NCs), without stabilizer, in the presence of the standard concentration of lipid (1× lipid) (filled circles), half the concentration of lipid (0.5× lipid) (filled triangles) and double the concentration of lipid (2× lipid) (filled squares) and shows that, at the two higher levels of lipid, there is complete dissolution of the NCs and that the dissolution rate is the same. In contrast, at the half lipid concentration, the release rate slows and is not complete, indicating saturation of the lipid phase.

We have focused on establishing the appropriate concentration of lipid to capture the intrinsic release rate of CsA in this experiment. Owing to prior optimization of the lipid formulation process, we have not explored maintaining constant lipid concentration and increasing the lipid surface area and number of particles, by reducing the mean particle size. As observed in §3c(ii), there is likely a lower limit on the size of a lipid particle that will efficiently be retained on the magnetized column, making the execution of such an experiment problematic. The previously cited work by Petersen et al. demonstrated that increasing lipid surface area and decreasing diffusion distances increases the rate of Nile red transfer to a lipid phase. While it follows that further decreasing lipid particle size at a fixed lipid concentration might increase the rate of transfer in either assay, practical limitations of either particle retention on the column or detection by flow cytometry present challenges to executing these additional experiments to determine the magnitude of the effect.

(iv). Typical release curves and a mass balance

Having demonstrated that the assay approach can maintain sink conditions with sufficient lipid and does not present mass transfer limitations, release curves for CsA and ITZ NCs, stabilized by PS_1.5k-b-PEG_5k-OH, were determined. Because the fraction of the drug in the NCs and released into the lipid phase can be independently measured, this assay uniquely allows for all of the drug mass in the system to be accounted for. The independently measured fraction of drug in the NCs and released drug are plotted for the CsA NCs and ITZ NCs in figure 4b and c, respectively. The sum of the two fractions is plotted to show the mass balance at each time point, relative to 100%, which is plotted as a dotted line.

(v). Effect of block copolymer layer on dissolution

The polymer composition provides a way to control in vivo circulation time [20]. It has sometimes been observed that hydrophobic NPs prepared by FNP have the potential for self-stabilizing through electrostatic interactions [21]. For this work, CsA NPs were prepared by FNP without a block copolymer stabilizer and with either the PS_1.5k-b-PEG_5k-OH or PLA_3.8k-b-PEG_5k-OCH₃. The NPs are stable with or without the block copolymer layer. Without a block copolymer, the NPs are 400 nm, and with the block copolymer, the diameter increases by approximately 40 nm (table 1 and figure 5b inset). The increase in diameter indicates that the CsA nucleation and growth is likely completed prior to block copolymer adsorption. The increase of 40 nm corresponds to the thickness of the block copolymer layer, which approximately corresponds to the thickness of the layer that has previously been seen for coating inorganic crystals with block copolymers by FNP [22].

The dissolution profiles for the three formulations are plotted in figure 5b. The dissolution rate for the CsA NC formulations which are coated with the block copolymers are nearly identical, with k_{PLA3.8k-b-PEG5k-OCH3}=−8.4×10⁻³±7×10⁻⁴ and k_{PS1.5k-b-PEG5k-OH}=−9.1×10⁻³±1×10⁻⁴. In contrast, the dissolution rate of a CsA NP without a block copolymer layer is significantly faster, with k=−3.8×10⁻²±3×10⁻³. The NPs that are 1 : 1 CsA/block copolymer have release rates that are slowed by 4×, because the hydrophobic block of the polymer provides a partial barrier which diminishes the rate of drug diffusion to the solid–liquid interface. While the release rate of CsA from polymer-coated NCs is reduced relative to the dissolution rate of CsA NPs without a polymer coating, the release is still rapid, with 50% of the drug released in approximately 100 min.

4. Conclusion

In this work, the viability of using magnetic lipid microparticles to serve as a sink for hydrophobic drug dissolution assays has been demonstrated. The lipid properties were optimized for the purposes of the assay, including high surface area, good column retention and recovery, and concentration. Filtering the initial emulsion through the magnetized column enables positive selection of retained particles, as well as resuspending particles directly in a medium other than water, such as buffer. The assay was validated internally by demonstrating that, when the drug has high solubility in the lipid phase, there is a sufficient concentration of lipid at 10 mg ml⁻¹ 1 : 3 CTO lipid microparticles, and sink conditions are maintained. The recovery of drug in this assay is good and mass balances can routinely be performed to track all drug molecules. Saturation of media as seen in dialysis is not an issue, time-consuming and high-energy separation steps such as centrifugation are unneeded, and this method is not limited to the study of fluorescent compounds. This assay opens the door to a wide range of future studies. One example carried out in this work was quantifying the extent to which block copolymer coatings on NCs control release. The rate CsA NP dissolution was reduced by a factor of four by polymer coatings, but the type of polymer coating was not found to significantly affect the rate of dissolution. Further studies are needed to clarify the agreement between in vitro and in vivo results.

Data accessibility

Additional information and data supporting this article have been uploaded as part of the electronic supplementary material.

Authors' contributions

S.M.D., A.A.B., C.R. and R.K.P. contributed to the conception of the novel assay design and execution of related experiments. M.D. and H.B. developed and executed the flow cytometry protocols. S.M.D., A.A.B., M.D., H.B. and R.K.P. performed analysis and interpretation of data. S.M.D. and A.A.B. drafted the article and all authors provided substantial critical review and final approval of the version to be published.

Competing interests

We have no competing interests.

Funding

S.M.D. was supported by Merck & Co. through the Merck Graduate Student Fellowship, M.D. by a grant from the Egyptian government.