Evaluation of the drug loading capacity of different lipid nanoparticle dispersions by passive drug loading

Accepted Manuscript Evaluation of the drug loading capacity of different lipid nanoparticle dispersions by passive drug loading Karin M. Rosenblatt, Heike Bunjes PII: DOI: Reference:

S0939-6411(17)30336-3 http://dx.doi.org/10.1016/j.ejpb.2017.03.010 EJPB 12465

To appear in:

European Journal of Pharmaceutics and Biopharmaceutics

Received Date: Revised Date: Accepted Date:

13 May 2016 28 November 2016 12 March 2017

Please cite this article as: K.M. Rosenblatt, H. Bunjes, Evaluation of the drug loading capacity of different lipid nanoparticle dispersions by passive drug loading, European Journal of Pharmaceutics and Biopharmaceutics (2017), doi: http://dx.doi.org/10.1016/j.ejpb.2017.03.010

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Evaluation of the drug loading capacity of different lipid nanoparticle dispersions by passive drug loading

Karin M. Rosenblatt1, Heike Bunjes1,2*

1

Friedrich-Schiller-Universität Jena, Institut für Pharmazie, Lehrstuhl für Pharmazeutische 2

Technologie, Lessingstraße 8, 07743 Jena, Germany, Technische Universität Braunschweig, Institut für Pharmazeutische Technologie, Mendelssohnstr. 1, 38106 Braunschweig, Germany

*Corresponding author: Heike Bunjes, Technische Universität Braunschweig, Institut für Pharmazeutische Technologie, Mendelssohnstr. 1, 38106 Braunschweig, Germany Tel. (++49) 0531 3915652, Fax (++49) 0531 3918108; [email protected]

1

Graphical Abstract

Drug

Filtration

+

Drug Time

Analysis

(Days) Lipid

Homogenized Dispersion

Incubation under Agitation

Concentration Drug/Lipid [%]

14 Data1_Emulsion Trimyristin Emulsion Secondary TM Secondary Suspension Suspension (b) (b) Primary TM Primary Data1_Suspension Suspension (b)(b) TS Suspension (a) Tristearin Suspension (a) Data1_Smectic Cholesteryl Myristate Smectic CM Crystalline Cholesteryl Myristate Crystalline

12 10 8 6 4 2 0 Bmv

Cmz

Dzp

Gsf

Ffa

Ibu

Q10

Rtp

2

Abstract When using lipid nanoparticles as drug carrier system it is important to know how much drug can be loaded to the nanoparticles. The mainly used drug loading procedure is an empirical approach dissolving the drug in the liquid lipid during preparation of the nanoparticles. This approach does not necessarily lead to the truly loadable amount, as the lipid can, e.g. be overloaded, in particular when it is processed in the heat. In this work, a different procedure, passive drug loading, was evaluated to determine the drug loading capacity of various lipid nanoparticles (supercooled trimyristin emulsion droplets, solid trimyristin nanoparticles, tristearin nanoparticles in the a-modification and cholesteryl myristate nanoparticles in the supercooled smectic as well as in the crystalline state). The nanoparticle dispersions were exposed

to

eight

different

model

drug

compounds

(betamethasone-17-valerate,

carbamazepine, diazepam, flufenamic acid, griseofulvin, ibuprofen, retinyl palmitate, ubidecarenone) in the bulk state, which varied in partition coefficient and aqueous solubility, and equilibrated over time. The passive loading procedure had no relevant impact on the particle sizes or the physicochemical state of the nanoparticles. The loadable drug amount differed distinctly for the different model compounds and also between the different types of lipid nanoparticles. For most compounds, the loaded amount was much higher than the aqueous solubility. Trimyristin-based dispersions generally had the highest loading capacity, the emulsion usually being equal or superior to the solid trimyristin nanoparticles. For betamethasone-17-valerate, however, solid lipid nanoparticles exhibited by far the highest drug load. The extremely lipophilic model drugs retinyl palmitate and ubidecarenone could not be loaded with the passive approach.

3

Keywords Solid lipid nanoparticles, nanoemulsions, supercooled smectic cholesteryl myristate nanoparticles, colloidal drug carriers, solubilization, poorly soluble drugs, drug loading, αmodification

4

1. Introduction Colloidal dispersions of lipid particles are under intensive investigation as drug carrier systems for poorly water-soluble drugs, in particular with regard to their intravenous administration. Such dispersions can be based on a broad variety of lipid compositions and particle structures such as emulsions of liquid oils, suspensions of crystalline glycerides or fatty acids (solid lipid nanoparticles), dispersions of thermotropic or lyotropic liquid crystalline phases, mixed micelles or liposomes. Some of these carrier systems have already been introduced to the pharmaceutical market [1, 2, 3]. Besides the general suitability of a potential drug carrier system for the intended way of administration (e.g., physiological compatibility of its ingredients, appropriate particle size distribution, chemical and physical stability) its ability to solubilize a sufficient amount of drug is an important prerequisite for its successful use as drug delivery system. Hitherto, there is, however, only little systematic knowledge on the solubilization capacity of the different types of colloidal lipid carrier systems. Thus, identifying a suitable carrier system for a given drug is still very much based on empirical approaches. In general, this requires to test a range of carrier systems with different drug loads in order to determine their solubilization capacity for the drug of interest. The commonly used technique for loading colloidal lipid dispersions with poorly soluble drugs is to process the drug with the lipids. During the preparation of colloidal lipid emulsions or solid lipid nanoparticles the drug is usually dissolved in the (molten) lipid prior to particle formation, e.g., by high pressure homogenization. If, for some reason, the solubility limit of the drug is exceeded during manufacturing of the dispersion, the drug will later precipitate in the dispersion. In particular when the systems are prepared at elevated temperature, the solubility in the lipid phase may be overestimated. In solid lipid nanoparticles, the crystalline nature of the solid particle core may distinctly reduce the drug incorporation capacity compared to the liquid state of the lipid [4]. Moreover, the drug partitions between the lipid and the aqueous phase. If the solubility limit in the aqueous phase is exceeded as a result of 5

the partitioning process the drug will precipitate even if the lipid phase could, in principle, still incorporate more drug [5]. As a complication, supersaturation of the aqueous phase may occur leading to crystallization of the drug from the aqueous phase only after a certain time of storage. A delay of precipitation processes may also be caused by slowly proceeding solid state transitions within the crystalline core of solid lipid nanoparticles leading to drug expulsion [4]. In addition to phenomena of delayed drug crystallization, detection of crystallized drug in colloidal dispersions may become a problem when drug crystals are hard to detect, e.g., by light microscopy, due to their shape (e.g. in the case of long, thin needles), a very small size or a low number of crystals in the overall sample. Taken together, these phenomena can make it very difficult to determine the true loading capacity of colloidal lipid dispersions. Moreover, the method is rather time and material consuming since samples with different drug concentrations may have to be prepared and investigated in order to find the solubilization limit of the respective system. This may be a serious obstacle when the amount of available drug is very low as it is usually the case in early pharmaceutical development [6, 7]. With respect to the development of carrier systems for new drugs it would thus be very helpful to find a more efficient way to reliably determine the drug loading capacity of the respective type of nanoparticle. One possibility into this direction might be the addition of drugs crystals to the dispersions and to observe the amount of dissolved drug as described in a study by Sznitowska et al. [8] who investigated the solubilizing effect of submicron emulsions for different drugs as well as a possible destabilization potential of the drug loading process. The aim of our study was to extend these investigations by incubating different types of colloidal lipid dispersions with drug crystals as a passive way of drug loading. This way, it should be possible to prevent overloading of the dispersions since the solubilized drug is always in equilibrium with its crystalline counterpart. For these investigations, four kinds of 6

colloidal lipid dispersions were selected to study the effect of properties such as the inner structure, shape, and the type of matrix lipid of the particles on the drug loading capacity. We investigated melt-homogenized, tyloxapol-stabilized trimyristin dispersions, which can either form a suspension of solid lipid nanoparticles or a nanoemulsion of supercooled droplets depending on storage conditions [9]. This allows to directly evaluate the impact of the physical state of the nanoparticles (i.e. crystalline solid vs. liquid) on their drug incorporation capacity [4]. Tristearin nanoparticles stabilized with poly(vinyl alcohol) were used as a further type of solid lipid nanoparticles since these particles can be kept in the metastable αmodification for a certain time after preparation [10]. Tristearin nanoparticles in the a-form exhibit a different inner structure and a different shape compared to the aforementioned tyloxapol-stabilized solid trimyristin nanoparticles, which rapidly transform into the stable βmodification [9]. The a-form of triglycerides is less tightly packed than the stable bmodification and there are indications that it thus might incorporate drugs more easily [4, 11]. Moreover, a-form tristearin nanoparticles stabilized with poly(vinyl alcohol) are spherical whereas triglyceride nanoparticles in the b-form usually have a platelet-like shape [12, 10, 13]. As a fourth type of dispersion, cholesteryl myristate nanoparticles in the supercooled smectic liquid-crystalline state were investigated. In the liquid-crystalline state, the molecules display a certain degree of order but are still mobile; the particle matrix is much more viscous than in a conventional emulsion. Supercoooled smectic nanoparticles were developed with the aim to combine postulated advantages of solid lipid nanoparticles (in particular, better control over drug release) with the often observed higher solubilization capacity of lipid emulsions [14]. Eight poorly water-soluble model drugs (Fig. 1), which differ in structure, water solubility and partition coefficient, were chosen for our studies. Besides the determination of the drug loading capacity, the time-course of the drug loading process, drug solubility in the aqueous phase and the influence of drug loading on the particle size of the dispersions (as an indication 7

for the physical stability) were investigated. In order to detect potential differences that might result from different drug loading procedures the drug loading capacity of trimyristin nanoemulsion and –suspension and the α-form tristearin nanoparticles was also determined with two of the drugs using the direct drug loading procedure (i.e. addition of the drug prior to preparation of the dispersion).

Figure 1: Molecular structures of the model drug substances used for passive loading

8

2. Materials and Methods 2.1 Materials As matrix lipids for the dispersions, tristearin (Dynasan 118, Hüls/Condea, Witten, Germany), trimyristin (Dynasan 114, Hüls/Condea, Witten Germany) or cholesteryl myristate (Sigma, USA) were used. The dispersions were either stabilized with poly(vinyl alcohol) (PVA, Mowiol 3-83, Clariant, Frankfurt/Main, Germany) or with tyloxapol (Sigma, Seelze, Germany). The aqueous phase contained sodium azide (Sigma, Seelze, Germany) and glycerol (Caelo, Hilden, Germany). For pH adjustment of the aqueous phase acetic acid (100 % DAB, Carl Roth, Karlsruhe, Germany) and sodium hydroxide (Carl Roth, Karlsruhe, Germany) were used. Drugs for passive loading were betamethasone-17-valerate (Bmv, Fagron, Barsbüttel, Germany), carbamazepine (Cmz, Novartis, Basel, Switzerland), diazepam (Dzp, Synopharm, Barsbüttel, Germany), flufenamic acid (Ffa, Lindopharm, Hilden, Germany), griseofulvin (Gsf, Euro OTC Pharma GmbH, Bönen, Germany), ibuprofen (Ibu, Bayer, Germany), retinyl palmitate (Rtp, Sigma, Seelze, Germany) and ubidecarenone (Q10, Kyowa Hakko Kogyo, Tokyo, Japan). Acetonitrile (ACN) for HPLC gradient grade (Fisher Scientific and VWR) and tetrahydrofuran (THF) for HPLC (Fisher Scientific and VWR) were used as solvents for high performance liquid chromatography. Purified water was prepared by filtration and deionization/reverse osmosis (Milli RX 20, Millipore, Schwalbach, Germany). All substances were used as received.

2.2 Methods 2.2.1 Preparation of the dispersions The dispersions were prepared by melt homogenization with a lipid content of 10% (w/w). The compositions of the different dispersions and the processing parameters are shown in 9

Table 1. All dispersions were prepared with two aqueous phases differing in pH: either unbuffered or with pH 4.5 (sodium acetate buffer 25 mM), respectively (pH measurement with pH electrode Pt Inlab 415, Mettler Toledo, Germany). The aqueous phase contained 0.05% (w/w) sodium azide as preservative and 2.25% (w/w) glycerol for isotonization. The lipid was heated to about 10 °C above its melting point. In case of the conventionally loaded dispersions the drug was added to the molten lipid. The emulsifier was dissolved in the aqueous phase and the aqueous phase was heated to the same temperature as the molten lipid. Both phases were combined and pre-homogenized with an ultraturrax (Ultra-Turrax T8, IkaWerke, Staufen, Germany) for 1 min. High pressure homogenization was performed in a Microfluidizer M110S (Microfluidics, USA) for 3 min. The resulting dispersions were either stored at 23 °C or cooled to ~ 5 °C (Table 1). Prior to the drug loading experiments, the nanoemulsion and the α-form nanoparticles were filtered with glass fiber filters. In case of the nanoemulsion filtration (GF/C, nominal pore size 1.2 µm, Whatman) was employed to remove a small fraction (< 1%) of crystalline matrix lipid that was detected by DSC. In αform nanoparticle dispersions filtration (GF/D glass fiber filters, nominal pore size 2.7 µm, Whatman) eliminated a few macroscopic particles. All other dispersions were used without initial filtration.

Table 1: Composition and preparation conditions of the unloaded dispersions Name

Composition of dispersions

Preparation conditions

Matrix lipid (10 Emulsifier /

Temperature

Pressure

Storage

%)

concentration (w/w)

Nanoemulsion

Trimyristin

Tyloxapol 5 %

70 °C

750 bar

23 °C

Nanosuspension

Trimyristin

Tyloxapol 5 %

70 °C

750 bar

5 °C

α-form nanoparticles

Tristearin

PVA 6 %

80 °C

750 bar

5 °C

Smectic cholesteryl

Cholesteryl

PVA 5 %

90 °C

1000 bar

23 °C

myristate nanoparticles

myristate

Crystalline cholesteryl

Cholesteryl

PVA 5 %

90 °C

1000 bar

5 °C

myristate nanoparticles

myristate

10

2.2.2 Physicochemical characterization 2.2.2.1 Photon correlation spectroscopy (PCS) The particle size of the dispersions was assessed by PCS after dilution with particle free water in a Zetasizer Nano ZS (Malvern, U.K.) at an angle of 173°. Four measurements with 300 s each were performed at 25 °C and the last three were used for calculation of the z-average and the polydispersity index (PDI).

2.2.2.2 Differential scanning calorimetry About 13 µl of the dispersions were accurately weighed into aluminum crucibles, sealed and placed in a Pyris 1 DSC (Perkin Elmer, USA). The dispersions were heated to 75 °C (trimyristin), 85 °C (tristearin) or 90 °C (cholesteryl myristate), respectively, with a heating rate of 10 °C/min, held at that temperature for 10 min, cooled to 0 °C (trimyristin, tristearin) or -11 °C (colesteryl myristate) (cooling rate 5 °C/min) and heated again (10 °C/min).

Drug Analysis

Filtration

+

Drug Time (Days)

HPLC / UV Dissolution in respective solvent HPLC / ELSD Lipid

Homogenized dispersion

Incubation under agitation

10 % Lipid 5-6 % Surfactant

Figure 2: Principle of the passive loading procedure with subsequent analysis.

2.2.3 Drug loading The principle of the passive loading procedure with subsequent determination of the solubilized drug (and lipid content) is schematically presented in Fig. 2.

11

2.2.3.1 Passive drug loading About 2 ml of the dispersions was added to between 10 – 80 mg of the bulk drug in 10 ml injection vials. The mixtures were orbitally shaken (75 rpm) in a water bath at 23 °C (Grant OLS 200, Grant Instruments, UK). Trimyristin nanoemulsion, α-form nanoparticles and the corresponding water phases of the dispersions (without emulsifier) were used to investigate time dependent drug loading (between 12 h and 4 weeks, in dependence on the drug). The trimyristin nanosuspension, the smectic nanoparticles and solutions of the emulsifiers in the water phase were investigated at one reasonable time point (7 days or 4 weeks, respectively). In the emulsifier solutions, tyloxapol concentrations were 0.2 mg/ml (suspension), 1.0 mg/ml (emulsion unbuffered) or 1.7 mg/ml (emulsion pH 4.5), respectively, and the PVA concentration was 2 mg/ml. The concentration of tyloxapol was determined by ultrafiltration of unloaded dispersions (cf section 2.2.4.3). The PVA concentration in the aqueous phases of tristearin suspensions (α) and smectic nanoparticles was deduced from the tyloxapol concentrations. Considering the roughly cylindrical shape of smectic particles and a spherical shape of tristearin nanoparticles (α) not more than 2 mg/ml free PVA was assumed. Additionally, one fraction of the trimyristin nanoemulsion and the smectic nanoparticles was put in the refrigerator for approx. 24 h after drug loading in the presence of excess drug to crystallize the matrix lipid. The resulting suspensions were termed “secondary suspensions” (particles were crystallized after passive loading) to differentiate them from the suspensions that were directly incubated with the drug (“primary suspensions”). After shaking, the samples were collected with a glass syringe and filtered with a glass fiber filter (Whatman, Dassel, Germany) in a stainless steel filter holder (Millipore, Schwalbach, Germany). For the nanoemulsion and the aqueous phases GF/C glass fiber filters (nominal pore size 1.2 µm) and for all other dispersions GF/D filters (nominal pore size 2.7 µm) were used. Macroscopically, excess drug crystals could be removed by filtration.

12

For the two acidic drugs, Ibu and Ffa, an acetic acid buffered water phase with a pH of 4.5 which is in the region of the pKa value was used to ensure that a fraction of drug was in the nonionic state. For all other drugs, dispersions with a non-buffered aqueous phase were used. Each incubation experiment (dispersion/emulsifier solution in the presence of bulk drug) was set up in triplicate.

2.2.3.2 Direct drug (over)loading To determine the drug loading capacity obtained by conventional drug loading the drug was added to the molten lipid before homogenization. The amount of drug was chosen with regard to the results obtained from the passive loading: the drug was used in excess to obtain drug overloading. Dzp (concentration added: 0.5%) and Ffa (1.5%) were used as model drugs, the latter with dispersions buffered at pH 4.5. Trimyristin emulsions, suspensions and α-form nanoparticles were investigated. After preparation by high-pressure homogenization, the dispersions were stored at the respective temperatures (Table 1) in the presence of additional drug crystals to provoke crystallization of excess drug. After 7 days and 4 weeks of storage the dispersions were filtered with glass fiber filters as described for the passive loading procedure, characterized with PCS and DSC and the drug and lipid concentration were determined.

2.2.4 Determination of drug and lipid content The drug and the lipid contents of all dispersions were determined with high performance liquid chromatography (HPLC) (BeckmanCoulter, Gold System, Software: 32Karat Ver. 5.0, Krefeld, Germany).

13

2.2.4.1 Drug concentrations A LiChrospher RP18, 5 µm column (Macherey-Nagel, Switzerland) 250 mm x 3 mm (for Dzp and Q10: 125 mm x 3 mm) was used, the flow rate was 1 ml/min. UV detection was carried out with a diode array UV detector 168 (Beckman Coulter, Gold System). The mobile phases for the drugs and the detection wavelengths are specified in Table 2. For sample preparation, 100 µl of trimyristin or cholesteryl myristate dispersion were diluted with 2 ml tetrahydrofuran (THF) and 5 ml of acetonitrile (ACN) were added. Tristearin dispersions were diluted with 5 ml THF and 2 ml ACN. The dilutions were concentrated by aerating with nitrogen to precipitate the lipid. The sample was filtered through a glass fiber filter (GF/C), the filter was washed with acetonitrile and acetonitrile was added to the filtrate up to 10.0 ml (except Rtp loaded trimyristin dispersions: no separation of lipid and thus no filtration required). Each loaded sample was diluted in triplicate and each diluted sample analyzed twice. The recovery rate was determined in the same way with unloaded dispersions, which were spiked with different drug concentrations. All drug concentrations were corrected for the corresponding recovery rates, which were always > 85%. Drug loaded water phases were measured either undiluted or were diluted with ACN in an appropriate way.

14

Table 2: Physicochemical properties and pH of the dispersions used for passive drug loading, mobile phase and detection wavelength of quantitative analysis of the drug with HPLC Drug

Abbr.

logP/logD

Aqueous

pH

solubility

HPLC mobile phase

Detection

[v/v]

Injection volume

[µg/ml] Betamethasone-

Bmv

4.1

1.1

n.b.

ACN/H2O 60/40

240 nm

20 µl

Carbamazepine

Cmz

1.9

220

n.b.

ACN/H2O 50/50

286 nm

20 µl

Diazepam

Dzp

2.8

51

n.b.

ACN/H2O 50/50

254 nm

20 µl

Flufenamic acid

Ffa

4.7 (pH 4)

18 (pH 4)

4.5

ACN/H2O/Acetic acid

288 nm

20 µl

3.9 (pH 5)

130 (pH 5)

17- valerate

65/35/0.1

Griseofulvin

Gsf

2.0

39

n.b.

ACN/H2O 50/50

292 nm

20 µl

Ibuprofen

Ibu

3.4 (pH 4)

270 (pH 4)

4.5

ACN/H2O/Acetic acid

220 nm

20 µl

2.8 (pH 5)

930 (pH 5)

Retinyl palmitate Rtp Ubidecarenone

Q10

14.3 19.1

65/35/0.1

7.9·10

-4

n.b.

ACN/THF 70/30

328 nm

100 µl

5.9·10

-5

n.b.

94 % (THF/ACN

275 nm

100 µl

35/65), 6 % H2O n.b.: not buffered The values for logP/logD and aqueous solubility are predicted properties (at 25°C) according to the SciFinder database [15]. For non-dissociating substances (Bmv, Cmz, Dzp, Gsf, Rtp, Q10) the logP values and the aqueous solubilities at pH 7 are given, for the weak acids (Ibu, Ffa) the logD values and the aqueous solubilites at the specified pH.

2.2.4.2 Lipid concentration The column was a LiChrospher 100 RP18, 5 µm, (Merck, Germany), 250 mm x 4 mm kept at 30 °C. An evaporative light scattering detector (ELSD) (Alltech Varex MKIII ELSD, USA) was used. The mobile phases and corresponding tube temperature and the nitrogen gas flow of the ELSD were: trimyristin (THF/ACN 45/55, 91 °C, 2.23 slpm), tristearin (THF/ACN 60/40, 88 °C, 2.30 slpm), cholesteryl myristate (THF/ACN 50/50, 90 °C, 2.25 slpm). All concentrations were v/v; the flow rate was 1 ml/min. Tristearin and cholesteryl myristate dispersions were diluted with a mixture of THF/ACN (90/10 or 80/20, v/v, respectively), filtered with a glass fiber filter (Whatman, Dassel, Germany, Gf/C, nominal pore size 1.2 µm). The filter was washed with the solvent and the solvent was added to the filtrate up to 10.0 ml. Trimyristin dispersions were diluted with THF/ACN (80/20, v/v) and injected without prior filtration. The recovery rate of tristearin and 15

cholesteryl myristate was determined in the same way with an aqueous PVA (6 %) solution spiked with lipid. The lipid concentration was only determined for some of the trimyristin emulsions and tristearin dispersions. As the lipid concentrations were all in a similar range, the concentration was extrapolated. In case of trimyristin emulsions, the lipid content was determined for all samples after 1 day, 7 days or 4 weeks (exception: Bmv: all dispersions, and Gsf: 12 h samples were investigated additionally). For calculation of 12 h and 2 days results, the 1 day value was used; for 4 days the 7 days results. The same applies to the tristearin results, were only the 1 and 7 day or 4 weeks dispersions were investigated, respectively. All samples were prepared in triplicate and each sample was injected twice.

2.2.4.3 Determination of tyloxapol concentration in the water phase of undiluted dispersions A small homemade stirred ultrafiltration cell with a regenerated cellulose membrane filter (diameter 25 mm, molecular weight cut off 100,000 Da, Millipore) was used to separate the aqueous phase from the nanoparticles. The membrane (pre-equilibrated in the aqueous phase of the corresponding dispersions) was placed into the ultrafiltration cell and about 1 ml dispersion was filtered by applying low nitrogen pressure (less than 1 bar). The first droplets were discarded and the subsequent filtrate was analyzed. The concentration of free tyloxapol in the aqueous phase of unloaded trimyristin dispersions (emulsions and suspensions) was determined with UV spectroscopy (DU640, Beckman Coulter, Krefeld, Germany) at a wavelength of 278 nm (if necessary after appropriate dilution of the filtrate). The filtration procedure was validated with regard to retention of micelles by the membrane with two different tyloxapol concentrations above the critical micelle concentration (0.081 mg/ml [16]). The recovery was more than 97% of tyloxapol at 0.3 mg/ml and 90% at 5 mg/ml after filtration of the 2 nd milliliter. No correction of the determined tyloxapol concentration in the aqueous phase of the dispersions was applied. 16

3. Results 3.1 Aqueous solubility of drugs

Figure 3: Solubility of drugs in the aqueous phases of the same composition as those of the dispersions (determined after 7 days except for Q10 and Rtp (4 weeks); the solubility in the emulsifier-free aqueous phase was not determined for Q10 and Rtp). The aqueous phases of Ffa and Ibu were buffered at pH 4.5. Emulsifier concentrations are specified in the figure legend. *The aqueous phase of the emulsions contained tyloxapol concentrations of 1.0 mg/ml (unbuffered emulsions) or 1.7 mg/ml (emulsions buffered at pH 4.5). **High variability observed for Rtp (see text).

All drugs under investigation were poorly soluble but they still differed distinctly in aqueous solubility (Fig. 3). The final concentration in the aqueous phase (without emulsifier) was usually achieved already after one day (Fig. S1; supplementary information). Q10 and Rtp were not detectable with the applied methods even after four weeks of incubation. The solubility of all other drugs differed by several orders of magnitude (in the range of 0.48 to 117 µg/ml). The rank order of solubility was Cmz > Ibu > Dzp > Gsf > Ffa > Bmv >> Q10 and Rtp. The addition of emulsifiers to the aqueous phases affected the solubility in a different way. For Bmv, the solubility increased by several orders of magnitude. For Cmz, Dzp, Gsf, Ffa, and Ibu the solubility was at most moderately increased, except for Ffa and Ibu in the aqueous phase containing 1.7 mg/ml tyloxapol that induced a considerable increase in solubility in particular for Ffa. This concentration is above the critical micelle concentration (CMC) of tyloxapol [16]. In the emulsifier-containing aqueous phases incubated with Q10, no 17

drug could be detected. For Rtp, the drug concentrations were usually very low even in the presence of emulsifiers, not more than 0.01 mg/ml. In a few cases, however, surprisingly high Rtp concentrations were observed. These are probably indicative for problems occurring during filtration in which Rtp may not be completely separated due to the semisolid character of this compound.

3.2 Passive drug loading 3.2.1 Trimyristin dispersions The passive drug loading capacity of three different types of trimyristin nanoparticles was examined. These types were nanoparticles of the supercooled liquid matrix lipid (emulsion) and of the solid lipid – either crystallized prior to drug loading (primary suspensions (β)) or crystallized after passive drug loading of emulsion particles by cold (refrigerator) storage for 24 h prior to filtration (secondary suspensions (β)). The time dependence of passive loading was investigated with trimyristin emulsions. The trimyristin dispersions used for passive loading usually had a z-average particle size around 100 nm and a PDI between 0.17 and 0.22 (Table 3). The particle size parameters remained within this range during the drug loading process (except for the Bmv-loaded dispersion where the PDI increased to 0.3). The physical state of the particles (liquid or crystalline) was confirmed by DSC; it was not affected by the loading procedure (Fig. S2; supplementary information).

18

Table 3: PCS particle sizes and polydispersity indices of unloaded and passively loaded dispersions. Trimyristin

Tristearin

Emulsion

Secondary suspension

Cholesteryl myristate

Primary suspension

Smectic

crystalline

nm

PDI

nm

PDI

nm

PDI

nm

PDI

nm

PDI

nm

PDI

Unloaded

103

0.22

103

0.21

99.7

0.21

136

0.17

131

0.10

136

0.09

Unloaded (7d)

97.3

0.17

102

0.20

101

0.22

136

0.15

134

0.10

141

0.12

a

Unloaded (4w)

99.3

0.17

106

0.23

102

0.21

138

0.16

136

0.10

N/A

N/A

Bmv (7 d)

99.9

0.19

112

0.30

100

0.20

128

0.15

129

0.10

138

0.11

Cmz (7d)

100

0.20

100

0.20

99.8

0.20

139

0.20

134

0.10

139

0.10

Dzp (7d)

98.2

0.17

105

0.23

98.7

0.19

142

0.16

134

0.10

144

0.13

Gsf (7d)

97.7

0.17

102

0.21

101

0.21

137

0.20

132

0.09

137

0.08

Q10 (4w)

97.7

0.17

103

0.22

98.5

0.20

138

0.15

137

0.11

140

0.09

0.10

b

0.08

a

Rtp (4w)

98.9

0.17

104 a

0.21

99.4

0.20

138

0.16

140

140

Unloaded pH 4.5

118

0.27

N/A

N/A

104

0.22

143

0.21

124

0.10

N/A

N/A

Unloaded pH 4.5 (7d)

101

0.17

105

0.22

106

0.23

136

0.15

128

0.10

131

0.10

Ffa (7d)

103

0.18

110

0.23

104

0.22

156

0.27

128

0.10

137

0.15

Ibu (7d)

102

0.21

101

0.20

103

0.21

133

0.16

129

0.11

142

0.09

a

No samples prepared, therefore no value available

b

Different initial batch: unloaded particle size: 137, nm PDI: 0.08, 4 weeks agitation: 136 nm, PDI: 0.09

19

Figure 4: Drug concentrations (in relation to the matrix lipid concentration) in trimyristin dispersions after passive loading: emulsion (over time: 12 hours up to 7 days), primary and secondary suspension (β). Suspensions were usually loaded over 7 days, except for Q10 and Rtp (4 weeks). Bars marked with *: significant (P = 0.95) difference between loading of primary and secondary suspension. **: Ffa loaded trimyristin suspensions were filtered a second time; the concentrations after the second filtration were: primary suspension: 11.9% ± 1.5%, secondary suspension: 12.0% ± 0.6%.

Passive loading led to pronounced loading of most drugs to the dispersions. The drug loading capacity of the nanoparticles differed distinctly between the different drugs (Fig. 4). The highest loaded amount was observed for Ffa and Ibu, which were in a similar range of about 12% for the emulsions. In contrast, the two most lipophilic drugs with the lowest aqueous solubility, Q10 and Rtp, were hardly detectable in the dispersions even after four weeks of loading (Q10 < 0.012%, Rtp < 0.04% related to the matrix lipid). Loading of emulsions with all other drugs led to moderate drug concentrations between about 0.5 and 3.0% of drug related to the matrix lipid in the following order: Dzp > Bmv > Cmz > Gsf. In general, the diffusion of the drug to the lipid nanoparticles was completed within 12 hours, except for Ffa which still showed an increase in drug concentration after 2 days that seemed to be completed after 7 days. In most cases, the drug concentration in the emulsions was higher compared to that in the suspensions (Ffa, Ibu and Dzp) or in a similar range (Cmz and Gsf). In contrast, the Bmv 20

concentration in the suspensions was more than twice as high as in the emulsion. In the Ffaloaded suspensions, some needle like crystals (obviously representing recrystallized drug) were found macroscopically several weeks after filtration. Thus, both suspensions were filtered once again. Afterwards, surprisingly, the determined drug/lipid ratio was higher than before, in the range of the loading of the emulsions (Fig. 4). The reason for this phenomenon is yet unclear. In order to check if the way of loading has an effect on the achievable drug concentration in the lipid nanosuspensions, the drug loading capacity of suspension particles that had been directly incubated with the drug (“primary suspension”) was compared to that of nanoparticles that had been crystallized after loading the drug to the corresponding emulsion (“secondary suspension”). The drug load tended to be higher in the secondary suspensions (Fig. 4) but the effect was usually rather small. Thus, the drug molecules are predominantly located at the surface of the solid nanoparticles and at most a small fraction can be incorporated in the particle matrix. Of the drugs under investigation only for Bmv (which in general exhibited an extraordinarily high solubilization capacity in the suspensions) a significantly higher amount of drug was found in the secondary compared to the primary suspension. This result suggests that a distinct amount of Bmv (~1% drug/lipid) is incorporated in the solid particles.

3.2.2 Tristearin nanoparticles (α-form) The PCS particle sizes of the unloaded dispersions were around 140 nm with a PDI of 0.17 or 0.21 (Table 3). Shaking or drug loading did not lead to pronounced alterations of the particle size parameters (except upon loading with Ffa where the PDI increased to 0.27). According to DSC, the particles were completely in the a-form at the start of the loading procedure, which usually had only minor effects on the polymorphic state of the particles (Fig. S2, S3). A distinct effect was only observed for Ibu-loaded nanoparticles, where the net enthalpy 21

corresponding to melting of the β-polymorph was roughly 20% of that of the α-polymorph after 7 days of drug loading.

Figure 5: Drug concentration in dispersions of tristearin nanoparticles in the a-form after passive loading (over time; n.d.: not detected)

The concentration of the loaded drugs in the a -form nanoparticles followed a similar order as in the trimyristin suspensions (Fig. 5). Ffa and Ibu showed the highest loading of roughly 2.4% related to the lipid. As observed for the trimyristin emulsions, there was an increase in Ffa concentration over 7 days. An increase in concentration over time was also observed for Bmv ranking third in loaded drug concentration. However, there were large differences between the three investigated samples, in particular for those investigated after 2 and 4 days. Cmz and Dzp were found in a concentration of about 0.5%, with decreasing concentration of Cmz over time. The Gsf content was even lower than that of Cmz. Q10 or Rtp were not detectable in the tristearin dispersions. Usually, the drug contents in the a-form nanoparticles were distinctly lower than those in all types of trimyristin dispersion (emulsion, primary and secondary suspension). Only Bmv showed a drug loading comparable to that of the trimyristin emulsion, but much lower than that of both suspensions. 22

3.2.3 Cholesteryl myristate nanoparticles (smectic and crystalline)

Figure 6: Results of passive drug loading of cholesteryl myristate nanoparticles (smectic particles and particles crystallized after loading). Samples were loaded for 7 days, except Q10 and Rtp (4 weeks). Bars marked with *: significant difference in drug loading between smectic and crystalline nanoparticles (P = 0.95), bars marked with **: samples showed significant degradation of cholesteryl myristate, thus the drug concentration was related to the lipid concentration of the cholesteryl myristate suspension.

The PCS particle sizes of the smectic particles used for loading were around 130 nm with a low PDI (0.10); the size of the crystallized cholesteryl myristate nanoparticles was a bit larger. The drug loading procedure did not affect the particle sizes much (Table 3). The smectic or crystalline state of the nanoparticles was confirmed by DSC (Fig. S2). The dispersions of smectic nanoparticles contained a minor fraction of recrystallized cholesteryl myristate (< 1% in the unbuffered, < 2% in the buffered dispersion). This fraction increased slightly during agitation (unbuffered, unloaded dispersion ~ 2% after 7 days, ~4% after 4 weeks; buffered dispersion 6% after 7 days). All drug-loaded dispersions had a similar crystalline fraction, except for the Ibu-loaded dispersion in which at maximum 9% crystalline cholesteryl myristate was found (the amount could not exactly be quantified due to an overlap of a thermotropic liquid crystalline transition). As for the other types of dispersions, Ffa and Ibu could be loaded at the highest concentrations to the cholesteryl myristate dispersions (about 4% related to the lipid in the smectic nanoparticles) (Fig. 6). The Bmv concentration was about 1.6%, whereas Dzp, Cmz 23

and Gsf were loaded at between 0.4% and 0.9% (drug loading concentrations: Bmv > Dzp > Cmz » Gsf). Q10 and Rtp were again hardly detectable. In most cases, the loading capacity of the smectic particles was higher than that of the subsequently crystallized cholesteryl myristate particles, except for Bmv and Ffa, which displayed almost identical loading of smectic and crystalline nanoparticles. In contrast to the situation with trimyristin nanoparticles, an increase in Bmv-load was thus not observed for crystalline cholesteryl myristate nanoparticles. In general, the drug loading capacity for a given drug was higher than the loading of the α-form nanoparticles and was mostly lower than in all types of trimyristin dispersions. In most cases, the content of cholesteryl myristate was determined some time after drug loading and determination of the drug content. At that time, in some of the samples an additional signal, probably indicating a degradation product of cholesteryl myristate, was observed besides the signal of cholesteryl myristate in chromatography. This signal emerged only in samples of smectic but not of crystalline particles. Apparently, the main degradation took place after filtering off the drug crystals from of the samples, i.e. between determination of the drug and the cholesteryl myristate content. This means that the determined drug concentration for the corresponding smectic particles would probably be related to an erroneously too low lipid concentration after degradation. Thus, the drug content was set in relation to the lipid content of the corresponding crystalline cholesteryl myristate particles for these samples.

3.3 Drug (over)loading – direct method To assess the potential influence of the way of drug loading some dispersions were also loaded by adding drug to the lipid phase during preparation. The tested dispersions were trimyristin emulsion and suspension and tristearin a-form nanoparticles. Dzp and Ffa were used as model drugs. In order to obtain dispersions loaded to a maximum content an excess 24

amount of drug was added to the molten lipid phase (overloading). As for the passively loaded dispersions, excess (precipitated) drug was removed by filtration. After filtration, the particle sizes of the Dzp-(over)loaded dispersions were in the same range as those resulting after passive loading. In contrast, overloading with Ffa led to smaller sizes of the dispersions (trimyristin: z-ave: ~ 85 nm, PDI ~ 0.17; tristearin: z-ave: ~ 120 nm, PDI: ~ 0.16). As usual, trimyristin emulsions did not lead to any thermal event upon heating in DSC whereas the common, structured melting event was observed for the trimyristin suspensions. For the tristearin α-form nanoparticles, direct loading with Dzp had virtually no effect on the polymorphic behavior and mainly the melting event of the α-form was observed during first heating in DSC even after several weeks of storage (Fig. S3). In contrast, direct loading with Ffa led to a pronounced recrystallization exotherm and β-form melting endotherm (Fig. S3). Both were much larger than those detected for the passively loaded dispersions. Thus, the polymorphic structure of the particles was not exactly comparable for passively and directly loaded Ffa-containing dispersions.

Figure 7: Comparison of drug loading of trimyristin emulsions and suspensions and tristearin a-form nanoparticles via the passive and the direct way. Left: Dzp, right: Ffa. Please note that the values given for the trimyristin suspension passively loaded with Ffa reflect the results after the first filtration (for data after a second filtration see caption to Fig. 4).

25

For Dzp, direct loading into trimyristin emulsions led to a significantly higher drug load than passive loading irrespective of the equilibration time (Fig. 7). Also for the trimyristin suspensions, a much higher drug concentration was observed for directly loaded Dzp (the drug load after 7 days was even higher than in the directly loaded emulsion). However, a distinct decrease in concentration was observed after 4 weeks in the suspension indicating slowly progressing drug crystallization. The drug concentration at this time point still remained above that of the passively loaded suspensions. A similar phenomenon was observed for the tristearin suspension where the Dzp concentration dropped to half its value between week 1 and week 4 (also remaining above the value for the passively loaded counterpart). The results of direct loading of Ffa into the trimyristin emulsion and suspension were comparable to that of the passive loading procedure without any indication of retarded drug crystallization between one and four weeks of storage (Fig. 7). In contrast, almost twice as high drug loads were found in directly loaded tristearin nanoparticles compared to passively loaded systems. Again, there was no sign of delayed drug precipitation but rather a trend towards increasing drug concentration over time. Tentatively, the higher drug load observed for directly loaded tristearin nanoparticles may explain the differences in polymorphic behavior compared to the passively loaded system in particular if the additional drug load is assumed to be localized preferentially in the particle core.

26

4. Discussion

Figure 8: Comparison of the drug concentration in the aqueous phases and the different types of nanoparticle dispersions (concentration in the overall dispersion) obtained after 7 days of drug loading (4 week for Q10 and Rtp).

4.1 Aqueous solubility The solubilities of the model drugs in the emulsifier-free aqueous phases used for preparation of the dispersions are in a similar range as calculated values (Table 2). Emulsifier-containing aqueous phases in some cases tend to solubilize higher amounts of the poorly water-soluble drugs, as particularly observed for Bmv. The high solubility of Bmv in each of the three emulsifier-containing aqueous phases suggests an affinity of this substance to amphiphilic structures. The drug concentration in the aqueous phases including the emulsifier-containing ones was much lower than in passively loaded dispersions (Fig. 8). It can thus be concluded that the drug was indeed loaded to the lipid nanoparticles. The drug concentrations determined in the aqueous phases are, however, presumably not directly comparable with drug concentrations in the aqueous phases of the dispersions (i.e., in the presence of the lipid) as that is rather partition and not (only) solubility controlled. 27

4.2 Passive drug loading

Figure 9: Comparison of passive loadability of the different types of lipid nanoparticles: Trimyristin emulsion, primary and secondary suspension (β), tristearin suspension (α) and cholesteryl myristate smectic and crystalline nanoparticles.

Passive loading proved to be a suitable method to load different kinds of lipid nanodispersions with poorly water-soluble drugs. The extent of loading depended distinctly on the type of nanoparticles and even more clearly on the properties of the drugs under investigation (Fig. 9). Ffa and Ibu always led to the highest drug loads whereas Cmz and Gsf did not seem to have much affinity to either of the nanoparticulate systems employed. The most lipophilic and least water-soluble drugs (Q10 and Rtp) could only be loaded in trace amounts with this method during the investigated time frame. The results obtained upon passive loading with Q10 and Rtp were rather unexpected as it is known that both drugs can be incorporated into solid lipid nanoparticles when using the direct loading procedure. In particular Q10 showed tremendous differences between direct and passive loading. This drug can be loaded to in virtually any relation to triglycerides via the direct way [4, 17]. In contrast, it was not possible to achieve at least a marginal loading with

28

passive loading. This is probably due to the very low aqueous solubility of this compound. The temperature during preparation may also play a role, as direct loading usually occurs above the melting point of Q10 which remains in the supercooled liquid state even after the attachment to solid lipid nanoparticles when loaded at higher concentrations [4, 17]. Rtp could also be directly loaded in much higher amounts (up to 5%) to solid lipid nanoparticles than observed here [18]. Obviously, the passive loading procedure does not work properly in the case of Q10 and Rtp. As indicated above, this may be related to an extremely high lipophilicity / low water solubility of these drug substances (Table 2) that does not allow the diffusion of a reasonable number of drug molecules through the aqueous phase. As such a phenomenon would set a clear limit to the applicability of the passive loading procedure, the underlying mechanisms require to be elucidated in more detail.

4.2.1 Trimyristin dispersions Generally, a higher loading capacity of emulsions compared to suspensions (β) could be expected because a liquid matrix of the particles is supposed to solubilize a higher amount of drug compared to a solid crystalline matrix. Most of the drugs indeed showed that behavior, with up to 50% higher loading in relation to that of the suspension (β) (Fig. 4). A much higher amount of drug in the suspension (β) (as seen for Bmv) compared to emulsions either suggests a better incorporation capacity of the solid matrix or a high affinity of the drug to the surface of the particles. A higher incorporation capacity of the suspension seems unlikely as the secondary suspension (β) was loaded in the emulsion state. Hence, the maximum incorporated amount could only be similar to emulsions. In particular, the higher loading of the primary suspension (β) as compared to the emulsion supports the assumption of drug adsorption to the particle surface. Due to the platelet-like shape of the solid lipid particles (β) [12, 13], their specific surface area is much larger than in spherical emulsion droplets. A 29

tendency of corticosteroids, in particular Bmv, to localize at the surfaces of solid lipid nanoparticles has already been shown previously [19]. The high solubility in surfactant containing aqueous phases of the dispersions compared to the plain aqueous phases supports the assumption of an affinity of Bmv to amphiphilic structures. Experiments with primary and secondary trimyristin suspensions (β) were performed to differentiate how much drug is incorporated in the particles and how much is adsorbed to the surface of the particles. It is inconceivable that drug molecules can be encapsulated into the already crystalline structure of the particles solidified prior to loading. Hence, loading of the primary suspension should only represent the extent of drug molecules localized on the surface of the solid nanoparticles. In contrast, when the drug is loaded into liquid nanoparticles that are subsequently crystallized some drug might theoretically become entrapped within the solidifying particle core. Thus, the amount in the secondary suspension (β) should reflect the sum of drug adsorbed and that incorporated. A significantly higher loading of the secondary suspension (β) compared to the primary one could only be observed for Bmv and Dzp; only Bmv seemed to be incorporated in a marked amount (~1%, Dzp ~0.3%). These results support the assumption of a very low drug incorporation capacity of solid lipid nanoparticles with a highly ordered matrix. Nevertheless, solid nanoparticles do show a distinct loading capacity, often being even in the range of emulsions, which most likely is caused by adsorption of the drug to the particles. While this would still make them suitable as drug delivery systems, a sustained release of adsorbed drug seems rather unlikely. Previous investigations on the localization of model drugs in lipid nanosuspensions also found a major fraction of drug to be adsorbed to solid lipid nanoparticles rather than incorporated [19,20]. For suspensions passively loaded with Ffa, drug recrystallization was observed after filtration for hitherto unknown reasons. Tentatively, a shift in pH causing a different solubility of Ffa

30

might play a role. The possibility of a pH shift is also supported by the longer loading time of the drug, which also might be result of a change in pH.

4.2.2 Tristearin nanoparticles (α) Tristearin particles in all cases had the lowest solubilization capacity. This may be related to the different structure of the tristearin compared to trimyristin nanoparticles (physical state, polymorph, shape) as well as to differences in chemical properties of the two triglycerides. Although the inner structure of the matrix of tristearin nanoparticles in the α-modification is less ordered than that of nanoparticles in the β-modification, it is still a solid matrix. Thus, it is uncertain whether an incorporation of drug molecules into the particles via passive loading is possible. The deviating behavior of Bmv, being loaded in a comparable concentration as in the trimyristin emulsion and both types of cholesteryl myristate particles, supports the idea of a surface adsorption-driven loading of Bmv assumed from the results with the trimyristin dispersions. As most of these aforementioned nanoparticles are spherically or cylindrically shaped, a comparable surface area and with that a similar amount of adsorbed drug can be assumed.

4.2.3 Cholesteryl myristate nanoparticles (smectic and crystalline) The liquid crystalline smectic structure of the cholesteryl myristate particles exhibits a higher mobility of the matrix molecules than a solid matrix, but is much more viscous than the liquid matrix in emulsion droplets [14]. Thus, incorporation of drug molecules in the particles by passive loading might be possible. Both types of cholesteryl myristate particles, smectic and crystallized, usually solubilized a lower amount of drug than trimyristin dispersions but the loadability was higher than that of tristearin suspensions (α). The latter obviously results from loading into the liquid crystalline 31

state (and not in the crystalline state) where the incorporation is apparently more easily possible. A lower loading capacity of smectic nanoparticles compared to trimyristin emulsions can be explained by a higher acceptance of foreign substances/ mobility in the liquid state than the liquid crystalline one. Although it could be expected that a chemically different matrix, which is in some cases closer to the structure of the model drugs (e.g. for Bmv) than triglycerides, might lead to a higher incorporation capacity in the crystalline state, the loading in trimyristin suspensions (β) was always higher, even of the primary suspensions (β). This might be attributed to the much larger specific surface area of the trimyristin suspensions in relation to the non-spherical smectic and crystalline cholesteryl myristate nanoparticles [13]. Moreover, the solubilization capacity of the smectic phase, which is less ordered than the trimyristin crystals, may not be as high as anticipated. Again, this may depend on the nature of the drug to be incorporated: smectic cholesteryl myristate nanoparticles were rather favorable in a preformulation study for a lipophilic drug candidate (trimyristin nanoparticles were, however, not included in this study) [21].

4.3 Direct loading With Dzp, all dispersions could be loaded in a higher amount in the direct way than with passive loading. This might be either due to incomplete equilibration of the overloaded Dzp dispersions or to a (higher) incorporation into the interior of the matrix as compared to passively loaded nanoparticles. As the direct loading led to a higher Dzp concentration even in the emulsion, an incomplete equilibration seems more likely. In contrast, direct loading of the trimyristin dispersions with Ffa resulted in similar concentrations as passive loading, whereas the loading of the tristearin suspension (α) was also higher. The differing behavior of the two drugs could be caused by different time periods necessary for equilibration of the system. This assumption is supported by the decrease in Dzp concentration between 7 days and 4 weeks of storage in both trimyristin dispersions and in the 32

tristearin suspension (α). Crystallization of overloaded drug was apparently retarded, even though the dispersions were stored in the presence of crystalline drug. It might be expected that direct drug loading of the a-form nanoparticles can lead to higher drug loads than passive loading. Such an effect was indeed observed for Ffa-loading of tristearin nanoparticles. The difference observed between passive and direct drug loading might give an estimate on the amount of drug located within the particles – more than 1% for the Ffa-loaded tristearin nanoparticles under investigation here. However, as loading with Ffa accelerated the transition of the a-form nanoparticles into the stable b-modification, the stability of drug loading remains to be investigated for this system. The fact that differences were observed between the results of passive and direct loading poses the question about the reliability of loading results obtained with both methods. The results obtained with the Dzp-loaded suspensions underline the difficulties that may arise due to delayed crystallization of the drug. Such processes – and thus the direct loading procedure as it was applied here – may easily lead to an overestimation of the achievable drug load. On the other hand, passive loading may underestimate the maximum drug concentration that can be obtained for the dispersions in particular in cases where drugs could, in principle, be accommodated in a solid particle core or where drug transport to the particles is hampered. Thus, the passive loading procedure seems to be particularly promising with carrier particles that provide a certain fluidity in their matrix. In any case, the results from passive loading can be assumed to lead to a more cautious estimate of the achievable drug load. This would put a developer on the safe side as it would avoid overloading of the dispersions. Compared to approaches that start from the investigation of the solubility of the drug in the respective matrix lipid passive loading can also be applied to lipids that are in a special physical state in the nanoparticles (like the supercooled liquid state of trimyristin particles or the supercooled smectic state of cholesteryl myristate). Moreover, the contribution of the nanoparticle surface

33

as a potential site of drug localization can be included into the considerations when using the passive loading method.

5. Conclusion The passive loading procedure can be used to evaluate the drug loading potential of different colloidal dispersions in an efficient way. Usually, the loading process is finalized within a rather short period of time and it can be carried out with small amounts of dispersion and drug. Thus, it has the potential to be used as screening method if the limitations of the process are taken into adequate consideration. In order to understand these limitations in more detail the mechanisms underlying the passive loading process need to be elucidated. The method can be employed to collect systematic knowledge on the interaction of drugs with lipophilic carrier systems that may form the database, which in the end may help to provide a more rational basis for carrier selection. Apart from its use as screening method, passive loading may also be applicable for small scale production if only little material is available and/or if containment is required for toxicological reasons.

Acknowledgements This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

34

References [1]

H. Bunjes, Lipid nanoparticles for the delivery of poorly water-soluble drugs, J. Pharm. Pharmacol. 62 (2010) 1637–1645.

[2]

H. Bunjes, J. Kuntsche, Lipid nanoparticles based on liquid crystalline phases, in: V.P. Torchilin, M.M. Amiji (Eds.), Handbook of Materials for Nanomedicine, Pan Stanford, Singapore, 2011, pp. 445–493.

[3]

P. van Hoogevest, X. Liu, A. Fahr, M.L.S. Leigh, Role of phospholipids in the oral and parenteral delivery of poorly water soluble drugs, J. Drug Deliv. Sci. Technol. 21 (2011) 5–16.

[4]

K. Westesen, H. Bunjes, M.H. Koch, Physicochemical characterization of lipid nanoparticles and evaluation of their drug loading capacity and sustained release potential, J. Controlled Release 48 (1997) 223–236.

[5]

C. Washington, Stability of lipid emulsions for drug delivery, Adv. Drug Deliv. Rev. 20 (1996) 131–145.

[6]

S. Balbach, C. Korn, Pharmaceutical evaluation of early development candidates; “the 100 mg-approach”, Int. J. Pharm. 275 (2004) 1–12.

[7]

J. Maas, W. Kamm, G. Hauck, An integrated early formulation strategy – From hit evaluation to preclinical candidate profiling, Eur. J. Pharm. Biopharm. 66 (2007) 1–10.

[8]

M. Sznitowska, S. Janicki, E. Dabrowska, K. Zurowska-Pryczkowska, Submicron emulsions as drug carriers, Eur. J. Pharm. Sci. 12 (2001) 175–179.

[9]

H. Bunjes, K. Westesen, M.H. Koch, Crystallization tendency and polymorphic transitions in triglyceride nanoparticles, Int. J. Pharm. 129 (1996) 159–173.

[10] K.M. Rosenblatt, H. Bunjes, Poly(vinyl alcohol) as emulsifier stabilizes solid triglyceride drug carrier nanoparticles in the α-modification, Mol. Pharmaceutics 6 (2009) 105–120. [11] V. Jenning, M. Schäfer-Korting, S. Gohla, Vitamin A-loaded solid lipid nanoparticles for topical use: drug release properties, J. Controlled Release 66 (2000) 115–126. [12] H. Bunjes, M.H.J. Koch, K. Westesen, Effect of particle size on colloidal solid triglycerides, Langmuir 16 (2000) 5234–5241.

35

[13] S. Petersen, F. Steiniger, D. Fischer, A. Fahr, H. Bunjes, The physical state of lipid nanoparticles influences their effect on in vitro cell viability, Eur. J. Pharm. Biopharm. 79 (2011) 150–161. [14] J. Kuntsche, K. Westesen, M. Drechsler, M.H.J. Koch, H. Bunjes, Supercooled smectic nanoparticles: A potential novel carrier system for poorly water soluble drugs, Pharm. Res. 21 (2004) 1834–1843. [15] Chemical Abstracts Service - CAS, SciFinder, 2016, www.cas.org/products/scifinder, accessed April 2016. [16] Sigma Aldrich, Tyloxapol, Sigma Aldrich product information (2008). [17] H. Bunjes, M. Drechsler, M.H.J. Koch, K. Westesen, Incorporation of the model drug ubidecarenone into solid lipid nanoparticles, Pharm. Res. 18 (2001) 287–293. [18] V. Jenning, S.H. Gohla, Encapsulation of retinoids in solid lipid nanoparticles (SLN), J. Microencapsulation 18 (2001) 149–158. [19] R. Sivaramakrishnan, C. Nakamura, W. Mehnert, H.C. Korting, K.D. Kramer, M. Schäfer-Korting, Glucocorticoid entrapment into lipid carriers — characterisation by parelectric spectroscopy and influence on dermal uptake, J. Controlled Release 97 (2004) 493–502. [20] T. Blaschke, L. Kankate, K.D. Kramer, Structure and dynamics of drug-carrier systems as studied by parelectric spectroscopy, Adv. Drug Deliv. Rev. 59 (2007) 403–410. [21] E. Kupetz, L. Preu, C. Kunick, H. Bunjes, Parenteral formulation of an antileishmanial drug candidate – Tackling poor solubility, chemical instability, and polymorphism, Eur. J. Pharm. Biopharm. 85 (2013) 511–520.

36