J. Phys. Chem. B 1999, 103, 10373-10377
10373
Observation of Size-Dependent Melting in Lipid Nanoparticles Tobias Unruh,* Heike Bunjes, and Kirsten Westesen Institut fu¨ r Pharmazie, Pharmazeutische Technologie, Friedrich-Schiller-UniVersita¨ t Jena, Lessingstrasse 8, 07743 Jena, Germany
Michel H. J. Koch European Molecular Biology Laboratory, Hamburg Outstation, EMBL c/o DESY, Notkestrasse 85, 22603 Hamburg, Germany ReceiVed: April 19, 1999
The differential scanning calorimetry thermograms of suspensions of trimyristin nanoparticles in the stable β-polymorph display several melting events. Particle thickness distributions obtained by analysis of difference X-ray diffraction patterns recorded during melting reveal that these discrete events correspond to the melting of classes of particles differing in thickness by only one unit cell. A large increase in lattice constants with decreasing particle thickness was observed, indicating that classical surface thermodynamics is not applicable to particles consisting of only a few molecular layers.
Introduction The formulation of poorly-water-soluble drugs for intravenous administration is a demanding task in pharmaceutical technology. Despite various approaches such as solubilization utilizing micelles, emulsions, liposomes, and suspensions, for many drugs formulations with high bioavailability and minimal side effects are not available. The formulation problem becomes even more serious when the drugs have to be targeted to defined tissues. Targeting is strongly desirable for unspecifically acting, highly toxic drugs such as cytostatics, which are often only poorly soluble in water (e.g., taxol). Suspensions of solid lipid nanoparticles (i.e., aqueous colloidal suspensions of, for example, triglycerides) have been proposed as a novel type of drug carrier device for intravenous administration.1 Due to their solid state, the release of incorporated drugs should be slow. It has been observed that some colloidally dispersed triglycerides exhibit an uncommon melting behavior expressed by multiple discrete peaks in their differential scanning calorimetry (DSC) thermograms over a range of about 10 °C. In the present study trimyristin nanoparticles were investigated. The meltingtemperature range of these nanoparticles is clearly shifted below the melting temperature of bulk trimyristin at approximately 56 °C (cf. Figure 1). The well-resolved DSC peaks mark multiple discrete melting events. Extensive simultaneous smalland wide-angle X-ray diffraction studies on these suspensions indicate that the phenomenon is not due to polymorphic transitions since only the reflections of the β-modification are observed upon heating.2 It has been proposed that this phenomenon is due to particle-size effects. The present study was initiated to elucidate the mechanism of the stepwise melting process in more detail using a novel method of analysis of the X-ray diffraction data. Experiments and Methods Trimyristin nanosuspensions were prepared by high-pressure melt homogenization at 85 °C. The suspensions consisted of * Author to whom correspondence should be addressed. E-mail: toun@ pt.uni-jena.de.
Figure 1. DSC curve of a trimyristin suspension. The numbers indicate the thickness of the particles, expressed in number of unit cells, melting at the corresponding temperature.
10% (w/w) trimyristin (D114, Hu¨ls AG), 3.2% Lipoid S100 (Lipoid KG), 0.8% sodium glycocholate (Sigma, approximately 99%), and water containing 2.25% glycerol and 0.01% thiomersal (Synopharm). After preparation of a crude pre-emulsion by sonification at 85 °C, an APV Gaulin, Micron Lab 40 homogenizer was used for homogenization. Every sample was cycled 4 times at 1200 bar and once at 1500 bar. The PCS (photon correlation spectroscopy) measurements were performed on dilute samples using a ZetaPlus apparatus (Brookhaven Instruments Corporation) and yielded z-averages of the recrystallized particles around 130 nm. The DSC measurements were carried out using a Micro DSC III (Setaram) at a scan rate of 0.1 °C/min. The weight of both the sample and the reference (pure water) was about 150 mg. Electron micrographs were recorded using a Zeiss CEM 902A (LEO, 80 kV, objective aperture 17.3 mrad). Samples were frozen as thin films in liquid propane, freeze-fractured at 173 K and 5 × 10-6 bar (BAF 400, Bal-Tec) and shadowed with platinum/carbon at 45°. The replicas were stabilized by
10.1021/jp9912612 CCC: $18.00 © 1999 American Chemical Society Published on Web 11/06/1999
10374 J. Phys. Chem. B, Vol. 103, No. 47, 1999 vertical deposition of pure carbon. Samples for cryo-transmission electron microscopy were prepared by placing a drop of the dilute dispersion on an uncoated copper grid, removing excess liquid with filter paper. The sample was cryo-fixed by shooting it into nitrogen-cooled liquid ethane. Excess ethane was removed. The sample was transferred with a cryo-transfer unit into the precooled cryo electron microscope (beam current ≈ 1 µA, T ≈ 80 K). X-ray diffraction patterns were recorded with a Kratky camera (HECUS M. BRAUN-Graz X-ray Systems) on a conventional X-ray source (Seifert generator ID3003, CuKR1,2, Ni-Filter) and on the double-focusing synchrotron radiation small-angle camera of the EMBL on the storage ring DORIS of the Deutsches Elektronen Synchrotron (DESY) at Hamburg.3 Essentially the same results were obtained in both cases, but a much better temperature resolution was achieved with synchrotron radiation where measuring times are approximately 1000 times shorter.4 The data recorded with the line-focused Kratky camera (cf. Figure 2 A) were desmeared using a direct method of beamheight correction6 (cf. Figure 2 B). Difference diffraction patterns suitable for analysis were obtained by subtracting two patterns recorded at slightly different temperatures after correction by Lorentz and polarization factors for powder samples7 (cf. Figure 2 C). The difference patterns were fitted by Pearson VII functions using a Marquardt-Levenberg least-squares fit.8 Reflections recorded with the conventional source were fitted using a Pearson VII function for KR1 and another one for KR2. To eliminate the instrumental broadening, each of the fitted curves was deconvoluted using Stokes’ method9 with a fitted profile of a polycrystalline trimyristin sample sintered for 7 days at 50 °C (standard sample), where the line profiles should be almost entirely due to instrumental broadening (cf. Figure 2 D). Strain broadening was separated from size broadening by analyzing the corrected difference diffraction patterns using the Fourier single-line method7 (cf. Figure 2 E). This approach implies that the thickness of the crystals parallel to the reciprocal lattice vector of the reflecting (001)-plane of trimyristin is constant over one crystallite. As illustrated in Figure 3 A the trimyristin crystals fulfill this criterion because of their platelet shape, where the (001)-plane is parallel to the flat surfaces of the platelets as can be seen from the layered structure10 of the particles in Figure 3 B. A detailed description of the difference diffraction pattern analysis is given elsewhere.5 Particle Thickness Distribution The mean crystal size can be estimated from line broadening of X-ray Bragg reflections. From the peak shapes it is in general possible to estimate a particle-size distribution. It is, however, not possible to obtain reliable information about particle-size distribution functions from the analysis of peak broadenings in single diffraction patterns without making assumptions justified by other methods because small experimental errors lead to large uncertainties in the distribution curve.4 Considerably more reliable information is obtained when the nanosuspensions are examined during the melting process. The standard11 and modified7 Warren-Averbach analysis using the (001) and (003) reflections of trimyristin indicate that the melting temperature of the particles decreases with decreasing average particle thickness (along the crystallographic c-axis). To get information on the average thickness of the particles which are melting within a narrow temperature range [T1,T2] the diffraction patterns collected at T2 were subtracted from those collected at T1. The analysis of such difference diffraction patterns in subsequent temperature ranges offers important
Unruh et al. advantages over that of individual diffraction patterns. The main advantage is that difficulties with background determination12 are largely eliminated as background intensities and their fluctuations are largely canceled by the subtraction, especially when the temperature intervals are small. In most cases the remaining low background can be fitted by a linear function. Moreover, the thickness distribution function can be derived directly rather than through a cumulative sum and the fact that each difference diffraction pattern corresponds to a narrow distribution of thicknesses of the particles further simplifies the analysis. Deviations of the integral mean value of the Bragg angle of the difference diffraction peak from the equivalent peak in the diffraction pattern of the bulk material give the overall change in spacings of the corresponding reflecting planes. Thus displacements caused by a particle size dependence of the lattice constants can be recognized and corrected. The standard analysis methods lead to serious problems in the case of triglycerides where spacing changes up to 5% are observed. Results and Discussion The particle thickness distribution determined by X-ray diffraction on the same sample as investigated by DSC (Figure 1) and TEM (Figure 3) shown in Figure 4 clearly reveals classes of particles with distinct thicknesses melting in separate temperature ranges. The absolute values of the particle thickness are in good agreement with those found by semi-quantitative analysis of cryo-transmission electron micrographs (cf. Figure 3 A). The distribution functions obtained by statistical analysis of cryo-micrographs differ significantly from those estimated by X-ray diffraction, mainly because there is a bias toward side views of small particles caused by the surface tension of the thin liquid films before their freeze. The analysis of X-ray difference diffraction patterns is thus preferable for the determination of particle-thickness distributions.13 The results of the particle-thickness analysis are also superimposed on the DSC curve in Figure 1. Despite the lowtemperature resolution, it is obvious that successive intense peaks at increasing temperatures in the DSC thermogram can be related to the melting of classes of crystals with thickness increments of only one unit cell. Besides the correlation of the particle-thickness distribution with the intense DSC peaks, some additional small or split peaks can be observed in the DSC curve. Further investigations with improved temperature resolution may clarify if this “fine structure” is related to a particle-size effect too. A first approach, however, indicates that some other effects (possibly recrystallization) influence the DSC curves as well as the difference diffraction patterns.5 The smearing of the DSC peaks at low temperatures should be expected because the thin particles are also small perpendicular to the (001) reciprocal lattice vector so that the influence of the size distribution in these directions is superimposed on the melting effect described above. The broadening of the (21h4)- and (302)-reflections of suspensions with decreasing mean particle sizes was indeed observed (cf. Figure 5). Because of the platelet-like shape of the particles, their size perpendicular to the (001)-vector increases rapidly with increasing mean particle size and the influence of this dimension on the melting temperature of the particles should rapidly vanish. No thermodynamic model describing the correlation between particle size and melting temperature is yet available. Thomson’s equation14 gives a crude approximation at best. It was observed that the (001)-interplanar distance of trimyristin nanosuspensions increases with decreasing particle thickness (expressed by a decrease in d-values of the difference (001)-reflections with
Size-Dependent Melting in Lipid Nanoparticles
J. Phys. Chem. B, Vol. 103, No. 47, 1999 10375
Figure 2. Analysis of the (001)-Bragg reflection of a trimyristin nanosuspension (2θ: scattering angle). T1 - T2 designates the difference diffraction peak calculated by subtraction of the diffraction pattern measured at T2 °C from the pattern measured at T1 °C. (A) Data recorded at 42 °C (O), 46 °C (×), 48 °C (+), and 52 °C (0). The solid line represents the corresponding pattern of the standard sample (see text). (B) Desmeared patterns of the standard sample (solid line) and of the suspension at 48 °C (+). (C) Difference diffraction peak 48-49 (O). The solid line represents the least-squares fit. (D) Fits of the (001)-reflection of the standard sample (dashed line) and of the difference diffraction peak 48-49 (dotted line). The solid line represents the deconvoluted profile. X-scale: h′3 ) sin θ/[2(sin θ2 - sin θ0)] (θ0: peak maximum; θ2: right-hand border of the peak). (E) Corrected Fourier coefficients Acn of deconvoluted difference diffraction peaks vs L (n: harmonic number). L ) na′3 is a real distance along the columns of (fictitious) unit cells with a cell constant a′3 perpendicular to the reflecting planes.11 The dashed lines represent the tangents at the points with the most negative slope. For normalization the Fourier coefficients were divided by the coefficient value obtained from extrapolation of the corresponding tangents to L ) 0 nm. The particle thickness d was estimated from the intersection of a tangent with the abscissa.
temperature). This implies, that missing neighbors of the crystalline arrangement lead to weaker van der Waals interactions and thus to a reduced lattice Gibbs free energy, which could explain the drop in melting temperature with decreasing
particle size. Classical surface thermodynamics seems not to be applicable because the crystals under investigation are too thin to possess a bulk phase. The influence of the emulsifiers on the observed effect is, however, still unknown and must be
10376 J. Phys. Chem. B, Vol. 103, No. 47, 1999
Unruh et al.
Figure 3. (A) Cryo-transmission electron micrograph of a trimyristin suspension. Both top views (1) and side views (2) of the crystals can be seen. The side view (parallel to the (001) reciprocal lattice vector) directly gives the particle thickness. Decreasing amounts of particles with interplanar distances of 1-5 times the spacing of the (001)-reflection of polycrystalline trimyristin were found. (B) Transmission electron micrograph of a freeze-fractured trimyristin suspension. The molecular layers perpendicular to the platelet plane can be seen (crystal inside the white circle).
Figure 4. Distribution of particle thicknesses d perpendicular to the (001) reciprocal lattice vector of the crystals of a trimyristin nanosuspension determined by Fourier analysis of difference X-ray diffraction patterns recorded during the melting of the particles by raising the temperature in steps of 1 °C. The ordinate represents the integral intensity of the difference diffraction peaks. The width of the bar charts corresponds to the minimum and maximum values of the particle thicknesses in the specified temperature range. The top axis gives the particle thickness in units of the (001)-interplanar spacing of polycrystalline trimyristin.
clarified in future studies in order to gain fundamental information on the melting behavior of small particles in suspension. Conclusions The experiments above clearly demonstrate that the discrete melting events of triglyceride nanoparticles are due to a particlesize effect. The thickness distribution of platelet-shaped particles can be analyzed in terms of a set of distinct quasi-monomodal distributions using difference X-ray diffraction patterns. The dependence of both the melting temperature and the interplanar spacing on the total number of molecular layers of the nanocrystals can be measured. More complete information on particle sizes and shapes could be obtained by extending the method to the size analysis of three-dimensional particles by simultaneously processing reflections of other crystallographic planes.
Figure 5. (21h4)-reflection of a trilaurin suspension (PCS z-average: 123 nm) at 25 °C (O) and 40.5 °C (0) recorded using synchrotron radiation. The data at 40.5 °C were multiplied by a factor of 1.5 in order to get comparable maximum intensities of the two peaks. A constant background was subtracted. The peak recorded at 40.5 °C (lower mean particle thickness) is significantly sharper than the one recorded at 25 °C indicating a particle-size effect also in the direction parallel to the (21h4) reciprocal lattice vector. However, the observed peak broadening could also be due to some other effects, e.g., increasing close-range order with increasing temperature, and should be investigated in further studies.
The method and results described above are not only of interest from a fundamental point of view. The particle-sizedependent melting of solid lipid nanoparticles may offer interesting new possibilities for targeting of drugs to specific body sites: Melted particles will release their drug load much faster than solid particles due to the higher diffusion coefficient of the drug molecules inside the liquid carrier. When solid triglyceride particles circulating in the bloodstream after intravenous administration reach a site where the temperature is higher than the melting temperature of the particle matrixs e.g., due to an external warming of that sitesthey will suddenly melt and rapidly release their drug load. In this way, a much higher drug concentration is obtained in the heated organ or tissue compared to other sites, where slow release from the solid carriers will lead to much lower concentrations. Such drug delivery systems require carrier particles with a narrow meltingtemperature range around approximately 39-40 °C, which may be obtained by particle-size fractionation (e.g., by ultracentrifu-
Size-Dependent Melting in Lipid Nanoparticles gation) of triglyceride suspensions which exhibit the uncommon, stepwise melting behavior. Acknowledgment. We gratefully acknowledge M. Drechsler for taking and processing the electron micrographs. References and Notes (1) Westesen, K.; Siekmann, B. Microencapsulation; Benita, S., Ed.; Marcel Dekker: New York, 1996; p 213. (2) Bunjes, H. Thesis, Jena, 1998. (3) Koch, M. H. J.; Bordas, J. Nucl. Instrum. Methods 1993, 208, 461. (4) Unruh, T.; Bunjes, H.; Westesen, K.; Koch, M. H. J. Annual Report, EMBL Hamburg Outstation, 1998, p 391. (5) Unruh, T.; Bunjes, H.; Westesen, K.; Koch, M. H. J. Prog. Colloid Polym. Sci., submitted. (6) Singh, M. A.; Ghosh, S. S.; Shannon, R. F., Jr. J. Appl. Crystallogr. 1993, 26, 787.
J. Phys. Chem. B, Vol. 103, No. 47, 1999 10377 (7) Delhez, R.; de Keijser, Th. H.; Nittemeijer, E. J. National Bureau of Standards Publication 567, Proceedings of Symposium on Accuracy in Powder Diffraction held at NBS, Gaithersburg, MD, June 11-15, 1979 (Issued February 1980), pp 213-253. (8) Press: W. H.; Teukolsky, S. A.; Vetterling, W. T.; Flannery, B. P. Numerical Recipes in C; Cambridge University Press: New York, 1995. (9) Stokes, A. R. Proc. Phys. Soc. London 1948, 61, 382. (10) The thickness of the layers yields approximately 3.6 nm which corresponds to the (001)-spacing of 3.59 nm (Frede, E.; Precht, D. Fette Seifen Anstrichmittel 1977, 2, 69). (11) Warren, B. E. X-ray diffraction; Addison-Wesley, Reading, MA, 1969. (12) Delhez, R.; de Keijser, Th. H.; Nittemeijer, E. J. Fresenius J. Anal. Chem. 1982, 312, 1. (13) To get information on the particle dimensions in directions other than perpendicular to the (001)-plane, corresponding reflections have to be investigated using the same procedure. (14) Defay, R.; Prigogine, I.; Bellemans, A.; Everett, D. H. Surface Tension and Adsorption; Longmans Green: London, 1966.