Physicochemical Characterization of a Model Intravenous Oil-in-Water Emulsion KIRSTENWESTESEN" AND THOMASWEHLER' Received Ma 22, 1991, from the 'Institute of Pharmaceutical Technology, MendelssohnstraBe 1, 3300 Braunschweig, Germany, and the *Kabi Pharmacia Ab, Cardiovascular Section for Structure Analysis, Lindhagensgatan 133, 1 12 87 Stockholm, Sweden. Accepted for publication September 6, 1991.
Abstract 0 Photon correlation spectroscopy, nuclear magnetic resonance spectroscopy, electron microscopy, and small-angle X-ray scattering were used for the structuralcharacterizationof a modeloil-in-water emulsion containing particles in the submicrometer size range. Additionally, small-angle X-ray diffraction, wide-angle X-ray diff raction, and differential scanning calorimetry were applied to raw materials and to binarymixtures. The majority of emulsion dropletshave the characteristic of an ideal emulsion roplet, that is, a liquid lipid core covered by an emulsifier monolayer. owever, the system contains a certain excess of emulsifier. Pacicles with bi- and/or oligolayer structures can be deduced. Double-emulsion droplets were detected. Large unilamellar vesicles were not found; however, the existence of small unilamellar liposomes (also referredto as small unilamellar vesicles or SUVs) seems likely. The proportion of all small nonmonolayer structureswas quantifiedby nuclear magnetic resonance spectroscopy. No mixed micellar structures were detectable. Lysophospholipids were not detected in the aqueous phase, indicating their predominant incorporation into the emulsifier layers. Water-soluble phospholipid degradation products were found in the water phase. The existence of at least several monolayers of phospholipids does not seem to be a prerequisitefor a stable soybean oil-in-water emulsion, in general.
R
In the past, oil-in-water (o/w) emulsions have often been considered as two-phase systems consisting of a fluid oil phase dispersed in a continuous water phase, with the interface being stabilized by a monomolecular layer of surfactant molecules. The present International Union of Pure and Applied Chemistry definition of an emulsion is broader, because it includes the possibility of several phases in an emulsion, including liquid-crystalline phases: "In an emulsion liquid droplets and/or liquid crystals are dispersed in a liquid".' The change in this definition of an emulsion was largely due to the detection of (lamellar) liquid-crystalline phases in a number of systems.2-5 The phase equilibria of ternary surfactantroil-water systems containing phospholipids of different composition as emulsifiers and soybean oil as the dispersed oil phase were studied and related to the stability of o/w emulsions.6.7 The determination of the minimum amount of emulsifier that is needed for the preparation of a stable emulsion indicated that at least several monolayers of the emulsifier are required.7 The structural investigations cited concentrated on o/w emulsions with an excess of emulsifier and mean droplet sizes in the micrometer range. However, the inner structure of conventional emulsions intended for intravenous administration (iv emulsions) has not been studied in much detail. Description of the systems has focused mainly on particle size distributions and electrokinetic properties. The reason the inner structures of the phospholipid-containing particles has not been investigated in detail probably is that many techniques are not applicable in the submicrometer range. Phospholipid-stabilized iv emulsions have been widely used for parenteral nutrition8 and have been studied and intro0022-3549/9fl08OO-0777$02.50/0 0 1992, American Pharmaceutical Association
duced as drug carrier systems, especially for lipophilic compounds.Cl1 The chemical composition of most of the commercially available iv emulsions is quite simple and similar for different preparations. These emulsions often contain soybean oil as the fat phase; fractionated phospholipids or lecithin as the emulsifier; nonelectrolytic compounds, such as glycerol or xylitol, to achieve blood isotony; and small quantities of electrolytes to achieve a suitable pH (prior to sterilization). Common products contain 10-208 (w/w)oil and 1.5%(w/w) or less emulsifier. For most of the commercial products, the amount of emulsifier is not varied in relation to the oil content. Commercial iv emulsions such as Intralipid 10:20%, Lipofundin MCT 10:20%, Lipohorm 10:20%, and Lipoven65 10:20%contain the same amount of phospholipids independent of the fat load of the system. In those products, the phospho1ipid:oil ratio varies significantly between preparations containing either 10 or 20%oil by weight. In contrast Lipofundin S 10:20%shows a constant phospho1ipid:oil ratio. Independent of the variability or constancy of the phospho1ipid:oilratio in the various iv emulsions, an extensive excess of emulsifier does not seem to be probable because of the high degree of dispersity of the oil phase. This aspect is important in comparing these preparations with the cruder systems discussed previously. An influence of emulsion properties on drug distribution and elimination rates from plasma in animals has been described inter alia after iv administration of phospholipidstabilized emulsions containing coenzyme QI0.1L14 The authors discussed influences on the uptake of the drug by the heart and rates of elimination from plasma of the emulsifier chemistry and total amount of emulsifier in systems with fixed fat load. No complete physicochemical characterization of the application systems was performed in parallel. However, physicochemical characterizations, especially of the inner structure of these systems, may explain results from animal experiments and aid the development of the convenient drug carrier and targeting systems. Therefore, the aims of this study were to find a useful combination of independent methods for characterizing iv emulsions in general and to determine the inner structure of a model iv emulsion stabilized by phospholipids. The structural parameters considered in detail in this publication are related to the internal structure of the particles. An emulsion with an extremely high fat load and a classical emulsifier was chosen as a model iv system. The analysis of the particle size distribution will be reported separately.
Experimental Section Materials-Emulsion system 1 contained fractionated soybean oil (250.0 mg), fractionated egg phospholipids (18.0mg), glycerol USP (22.4 mg), and water for injection (1.0 mL). Emulsion system 2 contained fractionated soybean oil (250.0 mg), sodium caseinate (30.0 mg), glycerol USP (22.4 mg), and water for injection (1.0 mL).In both systems, 1 N NaOH was used to adjust pH. Journal of Pharmaceutical Sciences I 777 Vol. 81, No. 8, August 1992
The following raw materials and chemicals were used: soybean lysophosphatidylcholine (LPC)L 21005 (Lipoid KG),soybean phosphatidylcholine (PC)NC 95 (Nattermann), fractionated egg phospholipids, sodium glycocholate G-7132/88F-5023, (Sigma), L-aglycerophosphorylethanolamine (GPE); (lot no. 87F8385; Sigma), L-a-glycerophosphorylcholine(GPC; lot no. 89F8446; Sigma), praseodymium(1II) nitrate pentahydrate (Janseen Chimica),trimethylphosphate (Merck), deuterated water (99.988, SStudsvik Energiteknik AB), methanol (analytical grade; Merck), chloroform (analytical grade; Merck), chloroformd, (Merck), and dialysis membranes (standard membranes part no 105-10; Technicon Ltd.). M e t h o d e P r e p r a t i o n of Multilamellar Liposomes-Phoepholipids (18 mg/mL) in dust-free water were swirled in a vortex mixer (Ultra Turrax TP 18/10) for 10 min at 20 000 rpm. During swirling, the system was cooled in an ice-water bath. Preparation of Small Unihmellar L i p o s o m e 4 m a l l unilamellar lipsomes (also referred to as small unilamellar vesicles or SUVs) were prepared as follows. Fractionated egg phospholipids (18 mg/mL) in dust-free water were sonicated (Soniprep 150) for -30 min. An average particle size of -40-60 nm was desired. Sonication efficiency was controlled by dynamic light scattering measurements. Preparation of Large Unilamellar Vesicles (L Ws)-Egg phospholipids (18 mg/mL; Ovothin 170, Lucas Meyer) were dispersed in dust-free water with a magnetic stirrer. The resulting largemultilamellar-vesicle dispersion was subjected to a f r e e s t h a w protocol consisting of five cycles of freezing in liquid nitrogen for 30 s and thawing in a water bath at 40°C. The frozen and thawed multilamellar vesicle dispersions were rapidly passed through two stacked 200-nm-pore-size polycarbonate filters (Nuclepore) with an extruder (Lipex Biomembranes Inc.). The extrusion was carried out under a nitrogen pressure of 175 lb/in., and was repeated five times. Prepamtion of Mixed Micellar Systems-For the LPC mixed micellar solution, LPC (3.0 mg/mL) and sodium glycocholate (12.0 mg/mL) were dissolved in methanol (-3 mL), and the solvent was subsequently evaporated on a rotary evaporator. The semisolid concentrate was dissolved in methanol again and dried on the rotary evaporator. This procedure was performed three times. Finally, the dry residue was dissolved in D20(6.6 mL) and allowed to equilibrate for -3 h. The same method was applied to the PC mixed micellar solution, but with PC (3.0 mg/mL) and sodium glycocholate (12.1 mg/mL), and to the mixed micellar solution with fractionated egg phospholipids, but with fractionated egg phospholipids (3.1 mg/mL) and sodium glycocholate (12.4 mg/mL). A highly concentrated mixed micellar solution was also prepared containing fractionated egg phospholipids (157.4 mg/mL) and sodium glycocholate (89.5 mg/mL). Dialysis-The emulsion was dialyzed for 16 h at 40°C against bidistilled water containing glycerol (22.4 mg/mL). The equilibrium dialysis was performed with equal volumes (0.8 mL) of donor and acceptor media. Dynamic Light Scattering4amples for dynamic light scattering (or PCS, photon correlation spectroscopy) experiments were prepared or diluted with dust-free water. Studies were performed on a Malvern Autosizer IIc (Malvern Instruments) at a fixed angle of 90". The instrument was run in the "automode" as well as in "manual mode", at various sample times, experimental durations, distribution widths, estimated sizes, and other parameters. Number distributions were used for data interpretation. Transmission Electron Microscopy (TEM)-Freeze fracturing and freeze etching of samples were performed with the Refreeze fracture unit (BAF 400; Balzers AG, Lichtenstein). Cryoprotectants were not wed. The samples were sandwiched between two flat golden holders. Fast freezing was done by one of the following methods: slush method in melting nitrogen, slush method in liquid Freon (liquid fluorinated hydrocarbons), slush method in melting propane, or jet freezing with liquid propane. The fracturing temperature was 173 K. Specimens prepared by one of the freezing techniques were sometimes etched for 30 or 60 s (173 K, lO-'-lO-' mPa). The fractured (and etched) frozen samples were shadowed under 45" with platinum/ carbon (layer thickness, 2 nm) and under 90"with pure carbon for replica production. Replicas were cleaned with ch1oroform:ethanol (l:l,v/v) and picked up on uncoated microscope grids. Replicas were viewed with an electron microscope (EM 300, Philips, Kassel, Germany). For replica preparation, the original emulsions or diluted systems were used. Dilution was performed with bidistilled water containing 2.25%(w/v) glycerol USP. 778 / Journal of Pharmaceutical Sciences Vol. 81, No. 8, August 1992
Nuclear Magnetic Resonance (NMR) Spectroscopy-NMR samples were undiluted, except for paramagnetic-shifted samples, and kept in the native form. Trimethylphosphate [S 2.00 ppm, 6.3 mM, CDC1,:MeOH (2:1)] was used as an external reference in a Wilmad NMR reference tube (5 mm), which was placed inside a 10-mm NMR sample tube. In the paramagnetic-shifted samples, p&ymiuy(IID nitrate pentahyhte (50 mM D,O solution) was added in 1:5 pohons. 31PNMR spectra were obtained on a JEOL GX-400 spectrometer operating at a Larmor frequency of 161.70 MHz. The sample temperature was 25 "C. A 45" pulse (11 p )was applied, and 5000 free induction decays were acquired with quadrature detection and Fourier transformed with either 1-, 5-, or 10-Hz exponential line broadening. The spectral width was 60 lrHz (for the whole spectrum) or 8 lrHz (optimized), resolved into 32 000 data points. The pulse repetition time was 32 s. A proton decoupling with a power of 9 W was gated on only during acquisition of free induction decays. The integral was obtained aRer phase a4ustment and baseline correction. The relaxation time was measured by the inversion recovery method, and a value of 11 s for the trimethylphosphate signal (not deaerated) was calculated with a three-parameter nonlinear fit. Small-Angle X-my Scattering and Small-Angle X-ray Diffmction-Small-angle X-ray scattering (SAXS) measurements and small-angle X-ray diffraction studies were performed in a compact small-angle system (Kratky KCLC) connected to a Philips PW 2253111 X-ray tube with copper anode. A nickel foil served as a KB-filter. The following operating conditions were chosen: a tube voltage of 40 to 50 kV and an anode current of 35 to 40 mA. An OED 50 position sensitive detector (Braun GmbH, Munich, Germany) was used,with argon:methane (9:l) as filling gas. The linked multichannel analyzer was a Canberra MCA 8100. Liquid samples were measured in a capillary holder (K-PR) with Peltier temperature control device (Anton Paar, Graz, Austria). Semisolid samples were measured in a homemade holder placed between two Mylar (a kind of polyethylene terephthalate) foils. Exposure times of 3000 or 5000 s were chosen for SAXS measurements. Exposure times of 300-400 s were suflicient for diffraction studies. Scattering curves were analyzed with the computer program collection of 0. Glatter called Fortran IV-Programmpaket ITP-81. Wide-Angle X-ray Dimtion-Studies were performed in a Philips PW 1060/25 X-ray goniometer, equipped with a stepmotor, a Xeproportional Geiger counter (PW 1711/10), and a PW 1710 controller. X-ray tube, anode, and filter material, as well as operating conditions, were the same as for SAXS studies. Samples were measured in aluminum holders. Differential Scanning Calorimetry (DSC)-A Perkin-Elmer DSC 2 4 calorimeter with thermal analysis data station was used for calorimetric measurements. Samples (-6-9 mg) were weighed into Perkin-Elmer, standard, volatile-sample aluminum pans. An empty pan was used as reference. A scan rate of 2.5 or 5 Wmin was chosen. DSC samples were prepared at mom temperature. Each sample was run three times. The samples were heated from 250 to 400 K at a .wan rate of 5 K/min (first run) and held for 5 min at 400 K. Then, the samples were cooled from 400 to 275 K at a scan rate of 5 Wmin (second run) and held for 60 min at 275 K. ARer that, the samples were heated to 400 K again at a scan rate of 5 Wmin (third run).
Results TEM Studies-Undiluted Samples Prepared with Melting Nitrogen (Slush Technique)-The TEM micrographs indicate the original arrangement of particles in the external phase of the emulsion (Figure 1). The particles seem to be randomly distributed in the fracture plan, and their tight packing is obvious. The majority of particles have structured cores. A layered structure is typical for lamellar liquid-crystalline or crystalline materials. Obviously, the oil phase crystallized during freezing of the sample in melting nitrogen. The surrounding water phase, however, appears amorphous. Faster freezing resulted in structure-free, amorphous fractures (Figure 2). For structural analysis, the crystallization artifact of the oil phase (Figure 1)was a n advantage, because samples prepared with melting nitrogen could be used to differentiate between oil-filled, structured fractures and amorphous, water-filled
Flgure 1-TEM micrograph of a replica of the native model iv emulsion showing the tight packing of particles and crystallization artifacts of the oil phase ( x 1 1 700).
Flgure S T E M micrograph of a phospholipid-freesoybean oil emulsion stabilized with casein “micelles”(x49 140).All particles look “cut”and amorphous. No layer structures are detectable. The surface lines are clearly visible.
globules. The micrographs show that particles >lo0 nm are normally oil filled. Larger numbers of LUVs were not detected in the model emulsion. The replicas showed a few particle fractures containing holes or cavities. All particles of this type were >lo0 nm in diameter. The holes in fatty cores of o/w emulsions could be freezing artifacts caused, for example, by differences in density changes of water and oil during the freezing process. Therefore, the surface characteristics of these cavities were studied by water condensation on the fractures. Water condenses predominately on hydrophilic surfaces. On micrographs showing water contamination of the specimen, particles containing holes filled with water drops are detectable (Figure 3). Obviously, the holes have hydrophilic surfaces that result from phospholipids covering and stabilizing the incorporated water. The particles are typical double-emulsion droplets. They consist of small water droplets dispersed in the lipid core of an olw emulsion drop. The water droplets are stabilized by a surfactant layer. However, only a few of these particle-in-particle structures exist in the model iv emulsion. Freon Slush of a Phospholipid-Free Test System-The Freon slush method was used to prepare a phospholipid-free test system. The test system contained casein “micelles”(3w t %) instead of fractionated egg phospholipids as a surface
Flgure S T E M micrograph of a replica of a fractured emulsion droplet containing a highly water contaminated spherical area (x23 400).
stabilizer of a soybean oil emulsion containing -25% (w/w) oil. Some samples show particles with amorphous oil cores (Figure 2). The samples demonstrate that the soybean oil exists in its liquid state at room temperature and that the samples prepared in melting nitrogen contain crystallization artifacts. Although the freezing artifacts of the oil phase were helpful for the distinction between oil- and water-filled structures, crystallization of fat becomes a problem if phospholipid layers are to be detected and quantified. Crystallized fat can occur in layered structures similar to liquid-crystalline lamellar phospholipid layers16 and can mimic phospholipid oligo- or multilayers. The micrograph of the casein-stabilized emulsion shows an area of the specimen containing completely “layer-free” particle fractures, and all structures look “cut” (Figure 2). The surface line of the particles is clearly visible. Therefore, all layered surface structures in an artifact-free sample of the model iv emulsion represent phospholipid layers saturated with compounds from the oil phase. Diluted Samples Prepared with Melting Propane (Slush Technique)-Micrographs of undiluted samples show the extremely tight packing of particles in the model iv emulsion (Figure 1).Because smaller particles are hidden in highly concentrated systems, the emulsion was diluted (l:l), and fast freezing was done by the slush method in melting propane (Figure 4). Freezing in melting propane was fast enough to yield clear cuts through the particles in the fracture plan. The cores of the fractures look nearly amorphous; thus no further distinction between oil-filled and water-containing globules was possible. The majority of fractures show no typical layered structures on the surface (Figure 4). No (dispersed) lamellar phase was found. Multilayer structures, as often discussed for olw emulsions with an excess of emulsifier, could not be detected in the model iv emulsion. Only a few bi- and oligolayer structures were detectable. Most of the fractured oil droplets seem to be surrounded by emulsifier monolayers. The majority of emulsion droplets show the typical characteristics of a n ideal emulsion droplet, that is, they have a liquid lipid core covered by a n emulsifier monolayer. Large numbers of fractures with diameters of C60 nm are visible in the diluted sample. Dispersions of small unilamellar vesicles (SUVs) can be produced by the same techniques used to produce phospholipid-stabilized iv emulsions. However, the micrographs do not allow a clear distinction between SUVs and small emulsion droplets. The existence of small unilamellar liposomes cannot be ruled out by TEM studies nor Journal of Pharmaceutical Sciences I 779 Vol. 81, No. 8, August 1992
30
-
0
0
100
200
300
500
400
700
800
800
r s L chnnnslnr.
Figure &TEM micrograph of fractured, diluted (1:l) model emulsion
showing amorphous oil cores after fast freezing (x31 590). Large
0.0035
numbers of small particles are visible.
*
verified, because SUVs and small emulsion droplets (80-100 nm in diameter are normally oil-filled droplets. SAXS was used to get additional information about a possible coexistence of bilayer arrangements, such as SUVs or L W s , and ideal emulsion droplets in the model iv emulsion. SUVs produce no small-angle X-ray diffraction, but diffuse smallangle X-ray scattering.1618 The scattering curve of an aqueous dispersion of small unilamellar liposomes with a broad and continuous size distribution is smooth and reflects the scattering in the single bilayers.16-ls Typical smooth scattering curves were gained from SUV dispersions with 0.5-6.0% (wlw) phospholipids and a broad size distribution, as determined by PCS measurements (Figure 5, Top). Information about the bilayer thickness was obtained with the flat-sheet approximation for computeradded data analysis. The thickness was obtained from the correlation function (y(J along a line perpendicular to the bilayer plane. The point where finally approaches zero correspondsto the bilayer thickness, which was determined to be 4.5 nm (Figure 5, Bottom). The undiluted emulsion (Figure 6a) and the diluted samples showed no comparable scattering behaviors. Samples of the model emulsion were mixed with a dispersion of SWs. The addition of a vesicle dispersion to the emulsion caused a detectable diffuse scattering typical of scattering in single bilayers if the system contained 0 . 5 4 7 5 % phospholipids bound to vesicle structures. A typical mixture contained 1.05% (w/w) phospholipids originating from the emulsion and -0.75% (w/w) originating from the vesicle dispersion (Figure 6b). The scattering curve of this mixture had a shape comparable with the shape of the curve of the pure vesicle dispersion containing 0.75% (w/w)phospholipids. Larger amounts of single bilayer structures, such as SWs, can be detected directly by SAXS measurements, even in mixtures with ideal submicrometer-sized emulsion droplets. In contrast to the NMR spectroscopic studies, the amount of bilayer arrangements with diameters >80-100 nm was detectable in SAXS experiments, as demonstrated by SAXS measurements on a dispersion of LUVs. If the model iv emulsion contains SUVs or larger bilayer arrangements, the concentration of all these structures in the native emulsion should be below or close to the detection limits of the SAXS measurements. Therefore, less than -28% (w/w)of the total amount of emulsifier is involved in the formation of single bilayer structures, such as vesicular structures.
-
780 I Journal of Pharmaceutical Sciences Vol. 87, No. 8, August 1992
0.0025
0.0015 0.0005 0
--0.0°15 0.0025
~
\
I\
1\
............ .......
.................
I--,
0
10
20
30
.
40
,
.
50
.
,
60
R [A1
Figure W o p ) Scatteringcurve and (Bottom) correlationfunction, y(r), of a SUV dispersion containing 1% (w/w) phospholipids (LaGrange multiplication, 7.00; 20 splines).
NMR S p e c tro s ~ o p y - ~ ~NMR P spectroscopy was used to characterize the conformation and motional properties of phospholipids in various model systems.19 31P NMR line shapes are, inter alia, not only dependent on the sizes of phospholipid particles20 but also affected by a number of other parameters, such as temperature, magnetic field strength, viscosity, aqueous phase pH, dispersion state, and methods of emul~ification.~~-2s Small phospholipid-containing structures, such as micelles and sonicated vesicles, show relatively narrow, symmetrical 'lP NMR lines, reflecting isotropic motional averaging and arising mostly from lateral diffusion and Brownian tumbling. Unsonicated phospholipid bilayers, on the other hand, exhibit broad, unsymmetrical lines (powder-type lines), which are caused by restricted anisotropic motion of the phospholipid molecules. In the 'lP NMFt spectrum of the model iv emulsion, three types of line shapes are seen (Figure 7a): narrow signals with a half-line width (Avl12) at 2 Hz and a chemical shift (6) at 6 0.24 and S -0.33 ppm; broader (Avl12 = 83 Hz),but still symmetrical, signals a t 6 -0.07, 6 -0.62, and 6 -1.06 ppm (Figure 7c); and a n underlying asymmetrical signal with a chemical shift anisotropy (Am) of 32 ppm (where A u = 9 - q, and 9 and q,are components of u that are, respectively, parallel and perpendicular to the molecular symmetry axis; cq = 23 ppm; 7,= -9 ppm; Figure 7b). Possible other underlying signals with various line shapes are not resolved. In 31P NMR spectra of one system (i.e., emulsions), the parameters that should influence line shapes are particle size and particle structure (all other parameters are constant). Different line shapes in the emulsion spectrum indicate a nonhomogeneity in particle structures and sizes. For correct
-
O 80 O\
6oT
.
401
I
0
E g
90 nm in diameter. This size distribution delivered about the same line shapes and chemical shifts as the broad symmetrical signals in the emulsion spectrum (Figure 8a). A mixture of small unilamellar liposome dispersion and emulsion (9:l) had liposome signals that overlapped with the broad symmetrical signals in the emulsion spectrum. This result supports the TEM results that indicate larger amounts of particles in the system with diameters below -100-120 nm. In contrast, preparations containing either large multilamellar structures or large unilamellar liposomes gave line shapes similar to those of the asymmetrical emulsion signal (Figure 9). Taking TEM and SAXS results into account, we concluded that the broad, asymmetrical signal of the emulsion does not result from a dispersed lamellar phase but arises predominately from phospholipid molecules stabilizing large, ideal emulsion particles. Mixed micellar systems of different compositions were also prepared. The observed line width for the micelles varied from -5 to 16 Hz, depending on the phospholipid used for the
2
1
0
I""I -1
-2
"'""'I"' -3 -4
I
I
I T
-5
PPm Figure 7-31P NMR spectrum of model iv emulsion (a) with 1BkHz spectral width and 10-Hz line broadening, (b) as in a, but enhanced, and (c) with 1.5-kHz spectral width and 1-Hzline broadening.
preparation. LPC-containing mixed micelles had line widths of 5 Hz (Figure 10a) and PC-containing mixed micelles had line widths of 16 Hz (Figure lob), reflecting the difference in mobility and size of the micelles. A narrowing of the line widths can be seen in the spectra of mixed micelles containing fractionated egg phospholipids (Figure 1Oc). Drastic differences in total micellar concentration had only a minor influence on the line width and the chemical shift. To elucidate if lysophospholipids are dissolved in the aqueous phase or incorporated into the emulsifier layers, the emulsion was dialyzed, and the dialysate was examined with 'lP NMR spectroscopy. Signals with much smaller half-line widths ( A V , , = ~ 1Hz)are observed, and chemical shifts are not in accord with the observed micellar chemical shifts (i.e., lysophospholipids are predominantly incorporated into the emulsifier layers). No micelles or mixed micelles formed by emulsifier and/or degradation products could be detected in the 'lP NMR spectra of the model iv emulsion. The asymmetrical and the broad, symmetrical signals of the emulsions are clearly broader than the observed signals for the micelles ( A Y ~ /
Abstract 0 Photon correlation spectroscopy, nuclear magnetic resonance spectroscopy, electron microscopy, and small-angle X-ray scattering were used for the structuralcharacterizationof a modeloil-in-water emulsion containing particles in the submicrometer size range. Additionally, small-angle X-ray diffraction, wide-angle X-ray diff raction, and differential scanning calorimetry were applied to raw materials and to binarymixtures. The majority of emulsion dropletshave the characteristic of an ideal emulsion roplet, that is, a liquid lipid core covered by an emulsifier monolayer. owever, the system contains a certain excess of emulsifier. Pacicles with bi- and/or oligolayer structures can be deduced. Double-emulsion droplets were detected. Large unilamellar vesicles were not found; however, the existence of small unilamellar liposomes (also referredto as small unilamellar vesicles or SUVs) seems likely. The proportion of all small nonmonolayer structureswas quantifiedby nuclear magnetic resonance spectroscopy. No mixed micellar structures were detectable. Lysophospholipids were not detected in the aqueous phase, indicating their predominant incorporation into the emulsifier layers. Water-soluble phospholipid degradation products were found in the water phase. The existence of at least several monolayers of phospholipids does not seem to be a prerequisitefor a stable soybean oil-in-water emulsion, in general.
R
In the past, oil-in-water (o/w) emulsions have often been considered as two-phase systems consisting of a fluid oil phase dispersed in a continuous water phase, with the interface being stabilized by a monomolecular layer of surfactant molecules. The present International Union of Pure and Applied Chemistry definition of an emulsion is broader, because it includes the possibility of several phases in an emulsion, including liquid-crystalline phases: "In an emulsion liquid droplets and/or liquid crystals are dispersed in a liquid".' The change in this definition of an emulsion was largely due to the detection of (lamellar) liquid-crystalline phases in a number of systems.2-5 The phase equilibria of ternary surfactantroil-water systems containing phospholipids of different composition as emulsifiers and soybean oil as the dispersed oil phase were studied and related to the stability of o/w emulsions.6.7 The determination of the minimum amount of emulsifier that is needed for the preparation of a stable emulsion indicated that at least several monolayers of the emulsifier are required.7 The structural investigations cited concentrated on o/w emulsions with an excess of emulsifier and mean droplet sizes in the micrometer range. However, the inner structure of conventional emulsions intended for intravenous administration (iv emulsions) has not been studied in much detail. Description of the systems has focused mainly on particle size distributions and electrokinetic properties. The reason the inner structures of the phospholipid-containing particles has not been investigated in detail probably is that many techniques are not applicable in the submicrometer range. Phospholipid-stabilized iv emulsions have been widely used for parenteral nutrition8 and have been studied and intro0022-3549/9fl08OO-0777$02.50/0 0 1992, American Pharmaceutical Association
duced as drug carrier systems, especially for lipophilic compounds.Cl1 The chemical composition of most of the commercially available iv emulsions is quite simple and similar for different preparations. These emulsions often contain soybean oil as the fat phase; fractionated phospholipids or lecithin as the emulsifier; nonelectrolytic compounds, such as glycerol or xylitol, to achieve blood isotony; and small quantities of electrolytes to achieve a suitable pH (prior to sterilization). Common products contain 10-208 (w/w)oil and 1.5%(w/w) or less emulsifier. For most of the commercial products, the amount of emulsifier is not varied in relation to the oil content. Commercial iv emulsions such as Intralipid 10:20%, Lipofundin MCT 10:20%, Lipohorm 10:20%, and Lipoven65 10:20%contain the same amount of phospholipids independent of the fat load of the system. In those products, the phospho1ipid:oil ratio varies significantly between preparations containing either 10 or 20%oil by weight. In contrast Lipofundin S 10:20%shows a constant phospho1ipid:oil ratio. Independent of the variability or constancy of the phospho1ipid:oilratio in the various iv emulsions, an extensive excess of emulsifier does not seem to be probable because of the high degree of dispersity of the oil phase. This aspect is important in comparing these preparations with the cruder systems discussed previously. An influence of emulsion properties on drug distribution and elimination rates from plasma in animals has been described inter alia after iv administration of phospholipidstabilized emulsions containing coenzyme QI0.1L14 The authors discussed influences on the uptake of the drug by the heart and rates of elimination from plasma of the emulsifier chemistry and total amount of emulsifier in systems with fixed fat load. No complete physicochemical characterization of the application systems was performed in parallel. However, physicochemical characterizations, especially of the inner structure of these systems, may explain results from animal experiments and aid the development of the convenient drug carrier and targeting systems. Therefore, the aims of this study were to find a useful combination of independent methods for characterizing iv emulsions in general and to determine the inner structure of a model iv emulsion stabilized by phospholipids. The structural parameters considered in detail in this publication are related to the internal structure of the particles. An emulsion with an extremely high fat load and a classical emulsifier was chosen as a model iv system. The analysis of the particle size distribution will be reported separately.
Experimental Section Materials-Emulsion system 1 contained fractionated soybean oil (250.0 mg), fractionated egg phospholipids (18.0mg), glycerol USP (22.4 mg), and water for injection (1.0 mL). Emulsion system 2 contained fractionated soybean oil (250.0 mg), sodium caseinate (30.0 mg), glycerol USP (22.4 mg), and water for injection (1.0 mL).In both systems, 1 N NaOH was used to adjust pH. Journal of Pharmaceutical Sciences I 777 Vol. 81, No. 8, August 1992
The following raw materials and chemicals were used: soybean lysophosphatidylcholine (LPC)L 21005 (Lipoid KG),soybean phosphatidylcholine (PC)NC 95 (Nattermann), fractionated egg phospholipids, sodium glycocholate G-7132/88F-5023, (Sigma), L-aglycerophosphorylethanolamine (GPE); (lot no. 87F8385; Sigma), L-a-glycerophosphorylcholine(GPC; lot no. 89F8446; Sigma), praseodymium(1II) nitrate pentahydrate (Janseen Chimica),trimethylphosphate (Merck), deuterated water (99.988, SStudsvik Energiteknik AB), methanol (analytical grade; Merck), chloroform (analytical grade; Merck), chloroformd, (Merck), and dialysis membranes (standard membranes part no 105-10; Technicon Ltd.). M e t h o d e P r e p r a t i o n of Multilamellar Liposomes-Phoepholipids (18 mg/mL) in dust-free water were swirled in a vortex mixer (Ultra Turrax TP 18/10) for 10 min at 20 000 rpm. During swirling, the system was cooled in an ice-water bath. Preparation of Small Unihmellar L i p o s o m e 4 m a l l unilamellar lipsomes (also referred to as small unilamellar vesicles or SUVs) were prepared as follows. Fractionated egg phospholipids (18 mg/mL) in dust-free water were sonicated (Soniprep 150) for -30 min. An average particle size of -40-60 nm was desired. Sonication efficiency was controlled by dynamic light scattering measurements. Preparation of Large Unilamellar Vesicles (L Ws)-Egg phospholipids (18 mg/mL; Ovothin 170, Lucas Meyer) were dispersed in dust-free water with a magnetic stirrer. The resulting largemultilamellar-vesicle dispersion was subjected to a f r e e s t h a w protocol consisting of five cycles of freezing in liquid nitrogen for 30 s and thawing in a water bath at 40°C. The frozen and thawed multilamellar vesicle dispersions were rapidly passed through two stacked 200-nm-pore-size polycarbonate filters (Nuclepore) with an extruder (Lipex Biomembranes Inc.). The extrusion was carried out under a nitrogen pressure of 175 lb/in., and was repeated five times. Prepamtion of Mixed Micellar Systems-For the LPC mixed micellar solution, LPC (3.0 mg/mL) and sodium glycocholate (12.0 mg/mL) were dissolved in methanol (-3 mL), and the solvent was subsequently evaporated on a rotary evaporator. The semisolid concentrate was dissolved in methanol again and dried on the rotary evaporator. This procedure was performed three times. Finally, the dry residue was dissolved in D20(6.6 mL) and allowed to equilibrate for -3 h. The same method was applied to the PC mixed micellar solution, but with PC (3.0 mg/mL) and sodium glycocholate (12.1 mg/mL), and to the mixed micellar solution with fractionated egg phospholipids, but with fractionated egg phospholipids (3.1 mg/mL) and sodium glycocholate (12.4 mg/mL). A highly concentrated mixed micellar solution was also prepared containing fractionated egg phospholipids (157.4 mg/mL) and sodium glycocholate (89.5 mg/mL). Dialysis-The emulsion was dialyzed for 16 h at 40°C against bidistilled water containing glycerol (22.4 mg/mL). The equilibrium dialysis was performed with equal volumes (0.8 mL) of donor and acceptor media. Dynamic Light Scattering4amples for dynamic light scattering (or PCS, photon correlation spectroscopy) experiments were prepared or diluted with dust-free water. Studies were performed on a Malvern Autosizer IIc (Malvern Instruments) at a fixed angle of 90". The instrument was run in the "automode" as well as in "manual mode", at various sample times, experimental durations, distribution widths, estimated sizes, and other parameters. Number distributions were used for data interpretation. Transmission Electron Microscopy (TEM)-Freeze fracturing and freeze etching of samples were performed with the Refreeze fracture unit (BAF 400; Balzers AG, Lichtenstein). Cryoprotectants were not wed. The samples were sandwiched between two flat golden holders. Fast freezing was done by one of the following methods: slush method in melting nitrogen, slush method in liquid Freon (liquid fluorinated hydrocarbons), slush method in melting propane, or jet freezing with liquid propane. The fracturing temperature was 173 K. Specimens prepared by one of the freezing techniques were sometimes etched for 30 or 60 s (173 K, lO-'-lO-' mPa). The fractured (and etched) frozen samples were shadowed under 45" with platinum/ carbon (layer thickness, 2 nm) and under 90"with pure carbon for replica production. Replicas were cleaned with ch1oroform:ethanol (l:l,v/v) and picked up on uncoated microscope grids. Replicas were viewed with an electron microscope (EM 300, Philips, Kassel, Germany). For replica preparation, the original emulsions or diluted systems were used. Dilution was performed with bidistilled water containing 2.25%(w/v) glycerol USP. 778 / Journal of Pharmaceutical Sciences Vol. 81, No. 8, August 1992
Nuclear Magnetic Resonance (NMR) Spectroscopy-NMR samples were undiluted, except for paramagnetic-shifted samples, and kept in the native form. Trimethylphosphate [S 2.00 ppm, 6.3 mM, CDC1,:MeOH (2:1)] was used as an external reference in a Wilmad NMR reference tube (5 mm), which was placed inside a 10-mm NMR sample tube. In the paramagnetic-shifted samples, p&ymiuy(IID nitrate pentahyhte (50 mM D,O solution) was added in 1:5 pohons. 31PNMR spectra were obtained on a JEOL GX-400 spectrometer operating at a Larmor frequency of 161.70 MHz. The sample temperature was 25 "C. A 45" pulse (11 p )was applied, and 5000 free induction decays were acquired with quadrature detection and Fourier transformed with either 1-, 5-, or 10-Hz exponential line broadening. The spectral width was 60 lrHz (for the whole spectrum) or 8 lrHz (optimized), resolved into 32 000 data points. The pulse repetition time was 32 s. A proton decoupling with a power of 9 W was gated on only during acquisition of free induction decays. The integral was obtained aRer phase a4ustment and baseline correction. The relaxation time was measured by the inversion recovery method, and a value of 11 s for the trimethylphosphate signal (not deaerated) was calculated with a three-parameter nonlinear fit. Small-Angle X-my Scattering and Small-Angle X-ray Diffmction-Small-angle X-ray scattering (SAXS) measurements and small-angle X-ray diffraction studies were performed in a compact small-angle system (Kratky KCLC) connected to a Philips PW 2253111 X-ray tube with copper anode. A nickel foil served as a KB-filter. The following operating conditions were chosen: a tube voltage of 40 to 50 kV and an anode current of 35 to 40 mA. An OED 50 position sensitive detector (Braun GmbH, Munich, Germany) was used,with argon:methane (9:l) as filling gas. The linked multichannel analyzer was a Canberra MCA 8100. Liquid samples were measured in a capillary holder (K-PR) with Peltier temperature control device (Anton Paar, Graz, Austria). Semisolid samples were measured in a homemade holder placed between two Mylar (a kind of polyethylene terephthalate) foils. Exposure times of 3000 or 5000 s were chosen for SAXS measurements. Exposure times of 300-400 s were suflicient for diffraction studies. Scattering curves were analyzed with the computer program collection of 0. Glatter called Fortran IV-Programmpaket ITP-81. Wide-Angle X-ray Dimtion-Studies were performed in a Philips PW 1060/25 X-ray goniometer, equipped with a stepmotor, a Xeproportional Geiger counter (PW 1711/10), and a PW 1710 controller. X-ray tube, anode, and filter material, as well as operating conditions, were the same as for SAXS studies. Samples were measured in aluminum holders. Differential Scanning Calorimetry (DSC)-A Perkin-Elmer DSC 2 4 calorimeter with thermal analysis data station was used for calorimetric measurements. Samples (-6-9 mg) were weighed into Perkin-Elmer, standard, volatile-sample aluminum pans. An empty pan was used as reference. A scan rate of 2.5 or 5 Wmin was chosen. DSC samples were prepared at mom temperature. Each sample was run three times. The samples were heated from 250 to 400 K at a .wan rate of 5 K/min (first run) and held for 5 min at 400 K. Then, the samples were cooled from 400 to 275 K at a scan rate of 5 Wmin (second run) and held for 60 min at 275 K. ARer that, the samples were heated to 400 K again at a scan rate of 5 Wmin (third run).
Results TEM Studies-Undiluted Samples Prepared with Melting Nitrogen (Slush Technique)-The TEM micrographs indicate the original arrangement of particles in the external phase of the emulsion (Figure 1). The particles seem to be randomly distributed in the fracture plan, and their tight packing is obvious. The majority of particles have structured cores. A layered structure is typical for lamellar liquid-crystalline or crystalline materials. Obviously, the oil phase crystallized during freezing of the sample in melting nitrogen. The surrounding water phase, however, appears amorphous. Faster freezing resulted in structure-free, amorphous fractures (Figure 2). For structural analysis, the crystallization artifact of the oil phase (Figure 1)was a n advantage, because samples prepared with melting nitrogen could be used to differentiate between oil-filled, structured fractures and amorphous, water-filled
Flgure 1-TEM micrograph of a replica of the native model iv emulsion showing the tight packing of particles and crystallization artifacts of the oil phase ( x 1 1 700).
Flgure S T E M micrograph of a phospholipid-freesoybean oil emulsion stabilized with casein “micelles”(x49 140).All particles look “cut”and amorphous. No layer structures are detectable. The surface lines are clearly visible.
globules. The micrographs show that particles >lo0 nm are normally oil filled. Larger numbers of LUVs were not detected in the model emulsion. The replicas showed a few particle fractures containing holes or cavities. All particles of this type were >lo0 nm in diameter. The holes in fatty cores of o/w emulsions could be freezing artifacts caused, for example, by differences in density changes of water and oil during the freezing process. Therefore, the surface characteristics of these cavities were studied by water condensation on the fractures. Water condenses predominately on hydrophilic surfaces. On micrographs showing water contamination of the specimen, particles containing holes filled with water drops are detectable (Figure 3). Obviously, the holes have hydrophilic surfaces that result from phospholipids covering and stabilizing the incorporated water. The particles are typical double-emulsion droplets. They consist of small water droplets dispersed in the lipid core of an olw emulsion drop. The water droplets are stabilized by a surfactant layer. However, only a few of these particle-in-particle structures exist in the model iv emulsion. Freon Slush of a Phospholipid-Free Test System-The Freon slush method was used to prepare a phospholipid-free test system. The test system contained casein “micelles”(3w t %) instead of fractionated egg phospholipids as a surface
Flgure S T E M micrograph of a replica of a fractured emulsion droplet containing a highly water contaminated spherical area (x23 400).
stabilizer of a soybean oil emulsion containing -25% (w/w) oil. Some samples show particles with amorphous oil cores (Figure 2). The samples demonstrate that the soybean oil exists in its liquid state at room temperature and that the samples prepared in melting nitrogen contain crystallization artifacts. Although the freezing artifacts of the oil phase were helpful for the distinction between oil- and water-filled structures, crystallization of fat becomes a problem if phospholipid layers are to be detected and quantified. Crystallized fat can occur in layered structures similar to liquid-crystalline lamellar phospholipid layers16 and can mimic phospholipid oligo- or multilayers. The micrograph of the casein-stabilized emulsion shows an area of the specimen containing completely “layer-free” particle fractures, and all structures look “cut” (Figure 2). The surface line of the particles is clearly visible. Therefore, all layered surface structures in an artifact-free sample of the model iv emulsion represent phospholipid layers saturated with compounds from the oil phase. Diluted Samples Prepared with Melting Propane (Slush Technique)-Micrographs of undiluted samples show the extremely tight packing of particles in the model iv emulsion (Figure 1).Because smaller particles are hidden in highly concentrated systems, the emulsion was diluted (l:l), and fast freezing was done by the slush method in melting propane (Figure 4). Freezing in melting propane was fast enough to yield clear cuts through the particles in the fracture plan. The cores of the fractures look nearly amorphous; thus no further distinction between oil-filled and water-containing globules was possible. The majority of fractures show no typical layered structures on the surface (Figure 4). No (dispersed) lamellar phase was found. Multilayer structures, as often discussed for olw emulsions with an excess of emulsifier, could not be detected in the model iv emulsion. Only a few bi- and oligolayer structures were detectable. Most of the fractured oil droplets seem to be surrounded by emulsifier monolayers. The majority of emulsion droplets show the typical characteristics of a n ideal emulsion droplet, that is, they have a liquid lipid core covered by a n emulsifier monolayer. Large numbers of fractures with diameters of C60 nm are visible in the diluted sample. Dispersions of small unilamellar vesicles (SUVs) can be produced by the same techniques used to produce phospholipid-stabilized iv emulsions. However, the micrographs do not allow a clear distinction between SUVs and small emulsion droplets. The existence of small unilamellar liposomes cannot be ruled out by TEM studies nor Journal of Pharmaceutical Sciences I 779 Vol. 81, No. 8, August 1992
30
-
0
0
100
200
300
500
400
700
800
800
r s L chnnnslnr.
Figure &TEM micrograph of fractured, diluted (1:l) model emulsion
showing amorphous oil cores after fast freezing (x31 590). Large
0.0035
numbers of small particles are visible.
*
verified, because SUVs and small emulsion droplets (80-100 nm in diameter are normally oil-filled droplets. SAXS was used to get additional information about a possible coexistence of bilayer arrangements, such as SUVs or L W s , and ideal emulsion droplets in the model iv emulsion. SUVs produce no small-angle X-ray diffraction, but diffuse smallangle X-ray scattering.1618 The scattering curve of an aqueous dispersion of small unilamellar liposomes with a broad and continuous size distribution is smooth and reflects the scattering in the single bilayers.16-ls Typical smooth scattering curves were gained from SUV dispersions with 0.5-6.0% (wlw) phospholipids and a broad size distribution, as determined by PCS measurements (Figure 5, Top). Information about the bilayer thickness was obtained with the flat-sheet approximation for computeradded data analysis. The thickness was obtained from the correlation function (y(J along a line perpendicular to the bilayer plane. The point where finally approaches zero correspondsto the bilayer thickness, which was determined to be 4.5 nm (Figure 5, Bottom). The undiluted emulsion (Figure 6a) and the diluted samples showed no comparable scattering behaviors. Samples of the model emulsion were mixed with a dispersion of SWs. The addition of a vesicle dispersion to the emulsion caused a detectable diffuse scattering typical of scattering in single bilayers if the system contained 0 . 5 4 7 5 % phospholipids bound to vesicle structures. A typical mixture contained 1.05% (w/w) phospholipids originating from the emulsion and -0.75% (w/w) originating from the vesicle dispersion (Figure 6b). The scattering curve of this mixture had a shape comparable with the shape of the curve of the pure vesicle dispersion containing 0.75% (w/w)phospholipids. Larger amounts of single bilayer structures, such as SWs, can be detected directly by SAXS measurements, even in mixtures with ideal submicrometer-sized emulsion droplets. In contrast to the NMR spectroscopic studies, the amount of bilayer arrangements with diameters >80-100 nm was detectable in SAXS experiments, as demonstrated by SAXS measurements on a dispersion of LUVs. If the model iv emulsion contains SUVs or larger bilayer arrangements, the concentration of all these structures in the native emulsion should be below or close to the detection limits of the SAXS measurements. Therefore, less than -28% (w/w)of the total amount of emulsifier is involved in the formation of single bilayer structures, such as vesicular structures.
-
780 I Journal of Pharmaceutical Sciences Vol. 87, No. 8, August 1992
0.0025
0.0015 0.0005 0
--0.0°15 0.0025
~
\
I\
1\
............ .......
.................
I--,
0
10
20
30
.
40
,
.
50
.
,
60
R [A1
Figure W o p ) Scatteringcurve and (Bottom) correlationfunction, y(r), of a SUV dispersion containing 1% (w/w) phospholipids (LaGrange multiplication, 7.00; 20 splines).
NMR S p e c tro s ~ o p y - ~ ~NMR P spectroscopy was used to characterize the conformation and motional properties of phospholipids in various model systems.19 31P NMR line shapes are, inter alia, not only dependent on the sizes of phospholipid particles20 but also affected by a number of other parameters, such as temperature, magnetic field strength, viscosity, aqueous phase pH, dispersion state, and methods of emul~ification.~~-2s Small phospholipid-containing structures, such as micelles and sonicated vesicles, show relatively narrow, symmetrical 'lP NMR lines, reflecting isotropic motional averaging and arising mostly from lateral diffusion and Brownian tumbling. Unsonicated phospholipid bilayers, on the other hand, exhibit broad, unsymmetrical lines (powder-type lines), which are caused by restricted anisotropic motion of the phospholipid molecules. In the 'lP NMFt spectrum of the model iv emulsion, three types of line shapes are seen (Figure 7a): narrow signals with a half-line width (Avl12) at 2 Hz and a chemical shift (6) at 6 0.24 and S -0.33 ppm; broader (Avl12 = 83 Hz),but still symmetrical, signals a t 6 -0.07, 6 -0.62, and 6 -1.06 ppm (Figure 7c); and a n underlying asymmetrical signal with a chemical shift anisotropy (Am) of 32 ppm (where A u = 9 - q, and 9 and q,are components of u that are, respectively, parallel and perpendicular to the molecular symmetry axis; cq = 23 ppm; 7,= -9 ppm; Figure 7b). Possible other underlying signals with various line shapes are not resolved. In 31P NMR spectra of one system (i.e., emulsions), the parameters that should influence line shapes are particle size and particle structure (all other parameters are constant). Different line shapes in the emulsion spectrum indicate a nonhomogeneity in particle structures and sizes. For correct
-
O 80 O\
6oT
.
401
I
0
E g
90 nm in diameter. This size distribution delivered about the same line shapes and chemical shifts as the broad symmetrical signals in the emulsion spectrum (Figure 8a). A mixture of small unilamellar liposome dispersion and emulsion (9:l) had liposome signals that overlapped with the broad symmetrical signals in the emulsion spectrum. This result supports the TEM results that indicate larger amounts of particles in the system with diameters below -100-120 nm. In contrast, preparations containing either large multilamellar structures or large unilamellar liposomes gave line shapes similar to those of the asymmetrical emulsion signal (Figure 9). Taking TEM and SAXS results into account, we concluded that the broad, asymmetrical signal of the emulsion does not result from a dispersed lamellar phase but arises predominately from phospholipid molecules stabilizing large, ideal emulsion particles. Mixed micellar systems of different compositions were also prepared. The observed line width for the micelles varied from -5 to 16 Hz, depending on the phospholipid used for the
2
1
0
I""I -1
-2
"'""'I"' -3 -4
I
I
I T
-5
PPm Figure 7-31P NMR spectrum of model iv emulsion (a) with 1BkHz spectral width and 10-Hz line broadening, (b) as in a, but enhanced, and (c) with 1.5-kHz spectral width and 1-Hzline broadening.
preparation. LPC-containing mixed micelles had line widths of 5 Hz (Figure 10a) and PC-containing mixed micelles had line widths of 16 Hz (Figure lob), reflecting the difference in mobility and size of the micelles. A narrowing of the line widths can be seen in the spectra of mixed micelles containing fractionated egg phospholipids (Figure 1Oc). Drastic differences in total micellar concentration had only a minor influence on the line width and the chemical shift. To elucidate if lysophospholipids are dissolved in the aqueous phase or incorporated into the emulsifier layers, the emulsion was dialyzed, and the dialysate was examined with 'lP NMR spectroscopy. Signals with much smaller half-line widths ( A V , , = ~ 1Hz)are observed, and chemical shifts are not in accord with the observed micellar chemical shifts (i.e., lysophospholipids are predominantly incorporated into the emulsifier layers). No micelles or mixed micelles formed by emulsifier and/or degradation products could be detected in the 'lP NMR spectra of the model iv emulsion. The asymmetrical and the broad, symmetrical signals of the emulsions are clearly broader than the observed signals for the micelles ( A Y ~ /
