Phase Behavior of the Quaternary Poly (dl-lactide-co-glycolide

Nov 1, 2001 - Long-Acting Injectable Hormonal Dosage Forms for Contraception. Linfeng Wu , Dileep R. Janagam , Timothy D. Mandrell , James R. Johnson ...
0 downloads 0 Views 90KB Size
J. Phys. Chem. B 2001, 105, 12157-12164

12157

Phase Behavior of the Quaternary Poly(DL-lactide-co-glycolide)/Monoolein/ 1-Methyl-2-pyrrolidinone/Water System: An Experimental and Theoretical Study Anna K. Johansson,*,† Per Linse,‡ Lennart Piculell,‡ and Sven Engstro1 m† Pharmaceutical Physical Chemistry, Department of Pharmaceutics, Uppsala UniVersity, P.O. Box 580, SE-751 23 Uppsala, Sweden, and Physical Chemistry 1, Center for Chemistry and Chemical Engineering, Lund UniVersity, P.O. Box 124, SE-221 00 Lund, Sweden ReceiVed: May 8, 2001; In Final Form: September 4, 2001

The phase behavior of the poly(DL-lactide-co-glycolide) (PLG)/monoolein (MO)/1-methyl-2-pyrrolidinone (NMP)/water system has been studied. The particular system was selected because it was a candidate for being a suitable system for creation of lipid-polymer hybrid particles. Phase diagrams of the four ternary subsystems, as well as phase behaviors of selected quaternary mixtures, have been determined. The PLG/ NMP/water system exhibits a behavior similar to a polymer/solvent/nonsolvent system, and the PLG/MO/ NMP system shows a segregative behavior similar to a polymer 1/polymer 2 (or surfactant)/solvent system. The ternary MO/NMP/water subsystem exhibits a rich, mainly mesomorphic phase behavior, including lamellar and cubic phases, and offers thus possibilities to create both liposomes and cubosomes. At high NMP content, the system gives rise to several liquid phases, such as a sponge phase and MO-poor and MO-rich liquid phases. Addition of PLG to the MO/NMP/water subsystem reveals that both liquid crystals and liquid phases may exist in equilibrium with PLG-rich liquids, the latter being highly viscous. At high water content, the PLG phase may be caught in a glassy state at room temperature. The Flory-Huggins theory has been used to model three of the ternary subsystems and some parts of the quaternary system.

Introduction Poly(DL-lactide-co-glycolide) (PLG) is a linear polyester, which because of its biocompatibility has received much interest within the pharmaceutical field. PLG is insoluble in water and degrades slowly (weeks to years depending on polymer composition and temperature) under formation of glycolic and lactic acid, which in vivo transform into carbon dioxide and water. The most common use of PLG as a drug delivery matrix is in the form of (injectable) particles of micrometer size, but other dosage forms have been presented as well.1-5 Polymer particles can be produced by double emulsification,6-10 spray-drying,11-15 and supercritical fluid techniques.16,17 One drawback with the PLG system is that it offers a relatively harsh inner environment for sensitive drugs because of the decrease in pH during degradation.18 A number of methods have been introduced to overcome this difficulty.3,19 Lipids have been used for a long time for drug delivery in the form of emulsions and, more recently, liposomes. A liposome, which consists of one or several concentric spherical bilayers, has a nonlamellar analogue in the cubosome. The latter is made from a bicontinuous cubic liquid crystal. The most studied system in this case is the monoolein (MO)/water system20-22 in which the cubic phase forms spontaneously in excess water. The cubic phase has been shown to be able to accommodate and stabilize both water-soluble23-29 and membrane30-37 proteins. Recently, it was shown that the cubic phase can be transformed into a so-called sponge phase (L3) by addition of a solvent such as dimethyl sulfoxide, ethanol, * To whom correspondence should be addressed. [email protected]. Fax: +46 18 4714377. † Uppsala University. ‡ Lund University.

E-mail:

propylene glycol, poly(ethylene glycol),38,39 or 1-methyl-2pyrrolidinone (NMP).40 The sponge phase is a liquid, which in this particular case is transformed to a cubic phase in excess water. Here, NMP is of interest because it is a good solvent for PLG and because it is claimed to be biocompatible and as such is used in pharmaceutical formulations with PLG.41 The motivation of this study is to explore the basis for the formation of lipid-polymer ”composite” particles with welldefined microdomains rich in polymer or lipid. To be able to decide on a process for creation of lipid-polymer composite particles, the phase separation occurring during the considered process must be known. Such particles, if formed, may offer a gentler environment for proteins and peptides if these could be forced to reside in the lipid domain. Lipid-polymer hybrid particles seem to offer possibilities for, for example, oral delivery of poorly absorbed drugs.42 As an initial stage, we have undertaken a phase study of the quaternary PLG/MO/NMP/ water system. Here, we present experimentally determined phase diagrams for the four ternary subsystems as well as some aspects of the phase behavior of the quaternary system and results from model calculations using the Flory-Huggins theory. We have made a detailed phase study of the MO/NMP/water system, and in particular, the regions at high-NMP contents have been studied. The results should be useful for the future choice of conditions for particle formation and for the prediction of the fate of such particles during storage and in vivo administration. Experimental Section Materials. Poly(DL-lactide-co-glycolide) (PLG, RG504H, Mw ) 50 000, Mn ) 11 000) was purchased from Boehringer Ingelheim (Ingelheim am Rhein, Germany), and monoolein (MO) was purchased from Danisco Ingredients A/S (Brabrand,

10.1021/jp011750d CCC: $20.00 © 2001 American Chemical Society Published on Web 11/01/2001

12158 J. Phys. Chem. B, Vol. 105, No. 48, 2001

Johansson et al.

Denmark) (batch 1876, lot nos. 88 and 115). The monoglyceride content of lot nos. 88 and 115 was 94.9% and 94.0%, respectively, and the rest was mainly diglycerides and free glycerine. The monoolein part of the monoglyceride content was 88.4% and 89.8% for lots 88 and 115, respectively. Anhydrous grade 1-methyl-2-pyrrolidinone (NMP) was purchased from Sigma-Aldrich Chemie (Steinheim, Germany). Water filtered through a Millipore purification system was used throughout the study. All ingredients were used as received. Sample Preparation and Phase Characterization. The samples were prepared (0.8-3.0 g of material per sample) in glass vials, which were sealed and held at the temperature of interest until equilibrium was reached (days to weeks). To separate different phases, the samples were centrifuged (at 9400g, 10-30 min, 20 °C) in a Beckman Avanti 30 compact centrifuge (Palo Alto, California). An F0650 rotor with a fixed angle of 25° was used in all centrifugations. Phase characterization was performed by visual inspection. Anisotropy was detected by using crossed polarizers. Effects of the polydispersity of PLG on the compositions of coexisting phases have been neglected. Differential Scanning Calorimetry (DSC). The glass transition temperatures were determined by means of DSC (220C Seiko Instruments, Japan). The software Xstar 6000, version 3.4 A (Seiko Instruments, Japan) was used for instrument control and data analysis. Hermetically sealed coated aluminum pans (TA Instruments, New Castle, DE) were used throughout the study. Glass transition temperatures were generally estimated as the temperatures at which the heat capacity changed rapidly during heating. Where nothing else is written, the heating/ cooling rate was 3 °C/minute. All transitions were checked for reversibility. Theoretical Section The Flory-Huggins theory43,44 was used to theoretically describe the phase behaviors of the three ternary PLG/MO/NMP, PLG/NMP/water, and MO/NMP/water systems. Experimentally determined data were used to fit the interaction parameters, and predicted phase diagrams were used as guidance for selecting new sample compositions. In the Flory-Huggins theory, it is assumed that the polymer is linear and flexible. Furthermore, the components are assumed to mix in a fully random pattern, and the interactions are assumed to be restricted to the nearest neighbors. By minimization of the total free energy, the number of phases and the distribution of the components between the different phases at equilibrium can be determined. The total free energy per lattice site, Atot, is given by

Atot )

∑R φRAR

(1)

where R is the phase index, φR is the volume fraction of phase R, and AR is the free energy per lattice site of phase R. In the Flory-Huggins theory, AR is given by

AR ) kT

[

φiR 1 lnφiR + 2 i

∑i r

∑i ∑j χijφiRφjR

]

(2)

where ri is the number of lattice sites occupied by component i, φiR is the volume fraction of substance i in phase R, and χij is the interaction parameter defined as

χij )

[ (

)]

wii + wjj z wij kT 2

(3)

Here, wii, wjj, and wij denote the (free) energies of interaction between components ii, jj, and ij, respectively, of neighboring positions in the lattice. The interaction parameters were hence concentration-independent as in the original Flory-Huggins theory. Finally, z denotes the number of nearest neighbors, and kT is the Boltzmann constant times the temperature. In a binary solution, the theory predicts a phase separation if χij > χcij where

χcij )

(

1 1 1 + 2 r x i xrj

)

2

(4)

From the sets {ri} and {χij} and the relative amounts of each component, the number of phases and their compositions, {φiR}, can be calculated. Because the densities of the components are almost equal, substituting weight fractions for volume fractions gives rise to negligible errors as compared to the other approximations in the Flory-Huggins theory. Therefore, the results of the calculations in volume percentages are directly compared with the results of the experimental study in weight percentages. The temperature enters the model only through the interaction parameters χij. Three ternary subsystems were modeled in detail, and the following procedure was used: On the basis of the relative molecular weights, the set of r parameters, rw ) 1, rNMP ) 2, rMO ) 20, and rPLG ) 611, were assigned to the four components. The precise values are not critical. With four components, there are six independent interaction parameters. Three of them were determined by fitting the calculated PLG/ NMP/water phase diagram to the experimental one. The χPLG/NMP parameter thus determined was subsequently used in the fitting of the PLG/MO/NMP phase diagram by adjusting χPLG/MO and χMO/NMP. The χNMP/w and χMO/NMP parameters determined from the first and second fits, respectively, were finally used in the fitting of the MO/NMP/water phase diagram in which only χMO/w was adjusted. Results and Discussion This section presents the experimentally obtained phase behavior and the results of the modeling of the phase behavior according to the Flory-Huggins theory. The subsystems as well as the quaternary system are presented and discussed further below. Generally, a quaternary system is rather complicated, so we have focused on the phase behaviors of the four ternary subsystems composing the four sides of the tetrahedron-shaped phase diagram. Figure 1, with the sides of the tetrahedron foldedout, presents the resulting phase diagram. The overall features of the phase diagram are determined by the following properties of the system. (1) NMP is totally miscible with each of the other three substances, while the latter are more or less immiscible in pairs. (2) Owing to its amphiphilic properties, MO forms bilayers in mixtures in which water is the dominating solvent. This results in the formation of structured (liquid crystalline) phases. Note that the comparatively high miscibility of MO with water is entirely due to these structured phases; the molecular solubility of MO in water is very low. As a consequence of the above, the quaternary system contains one single tripod-shaped isotropic phase, denoted L, that emanates from the NMP corner of the phase tetrahedron and extends in three narrow branches along the three axes describing the binary mixtures of NMP with polymer, lipid, or

Quaternary PLG/MO/NMP/H2O System

J. Phys. Chem. B, Vol. 105, No. 48, 2001 12159

Figure 1. Experimentally determined phase diagrams of the ternary PLG/MO/NMP/water subsystems at 20 °C (in wt %). The four ternary phase diagrams constitute the four surfaces of the phase tetrahedron, which describes the phase behavior of the quaternary system. Each experimental sample is marked by a dot. One-phase regions are bounded by solid curves, and in the MO/NMP/water and PLG/MO/water systems, dashed lines separate two- and three-phase regions. In the PLG/NMP/water and the PLG/MO/NMP phase diagrams, dashed lines denote tie lines.

water. In the following, these branches will be denoted LPLG, LMO, and Lw, respectively. In addition to the isotropic phase, the phase tetrahedron contains three phases, all involving lipid bilayers that are confined to the MO/NMP/water “bottom” plane. These phases are the lamellar phase (LR), the cubic phase (Q), and the “sponge” phase (L3). Brodbeck et al. have recently studied the phase behavior of the ternary PLG/NMP/water system.45 The phase behavior of the ternary MO/NMP/water system has partly been studied before by Ekelund.40 The Binary PLG/NMP System. PLG of the quality used has a glass transition temperature of about 41 °C. PLG is completely miscible with NMP, but the glass transition can give rise to slow kinetics. To be sure that the samples containing PLG had not been locked in nonequilibrium states, the glass transition temperature, Tg, versus PLG content was determined by means of DSC. Figure 2 illustrates the result from the DSC measurements, which reveals a linear relationship between weight percentage of polymer and the transition temperature for the interval investigated. This behavior is similar to that of the polystyrene/trans-decalin system.46 Figure 2 shows that Tg falls below room temperature when the PLG content is below ca. 90%. The Ternary PLG/NMP/Water System. The phase behavior of the PLG/NMP/water system is illustrated in Figure 1. We observe a phase behavior that is typical for polymer (PLG)/ solvent (NMP)/nonsolvent (water) systems.47 The four determinations of the tie line at two different preparation temperatures, using two different methods (described below), agreed, and the NMP contents of the water-rich isotropic phases are given in Table 1. This agreement implies that (i) the polymerrich phase is not a glass at 20 °C and (ii) no significant amounts of hydrolysis products are present in the water-rich isotropic

Figure 2. Glass transition temperature versus PLG content of the binary PLG/NMP system.

TABLE 1: NMP Contents of the Water-Rich Phase of Ternary PLG/NMP/Water Samplesa sample

temperature during dissolution (°C)

NMP content (%) (refractive index)

NMP content (%) (lever rule)

1 2

50 20

76 76

75 74

a

8.0% PLG, 70.0% NMP, and 22.0% water.

phase. The second conclusion is based on the fact that if hydrolysis products were present in the water-rich isotropic phase, the refractive index would have increased because of formation of DL-lactic acid and glycolic acid, which in turn would have given a tie line disobeying the lever rule. Keeping the sample at a temperature higher than the glass transition temperature, 41 °C, significantly reduces the PLG

12160 J. Phys. Chem. B, Vol. 105, No. 48, 2001 dissolution time, but the disadvantage is an expected increase of the rate of PLG degradation. Therefore, the determination of one tie line of the PLG/NMP/water system was performed by using two identical samples prepared in a region with still rather short dissolution times. To detect insufficient dissolution or increased hydrolysis, one sample was kept at 20 °C for 15 h and the other one at 50 °C for the same time. Finally, both samples were kept at 20 °C for 11 h. The sample compositions and temperatures during dissolution are given in Table 1. The phases were separately weighted, and the refractive index of the water-rich isotropic phase was determined at 20 °C. Brodbeck et al. have determined tie lines for the system PLG (Resomer 502, Mw ) 12 000)/NMP/water by means of HPLC and Karl-Fisher titration.45 From the tie lines reported,45 one can conclude that the PLG content of the water-rich phase is negligible. The assumption that the water-rich isotropic phase does not contain PLG is also confirmed by the phase boundary determined in our study. Under this assumption, the NMP content of the water-rich isotropic phase was extracted from a standard curve of the refractive index versus NMP content of the binary NMP/water system. The samples used for creating the standard curve were treated in exactly the same way as the samples used to estimate the tie line. All refractive indexes were measured at 20 °C. The tie line was constructed by drawing a straight line from a point on the diagram with zero MO content and with the NMP/water ratio of the water-rich isotropic phase through the point describing the total composition of the twophase sample. The tie line ends when the line crosses the phase boundary. To validate the assumption made regarding the refractive index, the tie line was also determined by using the lever rule, which means that the total composition of the twophase sample and the corresponding weights of the phases were used in the calculations. Table 1 contains the resulting NMP contents from the refractive index tests on the liquid phase and the NMP contents estimated by the lever rule. The slope of the tie line is steep, indicating that the polymer-rich phase is indeed concentrated and the NMP/water ratio is, as expected, higher for the polymer-rich phase. The Ternary PLG/MO/NMP System. The phase behavior of the PLG/MO/NMP system is similar to that of the PLG/ NMP/water system and is given in Figure 1. We found a complete miscibility at all compositions of the binary MO/NMP system, whereas PLG and MO are practically immiscible. Moreover, there is a critical NMP concentration of 78% w/w below which the system may separate into one PLG-rich isotropic phase and one MO-rich isotropic phase. A tie line in the PLG/MO/NMP system was also determined using equilibrated centrifuged samples. The weights of the phases were determined, and the tie line was calculated according to the lever rule. According to the slope of the tie line, the NMP content is almost the same in the two coexisting phases, displaying only a weak preference of NMP for MO over PLG. These findings are all in agreement with the segregative phase behavior often observed in polymer 1/polymer 2/solvent systems44 as well as in polymer/surfactant/solvent systems.48,49 In the latter type of systems, the surfactants are assumed to associate forming micelles, thus giving rise to macromolecular entities with similar phase properties as polymers. In the present case, we have no experimental evidence for association of the lipid (MO) in the solvent (NMP), so this ternary system can be regarded as a polymer/solvent/nonsolvent system as well. Pure MO melts at 35 °C, which means that it in our study is in crystalline form at room temperature.

Johansson et al.

Figure 3. Phase diagram of the ternary MO/NMP/water system with squares showing the locations of the samples used for the addition of PLG. Numbers within brackets denote the sample name according to Table 2.

The Ternary MO/NMP/Water System. The phase behavior of the MO/NMP/water system is also given in Figure 1. A crude phase map of this system was determined by Ekelund focusing on the sponge-phase region.40 In our work, a more thorough investigation of the phase behavior was undertaken. In comparison with the two ternary subsystems presented above, the present system shows a rich mesomorphic behavior with a multitude of different structures. Because of its complexity, an enlarged version of this diagram is presented in Figure 3. We have found a few major differences compared to the published diagrams by Ekelund.40 The true lamellar one-phase region does not extend as far toward low-MO contents as was previously reported. Instead, two three-phase regions are found in that region. Samples from these three-phase regions consist of waterrich and lamellar phases, with the third phase being either the sponge phase (L3) or the isotropic MO-rich phase. Yet another three-phase region has also been found, containing the lamellar, cubic, and L3 phases in coexistence. One way to describe the phase behavior, at less than 34% MO content (the more interesting part of the phase diagram from a particle formation point of view), is to explain the effect of adding MO to NMP/water mixtures with increasing NMP/ water ratios. In NMP/water mixtures with NMP/water ratios up to 23:77, addition of MO gives rise to formation of the cubic phase (Q) coexisting with a water-rich isotropic liquid phase. Moreover, adding MO to mixtures with NMP/water ratios between 23:77 and 32:68 can result in formation of both a sponge (L3) and a cubic phase. If both are formed, they coexist with a water-rich isotropic liquid phase consisting of approximately 25% NMP and 75% water. Adding MO when the NMP/water ratio is between 32:68 and 55:45 brings the sample into a two-phase region where sponge phase is in equilibrium with a water-rich isotropic liquid phase, into a one-phase region with sponge phase, or into a two-phase region where sponge phase is in equilibrium with a lamellar phase (LR). Thus, MO is essentially insoluble in NMP/water mixtures with NMP/water ratios between 0:100 and 55:45 but may swell and form bicontinuous lipid bilayer structures. When MO is added to NMP/water mixtures with NMP/water ratios between 55:45 and 65:35, a sponge phase and a lamellar phase (LR), only a lamellar phase, or both an isotropic MOrich phase (LMO) and a lamellar phase can be formed in addition

Quaternary PLG/MO/NMP/H2O System to the water-rich isotropic liquid phase. In other words, in this part of the phase diagram, two different three-phase regions exist with a narrow two-phase region between them where a lamellar phase is coexisting with an NMP/water mixture. Addition of MO to NMP/water mixtures with NMP/water ratios between 65:35 and 75:25 gives rise to formation of one MO-rich and one MO-poor isotropic liquid phase, similarly to the two PLG-containing ternary systems described above. At higher NMP/water ratios, unlimited amounts of MO can be dissolved. On the subject of applications, it might be possible to create liposomes from the narrow region where LR is coexisting with Lw. Even though liposomes have not been created here, it seems likely that homogenization of a two-phase sample with an MOcontent of less than 5% from the above-mentioned region could result in formation of more or less stable liposomes. We have made some initial experiments where cubosomes have been formed (confirmed by cryotransmission electron microscopy (TEM)) by spraying the L3 phase into excess water. To obtain increased stability of the cubosome dispersion, a small amount (corresponding to a final content of about 0.5-1 wt %) of poloxamer 188 (pluronic F68) can be dissolved either in the original L3 phase or in the excess water. The Ternary PLG/MO/Water System. To understand the ternary PLG/MO/water system, PLG was added to two binary MO/water samples, one containing cubic phase and the other lamellar phase, until the total PLG content was 0.2% in each sample. After a week, nothing had happened with the white PLG granules, and the same held after bringing the samples to 42 °C overnight and then keeping them at room temperature for another week. Hence, at 20 °C PLG is not very soluble in either the lamellar or the cubic phases. To further investigate the ternary PLG/MO/water system, a sample from the LMO/LR region of the binary MO/water system was centrifuged thoroughly and PLG was then added to the lighter of the two phases in the vial, the LMO phase. The sample was held at 20 °C, and the PLG granules were observed for 5 min until they had extracted some MO and water and formed a yellowish concentrated phase. Addition of PLG to pure water or pure MO (in melted or crystalline form) did not results in a yellowish polymer-rich phase, which indicates that both MO and water were extracted when PLG was added to the LMO/LR sample. The sample then consisted of three phases: LMO, LR, and the PLG-rich isotropic phase, proving that the system PLG/ MO/water contains a one-phase region, as shown in Figure 1. To get an idea of the composition of the PLG-rich isotropic phase, PLG was added gradually to the three-phase sample kept at 35 °C (the melting point of the lipid, a temperature below the glass transition temperature for the pure polymer). The lamellar phase disappeared, which implies that the polymerrich phase has a greater water/MO ratio than the corresponding initial ratio for the sample. During these tests, all PLG was added in less than 30 min, which means that the hydrolysis and the corresponding water consumption were negligible, even though the test was performed at 35 °C. The phase boundary of the PLG-rich one-phase region, according to Figure 1, is schematic. Also evident from Figure 1 is the large three-phase region where the PLG, Q, and water phases are coexisting. The Quaternary PLG/MO/NMP/Water System. Selected parts of the quaternary system were examined by adding (i) PLG to samples in the ternary MO/NMP/water subsystem and (ii) water to samples in the ternary PLG/MO/NMP subsystem. In the former case, the polymer swells, and in the latter case, the polymer-rich phase concentrates.

J. Phys. Chem. B, Vol. 105, No. 48, 2001 12161 TABLE 2: Composition and Phase Behavior of Quaternary PLG/MO/NMP/Water Systems initial phasea

PLG (%) MO (%) NMP (%) water (%)

(1) LMO

32.0

8.9

44.0

15.1

(2) LMO

19.9

38.0

28.6

13.5

(3) LMO/LR

6.2

45.6

30.9

17.3

(4) L3

1.1

5.9

51.5

41.5

(4) L3

22.0

4.7

40.6

32.7

(5) L3

1.1

36.0

21.4

41.5

(6) Lw

3.4

0.9

65.2

30.5

phase behavior LMO and PLG rich LMO and PLG rich LMO/LR and PLG rich L3 and PLG rich L3 and PLG rich L3 and PLGb LMO/Lw and PLG rich

a Initial phase of the MO/NMP/water system. The numbers within the parentheses are used for illustrating the location of the samples in the MO/NMP/water phase diagram, see Figure 3. LMO/LR denotes a sample from the LMO and LR two-phase region. b PLG did not swell at 20 °C.

The PLG addition was made to the samples indicated in Figure 3 and listed in Table 2. A general observation is that PLG is not soluble to any substantial amount in any of the samples studied. PLG swelled and formed highly viscous liquids, except in the sample with the lowest concentration of NMP. Because all samples are close to phase borders, we expected to be able to draw some conclusions about the composition of the liquid extracted by PLG from changes in the phase behavior of the original phases. Table 2 reveals that it is reasonable to conclude that PLG extracts more than NMP (or a NMP/water mixture greater than 9/1 w/w) from the original samples. If this was not the case, one would expect the remaining lipid systems to move away from the NMP corner into new areas of the phase diagram. For example, the original L3 phase with high amounts of NMP should be depleted of NMP, which in turn would lead to the formation of an Lw phase. According to our results, however, the L3 phase persists which implies that PLG extracts, besides water, also some MO. A similar reasoning can be applied to the other samples in Table 2 as well. For the L3 phase with low amounts of NMP, PLG did not swell, which may be understood from the very steep tie lines in the PLG/NMP/water system leading to a glass transition temperature above room temperature when the NMP content becomes too low, see Figure 2. Support for this conclusion comes from the fact that when this sample was heated to 42 °C for 14 days PLG formed a clear yellowish PLG-rich phase. Water was added in a stepwise manner to three samples in the PLG/MO/NMP subsystem with 75% NMP content. Two samples, A and B, were located in the two-phase region, and sample C was located in the one-phase region (see Figure 4a). Figure 4b summarizes the phase behavior resulting from adding water to the samples. Up to approximately 20% water content, the samples were in a two-phase region containing two isotropic liquids in which one of them is a yellowish concentrated PLGrich isotropic phase and the other one is a diluted L phase. On further addition of water, the same sequence of phases appeared (in addition to the PLG-rich isotropic phase, which was always present) as is found with increasing water content in the ternary MO/NMP/water system. However, because the time to reach equilibrium was long and because we tried to minimize the hydrolysis, the narrow LR/Lw/PLG-rich and Q/L3/Lw/PLG-rich regions were not studied. Common to all three samples, as more water was added, the yellowish color of the PLG-rich isotropic

12162 J. Phys. Chem. B, Vol. 105, No. 48, 2001

Johansson et al. TABLE 4: Interaction Parameters Used in the Modeling Using the Flory-Huggins Theorya χij

MO

NMP

water

PLG MO NMP

0.160 (0.035)c

0.100 (0.279)b 0.070 (0.433)c

2.400 (0.541)b 1.755 (0.749)d 0.005 (1.457)b

ar PLG ) 611, rMO ) 20, rNMP ) 2, and rwater ) 1. The corresponding critical interaction parameters, calculated from eq 4, are given within parentheses. b From the fit to the experimental ternary PLG/NMP/water phase diagram. c From the fit to the experimental ternary PLG/MO/ NMP phase diagram. d From the fit to the experimental ternary MO/ NMP/water phase diagram.

Figure 4. Phase diagram of the ternary MO/PLG/NMP system with squares showing the locations of samples before addition of water (a) and a cut through the tetrahedron illustrating the experimental phase behavior from water addition to samples A, B, and C (symbols) and the calculated border between the L/PLG-rich two-phase region (below the dashed line) and the LMO/LW/PLG-rich three-phase region (above the dashed line) as predicted by the Flory-Huggins theory (b).

TABLE 3: Glass Transition Temperature, Tg, of PLG-Rich Phases from Sample B phases

water content (%)

Tg (°C)

timea (days)

L3/PLG-rich L3/Lw/PLG-rich L3/Lw/PLG-rich Lw/Q/PLG-rich

37.3 47.3 54.3 69.4

4 10 14 17

1 1 1 1

a

Time between preparation and analysis.

phase gradually turned white. This indicates that the PLG-rich isotropic phase gets more concentrated. The simplest explanation for this trend is that the added water extracts NMP. Determinations of the glass transition temperature were made on small amounts of the polymer-rich phase of sample B to investigate whether the polymer was in a glassy state. Table 3 contains the results from the DSC measurements on the PLGrich isotropic phases. During these measurements, the more distinct transition during cooling was used to estimate Tg. This means that because of hysteresis effects the true Tg is probably two or three degrees higher. The results show that the sample

with 69.4% water content might be a glass at 20 °C and that the other samples with high-water contents are close to being glasses at 20 °C. Investigations were also performed on the PLGrich/Q/Lw sample to determine the influence of hysteresis as the cooling/heating rate varies. As can be seen in Table 3, Tg was determined to 17 °C with a cooling/heating rate of 3 °C/ min. Increasing the cooling/heating rate to 5 °C/min gave a measured value of 16 °C. When finally the cooling/heating rate was increased to 7 °C/min, Tg was determined to 15 °C. The conclusion is that the influence of hysteresis decreases with decreasing cooling/heating rate. In summary, the best way to understand what the phase diagram of the quaternary system looks like is to focus on the PLG-rich phase. Because added PLG swells and does not get incorporated into any of the phases of the MO/NMP/water system, the other phases found in the quaternary system are the same as in the MO/NMP/water system. Tie lines stretch from the MO/NMP/water surface of the tetrahedron-shaped phase diagram to different points on the surface defining the PLG-rich one-phase region. Figure 4b illustrates how all of the phases of the MO/NMP/water surface are found in a cut through the tetrahedron. Flory-Huggins Modeling of the Systems. First, the PLG/ NMP/water system was considered. The value of χPLG/w was determined to yield a strong phase separation in the binary PLG/ water system. Then, the χPLG/NMP and χNMP/w parameters were obtained from fits to the experimentally determined tie line and the location of the critical point. Thereafter, the PLG/MO/NMP system was considered, for which χPLG/MO and χMO/NMP were determined. Finally, calculations were performed in the segregative region appearing at high-NMP contents of the MO/NMP/ water system for which χMO/w was determined. All interaction parameters fitted are collected in Table 4, and the fitted phase diagrams are compared with the experimental ones in Figure 5. For the PLG/NMP/water system, the model calculations are in a reasonable agreement with the experimental phase behavior. Thus, the phase behavior of the PLG/NMP/water system can be described within the Flory-Huggins model. Obviously, χPLG/NMP and χNMP/w are smaller than the corresponding critical interaction parameters, whereas χPLG/w is higher than the critical one (see Table 4). Figure 5 also shows that the calculated phase behavior of the PLG/MO/NMP system is in excellent agreement with the experimentally determined phase behavior. Here, χPLG/NMP and χMO/NMP are smaller than the corresponding critical interaction parameters, whereas χPLG/MO is higher than the critical one (see Table 4). At NMP contents below 70% and at NMP/water ratios higher than ∼65:35, the experimentally determined ternary MO/NMP/ water system exhibits a region where phase separation results in the formation of two liquids. With only χMO/w left to be

Quaternary PLG/MO/NMP/H2O System

Figure 5. Calculated phase boundaries (dashed curves) and tie lines (dashed lines) from the Flory-Huggins theory and experimentally determined one-phase boundaries (solid curves) and two-phase and three-phase boundaries and tie lines (solid lines) for three of the four ternary phase diagrams.

adjusted, also this region of the phase diagram was well-captured by the Flory-Huggins theory. The slope of the calculated tie line is in good agreement with the slope of the tie lines near the three-phase region in the upper part of the phase diagram. The main deviation between the experimentally determined phase behavior and the calculated one is the location of the phase boundary that divides the one-phase MO-rich region from the two-phase region. The effective PLG/water interaction is unfavorable which, together with the fact that PLG is a polymer, results in a very strong segregation, even in the presence of NMP. MO and water are more miscible, mainly because MO is smaller than PLG (the interaction parameter is still large). MO and PLG, finally, are more miscible than water and PLG, even though MO is a larger molecule than water. According to the model, the PLG/ MO interaction is much less unfavorable than the PLG/water interaction, which seems very reasonable. The Flory-Huggins model was also applied to some extent to the quaternary system. The calculations were carried out using the values of the interaction parameters determined for the subsystems (see Table 4). The calculated phase behavior of the quaternary system was in reasonable agreement with the experimental one, see Figure 4b, with respect to the amount of water needed to form a third phase. The deviation between the slope of the phase boundary estimated from the Flory-Huggins calculations and that from the experimental findings (see Figure 4b) reflects the deviation between theory and experiment in the MO/NMP/water system. Approaches to Particle Formation. Taking a starting point in the ternary PLG/MO/NMP system with a liquid containing more than 75% NMP and with a PLG/MO ratio of approximately 4:1 and spraying such a liquid into excess water could be a possible way to create lipid-polymer hybrid particles. However, one main complication has to be considered. The interaction between MO and PLG might be unfavorable in the sense that the phase separation is strong when NMP leaves the droplets about to become particles. As NMP is leaving a droplet, MO and PLG tend to separate, which might yield welldefined regions of MO within a polymer matrix, but if the separation is too strong, MO might leave the polymer matrix completely. On the other hand, if some water enters the droplet, MO might be trapped within the polymer matrix in the form of the concentrated PLG-rich phase found in the MO/PLG/water system. Considering the fact that the surface of the droplet will be the first part experiencing high water concentration, a polymer shell will probably form quickly. If water enters inside

J. Phys. Chem. B, Vol. 105, No. 48, 2001 12163 the shell, the concentrated PLG-rich phase will be found there and, depending on the remaining MO content and the water content, also different MO-rich phases. If no water at all enters inside the polymer shell, separate regions of pure MO and pure PLG will be found inside the particle. The kind of particles created is a result of a competition in which (i) NMP leaves the droplets and in which (ii) water affects at least the surfaces of the droplets at the same time as (iii) MO and PLG separate. One would like to be able to adjust the interaction between MO and PLG to get the desired phase separation. Qualitatively, this can be done by using different lengths of PLG molecules, thus affecting the entropy of mixing. Another approach is based on the emulsification technique, frequently used when making PLG particles. One important result from the present study is that lipid and polymer tend to separate into different phases in the common solvent NMP. This gives one the opportunity to disperse one liquid phase into the other, starting, for example, in the two-phase region of the MO/ PLG/NMP system at an NMP content of approximately 65% and emulsifying the MO-rich phase into the polymer-rich phase. The resulting emulsion may then be poured under vigorous stirring into an aqueous solution containing some NMP. By the choice of a proper NMP/water ratio of the aqueous solution, it is possible to succeed with the otherwise tricky homogenization process. In this way, a double-emulsion is likely created. To turn the emulsion droplets into hard particles, more water can be added. Conclusions The phase behavior of the quaternary PLG/MO/NMP/water system has been examined, and three of its subsystems have been investigated in detail. The particular system was selected because it was a candidate for being a suitable system for creation of lipid-polymer hybrid particles. The phase behavior of the ternary PLG/NMP/water system was found to be in agreement with the phase behavior reported by Brodbeck et al.45 It is also in agreement with other polymer/ solvent/nonsolvent systems.43,47 The slope of the tie line shows that the fraction of NMP in the water-rich phase is larger than that in the PLG-rich phase. According to the Flory-Huggins analysis, this is due to the fact that water is a smaller molecule than PLG and that χPLG/NMP > χNMP/w. The PLG/MO/NMP system separates into two different phases, i.e., one rich in PLG and the other rich in MO. The phase behavior thus resembles the segregative phase behaviors recently described50 for systems consisting of two different polymers and one good solvent. The MO/NMP/water system shows a very rich phase behavior. It offers opportunities for creation of both liposomes and cubosomes. The former might be formed from the small two-phase region located between the two three-phase areas in the upper part of the phase diagram, and by spraying the sponge phase into excess water, the latter is formed. A one-phase region exists in the ternary PLG/MO/water system near the PLG corner. Even though the polymer is immiscible with both water and MO, it swells in a mixture of the two. Regarding the quaternary system (at 20 °C), the conclusions are that the L3, LR, and Q phases do not substantially incorporate PLG. Furthermore, PLG is in some cases in its glassy state, and in other cases, PLG extracts a mixture and forms a concentrated yellowish phase. We also conclude that when PLG swells, it can extract mixtures with different compositions. At high NMP/water ratios, it is mainly an NMP/water mixture that is extracted by PLG. Three of the ternary subsystems and some aspects of the quaternary system were modeled using the Flory-Huggins

12164 J. Phys. Chem. B, Vol. 105, No. 48, 2001 theory. The relative sizes of the molecules were based on the relative molecular weights, and the interaction parameters were systematically fitted from three of the ternary subsystems. The interaction parameters obtained describe qualitatively the strength of the interaction among the components, but should be treated as effective parameters adapted for the present system. The present study was undertaken to get some guidance for the formation of lipid-polymer ”composite” particles. Although particle formation, irrespective of technique, deals with nonequilibrium conditions, a phase diagram describing equilibrium conditions gives one the basis from which a process can be designed. Moreover, the equilibrium properties of the system may also give valuable information of the fate of hybrid particles, both during storage and in an aqueous environment in vitro and in vivo. There are a number of methods available for making microparticles such as emulsion techniques, spraying, and supercritical fluid methods. Here, spraying techniques, such as atomization, may be useful when starting from an NMP-rich liquid. Acknowledgment. We are grateful to Gerd Olofsson for technical support. This work was financed by the Council of Strategic Research (SSF) within the Colloid and Interface Technology (CIT) program. References and Notes (1) Jain, R. A.; Rhodes, C. T.; Railkar, A. M.; Malick, A. W.; Shah, N. H. Pharm. DeV. Technol. 2000, 5, 201-207. (2) Lu, L.; Peter, S. J.; Lyman, M. D.; Lai, H. L.; Leite, S. M.; Tamada, J. A.; Uyama, S.; Vacanti, J. P.; Langer, R.; Mikos, A. G. Biomaterials 2000, 21, 1837-1845. (3) Zhu, G.; Schwendeman, S. P. Pharm. Res. 2000, 17, 351-357. (4) Eliaz, R. E.; Kost, J. J. Biomed. Mater. Res. 2000, 50, 388-396. (5) Murakami, H.; Kobayashi, M.; Takeuchi, H.; Kawashima, Y. J. Controlled Release 2000, 67, 29-36. (6) Lamprecht, A.; Ubrich, N.; Perez, M. H. Int. J. Pharm. 2000, 196, 177-182. (7) Igartua, M.; Hernandez, R. M.; Esquisabel, A.; Gascon, A. R.; Calvo, M. B.; Pedraz, J. L. Int. J. Pharm. 1998, 169, 45-54. (8) Hernandez, R. M.; Igartua, M.; Gascon, A. R.; Calvo, M. B.; Pedraz, J. L. Eur. J. Drug Metab. Pharmacokinet. 1998, 23, 92-96. (9) Celebi, N.; Erden, N.; Turkyilmaz, A.; Int. J. Pharm. 1996, 136, 89-100. (10) Uchida, T.; Goto, S. Biol. Pharm. Bull. 1994, 17, 1272-1276. (11) Blanco-Prieto, M. J.; Besseghir, K.; Zerbe, O.; Andris, D.; Orsolini, P.; Heimgartner, F.; Merkle, H. P.; Gander, B. J. J. Controlled Release 2000, 67, 19-28. (12) O’Hara, P.; Hickey, A. J. Pharm. Res. 2000, 17, 955-961. (13) Prior, S.; Gamazo, C.; Irache, J. M.; Merkle, H. P.; Gander, B. Int. J. Pharm. 2000, 196, 115-125. (14) Baras, B.; Benoit, M.-A.; Gillard, J. J. Microencapsulation 2000, 17, 485-498. (15) Bain, D. F.; Munday, D. L.; Smith, A. J. Microencapsulation 1999, 16, 453-474. (16) Ghaderi, R.; Artursson, P.; Carlfors, J. Pharm. Res. 1999, 16, 676681. (17) Ghaderi, R.; Artursson, P.; Carlfors, J. Eur. J. Pharm. Sci. 2000, 10, 1-9.

Johansson et al. (18) Fu, K.; Pack, D. W.; Klibanov, A. M.; Langer, R. Pharm. Res. 2000, 17, 100-106. (19) Brunner, A.; Mader, K.; Gopferich, A. Pharm. Res. 1999, 16, 847853. (20) Hyde, S. T.; Andersson, S.; Ericsson, B.; Larsson, K. Z. Kristallogr. 1984, 168, 213-219. (21) Larsson, K. Nature 1983, 304, 664. (22) Engstro¨m, S. Lipid Technol. 1990, 2, 42-45. (23) Sadhale, Y.; Shah, J. C. Int. J. Pharm. 1999, 191, 65-74. (24) Sadhale, Y.; Shah, J. C. Int. J. Pharm. 1999, 191, 51-64. (25) Nylander, T.; Mattisson, C.; Razumas, V.; Miezis, Y.; Hakansson, B. Colloids Surf. 1996, 114, 311-320. (26) Razumas, V.; Kanapieniene, J.; Nylander, T. Anal. Chim. Acta 1994, 289, 155-162. (27) Wallin, R.; Engstro¨m, S.; Mandenius, C. F. Biocatalysis 1993, 8, 73-80. (28) Portmann, M.; Landau, E. M.; Luisi, P. L. J. Phys. Chem. 1991, 95, 8437-8440. (29) Ericsson, B.; Eriksson, P. O.; Lofro¨th, J. E.; Engstro¨m, S. Cubic phases as delivery systems for peptide drugs. In Polymeric Drugs and Drug DeliVery Systems; Dunn, R. L., Ottenbrite, R., Eds.; ACS Symposium Series 469; American Chemical Society: Washington DC, 1991. (30) Chiu, M. L.; Nollert, P.; Loewen, M. C.; Belrhali, H.; PebayPeyroula, E.; Rosenbusch, J.; Landau, E. Acta Crystallogr. 2000, 56, 781784. (31) Pebay-Peyroula, E.; Neutze, R.; Landau, E. M. Biochim. Biophys. Acta 2000, 1460, 119-132. (32) Ai, X.; Caffrey, M. Biophys. J. 2000, 79, 394-405. (33) Lindblom, G.; Quist, P. O. Curr. Opin. Colloid Interface Sci. 1998, 3, 499-508. (34) Rummel, G.; Hardmeyer, A.; Widmer, C.; Chiu, M.; Nollert, P.; Locher, K.; Pedruzzi, I.; Landau, E. M.; Rosenbusch, J. P. J. Struct. Biol. 1998, 121, 82-91. (35) Ostermeier, C.; Michel, H. Curr. Opin. Struct. Biol. 1997, 7, 697701. (36) Landau, E. M.; Luisi, P. L. J. Am. Chem. Soc. 1993, 115, 21022106. (37) Landau, E. M.; Rosenbusch, J. P. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 14532-14535. (38) Engstro¨m, S.; Alfons, K.; Rasmusson, M.; Ljusberg-Wahren, H. Prog. Colloid Polym. Sci. 1998, 108, 93-98. (39) Alfons, K.; Engstro¨m, S. J. Pharm. Sci. 1998, 87, 1527-1530. (40) Ekelund, K. Lipid Bilayers Versus Monolayers, Sponge Phases, and Skin Lipid Domains. Ph.D. Thesis, Lund University, Lund, Sweden, 2000. (41) Dunn, R.; English, J.; Cowsar, D.; Vanderbilt, D. Biodegradable in-situ forming implants and methods of producing the same. U.S. Patent, 4,938,763, July 3, 1990. (42) Langstone, M.; Flores, M.; Sankaram, M. B. Proc. Int. Symp. Controlled Release Bioact. Mater. 1999, 26, Poster 6459. (43) Flory, P. J. Principles of Polymer Chemistry, 12th ed.; Cornell University Press: New York, 1953. (44) Scott, R. L. J. Chem. Phys. 1949, 17, 279-284. (45) Brodbeck, K. J.; DesNoyer, J. R.; McHugh, A. J. J. Controlled Release 1999, 62, 333-344. (46) Arnauts, J.; Berghmans, H.; Koningsveld, R. Makromol. Chem. 1993, 194, 77-85. (47) Scott, R. J. Chem. Phys. 1949, 17, 268-279. (48) Piculell, L.; Lindman, B. AdV. Colloid Interface Sci. 1992, 41, 149178. (49) Jo¨nsson, B.; Lindman, B.; Holmberg, K.; Kronberg, B. Surfactants and Polymers in Aqueous Solution; John Wiley & Sons: New York, 1998. (50) Bergfeldt, K.; Piculell, L.; Linse, P. J. Phys. Chem. 1996, 100, 3680-3687.