Glycolipid based cubic nanoparticles: preparation and structural aspects

11
Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117 Glycolipid based cubic nanoparticles: preparation and structural aspects Thomas Abraham a,, Masakatsu Hato b , Mitsuhiro Hirai c a Department of Biochemistry, University of Alberta, 3–39 Medical Sciences Buildings, Edmonton, Alta., Canada T6G 2H7 b Nanotechnology Research Institute, AIST, Tsukuba Central 5, Tsukuba 305-8565, Japan c Department of Physics, Gunma University, 4-2 Aramaki, Maebashi 371-8510, Japan Received 1 December 2003; accepted 19 February 2004 Abstract Kinetically stable cubic colloidal particle dispersion was produced from a glycolipid using a novel preparation strategy based on the dialysis principle. The use of synchrotron small-angle X-ray diffraction (SSAXD) permitted the identification of exact structure of these dispersed particles in the colloidal state. Dynamic light scattering methods were used to obtain size and size distributions. A glycoside, 1-O-phytanyl--d-xyloside (-XP), that exhibits Pn3m cubic phase in an excess aqueous medium, was used as the lipid material. The dialysis technique includes controlled stirring action both inside and outside of the dialysis membrane tube. Initially, a mixed micellar system composed of -XP, n-octyl--d-glucopyranoside (-OG) and a triblock copolymer, Pluronic F127 (PL) was prepared in the aqueous medium. About 10 wt.% of PL to lipid weight was found to be sufficient to produce stable colloidal dispersions. The mean volume diameter of these colloidal particles was found to be in the range of 0.85 ± 0.05 m. The cubic phase structure of these colloidal particles is greatly depended on the final -OG concentration level in the system. Coexistence of Im3m and Pn3m cubic structures has been identified in these colloidal particles. This coexistence has the characteristics of Bonnet relation, which forms a compelling case for the infinite periodic minimal surface (IPMS) descriptions. These colloidal particles could restore pure Pn3m phase structure, but a longer dialysis time was needed. This work, in general, will open up new possibilities for membrane protein reconstitution and other relevant biological applications using colloidal cubic lipid particles. © 2004 Elsevier B.V. All rights reserved. Keywords: Lipids; Bicontinuous cubic phases; Membranes; Nanoparticles; X-ray; Bonnet transformation 1. Introduction Lipids that exhibit bicontinuous cubic phases have re- ceived considerable attention in biophysical and biomedi- cal fields, because of their unique role in membrane protein crystallization [1–4], their relationship to structures seen in biological membranes [5–7] and their potentials as drug car- riers [8,9]. Glycolipids, the amphiphiles that bear sugar head groups, have importance in biotechnology domains because of their biological functions such as molecular recognition as well as the stabilizing effects on protein functions and structures. Therefore, the lipid nanoassemblies that combine the unique Corresponding author. Tel.: +1-780-492-2412; fax: +1-780-492-0095. E-mail address: [email protected] (T. Abraham). properties of bicontinuous cubic phase and the biological functions of sugar groups can be a very promising new class of biomaterials. The lyotropic cubic liquid crystalline phases may be clas- sified into two distinct classes: bicontinuous cubic phases and micellar or discontinuous cubic phases. The interfaces of these phases may either have curvature towards or away from the water. The inverse forms of bicontinuous cubic phases, which have interfacial curvature towards the wa- ter are the ubiquitous forms found in the biologically rele- vant amphiphiles. In the bicontinuous phases, the lipid bi- layer divides the entire space into two intercrossing water channels. The bilayer is with zero mean curvature and no intersection. Two alternate structural representations have been proposed to describe the bicontinuous cubic phases [10], one in terms of rod-like elements and the other of folded surfaces (infinite periodic minimal surfaces (IPMS)), 0927-7765/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.colsurfb.2004.02.015

Transcript of Glycolipid based cubic nanoparticles: preparation and structural aspects

Page 1: Glycolipid based cubic nanoparticles: preparation and structural aspects

Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117

Glycolipid based cubic nanoparticles:preparation and structural aspects

Thomas Abrahama,∗, Masakatsu Hatob, Mitsuhiro Hiraic

a Department of Biochemistry, University of Alberta, 3–39 Medical Sciences Buildings, Edmonton, Alta., Canada T6G 2H7b Nanotechnology Research Institute, AIST, Tsukuba Central 5, Tsukuba 305-8565, Japan

c Department of Physics, Gunma University, 4-2 Aramaki, Maebashi 371-8510, Japan

Received 1 December 2003; accepted 19 February 2004

Abstract

Kinetically stable cubic colloidal particle dispersion was produced from a glycolipid using a novel preparation strategy based on thedialysis principle. The use of synchrotron small-angle X-ray diffraction (SSAXD) permitted the identification of exact structure of thesedispersed particles in the colloidal state. Dynamic light scattering methods were used to obtain size and size distributions. A glycoside,1-O-phytanyl-�-d-xyloside (�-XP), that exhibitsPn3m cubic phase in an excess aqueous medium, was used as the lipid material. The dialysistechnique includes controlled stirring action both inside and outside of the dialysis membrane tube. Initially, a mixed micellar system composedof �-XP, n-octyl-�-d-glucopyranoside (�-OG) and a triblock copolymer, Pluronic F127 (PL) was prepared in the aqueous medium. About10 wt.% of PL to lipid weight was found to be sufficient to produce stable colloidal dispersions. The mean volume diameter of these colloidalparticles was found to be in the range of 0.85± 0.05�m. The cubic phase structure of these colloidal particles is greatly depended on thefinal �-OG concentration level in the system. Coexistence ofIm3m andPn3m cubic structures has been identified in these colloidal particles.This coexistence has the characteristics of Bonnet relation, which forms a compelling case for the infinite periodic minimal surface (IPMS)descriptions. These colloidal particles could restore purePn3m phase structure, but a longer dialysis time was needed. This work, in general,will open up new possibilities for membrane protein reconstitution and other relevant biological applications using colloidal cubic lipidparticles.© 2004 Elsevier B.V. All rights reserved.

Keywords: Lipids; Bicontinuous cubic phases; Membranes; Nanoparticles; X-ray; Bonnet transformation

1. Introduction

Lipids that exhibit bicontinuous cubic phases have re-ceived considerable attention in biophysical and biomedi-cal fields, because of their unique role in membrane proteincrystallization[1–4], their relationship to structures seen inbiological membranes[5–7] and their potentials as drug car-riers [8,9].

Glycolipids, the amphiphiles that bear sugar head groups,have importance in biotechnology domains because of theirbiological functions such as molecular recognition as wellas the stabilizing effects on protein functions and structures.Therefore, the lipid nanoassemblies that combine the unique

∗ Corresponding author. Tel.:+1-780-492-2412;fax: +1-780-492-0095.

E-mail address: [email protected] (T. Abraham).

properties of bicontinuous cubic phase and the biologicalfunctions of sugar groups can be a very promising new classof biomaterials.

The lyotropic cubic liquid crystalline phases may be clas-sified into two distinct classes: bicontinuous cubic phasesand micellar or discontinuous cubic phases. The interfacesof these phases may either have curvature towards or awayfrom the water. The inverse forms of bicontinuous cubicphases, which have interfacial curvature towards the wa-ter are the ubiquitous forms found in the biologically rele-vant amphiphiles. In the bicontinuous phases, the lipid bi-layer divides the entire space into two intercrossing waterchannels. The bilayer is with zero mean curvature and nointersection. Two alternate structural representations havebeen proposed to describe the bicontinuous cubic phases[10], one in terms of rod-like elements and the other offolded surfaces (infinite periodic minimal surfaces (IPMS)),

0927-7765/$ – see front matter © 2004 Elsevier B.V. All rights reserved.doi:10.1016/j.colsurfb.2004.02.015

Page 2: Glycolipid based cubic nanoparticles: preparation and structural aspects

108 T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117

corresponding to Schoen’s skeletal graphs[29] and infinitelabyrinths, respectively. Lately, the structural classificationsbased on the IPMS are most commonly used to describe cu-bic phase lipid structures. Very recently, the representationsin terms of nodal surfaces have been used to describe thedynamic structure of cubic phases[11]. Three different in-verse bicontinuous cubic lipid phases are known at present,having space groupsPn3m, Ia3d, andIm3m, correspondingto the infinite periodic minimal surfaces, the diamond type(D-surface), the gyroid type (G-type) and the primitive type(P-surface), respectively[12].

To date, glyceryl monooleate (GMO) has been only thelipid of choice in formulating bicontinuous cubic phase forbiological applications[13]. GMO in equilibrium with wa-ter can be dispersed into particles, like liposomes. Landhand Larsson[14] produced cubic phase dispersions frommonoolein lipid. In this process, monoolein was combinedwith water at 40◦C for 24 h. High-pressure homogenizerwas employed to produce cubic particles, requiring highpressures and numerous passes before homogeneous dis-persions were produced. Finally, the particles are stabilizedagainst flocculation by polymer addition. Several other gooddescriptions of the preparation of aqueous dispersions ofcubic–water phases based on GMO have been provided inrecent years[15–17].

The main objective of this work is to produce cubic par-ticle dispersions from a glycolipid, an alternative to GMO,and to offer a new process for the preparation cubic parti-cles that facilitates the reconstitution of membrane proteinsin such cubic particles. In this article, we report a novelpreparation strategy for the production of cubic particle dis-persions from a glycolipid and the characteristics of theseparticles. A glycoside, 1-O-phytanyl-�-d-xyloside (�-XP),which exhibits Pn3m cubic phase in an excess aqueousmedium[18] was used as the lipid material. An amphiphilictriblock copolymer was used to provide dispersion stability.This preparation method offers the advantage of direct in-

Fig. 1. Chemical structures of: (A) 1-O-phytanyl-�-d-xyloside (�-XP); (B) n-octyl-�-d-glucopyranoside (OG); (C) Pluronic F127 (PL). Geometricalshapes of�-XP and�-OG are shown on the right side of relevant chemical structures, in which the symbolsv, a andlc denote the volume of hydrophobicchain, the head group area and the chain length, respectively.

clusion of colloidal stabilizers and sensitive materials suchas membrane proteins.

2. Materials and methods

2.1. Materials

The synthesis and the relevant physical properties of 1-O-phytanyl-�-d-xyloside are described elsewhere[18–20].n-Octyl-�-d-glucopyranoside (�-OG) was purchased fromSigma–Aldrich. Pluronic F127 (PL), a triblock copolymerwith a composition of [PEO]99–[PPO]67–[PEO]99 was ob-tained from BASF, USA. The structure of these chemicals isshown inFig. 1. All other chemicals, including the cellulosedialysis membrane tube (20 mm diameter), were purchasedfrom Wako Pure Chemical Industries (Osaka, Japan). ELGApurified water of resistivity higher than 18.2 M� cm wasused in all the experiments.

2.2. Dialysis process

A dialysis method (Fig. 2A), which evolved after severaltrail and error methods, was found to produce nanoparticledispersions. In this dialysis technique, controlled stirring wasmaintained both inside and outside of the dialysis membranetube using magnetic stirrers. It was found that the stirringaction inside of the dialysis membrane was crucial for thefast process and the production of cubic nanoparticle dis-persions. The dialysis experiments with aqueous mixtures(composed of�-XP, �-OG and PL) were conducted againstlarge volume of water, approximately 700 times of the aque-ous mixture inside of the dialysis tube, at 22±1◦C. A treateddialysis tube was used in these dialysis experiments. Thetreatment procedure follows. Initially, the dialysis tube wassoaked in 50% ethanol aqueous solution for about 1 h (twotimes). The soaking was then continued in 10 mM NaHCO3

Page 3: Glycolipid based cubic nanoparticles: preparation and structural aspects

T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117 109

Fig. 2. (A) Dialysis setup. Controlled stirring was maintained both inside and outside of the dialysis membrane tube using magnetic stirrers. (B) Stepsinvolved in the preparation of polymer stabilized cubic liquid crystalline nanoparticles. See text for details.

solution for about 1 h (two times). The dialysis tube wassoaked again in 1 mM EDTA solution for about 1 h, beforefinally washed with the distilled water. Washed dialysis tubewas stored in sodium azide (NaN3) solution at 4◦C. Thedialysis experimental conditions were kept identical in theall cases.

2.3. Preparation of bulk β-XP mixtures

A syringe assembly was used to prepare the bulk mix-tures of �-XP with other components such as�-OG, PLand water. This mechanical mixing device comprises twomicrosyringes with threaded terminations (250�l; ITO Mi-crosyringe: MS-GAN025), and a stainless steel coupler witha small threaded bore and a Teflon o-ring at the center of thebore. The syringes were connected by the coupler throughwhich the sample components subjected to pass during themixing process. A mechanical mixing device, somewhatsimilar to this syringe assembly, has been reported in theliterature previously[30]. This assembly ensures thoroughmixing and eliminates the possibility of loss of water orchange in compositions. Samples of total weight ca. 100 mg

were used. Appropriate quantities of�-XP and other com-ponents were weighed into the syringe. The sample-loadedsyringe was then kept in a chamber saturated with watervapor at room temperature for about 3 days for incubation.To ensure the homogeneity of the system, the samples weresubjected to thorough mixing repeatedly during incubation.

2.4. Methods for identifying the phase structures

For preliminary examinations, an Olympus BX51 polar-ized microscope (PM), was used to inspect the optical tex-ture of lipid samples.

The phase structures of colloidal�-XP particles were de-termined using a synchrotron small-angle X-ray diffractionspectrometer (SSAXD), attached to the synchrotron radia-tion source, High Energy Accelerator Research Organization(KEK), Tsukuba, Japan. The�-XP dispersions of concentra-tions ca. 2–3% (w/w) were directly used in SAXD measure-ments. Sample solutions were filled in to a mica-window cellwith 1.0 mm path length. A small suction pump was used forthe filling. The cells were placed in a cell holder controlledat 25± 0.1◦C. The X-ray wavelength used was 0.149 nm.

Page 4: Glycolipid based cubic nanoparticles: preparation and structural aspects

110 T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117

A camera length of 80 cm was used. The exposure time was480 s (8 min) for each sample. A one-dimensional positionsensitive proportional counter was used to detect the scatter-ing intensities. The counter consisted of 512 channels, andit was calibrated using collagen. The scattering data directlytransferred to a computer, where the scattering profile (in-tensity versus channel number) was displayed.

In the case of bulk�-XP mixtures, SAXD measure-ments were performed with Ni-filtered Cu K� radiation(λ = 0.154 nm) generated by a Rigaku RU-200 X-ray gen-erator (40 kV, 100 mA) with a double pinhole collimator(φ = 0.5 and 0.3 mm). The bulk mixtures, prepared asdescribed earlier, directly transferred into quartz capillary(Glas, Berlin, 1.5 mm outer diameter). The sample-loadedcapillaries were then incubated at 25◦C for at least 2 days.The SAXD measurements were carried out at 25± 0.1◦C.The samples were equilibrated at this temperature for∼1 hprior to X-ray exposure. The sample temperature was con-trolled with a Mettler FP82HT hot stage within an accuracyof (0.1◦C). The sample to film distance was 205 mm. Theexposure time was 50–60 min. To ensure the homogene-ity of each system, three to four sample-loaded capillarieswere prepared from the same system. Such samples gavethe identical and reproducible of X-ray diffraction pattern.SAXD measurements were also performed on pure water atsimilar conditions. Thus, the contribution of water towardsthe scattering profiles was subtracted.

In diffraction experiments the Bragg peaks of cubic struc-ture are observed at:

dhkl = a√h2 + l2 + k2

(1)

where dhkl is the Bragg spacing,a the unit cell size andh, k and l the Miller indices. The Miller indices dependon the lattice type and the symmetry elements of the cubicstructure. In well ordered cubic structures, thePn3m latticeexhibits characteristics reflections (peaks) at (1, 1, 0), (1, 1,1), (2, 0, 0), (1, 1, 0), (2, 1, 1), (2, 2, 0), (2, 2, 1), (3, 1, 0),(3, 1, 1), (2, 2, 2) and (3, 2, 1), while the peak positions ofIm3m occur at (1, 1, 0), (2, 0, 0), (2, 1, 1), (2, 2, 0), (3, 1,0), (3, 1, 1), (2, 2, 2), (3, 2, 1), (4, 0, 0), (4, 1, 1), (3, 3, 0)and (4, 2, 0). In the diffraction plots, the scattering vectorq(=4π sinθ/λ) versus the intensity is plotted.

2.5. Particle size distribution: dynamic light scatteringmethods

The particle size distribution of cubic particle dispersionswas determined using UPA 150 Particle Size Analyzer. Thisinstrument is capable of measuring particle size rangingfrom 3.0 to 7.0�m. Concentrated dispersions were diluted to1.0 wt.% in order to adjust the signal level. All measurementswere performed at 25◦C. UPA uses the controlled referencemethod of dynamic light scattering. In principle, the particlesize distribution is computed directly from measured fre-quency spectrum recovered from the Doppler-shifted scat-

tered light, assuming the particle geometry is spherical. Thedata inputs for the computation include the particle refrac-tive index (1.37) and particle density (0.97 g/cm3).

The average size and size distribution of mixed micelleswas measured by ELS-800TS electrophoretic light scatter-ing apparatus (Otsuka Electronics, Osaka) using a squarecell in a dynamic light scattering mode. Calibration measure-ments were performed with the uniformly sized (0.11 and0.204�m diameters; S.D. = 0.01�m) polystyrene beads(Polysciences Inc., Warrington, PA) suspended in water.

3. Results and analysis

3.1. Solubilization and dialysis ofβ-XP/β-OG/water system

1-O-Phytanyl−�-d-xyloside is practically insoluble inwater. The aqueous solubility estimated from that of ho-mologous lipids, i.e. those with shorter isoprenoid-typehydrophobic chains, was found to be in the order of 10−6 Mor less. A sugar surfactant,n-octyl-�-d-glucopyranoside,was used to solubilize�-XP. Fig. 3shows the solubilizationcurve of �-XP by �-OG. The minimum amount of�-OGrequired to fully solubilize�-XP is linearly related to theamount of�-XP present in the solution. Their concentrationrelationship can be expressed as:

C�-OG = 2.62C�-XP + 0.019 (2)

whereC�-OG andC�-XP denote moles of�-OG and�-XPpresent in 1000 g of solution, respectively. As shown inFig. 3, the solubilization occurs only above 19.0 mM of�-OG, which is close to the cmc of pure�-OG (∼20 mM).This suggests that the�-XP is solubilized by the formation

Fig. 3. Phase diagram of�-XP/OG/water mixture, OG vs.�-XP. Insertedpictures show representatives of clear mixed micellar solution and cloudytwo phase solution. The inserted equationC�-OG = 2.62C�-XP + 0.019presents the linear relationship betweenC�-OG andC�-XP, which denotemoles of OG and�-XP present in 1000 g of solution, respectively.

Page 5: Glycolipid based cubic nanoparticles: preparation and structural aspects

T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117 111

Fig. 4. SAXD profile obtained from a sample collected after∼5 h ofdialysis. Inset shows polarized microscopy picture of a sample collectedafter 5 h of dialysis. See text for details.

of mixed micelles from�-XP and�-OG molecules, in con-sistent with similar systems reported earlier by others[21].Using the dynamic light scattering measurements, the hy-drodynamic radius of these mixed micelles was determinedas 23 nm and their size distribution was found to be verysimilar to that of standard monodispersed polystyrene beads.

In a preliminary dialysis experiment, 1 g of 1.96 wt.%(0.046 M)�-XP mixed micelle system, was dialyzed as de-scribed inSection 2. Upon dialysis, the initial clear andtransparent solution became slightly turbid after about 1 h ofdialysis. As dialysis proceeded, the turbidity increased andthe samples started to exhibit birefringence. The samplescollected after∼5 h of dialysis consisted of dispersed spher-ical particles which are a lamellar phase (L�), as evidencedby the SAXD data (Fig. 4). The diffraction peaks are,d1 =4.4 nm,d2 = 2.2 nm and with the ratio ofd1/d2 = 0.50. Thelamellar repeat distance of 4.4 nm is close to that observedin a fully hydrated L� phase of the pure�-XP/water system.

Lipid in the form of lumps floated in the water after 12 hof dialysis. No birefringence regions are seen in these lumpsunder the polarized microscope. The lipid lumps thus pro-duced was examined in excess water using small-angle X-raydiffraction (curve A inFig. 5). At least seven well-definedpeaks (reflections) of varying intensities are seen. Clearly,these reflections or electron density fluctuations are thoseassociated with the ordered apolar–polar interfaces and CH3end-groups, which are located at the different planes of re-flection. The indexing of diffraction pattern confirms cubicPn3m phase and unit cell size (a) of 9.22 nm (Table 1). Thisdiffractogram is indistinguishable from thePn3m structureof a fully hydrated�-XP/water (32.4/67.6) binary system(curve B in Fig. 5). This bicontinuous cubic liquid crys-talline phase (Pn3m), as others, consists of twisted bilayermembrane. The center of the bilayer, whose locus is the

Fig. 5. SAXD profiles. (A) Lipid aggregate (lumps) in excess water(∼60 wt.% �-XP) obtained in the absence of the polymer stabilizer (PL)from the modified dialysis method. At least seven well-defined peaks(reflections) of varying intensities are seen. Indexing confirms cubicPn3mphase and unit cell size (a) of 9.22 nm (seeTable 1for diffraction data).(B) Fully hydrated pure�-XP/water (32.4 wt.%�-XP) binary system. Seetext for details.

minimal surface, and the interconnecting congruent waterchannels reside in the apolar and regions, respectively. Thebilayer can be considered as folded in a space accordingto an infinite periodic minimal surface, the D-surface, withzero mean curvature[7].

3.2. Making glycolipid cubosomes: dialysis ofβ-XP/OG/PL/water system

As described earlier, the dialysis of�-XP/�-OG mixedmicelle system can recover the originalPn3m cubic phasebut in a form of lipid lumps rather than in the dispersedcubic particles. Thus, an amphiphilic triblock copolymer,Pluronic 127, was used as a stabilizer to facilitate the forma-tion of cubic nanoparticles[14]. Fig. 2B features essential

Table 1Diffraction data (Fig. 5)

(h k l) (Pn3m) Lipid lumps obtainedafter dialysis, withoutPL (curve A in Fig. 5)

Bulk �-XP/watermixture, 32.4 wt.%�-XP (curve B inFig. 5)

d (nm) a (nm) d (nm) a (nm)

(1 1 0) 6.47 9.15 6.53 9.23(1 1 1) 5.32 9.22 5.32 9.22(2 0 0) 4.63 9.25 4.57 9.14(2 1 1) 3.74 9.16 3.72 9.11(2 2 0) 3.24 9.18 3.23 9.12(2 2 1) 3.08 9.24 3.05 9.14(2 2 2) 2.70 9.36 2.64 9.14

Average 9.22 9.16

Page 6: Glycolipid based cubic nanoparticles: preparation and structural aspects

112 T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117

Fig. 6. Particle size distribution in a colloidal dispersion measured usingUPA particle size analyzer. The colloidal system contained 2.8 wt.%�-XPand 10.1 wt.% polymer to lipid weight, and was obtained after 4 days ofdialysis. Particle sizes ranging from 0.17 to 3.3�m are present. See textfor details.

steps involved in the preparation of polymer stabilized cubiccolloidal dispersions.

The addition of PL to�-XP/�-OG mixed micelle system,up to 10 wt.% PL to�-XP weight, did not make signifi-cant difference in solubilization behavior. The dynamic lightscattering measurements show that the mixed micelle sizeremains nearly unaffected. The addition of PL facilitates theformation of dispersed cubic micro-/nanoparticles in excesswater, providing milky white dispersions. The low polymerconcentration, for instance 3.0 wt.% to�-XP, appears to im-part only the limited stability to the particles. Small numberof large aggregates, for example aggregates size larger than50�m, exist in such systems. The increase of PL contentto ∼10 wt.% prevents the aggregate formation and signifi-cantly enhances the dispersion stability.

We, here onwards, employed∼3 g of mixed micelle solu-tions containing 2.5±0.3 wt.%�-XP (0.058±0.007 M) andPL/�-XP weight fraction of∼10%, unless otherwise stated.Due to the increased mixed micelle solution volume and theincreased amount of�-OG to be removed,∼4 days dialysiswas required to recover the cubic lipid particles.

3.3. Particle size and distributions

Fig. 6 reports particle size distributions for the samplesobtained after 4 days of dialysis. At 10 wt.% polymer, par-ticle sizes ranging from 0.17 to 3.3�m are present. As ev-ident, the majority of particles are smaller than 1.0�m andthe mean volume diameter1 is determined as 0.85±0.05�m.

1 Mean diameter diameter= (∑

nivi × di)/(∑

nivi), whereni, vi, di

are number, volume and diameter ofith species.

Fig. 7. Polarized microscope pictures showing colloidal cubic particlesat a magnification level of 400 times. The colloidal system contained2.8 wt.%�-XP and 10.1 wt.% polymer to lipid weight, and was obtainedafter 4 days of dialysis. Larger particles in a range 1–4�m are seen. Asappears, the dispersed particles are isotropic and non-spherical.

The volume fraction of particles larger than 1.0�m is ap-proximately 30% of the total volume of particles.

Fig. 7 shows microscope scans of a colloidal dispersion,prepared at 10.1 wt.% PL to lipid weight, at a magnificationlevel of 400 times. Particle of size ranging from 1 to 4�mare clearly seen. Though the technique can give only the in-formation of larger particles (above∼1�m), two importantpoints become clear. Firstly, the dispersed particles are notspherical in shape, but appear to have crystal-like straightedges. Tumbling of their facets arising from the Brownianmotion of the particles was clearly observed under a micro-scope, a characteristic behavior of cubic phase dispersionswhich is not observed for a dispersed L� phase. Secondly,the dispersed particles appeared isotropic, which is again atypical characteristic of the cubic phase.

3.4. Dispersion stability

These dispersions are stable for at least several months atroom temperature. These dispersion exhibited slight cream-ing phenomena, i.e. a tendency of colloidal particles to floc-culate into multiple aggregates, after 2 weeks of storageat the room temperature. This creaming tendency increaseswith increasing�-XP concentration. The re-dispersion ofcreamy layer, however, was readily achieved by manualshaking of the solution. This implies that the creaming phe-nomena (flocculation) occurred at the shallow secondaryminimum position in the inter-particle potential, rather thanin the deep primary minimum where the colloidal parti-cles could be held tightly by van der Waals attractions.In this case, the presence of the adsorbed polymer (or an-other phase) in the form of a steric layer could increase the

Page 7: Glycolipid based cubic nanoparticles: preparation and structural aspects

T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117 113

potential barrier between primary and secondary minimaconsiderably. Thus, the significant flocculation of colloidalparticles into the primary minimum could be unlikely.

Another interesting feature of these cubic particle disper-sions is their stability in salt solutions. These dispersions arefound to be stable even in 300 mM 1:1 monvalent (NaCl)and 150 mM 1:2 divalent (MgCl2) salt solutions, signifyingagain the role of added triblock polymer (PL) as a steric sta-bilizer. Note that the steric layers such as the polyethyleneoxide layers (PEO), as in this case, are insensitive to theadded salt conditions[22]. Since the steric repulsive force isoriginated purely from the excluded monomer volume inter-actions, this form of force is unaffected by added salts. Thus,these dispersion systems exhibit good stability and can beconsidered stable in salt solutions. This has significance inmany biological applications.

3.5. Structure of the dispersed particles

Before we explore the structure of these polymer stabi-lized colloidal particles, it is our interest to see the effects ofthe polymer content (PL) on the structure of�-XP/PL/waterternary system. The bulk mixtures of�-XP/water/PL atthree different PL concentrations were prepared by the directmixing method. In these bulk mixtures, the water contentmaintained at 66.5 ± 0.5 wt.% and the PL content variedfrom 10–19.8% (w/w) with respect to�-XP weight.Fig. 8compares diffractograms of�-XP/water and�-XP/water/PLmixtures. Table 2 features the X-ray diffraction data ofthese mixtures. The indexing of diffraction pattern of the�-XP/water (32.4/67.6) mixture confirms a cubicPn3mphase with a unit cell size (a) of 9.16 nm (curve A). As fea-tured inFig. 8andTable 2, the�-XP/water/PL mixture, with10.0 wt.% polymer to lipid weight, exhibits cubicPn3mphase structure with a unit cell size (a) 9.15 nm (curve B),which is identical to that for the�-XP/water system. Furtherincrease of polymer content, i.e. 19.8 wt.% polymer to lipidweight, appears to induce slight changes in the diffraction

Table 2Diffraction data of bulk�-XP/water/PL mixtures (Fig. 8)

(h k l) (Pn3m) �-XP/W (32.4/67.6),0.0 wt.% PL (curve Ain Fig. 8)

�-XP/PL/W(30/2.99/67.0), 10.0 wt.%PL (curve B inFig. 8)

�-XP/PL/W(27.875/5.53/66.62),19.8 wt.% PL (curve Cin Fig. 8)

(h k l) (Im3m) �-XP/PL/W(27.875/5.53/66.62),19.8 wt.% PL (curve Cin Fig. 8)

d (nm) a (nm) d (nm) a (nm) d (nm) a (nm) d (nm) a (nm)

(1 1 0) 6.53 9.23 6.44 9.11 6.47 9.15 (1 1 0) 9.56 13.52(1 1 1) 5.32 9.22 5.29 9.16 5.29 9.16 (2 0 0) 6.47 12.94(2 0 0) 4.57 9.14 4.56 9.12 4.57 9.14 (2 1 1) 5.29 12.95(2 1 1) 3.72 9.11 3.75 9.19 3.71 9.09 (2 2 0) 4.57 12.92(2 2 0) 3.23 9.12 3.24 9.16 3.22 9.11 (3 1 0) Not shown –(2 2 1) 3.05 9.14 3.05 9.15 3.08 9.23 (2 2 2) 3.71 12.85(2 2 2) 2.64 9.14 2.65 9.18 6.47 9.15 (3 2 1) Not shown –

(4 0 0) 3.22 12.88(4 1 1) (3 3 0) 3.08 13.05

Average 9.16 9.15 9.15 13.01

Fig. 8. Effects of PL content on the structure of bulk�-XP/PL/water mix-tures. In these mixtures, the water content maintained 66.5 ± 0.5 wt.%.(A) Without polymer (PL); (B) 10.0 wt.% PL; (C) 19.8 wt.% PL. Insetshows the effect of polymer content on cubic unit cell size. X-ray diffrac-tion data are given inTable 2. Indexing of these profiles confirms cubicPn3m phase. See text for details.

curve (curve C). An additional weak reflection was seen inthis case, left to the (1 1 0) plane. A plausible explanation isthat it may result from the formation of additional new cubicphase structures. One may analyze this observation in termsof epitaxial relationships that could exist between differentcubic topologies. Note that epitaxial relationships mean therelationships of orientation and geometry between two adja-cent phases[31]. The peak positions of this diffraction pro-file (curve C), including the additional one, can be assignedto Im3m phase, which provide a unit cell size of 13.01 nm(Table 2). This shows that the overlapping of peak positions,corresponding toPn3m and Im3m phases, could occur atvarious planes, and that the epitaxial relationships couldexist at variousd-spacings:d110(Pn3m) = d200(Im3m),

Page 8: Glycolipid based cubic nanoparticles: preparation and structural aspects

114 T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117

d111(Pn3m) = d211(Im3m), d200(Pn3m) = d220(Im3m),d211(Pn3m) = d222(Im3m), d220(Pn3m) = d400(Im3m),d221(Pn3m) = d411(330)(Im3m). These planes might havethe characteristics of the epitaxial relations. Thus, the epi-taxially relatedPn3m andIm3m cubic phases might presentat this polymer content (19.8 wt.% polymer to lipid weight).

It is equally interesting to see the effects of�-OG in�-XP/water/PL mixtures, especially at their low mol%levels with respect to�-XP. The bulk mixtures of�-XP/water/PL/�-OG were prepared by the direct mixingmethod as described inSection 2. In these direct mixtures,the water content maintained at 66.0 ± 1.0 wt.%, the poly-mer (PL) content kept constant at 10± 0.04 wt.% (PL to�-XP weight) level, and the�-OG content varied from∼0.94 – 10.1 mol% with respect to�-XP. Fig. 9 comparesdiffractograms of these�-XP/water/PL/�-OG mixtures.Table 3features the X-ray diffraction data of these mixtures.At 0.94 mol%�-OG, the mixture exhibits the cubicPn3mphase structure, same as that of the�-XP/water binary mix-ture, but with a higher unit cell size (a) of 9.4 nm (curve A).The cubic phase structure remains the same when the�-OGconcentration increased to 2.4 and 3.5 mol%, but the unitcell size (a) increased to 9.5 nm (curve B) and 10.00 nm(curve C), respectively. A further increase of�-OG contentto 5.0 mol% alters the diffraction curve dramatically. Theindexing of this diffraction pattern confirmsIm3m cubicphase unit cell size (a) of 13.8 nm (curve D). According toan IPMS description, in this cubic phase, the center of thebilayer mimics the P-surface[7]. Such structural transitionfrom Pn3m to Im3m is known to be of low energy[7]. Thus,it is clear that the presence of relatively small quantities of�-OG could induce the structural changes in these systems.This can be explained in terms of a simple geometric con-sideration of the shape of the lipid and surfactant molecules[23]. �-XP is a practically water insoluble lipid with a bulkyphytanyl chain (Fig. 1A). Therefore,�-XP is considered tobe inverted truncated cone or wedge shaped, with a criticalpacking parameter (v/alc) more than one (Fig. 1A). Here,vis the volume of hydrophobic chain,a the head group area

Table 3Diffraction data of�-XP/PL/�-OG/water mixtures (Fig. 9)

(h k l) (Pn3m) �-XP/PL/�-OG/W0.94 mol%�-OG(curve A in Fig. 9)

�-XP/PL/�-OG/W2.4 mol%�-OG(curve B in Fig. 9)

�-XP/PL/�-OG/W3.5 mol%�-OG(curve C inFig. 9)

(h k l) (Im3m) �-XP/PL/�-OG/W5.0 mol%�-OG(curve D in Fig. 9)

d (nm) a (nm) d (nm) a (nm) d (nm) a (nm) d (nm) a (nm)

(1 1 0) 6.61 9.35 6.64 9.39 7.11 10.06 (1 1 0) 10.08 14.25(1 1 1) 5.42 9.38 5.43 9.40 5.86 10.14 (2 0 0) 7.23 14.45(2 0 0) 4.69 9.38 4.70 9.41 5.01 10.03 (2 1 1) 5.83 14.27(2 1 1) 3.81 9.33 4.02 9.85 4.05 9.91 (2 2 0) Not shown –(2 2 0) 3.31 9.35 3.35 9.48 3.49 9.87 (3 1 0) 4.41 13.95(2 2 1) 3.13 9.38 3.15 9.45 3.34 10.02 (2 2 2) 3.86 13.36(2 2 2) 2.70 9.36 2.73 9.46 2.89 9.99 (3 2 1) 3.73 13.95

(4 0 0) 3.44 13.75

Average 9.36 9.49 10.00 14.00

Fig. 9. Effects of �-OG content on the structure of bulk�-XP/PL/�-OG/water mixtures. In these mixtures, the water content maintained66.0± 1.0 wt.% and the PL content kept constant at 10± 0.04 wt.%. (A)0.94 mol%�-OG; (B) 2.4 mol%�-OG; (C) 3.5 wt.%�-OG; (D) 5.0 wt.%�-OG; (E) 10.1 wt.%�-OG. Inset shows the effect of�-OG content oncubic unit cell size. X-ray diffraction data are given inTable 3. See textfor details.

andlc the chain length. On the other hand,�-OG is a highlywater soluble cone shaped surfactant with a critical packingparameter (v/alc) far less than one (Fig. 1B). Therefore, theaddition of�-OG, to the�-XP systems havingPn3m phase,should impart a positive curvature to the lipid bilayer. Thisincrease in the curvature leads to the increase of the unitcell size and eventually results in the formation of cubicphase with a less negative curvature such as theIm3m cubicphase[7]. When the�-OG content increased to 10.1 mol%,the cubic phase ceases to exist and the mixture exhibitsliquid crystalline lamellar phase structure with a diffractionpeak atd1 = 4.4 nm (curve E).

Page 9: Glycolipid based cubic nanoparticles: preparation and structural aspects

T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117 115

Thus, it has been shown that the cubicPn3m phasestructure of the bulk�-XP/water binary system remainsessentially the same at 10.0 wt.% PL to�-XP. However,the presence of marginal amount of�-OG (i.e. 5.0 mol% of�-OG to �-XP) transforms the cubic structure fromPn3mto Im3m phase.

Let us now turn our attention to the structure of thedispersed particles produced by the dialysis method. Asan effort to determine the structure of the dispersed par-ticles, synchrotron SAXD measurements were performeddirectly on a colloidal dispersion, produced at 2.8 wt.%�-XP and 10.1 wt.% polymer to lipid weight, and ob-tained from 3.50 g of mixed micelle solution system after 4days of dialysis. This colloidal dispersion gives at least 10well-defined peaks (curve A inFig. 10). These peak posi-tions (reflections), however, do not conform to any particularbicontinuous cubic phase structures. Seemingly, the col-loidal dispersion, at this stage, exhibits more than one formof bicontinuous cubic phase structures. Clearly, longer dial-ysis time is required to restore thePn3m cubic phase. InFig. 10, we also show the scattering profile of a similarcolloidal dispersion, produced when the dialysis processcontinued for 6 days (curve B inFig. 10). Note that thiscolloidal dispersion was, produced at 2.3 wt.%�-XP and10.0 wt.% polymer to lipid weight, and was prepared from3.55 g mixed micelle solution system. In this case, at leastseven well-defined reflections are seen. Indexing of thesepeak positions confirmsPn3m phase structure with a unitcell size of about 9.4 nm (diffraction data inTable 4), whichis close to the one exhibited by�-XP/water mixture contain-ing 10% polymer and 0.94 mol%�-OG (curve A inFig. 9).Surprisingly, these peak positions, which correspond to thePn3m cubic phase (curve B inFig. 10), match perfectlywith some of those on curve A (indicated by dotted linesin Fig. 10). Attempt has been made assigning the extrapeak positions (indicated by downward arrows inFig. 10)to other bicontinuous cubic symmetries. Indexing of these

Table 4Diffraction data of colloidal dispersions (Fig. 10)

2.8 wt.% �-XP colloidal dispersion, 10.1 wt.% PL (curve A inFig. 10) 2.3 wt.% �-XP colloidal dispersion,10.0 wt.% PL (curve B inFig. 10)

Pn3m Im3m Pn3m

(h k l) d (nm) a (nm) (h k l) d (nm) a (nm) (h k l) d (nm) a (nm)

(1 1 0) 6.65 9.40 (1 1 0) 9.11 12.88 (1 1 0) 6.72 9.49(1 1 1) 5.46 9.46 (2 0 0) 6.28 12.57 (1 1 1) 5.46 9.46(2 0 0) 4.72 9.45 (2 1 1) 5.09 12.46 (2 0 0) 4.72 9.44(2 1 1) 3.85 9.44 (2 2 0) Not shown – (2 1 1) 3.85 9.44(2 2 0) 3.32 9.38 (3 1 0) Not shown – (2 2 0) 3.32 9.40(2 2 1) 3.12 9.35 (2 2 2) 3.59 12.44 (2 2 1) 3.14 9.42(2 2 2) Not shown – (3 2 1) Not shown – (2 2 2) 2.98 9.43

(4 0 0) Not shown –(4 1 1) (3 3 0) 2.90 12.28(4 2 0) 2.77 12.38

Average 9.40 12.50 9.44

Fig. 10. Synchrotron SAXD profiles showing the effects of�-OG con-tent (or dialysis time) on the structure of cubic particle dispersions. (A)Colloidal dispersion, 2.8 wt.%�-XP, containing 10.1 wt.% polymer, ob-tained after 4 days of dialysis. (B) Colloidal dispersion, 2.3 wt.% b-XP,containing 10.0 wt.% polymer to lipid weight, obtained when the dialysisprocess prolonged for 6 days. The diffraction data are shown inTable 4.See text for details.

extra peak positions confirms a cubicIm3m phase with aunit cell size (a) of 12.5 nm. Thus, it becomes clear thatthese colloidal particles could restorePn3m phase structure,same as that of�-XP/water binary mixture, from the mixedmicelles in the dialysis process, but a longer dialysis time isneeded. Also, it is evident that, during the process of dial-ysis (or during the removal process of�-OG from mixedmicelle system), the phase transformation could occur asfollow: mixed micelles→ liquid crystalline lamellar phase(L�) → mixture of cubicIm3m andPn3m phases→ cubicPn3m phase. This is consistent with the phase behavior

Page 10: Glycolipid based cubic nanoparticles: preparation and structural aspects

116 T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117

of �-XP/�-OG/PL/water bulk mixtures, where the phasetransformation sequence, with increasing�-OG content,determined as:Pn3m → Im3m → L�.

4. Discussion

The results in general illustrate the following. A dialy-sis method, which includes stirring action both inside andoutside of the dialysis tube, can successfully produce stablecolloidal dispersions of cubic phase particles at the submi-cron level. A triblock amphiphilic copolymer could providedispersion stability, possibly steric, to such systems. Thesecolloidal particles exhibit good stability in salt solutions. Thepresence of polymer stabilizer, even at 10 wt.% polymer tolipid weight, did not appear to bring any significant changesthe cubic phase structure or structural parameter of thesecolloidal lipid particles. These phase structural aspects are,however, greatly controlled by the amount of�-OG presentin these lipid colloidal systems.

As evidenced from the structural investigations, the pres-ence of small amounts of�-OG (∼5 mol%) could causesignificant changes in the phase structure of colloidal cubicparticles, i.e. the increase in the unit cell size of thePn3mphase and the transformation into theIm3m cubic phase. Itis also very interesting to notice the coexistence ofPn3m(D-surface) andIm3m (P-surface) cubic phases in these col-loidal particles (seeFig. 10). Note that according to the IPMSdescriptions,Pn3m andIm3m cubic phases correspond to D-and P-type minimal surfaces, respectively. There is a math-ematical relationship between the P- and D-type surfaces,defined by the so-called Bonnet transformation[7]. Thesesurfaces are Bonnet related, which means that they can betransformed each other by bending with constant Gaussiancurvature. According to the minimal surface theories, whenthe P and D cubic phases are in coexistence, the ratio be-tween the lattice parameters (a) of the two cubic phasesshould be 1.28[7,32]. This metric relation arises becausethe differential geometries of the P and D minimal surfacesare identical and are simply related by a mathematicalmapping (or by the Bonnet relation). In our case, the met-ric relation between the lattice parameters of coexistingIm3m (P-surface) andPn3m (D-surface) cubic phases isdetermined as:

aIm3m

aPn3m

= 12.5 nm

9.4 nm= 1.32 (3)

which is close to the theoretical value of 1.28, signifyingBonnet relation between these two phases. As mentionedearlier, the Bonnet relation means that the Gaussian curva-ture at all the points on the surface remains unchanged, al-though the surfaces may appear to be quite different after themapping. This transformation is driven by entropy and doesnot involve an enthalpy change (�H = 0). This means thatthe relative geometry and energy of these two cubic phasesare fixed as long as the monolayer interface is parallel to

the underlying minimal surfaces. Thus, our experimentallydetermined ratio in lattice parameter has the characteristicsof Bonnet relation, which essentially provides the convinc-ing evidence on the existence of cubic phases with minimalsurface (IPMS) structures in these colloidal particles.

In a closely related study, Gustafsson et al.[15] reportedthe structures of Pluronic 127 stabilized cubic phase aqueousdispersions (∼5 wt.%) produced from glyceryl monooleateusing the high-pressure homogenizer. These dispersed cu-bic particles gave significantly broadened SAXD reflections.They attributed this to the disorder in particle structure aswell as to the small particle size. In addition, the particlesstructure, for particles stabilized with 7.4 wt.% PL to lipid,identified asIm3m cubic phase which was different fromthe Pn3m cubic phase for the GMO/water binary mixturesin excess water. This phase transition, induced by the ad-dition of PL was in agreement with the phase diagram ofGMO/water/PL systems reported by Landh[24]. Gustafs-son et al.[15] suggested the penetration of polymer into thelipid layer as possible reason for such structural changes.

Contrary to these observations, we did not see any sig-nificant structural changes in the stabilized cubic particlesinduced by polymer even at 10 wt.% PL to lipid weight(Fig. 10). In addition, the dispersed cubic particles, producedby dialysis method, show well-defined SAXD reflections.This is most likely in part due to the slightly larger parti-cle size for�-XP particles than that of GMO cubosomes[15], and in part due to structural differences between theGMO and �-XP. Note that GMO is an unsaturated lipidthat exhibits less packing order compared to the saturatedlipid such as�-XP. Therefore, the penetration of polymersegments in the bilayer hydrophobic regions seems to bemuch easier in monoolein compared to that occurs in�-XP.

At the 10 wt.% polymer to lipid weight, although the inter-action between the bilayer and the polymer segments is notsufficient to induce the structural changes in thePn3m cubicphase, the cubic particle dispersions exhibit good stabilityover several months. The triblock copolymer, PL, is knownto stabilize various micro dispersions including liposomes[25]. In the case of liposomes, the experimental evidencessuch as the increase of the hydrodynamic radii and the re-duction of zeta potential are indications of some level of in-corporation of polymer segments in the bilayer[26]. Clearly,the stabilizing effect observed in our system emerges pos-sibly from the physical interaction between PPO blocksand apolar region of cubic assembly. The hydrophobic PPOsegment is expected to have collapsed conformation in poorsolvent like water. The linear chain size of PPO segment,i.e. R = bN1/3, whereb is the size of the repeating unit andN the number of repeating units, is estimated as 1.0 nm.On the other hand, the hydrophilic segment PEO has theexpanded conformation accordingR = bN0.6 and the chainsize in this case is estimated as 3.9 nm. The linear dimen-sion of triblock polymer PL consequently is 8.9 nm, whichis comparable to the hydrodynamic size of PL adsorbedonto perflurocarbon droplets[27]. The thickness of lipid

Page 11: Glycolipid based cubic nanoparticles: preparation and structural aspects

T. Abraham et al. / Colloids and Surfaces B: Biointerfaces 35 (2004) 107–117 117

bilayer, i.e. the twice of the thickness of phytanyl chains, isdetermined as 2.6 nm[28]. Considering these dimensionalaspects, the bilayer region of the dispersed particles canvery well accommodate hydrophobic PPO segments. Thus,the polymer chains can be considered to be anchored onthe hydrophobic region of the dispersed particles throughthe hydrophobic segments (PPO blocks) of PL. The hy-drophilic PEO blocks, however, would dangle in water thatoffers steric stability to the cubic particles. Such adsorptiontopology where the other two hydrophilic segments (PEO)extended into aqueous medium could provide steric stabilityto the particles, thus preventing the formation of colloidalaggregates.

5. Conclusions

A novel preparation strategy based on the dialysis prin-ciple was used to produce kinetically stable glycolipid cu-bic particle dispersions. A triblock amphiphilic copolymercould function as a steric stabilizer in such systems. Thispreparation method eliminates the detrimental effects ofhigh-energy input and offers the advantage of direct in-clusion of colloidal stabilizers and sensitive materials suchas membrane proteins. A 10 wt.% of a triblock copolymer(PL) to lipid weight was found to be sufficient to producethe stable cubic particle colloidal dispersions. The pres-ence of polymer stabilizer, at this concentration, did notappear to bring any apparent changes the cubic structureor structural parameter of these colloidal particles. The cu-bic phase structure of these colloidal particles is, however,greatly depended on the final�-OG concentration level inthe system. Bonnet relation that has been identified in thestructural determination provided convincing evidence forthe minimal surface description of the cubic phase struc-tures. The mean volume diameter of these colloidal parti-cles was found to be 0.85± 0.05�m. These colloidal dis-persions are found to be stable for several months at roomtemperature and show good stability in 1:1 and 1:2 saltsolutions.

Acknowledgements

T.A. thankful to the Japan Society for Promotion of Sci-ence for JSPS award and the full financial support for thisresearch work. The authors are grateful to Drs. D. Kato, D.Negishi, and Y. Abe at Lion Corporation, Process Devel-opment Research Center, for providing UPA 150 ParticleSize Analyzer facility for the particle size measurements.Synchrotron SAXD was performed under the approval of

the Photon Factory Program Advisory Committee (ProposalNos. 2001G359 and 2003G137).

References

[1] E.M. Landau, J.P. Rosenbusch, Proc. Natl. Acad. Sci. U.S.A. 93(1996) 14532.

[2] M. Caffrey, Curr. Opin. Struct. Biol. 10 (2000) 486.[3] B. De Kruijff, Nature 386 (1997) 129.[4] V. Gordeliy, J. Labahn, R. Moukhametzianov, R. Efremov, J. Granzin,

R. Schlesinger, G. Buldt, T. Savopol, A.J. Scheidig, J.P. Klare, M.Engelhard, Nature 419 (2002) 484.

[5] K. Larsson, J. Phys. Chem. 93 (1989) 7304.[6] T. Landh, Cubic cell membrane architectures, Thesis, Lund Univer-

sity, Sweden, 1996.[7] S. Hyde, A. Andersson, K. Larsson, Z. Blum, T. Landh, S. Lidin,

B.W. Ninham, The Language of Shape, first ed., Elsevier, New York,1997.

[8] C.J. Drummond, C. Fong, Curr. Opin. Colloid Interface Sci. 4 (2000)449.

[9] J.C. Shah, Y. Sadhale, D.M. Chilukuri, Adv. Drug Deliv. Rev. 47(2001) 229.

[10] V. Luzzati, H. Delacroix, A. Gulik, T. Gulik-Krzywicki, P. Mariani,R. Vargas, Curr. Top. Membr. 44 (1997) 3.

[11] M. Jacob, S. Andersson, Z. Kristallogr. 212 (1997) 486.[12] R.H. Templer, Curr. Opin. Colloid Interface Sci. 3 (1998) 255–263.[13] K. Larsson, Curr. Opin. Colliod Interface Sci. 5 (2000) 64.[14] T. Landh, K. Larsson, US Patent 5,531,925 (1993).[15] J.H. Gustafsson, H.L. Wahren, M. Almgren, K. Larsson, Langmuir

13 (1997) 6964.[16] P.T. Spicer, K.L. Hayden, M.L. Lynch, A. Ofori-Boateng, J.L. Burns,

Langmuir 17 (2001) 5748.[17] M. Nakano, A. Sugita, H. Matsuoka, T. Handa, Langmuir 17 (2001)

3917.[18] M. Hato, H. Minamikawa, R.A. Salkar, S. Matsutani, Langmuir 18

(2002) 3425.[19] H. Minamikawa, T. Murakami, M. Hato, Chem. Phys. Lipids 72

(1994) 111.[20] H. Minamikawa, M. Hato, Langmuir 13 (1997) 2564.[21] X. Ai, M. Caffrey, Biophys. J. 79 (2000) 394.[22] T.L. Kuhl, D.E. Leckband, D.D. Lasic, J.N. Israelachvili, Biophys.

J. 66 (1994) 1479.[23] J.N. Israelachvili, Intermolecular and Surface Forces, second ed.,

Academic Press, London, 1992.[24] T. Landh, J. Phys. Chem. 98 (1994) 8453.[25] M. Johnsson, N. Bergstrand, K. Edwards, J.J.R. Stålgren, Langmuir

17 (2001) 3902.[26] M. Silvander, Prog. Colloid Polym. Sci. 120 (2002) 25.[27] C. Washington, S.M. King, R.H. Heenan, J. Phy. Chem. 100 (1996)

7603.[28] M. Hato, H. Minamikawa, K. Tamada, T. Baba, Y. Tanabe, Adv.

Colloid Interface Sci. 80 (1999) 233.[29] A.H. Schoen, Infinite periodic minimal surfaces without

self-intersections, Technical Note D-5541, NASA, 1970.[30] A. Cheng, B. Hummel, H. Qiu, M. Caffrey, Chem. Phys. Lipids 95

(1998) 11–21.[31] J.M. Seddon, Biochim. Biophys. Acta 1031 (1990) 1–69.[32] S.T. Hyde, in: K. Holmberg (Ed.), Handbook of Applied Surface

and Colloid Chemistry, Wiley, 2001 (Chapter 16).