Overcoming Semipermeable Barriers, Such as the ... - ACS Publications

Nov 21, 2003 - skin, the stratum corneum (see left panel in Figure 1). The latter is ... example, in a “Franz cell”.8 Nanoporous systems are neede...
0 downloads 0 Views 1MB Size
Langmuir 2003, 19, 10753-10763

10753

Overcoming Semipermeable Barriers, Such as the Skin, with Ultradeformable Mixed Lipid Vesicles, Transfersomes, Liposomes, or Mixed Lipid Micelles Gregor Cevc,*,† Andreas G. Scha¨tzlein,† Holger Richardsen, and Ulrich Vierl IDEA AG, Frankfurter Ring 193a, D-80807 Mu¨ nchen, Deutschland, EU Received September 20, 2002. In Final Form: August 29, 2003 We studied lipid aggregate penetration through nanoporous, semipermeable barriers by direct transport measurements in vitro and with the confocal laser scanning microscopy of the skin in vivo. We found that it is necessary to use mixed lipid bilayers with a low resistance to permeabilization and high flexibility to overcome narrow, normally confining pores. Partial molecular demixing in the stressed vesicle bilayer serves both purposes. An aggregate comprising a suitable blend of amphipats (Transfersome, Tfs) is, therefore, extremely deformable and easily crosses even very narrow pores (rTfs g 10rpore, and possibly more). Each such vesicle then behaves as a responsive, self-optimizing, nanorobotic transport device. The mixed micelles with identical components or the simple vesicles (liposomes) with a similar size as that of unusually deformable vesicles do not share this quality. Liposomes only traverse barriers when rlipos e 1.5rpore; they clog narrower pores, unless they get fragmented in/before the orifice. Mammalian skin is perforated by a very large number (g107 cm-2) of very narrow (rpore ∼ 0.3 nm) intercellular hydrophilic pores. These can be widened into the barrier-spanning, hydrophilic transcutaneous pathways (rpathway ∼ 20-30 nm) by ultradeformable vesicles. Mixed micelles or liposomes do not activate such pores because they are respectively too small or too undeformable (κlipos > 10κTfs) and large (2rlipos/nm g 45 . 20) for the purpose. The outer two-thirds of the skin barrier also contain fewer but wider openings (rpore g 3 µm), which encircle groups of cells in the stratum corneum. The resulting sparse, low-resistance intercluster pathway can accommodate various sufficiently small aggregates (ra e 2 µm), including liposomes and micelles. All the tested lipidic particles can, therefore, reach locally ∼60% of the skin barrier depth, on the average. Ultradeformable vesicles move through the skin most uniformly and to the greatest relative depth, however. Locally or near the skin surface the distribution of different lipid aggregates that penetrate a barrier can be similar.

1. Introduction The skin (cutis) is one of the best biological transport barriers. This is mainly due to the outermost layer of the skin, the stratum corneum (see left panel in Figure 1). The latter is 10-30-µm thick1 and made of stacks of dead or dying keratinized cells, so-called corneocytes. Corneocytes form laterally intercalated stacks that are organized in columns (middle panel in Figure 1). Each column is oriented perpendicular to the skin surface and contains a few dozen corneocytes “glued” together with specialized, very hydrophobic lipids.2 Intercellular lipids in the skin are mainly located in crystalline lipid multilamellae3 (topright panel in Figure 1) and are covalently bound to the corneocyte envelope membranes.4 This increases the skin tightness to small molecules, such as water, and mechanically strengthens the barrier on a short length scale. To keep the skin flexible on a longer scale, and to allow a good fit between the planar lipids and the imperfectly flat cell envelope membranes, lipid multilamellae are merged through regions of less well-organized lipids,2 sometimes in all directions. * Author to whom correspondence should be addressed. † Present address: Cancer Research Unit, Department of Medical Oncology, Garscube Estate, Switchback Rd., Glasgow G61 1BD, U.K. (1) Christophers, E. In The Skin of Vertebrates; Spearman, R. I. C.; Riley, P. A., Eds.; Academic Press: London, 1980; pp 137-139. (2) Wertz, P. W. In Phospholipids: Characterization, Metabolism and Novel Biological Applications; Cevc, G., Paltauf, F., Eds.; AOCS Press: Champaign, 1995; pp 139-158. (3) Bouwstra, J. A.; Goris, G. S.; van der Spek, J. A.; Bras, W. J. Invest. Dermatol. 1991, 97, 1005-1012. (4) Swartzendruber, D. C.; Wertz, P. W.; Madison, K. C.; Downing, D. T. J. Invest. Dermatol. 1987, 88, 709-713.

Three to seven adjacent corneocyte columns in the skin typically form a cluster of cells.8 Adjacent cell groups near the stratum corneum surface are separated by 4-6 µm wide valleys, or clefts.5 These groves are open and “channel-like” near the skin surface and narrower toward the stratum corneum center.6 Each cleft, at the bottom, is filled with relatively amorphous lipids, which probably do not match the quality of proper intercellular seals. Lipid packing is generally the densest in the central stratum corneum region.6 Here, very few, if any, intercorneocyte contacts can reach a width of more than approximately 20 nm.7 The skin barrier structure described in previous paragraphs suggests that different skin surrogates should be used for in vitro tests. To simulate in the simplest possible fashion the skin permeability barrier, a silastic membrane is often exploited; more trustworthy is the employment of excised skin or of its outermost part, the epidermis, for example, in a “Franz cell”.8 Nanoporous systems are needed to model the skin penetration barrier and to mimic, at least in the first approximation, the size and shape of pores shown in the right bottom panel of Figure 1. Lipid vesicles, liposomes, were first used as potential (trans)dermal drug carriers in the early 1980s.9-10 It (5) Grove, G. L.; Grove, M. J. In Noninvasive topography assessment in cutaneous investigation in health and disease. Noninvasive methods and instrumentation; Leveque, J.-L., Ed.; Marcel Dekker, Inc.: New York, 1989; pp 1-32. (6) Scha¨tzlein, A.; Cevc, G. Br. J. Dermatol. 1998, 138, 583-592. (7) Aguiella, V.; Kontturi, K.; Murtoma¨ki, L.; Ramı´rez, P. J. Controlled Release 1994, 32, 249-257. (8) Hadgraft, J., Guy, R. H., Eds. Transdermal Drug Delivery. Developmental Issues and Research Initiatives; Marcel Dekker: New York, 1989.

10.1021/la026585n CCC: $25.00 © 2003 American Chemical Society Published on Web 11/21/2003

10754

Langmuir, Vol. 19, No. 26, 2003

Cevc et al.

Figure 1. Mammalian skin barrier at different magnifications as reported in the literature: full thickness skin (left); epidermis with the stratum corneum and epidermal/dermal junction (middle); and two corneocytes in the stratum corneum with intercellular multilamellar lipid seals (upper right), as viewed with electron microscopy, and the corresponding intercellular hydrophilic pathways (lower right; indicated by white arrows), visualized with the CLSM.

became clear later that stable liposomes, such as those that are being used for parenteral drug delivery, cannot cross the stratum corneum barrier.11-14 It is now generally agreed13,15,16 that different lipid vesicles [conventional liposomes, fluid or elastic liposomes, shape-adaptable vesicles]17 differ widely in their ability to penetrate the skin. Only a certain kind of mixed lipid vesicles can actually enter the skin barrier.15,18-20 The related mixed lipid micelles cannot penetrate the skin.21,22 In this work, we address in detail the question of differential semipermeable barrier penetration by various lipid aggregates [including simple lipid vesicles (liposomes), mixed lipid micelles, and shape-adaptable mixed lipid vesicles (Transfersomes, developed at and trademarked by IDEA AG)]. We first present the results of measurements with the artificial semipermeable barriers that resemble the skin penetration barrier and provide a theoretical rationale for our observations. We then discuss the confocal laser scanning microscopy (CLSM) results obtained with various fluorescently labeled lipid aggregates on the skin of living mice and deduce the corresponding material penetration profiles in the organ. Last but not least, we discuss the significance of various types of penetrability pathways in the skin. Our data thus (9) Mezei, M. In Liposomes as Drug Carriers: Trends and Progress; Gregoriadis, G., Ed.; Wiley & Sons: New York, 1988; pp 663-677. (10) Weiner, N.; Martin, F.; Riaz, M. Drug Dev. Ind. Pharm. 1989, 15, 1523-1541. (11) Cevc, G.; Blume, G.; Scha¨tzlein, A.; Gebauer, D.; Paul, A. Adv. Drug Delivery Rev. 1996, 18, 349-378. (12) Cevc, G.; Scha¨tzlein, A.; Blume, G. J. Controlled Release 1996, 36, 3-16. (13) Schreier, H.; Bouwstra, J. J. Controlled Release 1994, 30, 1-15. (14) Cevc, G.; Gebauer, D.; Scha¨tzlein, A.; Blume, G.; Stieber, J. Biochim. Biophys. Acta 1998, 1368, 201-215. (15) Schubert, R.; Joos, M.; Deicher, M.; Magerle, R.; Lasch, J. Biochim. Biophys. Acta 1993, 1150, 162-164. (16) Cevc, G. Exp. Opinion Invest. Drugs 1997, 6, 1887-1937. (17) This trademark of IDEA describes aggregates with very flexible/ permeable membranes and extreme shape deformability. (18) Cevc, G.; Blume, G. Biochim. Biophys. Acta 1992, 1104, 226232. (19) Lasch, J.; Bouwstra, J. J. Liposome Res. 1995, 5, 543-569. (20) Van den Bergh, B. A. I.; Vroom, J.; Gerritsen, H.; Junginger, H.; Bouwstra, J. A. Biochim. Biophys. Acta 1999, 1461, 155-173. (21) van Kuijk-Meuwissen, M. E. M. J.; Mougin, L.; Junginger, H. E.; Bouwstra, J. A. J. Controlled Release 1998, 56, 189-196. (22) van Kuijk-Meuwissen, M. E.; Junginger, H. E.; Bouwstra, J. A. Biochim. Biophys. Acta 1998, 1371, 31-39.

complement and our paper constructively comments on the results of the electron microscopy and two-photon fluorescence microscopy investigations done by Bouwstra et al. with elastic vesicles.20 2. Definitions, Materials, and Methods Liposome in this work describes a vesicle optimized in terms of physical stability and drug-retention capability. Most conventional liposome formulations comprise semi-/synthetic phosphatidylcholine(s), often mixed with cholesterol, in a gel phase. Mixed micelle, as used in this work, has the same components as the corresponding mixed lipid vesicle but is typically much smaller than the latter, and more compact, as a result of the absence of an aqueous core. Relatively high surfactant concentration in a micelle, which solubilizes the mixed lipid bilayer, is responsible for this. Phosphatidylcholine/sodium cholate (SC) micelles were long believed to be disklike, with a thickness of approximately 4 nm23 and a diameter of 17 nm.24 Recent cryo-electron microscopy suggest that bile salts first solubilize such lipid bilayers in the form of threads,25 the final structure then being disklike.24 Under the experimental conditions used in our study, the phosphatidylcholine/bile salt mixed micelle, according to the dynamic light scattering, has a diameter around 20 nm. Penetration is the process in which a transported entity changes the resistance of a barrier, itself, or both to trespass. This typically occurs by enforced penetrant insertion into the barrier. Permeation describes diffusion-based transport through a semipermeable barrier. Sometimes, the term “penetration” is used, incorrectly, instead. Transfersome is a composite lipid aggregate, typically in vesicular form, with high shape adaptability and good colloidal stability.26 The aggregate consists of at least one bilayer-forming lipid (e.g., a phosphatidylcholine) mixed with a bilayer softening compound, for example, a surfactant (bile salts or highly soluble phospholipids are examples for the latter). An ultra-adaptable Transfersome must have the following characteristics: (1) a very flexible membrane; (2) the capability to avoid the stress of volume and area incompressibility by exchanging volume with the surroundings and the membrane lipids between the two halves (23) Lichtenberg, D.; Robson, R. J.; Dennis, E. A. Biochim. Biophys. Acta 1983, 737, 285-304. (24) Egelhaaf, S. U.; Schurtenberger, P. Phys. Rev. Lett. 1999, 82, 2804-2805. (25) Walter, A.; Vinson, P. K.; Kaplun, A.; Talmon, Y. Biophys. J. 1991, 60, 1315-1325. (26) Cevc, G. In Liposome Technology, 2nd ed.; Gregoriadis, G., Ed.; CRC Press: Boca Raton, FL, 1992; pp 1-36.

Ultradeformable Mixed Lipid Vesicles

Langmuir, Vol. 19, No. 26, 2003 10755

of a bilayer; and (3) high bilayer surface hydrophilicity. A combination of these qualities facilitates vesicle shape transformation and makes the aggregate unusually resistant to fusion and coalescence. Lipid Suspensions. Phospholipid suspensions (10 wt % in a phosphate buffer with pH ) 7.4) were labeled for fluorescence measurements with 0.2-1.5 mol % rhodamine-DHPE (1,2dihexadecanoyl-sn-glycero-3-phosphoethanolamine-N-Lissamine rhodamine B sulfonyl, triethylammonium salt) obtained from Molecular Probes (Eugene, OR). In some experiments, diphenylhexatriene, purchased from Sigma (Neu-Ulm, Germany) was used instead. Lipid vesicles were prepared from phospholipids [soybean phosphatidylcholine (dSPC), 99%; Nattermann Phospholipids, Cologne, or Lipoid KG, Ludwugshafen, both Germany] with or without the addition of membrane softener [typically dSC, p.a.; Merck, Darmstadt, Germany], as described in ref 14. For ellipsometric experiments, phosphatidylglycerol (dPG, 99%), prepared from SPC by Nattermann Phospholipids, was used. The specific phospholipid/biodetergent weight/weight ratios were 100:0 (SPC/SC) for liposomes, 87:13 (SPC/SC) and 146:53 (SPC/ SC) for the highly deformable mixed lipid vesicles with a relatively low and high cholate concentration, respectively, and 65:35 (SPC/ SC) for the mixed lipid micelles. The final vesicle size was between 100 and 150 nm and was achieved by sonication or extrusion. This was checked with photon correlation spectroscopy (90°, ALV5000 ALV-Laser Vertriebsgesellschaft, Langen, Germany) to an accuracy of 5 nm, with 2-nm reproducibility. Typically, the sonication time of 5-10 min (50% duty cycle with a microtip using a Branson Sonifier at room temperature) or two to three passes through an 80 nm pore filter were sufficient to achieve this goal. Aggregate Transport across Nanoporous, Semipermeable Barriers with Fixed Size Openings. Aggregate transport across nanoporous, semipermeable barriers with fixed size openings (“the skin surrogate”) was assessed essentially as outlined in refs 14 and 27. In brief, a known volume of a lipid suspension (1-10 mL) was applied on one side of a set of narrow pores with the specified size under different hydration (i.e., osmotic) or hydrostatic pressure to induce a dynamic adaptation of membrane flexibility, if possible. When using hydrostatic pressure, the flux was normalized using trans-barrier water vapor evaporation data, to allow for different barrier porosity and then to calculate relative barrier penetration ability: Prel ≡ Psuspension/ Pwater e 1. In this publication, we refer to a such quantity, in short, as penetrability. Experimentally, we first established that none of the tested suspensions can pass through 10-nm pores. We, therefore, mainly used trans-barrier 20- or 30-nm pore filters under a systematically varied hydrostatic pressure to study trans-barrier transport. This yielded penetrability data that are described with eq 5 given further in the text. In a separate test series, the motion of various size aggregates through a broader range of pores (20-100 nm diameter) was driven by constant trans-barrier hydration pressure to assess the effect of the vesicle/pore dimension mismatch. Transport sensitivity to such applied pressure is discussed in ref 27. We furthermore observed that the size of ultradeformable vesicles is practically unaffected by fine pore crossing. In contrast, conventional liposomes break if they are forced through comparably narrow pores. We and others confirmed this with the dynamic light scattering, as is described in greater detail in ref 28. Ellipsometric Bilayer Elasticity Measurements. When linearly polarized light is reflected on a surface, the reflectance has a minimum at an angle called the principal incidence angle (pseudo-Brewster angle). At this angle, the reflected light intensity increases as (vesicle) material is adsorbing onto the surface. Simultaneously, the reflected light polarization changes. This variation is typically expressed as the complex ratio F ≡ tan Ψ exp i∆. By measuring and analyzing the reflected light de/ polarization F in an ellipsometer, one gets two independent

parameters: ψ and ∆. The refractivity index of the surface adsorbed lipid bilayer is well-known to be 1.5.29 This allows the further mathematical transformation of the ψ and ∆ values within the framework of the vesicle adsorption and deformation model. A convenient such model was developed by Seifert and Lipowsky,30,31 also to calculate the shape of a vesicle pressed against a flat wall. This can be done as a function of the bilayer elastic bending energy κ and size rves in terms of a reduced, normalized adhesion potential w/w0 ) 2rves/κ.31 By integrating the various vesicle forms, as given numerically by Seifert (personal communication), we were able to calculate the relative bending elasticity of the bilayer δ ) κ(composition)/κ ≡ κ(d)/κ0 ) w0/w(composition). The proviso for this was to keep the interaction potential w and vesicle size rves experimentally constant in the following fashion. We used a laboratory-built, fully automatic null-ellipsometer29 to monitor the adsorption and to analyze the shape of anionic lipid bilayer vesicles attached electrostatically to a cationic lipid monolayer, as is described in the previous paragraph and in greater detail in ref 29. In short, we ensured that both the positively charged monolayer and the negatively charged vesicles have a constant and uniform charge density, to fix the interaction potential w. For the purpose, we used bilayers prepared from a mixture of neutral phosphatidylcholine and of a sufficient amount of negatively charged PG to replace the varied relative amount of the negatively charged cholate. The resulting degree of aggregate deformation under conditions of constant electrostatic attraction and changing bilayer elasticity was thus assessed. Aggregate Transport across the Skin in Vivo. Aggregate transport across the skin in vivo was studied as described in ref 6. In short, anaesthetised hairy NMRI (for biodistribution studies) and nude balb/c mice (for microscopic studies) received 2.5 µL of the test suspension on an area of approximately 1 cm2 on the upper back skin. If needed, this was done after having previously manually clipped hair. During the following 12 h, the mice (2025 g) were kept on a warming pad under general anaesthesia. The animals were then killed and the treated skin was carefully excised. The residual lipid suspension was removed with a dry cotton swab, taking great care to avoid contamination of the untreated sites with the label. The skin samples were finally trimmed to approximately 0.5 cm2 and mounted on a glass slide with the stratum corneum facing the coverslide. CLSM. CLSM was done as described in ref 5. In brief, a LSM 410 invert CLSM system (Zeiss, Heidelberg, Germany) with high numeric aperture lenses (Plan-Neofluar oil immersion lenses: 40/1.3, 63/1.4, 100/1.3) and a minimized pinhole size (e10% of the maximum) were used for optimum depth-resolution. The 543-nm line of the helium/neon laser was used for the excitation of rhodamine fluorescence labels. Dichroitic beam splitters and the long pass filter minimized the nonspecific signal and the crossover noise. The photomultiplier tube (PMT) amplification/ sensitivity setting was determined at short scan times (t ) 1.08 s for the 512 × 512 pixel image), with slightly overmodulated signal scaling. This overmodulation was compensated by the averaging over 4-16 frames or lines, respectively, for the final gray-scale images. Identical settings were used for all sample comparisons at the same depth, to prevent any bias due to the nonlinear gain/contrast modulation of the PMT response. In some cases, the sensitivity was increased in the lower part of the stratum corneum to gain more information. Data acquisition and basic image processing were done with the original Zeiss CLSM software (Zeiss, Oberkochen, Germany). More complex processing and visualization problems, as well as image-format conversions, were solved with the program packages AVS 5.1 (Advanced Visualizations System, Inc., Waltham, MA), IDL 3.9 (Research Systems, Inc., Boulder, CO), respectively, under Unix on Convex, Silicon Graphics, and SUN computers. To obtain a reliable picture of the skin penetration profiles for various aggregates, the results of 6-10 depth scans were averaged laterally, after excluding all the data stemming from the pilosebaceous units in the measured areas. Experimental data

(27) Cevc, G.; Gebauer, D. Biophys. J. 2003, 84, 1010-1024. (28) Cevc, G.; Richardsen, H.; Scha¨tzlein, A. Biochim. Biophys. Acta 2002, 1564, 21-30.

(29) Richardsen, H. Ph.D. Thesis, The University of Munich, 1996. (30) Seifert, U.; Lipowsky, R. Phys. Rev. A 1990, 42, 4768-4771. (31) Seifert, U. Adv. Phys. 1997, 46, 13-137.

10756

Langmuir, Vol. 19, No. 26, 2003

Figure 2. Penetrability of a nanoporous, semipermeable barrier to a suspension of ultradeformable vesicles (Transfersomes) or liposomes as a function of the relative aggregate/pore size ratio. The curve for highly deformable vesicles was calculated using the Darcy-Bruinsma model with a ) 1.8 and n ) 1000. Using a ) 2.2 and n ) 400 gives nearly the same result, however.

were fitted with a monoexponential curve, to simplify and to improve the clarity of the final graphic data representation. Characteristic microscopic pictures were selected to give an impression of the lateral fluorescence distribution at different positions in the skin.

3. Results and Discussion We investigated the relative ability of small mixed lipid micelles and of various lipid vesicles to traverse pores of different widths in semipermeable, synthetic membranes. We also inspected the penetration of similar lipid aggregates through an intact mammalian skin. Fluorescence measurements or gravimetry and the CLSM or biodistribution studies were used for this purpose, respectively. To rationalize our observations quantitatively, we furthermore compiled the first complete set of basic mathematical relationships that describe all the experimental data in adequate detail. None of the tested lipid aggregates passes through barriers with 10-15-nm pores (data not shown). This suggests that even small mixed lipid micelles or maximally deformed mixed lipid vesicles are unable to match such small dimensions at least in one direction. In contrast, small micelles and certain types of vesicles do overcome semipermeable barriers with g20-nm pores, as is illustrated in Figure 2. Specifically, the passage of highly deformable mixed lipid vesicles through a semipermeable barrier with narrow pores is nearly unrestricted as long as rTfs > 8rpore and probably more. In absolute terms, ultradeformable vesicles are little sensitive to the pore size limitation down to 2rpore ∼ 20 nm. Evidence for this is given in ref 24 and in Figure 2. Phosphatidylcholine vesicles with a fluid-phase bilayer cross a barrier if their size is up to 50% greater than the pore size. According to Figure 2, the flux of the liposome suspensions outside this range is more than 3 orders of magnitude lower than that measured for rlipos/rpore e 1.5. In absolute terms, liposomes require at least 30-nm-wide pores and a high driving pressure (see further discussion) to cross a barrier. The difference between the expected start of exclusion at rlipos/rpore g 1 and the observed exclusion range rlipos/rpore g 1.5 is arguably due to the finite

Cevc et al.

Figure 3. Permeability of a nanoporous barrier (2rpore = 20 nm) to water (+) or penetrability of a similar membrane to suspensions of mixed micelles (dots), to mixed lipid vesicles with a relatively high (upward-pointing triangle) or low (downward-pointing triangle) softener concentration, and to simple liposomes (circles) under the influence of an increasing trans-pore pressure difference. Curves give the results of data fits obtained for the mixed lipid vesicles within the framework of the Maxwell approximation defined in the main text body. The relative size for conventional and ultradeformable vesicles is rves/rpore > 5, the value for mixed micelles being rmic/rpore ∼ 0.7. In the low-pressure range, the flux of all tested vesicles is practically negligible. Standard errors are smaller than symbol sizes.

spread of the liposome and pore size distributions. These are on the order of 15% and 30%, respectively. Consistent with the finding, previous researchers in the field have found that the liposome size after extrusion through fine pores is generally similar to but up to 50% larger than the pore size.32-34 Figure 3 highlights the effects of pressure on simple or mixed lipid vesicle motion through nanoporous semipermeable barriers. The data suggest that the pressure dependency of barrier penetrability to simple phosphatidylcholine vesicles is essentially steplike. For the mixed lipid vesicles, the dependency is strongly nonlinear, however, and nearly sigmoidal: suspensions of highly deformable vesicles have unusual rheological characteristics resembling those of non-Newtonian fluids. We previously explored the phenomenon by using relatively wide pores (30 nm < 2rpore < 80 nm).11,14 Here, we extend the measurements to 2rpore = 20 nm, and generalize and rationalize all our observations with the following theoretical considerations. Fluid and Suspension Flow through Nanoporous, Semipermeable Barriers with Fixed Size Openings. The laminar flow of a viscous fluid through a semipermeable barrier, according to Darcy’s law, increases linearly with the flux driving force ∆f or with the corresponding pressure ∆p:

jfluid ) P∆p

(1)

The proportionality factor for a fixed pore size and geometry, which is the permeability, or better to say penetrability, value P is proportional to the pore density npores and to the average pore area in a given barrier. Factor (32) Mayer, L. D.; Hope, M. J.; Cullis, P. R. Biochim. Biophys. Acta 1986, 856, 161-168. (33) Hunter, D. G.; Frisken, B. J. Biophys. J. 1998, 74, 2996-3002. (34) Nayer, R.; Hope, M. J.; Cullis, P. R. Biochim. Biophys. Acta 1989, 986, 200-206.

Ultradeformable Mixed Lipid Vesicles

Langmuir, Vol. 19, No. 26, 2003 10757

P is also inversely proportional to the pore length L and to fluid viscosity η, due to the fluid friction in the pore proper:

Pfluid ) πrpore2npores/8ηL

(2)

Moving a suspension of particles through a barrier brings an extra energetic cost of pushing the particles into the pores, which changes Pfluid to Psuspension. In the first approximation, this is an activation process:

Psuspension ) Pfluid exp(-nG#/kBT) e Pfluid

(3)

the activation energy nG# defining the total cost of inserting n particles into a pore. Experimentally, we find that the Psuspension value is always below Pfluid by a constant factor on the order of 1, as long as the typical aggregate size is below the average pore diameter in a barrier. Specifically for the phosphatidylcholine-based, nonionic mixed micelles with rmic/rpore e 1, we have measured the fluid/suspension flux ratio to be Pmic/Pfluid ) 0.45 (cf. Figure 3). The maximum flux of the ultradeformable vesicles suspension is even somewhat lower, Pves/Pmic ∼ 0.75, indicative of the fact that even maximally stressed/deformed vesicles are larger than mixed micelles rmic < rTfs. Suspensions of particles that are not highly deformable and have a diameter greater than the pore diameter do not flow through a barrier. Formally speaking, their partition coefficient in a pore is 0 [exp(-G#/kBT) ) 0] and Darcy’s law (eq 1) and our generalization thereof, leading to eq 3, do not apply. Barrier penetrability to a given suspension is potentially lowered further by friction between the pore wall and the penetrating elongated vesicle. For a spherocylinder adjusted to the pore width, Bruinsma35 calculated the extra reduction of the effective penetrability (“permeability”) to be

Pves/Psuspension ) rpore2/[8 + 0.233nL*(L*/rpore)2] (4) where L* is the spherocylinder length and n is the number of tensionless vesicles per unit pore length.36 Relative barrier penetrability thus depends just on the pore filling fraction, reflected in the ratio Psuspension e ∼Pmic and on the vesicle and pore dimensions. The flow equation including the Bruinsma correction resembles Darcy’s law (eq 2) with a friction-dependent additional term in the denominator.35 Our generalization furthermore contains an activation-dependent, intrapore partitioning factor in the numerator of the penetrability function (eq 3). This allows the driving pressure dependency to be expressed in terms of the activation energy, which for quasi-spherical vesicles moving through a cylindrical pore corresponds to the energetic cost of the (quasi)sphere-to-spherocylinder transformation. Using Maxwell’s approximation, as is described later in the text, to express the energy transfer leads to

Pves(∆p) )

{

Pves,max 1 - erf

(x ) x p* + ∆p

]} (5)

p* 4p* exp π∆p ∆p

[

where erf is error function. Pves,max e Pfluid is the maximum barrier penetrability and can be measured directly using (35) Bruinsma, R. Physica A 1996, 234, 249-270. (36) For better practicability, we use slightly different notation: Pves/ Psuspension ) a2[8 + 0.0026n(rves2/a2 - 1)a]-1, in which the parameters a and rves measure the pore and vesicle radius, respectively.

a high driving pressure. p* is the only adjustable model parameter and describes the resistance of the tested system to trans-barrier transport, in pressure units. Specifically, p* gives the pressure at which an average vesicle energy is equal to the activation energy for pore penetration. The inverse value of p* thus defines the system’s characteristic penetration ability (penetrability). For pores of constant size and noninteracting vesicles, this quantity depends on merely the penetrant’s properties. The barrier penetrability, 1/p*, to a given suspension is then exclusively controlled by the aggregate’s shape adaptability. The vesicle shape transformation that determines the activation term exp(-nG#/kBT) can involve several energetic contributions. For example, any vesicle fragmentation is preceded by lipid bilayer rupture. The pressure needed for this is given by the Laplace equation

∆pmin ) 2γrupture[1/rpore - 1/rves] = 2γrupture/rpore (6) and needs to exceed approximately twice the bilayer rupture tension39 divided by the pore radius. Smaller pores are, therefore, relatively more difficult to overcome. The activation energy for the process is pore size independent, however: # ) γrupture2/KA,effNm Grupture

(7)

KA,eff and Nm describe an effective lateral bilayer compressibility modulus and the number of molecules in a bilayer per unit area, respectively. The activation energy is given by the work of the membrane rupture as long as bilayer bending is energetically inexpensive. The following consideration can be made to check whether such a bending counts energetically. Elastic deformation of an unconstrained vesicle is governed by the elastic energy of bilayer bending κ,40 neglecting the dynamic effects.41 The minimum pressure, or force, needed to induce an elastic vesicle deformation is, thus, proportional to κ and also to the relative vesicle size raised to a certain power,42

∆pmin,bend ∝ κrvesrpore-R ∼ κrpore-R, R ∼ 1

(8)

The activation energy for bilayer bending in the simplest approximation is proportional to the relative surface area of each vesicle43 (rpore/rves)2 and to the elastic membrane bending modulus κ:40 # ) ∆Gelast ∝ κ Gelast

(9)

# If Gelast . 0, the full activation energy is given by the sum # # Grupture + Gelast . High vesicle elasticity is poorly compatible with adequate vesicle stability. This problem is minimized by using stress-responsive bilayers. The best example for these are the mixed lipid bilayers with a stress-dependent

(37) Kwok, R.; Evans, E. Biophys. J. 1981, 35, 637-652. (38) Frisken, B. J.; Asman, C.; Patty, P. J. Langmuir 2000, 16, 928. (39) Evans, E.; Rawicz, W. Phys. Rev. Lett. 1997, 79, 2379-2382. (40) Helfrich, W. Z. Naturforsch. 1973, 28c, 693-703. (41) Evans, E.; Ludwig, F. J. Phys.: Condens. Matter 2000, 12, A315A320. (42) Gommper, G.; Kroll, D. M. Phys. Rev. E. 1995, 52, 4198-4208. (43) Cevc, G. In Handbook of Physics of Biological Systems; Lipowsky, R., Ed.; Elsevier Science: New York, 1994; Vol. 1, Chapter 9, pp 441466.

10758

Langmuir, Vol. 19, No. 26, 2003

Cevc et al.

γrupture(composition) f γrupture(local stress) ) δγrupture e γrupture, δ e 1 (11)

Figure 4. Lipid bilayer elasticity as a function of relative SC concentration, measured with ellipsometry on electrically deformed vesicles with similar sizes and different bilayer compositions. The increasing deformation shown in the lower panel reflects the increase in the bilayer flexibility, described by parameter δ˜ , also shown directly in inset (see eq 10).

composition and elasticity.44-46 The elastic bending energy of such mixed bilayers is formally written as

κ(composition) f κ(local stress) ) δ˜ κ e κ, δ˜ e 1

(10)

The elastic energy of deformation of a given bilayer part is, thus, lowered by the stress-dependent change in local composition of such a bilayer part. The importance of the overall bilayer composition is indirectly illustrated in Figure 3. More direct evidence is provided in Figure 4, which reveals an increase in the bilayer elasticity upon SC incorporation into mixed lipid vesicles. In earlier publications, we argued that components with a high affinity for strongly curved bilayers (edge-active substances) lower the energetic cost of membrane deformation.43,44 This is due to the accumulation of edge-active additives at the most deformed site in an aggregate. Using anomalous membrane deformability theory,45,46 we then concluded11 that the Pves value reflects the increased bilayer deformability, which is sensitive to the bilayer composition. The conclusion is in accord with the data presented in Figure 4. We now also propose that a dynamic redistribution of the edge-active components in a bilayer lowers the activation energy for the local bilayer poration; this is essential to dissipate the internal stress of vesicle deformation. More specifically, a strongly deformed vesicle has a much smaller surface-to-volume ratio than a spherical vesicle. Impermeable vesicles filled with an incompressible fluid, thus, have to break to adjust the mismatch. An alternative is to allow a more gentle exchange of water and bilayer ingredients through the lipid bilayer. This goal is reached by the spontaneous lateral demixing of bilayer components, such that it accumulates membrane softening ingredients in the most strongly curved parts of highly deformable vesicle bilayers and catalyzes trans-bilayer pore formation at such sites. As a result of the interdependence of the local stress and bilayer composition, the pressure needed for (local) membrane permeabilization is given by (44) Cevc, G. Crit. Rev. Adv. Drug Delivery Systems 1996, 18, 349378. (45) Seifert, U. Phys. Rev. Lett. 1993, 70, 1335-1338. (46) Leibler, S. J. Phys. (France) 1986, 47, 507-516.

as deduced from the Laplace equation. Vesicle tension upon rupture is measurable for the large lipid aggregates directly and can be assessed for the small vesicles indirectly.38 The results of published bilayer lysis tension measurements37,38 are given for simple vesicles in Figure 5 as open symbols. The corresponding results for the highly deformable mixed lipid vesicles are presented as closed symbols. We calculated the latter by analyzing the penetrability data given in Figure 3 under the assumption # that G# ) Grupture . It seems that all liposomes larger than 50 nm break at a comparable size-adjusted tension. It is also obvious that the rupture tension is one order of magnitude lower for the highly deformable vesicles, as is seen from the low values of δ given in the inset to Figure 5. Comparison of Experimental and Theoretical Results. Figure 2 combines our experimental and theoretical findings pertaining to vesicle motion through narrow pores. It demonstrates filtration of conventional, relatively large lipid bilayer vesicles (liposomes) by a nanoporous barrier and the lack of an appreciable relativesize effect for the highly deformable mixed lipid vesicles (Transfersomes). The finite slope of the flux versus relative vesicle size curve, observed in Figure 2, is due to friction experienced by vesicles moving through narrow pores, rather than being a sign of size exclusion. In this context, it is noteworthy that the initial flux versus relative vesicle size curve has practically identical slopes for liposomes and ultradeformable vesicles. Including all data measured for ultradeformable penetrants yields 0.21-0.25 for rves/rpore (R ) 0.919). This is not significantly different from the slope deduced for phosphatidylcholine liposomes: 0.14-0.332 for rves/rpore (R ) 0.746; rves/rpore e 1.6). The bulk viscosity dependency is, therefore, dominant in either case, as one would expect in the unconfined range. Friction acting on a vesicle inside a narrow pore adds relatively little to the viscous flow resistance felt by the suspending fluid. Several pieces of evidence support the conclusion: first, the fact that the Darcy-Bruinsma curve for the flux versus relative ultradeformable vesicle size dependency is nearly linear in the entire studied range (rTfs/rpore ∼ 8; cf. Figure 2) and, second, the similarity of the linear slopes for liposomes and Transfersomes in the range rves/rpore = e1, where the Darcy-Bruinsma equation does not apply. The fit-deduced model parameters, as given in the caption to Figure 2, are also of the right order of magnitude. In contrast, within the framework of the Darcy-Bruinsma model, a several orders of magnitude higher friction parameter would have to be used to describe the conventional liposome flux for 1.5 < rlipos/rpore < 1.9. This reflects the flux resistance originating in vesicle breaking and implies that eq 4 is not useful in such a range. A jump in the flux or penetrability versus pressure curves is indicative of bilayer rupture and vesicle fragmentation. A nonlinear but smooth pressure response is diagnostic of bilayer bending and possibly of dynamic bilayer component demixing under stress. We think that the sigmoidal pressure dependency of pore penetration ability, measured with ultradeformable vesicles suspension (closed symbols in Figure 3), mirrors such a demixing. Our choice of eq 5 for the phenomenological data description (curves in Figure 3) is motivated by the assumption.

Ultradeformable Mixed Lipid Vesicles

Figure 5. Apparent pressure of lipid bilayer rupture, or permeabilization, as a function of the stress-influencing pore size. The results for small and very large liposomes were calculated from the data given in refs 38 and 37 respectively, in the former case assuming that the starting vesicle size is much greater than the pore size and using eq 6. The results for ultradeformable vesicles were calculated from barrier penetrability data, comparable to those given in Figure 3 with eq # 9, assuming Gelast ) 0. Error bars give standard deviations of the mean.

An expanded version of eq 5 must be used to model the said steplike pressure dependency of liposome transport. Equation 11 can be used to gain more molecular perspective. This equation also quantitatively explains another previously mentioned10 but not thoroughly analyzed difference between ultradeformable and conventional liposomes. As a result of the fact that the value of the stress responsiveness parameters δ and δ˜ are unity at most, suitable mixed lipid bilayers yield locally and deform easier than conventional liposome membranes. Equations 10 and 11, thus, unify the concept of composition and stress-dependent bilayer vesicle adaptability by addressing bilayer bending and permeabilization in parallel. The stress-dependent rigidity factors are, therefore, also composition (c)-sensitive. The factors δ˜ (c) and δ(c), thus, scale the elastic bilayer energy, Gelast ) δ˜ (c)κ/2,46 as well as the bilayer poration energy, Grupture ) δ(c)γrupture/ KA,effNm, respectively, relative to the elastic and rupture energy of a simple, single-component bilayer. The former dominates in very narrow and the latter in relatively wide pores. Small δ˜ (c) and δ(c) values diminish the energetic cost of strong vesicle deformation. Highly flexible and permeable membranes, therefore, accommodate mechanical stress best. The resulting membrane-confined bodies are consequently most deformable, in a stress- and composition-dependent manner. The pressure needed to push a large, ultradeformable vesicle with rTfs/rpore g 2 through a semipermeable barrier decreases nonlinearly with the edge-active ingredient concentration in the bilayer. For the ultradeformable vesicles containing approximately 31 relative % of such an additive (compared to the biosurfactant concentration that causes complete aggregate solubilization), the transbarrier flux approaches the maximum value Pves,max for ∆p . 1 MPa (cf. Figure 3). Increasing the bilayer softener concentration to ∼75 relative % lowers the necessary pressure to 0.1 MPa. Going to >100 relative % creates mixed micelles and brings the required pressure value down to 0 MPa. This explains the size (in)dependency of

Langmuir, Vol. 19, No. 26, 2003 10759

the flux of such mixed lipid vesicles, or micelles, driven by the high trans-barrier hydration pressure (cf. Figure 2). Additional early data can be found in ref 43. Reanalysis of the suspension filtration data published by Frisken et al.38 confirms that the elasticity of the lipid bilayers starts to play an important role for the simple phosphatidylcholine vesicles when the pore diameter is close to or below 50 nm. This value is reasonably close to our limit estimate of 100 nm based on the combination of eqs 6 and 8. The difference is probably due to the unavoidable neglect of the term describing the effect of (unknown) starting vesicle size in Darcy’s law (eq 1) in ref 38. Simple liposomes cannot pass efficiently through 20 nm pores with the diameter comparable to that of the hydrophilic pathways in the skin which we believe is around 30 nm. Lipid vesicles rather break on than pass through the barriers perforated by such pores.32,33 Specifically, when 30-50-nm pores are used in combination with a rather high pressure above 3 MPa, liposomes are broken through such pores. Intermediate pressures in the range 0.5-3 MPa typically lead to pores clogging (see also refs 33 and 34). The minimum required driving pressure increases with the vesicle concentration in the suspension; vesicle crowding, pore clogging, and the increased local viscosity effects all may contribute to this. Frisken et al. independently came to a similar conclusion.38 Taken together, the data given in Figures 2 and 3 highlight the transport of mixed lipid vesicles and simple liposomes through fixed-pore barriers. Liposomes have a quasi steplike and ultradeformable vesicles a more sigmoidal pressure-response curve. The turning point in the sigmoidal curve for ultradeformable vesicles moves downward with increasing concentration of the edge-active bilayer ingredient. This happens in a nonlinear fashion until all lipid bilayers are solubilized in the form of mixed lipid micelles suspended in the fluid. At this point, the characteristic pressure parameter p* attains zero value. Suspensions of mixed lipid micelles behave essentially as simple fluids and differ only in the magnitude of the penetrability (“permeability”) prefactor from the result measured with water or a buffer (cf. eq 1 and Figure 3). Ultradeformable mixed lipid aggregates, Transfersome vesicles, reach a similar maximum penetrability value Pmic = Pves,max at such a driving pressure (∆p ∼ 1 MPa . p*) that enforces, and ensures, vesicle deformation into a spherocylinder fitting into the pore. Prerequisites for this are strong bilayer curving (at least in one dimension) and transient bilayer rupture (arguably at one of the tips of a spherocylinder, where bilayer softening components accumulate spontaneously).48 The two basic principles, by which changes in the local bilayer composition affect elastomechanical bilayer properties, are illustrated in Figures 4 and 5. Skin in Vivo as an Adaptable Semipermeable Barrier. The skin penetration by lipid aggregates resembles the suspension transport through an artificial semipermeable barrier but also shows essential differences. This is mainly due to the fact that hydrophilic pores in an unperturbed skin are just around 0.3 nm wide; this only allows the motion of very small chemical entities, such as water. Lipid aggregates, therefore, cannot diffuse (47) Pikal, M. J. Pharm. Res. 1990, 7, 118-126. (48) The possibility that vesicles first break into fragments of similar size as the mixed micelles and then reform into large aggregates after the pore is ruled out by the data published in ref 28: the maintenance of the original aggregate diameter 2rves,0 during pore crossing and its independence of the changing vesicle pore size ratio rves,final/rpore = rves,0/ rpore g 1 are both incompatible with such hypothesis.

10760

Langmuir, Vol. 19, No. 26, 2003

Figure 6. Comparison of different aggregate transports through semipermeable barriers in vitro and in vivo. The skin model used in vitro has open pores of a constant width; pathways through the skin are originally too narrow to allow aggregate passage but are arguably widened by ultradeformable vesicles.

across the skin. More generally, in the absence of a strong external driving force, no vesicle transport through the skin is observed (see ref 18 and unpublished data). Vesicle transport is achieved, however, by using suitable vesicle compositions without occlusion. Such vesicles then apparently widen passages in the skin and cross the latter. We estimate that this happens for rpore/rTfs g 1/10, or higher, and 2rpore > 20 nm. For conventional aggregates, the conditions rpore/rlipos g 2/3 and 2rpore > 30-40 nm must be fulfilled. These conditions are relatively easily met by ultradeformable mixed lipid vesicles and only in a few places on the skin by conventional liposomes, as is evident from Figure 6. Under nonocclusive conditions, a strong hydration gradient exists across the stratum corneum.49 Suitably designed mixed lipid vesicles respond to this gradient by (1) shape adjustment, (2) transcutaneous pathway formation, and (3) escaping dehydration through the skin barrier. In contrast, the conventional mixed lipid micelles and liposomes remain confined to the skin surface, whether they are exposed to the stress of partial dehydration or not. Evidence for this is given in the right panel of Figure 6. We surmise that this is due to the small number of sufficiently wide pathways through the skin fulfilling the requirement rpore g 0.7rves. Aggregate transport through the skin is 2-3 orders of magnitude below that through an artificial semipermeable barrier with comparably wide, fixed pores (2rpore ∼ 20 nm). The best explanation for this is that pores in the skin are affected, and probably activated, by the force exerted on the skin through the aggregates pushed against the barrier by a strong external pressure acting in parallel on numerous molecules in each vesicle with a large aggregation number. It is plausible that the much smaller mixed micelles fail to achieve this as a result of their too-low aggregation number, as is explained in greater detail in refs 16 and 27. The previously mentioned process resembles the channel opening in the stratum corneum by a transdermal electrical potential. This has been observed in a voltage range corresponding to an approximately 2 MPa pressure difference and was assessed by streaming potential measurements; the calculated final channel diameter was between 13 and 26 nm.7 This is a large value compared (49) Warner, R. R.; Myers, M. C.; Taylor, D. T. J. Invest. Dermatol. 1988, 90, 218-224.

Cevc et al.

Figure 7. Representative fluorescence distribution profiles perpendicular to nude mouse skin measured 12 h after an epicutaneous administration of fluorescently labeled mixed micelles (dotted line), liposomes (dashed line), and ultradeformable vesicles, Transfersomes (solid line). The average fluorescence intensity calculated by averaging the results from 10 individual pictures is an approximate measure of the total amount of material delivered to a given depth in the skin. Among the three aggregate types tested, only ultradeformable mixed lipid vesicles (Transfersomes) cross the skin barrier in an appreciable quantity.

to the starting pore diameter for cationic, neutral, and anionic permeants, which is 0.67 nm through 1.35-2.7 nm, respectively;47 the pore diameter is rather narrow, however, compared with the diameter of even small lipid aggregates. The narrowness of the pathways in the stratum corneum explains why liposomes and mixed micelles are mainly confined to the upper half of the stratum corneum (see the representative integrated intensity profiles, calculated by averaging 10 original CLSM pictures, in Figure 7). Such lipid aggregates are only occasionally found deep in the stratum corneum. This occurs at the rare sites at which the widest passages at the high end of the finite range of the transdermal pathway width distribution (seen in Figure 8) reach deep into the skin barrier. Numerical elimination of the fluorescence contributions from such cleftlike passages affects the integrated penetration profiles quantitatively (see also ref 6) but does not change the general conclusion. The maximum penetration depth for different pathways through the barrier is similar, starting with the local skin surface (cf. ref 6). This justifies the integration of the fluorescence intensity perpendicular to the skin up to a certain depth for simple liposomes, for ultradeformable mixed lipid vesicles, and for mixed lipid micelles. The resulting quasi-exponentially decaying profiles then highlight the various relative capabilities of different lipid aggregates to overcome the skin barrier. Such profiles are not useful for reading-off the maximum depth of penetration, however, as is briefly discussed in previous paragraph. The aggregate-derived fluorescence in the skin is often rather diffuse as a result of the imperfect skin permeability barrier. The skin permeation enhancers (surfactants) present in the mixed lipid micelles strengthen the effect. Either of the two mechanisms can facilitate the transport of amphipatic, low-molecular-weight fluorophores detached from a carrier through the skin during sample preparation or the test stage. Whereas ultradeformable vesicles also often contain bilayer-softening surfactants, such vesicles deliver lipophilic fluorescent labels into the skin differently. According to Figure 8, this not only involves the superficial but also the lower part of the stratum corneum. A small proportion of fluorescent ultradeformable vesicles can even reach the viable murine

Ultradeformable Mixed Lipid Vesicles

Langmuir, Vol. 19, No. 26, 2003 10761

Figure 8. Fluorescence distribution patterns in nude mice skin (approximately 130 × 130 µm2) after an epicutaneous administration of phosphatidylethanolamine-N-rhodamine labeled ultradeformable mixed lipid vesicles (A series), similarly labeled liposomes (B series), and mixed lipid micelles (C series), as visualized with the CLSM at different depths in the skin relative to the average surface plane. All aggregates enter the skin between quasi-hexagonal corneocytes, most often between intercellular lipid (multi)lamellae and membrane envelopes (see also Figures 1 and 9). Transfersomes reach greatest depth and also penetrate significantly between cells in the lower stratum corneum (which ends at approximately 8 µm). Small (4 × 25 nm2) micelles and liposomes, in contrast, transport the fluorescent label nearly exclusively into the upper half of the skin barrier, except in the regions between corneocyte clusters.

epidermis, as is indirectly seen in Figure 8, which reproduces a series of CLSM scans. The biodistribution data given in the right panel of Figure 6 complement the CLSM data shown in Figure 8. Together, these data support the view that simple liposomes or mixed lipid micelles, in contrast to ultradeformable mixed lipid vesicles, do not transport large lipophilic labels deep in the skin in an appreciable quantity. Skin Penetration Kinetics. The skin penetration kinetics by various suspensions were found to diverge after the initial redistribution process. The uppermost corneocytes involved in the desquamination process offer little or no penetration resistance to a nonocclusively applied penetrant. The spreading of various lipid aggregates on the outermost stratum corneum layers (stratum corneum disjunctum) is, therefore, rather fast and is finished within less than 1 h. At greater depth in the skin and at all later stages of penetration, the different deformation capabilities of the tested vesicles or their different sensitivities to the transport driving pressure become influential. Differential Skin Penetration through Various Pathways. The highest fluorescence intensity, which corresponds to the highest local carrier-mediated label

concentration, is always observed at the positions in the skin that correspond to the lowest density of intercellular lipid packing. The most brilliant fluorescence is, thus, detected in the wide intercellular space between clusters of corneocytes, especially close to the surface (cf. Figure 9). Inside each cluster of cells the penetration is most probable between the lipids in intercorneocyte space and one or both of their proximal corneocyte envelope(s) (see Figure 1). The intercluster pathway, near the skin surface, coincides with the 4-6 µm deep intercluster clefts that are seen in the reflectance as well as the fluorescence mode.6 A similar skin surface picture is obtained for the human skin50 and can be visualized with the scanning electron microscope.5 The alignment of, but not the space between, corneocyte clusters in the skin, therefore, varies more between the different body sites than between various animal species:51 intercluster gorges always show a comparable distribution of widths and depths. (50) Fesq, H.; Hutzler, P.; Richardsen, H.; Cevc, G.; Ring, J.; Abeck, D. Arch. Dermatol. Res. 1999, 291, 130. (51) Different cluster shapes and alignments are probably a result of different underlying muscular arrangements or predominant activities.

10762

Langmuir, Vol. 19, No. 26, 2003

Cevc et al.

Figure 9. Left: surface picture of living nude mouse skin, as measured with the CLSM reflectometry. Right: characteristic lateral fluorescence distribution in the intercluster (black) and intercorneocyte (grey) space of the skin, which have a relatively low and high penetration resistance, respectively. The skin was stained epicutaneously with the fluorescently labeled ultradeformable vesicles that were applied without occlusion. Field dimensions: approximately 200 × 200 µm2.

Ultradeformable vesicles uniquely activate and traverse the intercorneocyte pathway through the skin as a result of their extremely small resistance to permeabilization and deformation. This pathway is much narrower than the intercluster route and leads between intercellular lipid (multi)lamellae and adjacent corneocyte envelopes in each cluster of corneocytes (cf. Figure 9, right). The intercellular penetration pathway “encircles” the corneocyte edges and consequently appears as a family of parallel stripes that delineate individual cell boundaries when viewed perpendicular to the skin surface (see right bottom panel in Figure 1). As a result of its greater tightness, the intercorneocyte pathway is much less brilliant and more difficult to observe than the intracluster pathway. The intercorneocyte pathway is consequently less prominent in Figure 5 than in our previous publication,6 which paid special attention to this particular route through the skin. The differentiation between intercluster and intercorneocyte pathways is somewhat arbitrary because it relies on a semiquantitative criterion of pathway width. It is reasonable to assume that this width is mainly between 20 and 30 nm but at rare places can exceed 200 nm. It is impossible to determine these values directly by means of CLSM as a result of the resolution limit of optical microscopy. Only some of the intercluster spaces stained by ultradeformable vesicles are sufficiently bright and distinct to give an impression of a channel similar to pores in the skin models that we used. The right-bottom panel in Figure 1 (see white arrows) and Figure 6 in ref 6 give examples for this. The intercluster channels are sufficiently wide, however, to accommodate even nearly nondeformed particles with diameters greater than 100 nm or at least >50 nm. Because these passages only represent a fraction of the intercluster pathway, they do not contribute appreciably to the transport of fluorescent liposomes or of micelles across the skin barrier. The majority of transdermal pathways is merely accessible to ultradeformable vesicles and, therefore, deserves the denomination “virtual channels”6 or “virtual pathways”. From the data given in Figure 3, we conclude that the relative penetrability of a narrow pore to liposomes is 3-4 orders of magnitude lower than that of ultradeformable vesicles in the high-pressure region. Ultradeformable

aggregates cross 20 nm constrictions approximately 3 times less efficiently than water. The latter is known to evaporate through the skin at the rate of approximately 0.4 mg cm-2 h-1. We, thus, surmise that the upper, theoretical limit for the liposome flux through open pores in the skin is e10 ng cm-2 h-1. However, intercorneocyte passages are normally not open to liposomes. A more realistic estimate, based on the closed pores model for liposomes, is, therefore, e1 ng cm-2 h-1. For ultradeformable vesicles, the corresponding value is g100 µg cm-2 h-1. The apparent penetrability value for micelles is expected to be in the intermediate range but closer to the liposome value. Limited aggregate ability to fit into a narrow intercellular passage in the stratum corneum conjunctum correlates with the relative width of individual penetration pathways near the skin surface. Many lipid vesicles partition readily into the outer part of intercluster pathways, where the space is wider than the average vesicle size. Penetration along this path is generally easy and reaches further than penetration in the narrow intercorneocyte space. Liposomes at the depth of 4-6 µm are, thus, typically found in the intercluster space. Ultradeformable aggregates also have a good chance to penetrate to an appreciable depth, crossing other intercellular junctions, and to bring a significant amount of material into the viable epidermis, using the intercellular pathway. The comparison of the integrated fluorescence intensity derived for various kinds of aggregates at different depths in the skin supports the conclusion (cf. Figure 7). The work by other researchers is compatible with our observations. The CLSM data from the Leiden group of Bouwstra et al.,20,52,53 as well as reports from Zellmer et al.,54 indicate that inelastic lipid vesicles seldom, if ever, cross the stratum corneum. Rather than doing this, such nonocclusively applied vesicles disintegrate at the stratum corneum surface15 and release lipids to interact with the (52) Hofland, H. E. J.; Bouwstra, J. A.; Spies, F.; Bodde, H. E.; Nagelkerke, J. F.; Cullander, C.; Junginger, H. E. J. Liposome Res. 1995, 5, 241-264. (53) Van den Bergh, B. A. I.; Bouwstra, J. A.; Junginger, H. E.; Wertz, P. W. J. Controlled Release 1999, 62, 367-379. (54) Zellmer, S.; Pfeil, W.; Lasch, J. Biochim. Biophys. Acta 1995, 1237, 176-182.

Ultradeformable Mixed Lipid Vesicles

barrier.52,54 This is true for liposomes54 and fluid liposomes,20 as well as niosomes.52 Only few,55 if any,22 vesicles are found deep in the skin, unless great care is taken to optimize not only the bilayer elasticity but also the vesicle stability.28 Too much elasticity, that is, too-high surfactant concentration, is even detrimental to the skin barrier.53 Conclusions The capacity of various lipid aggregates to cross narrow pores of fixed size is diagnostic of an aggregate deformation capability. The aggregate transport through the skin also involves pore opening in the skin, typically to a diameter of approximately 20 nm. To separate the two phenomena, we first investigated suspension transport through a semipermeable barrier with pores of relevant sizes and, more generally, determined the size exclusion limits for different aggregates. Micelles smaller than the average pore size cross all tested semipermeable barriers. Such aggregates get stuck in a barrier, however, when the pore diameter is comparable to or smaller than the aggregate size. Lipid vesicles pass through a semipermeable, nanoporous barrier only if they are sufficiently shape-adaptable. For example, mixed lipid vesicles with a very adjustable bilayer can cross pores wider than approximately 20 nm. Conventional vesicles, with a less deformable bilayer, can only overcome pores wider than approximately 0.7 of vesicle diameter. Such conventional liposomes, thus, behave as nondeformable entities and must be fragmented, in an energetically expensive process, to trespass g25 nm pores. One of the prerequisites for a sufficient bilayer adaptability, and for a successful vesicle transport through narrow pores, is the relaxation of the changing volumeto-area constraint during pore entry. It stands to reason that such a relaxation requires a concurrent, but transient or local, bilayer permeabilization. The presence of edgeactive components facilitates such bilayer permeabilization and simultaneously increases bilayer flexibility; the latter must be high enough for an energetically inexpensive elastic bilayer deformation. Both phenomena may be reflected in a nonlinear pressure dependence of the apparent barrier penetrability. Partial lateral demixing, notably at most deformed bilayer sites, arguably supports bilayer permeabilization (55) Bouwstra, J. A.; Honeywell-Nguyen, P. L.; de Graaff, A.; Groenink, W.; Junginger, H. E. Proceedings of 5th International Conference Liposome Advances; London, 2001.

Langmuir, Vol. 19, No. 26, 2003 10763

and softening; both greatly facilitate penetrant shape transformations in (front of) a pore. Such demixing is also responsible for the stress-dependent vesicle adaptation, which mainly involves the lowering of bilayer rupture tension for large pores and the diminishment of the work needed for elastic bilayer deformation in the case of very small pores. The results of in vivo experiments suggest that most lipid aggregates can penetrate the stratum corneum. Vesicles of different adaptabilities reach different depths, however. The ultradeformable mixed lipid vesicles, the mixed lipid micelles, and simple liposomes can fill the outermost third of the skin barrier relatively easily. Some of such aggregates even get deeper into the skin. This happens nearly exclusively at the sites of widest intercellular openings that are broad enough to accommodate even large aggregates but are relatively rare. The vast majority of potential passages between the skin cells is only accessible to highly deformable aggregates. This is partly due to the capability of such extremely adaptable aggregates to adjust their form to a very tight surrounding. Equally important is the ultradeformable vesicles’ ability to open and keep open the passages between epidermal cells that are normally closed. This effectively means that the skin penetrability to ultradeformable aggregates is increased by the vesicle transport. The CLSM experiments reveal that such aggregates activate passages in the skin and move through the barrier using a variety of virtual pathways between cells and cellular clusters. In contrast, liposomes can only penetrate wide spaces in the skin, such as outer parts of intercluster junctions and pilosebaceous units. Micellar components facilitate the diffusion of molecules with a high partition coefficient in the skin but, according to CLSM, do not promote small aggregate motion across a cutaneous barrier. Acknowledgment. We wish to thank Dr. D. Gebauer for contributing the data given in Figure 2. We also thank the staff of Institut fu¨r Pathologie (CLSM) and the members of the central computer unit of the Gesellschaft fu¨r Strahlen und Umweltforschung, GSF, Neuherberg, as well as of the Leibnitz Rechenzentrum, Mu¨nchen (picture analysis), who all contributed to the success of this study. We are also thankful to Prof. U. Seifert for sharing with us his vesicle deformation results. LA026585N