Reconstituting the Barrier Properties of a Water-tight Epithelial Membrane by Design of Leaflet-specific Liposomes* 210

To define aspects of lipid composition and bilayer asymmetry critical to barrier function, we examined the permeabilities of liposomes that model individual leaflets of the apical membrane of a barrier epithelium, Madin-Darby canine kidney type 1 cells. Using published lipid compositions we prepared exofacial liposomes containing phosphatidylcholine, sphingomyelin, glycosphingolipids, and cholesterol; and cytoplasmic liposomes containing phosphatidylethanolamine, phosphatidylserine, and cholesterol. The osmotic permeability of cytoplasmic liposomes to water (Pf ), solutes, and NH3 was 18–90-fold higher than for the exofacial liposomes (P f(ex) = 2.4 ± 0.4 × 10− 4 cm/s, P f(cy) = 4.4 ± 0.3 × 10− 3 cm/s;P glycerol(ex) = 2.5 ± 0.3 × 10− 8 cm/s,P glycerol(cy) = 2.2 ± 0.02 × 10− 6 cm/s; P NH3(ex) = 0.13 ± 0.4 × 10−4 cm/s,P NH3(cy) = 7.9 ± 1.0 × 10−3cm/s). By contrast, the apparent proton permeability of exofacial liposomes was 4-fold higher than cytoplasmic liposomes (P H+(ex) = 1.1 ± 0.1 × 10− 2 cm/s, P H+(cy) = 2.7 ± 0.6 × 10− 3 cm/s). By adding single leaflet permeabilities, we calculated a theoreticalPf for a Madin-Darby canine kidney apical membrane of 4.6 × 10− 4 cm/s, which compares favorably with experimentally determined values. In exofacial liposomes lacking glycosphingolipids or sphingomyelin, permeabilities were 2–7-fold higher, indicating that both species play a role in barrier function. Removal of cholesterol resulted in 40–280-fold increases in permeability. We conclude: 1) that we have reconstituted the biophysical properties of a barrier membrane, 2) that the barrier resides in the exofacial leaflet, 3) that both sphingomyelin and glycosphingolipids play a role in reducing membrane permeability but that there is an absolute requirement for cholesterol to mediate this effect, 4) that these results further validate the hypothesis that each leaflet offers an independent resistance to permeation, and 5) that proton permeation was enhanced by sphingolipid/cholesterol interactions.

and water through embedded proteins, are themselves often surprisingly permeable to a wide range of small molecules like water and urea. In contrast to this, there are certain cells and tissues, which, for functional reasons, exhibit exceedingly low permeabilities (1). These include the urinary bladder, which must maintain strikingly high concentration gradients for urea, NH 3 , protons, and CO 2 , when compared with the blood and surrounding interstitium (2); the stomach, the secretions of which have an extremely low pH; and the collecting duct in the kidney, which separates the urine being formed from the interstitium (1,3). Failure of the cell to maintain these barriers can lead to significant human pathology including renal tubular disorders, peptic ulcer disease, and cystitis (4,5).
The features of the lumenal surface that confer these remarkable properties are currently unclear, although it is now accepted that the barrier to permeation resides in the apical membrane of the superficial epithelial cells (3, 6 -10). Early physiological studies of mammalian bladder demonstrated an extremely high transepithelial resistance of up to 78,000 ohms cm 2 (11), and electrode impalement studies of the various epithelial cell layers demonstrated that the apical membrane of the most superficial layer was the location of the high electrical resistance (8,11,12). Creation of that barrier appears to depend on the cell's ability to synthesize an asymmetric bilayer, i.e. to apportion different lipid constituents to the outer or exofacial leaflet compared with the inner or cytoplasmic leaflet, and then to restrict the intermingling of those lipids (13). An earlier study from our laboratory demonstrated the functional consequences of losing bilayer asymmetry, when gastric apical vesicles, which possessed exceptionally low permeabilities to water, protons, and non-electrolytes, were made symmetrical by quantitative lipid extraction and reconstitution into liposomes. The newly formed liposomes, although made from identical apical membrane lipids, exhibited much higher permeabilities than the native membrane (10). Bilayer asymmetry is maintained by tight junctions between cells that not only isolate the basolateral and apical membranes, but also the exofacial leaflet from the cytoplasmic leaflet of the apical membrane (14,15). Thus, by means of lipid sorting in the trans-Golgi network and vectorial delivery of lipids to specific membrane domains, the cell can erect a membrane with low permeability. The precise structural features responsible for barrier function are unclear; however, it appears that certain combinations of lipids result in a more highly ordered membrane that has lower permeabilities for water and non-electrolytes. For example, sphingomyelin (SM) 1 and cholesterol have been shown to asso-* This work was supported by a research fellowship from the National Kidney Foundation and by National Institutes of Health Grant DK43955. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ciate closely due to van der Waals forces and hydrogen bonding, and these interactions induce tight packing in the bilayer (16). We, and others have previously shown that MDCK type 1 cells grown on permeable supports exhibit the permeability properties of a barrier epithelium (17). Because the lipid structure of the MDCK cell apical membrane is well defined, we aimed to reconstitute the low permeability properties of the MDCK type 1 cell, in an effort to understand the determinants that contribute significantly to this membrane's barrier properties (17). We also wished to further validate our earlier findings that each leaflet in a phospholipid bilayer offers an independent resistance to permeation (18,19). This hypothesis can be summarized by the following equation, in which the resistance offered by each leaflet is equivalent to the reciprocal of the permeability.
P AB is the permeability of a bilayer composed of leaflets A and B, P A is the permeability of leaflet A, and P B is the permeability of leaflet B. We have exploited this property in attempting to recreate the MDCK apical membrane: an asymmetric bilayer. Accordingly, we constructed liposomes with particular combinations of lipids in order to mimic the exofacial leaflet and the cytoplasmic leaflet of MDCK cells, and then measured their permeabilities to water, non-electrolytes, protons, and NH 3 . The cytoplasmic leaflets had dramatically higher permeabilities than the exofacial liposomes (20 -90-fold depending on the permeant). By the use of Equation 1, we derived a theoretical water permeability for the MDCK apical membrane that correlates extremely well with published permeabilities. Further experiments aimed at identifying the lipids responsible for barrier function implicate sphingolipid/cholesterol interactions as a key determinant in the formation of low permeability membranes.

EXPERIMENTAL PROCEDURES
Liposome Preparation-The following lipids from Avanti Polar Lipids Inc. (Alabaster, AL) were used to construct liposomes; bovine heart phosphatidylethanolamine (PE, catalog no. 830025), brain phosphatidylserine (PS, catalog no. 830032) bovine liver phosphatidylinositol (PI, catalog no. 830042), bovine heart phosphatidylcholine (PC, catalog no. 830052), cholesterol (catalog no. 700000), egg SM (catalog no. 860061), and brain cerebrosides (GSLs, catalog no. 131303). Natural lipids from mammalian tissues were chosen so that acyl chain heterogeneity and degree of unsaturation would reflect the likely chain composition of the canine cells we were modeling. In some experiments dipalmitoylphosphatidylcholine (DPPC) and Escherichia coli PE were used, and this is indicated in the text. Lipids were mixed by weight according to the proportions shown in Table I, dissolved in chloroform:methanol (2:1), aliquoted by volume into tubes, then dried down in a heating block set to 37°C under a stream of nitrogen. Lipids were then completely dried in an evacuated chamber for 2 h before storage at Ϫ20°C. On the day of liposome preparation, lipids were suspended in buffer containing 150 mM NaCl, 20 mM carboxyfluorescein (CF), 10 mM HEPES, pH 7.5. Lipids were vortexed for 2 min, heated to 37°C in a water bath for 10 min, vortexed for 2 min, probe sonicated three times (2 min each time) on ice at a medium power setting (with 2-min intervals) and then extruded through a 200 nm polycarbonate membrane. Liposomes were allowed to equilibrate on ice for 90 min, and then extravesicular CF was removed by passing vesicles over a Sephadex G50 column (Sigma). Vesicles were sized by quasi-elastic light scattering using a Nicomp model 270 submicron particle analyzer as described (3).
Water Permeability Measurements-Osmotic water permeability (P f ) was measured at 25 C as described (3,18,20). All other permeabilities were measured at 25 C also. Briefly, permeabilities were determined using a stopped-flow fluorimeter (SF.17 MV, Applied Photophysics, Leatherhead, United Kingdom) with a measurement dead-time of less than 1 ms. Liposomes containing 20 mM CF were rapidly mixed with an equal volume of an identical buffer that had 3 times the osmolality due to sucrose addition. The rate of water efflux from vesicles was measured as a function of vesicle shrinkage and CF fluorescence self-quenching. From parameters that included the initial rate of quenching, vesicle diameter, and applied osmotic gradient, P f was calculated using Math-Cad software (MathSoft Inc., Cambridge MA) (21).
Solute Permeability Measurements-Permeability measurements were performed as described using a stopped-flow fluorimeter (3,20,21). Briefly liposomes were equilibrated in buffer (500 mosmol/kg) containing 200 mM solute (glycerol, urea, formamide, or acetamide) for 30 -60 min at room temperature before the experiment was commenced, and then the liposomes were rapidly mixed with an equal volume of a solution with identical osmolality, containing 100 mM solute. Osmolalities of all solutions were confirmed and adjusted if necessary, by measuring freezing point depression on a Precision Instruments Osmette A osmometer. The concentration gradient results in solute efflux from liposomes followed by water efflux. Vesicle shrinkage can be monitored due to CF self-quenching. By use of parameters from the single exponential curve fit to the data, P solute was solved using MathCad software (3, 20 -22).
Proton Permeability-Apparent proton permeabilities were measured using pH-dependent quenching of fluorescence as described previously (3,10,19,20,22). Stopped-flow experiments were performed in which the liposomes were pretreated with 1 M valinomycin and then rapidly mixed with an identical buffer acidified to pH 6.50. Valinomycin, which was used to collapse any potential difference arising as a result of proton influx, did not appear to be necessary, as permeability measurements performed in its absence did not alter the results. Buffer capacity was determined on an SLM-Aminco 500C spectrofluorimeter by adding 10 mM acetate (final concentration) to liposomes as described (3). Fluorescence data from the stopped-flow device were fit to a single exponential curve, and fitting parameters were used to solve the following equation for P Hϩ .
J Hϩ ϭ ͑P Hϩ ͒͑SA͒͑⌬C͒ ϭ ͑⌬pH/t͒͑BCV) (Eq. 2) J Hϩ is the flux of protons, ⌬C is the initial difference in concentration of protons between the inside and the outside of the vesicle, ⌬pH is the change in pH when time equals , the time constant of the single exponential curve describing the initial change in fluorescence as a function of time, and BCV is the buffer capacity of an individual vesicle (3,10,20). NH 3 Permeability-NH 3 permeability was determined using stopped-flow fluorimetry by monitoring the pH-sensitive increase in fluorescence when vesicles equilibrated to pH 6.8 were rapidly mixed with the same buffer containing 20 mM NH 4 Cl as described (3,10,20). NH 3 in solution passes through the membrane and becomes protonated to NH 4 ϩ in the vesicle interior. By combining values for the rate of change of intravesicular pH, the final intravesicular pH and the buffer capacity (assessed in the same way as for proton permeability), P NH3 was calculated (20).
Statistics-Testing for significant differences was performed by the Bonferroni t test, which allows for multiple comparisons. Differences from the exofacial permeability were tested for, and a p Ͻ 0.05 was considered significant.

MDCK Apical Membrane Composition-
The predicted phospholipid composition of the MDCK type 1 apical membrane leaflets was derived from published analyses (15, 23) and by making a number of experimentally supported but simplifying assumptions about the asymmetric distribution of lipids in the bilayer. Explicitly, the assumptions were: (i) that the apical: basolateral membrane surface area ratio is 1:4 (24); (ii) that cholesterol and phospholipid in the apical membrane are present in a 1:1 ratio (13); (iii) that GSLs are distributed between apical and basolateral membranes in a 2:1 ratio (25); (iv) that half of the lipid moles are in each leaflet; (v) that all GSLs, SM, and PC in the apical membrane are in the exofacial leaflet (13); and (vi) that all of the PE, PI, and PS in the apical membrane are in the cytoplasmic leaflet (13). cial liposomes minus sphingomyelin; Exo-GSL, exofacial liposomes minus glycosphingolipid; Exo-Chol, exofacial liposomes minus cholesterol; S:U, saturated:unsaturated ratio; HPLC, high pressure liquid chromatography.
Hansson et al. (23) performed careful analyses of total lipids extracted from MDCK type 1 cells and tabulated in nanomoles/ filter the amount of each lipid present. We used these figures along with further data on the specific apical and basolateral membrane compositions found in Ref. 15. After allocating lipids on the basis of the published values and the above-listed assumptions, the remaining moles were balanced with cholesterol, for which no definitive information exists as to bilayer distribution. Resulting liposome compositions are shown in Table I.
Liposomes were prepared by suspending dried lipid mixtures in buffer containing 20 mM CF, heating to 37°C, vortexing extensively, probe sonicating, and then extruding through a 200-nm polycarbonate membrane using an Avanti mini-extruder (26). Phospholipid and cholesterol quantitation were performed by HPLC analysis (Avanti Polar Lipids Inc., Alabaster, AL) and confirmed that all the constituents in the cytoplasmic liposomes were present in the correct proportions. HPLC analyses of exofacial liposomes were performed by Avanti Polar Lipids Inc., and confirmed that PC, SM, and cholesterol were present in the correct relative ratios, but were unable to provide any information about the presence or concentration of GSLs. To confirm that GSLs were being incorporated at the correct molar ratio in the lipid mixture, liposomes were prepared from lipids in (chloroform:methanol), which had been spiked with BODIPY FL C 5 -glucocerebroside (Molecular Probes, Eugene, OR), [ 14 C]cholesterol (NEN Life Science Products), and [ 3 H]PC (NEN Life Science Products) prior to drying down under nitrogen. After the liposomes were prepared, they were washed by repeated centrifugation and then aliquots of the final liposome suspension were either scintillation counted or lysed with 0.025% (w/v) Triton X-100 and fluorimetry performed. From fluorescence standard curves and the known specific activity of the isotopes used, we confirmed that each component was present in the correct amount relative to the others.
Water Permeability of Exofacial and Cytoplasmic Leaflet Liposomes-Exofacial and cytoplasmic liposomes were tested for their permeability to water using stopped-flow fluorimetry. Vesicles with entrapped CF were rapidly mixed with an identical buffer to which sucrose (an impermeant solute) was added. Vesicles were thus rapidly exposed to an external solution osmolality 2-fold higher than inside. Water efflux in response to the imposed osmotic gradient causes vesicle shrinkage and CF self-quenching (Fig. 1C). Cytoplasmic liposomes exhibited a permeability to water that was approximately 20fold higher than that exhibited by exofacial liposomes (Fig. 1C). The rate of vesicle shrinkage displays single exponential decay kinetics and curves have been fitted to all stopped-flow profiles to show this. From parameters that include the initial rate of shrinkage, vesicle diameters, and applied osmotic gradient, the permeability coefficients were calculated. Fig. 1A shows the mean results from three separate liposome preparations. These experiments revealed an extremely low water permeability for exofacial liposomes of 2.4 Ϯ 0.4 ϫ 10 Ϫ4 cm/s, whereas cytoplasmic liposomes were 18.1 times more permeable with a P f ϭ 4.4 Ϯ 0.3 ϫ 10 Ϫ3 cm/s (p Ͻ 0.05).
Previous work from our laboratory has shown that the leaflets in a bilayer each offer independent resistances to water and solute permeation (18,19). From Equation 1 we can calculate a single leaflet permeability from the exofacial liposomes and a single leaflet permeability for the cytoplasmic liposomes (both symmetric bilayers) and combine them to arrive at a theoretical permeability for an MDCK apical membrane. Performing this calculation for water permeability yields a value of 4.6 ϫ 10 Ϫ4 cm/s, which is in good agreement with published permeabilities for MDCK type 1 apical membranes, which range from 1.8 ϫ 10 Ϫ4 cm/s to 10 ϫ 10 Ϫ4 cm/s (17,(27)(28)(29).
Water Permeability of Modified Exofacial Liposomes-One of the more striking features of the bilayer asymmetry observed in certain barrier membranes is the highly concentrated localization of sphingolipids in the exofacial leaflet (13). We wished to investigate whether GSLs and/or SM contribute significantly to the observed membrane impermeability and therefore carried out stopped-flow experiments on liposomes that lacked these components. We were also interested to explore the role of cholesterol in reducing membrane permeability. Accordingly, we prepared exofacial liposomes minus SM (Exo-SM), minus GSLs (Exo-GSLs), and minus cholesterol (Exo-Chol). In each case the missing component was replaced on a mol% basis with PC. In Fig. 1D stopped-flow profiles for each of these modifications is shown. When compared with the rate of vesicle shrinkage exhibited by complete exofacial liposomes, removal of either of the sphingolipids resulted in increased water permeability. Removal of cholesterol resulted in extremely fast water fluxes. The combined results from three experiments are shown in Fig. 1B. Removal of GSLs or SM resulted in 2.1 times higher or 2.4 times higher P f , respectively (p Ͻ 0.05). Thus, both SM and GSLs contribute to the low permeability properties of this membrane. Unsurprisingly, the removal of cholesterol also increased the permeability of this membrane. The magnitude of the increase is, however, remarkable. Water permeability was increased 37-fold even in the presence of a full complement of sphingolipids.
Effect of Changing Acyl Chain Composition on Water Permeability-As we had limited control over the acyl chain composition of these membranes, we chose lipids purified from mammalian cells. We anticipated this would yield bilayers with an "averaged" double bond and chain length heterogeneity (see Supplemental Material available in the on-line version of this article). To examine further the importance of acyl chain composition on the permeation behavior of these membranes, we constructed exofacial liposomes in which we replaced bovine heart PC (predominantly 16:0, 18:1, and 18:2 acylation) with DPPC (16:0 -16:0), and cytoplasmic liposomes that had bovine heart PE (predominantly 18:0 -20:4) with E. coli PE (predominantly 16:0 -18:1). The other lipids and their proportions were kept the same. Exofacial liposomes made with DPPC (DPPC/ Exo) showed a reduction in P f from 2.4 ϫ 10 Ϫ4 cm/s to 1.6 ϫ 10 Ϫ4 cm/s, whereas the cytoplasmic liposomes also exhibited a reduction in water permeability from 4.4 ϫ 10 Ϫ3 to 1.5 ϫ 10 Ϫ3 cm/s (Fig. 2B). These membranes exhibited significantly lower water permeability due to a reduction in the number of double bonds (p Ͻ 0.05). The influence of replacing bovine heart PC with DPPC is particularly interesting, given that this species represents only 9% of the total lipid and yet effects a one-third reduction in P f (Fig. 2A). We further tested these DPPC-exofacial liposomes by removing GSLs and replacing them on a mole for mole basis with DPPC (DPPC/Exo-GSLs in Fig. 2, A and C). There was no change in permeability ( Fig. 2A). As DPPC has a high melting point (T m ϳ41 C) and is structurally analogous to SM, the choice of artificial PC-lipid effectively masks the true effect of removing GSLs from the membrane. When bovine heart PE was replaced with E. coli PE, the result was a twothirds reduction in permeability (Fig. 2, B and D). As PE constitutes 40% of the lipid in this membrane, a reduction in double bond complement in PE would be anticipated to have a more profound effect. These data point to the need to carefully consider the appropriateness of lipids before conducting membrane modeling experiments. The acyl chain content can dramatically affect a membrane's biophysical properties. Solute Permeability of Exofacial and Cytoplasmic Leaflet Liposomes-The permeability of inner and outer leaflet liposomes to a range of solutes was examined. By abruptly exposing liposomes pre-loaded with 200 mM solute to an external buffer solution containing 150 mM solute, solute efflux occurs and vesicles shrink leading to CF self-quenching. In Figs. 3-6, it can be seen that the permeability of cytoplasmic liposomes to formamide (Fig. 3, A and C), acetamide (Fig. 4, A and C), urea (Fig. 5, A and C), and glycerol (Fig. 6, A and C) all exhibit much higher permeabilities than the exofacial liposomes (p Ͻ 0.05 for all). The permeability of cytoplasmic liposomes to formamide, acetamide, urea and glycerol were 22-, 35-, 52-, and 91-fold greater, respectively. These differences are greater than that observed for water permeability and demonstrate a rank order that equates with increasing molecular weight of the permeant. The data show remarkably low permeabilities associated with the exofacial leaflet and indicate that barrier function to solutes can clearly be accounted for by the specialized lipid composition of a single leaflet.
Solute Permeability of Modified Exofacial Leaflet Liposomes-When GSLs, SM, and cholesterol were omitted from the exofacial membranes, the stopped-flow experiments indicated a higher rate of permeation for all of the solutes occurred. This was especially true when cholesterol was absent and rates were much greater (panel D in Figs. [3][4][5][6]. The results of experiments performed in triplicate are shown in Figs. 3-6 (panel B). The increases in exofacial permeability when GSLs and SM were removed were fairly constant and did not appear to be solute-dependent. For example removal of GSLs resulted in 3.4 times higher (formamide, Fig. 3B), 6.1 times higher (acetamide, Fig. 4B), 4.4 times higher (urea, Fig. 5B), and 5.2 times higher FIG. 1. Water permeability of exofacial, cytoplasmic, and modified exofacial liposomes. Liposomes were prepared as described under "Experimental Procedures" with entrapped CF. Stoppedflow experiments were performed in which liposomes were rapidly exposed to a doubling of extravesicular osmolality. Vesicle shrinkage results in CF selfquenching. A, osmotic water permeability of outer leaflet (Exofacial) and inner leaflet (Cyto) liposomes; mean Ϯ S.E. for three separate liposome preparations. B, osmotic water permeability of exofacial liposomes modified by the removal of individual lipids. C and D, averaged stopped-flow traces (n ϭ 6 -10) from a single experiment. Fluorescence was converted to relative volume and curves show raw data plus superimposed fitted single exponential decay functions. *, p Ͻ 0.05 compared with exofacial.

FIG. 2. Effect of changing acyl chain composition on water permeability.
Liposomes were prepared as described under "Experimental Procedures" with entrapped CF. Stopped-flow experiments were performed in which liposomes were rapidly exposed to a doubling of extravesicular osmolality. A, osmotic water permeability of exofacial liposomes. Exofacial liposomes had bovine heart PC replaced with DPPC (DPPC/Exo), or DPPC/Exo liposomes had GSLs removed and replaced with more DPPC (DPPC/ Exo-GSLs). B, osmotic water permeability of cytoplasmic liposomes. Cytoplasmic liposomes had bovine heart PE replaced with E. coli PE (E. coli/Cyto). C and D, averaged stopped-flow traces (n ϭ 6 -10) from single experiments. Fluorescence was converted to relative volume, and curves were fitted with single exponential decay functions. Results shown in A and B are mean Ϯ S.E. for three separate liposome preparations. *, p Ͻ 0.05 compared with cytoplasmic liposomes.
(glycerol, Fig. 6B) permeabilities. For all four solutes, the removal of SM resulted in rates that were somewhat higher than the GSL-lacking liposomes. For formamide, acetamide, urea, and glycerol, the increases were 4.3, 7.5, 5.6, and 5.9 times, respectively. Doing a simple mean on these two sets of numbers shows that the membranes lacking SM were 22% more permeable than those lacking GSLs. As the mole fractions of these two lipids are approximately the same (see Table I), we conclude that both SM and GSLs contribute to the barrier of the exofacial leaflet, but SM is 20 -25% better at reducing permeability. The removal of cholesterol, as for water, resulted in a massive increase in the permeability of all four solutes. The increases for formamide, acetamide, urea, and cholesterol were 95, 136, 191, and 285 times, which again follows the rank order of molecular weights for these non-electrolytes. These data make it clear that sphingolipids alone will not significantly restrict solute or water permeation.
NH 3 Permeability of Leaflet Liposomes-The permeation kinetics of NH 3 were tested by exposure of liposomes to an NH 4 Cl solution clamped to pH 6.8. Free NH 3 gas in the buffer perme-ates across the membrane and upon protonation to NH 4 ϩ effects an increase in intravesicular pH, which can by measured by fluorescence changes. Fig. 7C shows that NH 3 flux is extremely rapid across the membrane of cytoplasmic leaflet liposomes. In contrast, the flux across exofacial membranes is much slower. This illustrates an important physiological phenomenon, namely that ordering a membrane to reduce the permeability of water and small non-electrolytes is also sufficient to hinder the permeability of a gas. Fig. 7A shows the quantitative difference in the permeability of the two membranes to NH 3 to be 60-fold. When GSLs were removed from exofacial liposomes the permeability increased 2.2 times, and when SM was removed the increase was 4.0 times. Removal of cholesterol increased the NH 3 permeability of the membrane 82-fold. These increases are similar to those seen for solutes and imply that NH 3 permeation occurs by a similar solubility-diffusion mechanism.
Apparent Proton Permeability of Leaflet Liposomes-The proton permeability of the two membranes was assessed by exposing liposomes equilibrated to pH 7.5 to an external buffer with pH 6.9. Therefore, our measurement follows the rate of FIG. 3. Formamide permeability of exofacial, cytoplasmic, and modified exofacial liposomes. Liposomes were prepared as described under "Experimental Procedures" with entrapped CF. Liposomes were preloaded with 200 mM formamide and then abruptly exposed to a solution of identical osmolality containing 150 mM formamide. Solute efflux and vesicle shrinkage results in CF self-quenching. A, formamide permeability of outer leaflet (Exofacial) and inner leaflet (Cyto) liposomes; mean Ϯ S.E. for three separate liposome preparations. B, formamide permeability of exofacial liposomes modified by the removal of individual lipids; Exo-GSL lack glycosphingolipids; Exo-SM lack sphingomyelin; Exo-Chol lack cholesterol. C and D, averaged stopped-flow traces (n ϭ 6 -10) from a single experiment. Fluorescence was converted to relative volume, and curves were fitted with single exponential decay functions. *, p Ͻ 0.05 compared with exofacial.
FIG. 4. Acetamide permeability of exofacial, cytoplasmic, and modified exofacial liposomes. Liposomes were prepared as described under "Experimental Procedures" with entrapped CF. Liposomes were preloaded with 200 mM acetamide and then abruptly exposed to a solution of identical osmolality containing 150 mM acetamide. Solute efflux and vesicle shrinkage results in CF self-quenching. A, acetamide permeability of outer leaflet (Exofacial) and inner leaflet (Cyto) liposomes; mean Ϯ S.E. for three separate liposome preparations. B, acetamide permeability of exofacial liposomes modified by the removal of individual lipids; Exo-GSL lack glycosphingolipids; Exo-SM lack sphingomyelin; Exo-Chol lack cholesterol. C and D, averaged stopped-flow traces (n ϭ 6 -10) from a single experiment. Fluorescence was converted to relative volume, and curves were fitted with single exponential decay functions. *, p Ͻ 0.05 compared with exofacial. dissipation of a pH gradient across the liposomal membrane. Although the precise mechanism of proton permeation is unclear, this approach allows us to measure the apparent proton permeability of the membranes in question. It should be noted that this technique does not allow us to necessarily distinguish between proton influx or hydroxide anion efflux. However, because proton flux appears to be independent of pH, there is good reason to believe that protons are in fact the permeating species (for a fuller discussion of this point, see Ref. 30). In Fig.  8B, typical pH-dependent fluorescence quenching profiles are shown for exofacial and exofacial-lacking-cholesterol liposomes. Cytoplasmic membrane permeability to protons was unexpectedly slower than that of exofacial membranes by a factor of four (Fig. 8A). Membranes with the lowest permeability to protons were cholesterol-depleted exofacial liposomes (Exo-Chol; Fig. 8A). This implies that cholesterol may contribute to the proton permeation pathway. The combined permeability data for all liposome variants is shown in Fig. 8A and reveals that the exofacial membranes had the highest proton permeability (0.011 cm/s) and that removal of sphingolipids reduced that permeability approximately 2-fold. DISCUSSION We have used published data on the lipid composition of MDCK type 1 apical membranes and in addition made a number of simplifying assumptions regarding its bilayer asymmetry, in order to construct vesicles that mimic each leaflet's composition. This analysis predicts that the apical membrane of these cells is comprised of an asymmetric bilayer, which is enriched in SM, GSLs, and PC in the exofacial leaflet, whereas the inner leaflet is composed mostly of the glycerophospholipids, PE, PS, and PI (Table I). Evidence to support these conclusions has come from a number of studies. For instance, in brush border membranes from rabbit intestine, it was shown that PE phospholipid was largely inaccessible to reagents from the outside (31). A similar accessibility study of rabbit kidney brush border membranes revealed that PE, PS, and PI were localized to the inner leaflet, whereas SM was external (32). In MDCK cells a fluorescent ceramide analogue, which was taken up and converted to SM and GSL in the Golgi, was completely extractable by bovine serum albumin from the apical side, demonstrating a preferential accumulation in the exofacial FIG. 5. Urea permeability of exofacial, cytoplasmic, and modified exofacial liposomes. Liposomes were prepared as described under "Experimental Procedures" with entrapped CF. Liposomes were preloaded with 200 mM urea and then abruptly exposed to a solution of identical osmolality containing 150 mM urea. Solute efflux and vesicle shrinkage results in CF self-quenching. A, urea permeability of outer leaflet (Exofacial) and inner leaflet (Cyto) liposomes; mean Ϯ S.E. for three separate liposome preparations. B, urea permeability of exofacial liposomes modified by the removal of individual lipids; Exo-GSL lack glycosphingolipids; Exo-SM lack sphingomyelin; Exo-Chol lack cholesterol. C and D, averaged stopped-flow traces (n ϭ 3-8) from a single experiment. Fluorescence was converted to relative volume, and curves were fitted with single exponential decay functions. *, p Ͻ 0.05 compared with exofacial.
FIG. 6. Glycerol permeability of exofacial, cytoplasmic, and modified exofacial liposomes. Liposomes were prepared as described under "Experimental Procedures" with entrapped CF. Liposomes were preloaded with 200 mM glycerol and then abruptly exposed to a solution of identical osmolality containing 150 mM glycerol. Solute efflux and vesicle shrinkage results in CF self-quenching. A, glycerol permeability of outer leaflet (Exofacial) and inner leaflet (Cyto) liposomes; mean Ϯ S.E. for three separate liposome preparations. B, glycerol permeability of exofacial liposomes modified by the removal of individual lipids; Exo-GSL lack glycosphingolipids; Exo-SM lack sphingomyelin; Exo-Chol lack cholesterol. C and D, averaged stopped-flow traces (n ϭ 3-8) from a single experiment. Fluorescence was converted to relative volume, and curves were fitted with single exponential decay functions. *, p Ͻ 0.05 compared with Exofacial.
leaflet (25). The predicted asymmetry also fits with well established rules, which indicate that choline-containing lipids are mostly exofacial, whereas the amino-containing phospholipids localize cytoplasmically (33). Cholesterol is present in both leaflets and was apportioned according to the quantitative analyses described in the reference papers (15,23).
Because the lipids of the exofacial leaflet of polarized epithelia are laterally constrained by the presence of tight junctions, whereas the lipids of the cytoplasmic leaflet are not, the inner leaflet of the apical domain is assumed to be the same as that of the basolateral membrane (13,34). As a consequence, it appears likely that the barrier function exhibited by "watertight" epithelia resides in the exofacial leaflet. Present studies directly tested this hypothesis and confirmed its validity. For water, small non-electrolytes, and NH 3 , all substances known to cross phospholipid bilayers by a "solubility-diffusion" mechanism, the permeability of the cytoplasmic leaflet was 1-2 orders of magnitude higher than the exofacial. These data indicated that the presence of appropriate lipids present in defined ratios was sufficient to generate a membrane with extremely low permeability to water, solutes, and NH 3 . Moreover, these experiments have allowed us to quantitate the magnitude of that difference in leaflet permeabilities for a range of biologically relevant permeant molecules.
For water, the cytoplasmic leaflet liposomes were 18 times more permeable than the exofacial with osmotic permeability coefficients of 4.4 ϫ 10 Ϫ3 cm/s and 2.4 ϫ 10 Ϫ4 cm/s, respectively. We have previously shown that each leaflet is able to offer an independent resistance to permeation (18,19). By analogy with electrical circuit theory, resistance is equal to the reciprocal of the permeability, i.e. R ϭ 1/P f (see Eq. 1). As our exofacial and cytoplasmic liposomes have symmetrical leaflets, knowing the membrane permeability of each bilayer allows us to calculate single leaflet permeabilities according to Equation 1. Combining the derived value for an exofacial leaflet and a cytoplasmic leaflet gave us a theoretical osmotic permeability coefficient (P f ) of 4.6 ϫ 10 Ϫ4 cm/s, which correlates well with the measured permeability of this membrane in cultured MDCK type 1 cells. Such close concordance of measured and theoretical values as determined by our model system argues that the assumptions made and the lipid compositions arrived at are probably reasonable representations of the real membrane and allow us to conclude that the outer leaflet of the MDCK type 1 apical membrane is probably 20-fold less permeable to water than the inner leaflet. These results also success- FIG. 7. NH 3 permeability of exofacial, cytoplasmic, and modified exofacial liposomes. Liposomes were prepared as described under "Experimental Procedures" with entrapped CF. Liposomes equilibrated to pH 6.8 were abruptly exposed to a buffer of identical pH containing NH 4 Cl. NH 3 entering the liposome becomes protonated to NH 4 ϩ , and, as a result, the pH and CF fluorescence increases. A, NH 3 permeability of outer leaflet (Exofacial) and inner leaflet (Cyto) liposomes; mean Ϯ S.E. for three separate liposome preparations. B, NH 3 permeability of exofacial liposomes modified by the removal of individual lipids; Exo-GSL lack glycosphingolipids; Exo-SM lack sphingomyelin; Exo-Chol lack cholesterol. C and D, averaged stopped-flow traces (n ϭ 6 -10) from a single experiment. Fluorescence was converted to pH, and curves were fitted with single exponential decay functions. *, p Ͻ 0.05 compared with exofacial.
FIG. 8. Apparent H ؉ permeability of exofacial, cytoplasmic, and modified exofacial liposomes. Liposomes were prepared as described under "Experimental Procedures" with entrapped CF. Liposomes were abruptly exposed to a buffer of lower pH such that the extravesicular pH was 6.9. Dissipation of the pH gradient reduces intravesicular pH and CF fluorescence. A, H ϩ permeability of outer leaflet (Exofacial) and inner leaflet (Cyto) liposomes and of exofacial liposomes modified by the removal of individual lipids; Exo-GSL lack glycosphingolipids; Exo-SM lack sphingomyelin; Exo-Chol lack cholesterol; mean Ϯ S.E. for three separate liposome preparations. B, averaged stopped-flow traces (n ϭ 6 -10) from a single experiment. Fluorescence was converted to pH, and curves were fitted with single exponential decay functions. *, p Ͻ 0.05 compared with exofacial. The conservativeness of the Bonferroni t test resulted in only one significant difference. Single comparison testing gave p ϭ 0.034 and p ϭ 0.013 for Exo-GSL and inner leaflet (Cyto) liposomes, respectively. fully apply Equation 1 to membranes of real cells for the first time.
Although we based our leaflet compositions on the best evidence available, there is little information known about the acyl chain composition of these various lipid classes in the MDCK apical membrane. Consequently, we chose mammalian lipids as the source for these experiments in an attempt to minimize artifacts that may have arisen from the use of bacterial or other lipid sources. Based on information provided by Avanti Polar Lipids, we know the acyl chain composition of our artificial membranes (see Supplemental Material available online). As expected the saturation ratio (i.e. saturated:unsaturated (S:U)) was quite different between the two membranes. Exofacial membranes had an S:U of 2.5 due to the presence of saturated acyl chains on SM and GSL. In contrast, cytoplasmic membranes had an S:U of 0.6, indicating a much higher proportion of double bond-containing fatty acids. Exofacial membranes contained high proportions of 16:0 (35.0%), 18:0 (8.1%), and 24:0 (16.4%) acyl chains. The predominant unsaturated acyl chains were 18:1 and 18:2 (11.1%) and 24:1 (8.4%). In comparison to this, the cytoplasmic membranes contained little 16:0 with the predominant saturated species being 18:0 (34.1%). Unsaturated chains were 18:1 (12.3%), 18:2 (12.3%), and 20:4 (27.1%). The acyl chain composition of these membranes is likely to be an important contributor to their permeation properties due to the different packing constraints imposed by the presence of double bonds. Double bonds introduce a rigid bend in an acyl chain, resulting in greater free volume space and higher fluidity in the membrane. It has been shown previously, for substances that cross membranes by a solubility-diffusion mechanism, that membrane fluidity and permeability are tightly correlated (20). The results presented here further support this principle.
We investigated the effect of using lipids with different fatty acid substituents. The use of disaturated DPPC in an exofacial membrane reduced the membrane permeability but more surprisingly appeared to substitute for GSLs upon their removal by maintaining low permeability. At 25 C DPPC is in the gel state and with a phosphocholine head group strongly resembles SM. Although there is no amide group for hydrogen bonding with cholesterol, the tight chain packing induced by cholesterol and saturated PC in the gel state appears sufficient to compensate for loss of GSLs. It is therefore extremely important that the most physiologically relevant lipids are used in membrane modeling studies. Clearly the complement of unsaturated acyl chains is an important determinant in the permeation properties of the membrane.
To further examine the validity of our leaflet-specific liposomes as a model for predicting reconstituted membrane permeabilities, it is informative to compare our results with permeabilities found in other barrier epithelia. Our theoretical MDCK membrane permeability to urea (employing Equation 1) is 6.2 ϫ 10 Ϫ8 cm/s. This compares to values of 2.1 ϫ 10 Ϫ7 cm/s (3-fold higher), 7.8 ϫ 10 Ϫ7 cm/s (13-fold higher), and 5.5 ϫ 10 Ϫ7 cm/s (9-fold higher) for toad bladder apical membrane, mammalian bladder apical membrane, and gastric apical membrane, respectively (3). Therefore, our reconstituted membrane system has a urea permeability approximately 1 order of magnitude lower than that seen for some other barrier membranes. We have, in addition, previously measured the diffusive urea permeability of MDCK cells to be 6.0 ϫ 10 Ϫ6 cm/s (17). This is approximately 1 order of magnitude higher than is seen for other barrier epithelia, and 2 orders of magnitude higher than our model liposomes would predict for this membrane. We believe this may indicate some form of facilitated urea uptake by MDCK cells. To the best of our knowledge, urea transport activity has not specifically been demonstrated in MDCKs; however, it is possible that cells derived from the distal nephron would possess such an activity. Further support for this conjecture comes from permeability studies performed by Rivers et al. (27), in which they measured the acetamide and formamide permeabilities of MDCK apical membranes to be 4.1 ϫ 10 Ϫ6 and 5.5 ϫ 10 Ϫ6 cm/s, respectively. The theoretical permeabilities for these two non-electrolytes as derived from our model liposomes was 3.8 ϫ 10 Ϫ6 and 5.5 ϫ 10 Ϫ6 cm/s, respectively, which are in very close agreement. Furthermore, liposomes composed of 25% SM, 40% cholesterol, and 35% PC have been shown to have a urea permeability of 4.4 ϫ 10 Ϫ8 cm/s (20), which is close to the value obtained for the exofacial liposomes used in this study (3.1 ϫ 10 Ϫ8 cm/s with a composition of 25% SM, 39% cholesterol, 36% PC and GSLs). Taken together these results indicate that urea permeability may be facilitated in MDCK cells and not reflect simple membrane permeant behavior. Measurements of water and other solutes, however, appear to validate our hypothesis and the use of Equation 1 to predict apical membrane permeability properties.
The theoretical MDCK membrane permeability to NH 3 is 2.6 ϫ 10 Ϫ4 cm/s, which is within an order of magnitude of values measured for other barrier epithelia (3). The proton permeability for our artificial membrane was 4.3 ϫ 10 Ϫ3 cm/s, which compares favorably with values of 8.0 ϫ 10 Ϫ3 cm/s, 3.0 ϫ 10 Ϫ3 cm/s, and 19 ϫ 10 Ϫ3 cm/s, obtained for medullary thick ascending limb apical membrane, whole mammalian bladder, and gastric apical membrane, respectively (3). These comparisons between our in vitro reconstituted membranes and real asymmetric barrier membranes lend support to the conclusion that these artificial liposomes reconstitute the biophysical characteristics of true barrier membranes to a wide range of biologically important permeant molecules.
We performed measurements of proton permeation that relied upon the dissipation of an applied pH gradient across the liposomal membrane with consequent quenching of an intravesicular pH-dependent fluorophore. Although this is a well established technique for monitoring proton permeation processes (3,10,19,20,22,30,37), the exact mechanism is not well understood. Measurements of proton permeability can yield very different values depending on the method used. If small pH gradients are allowed to dissipate around pH 7, values in the range of 10 Ϫ4 cm/s are obtained (35). However, if large pH gradients are applied to liposomes, proton fluxes with values of 10 Ϫ9 cm/s are recorded (36). This phenomenon was explored by Deamer and Nichols (37) and led them to conclude that decay of large pH gradients results in diffusion potentials that limit proton flux. Addition of valinomycin in K ϩ -containing buffers eliminated the formation of membrane potentials and therefore allowed proton fluxes to become rate-limiting. However, if proton fluxes were measured near pH 7 in the presence of small proton gradients, there was no formation of a limiting diffusion potential (37). Our experiments were carried out under the latter conditions, and as noted by Deamer and Nichols, the presence or absence of valinomycin did not affect the rate of proton flux. Although this indicates our measurements were not influenced by the formation of membrane potentials, it does appear to contradict rules of electroneutrality. Given that we cannot rule out that our proton flux may be accompanied by some other charge-dissipating process i.e. proton-anion co-permeation, we have chosen to refer to these measurements as demonstrating the apparent proton permeability of the liposomal membranes.
Our measurements of proton permeability in cytoplasmic and exofacial membranes yielded the unusual result that cyto-plasmic lipids had 4-fold lower proton permeability than the exofacial (Fig. 8). The mechanism by which protons translocate across phospholipid bilayers is not well understood, but it has clearly been established that, unlike water, solutes, and NH 3 , membrane fluidity is not a key determinant in permeation (19,20,38). In a study that examined the relationship between membrane fluidity and permeability, it was found that liposomes of differing composition and fluidities exhibited water permeabilities that varied up to 70-fold, whereas proton permeability varied no more than 3-fold. Moreover, within that 3-fold range, there was poor correlation between proton permeability and the fluidity of the membrane (20). Another set of experiments performed on DPPC liposomes revealed that protons permeated 4 times faster when the liposomes were heated above the phase-transition temperature, but that solute fluxes under the same conditions increased by 2 orders of magnitude (19). Brookes et al. (38) showed that liposomes made from phospholipids extracted from the mitochondrial inner membrane of eight different vertebrate species, with widely varying mitochondrial proton leak and degree of acyl chain unsaturation exhibited no differences in proton permeability. They concluded there was no correlation between liposome proton permeability and phospholipid fatty acid composition. The data presented here further support this. When cholesterol was removed from the exofacial membrane, water permeability increased nearly 40-fold. In dramatic contrast to this, proton permeability decreased 6.5-fold.
Proton permeation across membranes has been shown to occur at rates 5-6 orders of magnitude higher than for other monovalent cations. Several explanations have been proposed to account for this anomalous behavior. The first hypothesizes the presence of fatty acid "contaminants" within the membrane that are available for protonation. The protonated species translocates to the other side, releases a proton, and then cycles back as the anionic species (39,40). The second proposes the formation of transient defects in the interfacial region of the bilayer through which water "fingers" may project briefly and from which hydrated pores carrying protons may be pinched off. Water molecules present within the hydrophobic interior of the bilayer may then form transient hydrogen bonded chains or water wires, along which protons can move, thus allowing sequential water to water translocation across the membrane (30,39,41). Evidence for both mechanisms exist, and it is likely that both contribute to some degree depending on the nature of the membrane: its composition, thickness, the pH on either side of the bilayer, the relevant buffering agents present, and the temperature. Our proton flux data for the modified exofacial membranes reveal an additional intriguing possibility. When either of the sphingolipids was removed, proton permeation was reduced (Fig. 8). When cholesterol was removed entirely, the rate fell to one-sixth of that shown by the complete membrane. This phenomenon suggests that there is something specific about the presence of both cholesterol and sphingolipid in this membrane that facilitates proton transfer. One possibility is that the specific hydrogen bonding interaction, which occurs between the 3-OH group of cholesterol and the amide group present on GSLs and SM, provides a suitable early docking site for protons close to the aqueous/lipid headgroup interfacial region. The presence of this hydrogen bond may lower the free energy required for protons to partition into the proximal hydrophobic region of the membrane. The slow rate of permeation across cytoplasmic membranes may be consistent with this model. Although cholesterol constitutes 37 mol% of this bilayer, there are no sphingolipids present; therefore, the specific hydrogen bonding interaction we speculate may be important is not present. The effect of cholesterol and cholesterol/sphingo-lipid interactions on proton transport has not previously been studied. Therefore, further work will be needed to confirm this hypothesis.
Our model liposomes appear to recapitulate the barrier properties of the MDCK apical membrane to water, solutes (with the exception of urea), and NH 3 , to a degree consistent with that actually observed in these cells. The differences in permeability between the two membrane compositions is most likely due to the reduction in fluidity and increase in chain ordering, which occurs as a consequence of sphingolipid/cholesterol interactions in the exofacial liposomes. Specifically, it has been shown that there are likely to be hydrogen bonding interactions between the amide functional group of SM and the 3-OH group of cholesterol (16,42). In addition, there are likely to be strong van der Waals attractions as a result of the contribution of the planar steroid rings interacting with the saturated hydrocarbon chains of the sphingolipids. These cohesive forces result in greater chain ordering due to a reduction in trans-gauche isomerization about the carbon-carbon bonds and an increase in sphingolipid packing density (42). In addition to these interactions, GSLs are also known to self-associate through hydrogen bonding of their glycosyl head groups. Therefore, either in isolation or collectively, these molecular interplays may be responsible for reducing the fluidity and consequently the permeability of the exofacial leaflet.
To investigate which of the associative interactions between the lipids used in our exofacial liposomes were of greatest significance, we focused on the putative contribution of the sphingolipids. As GSLs are present in high concentration in the exofacial leaflet of epithelial apical membranes and appear to be almost exclusively localized to this membrane domain, we hypothesized that they might play a significant role in reducing membrane permeability. Indeed, this proved to be the case with both GSLs and SM contributing to reduced membrane permeability. The removal of approximately 20 mol% of either species and replacement with bovine heart PC resulted in 2-3-fold increases in membrane water permeability. For solutes, the increases were a little higher (3-7-fold). The data suggested that SM was better at reducing membrane permeability than GSLs. This is consistent with findings that show that cholesterol has a greater condensing effect on SM compared with galactosylceramide films when those sphingolipids were acyl chain-matched and in similar phase states (43). The greater cross-sectional area reduction exhibited by SM compared with GSL would certainly lead to tighter packing and hence lower permeability. When cholesterol was removed, the water permeability increased by 37-fold. A large increase in permeability was anticipated because of the known ability of cholesterol to condense and rigidify lipid bilayers. Cholesterol with its relatively planar structure orients itself with the polar 3-OH group toward the aqueous phase and its tetracyclic rings positioned between carbons 2 and 10 of the acyl chains (42). This spatial localization tends to increase local ordering of acyl chains in the proximal portion of the leaflet. In addition, as noted, the 3-OH can function as hydrogen bond donor and interact with the amide group of sphingolipids.
It is worth noting that the presence of sphingolipids in high concentrations in the outer leaflet of epithelial apical membranes can trap cholesterol in that leaflet by virtue of hydrogen bonding interactions and therefore may be contributing to the formation and functioning of detergent insoluble glycolipid rafts. These cholesterol and sphingolipid enriched microdomains have been implicated in a range of critical cellular processes including intracellular lipid and protein sorting, caveolae formation, and signal transduction (44,45). The cell, therefore, may not only be erecting a defensive shield at its apical pole but also be assembling functional lipid ensembles necessary for other specific molecular interactions.
These results demonstrate that MDCK type 1 cells can synthesize a low permeability apical membrane by the selective trafficking of cholesterol, SM, and GSLs to the outer leaflet of the plasma membrane. The cytoplasmic leaflet of this membrane, although possessing a relatively high cholesterol content (37.3 mol%), is nonetheless quite permeable to water and solutes, presumably due to the inability of cholesterol to form hydrogen bonds with glycerophospholipids like PE and PS. The successful application of the additive reciprocal equation for leaflet permeabilities validates our modeling of the MDCK apical membrane. The concordance of derived apical membrane permeability coefficients with actual permeabilities measured by others provides strong evidence that our liposomes had realistic lipid compositions and further validates the hypothesis that leaflets offer independent barriers to the permeation of small uncharged molecules.