246 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS Table II Permeability Coefficients of Various Markers Across BSA-Treated Artificial Membranes Marker Permeability coefficient (cm/h) Molecular Partition weight coefficient BSA-treated artificial Human skin (daltons) (octanol/water) membranes (n = 6) (from ref 15) Sucrose Cortisol Estradiol Progesterone 342 •0 0.000224 ñ 0.000063 •0 363 40 0.00255 ñ 0.00067 0.0003 272 490 0.00536 ñ 0.00273 0.03 315 5890 0.0266 ñ 0.0064 0.15 Values shown are average ñ SD. suggests that in order to obtain a model membrane that mimics skin permeability characteristics, a lipid matrix alone is not sufficient. In an attempt to simulate corneo- cyte-lipid bilayer interactions, heat-treated BSA was included in the model membrane to allow (i) the generation of hydrophobic protein microglobules or microspheres within areas of the membrane that may have previously constituted the polar pathways for markers such as sucrose and (ii) the creation of anchor sites similar to corneocytes in -4 ' I ' I ' I ' I ' I ' I -6 -4 -2 0 2 4 6 log PC (octanol/water) Figure 5. Correlation between permeability coefficients of markers across BSA membranes and octanol- water partition coefficients.

ARTIFICIAL MEMBRANES 247 -11 •, -2 E E -4 -9 -8 -7 -6 -5 -4 -3 -2 log P (human skin) Figure 6. Correlation between permeabilities of markers across BSA membranes and across human skin. stratum corneum with which the lipid bilayer matrix may interact and result in a more cohesive membrane. The BSA-treated membrane, after calcium chloride treatment, exhibited markedly higher resistance to the permeation of sucrose compared to non- treated membranes. It was also confirmed that heat treatment at 80øC of blank filters alone or of lipid membranes without BSA treatment did not lead to reduction of sucrose permeation across these membranes. It was also observed that both BSA and calcium chloride treatment protocols were essential to the generation of model membranes with high barrier properties to hydrophilic drug markers. Treatment of lipid-laden filters with either BSA only or calcium chloride alone did not suffice. This suggests that the model membrane obtained after the combined treatment protocols probably incorporates interactions between the protein and lipid bilayers via calcium-mediated linkages. It would be necessary to carry out extensive structural investigations by electron mi- croscopy and other spectroscopic methods to define the interactions at a molecular level. It is clear, however, that the model membranes prepared in this fashion possess excellent potential to screen a variety of compounds for permeation characteristics across human or animal skin. The excellent correlations of permeability values of several markers across the BSA-model membranes with octanol-water partition coefficients (Fig 5, r 2 = 0.92) and with available data on their permeability across human skin (Fig 6, r 2 = 0.97)