Development and Evaluation of a Human Skin Equivalent in a Semiautomatic Microfluidic Diffusion Chamber

Abstract

There is an increasing demand for transdermal transport measurements to optimize topical drug formulations and to achieve proper penetration profile of cosmetic ingredients. Reflecting ethical concerns the use of both human and animal tissues is becoming more restricted. Therefore, the focus of dermal research is shifting towards in vitro assays. In the current proof-of-concept study a three-layer skin equivalent using human HaCaT keratinocytes, an electrospun polycaprolactone mesh and a collagen-I gel was compared to human excised skin samples. We measured the permeability of the samples for 2% caffeine cream using a miniaturized dynamic diffusion cell ("skin-on-a-chip" microfluidic device). Caffeine delivery exhibits similar transport kinetics through the artificial skin and the human tissue: after a rapid rise, a long-lasting high concentration steady state develops. This is markedly distinct from the kinetics measured when using cell-free constructs, where a shorter release was observable. These results imply that both the established skin equivalent and the microfluidic diffusion chamber can serve as a suitable base for further development of more complex tissue substitutes.

Keywords: 3D printed device; electrospun mesh; microfluidic diffusion chamber; skin equivalent; transepithelial transport kinetic.

Figure 1

Skin equivalent sample holder device. (a) Scanning electron microscopic image of the electrospun PCL nanomesh consisting of filaments with sub-micron diameter. Scale bar: 3 μm. (b) Filament diameter distribution of the electrospun nanomesh, obtained from $n=4$ independent samples. (c) The assembled sample holder device, scale bar represents 1 mm. (d) Schematic drawing depicting the parts of the device. (e) Schematic cross section of the assembled sample holder device. The electrospun nanomesh serves as the substrate of the cells and is stretched across the sample holder containing collagen-I gel. The mesh is hold in place by a tightly fitting cone shaped ring.

Figure 2

Histology of skin equivalents. (a) Live/dead labeled image of a monolayer of HaCaT keratocytes, imaged on the electrospun membrane at day 25 in culture. Calcein (green) fluorescence shows live cells while dead cell nuclei are identified by ethidium (red) fluorescence. Co-localization of green and red fluorescence appears as yellow. (b) Toluidine blue-stained semi-thin section of a monolayer of cells at day 18 in culture. Blue arrows point to cell-cell junctions, asterisks indicate cell nuclei with apparent nucleoli, red arrowhead marks an apoptotic cell. (c,d) Frozen cross-sections of a keratinocyte culture kept in culture for 28 days and at the air-liquid interface (ALI) for 14 days. (c) Cells and nuclei are visualized with calcein (green) and NucBlue (blue) fluorescence. (d) Phase-contrast image of the same field shown in (c) merged with ethidium fluorescence (red), identifying dead cells. Dotted line indicates the border between the cell layer and the electrospun membrane. Scale bars: 20 μm.

Figure 3

Transdermal transport measurements using skin equivalents and human skin. A photo (a) and schematic drawing (b) shows the sample holder device fitted into a microfluidic Franz chamber inside an incubator box. The cap of the chamber immobilized the sample holder device and held the caffeine-containing cream in close contact with the skin equivalent. Medium was transfused through the microfluidic chamber, and fractions of the medium leaving the chamber were collected every 30 min. (c) Caffeine concentration in the collected fractions was measured by spectrophotometry and is shown as a function of time. Caffeine exposition was started at $t=0$. Red, magenta and teal colored symbols represent the average caffeine concentration obtained from the skin equivalent, human skin and cell-free samples, respectively. Error bars represent SEM, from $n=3$ independent experiments. The difference between concentration readings from the marked mesh-only and SE samples is significant ($p=1.2·{10}^{-7}$, $n=12$ in each group).