Selective adhesion of hepatocytes on patterned surfaces.

Successful development of cell-biased bioartificial liver devices necessitates the establishment of techniques and designs for long-term, stable hepatocellular function and efficient transport of nutrients and wastes within the device. Given the relatively large cell mass that one must consider, one possible solution involves the use of micropatterning technology to sandwich hepatocytes aligned in rows between two micropatterned surfaces. Rows of cells would alternate with hepatocyte-free areas, creating efficient transport channels for fluid flow and nutrient exchange. Ultimately, this type of device could also be used as a three-dimensional construct for investigating a variety of cell-surface, cell-extracellular matrix, and cell-cell interactions. To achieve this goal, one must develop techniques for selectively adhering hepatocytes to solid substrates. In this study, reproducible, selective adhesion of hepatocytes on a glass substrate with large regions of adhesive (AS) and nonadhesive (NAS) surfaces was obtained. The AS had hydrophilic characteristics, enhancing deposition of collagen molecules from an aqueous solution, and subsequent hepatocyte adhesion, whereas the NAS had hydrophobic properties and remained collagen-free and hepatocyte-free. In addition, a reproducible processing technique for obtaining patterns of hepatocytes was developed and optimized, using a surface with a single AS band as a first approximation to a micropatterned device. This was achieved by spincoating an aqueous collagen type I solution (0.1 mg/mL) on a banded surface at 500 rpm for 25 seconds. The morphology and long-term function of the hepatocytes attached to AS in nonbanded and banded surface configurations was assessed by mimicking sandwich culture and was shown to be similar to stable, differentiated sandwich cultures. Mathematical modeling was used to determine critical design criteria for the hypothetical micropatterned device. The oxygen distribution and viscous pressure drop were modeled along a typical microchannel and limited to in vivo values. An optimal channel length of 0.6 cm and a flow rate of 2.0 x 10(-6) mL/s were obtained for a channel of 100 microns in width and 10 microns in height. These values were reasonable in terms of practical implementation.