A collagen–chondroitin sulfate substrate cross-linked

with glutaraldehyde was used as a tissue matrix. Initially a thin layer of endothelial cells was grown in a culture dish. Keratocytes and support proteins were added before finally adding the final epithelial layer. The gross morphology, transparency and histology were reported to be similar to that of a natural cornea. Tests performed using mild detergents determined that the construct had a similar gene expression and wound-healing response when compared to human eye-bank corneas, albeit more sensitive. The stromal matrix was later modified to allow for recovery mechanisms following exposure to chemical treatments (Doillon et al., 2003), and this was later followed by the introduction of nerve–target cell interactions (Suuronen et al., 2004). Dorsal root

ganglia isolated from chick embryos were utilized as a CX-5461 concentration neural source, since optimal function, maintenance and wound healing of many tissues is dependent to some extent on peripheral sensory innervations (Suuronen et al., 2004). The innervated corneal constructs were reported to have lower cell death rates when exposed to test chemicals compared to non-innervated equivalents. This suggests that the presence of nerves protects the epithelium from chemical irritation and possibly explains why previous non-innervated MEK inhibitor corneal models have been deemed over-sensitive when used in toxicity studies. This

model still requires further development since many of the functional properties of the nerves remain unclear. These types of models may demonstrate more promise for clinical development as cadaveric alternatives for corneal transplantation rather than as models for toxicological testing. Reichl et al. (2005) manufactured a human corneal PIK-5 equivalent for in vitro drug permeation studies by culturing immortalized epithelial, endothelial and stromal cells in a collagen hydrogel matrix. Three reagents commonly used in ophthalmic drugs to treat glaucoma and inflammatory diseases were tested and permeation data obtained was compared with those from excised porcine cornea and a porcine cornea construct ( Reichl et al., 2004 and Reichl and Muller-Goymann, 2003). Porcine corneas were investigated due to their relatively similar anatomy and physiology to the human cornea. The human cornea construct had similar epithelial barrier properties to a native cornea with only small ultrastructural differences, possibly due to lack of tears and blinking. There was increased permeability in the corneal equivalents compared to the exercised porcine cornea for all reagents tested, although the differences were relatively minor. Unfortunately there was no data available to compare these corneal equivalents with an excised human cornea (as in the studies by Griffith et al. (1999).