Organotypic 3-D corneal epithelial cultures approximate corneal epithelium, forming stratified layers approximately 60 μm thick and expressing corneal-specific keratins . The layers consist of four to six cells, with flattened superficial cells and several layers of wing cells; ultrastructurally, the cells demonstrate surface microvilli, tight junctions, numerous intercellular interdigitations and desmosomes and a basement membrane . The 3-D cultures also have been shown to develop similar barrier properties to corneal epithelium as measured by transepithelial electrical resistance [13, 17].
The assays presented herein using 3-D corneal epithelial cultures demonstrated no significant differences in cytotoxicity between the BAC-containing latanoprost and olopatadine ophthalmic solutions and BAC-free travoprost ophthalmic solution at either of the time points (10 and 25 minutes) tested. Our results are in contrast to a study demonstrating latanoprost toxicity in monolayer cultures of HCE-T cells . The same group reported that up to 70% of the cells grown under these conditions died in 5 minutes in their control, which was treated with culture medium alone . However, HCE-T monolayer cultures have been found to exhibit cytotoxicity when exposed to concentrations of detergents well below those affecting the viability of 3-D cultures . Furthermore, an ophthalmic solution that was nontoxic in rabbit corneas and 3-D cultures was shown to cause breakage of cell junctions and vacuolization in HCE-T monolayers . These studies suggest that HCE-T monolayer cultures are not appropriate for evaluating the cytotoxicity of ophthalmic compounds.
The difference in cytotoxicity between the HCE-T monolayers and the 3-D cultures may reflect the extent of corneal differentiation. We demonstrated that 3-D cultures express significantly higher mRNA and protein levels of three corneal epithelial markers (EGFR, involucrin and TG1) compared to HCE-T monolayers. EGFR expression has been demonstrated in the basal layers of human corneal epithelium . Involucrin has been observed in suprabasal layers as is consistent with a role in differentiation . TG1 has been identified in the corneal epithelium, mostly in suprabasal cells and in the uppermost keratocyte layer, where it is found in association with fibrillin-containing microfibrils . The relatively higher expression level of EGFR in the monolayer system compared to the other markers may reflect its expression in the basal, less differentiated cell types.
The 3-D corneal cultures have been validated against the Draize rabbit irritation assay [15, 16]. In a multiple test-site study, the 3-D model gave an 80% overall concordance with rabbit irritancy data with 100% sensitivity . A range in exposure times (10, 20, 30 and 60 minutes) was investigated, with a 10-minute exposure found to be the most predictive of 24-hour rabbit irritancy scores. Cytotoxicity as measured by MTT reduction correlated highly with both histopathological evaluation of the cultures and the in vivo irritancy scores .
The duration of exposure of 3-D cultures to test compounds in other studies has varied widely among researchers, with some using 5-minute exposures [13, 17] and others using 20- and 60-minute exposures . In one study demonstrating toxicity of BAC in 3-D cultures at concentrations as low as 0.005%, exposure times were an extraordinarily long 6 and 24 hours . In our assays, exposure to BAC concentrations of 0.02% for 60 minutes demonstrated no toxicity relative to the saline control. Moreover, these durations far exceed those expected in human eyes, for which mean ocular surface residence times have been shown to be as low as 2.9 seconds .
In addition to the studies in the 3-D corneal cultures, we conducted a 1-year corneal pachymetry study in monkeys that demonstrated no effects of suprapharmacologic doses (67 times the recommended clinical dose of latanoprost and 4 times the associated dose of BAC) on corneal thickness. Corneal pachymetry has been shown to be a predictive measure of ocular irritation resulting in edema in the corneal epithelium . The absence of corneal thickness changes, irritation or histological changes implies a lack of a clinically significant or biologically functional effect.
The results of our assays demonstrated no significant differences in toxicity between commercially available BAC-containing latanoprost and olopatadine ophthalmic solutions and BAC-free travoprost ophthalmic solution using 3-D corneal epithelial cultures as an in vitro model of corneal cytotoxicity. Although 3-D cultures have some regenerative capacity, they lack certain components of the ocular surface that are involved in healing of corneal tissues, including the tear film, the corneal stroma and endothelium and the limbal vascular supply, which likely increases their sensitivity to toxic injury . Consistent with the results of our study, Townley and Reilly  recently reported that when patients experiencing ocular dryness and irritation while receiving latanoprost with BAC for at least 1 month were randomized to receive latanoprost in one eye and travoprost without BAC in the other eye, significantly more corneal staining was found in the eyes receiving travoprost (p = 0.025). In addition, the eyes receiving BAC-free travoprost showed a trend toward more dryness.