Validation of an in vitro cell culture model for predicting the clinical phototoxic potential of fluoroquinolone antibiotics

Nicola J. Traynor, Brian E. Johnson, *Neil K. Gibbs and James Ferguson

Photobiology Unit, Ninewells Hospital and Medical School, Dundee, Scotland DD1 9SY.

*Corresponding author: Neil K. Gibbs Tel: (44) 1382 632240 Fax: (44) 1382 646047 e.mail: n.k.gibbs@dundee.ac.uk


ABSTRACT

Many therapeutic drugs induce phototoxic skin responses following exposure to ultraviolet radiation (UVR). Although several in vitro model systems have been developed to predict drug phototoxicity, none have been validated against data from controlled clinical phototoxicity studies. Eight fluoroquinolone (FQ) antibiotics (ciprofloxacin, grepafloxacin, lomefloxacin, norfloxacin, ofloxacin, trovafloxacin, BAYy3118, BAY12-8039) were used to determine whether their phototoxic ability in an in vitro mammalian cell culture model, was indicative of their ability to photosensitise human skin in vivo. The in vitro phototoxicity of the FQ was determined by exposing V79 Chinese hamster fibroblasts to UVA radiation (0-11.25 J/cm2 glass-filtered Sylvania FR7T12 fluorescent source) in the presence of FQ (0-100 ug/ml) in complete medium. Cell damage was quantified with the MTT or neutral red assays and an in vitro phototoxic index calculated (PI = % cells damaged with FQ + UVA / % cells damaged with UVA alone ) for each end-point at 10 ug/ml FQ and 11.25 J/cm2 UVA. The in vivo photosensitising ability of the FQ was investigated using double-blinded, placebo and positive controlled, clinical skin phototesting of normal subjects. Minimal erythemal doses (MED) at 36530nm were determined before and after 6-7 days FQ ingestion. In vivo PI were calculated for the eight FQ and plotted against their respective in vitro PI from the MTT and NR assays. Linear regression analysis of the in vivo PI vs in vitro PI plot revealed that data from the MTT and NR end-points gave highly significant linear fits (p=0.0002 and p=0.0003 respectively) slopes of 1.6 and 1.5 respectively, correlation coefficients of 0.96 and 0.93 respectively, and R-squared values of 92% and 86% respectively. To our knowledge, this is the first occasion that an in vitro model designed to predict a particular toxicological response in humans has been directly validated against controlled clinical data. We have used eight FQ with a range of clinical PI (1-32) to demonstrated that a simple cell model and standard NR and MTT assays are good indicators of clinical FQ-induced phototoxic potential. The model does not offer any mechanistic information on FQ phototoxicity and it is not known whether it would be useful for other drug groups. The ability to predict potential drug phototoxicity at early stage of drug development is beneficial to the pharmaceutical industry and should reduce the risk of patients having a phototoxic episode during later phase clinical trials.

Abbreviations: DMSO, dimethyl sulphoxide, FQ, fluoroquinolone; MED, minimal erythemal dose; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide; PBS, phosphate buffered saline; UVR, ultraviolet radiation;