Arsenic is an established human carcinogen. However, there has been much controversy about the shape of the arsenic response curve, particularly at low doses. This controversy has been exacerbated by the fact that the mechanism(s) of arsenic carcinogenesis are still unclear and because there are few satisfactory animal models for arsenic-induced carcinogenesis. Recent epidemiological studies have shown that the relative risk for cancer among populations exposed to <or=60 ppb As in their drinking water is often lower than the risk for the unexposed control population. We have found that treatment of human keratinocyte and fibroblast cells with 0.1 to 1 microM arsenite (As(III)) also produces a low dose protective effect against oxidative stress and DNA damage. This response includes increased transcription, protein levels and enzyme activity of several base excision repair genes, including DNA polymerase beta and DNA ligase I. At higher concentrations (> 10 microM), As induces down-regulation of DNA repair, oxidative DNA damage and apoptosis. This low dose adaptive (protective) response by a toxic agent is known as hormesis and is characteristic of many agents that induce oxidative stress. A mechanistic model for arsenic carcinogenesis based on these data would predict that the low dose risk for carcinogenesis should be sub-linear. The threshold dose where toxicity outweighs protection is hard to predict based on in vitro dose response data, but might be estimated if one could determine the form (metabolite) and concentration of arsenic responsible for changes in gene regulation in the target tissues.