The impact of air pollution exposure on the level of total DNA adducts in human white blood cells (WBCs) was evaluated in two populations in the Czech Republic and compared to the exposure-DNA adduct relationship in other populations in the US and China in human lung cells and rodent lung tissue. The human populations examined were exposed to respirable particles (< 2.5 microm) (PM2.5) in urban, rural, and occupational settings where the particles originated from coal and petroleum fuel combustion, coke production, and other coal-tar aerosols (e.g., used in aluminum production). These particles contain carcinogenic polycyclic aromatic hydrocarbons (PAHs) that are known to form DNA adducts through covalent binding. Personal exposure to PM2.5 and PAHs were measured prior to collection of blood samples for DNA adduct analysis by 32P-postlabeling. Coke oven workers (n = 76), in 10 job categories on the top and side of a coke oven in Ostrava, CZ, were studied and compared to a different population exposed to environmental levels of PAHs from air pollution in Teplice, CZ. Personal exposures to airborne particles ranged from < 1 to more than 15,000 microg/m3 and carcinogenic PAHs exposure ranged from < 5 to > 200,000 ng/m3. At low to moderate environmental exposures to carcinogenic PAHs, DNA adduct levels in the WBCs were significantly correlated with exposure. However, at the higher occupational levels found on the coke oven, the exposure-DNA adduct relationship became non-linear. Under these high exposure conditions, the relative DNA adduct level per unit of exposure (DNA-binding potency) was significantly lower than measured at environmental exposures. This finding is consistent with observations in lung cells from bronchoalveolar lavage of humans exposed to a wide range of PAH. This same high exposure-dose non-linearity was also observed in lung DNA from rats exposed by inhalation to a coal-tar pitch aerosol. DNA adduct levels in all these cases show evidence of a form of non-linearity at high doses that has been described by Lutz (W.K. Lutz, Dose-response relationship and low dose extrapolation in chemical carcinogenesis, Carcinogenesis, 11 (1990) 1243-1247) as a superlinear dose response. This superlinear response may be due to saturation of metabolic activation enzymes, induction of either DNA repair processes or detoxification enzymes, or other mechanisms. Regardless of the mechanism, this decrease in the DNA-binding potency at moderate to high doses of PAH has important implications for dose-response extrapolation in risk assessment.