Variability of expression of the major glutathione S-transferases (GSTs) of liver, GSTA1 and GSTA2, is thought to affect the efficiency of detoxification of xenobiotics, including chemical carcinogens. Polymorphism of the GSTA1 regulatory sequence determines some of the variation of hepatic GSTA1 expression, but the polymorphisms in GSTA2 (exons 5 and 7) were not thought to affect GSTA2 activity. By examining GST protein expression for a set of human liver and pancreas samples (coupled with a cloning/polymerase chain reaction-restriction fragment length polymorphism strategy), we identified a novel substitution Pro110Ser (328C>T) and the corresponding novel variant GSTA2*E (Ser110Ser112Lys196Glu210), and confirmed the presence of variants GSTA2*A (Pro110Ser112Lys196Glu210), GSTA2*B (Pro110Ser112Lys196Ala210) and GSTA2*C (Pro110Thr112Lys196Glu210). GSTA2*C occurred at 30–60% (i.e. approximately 100-fold more frequent than previously reported) and GSTA2*E occurred (heterozygous) at approximately 11%. Hepatic expression of the Ser112 variants (GSTA2*A, GSTA2*B or GSTA2*E) was approximately four-fold higher than that of the Thr112 variant (GSTA2*C). Compared to any other variant, GSTA2E had lower rates of catalysis towards 1-chloro-2,4-dinitrobenzene (CDNB), 4-vinylpyridine, and cumene-, t-butyl- and arachidonic acid hydroperoxides, although kcat/Km for CDNB were similar for all four variants. Using a prostate cancer case–control population, it was found that GSTA1*A/GSTA2 C335 and GSTA1*B/GSTA2 G335 were in linkage disequilibrium in Caucasians but not in African–Americans. However, there were no significant differences in the distribution of these polymorphisms or resultant haplotypes by case status. Nevertheless, the rare genotypes, GSTA2*E/*E and GSTA1*B/*B + GSTA2*C/*C (potential low GSTA2 activity and low hepatic GSTA1 and GSTA2 expression, respectively) could increase the risk of adverse effects of xenobiotics via compromised efficiency of detoxification.