We investigated the accuracy of the insertion process in RNA chain elongation catalyzed by wheat germ RNA polymerase II. Error frequencies varied from 1 misinserted nucleotide per 250 polymerized correct substrates to less than 1 in 2 x 10(5), depending on template sequence and nature of the divalent metal cofactor. Higher error ratios were observed in the presence of Mn2+ compared to Mg2+, and with alternating poly[d(G-C)].poly[d(G-C)] compared to poly[d(A-T)].poly[d(A-T)]. In this latter case the eukaryotic RNA polymerase was as accurate as Escherichia coli RNA polymerase holo-enzyme. The fidelity of wheat germ RNA polymerase II was also examined in transcription of polynucleotide templates in the poly[d(G-C)] family adopting either the right-handed B or left-handed Z conformations. Error ratios for noncomplementary ATP increased markedly under experimental conditions favoring the B-to-Z conformational transition of the alternating copolymers. In accordance with the results of previous studies, the rate of productive elongation, i.e. the synthesis of poly[r(G-C)], was depressed, suggesting that the decreased accuracy of the enzyme derived from an altered competence of the enzyme to form elongation complexes on the left-handed DNA. As judged by the large difference in apparent Km values of the enzyme for complementary and noncomplementary nucleoside triphosphates, part of the discrimination between substrates seemed to take place at the initial binding step. Furthermore, the results indicate that wheat germ RNA polymerase II was able to elongate a primer with a 3'-terminal mismatch, and thus to incorporate the mismatched nucleotide stably in the nascent RAN. However, the probability of productive RNA chain elongation was much lower with noncognate than with the complementary substrates.