High fidelity of the polymerase chain reaction.

Abstract

The Thermus aquaticus ( Taq) polymerase chain reaction (PCR) is a rapid means for DNA amplification, with many applications both in population and evolutionary biology, and in other fields. We report here the results of an experiment that tests the fidelity of the reaction. We have cloned and sequenced the products of six PCR amplifications of a 2.5-kb fragment containing the SodcAl gene of Drosophila melanogaster. We have observed four substitutions and two single-nucleotide frameshifts. The observed cumulative error frequency is 3.0 X 1O-4 for substitutions and 1.5 X lop4 for frameshifts. The error.rates per nucleotide polymerized are estimated as 1.5 X lop5 and 0.7 X 10e5 for substitutions and frameshifts, respectively. The Taq polymerase is the enzyme of choice for PCR because it enables the amplification reaction to proceed at higher temperatures, which improves the specificity, yield, sensitivity, and length of products that can be amplified (Saiki et al. 1988). The fidelity of the reaction is, however, a cause of concern, since the Taq polymerase has no detectable 3’ + 5’ proofreading exonuclease activity (Tindall and Kunkel 1988). The Sod gene codes for CuZn superoxide dismutase (SOD). The SodcAl gene is a “null” allele, isolated from a natural population, that yields -5% of the wild-type level of the SOD protein (Graf and Ayala 1986). The reduction in enzyme level is associated with the insertion of a 0.68 kb-truncated P element within the gene, 47 bp downstream from the start of transcription (fig. 1) . We have previously isolated and sequenced a 2.5-kb EcoRI fragment containing the SodCA’ gene from an EMBL3 genomic library (Kwiatowski et al. 199 1). We have now amplified this fragment by six separate PCR reactions and have cloned in a pUC19 plasmid and sequenced one fragment from each reaction. We have sequenced a total of 13,513 bp ( 1,658 bp from one clone, 2,293 from another, 2,388 from each of two more, and 2,393 from each of the last two). Compared with the sequence previously obtained from directly cloned genomic DNA, two of these sequences exhibit no differences, the third exhibits a I-bp deletion (at about 505), the fourth a transition (G-A at 61 l), the fifth a transversion (T+A at 2322), and the sixth two transitions (T-C at 359 and G+A at 2076 ) and a 1 -bp insertion (at about 540) (fig. 2). The frameshifts are both associated with oligomers of the base inserted or deleted. The cumulative frequency of observed errors per nucleotide is 3.0 X 10 -4 substitutions and 1.5 X 10 -4 frameshifts. The error frequency observed is lo-30-fold lower than the incidence of nucleotidesite differences typically observed between randomly selected alleles from the same population [e.g., the estimated heterozygosity per nucleotide site in D. pseudoobscura is 0.97 X lo-* at the Xdh locus (Riley et al. 1989) and 0.84-1.17 X lo-* at the Adh locus (Schaeffer and Miller 199 1 )] . Whether this degree of error is acceptable as part of the experimental error in a particular study will depend on the objectives sought.