Statistical evidence for miscoding lesions in ancient DNA templates.

It is generally believed that sequence heterogeneity in PCR products from fossil remains are due to regular DNA polymerase errors as well as miscoding lesions compounded by damage in the template DNA (Pääbo 1990; Handt et al. 1994b, 1996; Höss et al. 1996; Krings et al. 1997). However, it has been difficult to test the frequency with which this assumption holds. First, DNA extractions from fossil remains rarely produce a yield large enough for pre-PCR analysis of postmortem modifications (Höss et al. 1996). Second, in most cases, it is not possible to determine whether nucleotide misincorporations by the DNA polymerase enzyme during amplification are caused by regular DNA polymerase errors or miscoding lesions in the template DNA sequences (Greenwood et al. 1999). Finally, the error rates of the DNA polymerase enzymes for PCR have proved to be highly unpredictable, making it difficult to account for regular DNA polymerase errors in amplified DNA sequences (Eckert and Kunkel 1991). Here, we present a statistical model for analyzing PCR-mediated base-misincorporations, catalyzed by the commonly used Thermus aquaticus (Taq) polymerase enzyme, in amplification products from fossil remains. The error rate of the Taq polymerase enzyme may vary more than 10-fold (;2 3 1024 to ,1 3 1025 per nucleotide per cycle) according to the precise DNA sequence and the in vitro conditions of DNA synthesis (Eckert and Kunkel 1991). Therefore, the tests of the model rely solely on the relative distribution of the distinct Taq polymerase errors, which, in contrast to the highly variable error rate, is nearly constant and independent of the starting template material and the conditions for the PCR, as shown in table 1. Hence, the tests are not affected by variations in PCR efficiencies and accuracy. The model compares the distribution of the regular Taq polymerase errors with the observed substitutions in amplification products from fossil remains under the hypothesis that any significant differences between the distributions are due to miscoding lesions in the template DNA sequences used for PCR. The model was applied to published multiple clone sequences of the mitochondrial (mt) DNA control region from three differently preserved specimens of Homo representing different ages: a ;600-year-old Hokokam Indian (VC15A) found in a cave in Arizona, southwestern United States (Handt et al. 1996), the ;5,000-