Patterns of nucleotide misincorporations during enzymatic amplification and direct large-scale sequencing of ancient DNA

Abstract

Whereas evolutionary inferences derived from present-day DNA sequences are by necessity indirect, ancient DNA sequences provide a direct view of past genetic variants. However, base lesions that accumulate in DNA over time may cause nucleotide misincorporations when ancient DNA sequences are replicated. By repeated amplifications of mitochondrial DNA sequences from a large number of ancient wolf remains, we show that C/G-to-T/A transitions are the predominant type of such misincorporations. Using a massively parallel sequencing method that allows large numbers of single DNA strands to be sequenced, we show that modifications of C, as well as to a lesser extent of G, residues cause such misincorporations. Experiments where oligonucleotides containing modified bases are used as templates in amplification reactions suggest that both of these types of misincorporations can be caused by deamination of the template bases. New DNA sequencing methods in conjunction with knowledge of misincorporation processes have now, in principle, opened the way for the determination of complete genomes from organisms that became extinct during and after the last glaciation.

Fig. 1.

Substitution pattern observed among the misincorporations seen in the PCR products from 29 ancient wolves (Left) and among modern dog and wolf mtDNA sequences (Right) from Savolainen et al. (33) (dogs) and Vila et al. (31, 32) (wolves). Numbers represent percentages, and the “ancestral” sequence for the dogs and wolves is approximated with the consensus sequence for all sequences studied.

Fig. 2.

Schematic illustration showing how 454 sequencing captures individual single-stranded mammoth molecules. When these are compared with elephant sequences, any reciprocal base-equivalent differences between the genomes will be present in statistically equal numbers across many reads. This is illustrated by the two G–C and C–G differences shown in green. Any modification in the ancient DNA, for example a deaminated C that yields U (orange), will yield an excess of differences relative to the reciprocal difference, as seen for the T–C vs. A–G differences below.

Fig. 3.

Number of base mismatches of each type, grouped into strand-equivalent pairs for the mammoth and elephant comparisons (Upper) and the human comparisons (Lower). The circle indicates the observed number of mismatches of each type. The lines above and below are the 99% confidence intervals obtained for 10,000 bootstrap replicates of the alignments. Note that positions where the mammoth carries a T and the elephant carries a C are significantly more numerous than positions where the mammoth carries an A and the elephant carries a G. Note also that the positions where the mammoth carries an A and the elephant carries a G are significantly more numerous than positions where the mammoth carries a G and the elephant carries an A. Among the 59 human sequences analyzed, there are no significantly elevated mismatches.

Fig. 4.

Summary of the oligonucleotide competition amplifications. Each bar represents the ratio of nucleotides found among clones from amplification products generated from a mixture of two oligonucleotides, given by the abbreviations below the bars. Misincorporated nucleotides are indicated by the black upper part of each bar, and numbers give the numbers of clones sequenced. The polymerase reagents are indicated as “Gold” for AmpliTaq Gold and “Platin” for Platinum Taq High Fidelity.