Reconstructing large regions of an ancestral mammalian genome in silico

Abstract

It is believed that most modern mammalian lineages arose from a series of rapid speciation events near the Cretaceous-Tertiary boundary. It is shown that such a phylogeny makes the common ancestral genome sequence an ideal target for reconstruction. Simulations suggest that with methods currently available, we can expect to get 98% of the bases correct in reconstructing megabase-scale euchromatic regions of an eutherian ancestral genome from the genomes of approximately 20 optimally chosen modern mammals. Using actual genomic sequences from 19 extant mammals, we reconstruct 1.1 Mb of ancient genome sequence around the CFTR locus. Detailed examination suggests the reconstruction is accurate and that it allows us to identify features in modern species, such as remnants of ancient transposon insertions, that were not identified by direct analysis. Tracing the predicted evolutionary history of the bases in the reconstructed region, estimates are made of the amount of DNA turnover due to insertion, deletion, and substitution in the different placental mammalian lineages since the common eutherian ancestor, showing considerable variation between lineages. In coming years, such reconstructions may help in identifying and understanding the genetic features common to eutherian mammals and may shed light on the evolution of human or primate-specific traits.

Figure 1.

Estimated reconstructability of ancestral mammalian sequences. Average base-by-base error rate in the reconstruction of each simulated ancestral sequence. The error rate shown is the sum of the percentages of bases that are missing, added, or mismatched as a result of errors in the reconstruction, averaged over 100 simulations of sets of orthologous sequences of length ∼50 kb. Error rates are given first for all regions, and in parentheses for nonrepetitive regions only. The Boreoeutherian ancestor, which is the ancestor that can best be reconstructed, is indicated by the arrow. Branches completely located inside the box are called “early branches” (see text). The species names at the leaves only indicate what organisms we simulated; no actual biological sequences were used here. The tree topology and branch lengths are derived directly from Eizirik et al. (2001).

Figure 2.

Estimated reconstructability of the Boreoeutherian ancestor. Fraction of the simulated Boreoeutherian ancestral sequence reconstructed incorrectly as a function of the number of extant species used for the reconstruction. For each number of species used, results are given counting all bases (left columns) and only nonrepetitive bases (right columns). Species are added in the following order: human, cat, chipmunk, sloth, manatee, rousette bat, mole, pig, beaver, tree shrew, horse, pangolin, mouse, armadillo, aardvark, okapi, dog, mole-rat, rabbit, and lemur.

Figure 3.

(A) Estimates of the expected number of substitutions per site between a repeat consensus C, it human descendent H, and the reconstructed ancestor A^*, based on a Kimura 2-parameter model and averaged over all human ancestral repeats of the region considered. The true ancestor A cannot be observed, but a distance of 0.026 substitutions per site between it and A^* is estimated from the three other distances. (B) Star phylogeny with n independent descendants. (C) A tree with bifurcating root. Irrevocable information loss occurs between R and its descendants A and B.

Figure 4.

Example of reconstruction of an ancestral Boreoeutherian sequence based on actual orthologous sequences derived from a MER20 retrotransposon. Arrows indicate positions where the reconstructed ancestor differs from the MER20 consensus. Longer arrows indicate the positions where the knowledge of the MER20 consensus sequence would have changed the ancestral base prediction. The position of the human sequence displayed is chr7:115,739,755-115,739,899 (NCBI build 34). The alignment of the flanking nonrepetitive DNA (data not shown) verifies that the sequences from the different species are, in fact,orthologous. The tree and branches are derived directly from Eizirik et al.(2001).

Figure 5.

Frequency of microdeletions (1–10 bp) (left) and microinsertions (right) during eutherian evolution. Indel rates for the branches shown with dashed lines cannot be accurately estimated. Estimates are based on a set of regions totaling about 280 kb, for which sequence data is available for all 19 mammals.