Mitochondrial DNA sequences in ancient Australians: Implications for modern human origins

The Ancient mtDNA Sequences Are Authentic.

We attempted to isolate mtDNA from six individuals from Willandra Lakes (LM in Fig. 1A) and six individuals at KS. We were not able to isolate DNA from two of the samples from the Willandra Lakes region. These samples were from cremation burials. We were successful with the 10 other samples from these two Pleistocene/Holocene deposits (Fig. 1A).

In isolating DNA from ancient bone samples, a major concern is the authenticity of any recovered sequence. Because PCR technologies are used, extreme care has to be taken to avoid the amplification of contaminating sequences that may, for example, be contributed by the investigators associated with the samples.

We are confident that the sequences we have obtained from the ancient bone samples are authentic. In each case we made two independent DNA isolates and carried out at least three independent PCR amplifications from each individual, with at least one from each of the two independent isolates, and obtained the same sequences. We used PCR primers to produce overlapping segments in the amplification reactions. In each case the sequences of the overlapping regions were identical for any given individual, indicating that the amplifications were derived from the same DNA source (Table 1). Each individual produced a different sequence in the region of mtDNA that we analyzed. All of the sequences differed from the sequence of the mtDNA segments of the two individuals handling the bone samples, and each of the sequences differed from an invariant mitochondrial insert sequence that occurs in the nuclear genome of many living humans (49).

The isolated DNA segments were short, as expected from degraded mtDNA. None of the samples from the ancient bones was capable of generating the entire 354-bp segment defined by the two terminal primers. In contrast, amplification of this same segment was achieved readily with the mtDNA isolated from the two individuals carrying out the DNA amplifications. To guard against amplification artifact, we sequenced the PCR products directly. Rare contamination was recognized readily and excluded from the analysis (see Methods).

We conclude that the mtDNA sequences we report from each of the 10 ancient bone samples are derived from the excavated remains.

The Ancient mtDNA Sequences Do Not Differentiate Gracile and Robust Morphologies.

Direct inspection of the alignment of the ancient Australian sequences with mtDNA sequences from living Aboriginal Australians (Fig. 1B) shows that the ancient sequences differ from the Cambridge Reference Sequence at between two (LM55) and 10 (LM3) nucleotide sites (Table 1). Two sequence sites (230, 278) separate the KS8, LM3, Feldhofer, and Insert sequences from all others, suggesting that two of the ancient Australian sequences (KS8 and LM3) diverged before the MRCA of sequences in living Aboriginal Australians (Table 1). Among the ancient sequences, three sites (263, 290, 355) indicate a separation of the LM3 and Insert sequences from all others (including KS8), suggesting that LM3 and the Insert sequences belong to a distinct early lineage.

Maximum-likelihood analysis shows that branches A, B, C, and D in Fig. 1B are present in all of the 13,000 most likely trees and in the weighted consensus tree derived from these. They are also present in the weighted consensus of all 105 possible trees involving bonobo, chimpanzee, Feldhofer, KS8, Insert, LM3, and a grouping of the remaining sequences. Branches C and A have approximately twice as much relative-likelihood support as branch B. If the lengths of branches A, C, or D are set to zero, then the resultant trees are significantly less likely (P < 0.0001 for each tree) than the maximum-likelihood tree, whereas the tree with the length of B set to zero was not significantly less likely (0.2 > P > 0.1). The A, B, C, and D branches were found in 96%, 26%, 76%, and 37%, respectively, of the neighbor-joining trees obtained by bootstrap resampling.

These results show that, with the possible exception of KS8 and LM3, the ancient Aboriginal sequences, including those from individuals with both robust and gracile morphologies, are within a clade that includes the sequences of living Aboriginal Australians, and that they therefore diverged after the MRCA of contemporary Aboriginal sequences. mtDNA lineages fail to differentiate individuals with clearly distinct morphologies.

Branch C indicates that the LM3 and Insert sequences belong to separate lineages. The existence of branch D indicates that these lineages diverged before the node that marks the MRCA of the living Aboriginal sequences (asterisk in Fig. 1B). There is relatively little support for branch B and therefore for a separate lineage leading to the KS8 sequence. We confirmed the grouping of LM3 and the Insert sequences relative to a larger, global sample of contemporary human sequences by likelihood mapping. The tree grouping the LM3 and the Insert sequences was the most likely in 100% of 140,000 quartets that included the LM3 and Insert sequences combined with two sequences sampled at random from the 3,453 human sequences. In all 140,000 comparisons there was absolute support for this grouping.

The LM3 Sequence Belongs to an Early Diverging mtDNA Lineage.

The divergence of the LM3 sequence before the MRCA of contemporary human sequences is indicated by its grouping with the Insert sequence (Fig. 1B), which other reports have suggested diverged before the MRCA of sequences in living humans (49, 53). Expansion of the maximum-likelihood analysis to include the Mezmaiskaya Neandertal sequence and sequences representing the major African lineages (53, 54) further confirmed the grouping of LM3 with the Insert sequence. The grouping was found in all trees with high likelihood values for all parameter configurations. The grouping of these two sequences would occur if the LM3 sequence were an allelic variant of the Insert sequence. This possibility can be ruled out because two of the primers (HA1 and HA2) used to amplify the LM3 DNA are complementary to regions of the mitochondrial genome that are not present in the Insert, and because there are 14 nucleotide differences between the regions sequenced for LM3 and the Insert. These differences are more than would be expected between two alleles at a locus in the human nuclear genome. Previous sequence analysis revealed no allelic variation of the Insert individuals from different parts of the world (49).

In the phylogenetic analysis, trees in which the LM3/Insert lineage branched before the MRCA of contemporary human sequences were not significantly more likely than trees in which this lineage diverged after the MCRA of contemporary human sequences. The branching order, and the basal topology of the tree generally, are highly sensitive to parameter configurations, and they could not be established with confidence. Median-joining network analysis of the ancient Australian and the Neandertal sequences with the representative contemporary African sequences also resulted in uncertainty about the basal topology of the tree. In the minimum-spanning tree the numbers of changes at sites 16,278, 16,311, 16,129, 16,230, 16,093, 16,189, 16,223, and 16,262 were 10, 9, 7, 7, 5, 5, 5, and 5, respectively.

Although this analysis did not reliably establish an early divergence of the LM3/Insert lineage, it demonstrated that the lineage is unusually long. We confirmed the latter conclusion by comparing the distribution of pairwise differences between the LM3 and 3,453 contemporary human mtDNA sequences, the distribution between the Insert sequence and the same sample of contemporary sequences, and the distribution of differences among the 3,453 contemporary sequences (Fig. 2A). The range of differences between the LM3 sequence and the contemporary sequences (6–21, mode = 12) is at the upper end of the range of differences among contemporary sequences (0–23, mode = 6). The range of differences between the Insert sequence and the contemporary sequences (16–28, mode = 21) extends well beyond the range of differences among contemporary sequences, indicating either that the Insert has evolved faster than sequences in the mitochondrial genome or that the LM3/Insert lineage diverged earlier. The first possibility is unlikely because in mammals the nuclear genome evolves much more slowly that the mitochondrial genome, and because a high rate of substitution at the Insert locus would be associated with a high level of sequence diversity within human populations, which is not observed (49). There is also no indication of an accelerated rate of evolution in other nuclear genome inserts from the mitochondrial genome (57). The more likely explanation is that the lineage leading to the Insert and LM3 sequences diverged before the MRCA of living human mtDNA sequences.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

(A) Distribution of pairwise differences (≈6 × 10⁶ comparisons) of mtDNA sites 12089–16387 among 3,453 contemporary human sequences from the mtDNA database (43), between these sequences and the Insert and LM3 sequences. The distributions are normalized to sum to 100. (B) Likely phylogeny of mtDNA sequences in ancient and contemporary humans.

Implications for Human Origins.

LM3, whose sequence is reported here, is the oldest individual dated accurately, and possibly the oldest human, from which DNA has been successfully recovered and analyzed. We conclude that his mtDNA and the Insert sequences form a monophyletic group relative to all of the other human sequences that probably diverged before the MRCA of living human mtDNA sequences (Fig. 2B). This conclusion implies that the most divergent known mtDNA lineage from an “anatomically modern” human is from an Australian individual. This finding does not imply that all living people originated in Australia, any more than previously described deep lineages in Africa demand a recent origin of humans on that continent. Deep lineages in Africa and our finding of an even deeper lineage in Australia are consistent with a number of possible models of the demographic and evolutionary history of our species.

Sequences from the lineage that includes LM3's mtDNA no longer occur in human populations, except as the nuclear Insert on chromosome 11. The fact that LM3's morphology is within the range of living indigenous Australians indicates that the lineages of the alleles contributing to this gracile phenotype have survived. In contrast, the mtDNAs of the robust KS individuals belong to the contemporary human lineage. Their distinct robust morphology has not survived intact, implying that the allelic lineages of many of the genes that contribute to this phenotype have been lost.

Lack of association between the survival of nuclear and mtDNA lineages is expected because they have different transmission patterns between generations. This point is emphasized by the high frequency of the Insert on chromosome 11 in many human populations (49). Despite having the Insert, none of these populations have the LM3/Insert mtDNA lineage from which the Insert must originally have been transferred. There must have been genetic exchange between people with mtDNA genomes from the LM3/Insert lineage and those with the contemporary lineage. Similar exchanges between people with other Pleistocene mtDNA lineages, like that of the Feldhofer Neandertal individual, may have occurred.

The genealogies of mtDNA sequences in most human populations, including Aboriginal Australians (41), characteristically have very little hierarchical branching structure. This pattern of sequence variation is consistent with a population expansion following a population bottleneck (58) and is generally taken as supporting the recent out of Africa model. Under this model, all contemporary sequences spread globally with an expanding population that replaced all other people and all other lineages. Africa has been postulated as the source of the expansion because some populations in Africa have more sequence diversity than populations anywhere else.

Our data present a serious challenge to interpretation of contemporary human mtDNA variation as supporting the recent out of Africa model. A separate mtDNA lineage in an individual whose morphology is within the contemporary range and who lived in Australia would imply both that anatomically modern humans were among those that were replaced and that part of the replacement occurred in Australia.

An alternative explanation is that the LM3/Insert mtDNA lineage was replaced by a spread of the “contemporary” mtDNA lineage through late Pleistocene human populations under directional selection pressure. A selective sweep of the contemporary mtDNA lineage through the early Australians after initial colonization is consistent with the pattern of mtDNA variation in Aboriginal populations (41). It explains the divergent LM3/Insert lineage in skeletally gracile people without implying the replacement of these people by a later wave of immigrants from Africa. If the mtDNA lineage was replaced as a result of selection for a new mtDNA sequence it would not follow that the lineages of nuclear genes would have changed. The lineages would have been retained and may well be represented in contemporary indigenous Australians. LM3 and his contemporaries, as well as the more recent robust KS individuals, all could have been ancestors of living indigenous Australians.

Our success in isolating and analyzing mtDNA from several late Pleistocene and early Holocene individuals should encourage further attempts to recover mtDNA from ancient human remains. Additional ancient mtDNA sequences may reveal other distinct mtDNA lineages in Pleistocene human populations. Our analysis has shown that anatomical features and the mtDNA of particular individuals may have different evolutionary paths, and some nuclear gene lineages have genealogical and/or geographical patterns that are different from those of mtDNA (8–10, 14). This difference limits the use of ancient DNA in tracing human evolutionary history because, in most cases, only mtDNA can be isolated and analyzed from ancient material. A fuller understanding of the genetic basis of recent human evolution will require more extensive investigation of nuclear genome variation.