When modern humans spread out of Africa and the Near East some 75,000–50,000 years ago, at least two archaic hominin groups, Neanderthals and Denisovans, inhabited Eurasia. While Neanderthals are known from an abundant fossil record in Europe and western and central Asia, Denisovan remains are currently only known from the Altai Mountains in southern Siberia3, 4. However, Denisovan ancestry is detected in present-day human populations from Oceania, mainland Asia and in Native Americans5, suggesting that they were once more widespread. High-quality genome sequences recovered from one Neanderthal and one Denisovan show that they were more closely related to each other than to modern humans6, 7 and that they diverged from a common ancestral population between 381,000 and 473,000 years ago7 if a mutation rate of 0.5?×?10?9 per site per year is used.
The Middle Pleistocene fossils from Sima de los Huesos (SH) are relevant for the question of when and where the ancestral populations of Neanderthals and Denisovans lived, but their relationship to these later archaic groups is unclear. They share some derived dental and cranial features with Late Pleistocene Neanderthals, for example, a midfacial prognathism and some aspects of the supraorbital torus, the occipital bone and the glenoid cavity1, 8. In apparent contrast to this, the mitochondrial (mt)DNA determined from one SH individual is more similar to an mtDNA ancestral to Denisovan than to Neanderthal mtDNAs2. However, the mtDNA is inherited as a single unit from mothers to offspring and does not necessarily reflect the overall relationship of individuals and populations. To clarify the relationships of the SH hominins to Neanderthals and Denisovans, we therefore set out to retrieve nuclear DNA from SH hominins. However, DNA preservation in these fossils is poor owing to their great age. Femur XIII, from which the SH mtDNA genome was sequenced, contains only small amounts of highly degraded endogenous DNA (30–45 base pairs (bp)) in a large excess of microbial DNA. To reconstruct its mtDNA genome, almost 2?g of bone had to be used to produce DNA libraries from which mtDNA fragments were isolated by hybridization capture. Furthermore, because of the presence of modern human DNA contamination, putatively endogenous sequences had to be identified on the basis of the presence of C to T substitutions that accumulate at the ends of DNA fragments over time owing to cytosine deamination9, which are largely absent in recent human DNA that contaminates fossils10, 11.
To retrieve nuclear DNA sequences from femur XIII, we generated approximately 2.6 billion sequence reads from the library with the highest frequency of terminal C to T substitutions (library A2021 (ref. 2)). In addition, between 600 million and 900 million reads were collected from each of four new specimens that were recovered from the site for molecular analyses (Extended Data Table 1). These were an incisor (AT-5482), a femur fragment (AT-5431), a molar (AT-5444) and a scapula (AT-6672).
In addition to sequencing random fragments from these specimens, we also isolated mtDNA fragments from the four new specimens by hybridization capture. Between 1,419 and 3,742 unique mtDNA fragments of 30?bp or longer were retrieved (Extended Data Table 2). To investigate whether they represented endogenous DNA or present-day human contamination, we determined the frequency of C to T substitutions relative to the human mitochondrial genome at each position in the fragments. The fragments from femur AT-5431 carry 44% C to T substitutions at the 5′ ends and 41% at the 3′ ends, compatible with the presence of endogenous ancient mtDNA. The other three specimens do not show discernible evidence of deamination-induced substitutions. Because there were too few DNA fragments to reconstruct the complete mtDNA genome of femur AT-5431, we restricted further analyses to ‘diagnostic’ positions in the mtDNA genome where each lineage in the mtDNA tree differs from the other hominin lineages and from the chimpanzee. At positions where modern humans differ from Neanderthals, Denisovans, SH femur XIII and the chimpanzee, 41% (17 out of 41) of the mtDNA fragments share the modern human state, indicating that they are derived from present-day human contamination (Extended Data Fig. 1). In contrast, among the five fragments that show evidence for deamination, none shares the human-derived state. Among the eight putatively deaminated fragments that overlap positions with variants specific to Denisovans and to femur XIII, all eight carry the derived variants present in both lineages. In addition, three out of five fragments carry variants specific to femur XIII only. In contrast, of nine mtDNA fragments overlapping positions diagnostic for Neanderthals, none carry the Neanderthal variants. We thus conclude that the mtDNA of femur AT-5431 is most closely related to the mtDNA of femur XIII.
The vast majority of endogenous DNA in the SH fossils is degraded to a size below 45?bp. To maximize the yield of DNA fragments we have therefore used fragments as short as 30?bp when reconstructing mtDNAs from these specimens2, 12, whereas DNA analyses from other archaic hominins have been restricted to fragments of 35?bp or longer6, 7. We explored whether it might be possible to use DNA fragments as short as 30?bp to study also the nuclear genome in SH specimens; however, this resulted in 9–67% of the aligned DNA fragments appearing to be of microbial rather than of hominin origin. In contrast, no spurious alignments were detected at a length cut-off of ≥35?bp (Supplementary Information, section 1). Using the latter cut-off, between 0.02% and 0.25% of the DNA sequences determined from the fossils align to the human reference genome (Table 1). With the exception of the scapula, all specimens show C to T substitution frequencies between 5% and 22% at the terminal alignment positions. When conditioned on C to T substitutions at the other ends of fragments, they increase to between 53% and 64% for the incisor, the femur fragment and the molar, to 33% and 42% for femur XIII, and to 4% and 10% for the scapula (Table 1 and Extended Data Fig. 2), indicating that all five specimens carry mixtures of highly deaminated endogenous nuclear DNA and less deaminated human contamination (Supplementary Information, section 2).
Owing to the extremely small amounts of data available, assessment of nuclear DNA contamination cannot be achieved using existing approaches that require multi-fold coverage in at least parts of the nuclear genome13, 14. We therefore used two alternative approaches to obtain estimates of contamination for the nuclear sequences (Supplementary Information, sections 2 and 3). The first approach, which compares deamination signals in all sequences to those carrying a C to T substitution at the opposing end, estimates human contamination to be >63% in all five specimens. The second approach estimates contamination as the percentage of sequences sharing the modern human state at sites where 90% or more of present-day humans differ from the chimpanzee and the two high-coverage archaic genomes. The contamination estimates from this approach are similarly high (Table 1), but decrease to 21% or less in three of the specimens (femur AT-5431, the incisor and the molar) when filtering for sequences showing terminal C to T substitutions indicative of deamination. Disregarding fragments without evidence for deamination, the amount of nuclear DNA sequence retrieved varies between 189?kb and 2.0?Mb for these three specimens (Table 1), all of which are male as inferred from the sequence coverage of chromosome X and the autosomes (Extended Data Fig. 3). The sex of the scapula and femur XIII cannot be confidently determined as a result of the high levels of present-day human contamination and the limited amount of data available.
To investigate how the SH hominins are related to modern humans, Neanderthals and Denisovans, we used the high-quality genome sequences of the Altai Neanderthal, the Denisovan finger bone and a present-day human individual from Africa (Mbuti, HGDP00982) to identify positions where one or more of these three genomes differ from those of the chimpanzee and other primates (bonobo, gorilla, orangutan, rhesus macaque) (Extended Data Fig. 4). We then estimated the percentages of all informative positions covered in each specimen that share the derived state for each branch in the tree relating the three genome sequences. For the femur fragment AT-5431 and the incisor, we find that 87% and 68%, respectively, of the positions on the common Neanderthal and Denisovan branch carry derived alleles; that 43% and 39%, respectively, of positions on the Neanderthal branch carry derived alleles; while 9% and 7%, respectively, on the Denisovan branch do so (Fig. 1). This indicates that the SH hominins are related to the ancestors of Neanderthals rather than Denisovans. The fraction of derived alleles shared with the Neanderthal genome is between two- and threefold smaller when sequences without terminal C to T substitutions are also included (Extended Data Table 3), confirming that the signal linking the SH hominins to Neanderthals is derived from endogenous DNA fragments. These results are stable when present-day human individuals other than the Mbuti are used in the analysis (Extended Data Table 4). Similar to femur AT-5431 and the incisor, the molar also shows a greater sharing of derived alleles with the Neanderthal than the Denisovan genome, although not statistically significantly so, probably because of the small amount of data available (Extended Data Fig. 5). By comparison, the fraction of derived alleles shared with the Neanderthal genome for several Late Pleistocene Neanderthals sequenced to low-coverage7, 13 is between 69% and 75% (Fig. 2). Thus, DNA sequences of the SH hominins diverged more than twice as far back along the lineage from the Altai Neanderthal genome to its ancestor shared with the Denisovan genome than DNA sequences of the Late Pleistocene Neanderthals from Europe and the Caucasus.
Ninety-five per cent binomial confidence intervals (CI) are indicated. The thickness of the branches is scaled by the extent of derived allele sharing. See Extended Data Fig. 4 for the total number of informative positions identified in the nuclear genome and Extended Data Table 3 for the number of sequences overlapping these positions.
Because nearly 30 hominin skeletons have been found in SH, it is likely that the specimens analysed here belong to different individuals. The nuclear DNA sequences of femur AT-5431 and the incisor show that they belonged to the Neanderthal evolutionary lineage, and the limited data available for the molar suggest that the same is true for this specimen. Thus, the results show that the SH hominins were early Neanderthals or closely related to the ancestors of Neanderthals after the divergence from a common ancestor shared with Denisovans. Although it is difficult to determine the age of Middle Pleistocene sites with certainty, geological dating methods1, as well as the length of the branches in trees relating the mtDNAs from femur XIII and an SH cave bear to other mtDNAs2, 12, suggest an age of around 400,000 years for the SH fossils. This age is compatible with the population split time of 381,000–473,000 years ago estimated for Neanderthals and Denisovans on the basis of their nuclear genome sequences and using the human mutation rate of 0.5?×?10?9 per base pair per year7. This mutation rate also suggests that the population split between archaic and modern humans occurred between 550,000 and 765,000 years ago. Such an ancient separation of archaic and modern humans is difficult to reconcile with the suggestion that younger specimens often classified as Homo heidelbergensis, for example Arago or Petralona, belong to a population ancestral both to modern humans and to Neanderthals15.
We further note that the SH hominins carry mtDNAs more closely related to those of Denisovans in Asia than Neanderthals, even though their nuclear genomes show that they are more closely related to Neanderthals. We have previously speculated that this discrepancy may be because the SH hominins carried two very divergent mtDNA lineages or that another hominin group contributed mtDNA both to the SH hominins and to Denisovans2. However, given that the SH hominins are early Neanderthals (or closely related to these), and assuming that the mtDNA they carried was typical of early Neanderthals, an additional possibility that appears reasonable is that the mtDNAs seen in Late Pleistocene Neanderthals were acquired by them later, presumably because of gene flow from Africa. It is possible that contacts between Africa and western Eurasia occurred in the Middle Pleistocene as indicated, for example, by the appearance of the Acheulean hand axe technology in Eurasia by 500,000 years ago16 and by the spread of the so-called ‘Mode 3’ technology around 250,000 years ago17. Gene flow from Africa may perhaps also explain the absence of Neanderthal-derived morphological traits in some Middle Pleistocene specimens in Europe such as Ceprano and Mala Balanica18, 19. Retrieval of further mtDNAs and, if possible, nuclear DNA from Middle Pleistocene fossils will be necessary to comprehensively address how Middle and Late Pleistocene hominins in Eurasia were related to each other.
联系客服
