Mal'ta Boy (ANE): Origins of Native Americans

Figure 3

Allentoft et al.6 have shown that in the early Bronze Age the Altai region was inhabited by a genetically West Eurasian population of the Afanasievo archaeological culture, most similar to the Yamnaya culture with about 50% of ANE ancestry6, and to the modern Avars (Suppl. file S2, Fig. 3, Suppl. Figs. 9.4, 9.5). In the late Bronze Age this population apparently had being gradually admixed with Siberians, giving rise to the Karasuk culture and the later cultures of the Iron Age6. The ancient genome ‘Iron Age Russia’, carbon-dated to 721–889 AD6, is most closely related to the typical modern Siberians: Nganasans according to the outgroup f3 statistic in the original study6 and Altaians or Koryaks according to the outgroup f3 statistics and their pairwise correlations on our datasets lacking Nganasans (Suppl. file S2). However, according to various analyses, the ‘Iron Age Altai’ (dated roughly to 2900-1500 YBP) and the Karasuk (carbon-dated to 3531-3261 YBP) populations of two and six genomes, respectively, are most closely related to each other and to Kets (Suppl. files S2, S3, Suppl. Figs. 7.5, 7.6, 8.17A, 9.1). The outgroup f3 statistics (Yoruba; Karasuk, X) on both genome-based datasets selected Kets as the best hit for Karasuk (Suppl. files S2, S3, Suppl. Fig. 7.5), although statistics for Mayans, Greenlanders, Mixe, Saqqaq, Mal’ta, Iron Age Russia, and Aleutian were not significantly different (|Zdiff| score < 3). Similarly, Native American, Beringian populations and Selkups were the best hits for Iron Age Altai and Karasuk according to the outgroup f3 statistics in the original study (Kets were lacking in the dataset, see Allentoft, et al.6). Importantly, the Karasuk culture has been tentatively associated with the Yeniseian-speaking people based on the toponymic evidence5, and the Altai region is considered to be the homeland of the Yeniseian language family2. As another piece to this puzzle, we observed genetic continuity between the Kets and the ancient genomes from the Altai.



Figure 4

Mal’ta (ancient North Eurasian) ancestry in Kets
The outgroup statistic f3 (Yoruba; Mal’ta, Ket) (Raghavan et al. 2014)21 was higher than statistics for all other Siberian and most Beringian populations. However, the statistic values were not significantly different within a large group of North Eurasians, according to Zdiff scores (Suppl. Figs. 7.13–7.16). Zdiff score for f3 (Yoruba; Mal’ta, Ket) vs. f3 (Yoruba; Mal’ta, Nganasan) equaled 7.4, 7.0 vs. f3 (Yoruba; Mal’ta, Yukaghir), and only 2.7 vs. f3 (Yoruba; Mal’ta, Selkup) (Suppl. Fig. 7.13). Thus, we suggest that, unlike the other members of the Nganasan-related clade (Fig. 2), Kets and, to a lesser extent, Selkups have a high proportion of Mal’ta ancestry, alternatively referred to as the ANE ancestry22. The Mal’ta ancestry in Kets was further supported by the TreeMix39 analysis, specifically by a migration edge connecting Mal’ta to the Ket-Karasuk clade, with a weight of 43% (Suppl. Fig. 9.1, Suppl. Table 7). Taking into account the admixture coefficients for the two sequenced Ket individuals (Ket891 and Ket884, Fig. 1A), we selected Ket891 as an individual with lower values of the North European and Siberian admixture components (in the K = 19 dimensional space). In addition, Ket891 was identified as non-admixed by reAdmix analyses (Suppl. Table 3). Ket891 demonstrated a slightly closer genetic affinity to Mal’ta (compare Suppl. Figs. 7.13 and 7.14).

These results were consistent with calculations of f4 statistic in two configurations: (X, Chimp; Mal’ta, Stuttgart) or (X, Papuan; Sardinian, Mal’ta), reproducing the previously used statistics22,26 (Suppl. Figs. 8.1–8.6). f4 (X, Chimp; Mal’ta, Stuttgart) analysis tests whether the population X has more drift shared with Mal’ta or with Stuttgart (an early European farmer, EEF22). Sardinians were used as the closest modern proxy for EEF22 in f4 (X, Papuan; Sardinian, Mal’ta). All possible population pairs (X,Y) were tested by f4 (Mal’ta, Yoruba; Y, X) on the genome-based dataset, including both Ket individuals (Fig. 5A). Compared to Kets, Mal’ta was significantly closer to none of the populations including Native Americans.


Figure 5

Based on all analyses, we can tentatively model Kets as a two-way mixture of broadly defined East Asians and ANE. Therefore, ANE ancestry in Kets can be estimated, using various f4-ratios, at 27% to 43% (depending on reference populations and datasets), vs. 25–53% in various Native American groups (Suppl. Table 6, see details in Suppl. Information, Section 8). Given non-significant Zdiff scores and f4 statistics (Fig. 5A) discussed above, it is difficult to identify the exact Eurasian population west of Chukotka and Kamchatka with the highest degree of the Mal’ta ancestry, but Kets are a good candidate. We speculate that ANE component was acquired by ancestors of Kets in the Altai region, where the Bronze Age Okunevo culture was located, with a surprisingly close genetic proximity to Mal’ta6. Later, the Yeniseian-speaking people occupied this region until the 16th–18th centuries3,4.

Based on previous studies16,25,28, the Saqqaq individual, and the Paleo-Eskimos in general25 may represent a separate and relatively recent migration into America. The Paleo-Eskimos have large proportions of Beringian (i.e. Chukotko-Kamchatkan and Eskimo-Aleut), Siberian, and South-East Asian ancestry. We have also shown that Kets and Selkups belong to a group of modern populations closest to an ancient source of Siberian ancestry in Saqqaq. This group also includes, but is probably not restricted to, Uralic-speaking Nganasans and Yukaghirs (the latter speak an isolated language). Unlike the other populations of this group, Kets, and, to a lesser degree Selkups, have a high proportion of Mal’ta (ancient North Eurasian) ancestry.



As shown previously28, Chipewyans, a modern Na-Dene-speaking population, have about 16% of Saqqaq ancestry. Thus, a gene flow dated at 5,000-6,000 YBP16 can be traced from the cluster of Siberian populations to Saqqaq, and from Saqqaq to Na-Dene. However, the genetic signal in contemporary Na-Dene-speaking ethnic groups is substantially diluted. The genetic proximity of Kets to the source of Siberian ancestry in Saqqaq correlates with the hypothesis that Na-Dene languages of North America are specifically related to Yeniseian languages of Siberia, now represented by Ket language only10. However, this genetic link is indirect and requires further study of population movement and language shifts in Siberia.

In addition, we show that Kets represent a modern Siberian population closest to ancient individuals of the Karasuk culture, spanning a period from about 3400 YBP to 2900 YBP (the individuals analyzed were dated to 3531-3261 YBP), and to few investigated Iron Age individuals of the Altai region (2900-1100 YBP)6. This genetic continuity correlates with historical linguistic data suggesting that the homeland of Yeniseian languages was located in the Altai region2,5.

Scientific Reports volume 6, Article number: 20768 (2016)

An overview of population relationships
First, we sought to obtain a broad overview of population genetic relationships and to gain insight into the levels of genetic heterogeneity of the populations using principal component analysis (PCA). We combined both publicly available DNA microarray genotyping data from 892 present-day human individuals (Li et al. 2008; The 1000 Genomes Project Consortium 2012; Yunusbayev et al. 2012; Fedorova et al. 2013; Khrunin et al. 2013; Zhou et al. 2013; Raghavan et al. 2014), representing 27 diverse Asian, European, Siberian and Native American populations (Fig. 1B; Supplemental Fig. S2; Supplemental Table S1), and 82 individuals from Mansi (N = 45) and Khanty (N = 37) groups genotyped for the first time, using high-density SNP microarrays. To provide relative placement of samples that were a focus of this study, we projected their sequenced genomes onto the principal components. All the sequenced genomes clustered with genotyped samples from corresponding populations, confirming that the sequenced genomes were good representatives of these groups.


Figure 1.
Geographic and genetic affinities of European and Siberian populations. (A) Map of Siberia, Western Russia, and Eastern Europe shows geographic locations of individuals sequenced as part of this work. Colored areas of the map show geographical areas traditionally occupied by corresponding populations (Peoples of Russia 1994). The Ural Mountain range, a natural boundary between Europe and Siberia, is shown as blue peaks. Western Siberian ethnic groups that include Mansi, Khanty, and Nenets people inhabit the vast regions of taiga stretching from the Ural Mountains in Western Siberia to the Yenisei River in the Siberian East. Northeastern European populations, including Komi, Karelians, and Veps, occupy areas to the west of Ural Mountains. The Eastern Siberian populations, including Altayans, Yakuts, Buryats, Evens, and Evenks, occupy the Siberian taiga region between the Yenisei River in the west and Sea of Okhotsk in the east. Kalmyks migrated in the 17th century to the lower Volga region from a region in northwestern China. (B) Principal component analysis (PCA) of 892 modern-day humans from 27 populations mainly from Siberia and Europe. PCA was performed using a common set of 137,637 SNPs. Sequenced samples are shown using an asterisk. Samples genotyped using DNA microarrays are shown with dots.

The PCA plot captured major differentiation of populations along the first principal component, accounting for 6.4% of the genetic variation. The first principal component corresponds to the west-to-east gradient across Eurasia. The second principal component captured the spread of populations along the north-to-south latitudinal cline, especially among Siberians, capturing 1% of the genetic variation. The PCA plot reveals substantial genetic differentiation between Siberian groups. This suggests that Siberian populations harbor significant genetic diversity not represented by other populations (e.g., European populations, which show only very modest levels of population-specific variation).

Admixture events between Western and Eastern Siberians
Next, we used TreeMix (Pickrell and Pritchard 2012) to construct a tree-based model of population genetic relationships and inter-population admixtures using a set of 30.1 million genome-wide autosomal SNPs inferred from the whole-genome sequencing data. The model placed European and Asian populations along the two early diverging branches (Fig. 2). Furthermore, the Western Siberian Mansi, Khanty, and Nenets groups formed an early diverging subclade with affinity to other Eastern European populations. Most Eastern Siberians, including the Even, Evenk, Buryat, and Yakut groups, formed a separate lineage more related to East Asian populations.


Figure 2.
Genetic relationships of modern-day Siberian, Asian, Native American, European, and African populations. Autosomal TreeMix admixture graph based on whole-genome sequencing data from 44 individuals primarily from Africa, Europe, Siberia, and America. The x-axis represents genetic drift, which is proportional to the effective population size Ne. Admixture events are shown with arrows, with admixture intensities indicated by percentages. Residuals are shown in Supplemental Figure S5.

The TreeMix model also revealed several important admixture events. In particular, we found that 43% (95% CI: 38%–47%) of the Western Siberian ancestry could be attributed to an ancient admixture, with the Eastern Siberian population most related to the modern-day Evenk people. The Northwestern Siberian Nenets people exhibited evidence of an additional admixture, with the Eastern Siberian population closely related to modern-day Evens. We estimated that 38% (95% CI: 31%–46%) of the Nenets’ ancestry could be attributed to this admixture event. To further confirm these predictions, we computed D-statistics (Durand et al. 2011; Patterson et al. 2012) to test for an excess of shared alleles between the Western Siberian and Eastern Siberian groups (Supplemental Figs. S7A–C, S8A–C, S18). In agreement with the TreeMix inference, we observed significant genetic affinity between the Mansi and Evenk, the Khanty and Evenk, and the Nenets and Even people based on the D-statistics tests (Supplemental Figs. S7A–C, 8A–C). These observations support the ancient admixture between ancestors of modern Western and Eastern Siberian populations.

Notably, TreeMix also inferred an admixture signal between common ancestors of the Mansi, Khanty, and Nenets and Native American Andean Highlanders, an indication of their shared genetic history. This event accounts for 6% (95% CI: 4%–8%) of the Western Siberian ancestry. As discussed below, this admixture can be explained by a genetic link of both Western Siberians and Native Americans to ANEs.

Phylogenetic trees provide simplified demographic models, lacking precise information about divergence times of ancestral lineages. The lengths of TreeMix branches are proportional to units of drift parameter (Pickrell and Pritchard 2012) and cannot be directly related to the time of population divergence. Furthermore, branch lengths can be affected by population bottlenecks, which further complicates the inference of population divergence time and limits interpretability of the model. In an attempt to provide more direct estimates of the divergence time between populations, we employed multiple sequentially Markovian coalescent (MSMC) analysis (Fig. 3A; Schiffels and Durbin 2014). We designated the time at which the relevant cross-coalescence rate between two populations becomes 0.5 as a proxy for their separation time (Fig. 3B). The MSMC method is particularly informative for events that are older than 10,000 yr, when each population is represented by only a single individual (Schiffels and Durbin 2014); estimates more recent than 10,000 yr ago are less precise, unless more individuals are included in the analysis.


Figure 3.
MSMC analysis of population separation times. (A) Separation times were calculated for each of the five reference individuals (French, Andreapol Russian, Mansi, Evenk, and Han). Reference populations are shown at the top (inside the orange rounded rectangles). Each reference population column shows MSMC separation time estimates between the reference population and other populations of interest. Populations of interest were sorted according to their separation time from diverging most recently (top) to the oldest (bottom). Separation times are shown on the left side of each graph. Populations with similar times were grouped together. (And) Andreapol; (Ust) Ustyuzhna; (Ob) Objachevo. MSMC separation times under 10,000 yr may be noisy, and therefore, such separation times and corresponding populations are shown in a gray-colored font. (B) MSMC cross-coalescence plots corresponding to divergence between Mansi and five other populations (Nenets, Evenks, French, Andean Highlanders, Dinka). The median separation value (y = 0.5; orange line) was used as a proxy for the population divergence time.

Our MSMC analysis suggested that the Mansi, Khanty, and Nenets separated relatively recently (4.8 kya), after the ancestral populations of Western Siberians experienced an admixture with an ancient group most related to the modern-day Evenk population (Fig. 2). This admixture can be dated to ∼6.8 kya based on the estimated separation time between the Mansi and Evenk populations. These time estimates are only approximate and should be interpreted with caution; however, the relative order of populations (Fig. 3A) is still informative for interpreting their relative genetic affinities. To further confirm these estimates, we performed a more computationally intensive, but more accurate eight-haplotype MSMC analysis using two individuals per population. The results from this analysis provided a similar estimate of 9.9 kya for the time of the Eastern Siberian admixture into Western Siberian populations (Supplemental Fig. S9). Therefore, the Eastern and Western Siberian populations likely came into contact within the last 5000–10,000 yr.