The genetic origin of the Indo-Europeans

Abstract

The Yamnaya archaeological complex appeared around 3300 BC across the steppes north of the Black and Caspian Seas, and by 3000 BC it reached its maximal extent, ranging from Hungary in the west to Kazakhstan in the east. To localize Yamnaya origins among the preceding Eneolithic people, we assembled ancient DNA from 435 individuals, demonstrating three genetic clines. A Caucasus-lower Volga (CLV) cline suffused with Caucasus hunter-gatherer¹ ancestry extended between a Caucasus Neolithic southern end and a northern end at Berezhnovka along the lower Volga river. Bidirectional gene flow created intermediate populations, such as the north Caucasus Maikop people, and those at Remontnoye on the steppe. The Volga cline was formed as CLV people mixed with upriver populations of Eastern hunter-gatherer² ancestry, creating hypervariable groups, including one at Khvalynsk. The Dnipro cline was formed when CLV people moved west, mixing with people with Ukraine Neolithic hunter-gatherer ancestry³ along the Dnipro and Don rivers to establish Serednii Stih groups, from whom Yamnaya ancestors formed around 4000 BC and grew rapidly after 3750-3350 BC. The CLV people contributed around four-fifths of the ancestry of the Yamnaya and, entering Anatolia, probably from the east, at least one-tenth of the ancestry of Bronze Age central Anatolians, who spoke Hittite^4,5. We therefore propose that the final unity of the speakers of 'proto-Indo-Anatolian', the language ancestral to both Anatolian and Indo-European people, occurred in CLV people some time between 4400 BC and 4000 BC.

Extended Data Figure 1:. The origin of Central Anatolian Bronze Age people.

(a) Models with eastern steppe sources (including CLV and Serednii Stih). Fitting models include Mesopotamian (Çayönü) and steppe ancestry. (b) Models with western sources, including Usatove and those from Southeastern Europe fail except those with Çayönü and either Mayaky or Boyanovo EBA (both of which are Yamnaya-derived). (c) The steppe (BPgroup)+Çayönü model fails all Chalcolithic/Bronze Anatolians except Central Anatolian Bronze Age. (d) Steppe (BPgroup) ancestry in the BPgroup+Çayönü model is observed in all individuals of the Central Anatolian Bronze Age (mean and ±3 s.e. estimated by qpAdm are shown for all Chalcolithic and Bronze Age individuals from Anatolia that fit the model at p>0.05) as well as in individual ART027_d from Chalcolithic Arslantepe in Eastern Anatolia. (e) BPgroup-related ancestry admixed with different substrata: Aknashen-related in the North Caucasus Maikop, Masis Blur-related in Chalcolithic Armenia, and Mesopotamian-related (Çayönü) in the ancestors of the Central Anatolian Bronze Age, following the route (f) from the North Caucasus to Anatolia; sites with BPgroup-related ancestry marked in bold. In all panels p-values estimated by qpWave are shown.

Extended Data Figure 2:. Admixture date estimates.

We estimate admixture dates for the Core Yamnaya as a mixture of European hunter-gatherer and West Asian populations (a), for the Don Yamnaya as a mixture of Core Yamnaya and UNHG (b), for the Bulgaria, Moldova, Romania, and Serbia (BMRS) Yamnaya as a mixture of Core Yamnaya and European Neolithic/Chalcolithic farmers (c), for the Corded Ware as a mixture of Core Yamnaya and Globula Amphora (d), and for a combined Caucasus-Anatolia population (Maikop-Armenia_C-TUR_C_BA) a mixture of European hunter-gatherer and West Asian populations which occurred ca. 4400 BCE (e). The Core Yamnaya were formed ca. 4000 BCE, followed by admixture ca. 3350 BCE with UNHG and European farmers in the east and west of the Dnipro-Don region and <3000 BCE in central-eastern Europe.

Extended Data Figure 3:. Population structure in people with a Yamnaya cultural affiliation.

Individuals are projected in the same space as in Fig. 1. (a) showing that the Core Yamnaya cluster (red fill symbols) from diverse sites is differentiated from the Don Yamnaya (blue fill) who tend towards the UNHG. (b) Yamnaya individuals in the West (Ukraine, Hungary, Slovakia, and Southeastern Europe) include a tight cluster of individuals as well as others that tend towards the direction of European Neolithic and Chalcolithic groups from Romania and Hungary. Individuals from Russia are shown in grey circles in panel (b). Coordinates of plotted points can be found in Online Table 6.

Extended Data Figure 4:

A 4-way model for the entire Dnipro-Don-Volga-Caucasus region. Mean and ±1 standard error estimated by qpAdm is shown.

Extended Data Figure 5:. The origin of Indo-Anatolian and Indo-European languages.

Genetic reconstruction of the ancestry of Pontic-Caspian steppe and West Asian populations points to the North Caucasus-Lower Volga area as the homeland of Indo-Anatolian languages and to the Serednii Stih archaeological culture of the Dnipro-Don area as the homeland of Indo-European languages. The Caucasus-Lower Volga people had diverse distal roots, estimated using the qpAdm software on the left barplot, as Caucasus hunter-gatherer (purple), Central Asian (red), Eastern hunter-gatherer (pink), and West Asian Neolithic (green). Caucasus-Lower Volga expansions, estimated using qpAdm on the right barplot, disseminated Caucasus Neolithic (blue)-Lower Volga Eneolithic (orange) proximal ancestries, mixing with the inhabitants of the North Pontic region (yellow), Volga region (yellow), and West Asia (green).

Figure 1:. Three Eneolithic clines and their neighbors in space and time.

(a) Map with analyzed sites. (b) PCA analysis using axes formed by a set of ancient West European hunter-gatherer (WHG), Siberian, West Asian, and European farmer populations. Selected individuals relevant to this study are projected (Methods). (c) qpAdm models fitted on individuals of the populations of the clines. The Volga Cline is generated by admixture between Lower Volga (BPgroup) people with upriver Eastern hunter-gatherers (EHG). People of the Dnipro Cline have UNHG or UNHG+EHG admixture relative to the Core Yamnaya (the hunter-gatherer source along this cline is significantly variable). The Caucasus-Lower Volga Cline is generated by admixture of lower Volga people with those from the Neolithic Caucasus (Aknashen-related).

Figure 2.. The three clines in the context of Eneolithic and Bronze Age admixture.

Six 3-source qpAdm models elucidate a complex history of admixture. (a) Caucasus and European hunter-gatherer admixtures in the “Old Steppe”: Krivyansky on the Lower Don received much more CHG-related admixture than upriver people of the Middle Don at Golubaya Krinitsa. In the Middle and Upper Volga and the Kama River, populations had negligible CHG-related influence. (b) The “Don-Volga” difference. On the Lower Volga and North Caucasus piedmont, the BPgroup received CHG-related ancestry like its western Lower Don counterpart at Krivyansky; but, it also received ancestry from Central Asia; this eastern influence was higher still in the Bronze Age Steppe Maikop. (c) The Volga basin Eneolithic populations vis-à-vis the Don: populations at Khvalynsk, Klopkov Bugor, and Ekaterinovka form a Volga Cline between the Berezhnvoka cluster on the Lower Volga and the upriver EHG-like populations of the Middle Volga (Labazy and Lebyazhinka). (d) the Volga basin Eneolithic populations vis-à-vis Central Asia: a slight excess of Central Asian ancestry in the Khi subset of Khvalynsk. (e) the “Dnipro” cline: the Core Yamnaya are on one end of a cline, that also includes the Don Yamnaya and Serednii Stih populations, formed by admixture from the “Caucasus-Lower Volga” (CLV) cline of differential admixture of Neolithic Caucasus and BPgroup people. The CLV Cline includes diverse people buried in kurgans at Berezhnovka, Progress-2, Remontnoye, and Maikop sites Klady and Dlinnaya-Polyana ~5000–3000 BCE. (f) “West Asian”: CLV ancestry first appears in the Chalcolithic population at Areni-1 in Armenia and is also present in the Bronze Age at Maikop. The majority of the ancestry is from West Asian sources from the Mesopotamia-Caucasus (or Çayönü-Masis Blur-Aknashen) cline. Chalcolithic and Bronze Age Anatolians lack CLV ancestry but traces of it can be found in Bronze Age Central Anatolians.

Figure 3.. Patrilineal succession.

Temporal distribution of key Y-chromosome haplogroups from Kazakhstan, Kyrgyzstan, Mongolia, Russia, Turkmenistan, Ukraine, Uzbekistan, and comparative regions of Europe and West Asia 6000–1000 BCE. The Early and Middle Bronze Age group includes the Yamnaya, Afanasievo, Poltavka, Catacomb, Chemurchek, and North Caucasus cultures; the Middle and Late Bronze Age group individuals of diverse cultures down to 1000 BCE including those of the Sintashta, Andronovo, Potapovka, and Srubnaya cultures. Information on which individuals are plotted can be found in Online Table 6.

Figure 4:. Trajectory of the Yamnaya expansion.

We use HapNe-LD to estimate the changes in effective population size over time of Yamnaya ancestors, performing the computation separately for the individuals from the earlier three hundred years of our sampling, and the later three hundred years; shading shows confidence intervals (dark: 50%, light: 95%). Jointly displaying these two trajectories reveals an extraordinary population expansion after 3642–3374 BCE (intersection of 95% confidence intervals for the two analyses for the minimum), from when the effective size is a few thousand to an order of magnitude larger. The offset on the x-axis is due to the difference in sampling time between the two groups.

Figure 5:. IBD analysis of the Yamnaya and their predecessors.

Pairs of individuals linked by at least one IBD segment ≥20cM in length reveal a sparse but highly connected network in the Pre-Yamnaya (Methods) (a) and Yamnaya (b) groups. No detectible IBD is found in the Pre-Yamnaya period beyond the scale of 1000km (c); Yamnaya share more IBD with each other at short distance scales but IBD sharing extends all the way to the ~6000km scale of their geographical distribution (d). However, closely related individuals only occur at short distance scales in both Pre-Yamnaya (e) and Yamnaya (f) groups, indicating that the IBD sharing in the Yamnaya was a legacy of their common origin. In panels (c-f) we display the two-sided 95% confidence interval as a vertical interval (at distance=0) or rectangle (at distance ranges>0) and in red the fraction of dividing the number of pairs of individuals sharing IBD (I)/total number of pairs of individuals (T). (g) In a set of 9 Yamnaya cemeteries, and a total of 25 kurgans, closely or distantly related individuals are virtually absent in inter-cemetery comparisons, more are found in inter-kurgan/within-cemetery comparisons, and more still in intra-kurgan comparisons; nonetheless, most Yamnaya individuals in all comparisons were unrelated. Kurgan burial of close kin was less common than in the case of a local patrilineal dynasty as at a Neolithic long cairn at Neolithic Hazleton North, but more common than in Neolithic monuments of Ireland. Two-sided 95% confidence intervals are shown.