Palaeogenomics of Upper Palaeolithic to Neolithic European hunter-gatherers

Abstract

Modern humans have populated Europe for more than 45,000 years^1,2. Our knowledge of the genetic relatedness and structure of ancient hunter-gatherers is however limited, owing to the scarceness and poor molecular preservation of human remains from that period³. Here we analyse 356 ancient hunter-gatherer genomes, including new genomic data for 116 individuals from 14 countries in western and central Eurasia, spanning between 35,000 and 5,000 years ago. We identify a genetic ancestry profile in individuals associated with Upper Palaeolithic Gravettian assemblages from western Europe that is distinct from contemporaneous groups related to this archaeological culture in central and southern Europe⁴, but resembles that of preceding individuals associated with the Aurignacian culture. This ancestry profile survived during the Last Glacial Maximum (25,000 to 19,000 years ago) in human populations from southwestern Europe associated with the Solutrean culture, and with the following Magdalenian culture that re-expanded northeastward after the Last Glacial Maximum. Conversely, we reveal a genetic turnover in southern Europe suggesting a local replacement of human groups around the time of the Last Glacial Maximum, accompanied by a north-to-south dispersal of populations associated with the Epigravettian culture. From at least 14,000 years ago, an ancestry related to this culture spread from the south across the rest of Europe, largely replacing the Magdalenian-associated gene pool. After a period of limited admixture that spanned the beginning of the Mesolithic, we find genetic interactions between western and eastern European hunter-gatherers, who were also characterized by marked differences in phenotypically relevant variants.

Fig. 1. Locations, dates and MDS plot of ancient Eurasian hunter-gatherers.

a, Geographic locations of newly reported individuals (filled symbols with black outline) and representative previously published individuals (outlined stars). Dotted lines delimit geographic regions described in the text. b, Calibrated radiocarbon dates of individuals plotted in a. The y axis shows the average of calibrated radiocarbon dates in thousands of years (kyr) (Supplementary Data 1.A). The horizontal dashed line marks the boundary between Late Pleistocene and Holocene. c, MDS plot of European hunter-gatherers based on 1 − f₃(Mbuti; pop1, pop2). The dimensions are calculated using newly reported and previously published hunter-gatherer groups or individuals with more than 30,000 SNPs. The detailed grouping of individuals shown with empty coloured circles is described in Supplementary Data 1.I.

Fig. 2. Genetic differences among Gravettian-associated populations.

a, MDS plot of pre-LGM individuals. The pre-40 ka group and the Fournol and Věstonice clusters are marked as shaded areas in different colours. Individuals and groups are plotted with the same colours and symbols as in Fig. 1 and names are indicated next to the symbols. b, Gravettian-associated individuals form two distinct groups, with central-eastern and southern European individuals as part of the Věstonice cluster and western and southwestern European individuals as part of the Fournol cluster. In central-western Europe, Gravettian-associated individuals from Goyet show affinity to both clusters. Error bars show 1× s.e.m. (black) or 3× s.e.m. (grey) of the f₄ values estimated from 5 cM-block jackknife analysis. c, Admixture graph modelling of the main pre-LGM European hunter-gatherer lineages created using qpGraph.

Fig. 3. The population substructure and diversity of Epigravettian-associated groups in southern Europe.

a, Population structure among Epigravettian-associated populations revealed by a neighbour-joining tree based on pairwise f₂ genetic distances. Branch labels show unique drift lengths; black dots refer to individuals with newly generated data and white dots refer to previously published genomes; the position of each node does not imply the location where the split took place. b, Population diversity shown by PMR between individuals in different groups. The grouping of Epigravettian-associated populations is shown with the same colour in a. The black-outlined diamond marked out in the Gravettian group shows the PMR between the two Gravettian-associated individuals from southern Italy (Paglicci 12 and Ostuni 1). In the box plot the centre line is the median, box bounds delineate the interquartile range and whiskers extend to maximum and minimum values, excluding outliers. The sample size of individual pairs included in each group is reported in Supplementary Data 3.A.

Fig. 4. Ancestry modelling of post-19 ka individuals in southwestern, western and central Europe.

a,b, The ancestries of individuals in the GoyetQ-2 cluster and Iberian hunter-gatherers (HGs) (a) and individuals in the Oberkassel cluster (b) were modelled using qpAdm, with Fournol 85 and Arene Candide 16 representing the Fournol and Villabruna ancestries, respectively. The length of the colour bar shows the proportion of each ancestry. The error bar shows the s.e.m. of estimates from 5-cM-block jackknife analysis. Details of the modelling are provided in Supplementary Data 3.C.

Fig. 5. Ancestry modelling of hunter-gatherers from 14–5.2 ka and their allele frequencies on phenotypic SNPs.

a, The genetic ancestry of hunter-gatherers dated between 14 ka and 5.2 ka modelled using qpAdm, with Oberkassel, Yuzhniy Oleniy Ostrov, Goyet Q-2 and Neolithic farmers from present-day Turkey (Barcın, Menteşe and Boncuklu sites) representing Oberkassel (WHG) (blue), Sidelkino (EHG) (red), GoyetQ2 (orange) and Anatolian Neolithic farmer (green) ancestries, respectively. The average calibrated date is shown, with pie charts indicating the estimated proportion of ancestry for each group or individual. Details of the modelling are provided in Supplementary Data 3.E,F. The expansion of farming by 9, 8, 7.5 and 7 ka is shown as green shades. Adapted from 10.5281/zenodo.5903165 (CC BY 4.0). b, Allele frequencies of different hunter-gatherer groups (coloured dots) on four SNPs related to skin colour (SLC24A5 and SLC45A2), eye colour (HERC2/OCA2) and lactase persistence (LCT). Dots are maximum likelihood estimates and error bars show 95% confidence intervals of the derived allele frequencies (n, the number of individuals in each group, is provided in Supplementary Data 3.G). Dashed lines show the frequencies estimated for the indicated present-day 1000 Genomes Project populations (CEU, Utah residents of northern and western European ancestry; GBR, British; IBS, Spanish; TSI, Tuscan ). Details on the allele frequency estimates are provided in Supplementary Information, section 12, Supplementary Figs. 27 and 28 and Supplementary Data 3.G.

Extended Data Fig. 1. MtDNA and Y-chromosome phylogenies.

Bold letters refer to (a) mtDNA and (b) Ychr haplogroups, whose boxes are coloured according to the legend in Extended Data Fig. 2. The labels in italic denote previously published individuals without new data generated in this study.

Extended Data Fig. 2. The distribution of mtDNA and Y-chromosome haplogroups among different hunter-gatherer groups.

The length of each coloured bar represents the fraction of individuals carrying the corresponding haplogroup (legend on the right of each panel). The number of individuals in each group is written to the right of each bar. We only plotted groups with more than two individuals and, for this reason, individuals from the GoyetQ116-1 cluster are included here into the Fournol cluster.

Extended Data Fig. 3. F₄-statistics comparing the affinity of pre-LGM European hunter-gatherers to Goyet Q116-1 and Sunghir.

The colours correspond to the grouping of tested populations, dots refer to the f₄-values and the dark and light error bars to 1*SE and 3*SE estimated from 5 cM block jackknife, respectively. This figure shows that the Gravettian-associated individuals from western Europe (Fournol cluster) are closely related to Goyet Q116-1 while the Gravettian-associated individuals from central-eastern and southern Europe (Věstonice cluster) are closely related to the Sunghir group, representative of the Kostenki cluster. Details are provided in Data S2.B.

Extended Data Fig. 4. Admixture graph modelling of pre-34 ka hunter-gatherer lineages.

In this admixture graph, the lineage related to Zlatý kůň splits more basally than the Bacho Kiro IUP group, who contributes to Tianyuan, Ust’Ishim and Goyet Q116-1 (indicated with red lines), but not to the Sunghir group.

Extended Data Fig. 5. Summary of ROH segments detected in Eurasian hunter-gatherers.

We visualize the total amount of ROH longer than 4 cM for (a) pre-LGM individuals, (b) Epigravettian- and Magdalenian-associated individuals, (c) individuals carrying high proportions of Sidelkino-related ancestry, and (d) individuals carrying high proportions of Oberkassel-related ancestry. Colour legend is shown in (e). Each bar represents one individual with the ROH grouped in four length categories (grouped by colour). The inferred pattern of short ROH (4–8 cM, visualized in blue) being common is in stark contrast to most later farmer populations, where the majority of individuals have no short ROH whatsoever (see), and evidences small effective population sizes across West Eurasian hunter-gatherer groups. A dashed line of 50 cM total ROH is drawn in each panel to help comparison between panels with different y-axis scales. Details of the grouping and ROH segments are provided in Data S3.B.

Extended Data Fig. 6. West Eurasian PCA showing the genetic positioning of post-LGM hunter-gatherers.

Present-day individuals (gray dots) genotyped on the Human Origins dataset are used to define the PCA variation onto which ancient genomes (coloured symbols) are projected. The newly reported individuals with over 15,000 SNPs on the Human Origins dataset are shown in black-outlined and filled symbols, as illustrated in the legend on the right, while representative ancient genomes are shown in outlined symbols, as illustrated in the legend at the bottom of the PCA.

Extended Data Fig. 7. Admixture dates between Oberkassel/Villabruna and ANE ancestries in the oldest individuals from the Sidelkino cluster.

The triangles show the average calibrated dates of the tested groups and the dots show the estimated admixture dates with the software DATES using Oberkassel (red dots) or Villabruna (blue dots) clusters as one source and ANE-related individuals as the other source population. The generation time is set to 29 years and the error bars show the SE of the admixture date estimated from jackknife resampling (n = 22 autosomal chromosomes). Additional details are provided in Supplementary Information, section 11 and Supplementary Table 3.

Extended Data Fig. 8. Changes of Sidelkino and Oberkassel ancestry proportions in post-14 ka hunter-gatherers.

The bivariate plots show the expansion of Oberkassel and Sidelkino ancestries through time in two European areas (longitude below and above 30 degrees). The x-axis shows the average age of each tested individual/group and the y-axis shows the proportion of Sidelkino ancestry, relative to the total hunter-gatherer ancestry (Oberkassel + Sidelkino) in each group. The three squares highlight Baltic HG groups associated with the Comb Ceramic Culture (CCC) that show a marked increase in Sidelkino-related ancestry compared to older Baltic HG groups.

Extended Data Fig. 9. Graphical summary depicting the main genetic transformations in post-40 ka hunter-gatherers from Europe.

This figure shows (a) the distribution of and interaction between hunter-gatherer genetic ancestries and (b) a simplified schematic representation of major chrono-cultural subdivisions of the European Upper Palaeolithic (green blocks) followed by a grouped Mesolithic to Neolithic block (in gray). The x-axes report the geographic regions as divided in Fig. 1a, and the y-axes report time in thousand years before present (kBP). In panel a, genetic affinity between different ancestries is indicated by thick lines or shades, while admixture is indicated with arrows. In panel b, the colour code does not imply archaeological similarities.