Long genetic and social isolation in Neanderthals before their extinction

Abstract

Neanderthal genomes have been recovered from sites across Eurasia, painting an increasingly complex picture of their populations' structure that mostly indicates that late European Neanderthals belonged to a single metapopulation with no significant evidence of population structure. Here, we report the discovery of a late Neanderthal individual, nicknamed "Thorin," from Grotte Mandrin in Mediterranean France, and his genome. These dentognathic fossils, including a rare example of distomolars, are associated with a rich archeological record of Neanderthal final technological traditions in this region ∼50-42 thousand years ago. Thorin's genome reveals a relatively early divergence of ∼105 ka with other late Neanderthals. Thorin belonged to a population with a small group size that showed no genetic introgression with other known late European Neanderthals, revealing some 50 ka of genetic isolation of his lineage despite them living in neighboring regions. These results have important implications for resolving competing hypotheses about causes of the disappearance of the Neanderthals.

Keywords: Europe; Neanderthal extinction; fossils; genetic and social isolation; genetic divergence; genomics; last Neanderthals; population structure; proteomics.

Graphical abstract

Figure 1

Map, plans, stratigraphy, and archeological 3D projections from Thorin in Grotte Mandrin (A) Map showing geographic location, ages, and coverage group of Neanderthal fossils with genome-wide data used in this study. (B) A 3D model of the disposition of the Thorin fossils during discovery. (C and D) Stratigraphic (C) and plan (D) views through Grotte Mandrin showing Thorin’s discovery location. See also Figure S2 and supplemental information.

Figure 2

The Thorin Neanderthal (Top) View of the mandible in situ when found in September 2019. (Bottom) Virtual reconstruction of the jaw and dental elements of Thorin in tilted (upper left), anterior (upper right), and lateral (bottom right and left) views. See also Figure S2 and supplemental information.

Figure 3

Bayesian modeling of Grotte Mandrin’s PNII levels containing Thorin (Top) The PNII phase of the Grotte Mandrin Bayesian model used to determine the age of Thorin. The direct ages on Thorin comprise 3 pooled mean hydroxyproline AMS ages, an ESR age, and a U-series age constrained as a minimum age (see Tables S1–S3). Outliers are included in the format [outlier:posterior/prior]. The boundaries are cross-referenced to the main Bayesian model established for the site, and represent the start and end dates of the PN2. (Bottom) Comparison of the Thorin modeled age against other late Neanderthals (see Table S12). Calibrated likelihoods in blue represent AMS dates obtained using HYP protocols. ∗∗ indicates high or low autosomal coverage of the dated specimen. Other likelihoods in black represent dates obtained on bulk or ultrafiltered/purified collagen samples. Dates are based on data from references 16 and 31–37. Related to Tables S1–S3 and S12.

Figure 4

Mitochondrial and Y chromosome phylogenies of Thorin and published archaics (A) A Bayesian phylogenetic tree of mitochondrial sequences restricted to the non-coding region from 24 Neanderthals, including Thorin, 1 Denisovan, and 73 modern humans. The tree is inferred from BEAST2, and the posterior probabilities are indicated for the deep divergences between the main Neanderthal clades. (B) Maximum likelihood tree of Y chromosome sequences based on 6,717 polymorphic SNPs from Thorin, 2 Denisovan, 8 Neanderthals including Thorin, and 4 modern humans, inferred using raxml-ng (Tables S11 and S14). Labels are colored based on region: yellow = Europe, green = Siberia, and brown = Caucasus. See also Figures S11 and S12 and Tables 11 and S14.

Figure 5

Allele sharing between Thorin and other known Neanderthals (Left) D-statistics of the form D (Altai Neanderthal, Vindija 33.19; Neanderthal X, Mbuti), showing that all Eurasian Neanderthal samples share more alleles with Vindija 33.19 than with the Altai Neanderthal. Thorin shares relatively more alleles with Vindija 33.19 than early European Neanderthals of the putative first radiation (HST, Scladina, Estatuas pit 1 layer 4), but less than the second radiation (Mezmaiskaya 1, Chagyrskaya 8). We used all reads obtained from each Neanderthal X case to maximize the resolution as contamination levels in general are low. Chagyrskaya 8 is, however, represented with diploid genotypes. (Right) D-statistics of the form D (Vindija 33.19, Thorin; Neanderthal X, Mbuti), showing that Neanderthals from Europe, the Caucasus, and Siberia younger than 80 ka share more alleles with Vindija 33.19 than with Thorin. The exception is the Gibraltar Neanderthal sample (FQ), which shows increased affinity with Thorin. We used all reads for each Neanderthal X and Thorin, while we represented Vindija 33.19 with pseudo-haploid genotypes. Error bars indicate 3× SE (|Z| = 3). Colors indicate the age of the individuals. Samples not directly dated or with uncertain dates are indicated with gray circles. See also Figures S10 and S16–S20 and Table S12.

Figure 6

Demographic history of the Thorin lineage (A) Best-fitting demographic model inferred using momi2 relating Thorin to other Neanderthal and Denisovan genomes. Blue branches show point estimates, whereas gray transparent branches indicate estimates obtained using 100 nonparametric bootstrap replicates. Results are based on a set of 2,454,271 transversion SNPs, with the lower-coverage samples Thorin and Mezmaiskaya 1 represented by pseudo-haploid genotypes. (B) Point estimates (dashed line) and density of 100 parametric bootstrap replicates for divergence time parameters of Thorin, Mezmaiskaya 1, and Chagyrskaya 8 Neanderthal genomes from the late European Neanderthal Vindija 33.19. See also Figures S13, S14, S22, and S26 and Tables 15 and S17.

Figure 7

Runs of homozygosity (ROH) in Neanderthals Bar plot showing cumulative total length of ROHs ≥5 Mb in Thorin and other Neanderthal genomes with average genomic coverage ≥1.5×. ROH length classes are distinguished by bar colors, with total length in each class indicated. See also Figures S23–S25.