Neanderthals and humans interbred in Europe for 5,400 years

Past human expansions shaped the spatial pattern of Neanderthal ancestry

Abstract
The worldwide expansion of modern humans (Homo sapiens) started before the extinction of Neanderthals (Homo neanderthalensis). Both species coexisted and interbred, leading to slightly higher introgression in East Asians than in Europeans. This distinct ancestry level has been argued to result from selection, but range expansions of modern humans could provide an alternative explanation. This hypothesis would lead to spatial introgression gradients, increasing with distance from the expansion source. We investigate the presence of Neanderthal introgression gradients after past human expansions by analyzing Eurasian paleogenomes. We show that the out-of-Africa expansion resulted in spatial gradients of Neanderthal ancestry that persisted through time. While keeping the same gradient orientation, the expansion of early Neolithic farmers contributed decisively to reducing the Neanderthal introgression in European populations compared to Asian populations. This is because Neolithic farmers carried less Neanderthal DNA than preceding Paleolithic hunter-gatherers. This study shows that inferences about past human population dynamics can be made from the spatiotemporal variation in archaic introgression.

INTRODUCTION
Sequencing of Neanderthal genomes has revealed that ~2% of the DNA of modern humans (MHs) outside of Africa is more similar to DNA from Neanderthals (NEs) than it is to DNA from sub-Saharan African populations (1, 2). Two main and nonexclusive hypotheses have been proposed to explain this pattern: (i) hybridization between NEs and MHs during their expansion out of Africa (OOA), leading to the introgression of Neanderthal DNA segments into MHs (1, 2), and (ii) incomplete lineage sorting resulting from ancestral population structure in Africa, with ancestors of non-Africans more closely related to NEs than to sub-Saharan Africans (3). Evidence in favor of hybridization has accumulated during the past decade (4–6). However, the number, timing, and locations of interbreeding events between MH and NE remain unclear. While early studies have suggested a single hybridization pulse in the Middle East (1, 2), a growing body of research supports the hypothesis of multiple hybridization events (7–11). In particular, it has been shown that multiple hybridization events over time and space in western Eurasia are consistent with the levels of Neanderthal ancestry observed in modern populations (7).
While NE ancestry is relatively uniform among modern Eurasian populations (1, 2), it is approximately 8 to 24% higher in East Asia than in Europe (5, 10, 12, 13). This observation is unexpected since the currently known geographic distribution of NEs was almost exclusively in the western part of Eurasia (14). Three major hypotheses have been proposed to explain the difference in NE ancestry between western and eastern Eurasian populations: (i) higher effective population size in Europeans compared to Asians, which led to a stronger effect of purifying selection acting on deleterious NE alleles in the former (15); (ii) dilution of NE ancestry in Europeans due to an input from a hypothetical “basal” (or “ghost” population) with little or no NE ancestry (16, 17); and (iii) multiple pulses of NE introgression, where the original Eurasian introgression pulse was supplemented by additional pulses after the divergence between the European and Asian populations, resulting in different NE ancestry levels (9, 10, 17, 18).
Recently, an additional hypothesis has been proposed, where different levels of NE ancestry between western Europe and eastern Asia are the result of the range expansion of MHs after the OOA event (19). Population range expansions have important evolutionary consequences, including (i) creating gradients of allele frequencies (20, 21); (ii) increasing the frequency of specific alleles, whether neutral (22, 23) or under natural selection (24, 25); (iii) decreasing genetic diversity along the axis of expansion (26, 27); and (iv) increasing mutational load in populations through the maintenance of deleterious alleles (28). In addition, when admixture with a local population occurs, population expansions tend to disproportionately increase the genetic contribution of the local population to the invasive genetic pool, even if interbreeding is limited (29). This latter effect is expected to result in the formation of spatial gradients of introgression along the axis of the biological invasion (see Fig. 1A). Under this assumption, introgression of local genes (i.e., NEs) increases in the invasive population (i.e., MHs) with the distance from the source of the expansion (i.e., Africa). This is due to the combined effects of (i) continuous hybridization events at the wave front of the range expansion, resulting in more interbreeding possibilities when moving away from the source; (ii) genetic surfing resulting from serial founder effects and population growth; and (iii) demographic imbalance between the growing expanding population and the local population at demographic equilibrium. This hypothesis proposes to explain the different levels of Neanderthal ancestry in Europe and East Asia by different geographical distances from the source of the MH expansion in Africa (19). This assumption of multiple hybridization events occurring continuously across time and space may not be distinguishable from a single hybridization pulse, as defined by Di Santo et al. (30). Computational simulations showed that the process of range expansion could explain the difference in NE ancestry between Europe and East Asia based on contemporary genomic information from the two extreme sides of Eurasia (west and east) (19). However, neither a detailed inspection of geographic introgression patterns (i.e., the existence of gradients) nor their change over time was included in this study. In addition, range expansions occurred not only during the OOA expansion (7) but also during other prehistoric periods (31). This includes the European Neolithic transition, when farmers coming from southeast Europe and Anatolia partially replaced hunter-gatherers (32–35), as well as the Bronze Age, with the spread of pastoralist populations from Eurasian steppes (36–38). Therefore, multiple population movements during recent human history could have contributed to shaping NE ancestry across time and space because distinct expanding populations can carry various levels of NE ancestry (20, 31, 39).

RESULTS AND DISCUSSION
Spatial gradients of Neanderthal ancestry in Eurasia
We analyzed an extended dataset of 4464 published ancient and modern genomes [from ~40,000 years before the present (BP) to present time] retrieved from the Allen Ancient DNA database (40). We associated each genome with one of the following population groups: Paleolithic/Mesolithic hunter-gatherers (HGs), Neolithic/Chalcolithic farmers (FAs), other ancient samples (OTs), or modern samples (MDs). We estimated NE introgression using F4 ratios for all genomes and averaged the values for genomes from the same geographic coordinates, time periods, and population group, leading to n = 2625 samples constituted of one or more genomes (Fig. 1B; see Material and Methods). We then explored the fixed effect of latitude, longitude, time (dates in years BP), continental area (Europe or Asia), and their interactions by using a linear mixed model (LMM) with log-transformed NE ancestry as the response variable. LMMs are particularly useful for dealing with hierarchical structures and the nonindependence of the dataset. We investigated the random effect of the population group, the period nested within these groups (in clusters of 500 years), and the spatial and temporal autocorrelation of data. This model was called “Full Eurasia” because it uses the whole dataset. On the basis of the lowest Akaike information criterion (AIC) value (41), the best LMM was retained (table S1).
By considering the average date of all samples as a time reference (~4200 years BP), we observed a linear relationship of NE ancestry with latitude and longitude, in both Europe and Asia (Fig. 2 and table S1). These geographic patterns support the hypothesis of spatial introgression gradients generated after population expansion with hybridization [(19), simulations in text S1 and data S1], schematically represented in Fig. 1A, in which the introgression of local genes (i.e., NEs) is expected to increase in the invasive population (i.e., MHs) with the distance from the source of its expansion (i.e., Africa). While a positive relationship with latitude is observed in Europe and Asia (Fig. 2A), a contrasting relationship is observed with longitude, positive in Asia and negative in Europe (Fig. 2B). The increasing NE introgression with latitude is compatible with the OOA expansion of MHs from southern to northern areas of Eurasia while hybridizing with NEs. The longitudinal pattern is compatible with a source of expansion in the Middle East, from which NE ancestry is expected to increase with longitude in Asia but decrease with longitude in Europe. Although alternative evolutionary forces may also be responsible for creating introgression gradients (e.g., spatially varying selective pressure), the specific pattern we observed with a source of all spatial gradients in the Middle East makes population range expansion the most parsimonious explanation.