Post by Admin on Jan 31, 2021 1:49:40 GMT
Tracing different HG ancestries in farmer groups Analogous to the test with HG individuals, we calculated f-statistics of the form f4(Mbuti, test; Villabruna, EHG) for all Neolithic groups (fig. S5A) to estimate the affinity of Neolithic groups (test) to the distal HG representatives of the east-west cline. Like the HG groups (fig. S2A), the f4-values from the Neolithic groups are mostly negative, indicating shared ancestry with Villabruna, but form a wide gradient reaching 0, which implies equal amounts of shared ancestry with EHG for some groups located in southeastern Europe. We used qpAdm to quantify these sources in MODEL B (text S9) for all Neolithic European groups. However, using these distal sources due to the very small amount of HG component in many early Neolithic groups, the model fails to reliably detect an EHG component (table S12). We therefore restricted the analysis to more proximal sources. We chose the deeply covered Mesolithic individual Loschbour from Luxembourg as representative of WHG ancestry (west of the Rhine) and therefore as a proxy for the Mediterranean route (34). We selected the Hungarian individual KO1, which looks genetically like a HG, though found in a farming context, for the HG ancestry of the continental route (10). We first replicated the f-statistics of the form f4(Mbuti, test; Loschbour, KO1) (fig. S5B). Here, positive f4-values indicate shared ancestry with KO1 as a proxy for the EHG-WHG cline. In turn, negative f4-values indicate excess shared ancestry with Loschbour and thus the WHG cluster. The results show a tendency for farmer groups east of the Rhine to share more ancestry with KO1 and farmer groups west of the Rhine with Loschbour. Notable exceptions are individuals from the Middle Neolithic Blätterhöhle group (6, 13), individual N22 from Poland (35), and our individual TGM009, all of which show a strong affinity to WHG-related individuals, although they are located east of the Rhine. On the basis of our geographic rationale, we then used qpAdm to model HG components in the Neolithic groups, including both HG sources that could be encountered en route for the proposed expansions (MODEL C; text S9, fig. S6, and table S13). We chose the best-fitting model given by qpAdm in cases where both HG sources were supported. Figure 4 summarizes these results, illustrating the diverse HG ancestries in time windows across major regions in Europe, adapting the models used to the targeted groups (table S14)."
Fig. 4 Maps of variable sources of HG and Anatolian Neolithic ancestries through time.
(A) Proportion of distal sources of Villabruna, EHG, Goyet_Q2, and Anatolia_Neolithic of post-LGM HG individuals (14000–4000 cal BCE) estimated according to MODEL B (table S12). The following five panels (B to F) show Neolithic farmer groups (6000–3500 cal BCE) modeled with the proximal sources Anatolia_Neolithic, KO1, and Loschbour according to MODEL C, in time increments of 500 years each. Transparent colors indicate individuals or groups not sufficiently supported by the models (P < 0.05). Note that not only N22 from the Polish BKG, TGM009, and Blätterhöhle, which carry a substantial proportion of HG ancestry, but also Anatolian_Neolithic ancestry have been modeled with MODEL B and were added to (E) and (F), respectively. See table S14 for model details.
We caution that because the amount of HG ancestry is very low in many Neolithic groups (<10%), it remains difficult to characterize the ultimate source reliably. Nevertheless, our admixture patterns from supported models show clear geographic signals. Neolithic groups associated with the LBK in central Europe (Hungary, Austria, and Germany) carry a small HG proportion, which was likely derived from admixture with HG individuals of the EHG-WHG cline and could have occurred in southeastern Europe during a preceding phase of the Neolithic expansion around 6000–5400 BCE. When using f-statistics of the form f4(Mbuti, test; BDB001, KO1) with our new Mesolithic genome (BDB001) from the Middle-Elbe-Saale region in central Germany as a geographically local HG proxy (instead of using Loschbour, which is located west of the Rhine), we do not find support for a local attraction for LBK groups, but the same pattern as for Loschbour (fig. S5C). This suggests that additional gene flow from neighboring Loschbour-like HG such as BDB001 in central Europe was negligible in the first Neolithic groups. However, the German Baalberge group (4000–3500 BCE) shows a marked increase of such HG ancestry, as well as individuals from the Blätterhöhle group, as has been suggested (5, 6), compared to a combination of both KO1-like and Loschbour-like ancestries for LBK groups (6). We can now show that this increase in WHG ancestry (up to 21.3 ± 1.5% in Baalberge; table S13) is driven by either local Loschbour-like ancestry or an expansion of farming groups from the west carrying this signal during the fifth millennium BCE, as suggested by archaeological data (36). For all studied Neolithic groups west of the Rhine, we observe a different pattern with a higher HG ancestry proportion, even for earliest groups that appears to be of a local (Loschbour-like) HG origin, consistent with archaeological data (2).
Late survival of HG ancestry in central Europe
In contrast to Middle-German Neolithic groups, the Late Neolithic individual TGM009 (~3300 BCE) found in an TRB/Elb-Havel group context shows a different pattern of HG ancestry. When modeled with distal sources, this individual carries 63.6 ± 5.2% Villabruna-related ancestry (see MODEL B; table S12 and Fig. 4). When we use proximal sources (MODEL C; table S13), we estimate the HG component to be split into 48.1 ± 6.4% KO1-related ancestry and 25.8 ± 6.1% Loschbour-like ancestry (P = 0.12). To test whether the subtle EHG-related signal observed in TGM009 (through KO1) could also come from Scandinavian HG, as suggested by the archaeological records of regional late contacts with the Scandinavian Mesolithic (37), we applied f-statistics of the form f4(Mbuti, test; Hungary_KO1, Sweden_Motala_HG). The f4 value is significantly negative (f4 = −0.0035, Z score = −6.383), indicating that TGM009 shares more ancestry with southeastern Europe HG than with Scandinavian HG (table S16). Consequently, qpAdm models with Sweden_Motala_HG in place of Hungary_KO1 received a poor fit (P = 1.32−04; table S16).
In the specific case of TGM009, we also tested whether individuals of the Pitted Ware Culture (PWC) group (~3200–2300 BCE) (38), considered to be “Neolithic HG,” would be a suitable contemporaneous proximal source with MODEL C set of outgroups (table S16). We find that the tests still support the three-way model with Anatolia_Neolithic, Luxembourg_Loschbour, and Hungary_KO1 as a third source (P = 0.12). However, when we add Sweden_Motala_HG to the outgroups, we find the best support for a four-way mixture model (P = 0.087) of 21.7 ± 2.4% Anatolia_Neolithic, 24.4 ± 6.2% Luxembourg_Loschbour, 12.6 ± 4.4% Sweden_PWC, and 41.3 ± 7.3% Hungary_KO1. The small but stable contribution of PWC groups at the northern fringes of the loess belt adds additional complexity to the modes of interaction between Late Neolithic and the last HG groups in central Europe.
Estimating admixture dates between HG and early farmers
To gain further insights into the timing of the admixture of the HG ancestry proportion, we used the software package DATES to estimate the timing of admixture events between early farmers and HG, given in generations in the past (Fig. 5, table S17, and text S10). The resulting date estimates, when ordered temporally and geographically, echo patterns of both groups that are also visible in the PCA (east versus west of the Rhine). It confirms that this pattern is dependent not only on the amount of HG ancestry but also on the qualitative signature of the HG ancestry, as supported by fig. S7. Together, the date estimates for the four southern French groups from sites PEN A and B and LBR B suggest admixture with local HG relatively soon (100 to 300 years) after the arrival of Neolithic farmers about 5850 cal BCE, with an admixture date about 5740–5450 cal BCE for all groups (table S17). These date estimates agree with local archaeological data for the establishment of early farming in this region (18), although we cannot exclude a scenario in which admixture occurred on the Italian Peninsula shortly before.
Fig. 5 Admixture dating per group.
Admixture date estimation according to DATES software, obtained in generations. Radiocarbon date intervals (given with 2-sigma) for each site are black lines; blue diamonds are the estimated admixture date (SE = 1). Admixture is estimated with two sources: European_HG and Anatolia_Neolithic. Number of years calculated on the basis of 28 years for one generation. Admixture time calculated according to the oldest date of the radiocarbon interval for each group (table S17 and text S10).
In the rest of France, post-LBK groups show an older admixture date, which, together with a strong local HG component, place the admixture event with WHG-related HG during the first phase of the local Neolithic (Fig. 4). In contrast, the site OBN reveals a heterogeneous genetic signal forming three distinguishable groups with different HG proportions and corresponding admixture dates (tables S7, S13, and S17). The differences between the date estimates for the three groups suggest that the OBN group shows a genetic substructure rather than a recent ongoing admixture event.
Tracing GoyetQ2 ancestry in France
We also used qpAdm with GoyetQ2, Villabruna, and Anatolia_Neolithic as sources to track Magdalenian-associated GoyetQ2-like ancestry in the new western Neolithic groups (MODEL D; text S9 and table S15) (15). The GoyetQ2-related component is observed in the sites of PRI (6 ± 3%) in western France and GRG (3.7 ± 1.3%) from the Paris basin. Similar to Iberian Neolithic groups, the PRI group (4300–4200 BCE) can be modeled successfully with GoyetQ2 as one of the HG components (fig. S8 and table S15), suggesting either admixture with local HG in western France who retained this post-Last Glacial Maximum component or genetic contacts with Neolithic Iberia in later phases. Of note, individuals from the site PRI on the Atlantic coast indicate a slightly more recent admixture date with European_HG (about 5200 BCE), which is consistent with a later arrival of Neolithic groups in westernmost France (table S17). We plotted these results in fig. S9 to complement the summary map with the GoyetQ2-related component in Neolithic groups, which adds another layer of complexity to the overall genetic picture of Neolithic Europe.
Connections with Britain and Ireland
To address questions at broader scale on the western fringe of Europe, we also investigated the relationship between Britain and Ireland and the European mainland during the Neolithic. Here, we performed f4-statistics of the form f4(Mbuti, Britain/Ireland groups; PRI, test), which measures whether Neolithic groups from Britain and Ireland share more genetic drift with the French Atlantic group PRI than with a test population (see fig. S10). As previously described (39, 40), we confirm that British Neolithic groups share affinities with the Mediterranean Neolithic (LBR_A, France_MN, and Iberia_MN), which is also visible in the WHG-rich HG proportion (Fig. 4). However, on the basis of the results from our French Neolithic sites, we suggest that English, Welsh, and Scottish groups are connected to the Mediterranean Neolithic sphere not only via the Atlantic coast but more plausibly also via Normandy (FLR), the Paris Basin (GRG), and southern France (LBR A, post-ICC groups).
DISCUSSION
The expansion of Anatolian Neolithic farmer–related ancestry across Europe has been described for many geographic regions (5–7, 9–11, 39, 40). A recurrent pattern of increasing HG gene flow in individuals associated with a farming lifestyle many centuries after initial contact and tentatively termed “HG resurgence” (5) has hitherto been observed in the Iberian Peninsula, northern/central Europe, the Carpathian Basin, and the Balkans (6). However, a comparison of genomic data in the light of the proposed main routes of Neolithic expansion has not yet been attempted. Our new results provide important insights for the region of modern-day France where both routes had intersected. This region has not been documented so far, but is ideally positioned to address these questions. The different Mesolithic genetic substratum in Europe (6, 7, 15, 33) enables us to track both Neolithic expansion routes in the form of the quantity and quality of the admixed HG component observed in Neolithic groups.
Neolithic groups from southern France, which are a part of the Mediterranean/ICC route of expansion, show a different genetic profile when compared to early periods of expansion in other regions, with a substantially higher HG component than groups associated with the continental route (up to 56 ± 2.9% of HG component for PEN B; table S10). Although the sites PEN and LBR date ~400 years (i.e., ~16 generations) later than the first local coastal settlements in Liguria, Provence, and Mediterranean Languedoc, they suggest a recent local admixture event (between three and six generations earlier; Fig. 5 and table S17). Archaeological research has argued for increased interaction between incoming farmers and indigenous HG in the western Mediterranean during a second stage of the Neolithization process and especially in areas with higher HG population densities, e.g., the Tosco-Emilian Apennine and Po plain (18). We are now able to confirm that these contacts left a traceable biological signal during the Neolithic expansion in southern France. From an archaeological perspective, this suggests that HG have contributed to the clear changes observed within the material culture postdating the pioneer phase. Note that ICC individuals from the eastern Adriatic coast have only a very small amount of HG ancestry with a greater affinity to central European groups (see table S8). This fits with the hypothesis of a differentiation of technical traditions within material cultures observed from both sides of Apennine Mountains in Italy: an Adriatic tradition connected to the Balkans and a Tyrrhenian one whose origin is still unknown (41). It is tempting to associate such a strong HG component on the Tyrrhenian side with the characteristic/specific pottery traditions observed in this same region and to consider these original traditions the result of a HG reinterpretation (41). However, the scarcity of genomic data available from central and southern Italy currently does not allow this hypothesis to be tested directly. Moreover, ICC individuals from the Iberian Peninsula also carry less HG ancestry. Together, this rejects the hypothesis that ICC-associated individuals represent a uniform genetic horizon per se and argues for more regionally nuanced scenarios of interaction.
Early central European farmers carried a very small percentage of HG component (about 5% on average), which is most likely nonlocal and instead derives from admixture in Transdanubia during an early stage of expansion. This not only matches general observations in the archaeological record (2) but could also explain why the earliest LBK lithic assemblages resemble those of the Late Mesolithic Blade and Trapeze Complex. It also confirms previous aDNA studies (5, 6), which argued for a fast spread of the first farmers across the German loess regions. LBK groups from southwestern and eastern Germany share more affinities with KO1 than with Loschbour (fig. S5). The admixture date estimates for southern German sites, SMH and SCH, 19.2 ± 3.8 and 12.3 ± 8.2 generations, respectively, are younger or contemporaneous to admixture date estimates from the Carpathian Basin and Austria (Fig. 5 and table S17). The temporal lag, and the subtle increase of shared KO1-like HG ancestry, allows us to trace LBK groups chronologically and matches well with the proposed model of LBK expansion from the core region in Transdanubia as suggested by archaeological research (14). However, the present picture does not explain the increasing amount of archaeological evidence of contact with HG and groups with southern influences in more western sites, particularly during the early LBK, such as the debated La Hoguette and Limburg phenomena (16).
In contrast to the situation in central Europe, regions west of the Rhine show a different profile during the following fifth millennium BCE. Here, first farmers carry a higher local Loschbour-related HG component, later increasing to up to 33.3 ± 3% in several individuals in Alsace (OBN B). Mitochondrial data support this finding with a higher average proportion of HG-affiliated haplotypes (U5 and U8) in all French groups from the fifth millennium BCE (15.5%; table S4 and text S4) compared to LBK groups east of the Rhine (1.4%). Although we do not have genomes from LBK individuals from northern France, we can infer from this HG component that admixture processes happened locally after the first farmers arrived. When modeling the European_HG component in the three groups from OBN with qpAdm, we observe an increase in the HG component of between 4.3 and 31.8%, which we attribute to local Loschbour-related sources (table S11). However, this approach cannot be directly applied to other contemporaneous French groups as potential additional movements linked to the Mediterranean route of Neolithic expansion that are supposed to have followed the first LBK farmers’ arrival in northern France could complicate the simple assumption of a two-way HG/farmer interaction (26, 42). The current genomic data do not allow us to distinguish whether the detected signal is confounded by a potential southern contribution. However, the estimated admixture dates for GRG and FLR in northern France indicate older admixture events that occurred more than 30 generations (840 to 930 years) earlier (Fig. 5 and table S17). In accordance with the established chronology of first Neolithic settlements in the French territory, the overlapping/synchronous date estimates obtained for southern ICC sites are consistent with the signal of a first HG contribution in the south of France, followed by a subsequent northward expansion of groups carrying this HG legacy (42).
The GoyetQ2-like HG component detected in individuals from the sites PRI and GRG suggests connections with the Iberian Peninsula where this post-LGM residual component is found, either by an admixture with HG (which could also be local) carrying this component or by contact with first farming communities from Iberia or because of exchanges with Neolithic groups from Iberia during the subsequent fifth millennium. To date, archaeological data are compatible with all three hypotheses (25, 43). However, while we can show affinities between PRI on the Atlantic coast and the Iberian Peninsula, the previously described affinities between the British Isles and Mediterranean Neolithic (39, 40) are currently best explained by sites mainly in Normandy and the Mediterranean area via the Paris Basin. English, Scottish, and Welsh groups show more genetic affinities to northern and southern France and Iberia than to western France (fig. S10). In contrast, Neolithic Ireland shows less affinities with the northern French coastline and the Mediterranean area than other British groups, which could be explained within an Atlantic framework. This whole pattern is in line with the hypothesis of two different phenomena and speeds of Neolithic expansion to the western and eastern British Isles, as proposed by archaeological data (44).
In summary, our study highlights a diverse pattern of cultural and biological interactions between first farmers and indigenous HG along the western Mediterranean coastline and west of the Rhine, which confirms a high variability of processes during the Neolithic expansion, distinct in the proportion of the HG component as well as processes and duration. The genetic structure among HG groups allowed us to track local admixture in early farmers, which not only is higher west of the Rhine compared to central and southeastern Europe but also is largely attributed to local and distinctly WHG-related sources. We show that nuanced sampling as well as increasing cohort numbers of both HG and early farmers can help to unravel the regional dynamics of the Neolithic transition and aid in refining our understanding of the underlying processes and developments over time, at both micro- and macroregional scales. On the basis of our observations, we find that broad-brush models are increasingly less likely to reconcile the full spectrum and details of the farmer-forager interactions and thus advocate the use of models with more regional focus in future studies.
Fig. 4 Maps of variable sources of HG and Anatolian Neolithic ancestries through time.
(A) Proportion of distal sources of Villabruna, EHG, Goyet_Q2, and Anatolia_Neolithic of post-LGM HG individuals (14000–4000 cal BCE) estimated according to MODEL B (table S12). The following five panels (B to F) show Neolithic farmer groups (6000–3500 cal BCE) modeled with the proximal sources Anatolia_Neolithic, KO1, and Loschbour according to MODEL C, in time increments of 500 years each. Transparent colors indicate individuals or groups not sufficiently supported by the models (P < 0.05). Note that not only N22 from the Polish BKG, TGM009, and Blätterhöhle, which carry a substantial proportion of HG ancestry, but also Anatolian_Neolithic ancestry have been modeled with MODEL B and were added to (E) and (F), respectively. See table S14 for model details.
We caution that because the amount of HG ancestry is very low in many Neolithic groups (<10%), it remains difficult to characterize the ultimate source reliably. Nevertheless, our admixture patterns from supported models show clear geographic signals. Neolithic groups associated with the LBK in central Europe (Hungary, Austria, and Germany) carry a small HG proportion, which was likely derived from admixture with HG individuals of the EHG-WHG cline and could have occurred in southeastern Europe during a preceding phase of the Neolithic expansion around 6000–5400 BCE. When using f-statistics of the form f4(Mbuti, test; BDB001, KO1) with our new Mesolithic genome (BDB001) from the Middle-Elbe-Saale region in central Germany as a geographically local HG proxy (instead of using Loschbour, which is located west of the Rhine), we do not find support for a local attraction for LBK groups, but the same pattern as for Loschbour (fig. S5C). This suggests that additional gene flow from neighboring Loschbour-like HG such as BDB001 in central Europe was negligible in the first Neolithic groups. However, the German Baalberge group (4000–3500 BCE) shows a marked increase of such HG ancestry, as well as individuals from the Blätterhöhle group, as has been suggested (5, 6), compared to a combination of both KO1-like and Loschbour-like ancestries for LBK groups (6). We can now show that this increase in WHG ancestry (up to 21.3 ± 1.5% in Baalberge; table S13) is driven by either local Loschbour-like ancestry or an expansion of farming groups from the west carrying this signal during the fifth millennium BCE, as suggested by archaeological data (36). For all studied Neolithic groups west of the Rhine, we observe a different pattern with a higher HG ancestry proportion, even for earliest groups that appears to be of a local (Loschbour-like) HG origin, consistent with archaeological data (2).
Late survival of HG ancestry in central Europe
In contrast to Middle-German Neolithic groups, the Late Neolithic individual TGM009 (~3300 BCE) found in an TRB/Elb-Havel group context shows a different pattern of HG ancestry. When modeled with distal sources, this individual carries 63.6 ± 5.2% Villabruna-related ancestry (see MODEL B; table S12 and Fig. 4). When we use proximal sources (MODEL C; table S13), we estimate the HG component to be split into 48.1 ± 6.4% KO1-related ancestry and 25.8 ± 6.1% Loschbour-like ancestry (P = 0.12). To test whether the subtle EHG-related signal observed in TGM009 (through KO1) could also come from Scandinavian HG, as suggested by the archaeological records of regional late contacts with the Scandinavian Mesolithic (37), we applied f-statistics of the form f4(Mbuti, test; Hungary_KO1, Sweden_Motala_HG). The f4 value is significantly negative (f4 = −0.0035, Z score = −6.383), indicating that TGM009 shares more ancestry with southeastern Europe HG than with Scandinavian HG (table S16). Consequently, qpAdm models with Sweden_Motala_HG in place of Hungary_KO1 received a poor fit (P = 1.32−04; table S16).
In the specific case of TGM009, we also tested whether individuals of the Pitted Ware Culture (PWC) group (~3200–2300 BCE) (38), considered to be “Neolithic HG,” would be a suitable contemporaneous proximal source with MODEL C set of outgroups (table S16). We find that the tests still support the three-way model with Anatolia_Neolithic, Luxembourg_Loschbour, and Hungary_KO1 as a third source (P = 0.12). However, when we add Sweden_Motala_HG to the outgroups, we find the best support for a four-way mixture model (P = 0.087) of 21.7 ± 2.4% Anatolia_Neolithic, 24.4 ± 6.2% Luxembourg_Loschbour, 12.6 ± 4.4% Sweden_PWC, and 41.3 ± 7.3% Hungary_KO1. The small but stable contribution of PWC groups at the northern fringes of the loess belt adds additional complexity to the modes of interaction between Late Neolithic and the last HG groups in central Europe.
Estimating admixture dates between HG and early farmers
To gain further insights into the timing of the admixture of the HG ancestry proportion, we used the software package DATES to estimate the timing of admixture events between early farmers and HG, given in generations in the past (Fig. 5, table S17, and text S10). The resulting date estimates, when ordered temporally and geographically, echo patterns of both groups that are also visible in the PCA (east versus west of the Rhine). It confirms that this pattern is dependent not only on the amount of HG ancestry but also on the qualitative signature of the HG ancestry, as supported by fig. S7. Together, the date estimates for the four southern French groups from sites PEN A and B and LBR B suggest admixture with local HG relatively soon (100 to 300 years) after the arrival of Neolithic farmers about 5850 cal BCE, with an admixture date about 5740–5450 cal BCE for all groups (table S17). These date estimates agree with local archaeological data for the establishment of early farming in this region (18), although we cannot exclude a scenario in which admixture occurred on the Italian Peninsula shortly before.
Fig. 5 Admixture dating per group.
Admixture date estimation according to DATES software, obtained in generations. Radiocarbon date intervals (given with 2-sigma) for each site are black lines; blue diamonds are the estimated admixture date (SE = 1). Admixture is estimated with two sources: European_HG and Anatolia_Neolithic. Number of years calculated on the basis of 28 years for one generation. Admixture time calculated according to the oldest date of the radiocarbon interval for each group (table S17 and text S10).
In the rest of France, post-LBK groups show an older admixture date, which, together with a strong local HG component, place the admixture event with WHG-related HG during the first phase of the local Neolithic (Fig. 4). In contrast, the site OBN reveals a heterogeneous genetic signal forming three distinguishable groups with different HG proportions and corresponding admixture dates (tables S7, S13, and S17). The differences between the date estimates for the three groups suggest that the OBN group shows a genetic substructure rather than a recent ongoing admixture event.
Tracing GoyetQ2 ancestry in France
We also used qpAdm with GoyetQ2, Villabruna, and Anatolia_Neolithic as sources to track Magdalenian-associated GoyetQ2-like ancestry in the new western Neolithic groups (MODEL D; text S9 and table S15) (15). The GoyetQ2-related component is observed in the sites of PRI (6 ± 3%) in western France and GRG (3.7 ± 1.3%) from the Paris basin. Similar to Iberian Neolithic groups, the PRI group (4300–4200 BCE) can be modeled successfully with GoyetQ2 as one of the HG components (fig. S8 and table S15), suggesting either admixture with local HG in western France who retained this post-Last Glacial Maximum component or genetic contacts with Neolithic Iberia in later phases. Of note, individuals from the site PRI on the Atlantic coast indicate a slightly more recent admixture date with European_HG (about 5200 BCE), which is consistent with a later arrival of Neolithic groups in westernmost France (table S17). We plotted these results in fig. S9 to complement the summary map with the GoyetQ2-related component in Neolithic groups, which adds another layer of complexity to the overall genetic picture of Neolithic Europe.
Connections with Britain and Ireland
To address questions at broader scale on the western fringe of Europe, we also investigated the relationship between Britain and Ireland and the European mainland during the Neolithic. Here, we performed f4-statistics of the form f4(Mbuti, Britain/Ireland groups; PRI, test), which measures whether Neolithic groups from Britain and Ireland share more genetic drift with the French Atlantic group PRI than with a test population (see fig. S10). As previously described (39, 40), we confirm that British Neolithic groups share affinities with the Mediterranean Neolithic (LBR_A, France_MN, and Iberia_MN), which is also visible in the WHG-rich HG proportion (Fig. 4). However, on the basis of the results from our French Neolithic sites, we suggest that English, Welsh, and Scottish groups are connected to the Mediterranean Neolithic sphere not only via the Atlantic coast but more plausibly also via Normandy (FLR), the Paris Basin (GRG), and southern France (LBR A, post-ICC groups).
DISCUSSION
The expansion of Anatolian Neolithic farmer–related ancestry across Europe has been described for many geographic regions (5–7, 9–11, 39, 40). A recurrent pattern of increasing HG gene flow in individuals associated with a farming lifestyle many centuries after initial contact and tentatively termed “HG resurgence” (5) has hitherto been observed in the Iberian Peninsula, northern/central Europe, the Carpathian Basin, and the Balkans (6). However, a comparison of genomic data in the light of the proposed main routes of Neolithic expansion has not yet been attempted. Our new results provide important insights for the region of modern-day France where both routes had intersected. This region has not been documented so far, but is ideally positioned to address these questions. The different Mesolithic genetic substratum in Europe (6, 7, 15, 33) enables us to track both Neolithic expansion routes in the form of the quantity and quality of the admixed HG component observed in Neolithic groups.
Neolithic groups from southern France, which are a part of the Mediterranean/ICC route of expansion, show a different genetic profile when compared to early periods of expansion in other regions, with a substantially higher HG component than groups associated with the continental route (up to 56 ± 2.9% of HG component for PEN B; table S10). Although the sites PEN and LBR date ~400 years (i.e., ~16 generations) later than the first local coastal settlements in Liguria, Provence, and Mediterranean Languedoc, they suggest a recent local admixture event (between three and six generations earlier; Fig. 5 and table S17). Archaeological research has argued for increased interaction between incoming farmers and indigenous HG in the western Mediterranean during a second stage of the Neolithization process and especially in areas with higher HG population densities, e.g., the Tosco-Emilian Apennine and Po plain (18). We are now able to confirm that these contacts left a traceable biological signal during the Neolithic expansion in southern France. From an archaeological perspective, this suggests that HG have contributed to the clear changes observed within the material culture postdating the pioneer phase. Note that ICC individuals from the eastern Adriatic coast have only a very small amount of HG ancestry with a greater affinity to central European groups (see table S8). This fits with the hypothesis of a differentiation of technical traditions within material cultures observed from both sides of Apennine Mountains in Italy: an Adriatic tradition connected to the Balkans and a Tyrrhenian one whose origin is still unknown (41). It is tempting to associate such a strong HG component on the Tyrrhenian side with the characteristic/specific pottery traditions observed in this same region and to consider these original traditions the result of a HG reinterpretation (41). However, the scarcity of genomic data available from central and southern Italy currently does not allow this hypothesis to be tested directly. Moreover, ICC individuals from the Iberian Peninsula also carry less HG ancestry. Together, this rejects the hypothesis that ICC-associated individuals represent a uniform genetic horizon per se and argues for more regionally nuanced scenarios of interaction.
Early central European farmers carried a very small percentage of HG component (about 5% on average), which is most likely nonlocal and instead derives from admixture in Transdanubia during an early stage of expansion. This not only matches general observations in the archaeological record (2) but could also explain why the earliest LBK lithic assemblages resemble those of the Late Mesolithic Blade and Trapeze Complex. It also confirms previous aDNA studies (5, 6), which argued for a fast spread of the first farmers across the German loess regions. LBK groups from southwestern and eastern Germany share more affinities with KO1 than with Loschbour (fig. S5). The admixture date estimates for southern German sites, SMH and SCH, 19.2 ± 3.8 and 12.3 ± 8.2 generations, respectively, are younger or contemporaneous to admixture date estimates from the Carpathian Basin and Austria (Fig. 5 and table S17). The temporal lag, and the subtle increase of shared KO1-like HG ancestry, allows us to trace LBK groups chronologically and matches well with the proposed model of LBK expansion from the core region in Transdanubia as suggested by archaeological research (14). However, the present picture does not explain the increasing amount of archaeological evidence of contact with HG and groups with southern influences in more western sites, particularly during the early LBK, such as the debated La Hoguette and Limburg phenomena (16).
In contrast to the situation in central Europe, regions west of the Rhine show a different profile during the following fifth millennium BCE. Here, first farmers carry a higher local Loschbour-related HG component, later increasing to up to 33.3 ± 3% in several individuals in Alsace (OBN B). Mitochondrial data support this finding with a higher average proportion of HG-affiliated haplotypes (U5 and U8) in all French groups from the fifth millennium BCE (15.5%; table S4 and text S4) compared to LBK groups east of the Rhine (1.4%). Although we do not have genomes from LBK individuals from northern France, we can infer from this HG component that admixture processes happened locally after the first farmers arrived. When modeling the European_HG component in the three groups from OBN with qpAdm, we observe an increase in the HG component of between 4.3 and 31.8%, which we attribute to local Loschbour-related sources (table S11). However, this approach cannot be directly applied to other contemporaneous French groups as potential additional movements linked to the Mediterranean route of Neolithic expansion that are supposed to have followed the first LBK farmers’ arrival in northern France could complicate the simple assumption of a two-way HG/farmer interaction (26, 42). The current genomic data do not allow us to distinguish whether the detected signal is confounded by a potential southern contribution. However, the estimated admixture dates for GRG and FLR in northern France indicate older admixture events that occurred more than 30 generations (840 to 930 years) earlier (Fig. 5 and table S17). In accordance with the established chronology of first Neolithic settlements in the French territory, the overlapping/synchronous date estimates obtained for southern ICC sites are consistent with the signal of a first HG contribution in the south of France, followed by a subsequent northward expansion of groups carrying this HG legacy (42).
The GoyetQ2-like HG component detected in individuals from the sites PRI and GRG suggests connections with the Iberian Peninsula where this post-LGM residual component is found, either by an admixture with HG (which could also be local) carrying this component or by contact with first farming communities from Iberia or because of exchanges with Neolithic groups from Iberia during the subsequent fifth millennium. To date, archaeological data are compatible with all three hypotheses (25, 43). However, while we can show affinities between PRI on the Atlantic coast and the Iberian Peninsula, the previously described affinities between the British Isles and Mediterranean Neolithic (39, 40) are currently best explained by sites mainly in Normandy and the Mediterranean area via the Paris Basin. English, Scottish, and Welsh groups show more genetic affinities to northern and southern France and Iberia than to western France (fig. S10). In contrast, Neolithic Ireland shows less affinities with the northern French coastline and the Mediterranean area than other British groups, which could be explained within an Atlantic framework. This whole pattern is in line with the hypothesis of two different phenomena and speeds of Neolithic expansion to the western and eastern British Isles, as proposed by archaeological data (44).
In summary, our study highlights a diverse pattern of cultural and biological interactions between first farmers and indigenous HG along the western Mediterranean coastline and west of the Rhine, which confirms a high variability of processes during the Neolithic expansion, distinct in the proportion of the HG component as well as processes and duration. The genetic structure among HG groups allowed us to track local admixture in early farmers, which not only is higher west of the Rhine compared to central and southeastern Europe but also is largely attributed to local and distinctly WHG-related sources. We show that nuanced sampling as well as increasing cohort numbers of both HG and early farmers can help to unravel the regional dynamics of the Neolithic transition and aid in refining our understanding of the underlying processes and developments over time, at both micro- and macroregional scales. On the basis of our observations, we find that broad-brush models are increasingly less likely to reconcile the full spectrum and details of the farmer-forager interactions and thus advocate the use of models with more regional focus in future studies.