aDNA preparation and extraction

46 individuals were excavated from Gougenheim necropolis and are currently stored at the Center of Conservation and Archaeology at Sélestat (Alsace) and available for study. We chose 22 individuals with exploitable teeth for DNA analyses, spatially distributed in all Gougenheim archaeological site sectors and representing all types of inhumation structures (structure grouping individuals in different positions, structures grouping several CV individuals, or structure with only one NCV individual—all details about samples are present in S1 Table). While the aDNA analyses were not anticipated before the excavation, the individuals were excavated without specific aDNA care. Consequently, the teeth sampled were systematically decontaminated, i.e., scraped, cleaned with bleach and subsequently exposed to UV radiation for 20 minutes on each side. All established aDNA guidelines were then followed to minimize contamination during all subsequent steps of analyses conducted in the aDNA facilities of the UMR PACEA (Bordeaux University). To trace the source of potential human contamination, all of the persons (N = 11) in contact with the human remains (from the excavation to the aDNA lab, each researcher considered has handled all human remains) were genotyped (S7 Table). The samples were then reduced to powder and incubated overnight in lysis buffer (0.5 M EDTA, pH 8, 25mg/mL proteinase K, and 0.5% N-Lauryl sarkosyl). The procedure of Mendisco et al. [37], which uses the DNA “NucleoSpin Extract II” kit (Macherey-Nagel, Düren, Germany), was followed to extract the DNA.

aDNA analyses

Eighteen mitochondrial SNPs and 10 Y chromosome SNPs were typed through one multiplex using MALDI-TOF MS-based SNP genotyping (iPLEX® Gold technology, Sequenom, Inc., San Diego, CA). This genotyping was conducted as previously described [37] to assess the DNA conservation for every sample and determine the mitochondrial (maternal) and Y chromosome (paternal) haplogroups. All primers used for these experiments and procedure details are available in Rivollat et al. [15]. Four overlapping fragments of the mtDNA HVR-I control region were also amplified and analyzed following the procedures described in Rivollat et al. [15] to determine the maternal haplotypes of the individuals (nps 16,024–16,380). All reported mutations were established according to the revised Cambridge Reference Sequence (rCRS) [38]. Each individual's haplogroup and haplotype were determined based on PhyloTree.org (mtDNA tree build 16 [39]). The sequences were deposited in the GenBank database (http://www.ncbi.nlm.nih.gov/genbank/; accession numbers KY485916-KY485933).

Descriptive analyses

Sequences from all published aDNA data dating from the Mesolithic to the Late Neolithic periods were compiled (14,500–2,600 cal. BC). We selected farmers’ groups attributed to well-identified waves of Neolithization, i.e., starting with the Stärcevo groups for Central Europe and the Danubian wave of Neolithization and starting from the Cardial groups for the southern Mediterranean wave (see S6 Table). The Gougenheim group was gathered with the data published for the Michelsberg site of Bruchsal-Aue, in western Germany [23]. The global sequence dataset was divided into 16 chronological, geographical and cultural groups in order to discuss (i) the implication of different anterior groups in the Michelsberg gene pool constitution and (ii) the relationships of the Michelsberg group with contemporaneous and later populations.

Haplotype diversities (S4 Table) and Fst values (S2 Table) between populations were calculated using Arlequin software (version 3.5.1.2 [40]). All HVR-I mitochondrial sequences available for nps 16,024–16,380 were used (shorter sequences were not included; N = 531). Classical PCA based on haplogroups frequencies was performed considering major European and near eastern haplogroups, that is, U5, U8, U(xU5,U8), H, HV, V, J, T, K, N1a, N(xN1a), X, I, W and others. Based on these haplogroups' frequencies, PCA was plotted using R version 3.1.2 (Pumpkin Helmet) (Fig 3) and Ward Clustering (using Euclidean Similarity Index and 5000 bootstraps) was performed with Past version 3.14 software [41] (S1 Fig). MDS based on Fst values (S2 Table) was plotted using R version 3.1.2 (Pumpkin Helmet) (S2 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Principal component analysis (PCA) of the ancient mtDNA dataset.

(a) PCA performed with haplogroup frequencies. (b) Circle of correlation. (c) Chronological overview of groups with the number of individuals between parentheses (see S6 Table for details). Colors are the same as in Fig 2. HGCN: Hunter-gatherers from Central Europe; HGW: Hunter-gatherers from West Europe; HGS: Hunter-gatherers from South Europe; STA: Starcevo-Körös; LBKT: Linearbandkeramik in Transdanubia; LBK: Linearbandkeramik; CRD: Cardial; EPI: EpiCardial Early Neolithic in Spain; RSG: Grossgartach/Planig-Friedberg/Rossen; MNF: Middle Neolithic in France; SCG: Schöningen; MICH: Michelsberg; BAC: Baalberge; LNS: Late Neolithic in Spain; SMZ: Salzmünde; BEC: Bernburg.

https://doi.org/10.1371/journal.pone.0179742.g003

We used Ancestral Shared Haplotype Analysis ASHA [42], as modified by Szécsényi-Nagy et al. [43], accounting for the temporal succession of cultures, to ascribe mtDNA haplotypes to particular cultures or time periods and to identify the number of ancestral lineages in each culture (S3 Fig). Thirteen Neolithic farmer datasets were set in a specific chronological order: STA, LBKT, LBK, RSC, SCG, MICH, BAC, SMZ, BEC for Central Europe (Hungary, Germany and Alsace) / CAR, EPI, MNF, LNS for western Europe (Spain and western France, see S5 Table). Because ancient European hunter-gatherers were represented by a very poor and certainly non-representative number of haplotypes (compared with farmers), we first considered the U/U2/U4/U5/U8 haplotypes to be "hunter-gatherer haplotypes" ("H-G" haplotypes S5 Table) [15, 44]. We then listed all haplotypes shared between ancient Neolithic groups (before 4,000 cal. BC) from both Neolithization waves (groups STA, LBKT, LBK, CAR, EPI) and considered them to be ubiquitous farmer haplotypes that were phylogeographically uninformative ("F" haplotypes, S5 Table). Each remaining lineage within a given cultural dataset was traced back to its earliest match in the chronology and regarded as an ancestral lineage that arose in this culture for the first time (S5 Table). This approach enabled us to estimate the number of mtDNA lineages directly transmitted from ancestral farmer groups to more-recent ones, thereby providing primary insight into each group’s maternal legacy and discussing the regional maternal (dis)continuity.

Michelsberg group origin

A striking diversity of haplogroups characterized the group of Gougenheim, including H, J, K, N1a, T, W, X and U5 lineages. The maternal data gathered for the Gougenheim group (N = 21) were then grouped with the mtDNA data previously published for 9 individuals from the Michelsberg site of Bruchsal Aue in current Germany [23] and permitted the more precise description of the maternal gene pool of Michelsberg communities ("MICH" group, S6 Table). We then compared the MICH maternal pool, both at the haplogroup frequencies (PCA (S3 Table and Fig 3) and Ward Clustering (S1 Fig)) and at the haplotype composition levels (F_ST values and MDS) with ancient H-G and Neolithic groups (see S2 Table and S2 Fig for details). An ancestral shared haplotype analysis (ASHA [42]), as modified by Szécsényi-Nagy et al. [43], was also conducted to identify the number of lineages inherited from previous groups in each culture (S5 Table and S3 Fig). Both PCA and MDS evidence showed that the diversity of the MICH group fitted well with the known ancient European Neolithic population diversity (Fig 3 and S2 Fig). Nevertheless, when going deeper in the ancient groups' affinities, an important genetic discontinuity could be highlighted in Alsace / western Germany between Middle Neolithic groups culturally affiliated with Roessen and Grossgartarch ("RSG") and the Late Neolithic MICH group. This maternal rupture could be linked to important differences in terms of genetic legacy from (i) Central European farmers, (ii) western European farmers and (iii) ancient European H-G. Indeed, descriptive analyses (Fig 3 and S2 Fig) consistently localized the RSG group among the Central Europe variability (blue), along with Starčevo groups ("STA") in Hungary or Linearbandkeramik groups ("LBKT" in Transdanubia and "LBK" in Germany) in Central Europe [5, 17, 45, 46]. Moreover, the ASHA permitted determining that 28.1% of the RSG mtDNA haplotypes were directly derived from the STA, LBKT and LBK groups (S3 Fig). At the opposite, the MICH group stood at a clearly more distant position in relation to Central European farmers (especially obvious in the MDS, S2 Fig) and presented far fewer haplotypes (7.4%) directly derived from STA, LBKT or LBK groups. The maternal differentiation characterized between RSG and MICH can also be explained by higher maternal affinities between the MICH group and farmers from the Paris Basin (France, MNF group). Indeed, all descriptive analyses showed small genetic distances between the MICH and Middle Neolithic groups from France (Paris Basin, "MNF" group). In the Ward Clustering conducted (S1 Fig), the MICH group formed a cluster with the MNF group, and this specific cluster was supported by the most important bootstrap value. Surely correlated to these affinities with farmers from the Paris Basin, a strong H-G legacy was noted in the MICH group, slightly visible on the PCA (Fig 3) and illustrated by the high frequency of typical H-G maternal lineages (14.8%, S3 Fig). This H-G legacy is far more important than H-G legacies measured in Early and Middle Neolithic farmer groups from Hungary (STA = 4.1%, LBKT = 5.1%), Germany (LBK = 2.7%, Schöningen group "SCG" = 9.1%) and Alsace (RSG = 6.3%), but similar to what could be measured for the same periods in southwestern and western Europe (EPI = 22.7%, MNF = 14.6%). In Germany, the resurgence of a significant number of H-G lineages was observed thereafter, around 3,100 cal. BC in the context of the Bernburg culture ("BEC") (S3 Fig) [5, 11].

Taken together, our data highlighted a major genetic break between Middle Neolithic LBK-derived groups in Alsace and the subsequent Late Neolithic Michelsberg groups. This genetic discontinuity appeared to be linked to new affinities with Middle Neolithic farmers from the Paris Basin, correlated to a new and important H-G legacy. All of the evidence gathered is in favor of a close genetic connection of the Michelsberg group with farmer communities from the Paris Basin region. When chronologically and geographically replaced, the data gathered suggest that the diffusion of the Michelsberg culture was linked to a movement of human groups from West (Paris Basin) to East (Rhine region). Since an admixture between H-G and farmers maternal components has been demonstrated in the Middle Neolithic groups of the Paris Basin (Gurgy "les Noisats", 5,000 to 4,000 cal. BC [15]), we can propose that the important frequencies of H-G haplogroups measured in the MICH group may have been indirectly inherited from the MNF group in the Paris Basin (see below). The occidental origin (in the Paris Basin) of the Michelsberg culture has been proposed for a long time, based on archaeological data [20–22]. This hypothesis considered that the Michelsberg culture did not appear in the Rhine region as alternatively supposed [47, 48] but emerged in the Paris Basin from the interactions between the Chasséen culture, itself originating from southern France, and the Noyen culture, itself deriving from the interaction between Cerny (Paris Basin) and Chasséen (S4 Fig). The major changes observed in pottery traditions between the Michelsberg culture and anterior cultures in Alsace and Germany, combined with the diffusion of specific ceramic features from southwestern traditions in the Paris Basin, led the author to consider that migrations may have been involved in such important transformations [20]. Consistently, the paleogenetic data provided by our study supported the involvement of farmer group migration to explain such a global cultural rupture. The groups involved in the Michelsberg diffusion originated in the Paris Basin and carried in Alsace few LBK-derived maternal lineages but a significant number of Paris Basin farmers and H-G maternal legacy.

New elements for discussing the dynamics of H-G/farmer interactions during the Neolithic period

The new mtDNA data gathered for the Michelsberg group also permit filling an important chronological and geographical gap to complete the scenario of H-G—farmer interactions during the Neolithic periods. ASHA analysis showed that important H-G legacy was first apparent in northern Spain during the Epicardial period (5,200–4,000 cal. BC, EPI = 22.7%; S3 Fig). The absence of H-G lineages in the Cardial group "CRD" may be due to the very poor quantity of data available (N = 11). Subsequently or simultaneously, such a H-G legacy appeared in the Paris Basin during the Middle Neolithic (5,000–4,000 cal. BC, MNF = 14.6%). In the meantime, very rare occurrences of H-G lineages could be observed in Central Europe, in accordance with recent genomic data [11]. The available mtDNA data then support the scenario of an increasing H-G admixture into farmers’ groups migrating farther West in Europe (in Spain and France) during the Neolithization processes. This hypothesis has already been proposed through the analysis of the maternal gene pool of the Paris Basin Middle Neolithic farmers [15] and has been recently reinforced by coalescent simulations [49].

At the beginning of the Late Neolithic, the important H-G legacy was still visible in Spain (3,700–3,500 cal. BC, "LNS" = 25.6%), whereas the MICH group was the only group in Central Europe to present a significant H-G maternal lineage frequency (14.8%). The first noticeable resurgence of H-G maternal lineages in Central Europe occurred only at the end of the fourth Millennium, in the context of the Bernburg culture (BEC = 29.4% [5, 11, 44]). This H-G legacy resurgence in Central Europe has already been largely described [5] and is illustrated in our PCA by the progressive and chronological displacement of Central Europe farmer groups toward H-G groups. The H-G legacy resurgence observed in the Bernburg context has been linked to the important admixture demonstrated between H-G and farmers in Scandinavia (in the North European Plain), one millennium earlier, in the context of the emergence of the Funnel Beaker Culture (TRB or TBK) [3, 4, 50]. The Bernburg groups, a late representative of the TRB groups in Central Europe, must have inherited their important frequencies of H-G haplogroups from their northern ancestors. Nevertheless, a recent genome-wide study conducted on 69 ancient Europeans [11] has provided arguments in favor of a different source of the H-G legacy resurgence detected in Central Europe. Indeed, the H-G component found for individuals from Central Europe and dated from the fourth Millennium showed more genomic affinities with H-G from western Europe (including La Braña 1- remains from Spain [51], Loschbour—H-G from Luxembourg [12] and KO1 H-G from Hungary [10]) than with H-G groups from eastern or northern Europe. Even if the genomic data available for the European H-G are still scarce and the exact source of H-G ancestry is consequently not fully established yet, these data suggest two possible sources of H-G legacy resurgence in Central Europe, i.e., from western Europe H-G and northern Europe/Scandinavian H-G. Evidently, further genomic analyses on H-G and Early/middle Neolithic farmers are necessary to delineate the process of H-G—farmer interactions during the Neolithic period in Europe, as well as in the regions discussed in the present paper. However, the gap observed between the appearance of the H-G legacy in Spain and in France between 5,000 and 4,000 cal. BC and the noticeable resurgence of the western H-G legacy in Central Europe from the fourth Millennium BC [5, 44] could be partly filled, both chronologically and geographically, by our Michelsberg group. Indeed, the migration of Michelsberg groups originating from the Paris Basin and carrying an important H-G legacy in Alsace and Germany, might have contributed to the subsequent resurgence of western H-G ancestry in these regions. Since previous analyses have demonstrated that the MNF group had inherited a noticeable fraction of its maternal gene pool from descendants of farmers associated to the Mediterranean neolithisation wave [47], we cannot exclude that part of the H-G legacy measured in MICH group finds its origin in southwestern farmers groups.

Gougenheim's mortuary practices

Among the 21 individuals providing mitochondrial profiles, 9 were found in the CV position, whereas 12 were found in the NCV position (Fig 1B). Only 2 genotyped individuals originated from some possible multiple “asymmetrical” deposit (structure 1029 grouping individual 1029–1 in the CV position and 1029–2 in the NCV position, but at two distinct levels), when all remaining genotyped individuals found in the NCV position originated from multiple “non-conventional” deposits and all remaining genotyped individuals found in the CV position originated from individual “conventional” deposits. Both the CV and NCV groups showed lineages originating from the (south)western regions as described above. Whereas all "H-G lineages" (of potential western origin) were found concentrated in the NCV group, the CV group contained a strong proportion of haplogroups H (H, H1 and H3) and X, which were more common in southern European and Paris Basin farmers (Fig 3). A similar phylogeographic signature could then be observed in both the CV and NCV groups, indicating a unique cultural and biological group at the local level but differently treated in death. Interestingly, isotope analyses demonstrated diet homogeneity between both groups and, as such, supported a shared geographical origin [52].

If we highlighted that the Gougenheim group represented a phylogeographically consistent biological group, we note differences in maternal gene pools of sub-groups CV and NCV. Although the small number of individuals in each group (and the very important mitochondrial diversity measured in each group, see below) did not allow any satisfactory statistical testing of differentiation, we could see that haplogroups W and X were specific to group CV, whereas haplogroups N1a, U5 and T were specific to group NCV (Fig 1B). Only two HVR-I haplotypes were found to be shared by both the CV and NCV groups: haplotypes J1_16069–16126 and H_CRS. Nevertheless, these haplotypes are very frequent in Neolithic populations (found respectively at 6.7% and 14.3% in the compiled database) and thus represent a poor indicator of group affinities or maternal kinship. We consider that the distinct maternal gene pools characterized for both the CV and NCV groups, submitted to distinct mortuary treatment, points to a relative genetic isolation of the two groups from each other. This genetic isolation may have been related to the social stratification of the community, with combined endogamy in each social group and the genetic isolation of distinct social stratum and/or to some geographic distance between separate communities. These assumptions nicely echo the hypothesis of sacrifices retained for the Gougenheim site. Ethnological and archaeological examples indeed show that the victims of human sacrifice are often slaves, sometimes specifically bought for the occasion, but more regularly abducted during ambushes or razzia in some enemy villages of the neighborhood or taken into captivity during warlike events [24]. In such circumstances, a genetic differentiation between distinct social groups corresponding to master vs. slave groups or between adversarial neighbor groups (allied communities are characterized by the exchange of women and genes) can be expected. In consequence, the obtained paleogenetic data fit well with the sacrifice hypothesis, even if they cannot be conclusive. However, the “peripheral accompaniment” assumption—archaeologically unlikely in Gougenheim—involving the same social stratum of subjugated people, is theoretically possible when considering only DNA data.

Both sub-groups CV and NCV presented an important maternal diversity (0.83 +/- 0.13 and 0.92 +/- 0.09, respectively), in total accordance with mitochondrial diversities generally measured for the Neolithic period in Europe [15, 17, 53] (S4 Table). As already proposed for other Neolithic farmer groups [43, 53, 54], this maternal diversity could be an indicator of a patrilocal social system (although Y chromosome data are necessary to directly test this hypothesis). Interestingly, each multiple structure also showed a striking maternal diversity. For instance, the structures 1076 and 1077, grouping 4 NCV individuals each, presented a total of 6 distinct mitochondrial sequences over the 7 individuals genotyped (S1 Table). Moreover, when considering all of the structures with aDNA data, no grouping or spatial structure could be highlighted according to maternal lineages (Fig 1A). Finally, concerning the individuals found in multiple “non-conventional” deposits, no shared haplotype could be detected. Only two individuals found in the multiple “non-conventional” structure 1077 shared the haplotype K_16224–16311, but once again, this very common haplotype in Neolithic groups (2.02%) may not be indicative of a maternal kinship. Considered together, the mtDNA data compiled for the Gougenheim site project the image of the random sampling of individuals inside both sub-groups CV and NCV. The biologically and socially differentiated sub-groups, submitted to very different mortuary treatments, must have been large enough to present such maternal diversities. Even if the presented arguments cannot be conclusive, the picture depicted at the genetic level is in total accordance with the idea of the random kidnapping of the individuals submitted to sacrifice. Whether these individuals were neighbors randomly abducted for religious needs or were slaves remains an open question.