We report genome-wide data for 33 Ashkenazi Jews (AJ), dated to the 14th century, following a salvage excavation at the medieval Jewish cemetery of Erfurt, Germany. The Erfurt individuals are genetically similar to modern AJ and have substantial Southern European ancestry, but they show more variability in Eastern European-related ancestry than modern AJ. A third of the Erfurt individuals carried the same nearly-AJ-specific mitochondrial haplogroup and eight carried pathogenic variants known to affect AJ today. These observations, together with high levels of runs of homozygosity, suggest that the Erfurt community had already experienced the major reduction in size that affected modern AJ. However, the Erfurt bottleneck was more severe, implying substructure in medieval AJ. Together, our results suggest that the AJ founder event and the acquisition of the main sources of ancestry pre-dated the 14th century and highlight late medieval genetic heterogeneity no longer present in modern AJ.


We have presented the first genome-wide data from historical AJ individuals. We used the data to refine the picture of early AJ origins. The ancestry of EAJ was closely related to that of modern AJ, as evidenced by the PCA, ADMIXTURE, and qpWave analyses, suggesting overall genetic continuity of AJ over the past ≈700 years. However, EAJ individuals had more variable ancestry than MAJ and were possibly stratified by the presence of a minor Eastern European ancestry component. Multiple lines of evidence suggest that the EAJ population had already experienced a “bottleneck” shared with MAJ: the high frequency of Ashkenazi founder mtDNA haplogroups; and the presence of Ashkenazi-specific pathogenic variants, other AJ-enriched alleles, and long runs of homozygosity. Carriers of the K1a1b1a mtDNA founder haplogroup seem to have descended from an even smaller set of founders. In agreement with previous studies [19, 23, 25], we date the onset of the expansion in AJ population size to about 20-25 generations ago (see additional discussion in SI 2). Our ancient DNA data allowed us to identify patterns in the history of AJ that would not have been otherwise detectable from modern genetic variation. Specifically, our genetic results suggest that the AJ population was structured during the Middle Ages. Within Erfurt, one group of individuals had an enrichment of Eastern European-related ancestry (Figure 1 and Figure 2B), while the other had ancestry very close to that of MAJ of Western European origin and modern Sephardi Jews (Figure 1 and Figure S10). The two groups also had significantly different levels of enamel δ18O (Figure 2C). Medieval AJ may have been structured even beyond Erfurt, based on our inferred demographic model (Figure 3E). In contrast, present-day AJ is a remarkably homogeneous population [17, 23, 33]. This suggests that even though the available under aCC-BY-NC 4.0 International license.was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made bioRxiv preprint doi: https://doi.org/10.1101/2022.05.13.491805; this version posted May 16, 2022. The copyright holder for this preprint (which18 overall sources of ancestry remained very similar between medieval and modern AJ, endogamy and within-AJ mixture since medieval times have contributed to the homogenization of the AJ gene pool. We found that a plausible model for the ancestral sources of EAJ (Figure 2A) include groups related to people in South-Italy (about 70%, who themselves plausibly might harbor Middle East-related ancestry), the Middle East (about 15%), and Eastern Europe (about 15%). Models with North-Italians as a source were also plausible, with an ancestry proportion of about 45% to each of North-Italians and Middle Easterners. The ancestry proportion estimates using North-Italians are closer to previous estimates using modern SNP and sequencing data [19, 35], but a North-Italian source was less favored by qpAdm (Table S7). While these results could be consistent with a model where the Middle Eastern ancestry in AJ has not been as large as previously thought, complicating the picture are (i) our inability to identify a satisfactory model for modern AJ; (ii) the historically variable levels of Middle Eastern ancestry in Italy [49, 88-90] (SI 2); and (iii) the possible problems when modeling an ancient population with modern sources using qpAdm [50] (although see our robustness tests in Table S7). Therefore, the direct contribution of ME sources to AJ ancestry may be higher than estimated (SI 2). Either way, the substantial Southern European ancestry we inferred adds weight to the evidence that early AJ descended, at least partly, from Italian Jews (SI 2). The estimate of about 15% Eastern European-related ancestry is consistent with a previous study [35]. The identification of this source as Eastern European relies on the f4 results (figs. S13, S14) and the qpAdm models (Table S7); however, this ancestry might derive from a broad area across Central or Eastern Europe, which may accord with recorded migration into Erfurt from Bohemia, Moravia, and Silesia (SI 1). For an additional discussion on the historical interpretation of these results, see SI 2. As with other ancient DNA studies, our historical inferences are based on a single site in time and space. This implies that our data may not be representative of the full genetic diversity of early AJ, as we have indeed inferred (Figure 3E). However, even for a single site, our sample size was relatively large (>30), and we were able to capture substructure not present in MAJ. Another limitation is the reliance of our demographic models on a relatively small number of runs of homozygosity, which are difficult to infer from pseudo-haploid data. In particular, several models were disqualified due to mismatch with observed counts of short ROH or IBD segments, which are difficult to accurately call (Figure 3D; Figure S25). Our inferred demographic model (Figure 3E) should not be interpreted as a complete and precise demographic reconstruction; rather, it should be viewed as a simplified model (perhaps one among many) that captures the main features of the observed genetic data. Similarly, our models for the ancestry of EAJ (Figure 2A) may not be the only plausible models, and the ancestral sources we inferred should be interpreted as proxies, distant in time and space, of the true ancestral population