Our results suggest that the two Y. pestis genomes from Warburg belonged to distinct strains. Due to their large genetic distance (Figure S7), it can be assumed that the younger Warburg_2 did not evolve from Warburg_1 and that both infections represent independent events. This is supported by the radiocarbon dates (Figure S1, Table S1) that placed the two events up to 400 years apart and by the observation that the genomes were detected in unrelated individuals who were buried in different gallery graves.
Interestingly, despite the screening of numerous specimens from each gallery grave (total n = 133), we found no further infections with Y. pestis or any other pathogen. In another collective WBC grave at the site of Niedertiefenbach (n = 42; present-day Germany), no signs of pathogens were detected either9. It must be acknowledged that pathogen-negative results do not necessarily mean absence of infection, as taphonomic processes may have degraded any microbial traces. In addition, both the Warburg and Niedertiefenbach samples consisted of petrous bones as well as teeth, the latter being the better source material for the detection of blood-borne viruses and bacteria11. These limitations notwithstanding, the arguments presented above (i.e., independent infection events at Warburg) and the overall small number of 2 positives among the 175 tested WBC individuals suggest that the collective graves were not used for the burial of victims of a plague outbreak or other epidemics, as previously suggested for the same period3. The few cases for the WBC are consistent with the results of other large-scale pathogen screenings that have so far revealed only single infections with human pathogens (Y. pestis, Salmonella enterica) or endemically occurring infections (hepatitis B virus, parvovirus B19, Helicobacter pylori) in Neolithic remains12. Also, the mortuary practice of single and multiple inhumations during that period does not indicate mass mortality, as would be expected in an epidemic. The findings from the Neolithic are thus in marked contrast to the short-term mass burials and the high pathogen load seen in the Middle Ages13,14.
The two Warburg genomes increase the number of Y. pestis genomes from the LN to four. All LN genomes were distinct from each other, reflecting separate lineages (Fig. 2, Figure S7). This diversity and the basal position of the LN Y. pestis lineages could have allowed them to survive in various environments and a wide host spectrum. According to the current phylogeny, the LN strains gave rise to two lineages, one from which the pathogens of the deadly Justinianic and medieval plagues emerged and another that lead to the LNBAs (Fig. 2). The LNBA clade went extinct sometime in the last millennium BCE4 (Fig. 3). For more than 2000 years, the LNBA strains were the dominant Y. pestis lineage in humans across Eurasia4. They may represent an adaptation to a very specialized Y. pestis ecology (e.g., host(s)), as reflected by the increasing pseudogenization of bacterial genes over time4. This process could have led to the evolutionary dead-end of the LNBA lineage and to a less severe, perhaps even chronic, manifestation of plague in humans that resembled an endemic rather than a pandemic disease2,4.
During the LN, woodland clearance increasingly created open landscapes in central and northern Europe15–18 that attracted a variety of new rodent and bird species (e.g., European hamster Cricetus cricetus19, white stork Ciconia ciconia20, grey partridge Perdix perdix21) originally native to the steppe further east or south. Some of these species could have been natural reservoirs and vectors of Y. pestis1, but if so, how could they have transmitted the bacterium to humans? An infection would have been feasible through close contact with a Y. pestis-positive animal22. However, this scenario would likely lead to single zoonotic events that might have caused limited outbreaks, and it does not explain the wide geographical distribution of the LN strains. Another route of transmission, well documented today (for a comprehensive overview of this topic, see1), would have relied on a domesticated carnivore (i.e., dog) that can bridge the distance between rodents/birds as primary Y. pestis hosts and humans. Dogs can develop primary pneumonic plague, which does not require flea adaptation, and can therefore directly infect humans1,23. Interestingly, the archaeological record during the LN shows increased numbers of dogs that were likely used for hunting and herding24,25. Dog teeth were also often made into pendants and jewelry24,25. To test the hypothesis that the Neolithic dog could have been an intermediate host for Y. pestis, we screened publicly available shotgun datasets generated from Eurasian Neolithic (n = 15) and Bronze Age dogs (n = 6)26–30. In a dataset of one canine (labelled C90), we were indeed able to detect Y. pestis-specific reads that map to the chromosome and all three plasmids and show the typical damage profiles for ancient DNA (Table S4). C90 consisted of a mandible found in the cultural layers of the Pitted Ware Culture settlement site Ajvide on the island of Gotland in present-day Sweden and dated to 4900 − 4500 cal BP (Fig. 1A)27. Although the data was originally only generated for canine population genetic analyses, the reads (not observed previously) demonstrate the presence of Y. pestis in the dog C90, for which we could reconstruct 8% of the 1 x bacterial genome. Due to the low coverage, we did not explore the placement of the canine Y. pestis in the phylogenetic tree.