Neolithic Yersinia pestis infections in humans and a dog

Plague, caused by Yersinia pestis, is a major historical pathogen with at least three known pandemics. Genetic evidence indicates human infections for more than 5000 years, with the oldest LN genomes previously detected in a hunter-gatherer from Rinnukalns, Latvia, and a farmer from Gökhem, Sweden. These early LN strains lacked key virulence factors and flea-adaptive mutations characteristic of later pandemic forms. Critical open questions include whether LN detections represent isolated zoonoses or the onset of a sustained, human-to-human pandemic across Eurasia, and how transmission to humans occurred prior to the evolution of efficient flea-borne mechanisms. Various rodents could have served as primary hosts, and human exposure may have been facilitated by dogs that hunted small mammals. This study presents two new LN Y. pestis genomes from human individuals at Warburg, Germany, and evidence of infection in a contemporaneous dog, to clarify early Y. pestis evolution and ecology.

Prior studies identified two LN Y. pestis genomes (Rinnukalns, Latvia; Gökhem, Sweden), showing early-stage evolution lacking flea-borne transmission adaptations, and a later Late Neolithic and Bronze Age (LNBA) lineage (ca. 4800–2500 cal BP) that spread widely across Eurasia. These findings suggested basal phylogenetic positions for LN strains, with questions remaining about transmission routes and epidemiological patterns (zoonoses versus pandemics). Broader ancient DNA work has documented ancient pathogen recovery from skeletal remains and outlined best practices for authentication and analysis. Research on dog domestication and population history provides context for increased canine presence in the LN, potentially affecting human–animal–pathogen interfaces. Together, the literature frames early Y. pestis as diverse, with limited virulence repertoire, and underscores the need to assess zoonotic reservoirs and cross-species transmission in pre-flea-adapted periods.

- Site and sampling: The Warburg necropolis (Germany; 5300–4900 cal BP), attributed to the Warburg Culture (WBC), comprises five gallery graves with remains of at least 198 individuals. Metagenomic screening was performed on skeletal elements from 133 individuals across the graves. Two Y. pestis-positive individuals were identified: Warburg_1 (5291–4971 cal BP; grave I) and Warburg_2 (5265–4857 cal BP; grave III). Radiocarbon modeling (OxCal v4.4.4) indicated an 88% likelihood that Warburg_1 predates Warburg_2 by ~200 years. - Ancient DNA processing and screening: Shotgun/metagenomic datasets were screened for Y. pestis DNA. Reads were mapped to Y. pestis references (chromosome and plasmids). Coverage and authenticity metrics were assessed; damage profiles were evaluated with standard aDNA tools. Data deposition: ENA PRJEB49618. - Genome recovery and coverage: Both positives yielded high coverage for Y. pestis, with mean coverage up to ~28× across plasmids, enabling downstream analyses. A third sample (C90; dog) had insufficient coverage for certain analyses. - SNP calling and filtering: VCFs for Warburg_1 and Warburg_2 were generated with stringent criteria: minimum 4× coverage per SNP, ≥90% allele support, quality ≥30, and exclusion of known homoplastic regions. For single-stranded libraries (downloaded LNBA data), C→T deamination-only SNPs were excluded. Yersinia pseudotuberculosis-specific SNPs were removed to simplify visualization. - Phylogenetic analysis: Maximum-likelihood phylogenies were built from a 10,743-SNP alignment including 226 modern and 62 ancient Y. pestis genomes, with Y. pseudotuberculosis as outgroup. Warburg genomes were placed relative to known LN and LNBA lineages. - SNP effect analysis: SNPs meeting the filters were annotated with SnpEff v4.3; results were screened for high-impact variants. C90 lacked sufficient coverage for SNP effect analysis. - Virulence gene analysis: Presence/absence and coverage of 15 chromosomal and 48 plasmid-associated virulence genes were assessed. Gene coverage was computed using bedtools v2.25.0 (intersect and coverage) and summarized for visualization (R v3.6.3; pheatmap). A custom awk script parsed coverage tables. - Molecular dating: A subset of 42 representative genomes (including Warburg_1 and Warburg_2) was analyzed in BEAST2 v2.6.4 with an uncorrelated lognormal relaxed clock, GTR+G4, and a coalescent constant population prior. Dates were provided as cal BP (1950=0). Two independent MCMC runs were combined with TreeAnnotator, ensuring ESS>200; Tracer used for convergence assessment. - Human population genetics: For the two Warburg individuals, reads were aligned to hg19; contamination was assessed and genetic sex inferred. Pseudo-haploid genotypes on ~1240k SNPs were merged with AADR references. PCA was conducted (smartpca) with ancient samples projected; ADMIXTURE (K=3–12) with bootstraps; qpAdm for two-way models. Kinship was assessed with READ using a dataset-derived normalization value. Both individuals showed typical WBC ancestry and were unrelated.

- Two new Late Neolithic Y. pestis genomes were recovered from Warburg (Germany): Warburg_1 (5291–4971 cal BP) and Warburg_2 (5265–4857 cal BP), likely separated by ~200 years with 88% probability that Warburg_1 is older. - The two genomes belong to distinct strains, indicating independent infection events within the same cemetery. - Across all currently known LN genomes (now n=4 including Warburg, Rinnukalns, and Gökhem), each represents a basal, separate lineage in the Y. pestis phylogeny, likely originating from different animal hosts. - Both Warburg genomes exhibited high coverage Y. pestis DNA, with mean plasmid coverages up to ~28× enabling phylogenetic and gene-content analyses; the dog sample (C90) had low coverage, precluding robust SNP effect and phylogenetic placement. - Virulence-factor screening is consistent with other LN strains that lack key flea-adaptive virulence traits seen in later pandemic lineages. - Population genetics show both infected individuals had typical Warburg Culture genetic ancestry and were not closely related. - Epidemiological context: Only 2 positives among 175 tested WBC individuals (133 from Warburg, 42 from Niderberchkan), and no clustering of additional pathogens, arguing against cemetery use for plague outbreak victims. - Zooarchaeological/ecological inference: Landscape opening during the LN likely increased small-rodent presence (potential reservoirs). Concurrent increases in dogs may have facilitated spillover; Y. pestis DNA was detected in an LN dog, supporting canine involvement.

The detection of two phylogenetically distinct Y. pestis strains in different Warburg graves, separated in time and without evidence for broader cemetery-wide infection, indicates sporadic, independent zoonotic events rather than a sustained epidemic or pandemic within this LN community. The basal and diverse positions of all known LN genomes imply multiple animal reservoirs and repeated spillovers into human settings before the evolution of efficient flea-borne transmission and high virulence. Archaeological and paleoecological evidence of landscape opening during the LN suggests increased contact with rodent reservoirs; simultaneously, rising dog populations could have amplified human exposure through hunting and scavenging behaviors. The finding of Y. pestis in an LN dog is consistent with this model. Collectively, the results address key questions about early Y. pestis ecology by showing frequent incursions into human settlements without clear signals of human-to-human transmission or mass mortality, refining our understanding of the pathogen’s pre-pandemic evolutionary phase.

This study contributes two new Late Neolithic Y. pestis genomes from Warburg and presents evidence of infection in a contemporaneous dog. Phylogenetic analyses place all known LN genomes as basal, distinct lineages, supporting a scenario of multiple animal reservoirs and repeated zoonotic spillovers rather than sustained human transmission. The lack of infection clustering and minimal positives across large screened cohorts argue against LN plague epidemics at these sites. Future research should target broader screening of contemporaneous humans and animals (especially rodents and dogs), improve recovery from low-coverage animal samples, and integrate paleoenvironmental data to identify specific reservoirs and transmission interfaces. Expanding temporal and geographic sampling will refine molecular dating and clarify the transition from early zoonoses to later flea-adapted pandemic lineages.

- Pathogen-negative findings do not confirm the absence of infection; aDNA degradation and taphonomic processes may erase microbial signatures, especially in bone compared to teeth. - The dataset combined different skeletal elements (bones and teeth), potentially affecting pathogen detection sensitivity. - The dog sample (C90) had insufficient Y. pestis coverage, precluding reliable SNP effect analysis and phylogenetic placement. - Limited number of LN-positive genomes restricts power to generalize about reservoir species, transmission dynamics, and geographic variability. - Potential biases from uneven preservation and sampling across graves and sites.