Archaeological evidence of mass burials at an ancient cosmopolitan city
Jerash was an important city during the Greco-Roman and Byzantine periods and is considered one of the largest and most well-preserved sites of Roman architecture in the world outside Italy 9. From the fourth through seventh centuries, Jerash was a thriving city in a prosperous part of the eastern Roman Empire, with a flourishing population, rich agricultural hinterland and extensive trade networks2,14–21. Among the many accoutrements of Roman rule was the installation of a hippodrome (chariot racetrack or Roman Circus) outside the city center. From the mid Sixth Century CE, major civic structures were repurposed to light industry, and non-funerary civic contexts suddenly employed to contain mass interments, with deposition dating somewhere within the mid-Sixth through mid-Seventh Centuries CE. In the sixth and seventh centuries CE, the abandoned hippodrome became a burial place and two mass graves dating to the first pandemic were excavated there from 1991-96 10 11–13. Our samples were obtained from these mass interments which held approximately 150 adults and 80 sub-adults (neonates, infants, and children) (Fig. 1B-D).
The excavation site was linked by land transport routes leading from the coast into the Jordanian uplands14. Trade flowed down the route, but in 541 CE so did the bubonic plague15. Procopius and John of Ephesus report the devastating effects of the plague, with modern historical studies estimating that somewhere between a third and half the population fell victim to this first recorded pandemic to sweep the region16–18 19,20
Around 550–650 CE, the mass burials was deposited in the western W2 &W3 storage chambers of the hippodrome, and this event was followed by a severe earthquake (possibly 659 CE) which collapsed the cavea (large stone block seats) on top of the skeletons (Fig. 2A). The archaeological dating being a possible 100 year period based on the chronological indicators of a recovered coin (dated to Constantine II or Constatine IV), and the recorded earthquake. These highly fragmented human skeletons were recovered from two disused chambers in the Hippodrome and the evidence suggests these burials were undertaken hastily 13. The human remains were sealed by earthquake debris from the collapsed chamber vault. The chambers were excavated in horizontal levels equivalent to the height of a course of blocks. Within those levels discrete “lots” were identified and collected 21. Our samples were collected from 2 chambers and several lots and the victims that can be identified spanned a large demographic section including adults and sub-adults 11,12 .
Stable isotope results indicate diverse childhood origins of victims
Stable isotope analysis was conducted on 11 individuals from the Jerash site, following well-established preparation methods22. Collagen was extracted from tooth roots and apatite from tooth enamel for 10 individuals, and from bone for 1 individual. All but one of the tooth root samples were M2s, representing diet during formation (ca. 8–15 years). Thus, our stable isotope data uncover the dietary patterns in childhood (enamel) to early adolescence life (collagen) stages of the victims. Our isotope values obtained for Jerash are in a similar range to those from the sites of Petra (1st century. BCE − 1st century CE)23 and Khirbet Qazone (1st-3rd century CE) 24 to the south.
In regards to the collagen, all samples had reliable yields and ratios of C and N. Except for individual W2-115-LM2 (green dot in Fig. 2B), the δ 13C range was narrow and averaged − 19.0‰, suggesting dietary protein based on C3 plants (e.g. wheat, barley and oats) and animals consumptions. That individual, with δ 13C value of -17.4, was consuming perhaps 10% of its protein diet from C4 plants such as millet. The δ 15N values for this individual was the second lowest of all, at 9.7‰, with the range for the 11 individuals from 9.1 to 13.2‰. At least three individuals (yellow dots in Fig. 2B), with higher δ 15N values, were either consuming freshwater fish or plants/animals with enriched values due to manure sodder (i.e. agricultural fertilizers of manure). Note that the determined N value ranges indicate that no apparent malnutritional status was found in any of the tested victim samples (Fig. 2B).
The tooth enamel δ 13C values, which represent the diet at the time of the tooth enamel formation (3–8 years), range from − 13.2 to -11.5‰, suggesting that C4 plants were consumed in significant quantities early in life for several individuals, including individual W2-115-LM2 (Fig. 2C).
A surprisingly wide range of tooth enamel δ 18O were found and varies from − 6.4 to -0.5‰, suggesting a variety of origins for these individuals (Fig. 2C This range is similar to that at Petra12,25. Two of the three individuals with the highest δ 15N values (Fig. 2B) also had the most negative δ 18O values (indicted by * in Fig. 1C), suggesting they may have grown up near a large freshwater system such as the Sea of Galilee or the River Jordan.
Proteomic results showed potential presence of the plague pathogen
Ancient proteomic analysis was conducted using a combination of mass spectrometry-based techniques. Protein extraction from well-preserved samples (n = 10) was performed under controlled laboratory conditions to minimize contamination. Extracted proteins were then subjected to liquid chromatography-mass spectrometry (LC-MS/MS) for identification and quantification. Spectral data were processed, with a focus on identifying pathogenic proteins. (Detailed protocols in supplemental materials and methods). We obtained 25–358 unique human peptides from the samples, and 4 samples having at least 3 Y. pestis peptides present based on MaxQuant v2.4 (Supplemental Table S1) (Fig. 2D). Our proteomic data confirmed the Jerash biomaterial is well preserved, and suitable for further genetic and genomic analysis, and provide possible evidence of the presence of Y. pestis in the burial sites.
Whole genome sequences captured the pandemic pathogen genome
A set of teeth from 8 individuals (represented by 8 teeth) were carefully selected for aDNA analysis based on criteria such as preservation status, bone density, and visual inspection for potential signs of contamination. Prior to extraction, external surfaces of bones were cleaned with a mild bleach solution and physically abraded to remove potential contaminants. aDNA extraction was performed in a dedicated clean room environment following established protocols optimized for ancient samples 26,27. Briefly, approximately 0.5 gram of bone powder was obtained from each sample using a sterilized dental drill. The powder was then decalcified using a solution of 0.5 M ethylenediaminetetraacetic acid (EDTA) at pH 8.0. To minimize the risk of contamination, stringent measures were implemented, including the use of negative controls, laminar flow hoods, and sterile equipment for each step of the extraction process (Detailed protocols in supplemental materials and methods).
To verify the presence of aDNA and assess its amplifiability, a series of targeted PCR tests were conducted using primer sets specific to conserved regions of Y. pestis genome (Fig. 2D) (Supplemental materials and methods). Microbial aDNA was successfully obtained in Cementum extraction C1. No signal was detected in either negative extraction control (-E) or negative PCR control (-P). Dentine (D) and Cementum (C) were used to generate a 1,465bp fragment of the 16S gene.
To obtain and authenticate aDNA, library preparation for whole genome sequencing was carried out using a modified protocol for aDNA, incorporating steps to mitigate potential DNA damage and fragmentation inherent to ancient samples. Briefly, end-repair, adapter ligation, and DNA indexing were performed using a combination of enzymatic and chemical processes. Unique barcoded adapters were used to allow for multiplexing of samples in a single sequencing run. Library amplification was conducted with a limited number of cycles to prevent over-amplification of damaged DNA molecules. Whole genome sequencing was carried out on an Illumina sequencing platform using paired-end sequencing reads. We used three independent aDNA authentication methods to verify the DNA was from the Jerash material, 1) obtaining short fragments (~ 100bp) due to ancient DNA degradation, 2) sequencing untreated (not gap-filled) DNA yielded gapped results, 3) computational validation of whole genome comparative analysis showing ancient lineages. We have obtained high coverage (100–400 million total reads, coverage ranges from 1.25x to 35x of Y. pestis genome) from 8 samples, and the Y. pestis genomes were constructed. The 8 Jerash isolates showed an average Nucleotide Identity (ANI) of 0.97 with whole genome variants comparisons (Supplemental Table S2) (Fig. 3A). Thus the 8 pathogen isolates genomes represent the same isolate with highly similar genomes, despite that the human host isotope signatures differ from each other. Our data shows that the demographically diverse victims are bearing the same Y. pestis isolates in this mass burial site.
For virulence factor analysis, we first removed low quality reads with BBMap and used BWA-MEM algorithm to generate aligned BAM files. We identified damage pattern in ancient DNA based on BAM quality scores rescaled with mapDamage v2.2 28. The Black Death strains possess a full set of virulence factors (Fig. 3B). In contrast, the earlier prehistorical strains29,30, except RT5/RT6 as previously reported, lack a set of virulence factors including the flea transmission related gene Ymt. For the 1st pandemic, our results show that the Jerash mass burial strain possesses a full set of virulence factor genes31, e.g. Ymt, Pla, F1 capsule, responsible for the potential transmission and pathology of Y. pestis (Fig. 3B).
Subsequently, we performed Maximum Likelihood (ML) phylogenetic analysis32 with the whole genome genetic variants (Fig. 3C). Genome level completeness and contamination were checked using checkM v1.0. Using the criteria of greater than 95% completion and less than 5% contamination, the genomes were further screened for phylogenomic analysis (Supplemental Table S3). The best fit model was chosen according to Bayesian Information Criterion (BIC)33. Tree topology bootstrapping was performed for 1000 replicates. We show that the Jerash strain belongs to the same highly supported cluster as strains recovered elsewhere from the same time frame as the first global pandemic (541–767 AD). Importantly, however, the Jerash strains are the first genomes recovered within the eastern Mediterranean region where the historical epicenter of the first global pandemic was located. A set of prehistorical strains in Eurasia showed phylogenetic affinity with the 1st pandemic strains, supporting the interpretation that the Jerash strain may represent one of the initial large scale transmissions within dense urban human populations in the 1st pandemic.