Y. pestis DNA Preservation in Skeletal Remains.

Complete information regarding skeletal screening is available in SI Materials and Methods. In total, we screened DNA extracts from 109 samples—53 bones and 46 teeth from the East Smithfield (ES) collection as well as 10 samples from St. Nicholas Shambles (SNS) that served as negative controls (Fig. 1 and Table S1). Total DNA content measured fluorometrically revealed less DNA in teeth than in bone, with dental extracts yielding on average 1,065 and 664 pg/μL for supernatant and pellet, respectively, compared with 1,956 and 2,280 pg/μL for bone. Quantitative PCR (qPCR) results showed the presence of amplifiable Y. pestis pla DNA in 5.7% of bones (3/53) and 37% of teeth (17/46) (Figs. S1 and S2). The caf1M assay yielded expected products in 5 of 17 teeth that contained amplifiable pla, although this finding is expected based on Y. pestis plasmid numbers where the pPCP1 plasmid outnumbers the pMT1 by an estimated 100 to 1 (19). No Y. pestis DNA was detected in any of the negative control samples. DNA sequences from the qPCR products contained both G to A and C to T transitions, damage patterns typical of ancient DNA (17). Despite their lower whole DNA content, dental samples were consistently a richer source for Y. pestis DNA than bone, and higher copy numbers were frequently observed in the EDTA supernatants as opposed to the pellets (Table S1). This finding is consistent with what one might expect from a blood-borne pathogen, where DNA likely resides in the desiccated blood vessels of the pulp chamber. All amplifiable DNA was extremely low in quantity, with a maximum estimated copy number of 30 copies/μL.

Multiplex PCR Data for the pla Gene.

Using a multiplex PCR approach, we were successful in sequencing 78.3% of the pla gene and its flanking intergenic spacers at a minimum of four times coverage (Dataset S1). Missing regions were caused by a lack of expected PCR product. Cloned sequence data revealed 63 sites showing DNA damage reflected in the predominance of C to T (53.97%) and G to A (30.16%) transitions, which accounted for 84.13% of all aberrant positions. No transversions were observed. To distinguish damage from potential polymorphisms, all sequences were confirmed by cloned data from two independent amplified libraries. Wherever products from these two libraries did not match, cloned sequence data from a third amplified library was sought. This process permitted us to resolve all transitions with the exception of three, where single nucleotide differences from published sequences were present in a subset of clones from two independently amplified libraries (Fig. S3). The PCR data did not contain the Microtus-specific C to T (20) at position 1,109 in Dataset S1, and it does not show the T to C transition previously reported in an ancient sequence for this region in the work by Raoult et al. (7) (Dataset S1, position 809).

Y. pestis Chromosomal PCR Assays.

Of 11 primer sets used to define branches 1 and 2 phylogenetic placement (9), only 1 primer set yielded expected amplification products, namely s19: it produced the shortest amplicon of all chromosomal primer sets (80 bp), and it corresponds to a region in the DNA helicase II gene. Sequence data showed the absence of the CTA motif common to extant branch 2 sequences, thus supporting the notion that medieval plague was not caused by an extant branch 2 variant (9); however, this region also revealed the presence of two synonymous point mutations that, to our knowledge, are not found in any Y. pestis sequences either modern or ancient (Fig. 2).

Enrichment Efficiency.

To evaluate the suitability of targeted enrichment for investigations of virulence in ancient pathogens, we attempted to capture one of the two pestis-specific virulence plasmids, namely the 9.6-kb pPCP1 along with complete human mitochondrial genomes from both Black Death and control teeth to evaluate endogenous DNA preservation. The total number of merged Illumina reads per sample that mapped to the target DNA (human mtDNA and pPCP1 DNA) varied between 0.3% and 49%, suggesting high enrichment efficiencies.

Human mtDNA Enrichment and Preservation.

Sequence clustering revealed between 228 and 84,244 unique fragments for the individual human teeth mapping to the reference mtDNA (Table 1), thus revealing high quantities of mtDNA fragments in all 10 human extracts with low quantities in two of the SNS human control extracts. The cave bear control sample yielded 15 mtDNA fragments, 6 of which aligned to the human mtDNA and 9 were identical or highly similar to the cave bear mtDNA genome (21). No human mtDNA fragments were found in the extraction blank. Complete or near complete mitochondrial genomes were constructed from fragments of 58 bp average length for all 10 human samples, with an average coverage of 0.7–280× (Table 1). All 10 mtDNA genomes were different from each other, and two samples (ES 3 and 4) extracted two times yielded the identical mtDNA genome sequence between duplicates. These data suggest that there was no cross-contamination between human samples. Nine of the ten were found to correspond to typical European mtDNA haplogroups based on phylotree (22) (Table 2). Only for SNS1 was the haplogroup not determinable due to the higher contamination level (Table 2). Comparisons of ancient mitochondrial consensus sequences against a worldwide dataset of 311 mtDNAs and the Cambridge reference sequence revealed between 1 and 17 positions that were either unique to the ancient samples or present at less than 1% frequency in the current modern DNA database. Such positions were used to calculate the extent of polluting human exogenous DNA contamination within our fragments (Table 2 and Fig. S4). To calculate the ratio of endogenous to exogenous human DNA, we selected all fragments that showed the rare variant(s) and considered them to be endogenous, whereas all fragments that showed a common human mtDNA variant were considered to be potential contaminants. In so doing, we found between 21 and 6,724 individual mtDNA fragments overlapping such contamination informative positions in the ES samples. The majority of these DNA fragments were found to be consistent (all unique fragments overlapping the position showed the same substitution), suggesting that the majority of mtDNA fragments from each sample is derived from a single biological source (Fig. S4) (15, 23). Individual mtDNA fragments were compared against the corresponding consensus sequence, and nucleotide substitutions were recorded for each position along the DNA fragment. Based on these estimates, levels of polluting human DNA could be calculated, although these rates were very low for all samples (Table 2). In support of this notion, fragments identified as endogenous DNA contained degradation patterns typical for ancient templates (Fig. S5), again suggesting preservation of endogenous human DNA in all 10 medieval teeth.

Y. pestis DNA Enrichment and Preservation.

We found between 42 and 36,986 unique fragments mapping to a portion of the Y. pestis pPCP1 plasmid reference genome in all libraries, including the pre-Black Death human control, the cave bear sample, and the extraction blank (Table 1). All reads in the non-ES extracts mapped to a region between positions 3,000 and 4,200 of the reference pPCP1 that shows a high similarity to expression vectors that are used for recombinant enzyme production. This region was found to be problematic in a previous Y. pestis pPCP1 study using shotgun sequencing (24); hence, it is likely that remnants of the expression vectors are being captured by our enrichment approach. Apart from these conserved motifs found across all samples and controls, the pre-Black Death human control samples (SNS) showed not a single unique fragment homologous to the pPCP1 sequence, whereas the five ES samples contained between 7 and 6,435 unique fragments that mapped to the Y. pestis pPCP1 plasmid (Table 1). Because all five individuals from the ES cemetery likely died from the same strain of Y. pestis, unique pPCP1 fragments were pooled to reconstruct a consensus sequence at maximum coverage. The pooled consensus has an average coverage of 43 bp, ranging from 0× to 702×, and an average fragment length of 36 bp (Table 1). This finding may be why a previous study using the ES collection that targeted regions in excess of 130 bp (8) was unsuccessful in replicating some original work with plague samples (7, 25). Not surprisingly, the three regions with the poorest coverage had a low GC content (%) (Fig. 3 and Fig. S6). Shotgun sequence data for ancient human samples (23) show similar patterns, suggesting a preservation bias to high GC content and thus likely ruling out enrichment artifacts. Despite lower coverage of AT-rich regions, the consensus sequence covers ∼99% of the entire pPCP1 at a minimum of 2× (Table 3).

We tested the authenticity of the pPCP1 DNA by analyzing the substitution patterns along the DNA fragments and found the same nucleotide misincorporation pattern that was observed within the mtDNA fragments from the five medieval samples of the ES site. The predominant C to T substitutions found at the 5′ ends and G to A substitutions at the 3′ ends are typical of ancient DNA (15, 26, 27). This pattern is not seen in the fragments mapping to the vector region of the pPCP1 plasmid, lending additional support to the notion that the majority of these fragments are indeed derived from contaminating expression vectors of the supplied reagents (Fig. 4 and Fig. S7). Fragment length distribution also indicates longer fragments obtained for this region, which is consistent with their coming from a modern contaminant source (Fig. 5).

Phylogenetic comparison of the 8,299-bp 2× consensus pPCP1 plasmid with modern Y. pestis reveals sequence identity with 11 of 14 extant Y. pestis strains; it did not match sequences typed as Microtus (AE017046.1), Angola (CP000900.1), and Medievalis (CP001611.1), although it does match the KIM pPCP1 (AF053945.1). It also resolves the three ambiguous positions identified by the multiplex approach of the pla gene and replicates the pla sequence described above. Considering indel polymorphisms, the ES strain does not contain the TT insertion common to the Orientalis biovar (AL109969.1 and CA88-4125).

Experiments Conducted at McMaster University.

Information on the archaeological site, skeletal sampling, and DNA extraction can be found in SI Materials and Methods. We note that the site from which the skeletons were recovered, the ES burial ground in London, is one of just a few excavated sites in the world that can be linked definitively and uniquely to the mid-14th century outbreak of the Black Death by archaeological and documentary evidence (34).

Experiments Conducted at the Max Planck Institute in Leipzig, Germany.