Ancient Yersinia pestis genomes provide no evidence for the origins or spread of the Justinianic Plague

Along with the publication of 137 ancient human genomes retrieved from archaeological remains of the Eurasian steppe, Damgaard et al., 2018 identified two individuals infected with Yersinia pestis, yielding one genome with 0.24x average coverage (DA147, 6th–9th c. AD) and another with 8.7x (DA101, 2nd–3rd c. AD). A phylogenetic analysis performed on the latter placed it in a position ancestral to a 6th-century Justinianic genome from Aschheim, Germany. These results are used to fuel an argument that the Justinianic Plague (541–544 AD) “was brought to Europe towards the end of the Hunnic period through the Silk Road along the southern fringes of the steppes” in contrast to the leading hypothesis of introduction via the Red Sea that is supported by historical accounts. In our reanalysis, we question the contested historical context of the presented genomes with the Justinianic Plague and show that the lower coverage genome might be rather related to the Black Death (1346–1353 AD).


29
The recent sequencing of dozens of pathogen genomes reconstructed from ancient DNA 30 enabled increased-resolution phylogeographic studies on the spread of infectious diseases in 31 prehistoric and historic times, especially in the context of human migration, mobility and trade 32 (Andrades Valtueña et al., 2017;Bos et al., 2016;Keller et al., 2019;Namouchi et al., 2018; 33 Rascovan et al., 2019;Rasmussen et al., 2015;Spyrou et al., 2019Spyrou et al., , 2016Vågene et al., 2018). 34 This is especially true for plague with its long and richly documented history and the abundance 35 of published ancient genomes of its causative agent, Yersinia pestis. Interpretation of 36 phylogenetic data in the context of human history requires careful assessment of tree 37 topologies, branch lengths and mutation rates as well as thoughtful consilient approaches in 38 integrating historical and archaeological data to prevent overly simplistic, deterministic or even 39 erroneous interpretations. 40 For the Second Pandemic, the geographic origin and the possible persistence within Europe 41 after the Black Death (1346-1352 AD) are the subject of ongoing scientific and scholarly 42 discussion (Bos et al., 2016;Namouchi et al., 2018;Schmid et al., 2015;Spyrou et al., 2016;43 Spyrou et al., 2019). Whereas the first comprehensive phylogenetic study on Y. pestis favoured 44 an East Asian origin (Cui et al., 2013), other scenarios assume an origin in Central Asia or the 45 Caucasus (Benedictow, 2004;Namouchi et al., 2018;Sussman, 2011). Similarly, the origin of 46 the Justinianic Plague (541-544 AD) has long been hypothesized to have originated in Africa 47 (Achtman et al., 1999;Cui et al., 2008;Sarris, 2002). More recent studies however agree that 48 the strains causing the First Pandemic (541-750 AD) likely emerged in Central Asia 49 (Eroshenko et al., 2017;Harper, 2017;Wagner et al., 2014). The fact that the first outbreak of SNPs are unique to DA101 compared to 95 in Aschheim, which is provisionally consistent 64 with Aschheim's younger age (reported in the SI and in Extended Data Fig. 9, though the latter 65 does not present the tree at full resolution). The identified shared ancestry was interpreted as 66 setting DA101 within the context of the "Justinian plague" (sic). The longer branch in the 67 Aschheim genome is explained by its younger age and a seemingly accelerated substitution 68 rate, which is supposedly indicative of an epidemic context.

69
Although the Justinianic Plague was previously thought to represent the first major onslaught  Spyrou et al., 2018). Here, we present a reanalysis of both DA101 and DA147 73 genomes which does not seem to support the arguments made by Damgaard et al., 2018. 74 Instead, the analysis of DA101 suggests it to be yet another example of a pre-Justinianic human 75 infection. Furthermore, in contrast to its suggested archaeological dating, we show DA147 to 76 occupy a phylogenetic position much closer to the Black Death (1346-1353 AD) than the 77 Justinianic Plague (Fig. 1C). As such, neither genome can address the origin of the First 78 Pandemic.

81
We reanalysed both presented genomes with a more extensive dataset of published modern and 82 ancient Y. pestis genomes (Fig. 1A, Table S1). We opted to include the genome from  Analysis of the DA101 genome revealed a minimum of 3 SNPs shared with Altenerding and a 90 minimum of 9 that are unique. By contrast, Altenerding has 51 unique SNPs (Table S2). This

94
Further to this, we attempted a molecular dating analysis, though the age of individual DA101 95 proved difficult to determine given discrepancies in the text and SI, ranging from  Table 2). Ultimately, we opted to use the calibrated radiocarbon interval, which 98 yielded a mean age of 154 BC (95% HPD: 527 BC to 153 AD) for the emergence of the shared 99 lineage and 9 BC (95% HPD: 318 BC to 221 AD) for their divergence time (Table S4). For 100 comparison, dating results without the recently published RT5 genome (Spyrou et al., 2018) 101 are shown in Table S4. This strongly supports a pre-Justinianic provenience for the DA101 102 genome. A number of shared or unique SNPs might be undetected for DA101 due to low 103 coverage, hence the estimated divergence dates are conservative and might be even older.

104
Regarding the substitution rate, we do not observe a notable acceleration on the Altenerding 105 branch (mean 2.67E-08) compared to the overall mean (1.48E-08) across the tested dataset 106 ( Fig. S2), particularly since both estimates show overlapping 95 % HPD intervals (Table S5).  (Namouchi et al., 2018;Spyrou et al., 2019Spyrou et al., , 2016 clearly indicate that this pathogen is able to 123 travel vast geographic expanses quickly, without accumulating genetic diversity in the process.  (Table S1), while omitting all private calls in DA147 144 since their vast majority will represent DNA damage and sequencing errors due to the 145 genome's low coverage. The remaining SNPs forming the branch of DA147 in Fig. S1 (red) 146 are an artefact caused by homoplastic or triallelic sites. We computed a maximum likelihood 147 phylogenetic tree that, unexpectedly, placed DA147 closest to the previously described 148 polytomy of Branches 1-4 (Fig. S1). The genomes's placement was further investigated by 149 visual inspection of all diagnostic SNPs separating Branches 1, 2, 3&4 and Branch 0 (see Table   150 S3). Our analysis reveals several potential placements for DA147: (1) it is one SNP ancestral 151 to the polytomy but derived with respect to the 0.ANT3 node, (2) it is directly on the polytomy,    Bolgar, (Spyrou et al., 2016)). A complete list of all Y. pestis genomes used is given in Table   214 S1. Previously identified problematic regions as well as regions annotated as repeat regions,