Ancient Yersinia pestis genomes from across Western Europe reveal early diversification during the First Pandemic (541–750)

Significance The first historically reported pandemic attributed to Yersinia pestis started with the Justinianic Plague (541–544) and continued for around 200 y as the so-called First Pandemic. To date, only one Y. pestis strain from this pandemic has been reconstructed using ancient DNA. In this study, we present eight genomes from Britain, France, Germany, and Spain, demonstrating the geographic range of plague during the First Pandemic and showing microdiversity in the Early Medieval Period. Moreover, we detect similar genome decay during the First and Second Pandemics (14th to 18th century) that includes the same two virulence factors, thus providing an example of potential convergent evolution of Y. pestis during large-scale epidemics.


Assessment of SNPEvaluation
To test the validity of the tool SNPEvaluation and the applied criteria for detection of false positive SNPs, we generated a set of artificial datasets representing clean and contaminated Y. pestis genomes ranging from ~3x to ~30x coverage. To mimic the contamination with environmental bacteria (background), we used the sample DIR002.A, since enrichment with a Y. pestis probeset cannot be simulated reliably in silico. The sample DIR002.A was captured and sequenced after initial qPCR screening, but was classified as negative for Y. pestis after sequencing and mapping. Therefore, the 8470 reads mapping with strict parameters are derived from environmental sources. For the target Y. pestis reads (foreground), we generated artificial sequencing data based on the reference genome CO92 aiming for 3-fold, 5-fold, 10-fold and 30-fold coverage in triplicates using the tool gargammel (1). We simulated paired-end 75 bp sequencing reads using the coverage flag (-c) and the length distribution of the paired-end 75 bp sequenced sample DIT003.B. The achieved coverages are ~6 % lower (on average 2. 83fold, 4.72-fold, 9.43-fold, 28.20-fold, see SI Appendix, Table S5) due to mapping quality filtering. The artificial reads were combined with the background and mapped with stringent (-n 0.1) and lenient (-n 0.01) mapping parameters. Since the capture probe set was constructed based on CO92 among others, a possible capture bias was neglected for the artificial CO92 reads. To test for the sensitivity of our method, we applied it to all false positive SNPs that were introduced by our background sample (SI Appendix, Fig. S2, Table S6). As expected, no SNPs were called in any of the simulated Y. pestis datasets without background. Using only the background (DIR002), a total number of 451 SNPs were called. One additional SNP was only called with spiked in Y. pestis reads (called as N in DIR002), summing up to a total of 452 possible false positive SNPs. With increasing coverage of spiked in Y. pestis reads, the number of false positive SNPs decreases from on average 139.67 for 3-fold, 40.33 for 5-fold, 5 for 10fold to 0 for 30-fold, showing that the environmental contamination can introduce a significant number of false positive SNPs for low coverage samples but is negligible for high coverage samples. The remaining positions of putative false positive SNPs are called as heterozygous (N) or as reference call in high coverage samples (in average 0.67 for 10-fold, 63.67 for 30fold). Using the same criteria as for the First Pandemic samples in this study, SNPEvaluation was able to filter 100 % of false positive SNPs from every artificial dataset. Moreover, all chosen filtering criteria were necessary for the determination as false positive SNPs: Regarding the ratio of mean coverage with low/high stringent mapping, this criterion was able to filter out almost all SNPs. Only 2 of 451 SNPs from the pure background, 1 of 147 SNPs from a 3x-fold and 1 of 39 SNPs from a 5-fold coverage contaminated Y. pestis samples passed the criterion (ratio 1.00), but were filtered out either due to a heterozygous SNP or uncovered positions surrounding the SNP. The lowest ratio for the false positive SNPs without other criteria applying ranges between 1.05 and 1.21, justifying the very strict criterion of a ratio of 1.00 to be accepted as a true positive SNP.
To determine the specificity of our method, we generated a SNP table based only on modern Y. pestis strains and extracted the SNPs that are shared by at least 34 genomes (~15 %). This cutoff was chosen due to the local maximum in the distribution of the numbers of genomes sharing SNPs. This resulted in a list of 418 SNPs, which constitute the backbone phylogeny of Y. pestis and are a good proxy for true positive SNP positions (see SI Appendix, Fig. S3, Table  S7). However, all positions appear as reference calls in our datasets, since the artificial reads are based on the same genome (CO92) that was arbitrarily chosen as reference genome. In the simulated CO92 datasets without background, in average 79.33 positions were called as reference for the simulated 3-fold coverage samples, 230.33 for 5-fold, 403.33 for 10-fold and all 418 for 30-fold. All remaining positions were called as N due to low coverage. The ratio of mean coverage with low/high stringent mapping never exceeded 1.00 in any datasets, showing that sequencing errors simulated by gargammel are not interfering with this parameter. A maximum of 2.63 % of the reference calls were filtered out due to uncovered bases surrounding the position in the simulated 3-fold or 5-fold coverage samples, but never for 10-fold or 30fold coverage. Using solely the background sample, only 2 positions were called as reference but filtered out due to a >1.00 ratio, giving evidence that the DIR002 sample is indeed negative for Y. pestis. For the simulated CO92 datasets contaminated with the background, in average 74.33 positions were called for the 3-fold samples, 227.67 for 5-fold, 401.33 for the 10-fold and all 418 for 30fold. This means that 0 -7.89 % of the positions are not called due to heterozygosity introduced by the background. After filtering with SNPEvaluation, in average 69.67 positions remained for the 3-fold samples, 216.67 for 5-fold, 385.00 for 10-fold and 401.67 for 30-fold, so 3.59 -8.57 % of called positions are filtered out. In summary, SNPEvaluation offers a maximum sensitivity (100 % of false positive SNPs filtered out), while retaining a high specificity (8.57 % or fewer of true shared positions filtered out). Although we used only one background sample in this analysis and were not able to model, i.e., the capture bias, the presented method is a powerful tool to filter low coverage datasets for false positive SNPs, a crucial step to build reliable bacterial phylogenies.

SNP Evaluation of the Aschheim Genome and SNP Effect Analysis
The Aschheim genome (2) was evaluated separately, given its peculiarly high number of potential false positive SNPs described previously (3). Our systematic evaluation verified previous classifications: all SNPs potentially unique to Aschheim that passed the criteria show a coverage lower than 5-fold, which was the threshold of their SNP calling. However, the high number of presumably shared SNPs that did not pass our stricter criteria underlines again the high 'heterozygosity' of the genome (see SI Appendix, Table S14) that might be explained not only by contamination by soil bacteria or sequencing errors but presumably also by PCR and capture artefacts, as previously discussed (3). Therefore, the Aschheim genome was excluded from subsequent analyses. be avirulent in humans. Therefore, the 12.9 kb deletion observed here in the First Pandemic genomes is presumably a similar case of genome decay. Another deletion of 14.8 kb was found on the pMT1 plasmid (positions 23,133 to 37,975 in CO92) of the genomes of the French clade (LVC_merged, LSD001.A, LSD023.A) and both genomes of Unterthürheim (UNT003.A, UNT004.A). All 21 genes affected by this deletion are hypothetical proteins including a putative ABC transporter ATP-binding protein, therefore the functional consequences of the deletion cannot be determined. The fact that this deletion is found in the French clade as well as in the Unterthürheim genomes but not in the Altenerding (AE1175) and Dittenheim genomes (DIT003.B) that are otherwise identical, suggests that this deletion is more common and mediated by a process similar to the transposable elements on the chromosome.

Phylogeographic Analyses
The new genomes and radiocarbon dates combined suggest an association of the British genome as well as the polytomy giving rise to the four lineages with the early phase of the First Pandemic or even the Justinianic Plague itself (541)(542)(543)(544). The accumulation of one (EDI001) or two (Altenerding cluster) SNPs from the basal node of all genomes could have happened on the way from Egypt to western Europe. The fact that the pandemic reportedly spread from Pelusium along the Mediterranean coastline in two independent waves, one heading west to Alexandria and the other east to Palestine, could explain the early branching event (11). Strikingly, such a diversification during the onset of a pandemic has not been found yet for the Black Death (1348)(1349)(1350)(1351)(1352), where the two genomes from London East Smithfield (12) and Barcelona were found to be identical (13). Besides differing mutation rates, this might be due to differences in propagation speed between the 6 th and 14 th century, related to changes in human mobility by land and sea: a significantly slower or less direct transmission over large distances would allow the pathogen to acquire more substitutions. The lineages found in Bavaria could have spread there by a 'western route' from Gaul, by a 'southern route' from Italy or by an 'eastern route' from Illyricum, which were all affected by plague in or around 543. The presence of plague in the British Isles even suggests a fourth 'northern route' upstream along the Rhine river. The 'western' and 'southern route' would have necessitated overland transport via the Roman road network that connected all of the relevant sites with the Mediterranean coastlines and was still functional in the 6 th century (SI Appendix, Fig. S9). The 'southern route' would have required crossing the Alps via different passes that had been used since Antiquity (14). Navigation along the Danube could have facilitated the 'eastern route'. The importance of rivers for the spread of plague has already been shown exemplarily for the Rhône during the First Pandemic (15) and for the Black Death (16). However, attempts to prove the preferential spread via rivers during Second Pandemic have recently been criticized (17,18). The site of Petting is geographically situated only 100 km southeast of Aschheim and Altenerding (SI Appendix, Fig. S9). However, it was located in the Roman province Noricum ripense whereas the sites with the distinct uniform lineage were situated in Raetia secunda (Aschheim, Altenerding, Unterthürheim) or close by (Dittenheim). Although the administrative system of the Western Roman Empire had broken down by the mid-6 th century, its political borders continued to be influential, not least because of the ecclesiastical system of dioceses that followed them. It is possible that these ancient boundaries influenced the spread of the two epidemic outbreaks in modern-day Bavaria. Since the river Inn separated Raetia secunda and Noricum ripense, this might suggest that rivers could serve as physical barriers to the spread of plague where river transport was negligible. This in turn would rather suggest the 'eastern route' or the 'southern route' for Petting as described above. Complementing the previous results from Aschheim and Altenerding, our new data from Unterthürheim and Dittenheim underline the epidemic extent of this plague outbreak in early medieval Bavaria, totalling 16 individuals with genomic evidence for Y. pestis and an additional five PCR-positive individuals in Aschheim (19). Far from the urban centres of the time and any recorded outbreak of plague, the new molecular evidence stands in strong contrast to Durliat's claim that the Justinianic Plague was merely an urban phenomenon. Instead, we view this data as being in line with ancient statements by Procopius, John of Ephesos and Paul the Deacon who reported that the countryside of the Levant and Italy were severely impacted.

Radiocarbon Dating
At least one individual per burial was sampled for radiocarbon dating for all burials that tested positive for Y. pestis, assuming simultaneity of interment for the multiple burials. Samples were dated at the CEZ Archaeometry gGmbH, Mannheim, Germany. For Saint-Doulchard, published radiocarbon dates of adjacent burials in the same trench are reported (20). These as well as the raw radiocarbon dates of Aschheim and Altenerding (2,3) were recalibrated for consistency. The raw radiocarbon dates were calibrated with IntCal13 (21) in OxCal v4. 3.2 (22). All raw and calibrated dates are given in SI Appendix, Table S12; Fig. S8 shows the respective probability distributions. Some of the intervals completely pre-date the onset of Justinianic Plague (541) which could be explained by a marine or freshwater reservoir effect (23,24) or human bone collagen offset (25). In the absence of C/N isotope data and a wellestablished method for addressing the human bone collagen offset, we report calibrated dates without any correction. A combination of the probability distributions using the function "Combine" in Oxcal was attempted for the trench inhumations of Lunel-Viel (LVC001, LVC002, LVC003, LVC005, LVC006, LVC007), the double burial 1175/1176 of Altenerding (AE1175, AE1176) and the quadruple burial 131-134 of Unterthürheim (UNT004, UNT005, UNT008). For Lunel-Viel, the resulting 2-sigma interval is 562-604 calAD (Acomb 148.6 %; A 109-127.9 %), for Altenerding 428-552 calAD (Acomb 107.1 %; A 104-105.9 %) and for Unterthürheim 428-544 calAD (Acomb 29.6 %; 14.5-100.9 %). The low agreement index for Unterthürheim is caused by the comparatively recent dating sample of UNT005 with an individual agreement index of 14.5 %. Therefore, the combination of uncorrected dates is deemed problematic in general and was dismissed for interpretation.

Archaeological Context Information
The following site descriptions present condensed information on all sites examined in this study. The classification of multiple burial types follows McCormick (2015). Sex and age determination is based only on morphological examination. For the age classification, the German system (27) is used as follows: Infans I (0-6 years old), infans II (7-12 years old), juvenile (13-20 years old), adult (20-40 years old), mature (40-60 years old), senile (more than 60 years old). Sites that were tested positive for plague are set in bold.
Alladorf (ALL; Markt Turnau, Landkreis Kulmbach, Germany): The Carolingian cemetery of Alladorf, dating roughly between 630 to 720, revealed 163 graves with remains of 276 individuals. However, the total size of the cemetery is unknown, since the excavation did not reach the borders of the burial area. B. Leinthaler classified three burials as double burials (type 1): 179/180 (ALL001, early adult female; ALL002, infans I), 184/185 (ALL003, infans II; ALL004, late adult to early mature male) and 203/204 (ALL005, late adult male; ALL006, infans I). The two burials 188/189 (infans I; late adult to early mature male) and 208/209 (infans I; early adult female) were classified as double burials with unclear simultaneity. (28,29) Dirlewang (DIR; Landkreis Unterallgäu, Germany): The Alemannic site of Dirlewang is a very small cemetery with 40 excavated burials and an expected number of 55 burials in total. It dates to the Late Merovingian period, from 650-700, based on the archaeological finds. Two double burials (type 1) were found on this site, 33/34 (DIR001, juvenile to early adult female; juvenile male) and 38/39 (adult male; DIR002, adult female). Graves 18 and 19 (mature male; early adult male) did not share the same grave pit but were buried very close to each other, indicating a connection. Burials 30/31, 36/37 and 2 were classified as non-simultaneous successive or additive double burials (type 2).

Dittenheim (DIT; Landkreis Weißenburg-Gunzenhausen, Germany):
The early medieval cemetery of Dittenheim recruited from a settlement on the site of the modern village. Only 6.5 km south of the limes, the settlement was probably well connected to the remaining Roman infrastructure. 2.5 km Southeast of the cemetery, remains of a Germanic fortification dating to the Migration Period were found, known as Gelbe Bürg. This structure went out of use around 500, just before the region fell under the rule of Franks. The cemetery was known already by 1937. The first excavation campaign in 1968 revealed the first 114 burials. Later campaigns took place in 1971 (burials 115-164) and 1972 (165-244). The excavations probably reached the borders of the cemetery. However, some burials might have been lost due to erosion from a nearby river and from plowing. Besides the early medieval burials, older settlement traces from the Linear Pottery culture, Pre-Roman Iron Age and the imperial Roman period as well as one Bronze Age burial were found on the site.
The cemetery of Dittenheim, dating from the middle of the 6 th century to the end of the 7 th century, revealed 238 graves containing 244 individuals and 10 cremations from the same period. Other remarkable finds were three horse burials and one circular pit. Four burials were classified as double burials type 1: 8A/B (DIT007, mature female; mature male; possibly an additive burial), 18A/B (DIT003, mature male; DIT004, adult female), 22A/B (DIT005, infans I; DIT006, adult female) and 188A/B (DIT001; DIT002, both infans II). There were also three exceptional single burials in prone position (180,201,204) deviating from the contemporary burial rites. (31) Edix Hill (EDI; civil parish Barrington, Cambridgeshire, United Kingdom): The Anglo-Saxon cemetery of Edix Hill, close to Barrington and Orwell was initially discovered in the 19 th century. Excavations between 1989 and 1991 revealed part of an inhumation cemetery comprising 149 individuals in 115 graves, dating to between 500 and 625/650, although one burial was believed to be possibly Iron Age. It is estimated that there may originally have been around 300 burials in total with a complete cross-section of the population by age and sex suggesting that the burials relate to a community of 50-65 people spanning around 150 years. The initial dating of the cemetery was based primarily upon artefact typologies of the various grave goods and seriation by correspondence analysis, with burials broadly divided into earlier and later groups and some evidence for spatial patterning over time. Subsequent radiocarbon dating has broadly confirmed and refined the dating of the cemetery. Palaeopathological evidence suggested the presence of tuberculosis, leprosy and cancers. The human remains are currently held by Cambridgeshire County Council, who generously provided access to the material. The cemetery contained a total of 18 multiple burials, of which 10 were classifiable as type 1: they comprise one quadruple burial with at least two of the individuals buried simultaneously and eight simultaneous double burials. Four double burials are of unclear simultaneity, another four and a triple burial are clearly understandable as type 2 (non-simultaneous). The site was initially not sampled for the purpose of plague screening, so samples from both multiple and single burials were screened. This includes the single graves 76 (Sk405, EDI001; juvenile), 69 (Sk359, EDI002, early adult female), 78 (Sk424, EDI003, juvenile), 46 (Sk146, EDI010, early mature male), 60 (Sk183, EDI011, adult female), 63 (Sk198, EDI012, early adult male), 83 (Sk436, EDI014, early adult female), 90 (Sk458, EDI017, senile female), 95 (Sk530, EDI018, late mature female), 97 (Sk551, EDI019, early adult male), 99 (Sk576, EDI020, mature male), 100 (Sk578, EDI021, early mature male), 105 (Sk592, EDI022, late mature female), the nonsimultaneous double burial grave 66 (Sk322A, EDI013, at least adult male; SK322B, early adult female) and triple burial grave 18 (Sk42A1, infans I; Sk42A2, early adult with indet. sex); Sk42B, EDI009, early adult female), the simultaneous double burials grave 96 (Sk547A, adult female; Sk537B, EDI004, infans II), 106 (Sk626A, EDI005, early adult female; Sk626B, EDI006, early adult male), 9 (Sk13A, EDI007, mature to senile male; juvenile male) and 84 (Sk440A, EDI015, adult female; Sk440B, infans I) as well as the complex burial grave 2, with two individuals buried simultaneously (Sk3B, adult to mature male; Sk3C, EDI007, early adult male) and two more individuals buried later (Sk3A1 and Sk3A2, both adult and possibly female and male respectively). (32,33) Forchheim (FOR; Gemeinde Pförring, Landkreis Eichstätt, Germany): Within the vestiges of a late medieval settlement, a quadruple burial was found without any context suggesting a larger burial ground. Two mature women (FOR002, individual 2; ind. 4), a mature man (FOR003, ind. 3) and a male juvenile (FOR001, ind. 1) were buried partially overlapping but all in East-West orientation. Remarkably, one of the women (2) was buried in the prone position. Grave goods such as a sax and a glass bead necklace date the burial to the second half of the 7 th century. A belt buckle was indicative of a clothed inhumation. (34) Grafendobrach (GRA; Gemeinde Kulmbach, Landkreis Kulmbach, Germany): The Carolingian cemetery of Grafendobrach revealed 85 burials; the excavations 1975-1976 did not reach the edges of the burial ground. The various garment artefacts found in the graves point to the late 9 th to early 10 th century. Besides several secondary burials (type 2), there were two remarkable graves: complex 83/84/85 consisted of two adjacent stone cists holding two women (83, GRA002, early mature; 85, GRA003 early adult) and an infant (84, ca. 1 year old) buried at the feet of 83. A single burial of a man (42, GRA001, early senile) was covered with stones, on top of which two neonates (43,44) were found without individual burial pits. (35) Kleinlangheim (KLH, Gemeinde Großlangheim, Landkreis Kitzingen, Germany): The Frankish cemetery of Kleinlangheim dates to the late 5 th to early 8 th century and was completely excavated from 1962 to 1969. It contained 244 burials and a remarkable number of 56 cremations; unlike other contemporary cemeteries, the graves showed signs of grouping, potentially indicating family groups. Nine multiple burials type 1 were identified, eight of them double burials: 35/36 (KLE005, adult male; late mature male), 41/42 (late adult male with skull fracture; juvenile to early adult), 147/148 (min. adult; neonate), 171/172 (two infans I), 208/209 (infans I; early adult female), 272/273 (late adult male; KLE004, senile male), 211 (KLE001, mature male; second individual cremated) and 218/219 (KLE002, mid-adult to mature male, KLE003, late adult to early mature female). The triple burial 100-102 contained remains of two infants (1-2 years old; 4-6 years old) and one mature individual of indeterminable sex. (36) Leobersdorf (LEO, Bezirk Baden, Austria): The Avar cemetery of Leobersdorf was excavated between 1977 and 1983 and was dated to 640-800. In total, 154 burials containing remains of 171 individuals were excavated with an exceptionally high number of 25 multiple burials. 16 burials were identified as double burials, of which 9 contained each an adult and a subadult (16: adult female and infans I; 57: mature male and infans I; 74: adult female and infant; 93: mature male and infans I; 100: mature male and juvenile female; 104: adult female and infans I; 114: adult male and infans I; 119: late mature male and infans II; 140: adult female and infans II). Four consisted of two adults (35: mature male and female; 86: adult female and mature male; 114: adult female and male; 144: senile and adult male) and two held only subadults (23: male juvenile and infans II; 103: two juvenile females; 145: infans I and II). The remains of infant individuals often were identified only after the morphological examination following the excavation. Three burials were identified as triple burials (21: senile male, mature female and infant; 67: adult female and two infants; 82: juvenile male, LEO001; adult female, LEO002; infans I, LEO003). A fourth (99) was interpreted as a double burial (mature male and infant) with a secondary burial (female). Two burials contained four individuals (105: mature female, juvenile female, infans I and neonates; 134: adult female, mature male, infans I and II). Burial 79 contained the remains of five individuals, identified as the burial of a mature female with secondary burials of four additional individuals (senile female and male, juvenile female and infans II) that had probably been buried on the same spot previously, and were reburied after the mature female. (37,38)

Lunel-Viel (LVH, LVC; Arrondissement Montpellier, Département Hérault, France):
The site of Lunel-Viel is equidistant (25 km) from the modern and Roman towns of Montpellier and Nîmes at the intersection of two paved roads that were active in late Antiquity: a later Roman secondary road running about 2 km southeast and parallel to the Roman Via Domitia that connected Iberia to Italy, and a road connecting the Via Domitia to the coastal lagoon l'Étang d'or, part of the complex of lagoons that included that of Lattes (anc. Lattara), famous already in Antiquity for its fishery (Pliny the Elder, Natural History, 9.29-32), and Maguelone, an island bishopric first mentioned in the 6 th century when abundant emerging archaeological evidence documents far-flung Mediterranean shipping connections. A brief Roman-era occupation in the 2 nd century BCE at Lunel-Viel was followed by continuous settlement beginning ca. 50-80 CE, although the ancient settlement area was abandoned in the 7 th century, and the exact location of the new dwelling zones between the 7 th and the 10 th century remains unknown. Three inhumation cemeteries received the deceased of this community in succession from the 4 th century. The earliest, at Le Verdier, functioned from the end of the 3 rd century down to the beginning of the 6 th century, and yielded 340 burials. The second, known as Les Horts, received burials of the late 5 th to 7 th centuries; 140 burials were excavated out of an estimated original total of ca. 200. The third is associated with the church of St. Vincent; burials began there ca. 520 and continued down to the 17 th century; 97 have been excavated. The cemetery Les Horts contained one clear case of a double burial, likely simultaneous (type 1), of two adult individuals (38A/B) placed head facing toes in a sarcophagus; the head and upper body of 38B were destroyed, along with the feet of 38A and part of the sarcophagus, during later leveling operations. Disarticulated remains of a third poorly preserved adult (38C, LVH001) were found pushed into the eastern end of the sarcophagus. Grave goods likely stemming from the original burial (one buckle, one plated buckle, one fibula, one pin and one clasp) date the burial of 38C to 475-550.
Outside the cemeteries, anomalous (deviant) burials of the remains of eight individuals were found during the excavation of the Gallo-Roman structures, called the Quartier central. Robbing of foundation stones in Antiquity had left empty trenches which were subsequently used for the summary interment of these individuals. A pin buckle, a spindle whorl and a knife found with the remains of the presumably clothed individuals date the interment to the 5 th to 6 th century. The positions and postures of the individuals clearly deviate from the burial customs seen in the contemporaneous cemeteries of Les Horts and Le Verdier: two adult women (3A: LVC003, adult; 3B: LVC004, mature) were placed under limestone slabs, with one resting her head on the thighs of the other. Westward of this group, a woman (1: LVC001, early adult) and another adult individual (2: LVC002, late adult to early mature, sex indeterminable) were found; the latter however had been disturbed by later agricultural activities. Whereas these individuals appear to have been laid rather carefully on the ground, the two men adjacent to the west (4: LVC005, mature, 5: LVC006, early adult) seem to have been carelessly dropped into the trench, the latter with his arms stretched over his head. Two more individuals, a young adolescent woman (6,LVC007) and an infant (7, sex indeterminable), were found in a second trench to the east. However, all individuals were interred more or less in supine East-West oriented position, as expected for early medieval Christian burials. (39,40) München-Aubing (AUB, Stadt München, Germany): The Baiuvarian cemetery of München-Aubing was excavated in two phases, 1938 and 1960-1963. It was used from the 5 th to 7 th century. A total of 896 graves was excavated. Four double burials (type 1) have been identified: 854/855 (AUB008, adult male; late mature male) 809/810 (AUB006; AUB007, two adult males), 724/725 (AUB004, adult male; AUB005, mature male; with possible later manipulation) and 676/677 (AUB002, mature female; AUB003, male individual). A third individual (675, infans I) was buried later on top of the double burial 676/677. (41) Neuburg an der Donau (NEU; Landkreis Neuburg-Schrobenhausen, Germany): The site of Neuburg an der Donau was occupied at least from the first half of the 2 nd century by a Roman military base associated with an urn cemetery with around 130 cremations. The burial ground Seminargarten is related to a later occupation period during the 4 th century. From an estimated number of 150 burials, this site revealed 130 burials with 133 individuals. One burial was identified as a double burial (32A/B, adult female and juvenile with indeterminable sex), one as a triple burial (34A/B/C: mature male; NEU001, mature male; NEU002, adult male). Both are part of zone 1, dating to 330-360. Two additional burials (14,28) were suggested to contain "mother and child" (neonate/infant). (42,43) Peigen (PEI; Markt Pilsting, Landkreis Dingolfing-Landau, Germany): The early medieval cemetery of Peigen dating to the mid-5 th to the 7 th century revealed around 274 burials. Among these, three could be identified as double burials. Double burial 18 contained an adult woman PEI001) and an early adult woman PEI002), buried in opposite orientation. Double burial 63 contained an adult woman (63-1, PEI004) and a senile man (63-2, PEI005). The burial of a late mature male (109-1) and a child (infans I, 109-2) form the third clear double burial. The burials 62-1 of a juvenile male and 62-2 of a late mature female (PEI003) were adjacent but at different depth and therefore of questionable simultaneity (type 2). (44) Petting (PET; Landkreis Traunstein, Germany): Although located in Germany today, Petting falls in a region that was in ancient times associated to the Roman city and later ecclesiastical center of Iuvavum (today Salzburg, Austria) and has therefore a different settlement history than other parts of southern Bavaria. The cemetery of Petting was discovered in 1991 and excavated within the following two years. A comprehensive publication of the site of Petting is still pending, not least because the archaeological findings were not available for research until the ownership situation was clarified in 2010. Studies so far concentrated on individual burials or specific artefacts. The cemetery was completely excavated, counting 721 burials. Only 24 burials were lost due to destruction prior to the excavation: around 50 % of the burials fell victim to ancient grave robbers. At least four multiple burials were observed in the cemetery, all presumably type 1. The double burial 172/173 contained a female and a male individual; the double burial 377/378 a late adult to mature and a late adult to early mature male (PET003/PET004), the triple burial 342/343/344 at least one early adult male (PET001) and an infans II (PET002). The double burial 630/631 of an early adult female (PET006) and an early mature female (PET007) was associated with a single burial of an infans II (PET005) on top. (45,46) Regensburg Fritz-Fend-Straße (RFF; Stadt Regensburg, Germany): Regensburg, Castra Regina, was an important Roman military base with a civilian settlement of the province Raetia. The extensive adjacent cemetery Kumpfmühl was excavated in several campaigns following the first excavation in 1872 by J. Dahlem. He documented 827 burials and cremations but left the vast majority of features undocumented, which led S. v. Schnurbein to the assumption that the first excavation uncovered around 3000 cremations and 2000 burials, mainly dating to the late 3 rd to mid 4 th century. More features were lost in this area due to later construction works without archaeological survey. An excavation in 1999 revealed 100 additional cremations and 50 burials. The excavations of Fritz-Fend-Straße and Im Güterbahnhof took place in 2011 and revealed 48 cremations/115 burials and 161 cremations/116 burials, respectively. The site Fritz-Fend-Straße contained two exceptional double burials of type 1: In burial 30, the second individual (30-1: RFF001, late adult to early mature male) was laid in prone position and opposite orientation on top of the first , who was buried in supine position. In 53, the two individuals, a juvenile female  and an early mature male RFF003), were buried in crouched position facing in the same direction but in opposite orientation. An archaeological or radiocarbon date for Fritz-Fend-Straße has not yet been published, but a similar date as for Kumpfmühl can be assumed. (47) Saint-Doulchard Le Pressoir (LSD; Arondissement Bourges, Département Cher, France): In the former village of Saint-Doulchard, 2 km from Bourges, an archaeological survey in 2007 revealed a dense funerary space, dated from the 7 th to 12 th centuries. The town is first mentioned as Sanctus Dulcardus in the 7 th century, named after the hermit Dulcardus (late 6 th century). It is likely that the cemetery was established on the site of Dulcardus' hermitage. However, a tumulus (6th to 4th c. BCE) and a Roman villa found in the vicinity of the excavated funerary space attest earlier occupation of the site. In 2009, the rescue excavation led by P. Maçon excavated a part of this cemetery measuring approximately 400 m², reaching one of its boundaries in the form of a ditch that borders the burial area and runs from the northeast to the southwest. This ditch was also used for burials: 57 individuals in 48 burials have been discovered within it whose organization and funerary practice may indicate a particular mortality crisis. However, some secondary burials and stratigraphically overlapping graves within the ditch could suggest that it was used on multiple occasions. Eleven multiple burials were found, including nine double burials F206-35/36 (mature; fetus), F206-68/69 (LSD008, mature to senile male; infans I), F206-71/72 (LSD009, senile male, infans I), F206-79a/79b (LSD011, mature to senile male; infans I), F206-93/145 (LSD012, early adult; early adult male), F206-152a/152b (infans I, infans II), F206-155a/155b (senile female; LSD019, mature to senile male), F206-156a/156b (LSD020, mature to senile male; infans I) and F206-204a/204b (LSD023, mature male; infans I-II), and two triple burials F206-91a/91b/91c (mature male, infans I, infans I), F 206-172a/172b/172c (infans I; LSD022, juvenile; LSD021, mature female). Three more double burials were found within the cemetery area delimited by the ditch: F206-75a/-75b (infans I to juvenile; late adult to mature female), F206-229a/-229b (late adult to mature female, infans I) and F206-235 (at least adult male; subadult). In addition to several multiple burials, a selection of single burials within the ditch were also sampled for this study: A palaeodemographic analysis highlighted similarities in the structure of the population buried in the ditch with mortality profiles frequently observed in epidemic contexts, including plague. This suspicion is reinforced by the taphonomic evidence arguing for simultaneous interments (type 1) in an easily available structure, separated from the rest of the community. Nevertheless, the presence of multiple burials in the space enclosed by the ditch might indicate a gradual shift in funerary practice at the beginning of an epidemic when mortality was still limited. Radiocarbon dating was performed on five individuals buried in the ditch. The calibrated 2sigma intervals cluster in two groups (ca. 650-880 and ca. 720-920), which might indicate the usage of the ditch as funerary space for multiple events. A combination of all five dates is rejected by Oxcal (df=4, T=11.2 [5% 9.5]), further substantiating the hypothesis of multiple events. (20) Sindelsdorf (SIN; Landkreis Weilheim-Schongau, Germany): The early medieval cemetery of Sindelsdorf consists of 331 burials with remains of 354 individuals dating from 500 to 720. It included three double burials type 1 (25/26 senile male and mature male; 51/52, senile and mature male (SIN006, SIN007); 154/155, infant and senile female) and a burial group of three individuals with one buried later on top (163: infant; 164: SIN002, late adult to early mature male; 165: SIN003, adult female; 162: SIN001, senile male). The double burial 25/26 with a later burial on top (24, mature female) may be identified as a probable homicide due to a perimortem skull fracture (26). (48) Straubing Azlburg I/II (SAZ; Stadt Straubing, Germany): Straubing, called Sorviodurum by the Romans, was an important castrum on the Danubian border of Raetia. The Late Roman cemeteries -around 200 m apart from each other -were excavated in 1981 (Azlburg I) and 1984 (Azlburg II). However, the borders of the cemeteries were not reached on all sides, so the original dimensions remain unclear. They were in use at roughly the same time, approximately between 300 and 450. Azlburg I contained 107 graves with 111 individuals and one cremation. The burial 54-1/-2/-3 was identified as a triple burial (SAZ002, male juvenile; SAZ003, male juvenile; third individual min. adult). The double burial 16/17 contained the remains of two children (SAZ001, infans I; neonate). Azlburg II contained additional 434 graves with 45 individuals. The double burial 5a/b of two adult men was interpreted as the burial of two soldiers.

Unterthürheim (UNT; Gemeinde Buttenwiesen, Landkreis Dillingen an der Donau, Germany):
The town of Unterthürheim is probably the site of the early medieval settlement that used the burial ground Buttenwiesen. Traces of an older Imperial Roman settlement (1 st to 5 th century) were found on the Thürlesberg hill, around two kilometers from the cemetery; it was near the Roman roads Via Iuxta Danuvium (following the Danube) and the Via Claudia Augusta (connecting Augusta Vindelicum/Augsburg with Italy) which intersect at the fort Submuntorium, around 10 km Northeast of Unterthürheim.

Valencia Plaça de l'Almoina (VAL; Ciudad de Valencia, Spain):
The site of Valencia, Plaça de l'Almoina is an intramural cemetery that was founded in the 5 th century and remained in use during the Visigothic period in the 6 th and 7 th centuries. It was excavated from 1985 to 1999. Contrary to the earlier Roman practice, the necropolis was not located outside the city walls but close to a shrine commemorating the martyrdom of St. Vincent, and adjacent to the cathedral. Many burials still follow the later Roman tradition that made use of tegulae and amphoras. Among the later burials, there are several monumentalizing collective tombs that were probably used as elite family tombs for successive burials. The excavated areas of the necropolis contained a number of multiple burials of differing types and dates, including 15 or 16 collective slab tombs that appear to reflect successive burials (type 2: details in McCormick 2016, 1024n15). At the moment of sampling, the human remains were comingled and only sorted by find numbers (corresponding the numbers VAL001-009), so the assignment to individuals is not possible for this site. Four or five features seemed to be of type 1, i.e., reflecting at least in part simultaneous burials. The quadruple burial tomb 4 was covered with tegulae and contained four individuals who were piled one on top of another in an east-west orientation (two samples were taken, VAL008). Tomb 28 was stone-lined and just north of an apse identified as a memorial shrine to the martyrdom of St. Vincent; it contained a minimum of 21 individuals, from whom sixteen samples were taken (VAL005, VAL006, VAL007). Tomb 50, just southwest of Tomb 28 (and actually under the wall of that apse), may well have been a continuation of Tomb 28; it contained at least seven individuals, from whom seven samples were taken (VAL003, VAL004). Tombs 28 and 50 are lower than and earlier than the wall of the apse. The apse itself seems well dated to the late sixth or early seventh century, based on pottery excavated in a pit under the apse pavement. Multiple tomb 40 was a pit burial containing four supine individuals (five samples, VAL002) oriented east-west in the entry from the east into the apse of the memorial shrine. Tomb 41 is south of the apse structure which might indicate a privileged burial space. It is a pit that seems to have used pre-existing Roman walls to the south and east for two sides; bricks were simply placed without mortar to form the other two sides of the burial space, which might suggest a hasty or improvised burial. It contained the very disturbed remains of at least 15 individuals, including 4 subadults, from which we took seven samples (VAL001). (38,(51)(52)(53)

Waging (WAG; Landkreis Traunstein, Germany):
The early medieval cemetery of Waging, close to the site of Petting, was discovered and excavated in 1987/1988. The excavators deduced a discontinuity between the previous Roman settlement (until 300) and the early medieval colonization related to the burial ground that was in use between 530 and 700. Similarly to the site of Petting, the resettlement was most likely dominated by Iuvavum (modern Salzburg, Austria) in the southeast. Although the 239 burials were subsequently analyzed and the artefacts went into a local exhibition after extensive conservation measures, the site is still not published in a comprehensive way. At least three burials were identified as multiple burials: the double burial 200/201 of an early mature male and infans II (WAG002; WAG003) were located on top of the burial of a potentially male infans II (WAG001). The double burial 37 contained remains of a potentially female infans II (WAG004) and a juvenile to early adult individual (WAG005). Burial 39 was identified as a double burial of a potentially female infans II (WAG006) and a late mature female (WAG007). (45,54) Westheim (WES; Landkreis Weißenburg-Gunzenhausen, Germany): The Merovingian cemetery of Westheim-Mehlbuck, excavated between 1979 and 1985, revealed remains of 255 individuals in 228 graves and dates to the 6 th to mid-7 th century. Five burials were clearly identified as double burials: 17a/b, 26a/b, 36a/b, 52a/b, 106a/b and 128a/b. Six additional burials were interpreted as presumably double burials of a child with a parent: 172, 190, 202, 206, 208 and 210. The triple burial I, uncovered during a previous excavation in the 1910s, held the remains of a late adult to early mature male (1-1, WES001), a late mature female (1-3, WES003) and of a third individual (1-2, WES002, min. adult, probably female). Burial 13a/b was reported by H. Lüdemann as a quadruple. However, according to R. Reiss, the burial contained only disturbed remains of two males (adult, mature). (38,55) PO 5.4, 7.4;8.3;11.1 (1910, 1911, 1912, 1915  The "Ns" are called due to heterozygosity (grey), all false positive SNPs were filtered out by our SNP evaluation (orange). With increasing coverage, the foreground has a high enough coverage to outshine the contamination and enable a reference call. See also SI Appendix, Table S6. Background Sensitivity (false positive SNPs) filtered out reference call Reference call filtered out SNP call N Fig. S3: Stacked bar chart for the 418 core positions for the backbone phylogeny of modern Y. pestis in the simulated datasets with ("Contaminated", right bar) and without ("Clean", left bar) background (DIR002.A). Whereas for the clean datasets, the number of "Ns" (grey) and filtered positions (orange) are only dependent on coverage and go to zero for the 30-fold coverage datasets, more positions are lost as "Ns" due to heterozygosity or filtered out due to contamination. See also SI Appendix, Table S7.       Fig. S9: Historical geography of the tested sites in Germany and Austria. A: Detailed map of all tested sites in Germany and Austria in relation to main rivers (blue), the Roman road network (orange) and the Roman provinces in their last state before the fall of the Western Roman Empire (shaded regions); all sources are given in the SI Appendix. The sites of published genomes are depicted by pink squares, the sites for the genomes presented here are represented by yellow squares. Sites tested negative are depicted in black upward-pointing triangles (burials dating before 541), squares (dating around 541-544), downward-pointing triangles (dating after 544). B: Relation of the detailed map (rectangular space) to the sites in Britain (Edix Hill), France (Lunel-Viel, Saint-Doulchard) and Spain (Valencia) presented in this study.  gives the archaeological IDs of simultaneously buried individuals or individual from the same collective burial (Valencia: for details see site description, SI Apendix). 'Dating' gives the range of the archaeological dating CE for each burial. 'Total' list the total number of individuals, further broken down to adults (>20 ya) and subadults (<20 ya). 'Arch. ID' and 'Lab ID ind.' are showing the individual ID with the morphological 'sex'(f=female, m=male, ?=indeterminable) and 'age' (i=infans 1, ii=infans 2, j=juvenile, a=adult, m=mature, s=senile, (a)=at least adult, e=early, l=late). Teeth are noted in the FDI system or letters (I=Incisor, C=Canine, P=premolar, M=molar, lower case letters for deciduous teeth, indet=indeterminable). The qPCR screening results for Y. pestis are given in the 'Scr' column, samples screened with MALT are indicated by a star. The 'Seq' column shows if samples were considered as positive after sequencing.       Table S7: Numbers of 418 core positions for the backbone phylogeny of modern Y. pestis in the simulated datasets with ("Contaminated") and without ("Clean") background (DIR002.A); see also SI Appendix, Fig. S3. Positions were either called as "reference call" or "Ns" representing positions not called due to low coverage or heterozygosity. "Positions filtered out" give the number of positions that did not pass our SNP evaluation. The positions "lost by heterozygosity" give the difference of reference calls (contaminated) to reference calls (clean). The positions "lost by contamination" give the difference of remaining calls after filtering (contaminated) to remaining calls after filtering (clean).  Table S8: Table with all non-shared chromosomal SNPs that were called in the investigated First Pandemic genomes, sorted by genome (second cell in header). 'Position' refers to the position in the reference genome CO92, followed by the respective SNP call and the reference call. The classification as potential false or true positive is based on the preceding parameters regarding a 50 bp window surrounding the SNP: Only SNPs with no uncovered position, no heterozygous SNP and an equal mean coverage with high and low stringent mapping (ratio of 1.00) are classified as true positive SNPs (highlighted in green). If a SNP is ambiguous due to conflicting classification in different samples, this is indicated with a question mark highlighted in orange. The 'comment' column with additional classification is followed by the respective calls in the other First Pandemic genomes (0=not covered, N=heterozygous, lower case letters=less than 3 reads coverage). The last column is giving the corresponding genomes, if the SNP appears as homoplastic in the SNP                      T T T N T T A T t  T t  N   231350  C  T  24  1  1.58  false positive  T T T T T T T N N T T g   3155190  C  T  0  1  1.91  false positive  T T T T T t  T T T T    . 'Position' refers to the position on the plasmid of the reference genome CO92, followed by the respective SNP call and the reference call. The classification as potential false or true positive is based on the preceding parameters regarding a 50 bp window surrounding the SNP: only SNPs with no uncovered position, no heterozygous SNP and an equal mean coverage with high and low stringent mapping (ratio of 1.00) are classified as true positive SNPs (highlighted in green). If a SNP is ambiguous due to conflicting classification in different samples, this is indicated with a question mark highlighted in orange. The 'comment' column with additional classification is followed by the respective calls in the other First Pandemic genomes (0=not covered, N=heterozygous, lower case letters=less than 3 reads coverage).       Table with all shared SNPs that were called in the First Pandemic genomes, giving the results of the SNP evaluation for each genome separately. Position is referring to the position in the reference genome CO92, followed by the reference call. For each genome, the call at the SNP position is given (0=not covered, N=heterozygous, lower case letters=less than 3 reads coverage). The classification as potential false or true positive is based on the preceding parameters regarding a 50 bp window surrounding the SNP: Only SNPs with no uncovered position, no heterozygous SNP and an equal mean coverage with high and low stringent mapping (ratio of 1.00) are classified as a candidates for true positive SNPs. SNPs that are supported by less than half of the genomes (<6, last column) are excluded as inconclusive (highlighted in orange). Although the WAG001.A sample was not included in this analysis due to low coverage, all positions were checked for a rough phylogenetic classification. Similarly, the pre-Justinianic DA101 sample was checked for all positions as well. Mean coverage ratio LS/HS 86824 A G 0 0 1.00 0 43 0 1.00 G 0 0 1.00 g 12 0 1.00 G 0 0 1.00 g 8 0 1.00 g 11 0 1.00 0 48 0 1.00 G 0 0 1.00 g 0 0 1.00 g 0 0 1.00 0 48 0 1.00 A 0 6 189912 A G 0 0 1.03 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 G 0 0 1.00 A 0 11 260148 C T 0 0 1.00 T 0 0 1.00 T 0 0 1.02 T 0 0 1.00 T 0 0 1.00 0 20 0 1.00 T 0 0 1.00 t 6 0 1.29 t 0 0 1.00 T 0 0 1.00 T 0 0 1.00 t 0 0 1.00 T 0 9 271114 C A 0 0 1.00 A 0 1 1.26 A 0 0 1.01 A 0 0 1.00 A 0 0 1.00 A 0 0 1.00 A 0 0 1.00 a 0 0 1.69 A 0 0 1.00 A 0 0 1.00 A 0 0 1.23 A 0 0 1.00 C a 8 485976 C T 0 0 1.00 T 0 0 1.00 T 0 0 1.02 T 9 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 t 1 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 C 0 9 557841 C T 0 0 1.00 T 0 0 1.03 T 0 0 1.02 0 11 0 1.00 T 0 0 1.00 t 0 0 1.00 T 0 0 1.00 T 0 0 1.06 t 0 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 C 0 8 727741 G A 0 0 1.00 A 0 0 1.00 A 0 0 1.00 a 0 0 1.00 A 0 0 1.00 0 25 0 1.15 a 0 0 1.00 A 0 0 1.00 a 0 0 1.00 A 0 0 1.00 a 0 0 1.28 A 0 0 1.00 G a 10 779365 C T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 0 16 0 1.00 t 0 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 T 0 0 1.00 C 0 11 898980 A T 0 0 1.00 t 0 0 1.00 T 0 0 1.00 t 0 0 1.00 t 0 0 1.00 T 0 0 1.00 t 0 0 1.00 T 0 0 1.00 t 0 0 1.00 T 0 0 1.12 t 0 0 1.00 t 0 0 1.00 A 0 11 Table S12: List of all modern and ancient Y. pestis genomes with accession number, sample origin and the corresponding publication.  Table S14: Table giving all unique and shared SNPs that were called in the re-analyzed Aschheim genome (A120), showing the classification as assessed to the other genomes. The classification is based on the preceding parameters regarding a 50 bp window surrounding the SNP: Only SNPs with no uncovered position, no heterozygous SNP and an equal mean coverage with high and low stringent mapping (ratio of 1.00) would be classified as true positive SNPs (highlighted in green, and blue if the coverage is below 5).