The York Gospels were probably made at Canterbury between about 990 and 1020, and they are held at York Minster. The book is a stack of animal skins, and three of its pages carry the DNA of the virus that causes sheeppox. Two of the three belong to the original book and one to a later medieval document bound in with it.

So do a page of a Latin glossary made in England around 800, a page from a copy of the Epistles of St Paul written about a century before that, and four leaves of a fifteenth-century volume once bound with the Vinland Map. Seventeen positive parchment folia in all, including ones from Britain, the Upper Rhine, Austria and France, plus one sheet of paper.

A disease that empties a flock

Sheeppox is caused by a large double-stranded DNA virus in the same family as smallpox, though in a different genus. Morbidity in an immunologically naive herd can reach 75 to 90 per cent, mortality runs around 50 per cent, and in young animals it can reach 100. Milk, wool and meat production all drop sharply in affected animals, and their hides are scarred by the lesions.

The genomes come from a study in Science Advances led by Louis L’Hôte and Kevin G. Daly of University College Dublin, published on 2 September. It reports 21 new genomes of the virus and, the authors say to their knowledge, the oldest genomes of any poxvirus recovered so far.

Sheeppox is a notifiable disease under the World Organisation for Animal Health. Europe eradicated it during the last century, and it has not stayed away: L’Hôte notes in a statement released by UCD that there is an outbreak running in Greece now.

Its written history is long and patchy. The Roman writer Columella describes “the pustule” in sheep in the first century AD, and outbreaks of pockes or variola in flocks recur in French, Spanish and English treatises between the thirteenth and sixteenth centuries. In Britain the references then stop, and when infected sheep brought the disease back in 1847 it was treated as new and foreign. None of those records describe the virus.

The screen behind the headline

To sample a manuscript page without damaging it, a conservator rubs it gently with an eraser and keeps the crumbs. The dust that comes off carries DNA from the animal whose skin the page is, and sometimes the DNA of what that animal was carrying.

The team screened 754 samples in total, drawn from material spanning more than 14,000 years across Eurasia, including sheep teeth from archaeological sites and parchment and paper from European libraries and archives.

Forty-one of those samples returned enough reads assigned to the capripoxvirus genus to clear the study’s thresholds, and in every case the closest match was sheeppox rather than its two relatives, goatpox virus and lumpy skin disease virus. Samples with more viral DNA were shotgun sequenced; thinner ones were enriched using RNA baits designed to fish capripoxvirus sequences out of the pool.

That produced 22 unique genomes, 21 of them new. Their depth varies enormously. The paper gives its range as 0.1 to 260.6 times coverage with a median of 1.5, and the inclusion threshold was a tenth of one time coverage with at least a tenth of the viral genome present. Only nine of the 22 clear five-fold coverage, and the phylogenetic claims rest on those nine.

Ten of the seventeen positive pages were calfskin

The paper puts the parchment production dates between the early eighth and the fourteenth centuries, although its own table runs several of the positive leaves into the fifteenth. The York Gospels are York Minster Library MS Add. 1; the Corpus Glossary, at around 800, is Corpus Christi College Cambridge MS 144, also probably Canterbury work; the Epistles of St Paul are Trinity College Cambridge MS B.10.5.

Then the unexpected part. Of the seventeen positive parchment folia, ten are calfskin, four are goatskin and three are sheepskin. Sheeppox has been reported in a small number of outbreaks in goats. In cattle, the authors write, there are no reported natural cases that they know of. Parchment was typically produced from calves six to eight weeks old, and the authors note that if the cattle had been exposed and had mounted an antibody response, their calves should have been protected by colostral immunity, and their skins should not be carrying the virus.

The study supplies its own qualification. Eight of the seventeen parchment samples carried mixed animal DNA, and the species recorded is simply the one most represented in the alignment. In six of those eight, sheep is the second most common signal, and one more gives a predominantly sheep reading.

The paper sets out four possibilities for the non-sheep positives and settles none. The virus may genuinely have crossed into those animals. Virions may have moved between skins in the lime pit where flayed hides were soaked, since the paper notes SPPV can persist outside the body on the scale of months, particularly from shed scabs. A positive sheep page may have contaminated the calf page pressed against it. Or sheep-derived products may have gone into the production, repair or modification of a non-sheep leaf. That last is close to the explanation the authors reach for with the single positive paper document, since paper of the period was often sized with gelatin boiled from leather scraps and parchment trimmings.

On one calfskin bifolium the team sampled the flesh side and the hair side separately and found tentative evidence of more virus on the flesh side, which would point to real infection. The difference was not statistically significant. That same leaf carries no sheep mitochondrial DNA at all, which the authors say makes contamination from a sheep source a less likely explanation for it.

One of the four positive leaves from the Yale volume complicates its own date. Folio 79RR is a repair. The volume’s ownership stamps were excised at the bottoms of several folios and the losses made good with replacement parchment, and the paper’s own table footnotes that leaf as a page “replaced at a later date” whose “sample age is potentially later than indicated”. At 5.2 times coverage it is nonetheless one of the nine genomes above the five-fold threshold that the tree-building rests on.

A tooth in central Kazakhstan

The two oldest genomes come from sheep molars at Late Bronze Age settlements on the Eurasian steppe. One is from Myrzhik in central Kazakhstan, radiocarbon dated to between 1875 and 1645 BC, and at 260.6 times coverage it is the deeper of the two Bronze Age genomes by a wide margin.

The other is from Rublevo-VI in the Altai Krai of Russia, dated to the later second millennium BC. Both were metallurgical production sites with dense populations of people and animals, and both ran herds managed to survive bad years. Sheeppox cuts straight across that, and the authors argue its unpredictable killing would have wrecked the scheduled culling of particular age and sex groups that steppe herders used to keep flocks viable.

Teeth kept the record. The same team screened the petrous bones of the French sheep in a seventeenth to nineteenth century mass burial at Louvres and found nothing, while four of the specimens taken from those nine animals were positive from tooth roots. Sheeppox occurs in the respiratory system, the oronasal tissue and the circulation, which the authors suggest can lead to long-term retention of viral DNA in dental remains.

How far back the Myrzhik molar pushes the record is stated three ways in the same paper. Its title says three thousand five hundred years. Its abstract says more than 3,700. Its introduction says at least about 3,800 years ago. The paper never reconciles them, and the UCD statement uses two of the three, four times over.

The radiocarbon determination underneath all of it is printed twice and not identically. The Oxford measurement on that molar, OxA-45742, appears as 3,340 plus or minus 19 years before present in the site provenance section and as 3,440 plus or minus 19 in the radiocarbon methods. That is a hundred-year gap on the study’s oldest specimen, and the paper does not flag it. This piece uses the calibrated range the paper prints for that molar, 1875 to 1645 BC, which is the figure it carries through its own results.

Genes that were already broken in the Bronze Age

Poxviruses tend to shed genes as they specialise on a host. Smallpox did it progressively, with more genes inactivated in modern strains than ancient ones.

Sheeppox did not. Of the 156 genes that work in lumpy skin disease virus, nine are inactivated in modern sheeppox, and all nine are already inactivated in the Bronze Age genome from Myrzhik. That constrains the timing of those inactivations to about 4,004 years before present at the latest, on a 95 per cent interval running from 5,107 to 3,662.

After that the paper reports no progressive gene inactivation through time, and what it calls a relatively constant number of active genes since the Bronze Age. Three individual ancient genomes do each carry further putative disruptions of their own, including the 1917 skin, in which an immune-modulating gene is cut short by a stop codon at amino acid 38. In one of the other two the authors decline to call the apparent truncation a pseudogene at all, because gaps in the coverage leave the reading frame uncertain, and the remaining genome carries mutations in two genes rather than one.

If losing the original nine was part of adapting to sheep, the paper’s reading is that the initial adaptation was already complete by about four thousand years ago. L’Hôte’s gloss in the UCD statement is that the sheeppox genome has been very stable over the last few millennia, unlike smallpox, so the changes before that point may be the ones that matter for fighting it in future.

The dates around all of this are model output, and the models disagree with each other. The 11,500 to 3,700 year window quoted for when the three capripoxviruses split is the widest range across the four models they tested, and one section heading in the paper compresses it to “within the last approximately 5200 years”. The split of the medieval European lineage from the ancestor of later sheeppox is dated to about 1,813 years before present, on a 95 per cent interval running from 2,541 to 1,348, an uncertainty of more than a millennium.

The shape of the family tree is less settled than the word the authors use for it. They write that their calibrated analyses “unambiguously” identify sheeppox as the earliest-diverging capripoxvirus. Left unconstrained, support for that root placement runs from 0.44 to 0.95 across clock models, a spread they attribute to their two Bronze Age genomes occasionally being thrown to the root of the whole genus. Forcing each of the three viruses to be monophyletic, which they justify from the maximum-likelihood placement, raises the support to about 1.0. A maximum-likelihood tree rooted on nine other poxviruses marginally favours lumpy skin disease virus as the first to diverge instead, which the authors say may reflect long branch attraction. They are clear that the only way to settle it is ancient goatpox or lumpy skin disease genomes, which nobody has.

Libraries as archives of animal disease

L’Hôte’s summary of the method is that the study shows parchment can preserve animal pathogen DNA, and that archives and libraries around the world may hold the genetic traces of past animal outbreaks. Every parchment collection is also a tissue bank of the livestock slaughtered to make it, dated by the handwriting on top.

The authors want ancient goatpox and lumpy skin disease genomes next, and say those too might come from animal skin documents.

The pages will not advertise which of them to sample. The positive folia show no visible lesions, which the authors say may not be unusual, since sheeppox lesions are often confined to the parts of a sheep that are not woolly and parchment-making scrapes the flesh away. They also name a manuscript from outside the study’s positives: there are indications of pox-like lesions on the membrane of the early eighth-century Codex Amiatinus, made from a combination of goat and sheep skin.