Following the results from the European Union project RESGEN CT96-78 on elm genetic resources, we studied the variability of chloroplast DNA in the two elm species native to Spain (the field elm, U. minor Miller sensu latissimo2, and the wych elm, U. glabra Huds.) and in elm samples from France, Greece, Italy and Britain (for methods, see supplementary information).

We analysed restriction fragment-length polymorphisms in chloroplast DNA after amplification by polymerase chain reaction and detected four major chloroplast lineages, which are maternally inherited in Ulmus (see supplementary information). Lineage A appears in Greece; lineage B in northern Italy; lineage C in Italy, the Iberian peninsula and Britain (but not in France); and lineage D throughout all of Spain, northern Italy, France and Britain. Lineage-D haplotypes appear in both U. minor and U. glabra samples. By contrast, in Spain and Britain, lineage-C haplotypes are observed only in U. minor, although they appear in both species in Italy (mainly in elms from the central region).

Except for the presence of lineage C in the Iberian peninsula and Britain, the distribution of lineages is very similar to those described for other trees3. Haplotype sharing among tree species is reckoned to indicate their presence in common glacial refugia, where chloroplasts would have been exchanged through hybridization and introgression4. Absence of lineage-C haplotypes in Spanish U. glabra highlights the lack of refugia for this lineage in Iberia, whereas its presence in both elms in Italy indicates that this lineage has an Italian origin.


Eighteen field elms with the same haplotype (from lineage C), collected from across Spain, Britain and central Italy, as well as five lineage-D individuals from Spain and Britain, were genotyped using seven nuclear microsatellites5 and two amplified fragment-length polymorphism (AFLP) primer combinations. The results revealed a widely distributed clone within lineage-C elms, represented here by five Spanish and three British samples (see supplementary information). We detected slight differences among the AFLP fingerprinting profiles of these eight samples, attributable to somatic mutations. The closest genetic similarity was observed between the clone and samples from Latium in Italy, which reinforces the idea that its origin was Roman.

These eight individuals were classified as English elm (U. minor var. vulgaris2, commonly called U. procera Salisb. by British botanists2). In Britain2 and in Spain6, English elm rarely sets seeds, but produces pollen normally and is very effective in vegetative propagation. These reproductive features, together with the deliberate plantation of this variety of elm by humans, could explain the maintenance of this genotype and its spread over Spain and Britain. But when, whence and why was it transported?

Although it has been suggested that the English elm was introduced during the Bronze Age by Celtic tribes2, our results support a hypothesis7 that it corresponds to the Atinian elm, which was used for vine-training by the Romans. In his treaty De Re Rustica8 (written in about AD 50), the Roman agronomist Columella advocates the use of elm for this purpose, recommending in particular a barren tree that was vegetatively propagated — the Atinian elm.

Columella owned three farms in Latium9, where most Italian lineage-C samples are found, and a fourth vineyard in Xerez10 (Andalusia, Spain). He and other farmers may have introduced different Italian elms to the Iberian peninsula, including the Atinian elm. Columella's writings influenced the subsequent establishment of vineyards to such an extent that the Roman emperor Domitian prohibited the plantation of new vines in Italy in AD 92 and ordered half of the vineyards in the provinces to be cut down11. In the meantime, the Atinian clone spread across Iberia and was probably transported to Britain in the form of root suckers, as indicated by the presence of elm pollen in a Roman vineyard in Britain12 and by the coincident distribution of suspected Roman vineyards12 and U. procera in Britain2.

The identification of the English elm with the Atinian clone was first proposed in the nineteenth century7. Our findings provide molecular support for this proposal and indicate that the English elm originated from the massive propagation of the Atinian clone by the Romans. This large-scale transformation of the elm's natural diversity became critical in the twentieth century, when most English elms succumbed to Dutch elm disease1, and should be taken into account in current European elm breeding and conservation strategies.