A rangewide phylogeography of Hermann's tortoise, Testudo hermanni (Reptilia: Testudines: Testudinidae): implications for taxonomy

Abstract

Hermann's tortoise (Testudo hermanni), the best‐known western Palaearctic tortoise species, has a rare natural distribution pattern comprising the Mediterranean areas of the Iberian, Apennine, and Balkan Peninsulas, as well as Sicily, Corsica and Sardinia. The western part of this range is traditionally considered habitat for T. h. hermanni, while T. h. boettgeri occurs in the Balkans. Taxonomy of this tortoise has been challenged in recent years, with the two subspecies being considered full species and the central Dalmatian populations of T. h. boettgeri being considered a third species, T. hercegovinensis. Using an mtDNA fragment approximately 1150 bp long (cytochrome b gene and adjacent portion of tRNA‐Thr gene), we investigated mtDNA diversity with regard to contrasting concepts of two subspecies or three species. Seven closely related haplotypes were identified from the western Mediterranean and 15 different, in part much‐differentiated, haplotypes from the Balkans. Western Mediterranean haplotypes differ from Balkan haplotypes in 16–42 mutation steps. One to seven mutation steps occur within western Mediterranean populations. Balkan haplotypes, differing in 1−37 nucleotides, group in parsimony network analysis into three major assemblages that display, in part, a similar degree of differentiation to that of western Mediterranean haplotypes relative to Balkan haplotypes. Rates of sequence evolution are different in both regions, and low divergence, palaeogeography and the fossil record suggest a slower molecular clock in the western Mediterranean. While monophyly in western Mediterranean haplotypes is well‐supported, conflicting evidence is obtained for Balkan haplotypes; maximum parsimony supports monophyly of Balkan haplotypes, but other phylogenetic analyses (Bayesian, ML, ME) indicate Balkan haplotypes could be paraphyletic with respect to the western Mediterranean clade. These results imply a process of differentiation not yet complete despite allopatry in the western Mediterranean and the Balkans, and suggest all populations of T. hermanni are conspecific. In the western Mediterranean no clear geographical pattern in haplotype distribution is found. Distribution of Balkan haplotypes is more structured. One group of similar haplotypes occurs in the eastern Balkans (Bulgaria, Republic of Macedonia, Romania and the Greek regions Evvia, Macedonia, Peloponnese, Thessaly and Thrace). Two distinct haplotypes, differing in eight to nine mutation steps from the most common haplotype of the first group, are confined to the western slope of the Taygetos Mts. in the Peloponnese. Yet another group, connected over between four and 23 mutation steps with haplotypes of the eastern Balkan group, occurs along the western slope of the Dinarid and Pindos Mts. (Istria, Dalmatia, western Greece). Taygetos haplotypes are nested within other haplotypes in all phylogenetic analyses and support for monophyly of the other Balkan groups is at best weak. We conclude that using the traditional two subspecies model should be continued for T. hermanni. Phylogeographies of T. hermanni and Emys orbicularis, another codistributed chelonian, are markedly different, but share a few similarities. Both were forced to retreat to southern refuges during Pleistocene glaciations. With the advent of Holocene warming, E. orbicularis underwent rapid range expansion and temperate regions of Europe and adjacent Asia were recolonized from refuges in the Balkans and the northern Black Sea Region. By contrast, T. hermanni remained more or less confined to refuges and nearby regions, resulting in a much smaller range, and allopatric and parapatric distribution of haplotype groups and clades. MtDNA lineages are more diverse in E. orbicularis than they are in T. hermanni on southern European peninsulas, indicating several distinct glacial refuges in close proximity and extensive intergradation during Holocene range expansion for E. orbicularis. In T. hermanni it is likely that only on the Balkan Peninsula was more than one refuge located, corresponding to the parapatric ranges of haplotype groups currently there. On the old western Mediterranean islands Corsica and Sardinia no differentiated (E. orbicularis) or only weakly differentiated haplotypes (T. hermanni) occur, even though there is evidence for the presence of both species on Corsica since at least the Middle Pleistocene. High mountain chains constitute major barriers separating distinct mtDNA clades or groups in each species.