Sudden replacement of cave bear mitochondrial DNA in the late Pleistocene

In the absence of interaction with genetically distinct populations, changes in frequencies of distinct mitochondrial DNA (mtDNA) sequences (haplotypes) within a population are caused by mutations and genetic drift, both of which are comparatively slow processes. By contrast, interactions between populations may lead to rapid changes in haplotype frequencies [1–3], for instance through extinction and recolonialisation. So far, the only case of a direct replacement of mtDNA sequences within a continuous population has been described in historical mouse populations [2]. To investigate the genetic stability of a Pleistocene population over time, we have investigated the mtDNA haplotype spectrum of cave bears from a location in southern Germany, continuously inhabited for thousands of years. We extracted 29 cave bear teeth from three geographically close caves from the Ach Valley (Supplemental data), 20 of which yielded amplifiable mtDNA. Altogether, we found five mtDNA haplotypes, four of which are closely related, differing by one to three substitutions from each other, whereas the fifth one differs from the other four by ten to thirteen substitutions (Supplemental data). All samples that yielded DNA were radiocarbon dated by accelerator mass spectrometry from purified bone collagen, either at BETA Analytic Inc. (17 samples) or at the Oxford dating unit (three samples). The dates range from 38,000 years before present (B.P.) to ~25,500 years B.P. (Figure 1; Supplemental data). Interestingly, the four closely related haplotypes all originate from bears dated to between 28,000 and 38,000 years, whereas the fifth, highly divergent haplotype was found exclusively in bears about 28,000 years old or younger (Figure 1; Supplemental data). Due to the confidence intervals of radiocarbon dates, it is not possible to decide from these ages whether there was any temporal overlap between the two types of sequences. However, while the ages of several samples belonging to the different haplogroups overlap when standard deviations are taken into account (Supplemental data), only one of the bears carrying the single divergent haplotype shows an older point estimated carbon age (by 30 years) than the youngest one carrying one of the other four haplotypes. To investigate whether the observed pattern could also be detected when the different haplotypes were randomly distributed across time within the cave bear population we performed randomization simulations. The results show that it is highly unlikely (p < 0.001) to obtain the observed pattern if the haplotypes were in reality randomly distributed across time. Although we cannot exclude that the minority haplotypes existed at low frequencies outside the time range within which we could detect them (Supplemental data), we think that the most likely explanation for the observed pattern (Figure 1) is a replacement of one group of sequences by the second. In previous studies of ancient mtDNAs spanning several thousand years [3–6], genetic continuity [5], introgression [3] and in one case also a mtDNA sequence replacement have been observed [6]. However, in the latter case, this was accompanied by local extinction of the species in question for several thousand years. Our data thus document the first case of a Pleistocene mtDNA sequence replacement within a population without evidence for temporary extinction. This result differs strongly from three other cave bear populations studied to date where genetic stability over 15,000 to 20,000 years has been observed [5]. We estimated the most recent common ancestor (MRCA) of the four closely related haplotypes using a Bayesian approach [7], resulting in an age of 130,000 years (95% credibility interval 40,000 to 390,000 years). As each of these four haplotypes differs from the closest different one by a single substitution (Supplemental data), none of the haplotypes has been lost since they shared a common ancestor. Thus, this cave bear population