Once lost, twice found: Combined analysis of ancient giant panda sequences characterises extinct clade

This year, two studies have reported ancient DNA sequences from giant pandas (Ailuropoda melanoleuca). We present the combined analysis of these ancient panda sequences to reveal a lost clade which survived through the Last Glacial Maximum and became extinct sometime subsequent to the Middle Holocene. This result indicates a major loss of mitochondrial diversity occurring during the recent evolutionary history of pandas. The giant panda is an iconic species that currently occurs in fragmented populations across six mountain ranges on the eastern margin of the Tibetan Plateau (Figure 1). However, this represents a tiny fraction of their distribution during the Pleistocene, when they occurred across most of southern China as well as adjacent parts of Myanmar, Thailand, and Vietnam (Jablonski et al., 2012). Although pandas have frequently been the subject of genetic studies (Li et al., 2010; Zhao et al., 2012), until very recently, ancient DNA sequences have proved elusive (Basler et al., 2017), potentially reflecting advanced DNA degradation typical of ancient bones deposited in tropical and subtropical environments (Kehlmaier et al., 2017; Min‐Shan Ko et al., 2018). This year, a study by Sheng et al. (2018) reported the successful recovery of around 5,000 bp of mitochondrial sequence from one 8,470 ± 45‐year‐old individual and another 5,025 ± 35‐year‐old individual from a natural sinkhole in Jiangdongshan, Tengchong County, Yunnan Province, southwestern China, which is outside the current distribution of pandas. This study found the c. 5,000‐year‐old haplotype to be nested within the mitochondrial diversity of extant pandas, whereas the c. 8,500‐ year‐old haplotype formed a divergent sister lineage to this clade. A second study by Min‐Shan Ko et al. (2018) reported the successful recovery of the mitochondrial genome of a 21,703 ± 208‐year‐ old individual from Cizhutuo cave in Guangxi Province, southern China, also outside the current distribution of pandas. This study also found the ancient haplotype to form a divergent lineage that is sister to all sampled extant panda haplotypes. We investigated the interrelationships of these three ancient haplotypes. We cropped the c. 22,000 year mitochondrial genome to c. 5,000 bp overlapping the two Jiangdongshan sequences and replicated the analysis described in Sheng et al. (2018) using BEAST 1.8.2 (Drummond, Suchard, Xie, & Rambaut, 2012) (Supporting Information). We constrained the coalescence time of extant panda haplotypes using a normal prior with a mean of 72,000 years and SD 10,000 years based on the results of Min‐Shan Ko et al. (2018). The substitution rate of the relaxed clock model was estimated within an open, uniform prior of 0%–20% per million years based on this node calibration and the radiocarbon dates of the ancient samples. Consistent with previous studies (Min‐Shan Ko et al., 2018; Sheng et al., 2018), we did not find evidence of strong mitochondrial phylogeographic structuring in modern pandas. Nuclear genomic studies do, however, support a primary structuring of living pandas into three populations according to mountain range, comprising a Qinling Mountains population; a Minshan Mountains population; and a third population encompassing the Qionglai, Daxiangling, Xiaoxiangling, and Liangshan Mountains (Zhao et al., 2012). Our results do suggest some congruent patterns of mitochondrial diversity, since the northeastern Qinling population and the two western populations combined each possesses some exclusive mitochondrial lineages established prior to 10,000 years ago (Figure 1). Given the overall lack of reciprocal monophyly among populations, and several examples of more recent coalescence of haplotypes between populations, the mitochondrial phylogeography is consistent with incomplete post‐divergence lineage sorting and/or recent maternal gene flow. DOI: 10.1111/jbi.13486