Assembly and alignment
Mitogenome amplicon reads were quality filtered with Trimmomatic v.0.33 (Bolger, Lohse, and Usadel 2014) with parameters SLIDINGWINDOW: 4:15 and MINLEN: 36. Since the only publicly available mitogenome representing any Tupaia species (the northern treeshrew, T. belangeriNC_002521; Schmitz, Ohme, & Zischler, 2000) is highly divergent from our study species (Roberts et al., 2011), we first generated reference mitogenomes for the mountain treeshrew and 3 more closely related outgroup species: the pygmy treeshrew, large treeshrew, and ruddy treeshrew. For each species, we selected one individual with the highest number of sequencing reads (pygmy treeshrew, BOR 443; large treeeshrew, BOR 010, & ruddy treeshrew, UMMZ174429) and assembled sequencesde novo using the MIRA v1.0.1 plugin in Geneious v9.1.2 (Biomatters Ltd.). Quality filtered sequence reads were mapped to the appropriate reference using BWA-MEM v0.7.10 (Li 2013) with default parameters. We also assembled mitogenomes from UCE-enriched library sequences (Supplemental Information). Consensus sequences were generated with Geneious (lowest coverage to call a base 5×, and Highest Total Quality parameters) and aligned with the MAFFT v7.450 plugin (Katoh et al. 2002). We transferred annotations from the northern treeshrew reference to the consensus sequences. To rule out the presence of nuclear copies of mitochondrial genes (NUMTs), we translated all protein-coding genes to check for frame shifts or stop codons.