Is it inappropriate to ask for your age? Evaluating parameter impact on tree dating in a challenging clade (Macroscelidea)

Abstract

Sengis (order Macroscelidea) are small mammals endemic to Africa. The taxonomy and phylogeny of sengis has been difficult to resolve due to a lack of clear morphological apomorphies. Molecular phylogenies have already significantly revised sengi systematics, but until now no molecular phylogeny has included all 20 extant species. In addition, the age of origin of the sengi crown clade and the divergence age of its two extant families remain unclear. Two recently published studies based on different datasets and age-calibration parameters (DNA type, outgroup selection, fossil calibration points) proposed highly different divergent age estimates and evolutionary scenarios. We obtained nuclear and mitochondrial DNA from mainly museum specimens using target enrichment of single-stranded DNA libraries to generate the first phylogeny of all extant macroscelidean species. We then explored the effects of different parameters (type of DNA, ratio of ingroup to outgroup sampling, number and type of fossil calibration points) and their resulting impacts on age estimates for the origin and initial diversification of Macroscelidea. We show that, even after correcting for substitution saturation, both using mitochondrial DNA in conjunction with nuclear DNA or alone results in much older ages and different branch lengths than when using nuclear DNA alone. We further show that the former effect can be attributed to insufficient amounts of nuclear data. If multiple calibration points are included, the age of the sengi crown group fossil prior has minimal impact on the estimated time frame of sengi evolution. In contrast, the inclusion or exclusion of outgroup fossil priors has a major effect on the resulting node ages. We also find that a reduced sampling of ingroup species does not significantly affect overall age estimates and that terminal specific substitution rates can serve as a means to evaluate the biological likeliness of the produced temporal estimates. Our study demonstrates how commonly varied parameters in temporal calibration of phylogenies affect age estimates. Dated phylogenies should therefore always be seen in the context of the dataset which was used to produce them.