Dasypus, Linnaeus, 1758

DASYPUS View in CoL PHYLOGENY AND SPECIES LIMITS

Our phylogeny provides three important advances compared to the previous molecular phylogeny ( Gibb et al., 2016), mainly due to our more comprehensive geographic sampling. First, we confirm that the nine-banded armadillo D. novemcinctus from most of South America is a distinct lineage from populations from French Guyana and from North and Central America. Based on the phylogenetic position, genetic divergence and the species delimitation analyses, Dasypus specimens from French Guyana should be recognized as a species distinct from D. novemcinctus , as was also suggested by Gibb et al. (2016). Guiana Shield specimens are known to exhibit pronounced molecular ( Huchon et al., 1999; Gibb et al., 2016) and morphological ( Billet et al., 2017; Hautier et al., 2017; Feijó et al., 2018) disparity from other D. novemcinctus populations. On the other hand, the relationships of D. novemcinctus from North and Central America are still uncertain. Because the type locality of D. novemcinctus lies in Pernambuco, Brazil, specimens from South America, excluding those from the Guiana Shield, should keep that name (for nomenclatural discussion, see: Feijó et al., 2018). Therefore, Central and North American populations could be (1) D. novemcintus based on monophyly, gene flow between Central and South America ( Arteaga et al., 2012) and the lack of consistent morphological distinctions ( Feijó et al., 2018); or (2) a distinct and cryptic species based on the high intraspecific divergence (~ 2.9%), paranasal sinus morphology ( Billet et al., 2017) and 3D skull shape ( Hautier et al., 2017).

Despite those advances, the monophyly of Dasypus novemcinctus remains uncertain, because D. mazzai and D. sabanicola are found nested among South American specimens of D. novemcinctus . In the Gibb et al. (2016) phylogeny, D. mazzai and D. sabanicola were sister to D. novemcinctus from the United States and D. pilosus . Our phylogeny recovered D. pilosus as the sister lineage to a clade that includes D. novemcinctus (non-Guiana specimens), D. mazzai and D. sabanicola . This conflict might be due to the shorter gene sequences used in our analysis compared to the complete mitogenomes of Gibb et al. (2016). However, it seems more likely attributable to the poorer sampling of intraspecific variability and geographic coverage in their analysis, which might result in inaccurate phylogenetic inference ( Wiens & Servedio, 1997; Lohse, 2009; Bergsten et al., 2012; Lim et al., 2012; Avendano et al., 2017). Samples from both D. mazzai and D. sabanicola used by Gibb et al. (2016) clustered with geographically distant samples of D. novemcinctus . Dasypus mazzai from Argentina clustered with D. novemcinctus from Paraíba state, north-eastern Brazil, whereas D. sabanicola from the savannas of Venezuela clustered with D. novemcinctus from Amazonia of western Brazil. Recent speciation, a limited number of informative sites, ancestral polymorphism, hybridization and introgression, all common attributes of young species ( Kekkonen et al., 2015; Gibb et al., 2016), might have generated those patterns. Despite the inclusion of D. mazzai and D. sabanicola within D. novemcinctus and their conspicuous morphometric overlap, three datasets suggest their distinctness. Feijó et al. (2018) found consistent diagnostic carapace traits for the two taxa, Abba et al. (2018) showed distinctions in 212 bp rRNA between the holotype of D. mazzai and both D. novemcinctus and D. septemcinctus hybridus , and, finally, the widely disjunct geographic distributions of D. sabanicola and D. mazzai precludes any gene flow. Thus, we advocate recognizing these two taxa as distinct species until there is further evidence from faster-evolving genes and increased sample size for both species.

A second main finding of this work is genetic support for the three species of the Dasypus kappleri complex. Phylogenetic trees provide well-supported reciprocal monophyly for D. kappleri , D. beniensis and D. pastasae . Tree-based delimitation (GYMC and PTP) and recovery analyses also recovered the three species. In contrast, the ABGD clustering method recognized D. beniensis , the older species with a Miocene/ Pliocene origin and D. pastasae + D. kappleri , which diverged during the Pleistocene. Previous studies have demonstrated the impaired performance of ABGD relative to tree-based models, such as PTP and GMYC, in situations with poor intraspecific sampling ( Puillandre et al., 2012; Talavera et al., 2013; Zhang et al., 2013; Kekkonen et al., 2015).

The third major finding of this work is to reject the reciprocal monophyly of D. septemcinctus and D. hybridus on molecular grounds. All species delimitation analyses failed to recover any partition in D. septemcinctus (including hybridus ). Feijó et al. (2018) synonymized D.hybridus with D.septemcinctus due to a lack of clear diagnostic characteristics but maintained its subspecific status based on partial diagnosability by morphometric traits and biogeographical distribution. Gibb et al. (2016) reported almost identical mitogenome sequences involving one specimen from Argentina (coded in our tree as DhybAR) and one from Uruguay (DhybUR). Both specimens were clustered together with a third sequence from Rio Grande do Sul, southern Brazil, and this clade is separated from a clade with specimens from central and northern Brazil in our concatenated mitochondrial and nuclear tree (Supporting Information, Fig. S1 View Figure 1 ). Following Feijó et al. (2018), the former clade represents the subspecies D. septemcinctus hybridus distributed in southern Brazil, Uruguay and eastern, central and northern Argentina, and the second clade represents D. s. septemcinctus , occurring in Brazil, eastern Paraguay, Bolivia and northern Argentina.