4.1 Species diversification
This study represents the first phylogenetic analysis of the genusToxicodryas . Phylogenetic analyses of our two-locus Sanger data set and 2848-locus RADseq SNP data set reveal two deeply divergent, strongly supported lineages in T. blandingii and three inT. pulverulenta (Fig. 2; Fig. S2). Although today, the two recognized species are broadly sympatric, clades within each species are generally situated allopatrically across river barriers. The two clades within T. blandingii are separated either by the Sanaga River in Cameroon or the Congo River in the DRC. Both rivers have frequently been interpreted as population barriers in other terrestrial vertebrates (Blackburn, 2008; Jongsma et al., 2018; Leaché et al., 2019; Leaché & Fujita, 2010; Portik et al., 2017), but additional sampling and comparative analyses will be needed to determine which river played the most deterministic role in shaping genetic structure in this species. Of the three T. pulverulenta clades, one is distributed in West Africa (albeit with limited sampling) and two are distributed in Central Africa, separated by the western Congo River. Our population structure analyses are concordant with phylogenetic analyses supporting five distinct genetic clusters (Fig. 4). Minor levels of admixture appear to have occurred between the T. pulverulenta clades separated by the western Congo River, and between the two clades of T. blandingiiin the sample collected at the Sanaga River (Fig. 4). In both species, the Congo River barrier seems to be stronger in the west where the river is wider, and the current is stronger. In the eastern DRC samples of clades from both species can be found on either side of this river (Fig. 3)
Divergence time estimates from a time-calibrated phylogeny also fail to reject predictions derived from the river-barrier hypothesis.Toxicodryas blandingii and T. pulverulenta diverged in the early to mid-Miocene, and subsequent intraspecific diversification took place in the late Miocene to the Pliocene (Fig. 2). The Congo River, a barrier in the Central African T. pulverulenta (divergence time: ~4.1 Mya), and a potential barrier in T. blandingii (divergence time: ~8.6 Mya), dates back to the mid-late Miocene (Flügel et al., 2015; Stankiewicz & de Wit, 2006). The Sanaga River, another potential barrier in T. blandingii , has a poorly known geological history, but likely dates back to the formation of the Adamawa Plateau in the late Eocene-early Oligocene (Fagny et al., 2016). Similar mid to late Miocene divergence times have been noted for other widespread Central and West African taxa including frogs (Bell et al., 2017; Jongsma et al., 2018; Zimkus et al., 2017), and terrestrial snakes (Portillo et al., 2019), and similar West to Central African distribution splits have been seen in forest cobras (Wüster et al., 2018), frogs (Leaché et al., 2019), lizards (Allen et al., 2019), and shrews (Jacquet et al., 2015). The Congo river has been a well-known barrier to many species including primates (Harcourt & Wood, 2012; Mitchell et al., 2015; Telfer et al., 2003), shrews (Jacquet et al., 2015), and frogs (Charles et al., 2018). However, while the timing and locations of population divergences in this study correspond with river barriers, the Miocene was also a time of global climatic change characterized by dramatic cooling and vegetation shifts throughout sub-Saharan Africa (Herbert et al., 2016; Jacobs, 2004; Menegon et al., 2014). Although most research surrounding the role of refugia in driving diversification has focused on the dramatic climate oscillations of the Pleistocene, it is likely that refugia are able to form during any period of climatic change (Haffer, 1997; Hampe & Jump, 2011; Jansson & Dynesius, 2002), but the role of possible older refugia has received little attention in the literature (Hampe & Jump, 2011).
Migration analyses support the western Congo River and the Dahomey Gap as barriers to gene flow in the genus Toxicodryas (Fig. 5a). The Dahomey gap is a natural savanna region in West Africa that separates the upper and lower Guinean rainforests, and which has been previously identified as a dispersal barrier for arboreal species (e.g. Rödel, Emmrich, Penner, Schmitz, & Barej, 2014; Schunke & Hutterer, 2005). Both areas also support lower genetic diversity in Toxicodryasthan expected under a pure isolation-by-distance model (Fig. 5b), emphasizing the biological reality of this barrier for forest-associated, primarily arboreal vertebrates despite the fact that both Toxicodryas species have been found in forest patches within the Dahomey gap. Our demographic analyses further suggest that riverine dispersal barriers between clades are strong, indicating divergence without gene flow between the two T. blandingii clades and divergence with minor gene flow across the Congo River in the two Central African T. pulverulenta clades (Fig. 6). Contemporary gene flow was ruled out with high confidence in both species (Table S3). In light of the Miocene divergence times and lack of gene flow between these five clades, it is likely that they represent distinct evolutionary lineages and, thus, surveys of morphological data and analyses of phenotypic variation are underway to determine if formal taxonomic revision is justified.