1. INTRODUCTION
For more than two hundred years, scientists have pondered the most pervasive pattern in biogeography, the latitudinal diversity gradient, a striking pattern of increasing species diversity from the poles toward the equator (Allen & Gillooly, 2006; Darlington, 1957; Darwin, 1859; Hutchinson, 1959; Jablonski, Roy, & Valentine, 2006; Pianka, 1966; Ricklefs & Schluter, 1993; Rohde, 1992; Rosenzweig, 1995; Wallace, 1878). Potential explanations of this pattern have been provided from diverse perspectives, including that tropical regions have a wider array of niches (Buckley et al., 2010; Lamanna et al., 2014; Stevens, 2011), higher primary productivity (Hawkins, Porter, Diniz-Filho, & Alexandre, 2003; Jetz & Fine, 2012; Whittaker, Nogués‐Bravo, & Araújo, 2007), higher environmental heterogeneity (Janzen, 1967; Stein, Gerstner, & Kreft, 2014; Stevens, 1989), greater land area (Fine & Ree, 2006; Rosenzweig, 1995; Terborgh, 1973), and higher climatic stability (Harrison & Noss, 2017; Hawkins, Diniz-Filho, Jaramillo, & Soeller, 2007). In more recent studies of historical tropical biogeography, researchers have focused on speciation processes in major tropical rainforests to explain their high biodiversity (e.g. Cardillo, Orme & Owens, 2005; Jablonski et al., 2006; Ricklefs, 2006; Smith et al., 2017; Weir & Schluter, 2007; Wiens & Donoghue, 2004; Wiens, Sukumaran, Pyron, & Brown, 2009).
Three major allopatric diversification mechanisms have been proposed in the classical literature to explain species diversity in the Amazon: the “river hypothesis” in which species and populations diverged across river barriers (Ayres & Clutton-Brock, 1992; Bates, 1863; Hershkovitz, 1977; Mayr, 1942; Sick, 1967; Wallace, 1853); the “refuge hypothesis” in which forests fragmented during Earth’s cold or dry climate cycles (i.e. the Pleistocene glaciation cycles), causing isolation and divergence in small forest patches (Haffer, 1969, 1974, 1982; Prance, 1982; Vanzolini, 1973; Vanzolini & Williams, 1970); and an amalgamate “river-refuge hypothesis” in which speciation was promoted by a combination of river barriers and climate driven vegetation changes (Ayres & Clutton-Brock, 1992; Haffer 1992, 1993). These hypotheses have been widely used in the study of Neotropical biodiversity and the mechanisms of its production (e.g. Gascon et al., 2000; Haffer, 2008; Patton & Silva, 2005; Richardson, Pennington, Pennington, & Hollingsworth, 2001; Weir, 2006). However, because the early scientific focus was primarily on the Amazon (Amorim, 1991; Cracraft, 1985; DeMenocal, 2004; Haffer, 1969, 1997; Plana, 2004; but see Fjeldså, 1994; Mayr & O’Hara, 1986), and given political instability in tropical Africa (Greenbaum, 2017; Siddig, 2019; Tolley et al. 2016), rigorous testing of the predictions stemming from these hypotheses has been neglected for the West and Central African rainforests until only recently.
Based on pollen core records (Brenac, 1988; Bonnefille & Riollet, 1988; Girese, Maley, & Brenac, 1994; Maley, 1987, 1989, 1991; Maley & Brenac, 1987; Maley & Livingstone, 1983; Sowunmi, 1991) and species distribution data (Colyn, 1987, 1991; Rietkerk, Ketner, & De Wilde, 1995; Richards, 1963; Sosef, 1991), Maley (1996) proposed several rainforest refugia for sub-Saharan Africa that are still widely used today (e.g. Bell et al., 2017; Hughes, Kusamba, Behangana, & Greenbaum, 2017; Huntley, Castellanos, Musher, & Voelker, 2019; Jongsma et al., 2018; Larson, Castro, Behangana, & Greenbaum, 2016; Penner, Wegmann, Hillers, Schmidt & Rödel, 2011, Portik et al. 2017; Fig. 1). Many of these hypothesized refugia are located in highland areas (e.g., the Cameroon Volcanic Line and the Albertine Rift), however, a major fluvial refuge, located in the gallery forests around the Congo River, has been supported by pollen core data (Maley, 1996), and distributional patterns of multiple bird (Huntley, Harvey, Pavia, Boano, & Voelker, 2018; Levinsky et al., 2013), mammal (Colyn, Gautier-Hion, & Verheyen, 1991; Levinsky et al., 2013) and plant taxa (Robbrecht, 1996).
Major river barriers in West and Central Africa include the Volta, the Sanaga, the Ogooué, the Congo, the Niger and the Cross Rivers (Fig. 1). The exact ages of many of these rivers are unknown but are generally estimated to date back to the Late Mesozoic to the Early Cenozoic (80–35 mya; Goudie, 2005; Stankiewicz & de Wit, 2006). However, while the Congo basin is quite old (Flügel, Eckardt, & Cotterill, 2015; Stankiewicz & de Wit, 2006), the present course of the Congo River appears to be much younger, dating to the mid to late Miocene and corresponding to the uplift of the East African Rift (Flügel, et al., 2015; Stankiewicz & de Wit, 2006).
Numerous phylogeographic studies have supported the importance of rivers, refugia, or both as drivers of diversification across disparate plant and animal species. Rivers alone have been shown to be important barriers for some species of primates (Mitchell et al., 2015; Telfer et al., 2003), shrews (Jacquet et al., 2015), and frogs (Charles et al., 2018; Penner et al. 2011; Penner, Augustin & Rödel, 2019; Wieczorek, Drews & Channing, 2000; Zimkus, Hillers & Rödel, 2010), but do not appear to represent an important barrier for many plant species (Dauby et al., 2014; Debout, Doucet, & Hardy, 2011; Hardy et al., 2013; Ley et al., 2014; Lowe, Harris, Dormontt, & Dawson, 2010). Refugia are suggested to have played an important role in the diversification of rodents (Bohoussou et al., 2015; Nicolas et al., 2011; Nicolas, Missoup, Colyn, Cruaud, & Denys, 2012), primates (Clifford et al., 2004; Haus et al., 2013; Tosi, 2008), frogs (Bell et al., 2017; Jongsma et al., 2018), lizards (Allen, Tapondjou, Greenbaum, Welton, & Bauer, 2019; Leaché et al., 2017), birds (Fjeldså & Bowie, 2008), pangolins (Gaubert et al., 2016), and rainforest plants (Born et al., 2011; Budde, González-Martínez, Hardy, & Heuertz, 2013; Daïnou et al., 2010; Dauby, Duminil, Heuertz, & Hardy, 2010; Duminil et al., 2015; Faye et al., 2016; Gomez et al., 2009; Hardy et al., 2013; Ley et al., 2014; Ley, Heuertz, & Hardy, 2016; Lowe et al., 2010). In some cases, divergence patterns match both refugial and riverine predictions (Anthony et al., 2007; Barej et al., 2011; Bohoussou et al., 2015; Gonder et al., 2011; Jacquet et al., 2014; Jongsma et al., 2018; Leaché et al., 2019; Leaché & Fujita, 2010; Marks, 2010; Portik et al., 2017), suggesting that both may have played roles simultaneously—or in combination—in evolutionary diversification. However, because of the spatial overlap of refugia with montane and riverine systems (Hofer, Bersier, & Borcard, 1999, 2000), and the sparse pollen core and fossil records for the tropics (Colinvaux, De Oliveira, Moreno, Miller, & Bush, 1996; Maley & Brenac, 1998), distinguishing between these three hypotheses has been difficult, especially when relying on phylogeographic data alone.
The three major allopatric diversification hypotheses make the following predictions regarding species diversification patterns in tropical African forests (1) river hypothesis: boundaries between population distributions should correspond to riverine barriers and the ages of populations should be relatively old, corresponding to the ages of the rivers; (2) refuge hypothesis: population distributions should be concordant with locations of hypothesized rainforest refugia during cold, dry periods and populations are predicted to be relatively young, possibly corresponding to the Pleistocene glaciation cycles; (3) river-refugia hypothesis: population distributions should be correlated with the locations of rainforest refugia and bounded by rivers barriers, or will have been confined to refugial locations and additionally subdivided by rivers. Finally, the timing of population splits should correspond to ages of rivers but would be expected to show patterns of expansion and contraction dating to the Pleistocene.
In this study, we use the snake genus Toxicodryas as a model system to test the predictions of these hypotheses. The genusToxicodryas consists of two large, rear-fanged, venomous West and Central African species, T. blandingii and T. pulverulenta. The taxonomic placement of this genus is uncertain. They were originally placed in the Asian genus Boiga (Schmidt, 1923), and some authors still classify them as such, but recent phylogenetic analyses recover them as the sister genus to the African egg eating snakes, Dasypeltis , albeit with weak support (Pyron et al., 2013). Both species in this genus are primarily arboreal, feeding mainly on birds, bats, frogs and chameleons (Akani, Barieenee, & Luiselli, 1998; Chippaux & Jackson, 2019; Nagy et al., 2011; Spawls, Howell, Hinkel, & Menegon, 2018). Because of their general arboreality, these species are predicted to have distributions strongly correlated with forest distribution. In addition, Toxicodryas is widely distributed within the Congo River fluvial system and broadly across West and Central Africa, making this genus a suitable system for testing the competing predictions of the river, refugia and river-refugia hypotheses.
Recent advances in paleo-climate modeling and genome-scale DNA sequencing have opened new avenues to testing classic hypotheses of tropical rainforest speciation (Bell et al., 2017; Leaché et al., 2019; Portik et al., 2017). In this study, we integrate dated phylogeographic inference, population structure analyses, and machine learning-based demographic modeling to identify the timing of divergence as well as the location and permeability of past and present dispersal barriers. These genetic data are combined with paleo-distribution and climate stability modeling to determine the congruence of historical distributions with the refugial and river-refugial hypotheses. Our results demonstrate that, although population distributions alone could be congruent with any of the three hypotheses, diversification times predate the Pleistocene, a finding that aligns with predictions of the river-barrier hypothesis. Historical demographic analyses support models of no migration among populations since the time of divergence, and migration analyses suggest that the western Congo River represents one of the strongest barriers to recent dispersal. Species paleo-distribution and climate stability modelling show no suggestion of suitable habitat contraction during or since the Pleistocene, allowing us to soundly reject the predictions of refugia hypotheses in favor of the prevailing role of riverine barriers in shaping, structuring, and maintaining diversity in this generally arboreal, forest-associated group of endemic African snakes.