Journal of Biogeography (J. Biogeogr.) (2017) 44, 182–194 ORIGINAL ARTICLE Evolutionary history of the thicket rats (genus Grammomys) mirrors the evolution of African forests since late Miocene Josef Bryja1,2*, Radim
Sumbera3, Julian C. Kerbis Peterhans4,5, Tatiana Aghova1,2, Anna Bryjova1, Ond
rej Mikula1,6, Violaine Nicolas7, Christiane Denys7 and Erik Verheyen8,9 1 Institute of Vertebrate Biology of the Czech Academy of Sciences, 603 65 Brno, Czech Republic, 2Department of Botany and Zoology, Faculty of Science, Masaryk University, 611 37 Brno, Czech Republic, 3 Department of Zoology, Faculty of Science,
e University of South Bohemia, 370 05 Cesk Bud
ejovice, Czech Republic, 4College of Professional Studies, Roosevelt University, 60605 Chicago, IL, USA, 5Field Museum of Natural History, 60605 Chicago, IL, USA, 6 Institute of Animal Physiology and Genetics of the Czech Academy of Sciences, 602 00 Brno, Czech Republic, 7Institute of Systematics and Evolution, UMR7205 CNRS-MNHN-EPHESorbonne Universites, 75005 Paris, France, 8 Royal Belgian Institute for Natural Sciences, Operational Direction Taxonomy and Phylogeny, 1000 Brussels, Belgium, 9 Evolutionary Ecology Group, Biology Department, University of Antwerp, 2020 Antwerp, Belgium *Correspondence: Josef Bryja, Institute of Vertebrate Biology of the Czech Academy of Sciences, Research Facility Studenec, Studenec 122, 675 02 Kon
e
sın, Czech Republic. E-mail: bryja@brno.cas.cz ABSTRACT Aim Grammomys are mostly arboreal rodents occurring in forests, woodlands and thickets throughout sub-Saharan Africa. We investigated whether the divergence events within the genus follow the existing evolutionary scenario for the development of African forests since the late Miocene. Location Sub-Saharan African forests and woodlands. Methods We inferred the molecular phylogeny of Grammomys using Bayesian and maximum likelihood methods and DNA sequences of 351 specimens collected from across the distribution of the genus. We mapped the genetic diversity, estimated the divergence times by a relaxed clock model and compared evolution of the genus with forest history. Results Phylogenetic analysis confirms the monophyly of Grammomys and reveals five main Grammomys lineages with mainly parapatric distributions: (1) the poensis group in Guineo-Congolese forests; (2) the selousi group with a distribution mainly in coastal forests of southern and eastern Africa; (3) the dolichurus group restricted to the easternmost part of South Africa; (4) the macmillani group in the northern part of eastern and Central Africa with one isolated species in Guinean forests; and (5) the surdaster group, widely distributed in eastern Africa south of the equator. Every group contains well supported sublineages suggesting the existence of undescribed species. The earliest split within the genus (groups 1 vs. 2–5) occurred in the late Miocene and coincides with the formation of the Rift Valley which resulted in the east–west division of the initially pan-African forest. The subsequent separation between groups (2 vs. 3–5) also dates to the end of the Miocene and suggests the split between Grammomys from coastal to upland forests in eastern Africa followed by a single dispersal event into western Africa during the Pleistocene. Conclusions The evolutionary history of the genus Grammomys closely reflects the accepted scenario of major historical changes in the distribution of tropical African forests since the late Miocene. Keywords Arvicanthini, coastal forests, late Miocene, lowland forests, mountain forests, phylogeography, Plio-Pleistocene climate changes, Rodentia, tropical Africa INTRODUCTION Tropical forests in Africa contain rich biodiversity. For example, the Eastern Arc Mountains support c. 3300 km2 of 182 http://wileyonlinelibrary.com/journal/jbi doi:10.1111/jbi.12890 forest that harbours 211 endemic or nearly endemic vertebrate species (Rovero et al., 2014), whereas the Albertine Rift mountains host the largest suite of endemic mammals on the continent (Plumptre et al., 2007). However, biological ª 2016 John Wiley & Sons Ltd Evolution of climbing rats and African forests diversity is not equally distributed across the African tropics (e.g. De Klerk et al., 2002), but knowledge of its distribution is crucial in prioritizing conservation activity. A recent study of forest composition in tropical Africa identified six floristic clusters associated with particular environmental conditions (Fayolle et al., 2014; Fig. 1). The origin of these forest types is the outcome of a complex evolutionary history that started from a single continuous equatorial forest that covered sub-Saharan Africa during a period of humid climate of the Early and Middle Miocene (Plana, 2004). By the Late Miocene, tectonic uplift created the Rift Valley and split the pan-African rainforest into the GuineoCongolese forests in western and Central Africa and the forests situated east of the rift. The rift formation combined with declining global temperatures and changes in monsoon winds resulted in an arid climate that caused the disappearance of forests along the slope of the rift mountains, hence creating the so-called arid corridor that periodically connected the northern (Sudanian and Somalian) and southern (Zambezian) savannas (Bobe, 2006). However, some old mountain ranges (e.g. Albertine Rift and Eastern Arc mountains) served as long-term forest refugia allowing the evolution of species-rich communities (e.g. Loader et al., 2014). Throughout this period, West (=Guinean) and Central (=Congolese) African forests continued to exist as a single unit that underwent periodic fragmentation during the Pleistocene (Maley, 1996). Since the Middle Pleistocene, the forested mountain chains in eastern Africa also underwent fragmentation, as suggested by increasing proportions of C4 vegetation, most likely indicating the origin of the current tropical grasslands around these mountains (Cerling, 1992). Based on the concept of phylogenetic niche conservatism (Wiens & Donoghue, 2004), this study proposes to use a phylogeographical approach for forest-dwelling mammals to investigate the evolutionary history and past connections among African forests. Phylogeographical patterns for widely aridulus gazellae minnae (a) bunƟngi macmillani poensis kuru KH dryas ARM caniceps EAM selousi SRM surdaster (b) Coastal East Africa Upland East Africa Dry West Africa Wet-Moist West Africa Moist Central Africa Wet Central Africa cometes surdaster group macmillani group dolichurus group selousi group poensis group dolichurus Figure 1 (a) Distribution of sampled Grammomys specimens in sub-Saharan Africa. The five main genetic groups of Grammomys are represented by different symbols (see key). Black stars show type localities of currently valid species (except G. surdaster, which is considered a junior synonym of G. dolichurus) mentioned in the text. Main mountain blocks mentioned in the text are schematically demarcated by dashed lines: KH = Kenyan Highlands, ARM = Albertine Rift Mountains, EAM = Eastern Arc Mountains, SRM = Southern Rift Mountains. (b) Distribution of main forest types in sub-Saharan Africa. The dots represent localities downloaded from Fayolle et al. (2014). They correspond to the six floristic clusters defined by the analysis of 1175 tree species in 455 sampling sites of tropical African forests. Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd 183 J. Bryja et al. distributed taxa with specific ecological requirements can be used to test alternative hypotheses of African forest evolution. Although an increasing number of studies have used this approach on sub-Saharan vertebrates (e.g. Huntley & Voelker, 2016), so far few studies have targeted widespread taxa living in various forest types (for a rare example, see Couvreur et al., 2008). It is in this context that we have used DNA sequences to infer for the first time the phylogeny of thicket rats of the genus Grammomys. These partly arboreal rodents, belonging to the tribe Arvicanthini (Ducroz et al., 2001; Lecompte et al., 2008; Missoup et al., 2016), occur in a variety of forests and woodlands in sub-Saharan Africa. Although 11–14 Grammomys species are currently recognized, the monophyly of the genus remains uncertain and its taxonomic sampling incomplete (Musser & Carleton, 2005). Because these climbing rats are widely distributed in subSaharan forests and woodlands, they may represent a suitable model group to trace the evolutionary histories of the forested habitats in which they occur. Moreover, the fact that they represent a genus originating during the radiation of Arvicanthini c. 8 Ma (Ducroz et al., 2001) provides an opportunity to study their evolutionary history since the Late Miocene, a crucial era for the development of African forests. Over the past decades, we have collected material of Grammomys rats from a large part of their distribution for molecular sampling. We inferred for the first time the phylogeny of the genus that we used together with estimated divergence dates as a proxy for the evolutionary histories of the different forest types in tropical Africa in which they occur. Lastly, based on observed diversity, we identified the geographical areas and genetic clades in which future taxonomic studies are most likely to result in discovery of new species of Grammomys. MATERIALS AND METHODS Sampling The study is based on 351 specimens of Grammomys genotyped for at least one genetic marker (see Table S1 in Appendix S1 in Supporting Information). The tissue samples were stored in 96% ethanol, DMSO or liquid nitrogen until DNA extraction. All fieldwork complied with legal regulations in the respective African countries, and sampling was carried out in accordance with local legislation (see Acknowledgements). In total, the analysed dataset includes genetic information on specimens collected from 170 localities in 18 African countries (Fig. 1). DNA sequencing We collected the sequences for mitochondrial markers, either the cytochrome b gene (CYTB, 334 new sequences and 11 from GenBank), the 16S rRNA gene (16S, 164 new sequences) or both, for all 351 specimens. For 112 selected 184 specimens, we also obtained sequences of the nuclear gene for interphotoreceptor binding protein (IRBP, 110 new sequences and two from GenBank) to match detected mitochondrial diversity as far as possible with sequences from a nuclear locus (see Table S1 in Appendix S1). Primers and PCR protocols for DNA from fresh material are detailed in Table S1 in Appendix S2. PCR products were Sanger sequenced from both sides in a commercial laboratory. Genetic data obtained from fresh material were complemented by eight museum samples (mostly dry skins) (see Appendix S1) pyrosequenced on GS Junior using the CYTB mini-barcode protocol (Galan et al., 2012). This approach was used for samples from geographical areas that are difficult to access today or from the type localities of G. dryas and G. poensis (see more details in Bryja et al., 2014a). Phylogenetic reconstructions within Grammomys and genetic distances Sequences of CYTB, 16S and IRBP were edited and aligned in SeqScape 2.5 (Applied Biosystems, Foster City, CA, USA), producing final alignments of 1140, 575 and 1261 bp, respectively. We first reconstructed the mitochondrial phylogeny using the concatenated CYTB and 16S dataset, because preliminary separate analyses of these two loci provided very similar topologies (not shown). We performed the final phylogenetic analyses with a reduced mtDNA dataset of 157 specimens (155 sequences of CYTB and 115 of 16S) (see Appendix S1), representing the main mtDNA lineages identified by preliminary analyses (not shown). The remaining 194 specimens (identical and/or shorter sequences from the same or neighbouring localities) were unambiguously assigned to particular lineages by neighbour-joining analysis (bootstrap support > 90%; not shown) in Mega 6.06 (Tamura et al., 2013). These data were used to increase the precision with which we mapped the geographical distribution of phylogenetic clades and assigned type material to particular genetic groups. To assess the monophyly of Grammomys reliably, we used as outgroups 24 mitochondrial sequences of 13 genera within the tribe Arvicanthini (sensu Lecompte et al., 2008), eight sequences of species from other tribes of Murinae and one species of the subfamily Gerbillinae (see Table S2 in Appendix S1). We used PartitionFinder 1.0.1 (Lanfear et al., 2012) to detect partitions and the most suitable substitution models simultaneously. Using the Bayesian information criterion (BIC), the best scheme supported four partitions (see Table S2 in Appendix S2). Mitochondrial phylogeny was analysed by maximum likelihood (ML) and Bayesian inference (BI) approaches. ML analysis was performed using RAxML 8.0 (Stamatakis, 2014). Because simpler models are not available in RAxML, the GTR+G model (option -m GTRGAMMA) was selected for the four partitions (option -q). The robustness of the nodes was evaluated by the default bootstrap procedure with 1000 replications (option -# 1000). Bayesian analysis of evolutionary relationships was performed in MrBayes 3.2.1 (Ronquist Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd Evolution of climbing rats and African forests supported by only one analysis selousi group dolichurus group macmillani group su6 Dated phylogeny of Arvicanthini The ML and BI analyses of the concatenated mitochondrial dataset resulted in different phylogenetic positions for the poensis group (see below). The ML tree suggests that the poensis group represents a separate lineage within Arvicanthini and does not belong to Grammomys. As the basal divergences within this tribe were poorly supported (not shown), we attempted to increase their degree of support by adding more mitochondrial and nuclear sequences. The enhanced dataset contained four mitochondrial (CYTB, COI+COII+ATPase8, 16S, 12S) and five nuclear markers (IRBP, RAG1, GHR, BRCA1, AP5). In total, this multi-locus dataset included 34 species of Arvicanthini (sensu Lecompte et al., 2008) comprising 14 genera. The genus Grammomys was represented by sequences of representatives of the five groups that were identified by the mitochondrial phylogeny. As outgroups, we used representatives of six other tribes of Murinae (see Table S3 in Appendix S1). The total length of the concatenated dataset was 9458 bp with 46% missing data. We performed analyses in RAxML and MrBayes using the partitioned datasets (see Table S2 in Appendix S2) as described above. The same dataset was used to estimate the times to most recent common ancestors (TMRCAs) of the clades that were identified by earlier analyses. We used a relaxed clock model with branch rates drawn from an uncorrelated lognormal distribution in beast 1.8.2 (Drummond et al., 2012). Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd surdaster group BS >75 and PP >0.95 poensis group & Huelsenbeck, 2003). Three heated and one cold chains were employed in a partitioned analysis, and runs were initiated from random trees. Two independent runs were conducted with 5 million generations each, and trees and parameters were sampled every 1000 generations. Convergence was checked using Tracer 1.5 (Rambaut & Drummond, 2007). For each run, the first 25% of sampled trees were discarded as burn-in. Bayesian posterior probabilities (PP) were used to assess branch support of the Markov chain Monte Carlo (MCMC) tree. The number of base substitutions per site of CYTB averaging over all sequence pairs between and within groups was calculated as uncorrected p-distance as well as using the Kimura 2-parameter (K2P) model. The groups were defined on the basis of phylogenetic analysis (see below and Fig. 2). This analysis was conducted in Mega 6.06 and involved 155 CYTB sequences representing 28 mitochondrial lineages. For the phylogenetic analyses of 101 retained nuclear IRBP sequences from all but one of the mitochondrial lineages (m6 was missing because no IRBP sequence was obtained), heterozygous sequences were phased using fastPHASE (Scheet & Stephens, 2006) implemented in DnaSP 5.10 (Librado & Rozas, 2009). Using PartitionFinder 1.0.1 and BIC, the best scheme supported two partitions (see Table S2 in Appendix S2). Phylogenetic analyses were performed in RAxML and MrBayes as described above. 0.05 Figure 2 Mitochondrial Bayesian tree of Grammomys based on concatenated alignment of 1140 bp of CYTB and 575 bp of 16S. The circles indicate statistical support for nodes, specifically 1000 bootstraps in maximum likelihood analysis (BS)/posterior probability from Bayesian analysis (PP). Only values BS > 75 and PP > 0.95 are shown. More detailed version of the tree with precise values of statistical support, tip labels and outgroups is shown in Appendix S3. Calibration of the molecular clock was based on four fossil taxa. Three represent the oldest records of three Arvicanthine genera (Lemniscomys, Arvicanthis, Aethomys) from the Lemudong0 o locality 1, Kenya (Manthi, 2007; 6.12–6.08 Ma), for which we used exponential priors with mean = 1.0 and offset = 6.1 for TMRCA of these genera. The fourth calibration point was represented by the Mus/Arvicanthis split (Kimura et al., 2015; 11.1 Ma), for which we set an exponential prior with mean 1.0 and offset 11.1. For more details, see Table S4 in Appendix S2. For divergence dating analysis, we used the partitioned multi-locus dataset (see Table S2 in Appendix S2) with priors set to the Yule speciation process, and we constrained the tree topology based on the results of the previous ML analysis. We used a linked partition tree 185 J. Bryja et al. and unlinked clock and site models. The MCMC simulations were run twice with 20 million iterations, with genealogies and model parameters sampled every 1000 iterations. The outputs from beast were analysed as described above, following the removal of 25% trees as burn-in. All phylogenetic analyses were run on CIPRES Science Gateway (Miller et al., 2010). Species tree and dating of divergences within Grammomys We used the concatenated mitochondrial sequences (CYTB+16S) and unphased nuclear IRBP genes of the genus Grammomys to obtain a dated species tree under the fully Bayesian framework implemented in the *beast package (Heled & Drummond, 2010), an extension of beast 1.8.2 (Drummond et al., 2012). Alignments for mitochondrial and nuclear genes were given separate and unlinked substitution, clock and tree models (the latter was linked for two mitochondrial markers). The monophyly of the five main lineages was constrained and the tree was calibrated (relaxed log-normal clock, secondary calibration) using the TMRCAs of the main Grammomys lineages estimated from the primary divergence date analysis of Arvicanthini (see Table S4 in Appendix S2). Two independent runs were carried out for 20 million generations with sampling every 2000 generations in beast. The resulting parameter and tree files from the two runs were examined for convergence in Tracer 1.5 and combined in LogCombiner 1.8.2 (Drummond et al., 2012) after removing 10% burn-in. A maximum clade credibility tree was calculated in TreeAnnotator 1.8.2 (Drummond et al., 2012). Biogeographical analysis The dispersal–extinction–cladogenesis model of Lagrange (DEC model; Ree & Smith, 2008) estimates geographical range evolution using a phylogenetic tree with branch lengths scaled to time, geographical (habitat) areas for all tips, and an adjacent matrix of plausibly connected areas. We used the optimization on multiple trees (i.e. Bayes-Lagrange or SDEC model) implemented in the in Rasp 3.1 software (Yu et al., 2015) to take into account topological uncertainty. Rasp computes the likelihood values of all possible ancestral distributions in Lagrange and, relying on a composite Akaike weight, it summarizes the biogeographical reconstructions across trees. Using the distribution data for particular lineages (Fig. 3), we assigned the distribution of tips on the species tree to six main forest types defined by Fayolle et al. (2014; see Fig. 1b). In S-DEC analysis, the maximum number of current and ancestral ranges was set at two (as currently no lineage occurs in more than two main forest types) and all six areas were allowed to be mutually connected in the past. For background phylogenetic information, we used 18,000 trees from the species tree analysis in *beast. The probability of ancestral areas was plotted in the form of pie-charts along the species tree. (b) (a) Bioko Island p1 p2 (kuru?) p3 p4 (poensis) se1 se2 se3 (cometes) se4 (selousi) se5 m1 (dryas) m2 m3 (bunƟngi) m4 (macmillani) m5 (cf. gazellae) m6 m7 m8 dolichurus su1 su2 su3 su4 su5 su6 su7 su8 su9 su10 Figure 3 Geographical distribution of genetic lineages within the five main Grammomys groups. Different groups are shown by different symbol shapes and different lineages by different symbol colours. The names of lineages correspond to those in Fig. 2 and putative species names for some are in parentheses (see text for more details). (a) Poensis (squares), selousi (circles) and macmillani (stars) groups; (b) dolichurus (stars) and surdaster (triangles) groups. 186 Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd Evolution of climbing rats and African forests RESULTS Phylogenetic analysis of the mitochondrial dataset and distribution of genetic variability The topology of mitochondrial Grammomys trees was similar in ML and BI analyses, except for the position of the poensis group (see below). Based on the topology and statistical support for the branches of the inferred tree, we defined five main genetic groups within the genus (Fig. 2; for the tree with tip labels and outgroups, see Appendix S3). These groups have largely parapatric distribution ranges with up to three groups partially overlapping in northeastern Tanzania and southeastern Kenya (Fig. 1). The group names are based on the ongoing taxonomic revision of the genus (J. Bryja et al., unpublished data). (1) The poensis group includes specimens from GuineoCongolese forests on the north bank of the Congo River, including montane forests of the Cameroon volcanic line (Fig. 1). In BI analysis, the poensis group formed a sister clade to the remaining Grammomys taxa (Fig. 2), but in ML topology, it formed a deeply divergent lineage with unresolved relationships to other genera of Arvicanthini. The group can be subdivided into four lineages (p1–p4; Fig. 2) with parapatric distributions. The most distinct populations (=p1) are found in Gabon, isolated by the river Ogooue (Fig. 3a). The lineage p2 may correspond to G. kuru (Thomas & Wroughton, 1907), described from northeastern Democratic Republic of the Congo (DRC). Grammomys poensis was described from Bioko Island and corresponds to lineage p4 (Eisentraut, 1965). (2) The selousi group is named after a recently described species, G. selousi Denys et al., 2011; from southeastern Tanzania, for which CYTB sequence of type material was included in the analysis. The group is subdivided into five lineages with allopatric or parapatric distribution ranges within a narrow belt along the East African coast (se1–se5; Figs 2 & 3a) and appears to prefer lowland forests, e.g. coastal forests inhabited by se4 and se5 (but the latter also occurs in the Usambara Mts and hills of southeastern Kenya; Fig. 3a). The only lineage within this group that is restricted to highlands is se1 in the Southern Rift Mountains (SRM) of southern Tanzania and northern Malawi. The South African lineage se3 may represent G. cometes (Thomas & Wroughton, 1908). (3) The dolichurus group occurs south of the Zambezi (Fig. 3b). Our sample size was too small for detailed analysis of internal genetic structure, but the three lineages seem to correspond to populations distributed along a north–south trajectory (not shown). (4) The macmillani group is composed of eight highly divergent genetic lineages (m1–m8; Figs 2 & 3a). Based on mostly non-overlapping distributions, three lineages can be assigned to earlier species descriptions, although comparisons with type material are required to confirm our current taxonomic interpretation. The m4 lineage is probably G. Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd macmillani (Wroughton, 1907) described from Wouida, north of Lake Turkana in Ethiopia; m1 corresponds to G. dryas (Thomas, 1907) described from the Ruwenzori Mts in Uganda, and m3 to G. buntingi (Thomas, 1911), which is the only Grammomys species in our material occurring west of the Dahomey gap. Furthermore, m5 may represent G. gazellae (Thomas, 1910), a taxon described from South Sudan and synonymized with G. macmillani (Hutterer & Dieterlen, 1984). (5) The surdaster group is named after G. surdaster (Thomas & Wroughton, 1908), a synonym of G. dolichurus (Musser & Carleton, 2005). However, if the dolichurus group is an exclusively southern African clade (see above), we recommend applying the name surdaster to populations north of the Zambezi as has been suggested by Musser & Carleton (2005). The surdaster group is sister to the macmillani group in all mitochondrial trees. Both groups have largely parapatric distribution ranges with a relatively narrow overlap in northern Tanzania and in the Albertine Rift. The surdaster group is widespread in the eastern African highlands between the equator and the Zambezi River (except for a single locality in central Mozambique; Fig. 1) and may also occur in Angola and southern DRC as suggested by su5 from the Kikwit area in southwestern DRC (see also the distribution map in Monadjem et al., 2015 under the name G. dolichurus). The group can be divided into 10 well-supported mitochondrial lineages with mostly parapatric distribution ranges (su1–su10; Figs 2 & 3b). The relationships among them are unresolved, although in most topologies su1 is sister to all the other lineages and su5–su7 and su8–su10 are monophyletic clades. Genetic distances Genetic distances for CYTB within and among mitochondrial lineages of Grammomys are summarized in Table S3 in Appendix S2. Uncorrected p-distances (and similarly K2Pcorrected distances) among lineages belonging to different groups were high and ranged from 8.4% (m59su2) to 18.7% (p29se5). The genetic distances among lineages within each group ranged between 6% and 12% (Table 1), except for the surdaster group, in which 11 of 45 lineage pairs differed by < 5% (see Appendix S2). Analysis of nuclear IRBP gene The phylogenetic analysis of phased IRBP sequences provided a less resolved tree (see Fig. S1 in Appendix S2). Of five major mitochondrial clades, only two (poensis and selousi) were reliably recovered by IRBP. The poensis group formed a clade with the genus Thallomys exclusive of the other Grammomys clades. In the selousi group, only se1 and se3 were significantly supported. In the macmillani group, the geographically adjacent m1 and m2 clades from the Albertine Rift Mts differed substantially in IRBP sequences, while m3 187 J. Bryja et al. Table 1 Minimum and maximum genetic distances (K2P corrected and uncorrected p distances) among lineages in four main Grammomys groups. Genetic variation within the dolichurus group was not analysed because of the low number of available sequences. Minimum distance Maximum distance Groups K2P distance p distance Lineages K2P distance p distance Lineages selousi poensis macmillani surdaster 0.093 0.072 0.064 0.037 0.086 0.067 0.061 0.036 se2 9 se3 p3 9 p4 m7 9 m8 su5 9 su9 0.127 0.106 0.134 0.108 0.114 0.097 0.119 0.098 se1 9 se5 p1 9 p2 m2 9 m6 su1 9 su2 from western Africa was significantly supported as the sister taxon of m5 from Central Africa. There was no obvious structure in the surdaster group, and specimens assigned to different mitochondrial lineages often had very similar or identical IRBP sequences (see Fig. S1 in Appendix S2). Monophyly and phylogenetic position of Grammomys The multi-locus ML and BI phylogenies yielded very similar topologies that validated the Arvicanthini tribe (see Fig. S2 in Appendix S2). All Grammomys representatives clustered in a monophyletic clade, but with low support for the placement of the poensis group. Sister groups that diverged successively were Thallomys and Aethomys, although the nodes were weakly supported. Surprisingly, Grammomys was reconstructed as distantly related to Thamnomys, a genus that historically has been thought to be closely affiliated to it (Musser & Carleton, 2005). Thamnomys diverged at the beginning of the Arvicanthini radiation and appears to be the sister genus of Oenomys. The remaining arvicanthine genera formed three well-supported clades: (1) Hybomys+Stochomys, (2) Desmomys+Rhabdomys and (3) Arvicanthis+Pelomys+Lemniscomys; and two lineages with long and unresolved branches (Dasymys and Micaelamys). Divergence dating within Arvicanthini and species tree of Grammomys The time of divergence between Grammomys and its sister genus Thallomys was estimated as Late Miocene (median TMRCA = 8.83 Ma; see Fig. S2 in Appendix S2). Soon after their split, the poensis group diverged from the rest of the genus (TMRCA of Grammomys = 8.21 Ma). The selousi group then separated (6.58 Ma) from the three remaining groups, which diverged from each other in the Pliocene. Based on secondary calibration of the species tree, TMRCAs of lineages within the five main Grammomys groups are mostly Pleistocene in age, i.e. < 2.5 Ma (Fig. 4). Biogeographical analysis The most probable scenario of the S-DEC model proposed the continuous distribution of ancestral Grammomys in the Late Miocene forests that covered eastern and Central Africa, 188 followed by a vicariance event that separated the Central (the poensis group) and East African groups (Fig. 4). The poensis group subsequently diverged by vicariance to p1 (Wet Central Africa) and remaining lineages (Moist Central Africa), from where the lineage p4 dispersed into West Africa (Nigeria). In East Africa, the ancestors of the selousi group dispersed to coastal forests in the Late Miocene, but lineage se1 remained in the uplands and split by vicariance from the rest of the group. The ancestral areas of both the macmillani and surdaster groups are clearly situated in the East African mountain forests. From there, a single dispersal event to wet-moist West African forests followed by diversification occurred in the m3 lineage (Fig. 4). DISCUSSION Deep divergence in Grammomys and the fragmentation of Miocene forests The multi-locus phylogeny of Arvicanthini supports the monophyly of Grammomys. The > 8 Ma divergence between the poensis group and the remaining lineages makes it one of the oldest intrageneric divergences among African murids (assuming that the poensis group remains in the genus Grammomys, which could be re-evaluated using the data presented here). This finding fits the model of fragmentation of the African Miocene forest into the current Guineo-Congolese forests and coastal and mountain forests in East Africa at this time (Lovett, 1993; Plana, 2004). The formation of the Rift Valley and the decline in global temperatures during the Late Miocene resulted in greater rainfall seasonality, and the spread of grassy vegetation and fragmentation of forests (Bobe, 2006). An increasing number of studies have shown that the genetic diversification between animal and plant taxa occurring in both the central and eastern African forests started during the Late Miocene. For example, the splits between Congolese and eastern African species of the plant genera Uvariodendron and Monodora are dated to c. 8.4 Ma (Couvreur et al., 2008). Similarly, the contraction and fragmentation of the Pan-African forest at this time played a key role in the diversification of some groups of African chameleons (Tolley et al., 2013). Additionally, two rodent lineages, endemic to montane forests of East Africa (the denniae group of Hylomyscus and Praomys delectorum), split from their sister lineages living mostly in Guineo-Congolese forests Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd Evolution of climbing rats and African forests su4 su3 su7 su5 su6 su2 su8 su9 su10 su1 m8 m7 m4 m5 m3 m2 m1 d se5 se4 se2 se3 se1 p4 p2 p3 p1 A - Coastal Eastern B - Upland Eastern C - Dry Western D - Wet-Moist Western E - Moist Central F - Wet Central dispersal vicariance BE->BEF->B|EF Mya Figure 4 Ultrametric Grammomys species tree from *beast. The pie-charts indicate the most probable ancestral areas of particular clades as estimated by S-DEC model in Bayes-Lagrange (Ree & Smith, 2008). at the beginning of the Praomyini radiation dated to the end of the Miocene (Lecompte et al., 2005; Missoup et al., 2012; Demos et al., 2014). Palaeoendemism in coastal forests of East Africa The coastal forests of East Africa were recognized as a distinct phytogeographical unit by White (1983) and, more recently, by Fayolle et al. (2014). They exhibit a patchy distribution extending from southern Somalia to the Limpopo River in southern Mozambique and represent endangered centres of biodiversity. There is evidence that most of the coastal forest endemics, including mammals, are palaeoendemics (Burgess et al., 1998). Phylogenetic reconstruction of Grammomys revealed the split of the selousi group from other East African Grammomys c. 6.5 Ma (Fig. 4), indicating a Late Miocene separation of coastal and highland forests in eastern Africa (Fig. 5). This is concordant with the divergence time (c. 6.5 Ma) proposed by Mikula et al. (2016) between the genus Beamys (a rodent typical of African coastal forests) and its sister genus Cricetomys (widespread in various African forests). The Grammomys lineage se3 from east coastal South Africa suggests a historical connection between coastal forests in East Africa and those further south, which has not been reported before. Species inhabiting these coastal forests are able to reach higher altitude Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd forests (possibly via riverine gallery forests) as suggested by the presence of se2 in the Mulanje Mts, se5 in the Usambara Mts and the observation that Beamys occurs in coastal forests as well as in the Southern Rift Mountains (SRM) (Happold, 2013). The clear north–south structuring within the selousi group reflects the fragmented nature of coastal forests; this separation may be maintained by large rivers (e.g. Rufiji, Zambezi, Limpopo) as observed for other lowland species (Bartakova et al., 2015; McDonough et al., 2015). Alternative hypotheses of divergence within coastal forests include climatic changes in the Plio-Pleistocene or increases in sea level, shrinking suitable habitats into isolated fragments situated at higher elevations (Burgess et al., 1998). Evolution of the eastern Afromontane biodiversity hotspot during Plio-Pleistocene climatic oscillations A reversal of the cooling trend occurred in the Early Pliocene. This represented the warmest period over the last 5 Myr, leading to the suggestion that East African forests may have expanded at this time, especially at higher elevations (Feakins & deMenocal, 2010). More continuous forest cover probably facilitated the dispersion of the dolichurus group in southeastern Africa during that period. However, after 3.5 Ma temperatures decreased and the Plio-Pleistocene aridification events linked with significant expansion of grass-dominated 189 J. Bryja et al. to have descended from high, humid montane forest to drier, forested savanna habitats. We hypothesize that an increased ability to colonize drier habitats may have allowed Grammomys to colonize relatively large areas at the interface between the Guineo-Congolese forests and the Sudanian savanna, and consequently, the Guinean forests-savanna mosaic of West Africa (m3; see below). The diversification events within the surdaster group may also be linked to Pleistocene climatic changes. There is increasing evidence that during humid periods within the last 2 Myr, the currently fragmented mountain forests of the EAM and SRM were repeatedly united, allowing the periodic exchange of forest-dependent faunas. However, it is unlikely that a single spatiotemporal scenario applies for all faunal components, as even species with presumably similar ecological requirements may have different responses to the same environmental changes (Carleton & Stanley, 2012). For example, phylogenetic reconstructions of the forest-dependent rodent Praomys delectorum revealed two distinct lineages corresponding to the Usambara Mts in the north and Nguru Mts in the south, which are separated by the wide savanna belt in northeastern Tanzania (Bryja et al., 2014b). However, both sides of this belt are inhabited by a single mitochondrial Grammomys lineage (su10; Fig. 3). Such incongruent distribution patterns may be due to a lower dependency of Grammomys on the prevailing ecological conditions in humid montane forests. This would have allowed ecosystems in East Africa generated more diverse mosaic environments (Bobe, 2006). Within the genus Grammomys, these environmental changes are reflected by intensive radiations that occurred in the eastern Afromontane hotspot, especially in the Eastern Arc Mountains and Southern Rift Mountains (EAM+SRM; the surdaster group) and the Kenyan Highlands and Albertine Rift Mountains (KH+ARM; the macmillani group) (Fig. 5). The overlap in the distribution ranges of mammal species occurring in the main blocks of the eastern Afromontane region (i.e. EAM+SRM versus KH+ARM) is generally very low (e.g. Carleton et al., 2015), suggesting that the faunas of the EAM+SRM and the KH+ARM pursued longterm independent evolutionary trajectories. The distribution ranges for the macmillani and surdaster groups reported in this study appear to agree with this scenario (Fig. 1). Demos et al. (2014) provided evidence of repeated Pleistocene connections between small mammal taxa inhabiting forests of the Albertine Rift Mts and the Kenyan Highlands. This explains the sister-group relationship between two lineages restricted to high elevations of the Albertine Rift Mts (i.e. palaeoendemics m1 + m2) and the rest of the macmillani group, the geographical origin of which is presumed to be in the Kenyan highlands. It can be argued that during one of the humid Pleistocene periods, lineage m4 from the Kenyan highlands colonized the southern Kenyan and northern Tanzanian mountains (e.g. the volcanoes in the Rift Valley inhabited by m7 and m8). Subsequently, the lineage leading to m5 appears (a) (b) Late Miocene (9-7 Mya) Late Miocene (8-5.5 Mya) (d) (c) ? Pliocene (5-2.5 Mya) Pleistocene (2.5-0.5 Mya) Figure 5 Schematic illustration of major evolutionary events in Grammomys. (a) The fragmentation of Late Miocene pan-African forest into the ancestors of current Guineo-Congolese forests (green) and East African montane and coastal forests (purple). (b) The split between Grammomys inhabiting montane (red) and coastal (yellow) forests in East Africa. (c) During the Pliocene the ancestors of the dolichurus (orange), surdaster (red) and macmillani (blue) groups split along a south–-north trajectory. The long-term forest refugia for the surdaster and macmillani groups were probably located in the EAM + SRM for the former and in KH + ARM for the latter. (d) Pleistocene climatic cycles caused repeated fragmentations and expansions of forest habitats leading to diversification within all five main clades. One of the expansions of the macmillani clade involved the colonization of Guinean forests (m3 lineage) by the ‘northern route,’ i.e. north of the Congolese forests. Note that the ellipses at (a) and (b) show only schematically the positions of ancestral populations and do not indicate precise geographical locations. 190 Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd Evolution of climbing rats and African forests them to colonize both miombo woodlands (lineage su4) and savanna-forest mosaics on the southeastern edge of the Congolese forests (su5–su7). Such distribution patterns have not been observed in previously studied forest specialists restricted to the EAM and SRM (e.g. Lawson, 2010; Tolley et al., 2011; Bryja et al., 2014b; Loader et al., 2014). Long-distance dispersal along the northern edge of the Congo Basin In order to explain similarities between eastern and western African montane forests and grasslands, many authors have assumed that, during climatic changes and especially during colder humid periods, the mountain floras and faunas must have extended to the lowlands, which facilitated dispersal between mountain massifs (White, 1981). The zones characterized by the mosaic of forest and savanna north of the Congo basin are among the least known areas of Africa. However, our results concerning the distribution of Grammomys m5 suggest that there is a clear biogeographical connection between Uganda (+westernmost Kenya) and Central Africa (northeastern DRC, CAR, South Sudan). This link is not only indicated by this study, but also by earlier studies which revealed that identical genetic lineages of other rodents occur in this forest/savanna mosaic, e.g. Mus cf. bufo (Bryja et al., 2014a), or Aethomys hindei (Monadjem et al., 2015). The biogeographical scenario suggests that, during humid phases, the Pleistocene lowland forests of the Congo Basin extended further north than they do today. This situation may have allowed the ancestors of Grammomys m3 + m5 from eastern Africa to disperse along the northern margin of the Congolese forest and colonize northeastern DRC, CAR and South Sudan (Fig. 5). It seems plausible that after the northern edge of the lowland forests in the Congo Basin receded, some populations persisted in the resulting relict forests in forest-savanna mosaics (i.e. G. m5 in CAR), montane areas (probably G. aridulus in Jebel Marra region in Sudan; Fig. 1) or adapted to new environments, where Grammomys mice were previously absent (G. buntingi=m3 in West Africa). CONCLUSION This is the first phylogenetic study of Grammomys that includes samples from most of its distribution area in subSaharan Africa. Our results suggest that the genus is monophyletic and unrelated to Thamnomys, and that its intrageneric divergences are among the oldest in African murids (> 8 Ma). The majority of the five detected clades have parapatric distribution ranges, and the times of divergence estimated among these clades agree with accepted scenarios for the evolutionary history of the African forests since the Late Miocene. The distribution of these lineages does not agree with the current taxonomy. Our results suggest that a revision of this genus will lead to discoveries of new species, especially in highland and coastal forests in East Africa. Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd ACKNOWLEDGEMENTS This study was supported by the Czech Science Foundation, project no. 14-36098G. The French ANR – IFORA project and the INCO-DEV Project – TREATCONTROL allowed work in West-Central Africa. Fieldwork in DR Congo was supported by the Belgian Directorate for Development Cooperation (DGD) and the Flemish Inter-University Council – University Development Cooperation (VLIR-UOS). For help during the fieldwork, we acknowledge V. Mazoch, H. Konvi
ckova, J.
Sklıba, M. L€ ovy, G. Mhamphi, F. Sedla
cek, S.
Safar
cıkova, A. Kone
cny , S. Gambalemoke Mbalitini, B. Kadjo, F. Kourouma, M. Sylla, D. Mory, J. Mwanga, P. Kaleme, B. Ndara Rusziga, R. Kityo, A Wetsiba, E. Kemming, A.D. Missoup, R. Cornette, S. Moulin, E. Lecompte, A. Lalis and all other local field collaborators. The assistance of R. Makundi, A. Massawe, C. Sabuni, J. Mbau, W.N. Chitaukali, M. Colyn, B. Dudu Akaibe, L. Koivogui, C. Camara and the late W. Verheyen with project logistics and collection of samples is highly appreciated. H. Konvi
ckova and L. Pialek helped with genotyping, and F. Jacquet, C. Sabuni and J. Go€ uy de Bellocq provided unpublished sequences. For permission to carry out the research and to collect specimens, we are obliged to the National Research Council and Forestry Department in Malawi, the Uganda National Council for Science and Technology, the Kenyan Forest Service and the Kenyan Wildlife Service, COSTECH Tanzania, the Guinean and Cameroonian Ministries of Water and Forest Management, the Zambian Wildlife Authority, Sokoine University of Agriculture in Morogoro (Tanzania), Uganda Wildlife Authority, Institut National pour l’Environment et la Conservacion de la Nature (Burundi), Institut Nacional pour la Conservacion de la Nature and the scientific team at Lwiro (DRC), Wildlife Conservation Society, Rwanda Development Board, and National University of Rwanda (N. Ntare, D. Tuyisingize), Gorongosa Restoration Project & the Carr Foundation (Mozambique), National Museums of Kenya (B. Agwanda) and the ‘Centre de surveillance de la biodiversite’ in Kisangani (DR Congo). We also thank the SYNTHESYS programme (BE-TAF-5113 to OM and FRTAF-5799 to JB) and museum curators that allowed us to study collections in their care: W. Wendelen (RMCA), R. Hutterer (ZFMK), J. Phelps (FMNH), C. Conroy (MVZ), D. Lunde (USNM) and V. Volpato (SMF). B. van Vuuren, J. Masters, P. Linder and two anonymous referees provided very useful comments on previous version of the manuscript. REFERENCES Bartakova, V., Reichard, M., Bla
zek, R., Pola
cik, M. & Bryja, J. (2015) Terrestrial fishes: rivers are barriers to gene flow in annual fishes from the African savanna. Journal of Biogeography, 42, 1832–1844. Bobe, R. (2006) The evolution of arid ecosystems in eastern Africa. Journal of Arid Environments, 66, 564–584. 191 J. Bryja et al. Bryja, J., Mikula, O.,
Sumbera, R., Meheretu, Y., Aghova, T., Lavrenchenko, L.A., Mazoch, V., Oguge, N., Mbau, J.S., Welegerima, K., Amundala, N., Colyn, M., Leirs, H. & Verheyen, E. (2014a) Pan-African phylogeny of Mus (subgenus Nannomys) reveals one of the most successful mammal radiations in Africa. BMC Evolutionary Biology, 14, e256. Bryja, J., Mikula, O., Patzenhauerova, H., Oguge, N.,
Sumbera, R. & Verheyen, E. (2014b) The role of dispersal and vicariance in the Pleistocene history of an East African mountain rodent, Praomys delectorum. Journal of Biogeography, 41, 196–208. Burgess, N.D., Clarke, G.P. & Rodgers, W.A. (1998) Coastal forests of eastern Africa: status, endemism patterns and their potential causes. Biological Journal of the Linnean Society, 64, 337–367. Carleton, M.D. & Stanley, W.T. (2012) Species limits within the Praomys delectorum group (Rodentia: Muridae: Murinae) of East Africa: a morphometric reassessment and biogeographical implications. Zoological Journal of the Linnean Society, 165, 420–469. Carleton, M.D., Banasiak, R.A. & Stanley, W.T. (2015) A new species of the rodent genus Hylomyscus from Angola, with a distributional summary of the H. anselli species group (Muridae: Murinae: Praomyini). Zootaxa, 4040, 101–128. Cerling, T.E. (1992) Development of grasslands and savannas in East Africa during the Neogene. Palaeogeography, Palaeoclimatology, Palaeoecology, 97, 241–247. Couvreur, T.L.P., Chatrou, L.W., Sosef, M.S.M. & Richardson, J.E. (2008) Molecular phylogenetics reveal multiple tertiary vicariance origins of the African rain forest trees. BMC Biology, 6, e54. De Klerk, H.M., Crowe, T.M., Fjeldsa, J. & Burgess, N.D. (2002) Biogeographical patterns of endemic terrestrial Afrotropical birds. Diversity and Distributions, 8, 147– 162. Demos, T.C., Kerbis Peterhans, J.C., Agwanda, B. & Hickerson, M.J. (2014) Uncovering cryptic diversity and refugial persistence among small mammal lineages across the Eastern Afromontane biodiversity hotspot. Molecular Phylogenetics and Evolution, 71, 41–54. Denys, C., Lalis, A., Lecompte, E., Cornette, R., Moulin, S., Makundi, R., Machang’u, R., Volobouev, V. & Aniskine, V. (2011) Notes on a faunal survey in Kingu Pira (South Tanzania) with new karyotypes of several small mammals and description of a new Murid species (Mammalia, Rodentia). Zoosystema, 33, 5–47. Drummond, A.J., Suchard, M.A., Xie, D. & Rambaut, A. (2012) Bayesian phylogenetics with BEAUti and the BEAST 1.7. Molecular Biology and Evolution, 29, 1969–1973. Ducroz, J.F., Volobouev, V. & Granjon, L. (2001) An assessment of the systematics of arvicanthine rodents using mitochondrial DNA sequences: evolutionary and biogeographical implications. Journal of Mammalian Evolution, 8, 173–206. Fayolle, A., Swaine, M.D., Bastin, J.-F., Bourland, N., Comiskey, J.A., Dauby, G., Doucet, J.-L., Gillet, J.-F., Gourlet192 Fleury, S., Hardy, O.J., Kirunda, B., Kouame, F.N. & Plumptre, A.J. (2014) Patterns of tree species composition across tropical African forests. Journal of Biogeography, 41, 2320–2331. Feakins, S.J. & deMenocal, P.B. (2010) Global and African regional climate during the Cenozoic. Cenozoic mammals of Africa (ed. by L. Werdelin), pp. 45–55. University of California Press, Berkeley and London. Galan, M., Pages, M. & Cosson, J.F. (2012) Next-generation sequencing for rodent barcoding: species identification from fresh, degraded and environmental samples. PLoS ONE, 7, e48374. Happold, D.C.D. (2013) Mammals of Africa, Volume III Rodents, hares and rabbits. Bloomsbury Publishing, London and New York. Heled, J. & Drummond, A.J. (2010) Bayesian inference of species trees from multilocus data. Molecular Biology and Evolution, 27, 570–580. Huntley, J.W. & Voelker, G. (2016) Cryptic diversity in Afro-tropical lowland forests: the systematics and biogeography of the avian genus Bleda. Molecular Phylogenetics and Evolution, 99, 297–308. Hutterer, R. & Dieterlen, F. (1984) Zwei neue Arten der Gat€ tung Grammomys aus Athiopien und Kenia (Mammalia: Muridae). Stuttgarter Beitr€age zur Naturkunde Serie A (Biologie), 374, 1–18. Kimura, Y., Hawkins, M.T.R., McDonough, M.M., Jacobs, L.L. & Flynn, L.J. (2015) Corrected placement of Mus-Rattus fossil calibration forces precision in the molecular tree of rodents. Scientific Reports, 5, e14444. Lanfear, R., Calcott, B., Ho, S.Y.W. & Guindon, S. (2012) PartitionFinder: combined selection of partitioning schemes and substitution models for phylogenetic analyses. Molecular Biology and Evolution, 29, 1695–1701. Lawson, L.P. (2010) The discordance of diversification: evolution in the tropical-montane frogs of the Eastern Arc Mountains of Tanzania. Molecular Ecology, 19, 4046–4060. Lecompte, E., Denys, C. & Granjon, L. (2005) Confrontation of morphological and molecular data: the Praomys group (Rodentia, Murinae) as a case of adaptive convergences and morphological stasis. Molecular Phylogenetics and Evolution, 37, 899–919. Lecompte, E., Aplin, K., Denys, C., Catzeflis, F., Chades, M. & Chevret, P. (2008) Phylogeny and biogeography of African Murinae based on mitochondrial and nuclear gene sequences, with a new tribal classification of the subfamily. BMC Evolutionary Biology, 8, 199. Librado, P. & Rozas, J. (2009) DnaSP v5: a software for comprehensive analysis of DNA polymorphism data. Bioinformatics, 25, 1451–1452. Loader, S.P., Ceccarelli, F.S., Menegon, M., Howell, K.M., Kassahun, R., Mengistu, A.A., Saber, S.A., Gebresenbet, F., de Sa, R., Davenport, T.R.B., Larson, J.G., M€ uller, H., Wilkinson, M. & Gower, D.J. (2014) Persistence and stability of Eastern Afromontane forests: evidence from brevicipitid frogs. Journal of Biogeography, 41, 1781–1792. Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd Evolution of climbing rats and African forests Lovett, J.C. (1993) Eastern Arc moist forest flora. Biogeography and ecology of the rain forests of eastern Africa (ed. by J.C. Lovett and S.K. Wasser), pp. 33–55. Cambridge University Press, Cambridge. Maley, J. (1996) The African rain forest – main characteristics of changes in vegetation and climate from the Upper Cretaceous to the Quaternary. Proceedings of the Royal Society of Edinburgh, 104B, 31–73. Manthi, F.K. (2007) A preliminary review of the rodent fauna from Lemudong’o, southwestern Kenya, and its implication to the late Miocene paleoenvironments. Kirtlandia, 56, 92–105. McDonough, M.M.,
Sumbera, R., Mazoch, V., Ferguson, A.W., Phillips, C.D. & Bryja, J. (2015) Multilocus phylogeography of a widespread savanna-woodland-adapted rodent reveals the influence of Pleistocene geomorphology and climate change in Africa’s Zambezi region. Molecular Ecology, 24, 5248–5266. Mikula, O.,
Sumbera, R., Aghova, T., Mbau, J., Katakweba, A.A.S., Sabuni, C.A. & Bryja, J. (2016) Evolutionary history and species diversity of African Pouched Mice of the genus Saccostomus (Rodentia: Nesomyidae). Zoologica Scripta, in press. DOI:10.1111/zsc.12179. Miller, M.A., Pfeiffer, W. & Schwartz, T. (2010) Creating the CIPRES Science Gateway for inference of large phylogenetic trees. Proceedings of the Gateway Computing Environments Workshop (GCE), 14 Nov. 2010, pp. 1–8. New Orleans, LA. Missoup, A.D., Nicolas, V., Wendelen, W., Keming, E., Bilong Bilong, C.F., Couloux, A., Atanga, E., Hutterer, R. & Denys, C. (2012) Systematics and diversification of Praomys species (Rodentia: Muridae) endemic to the Cameroon Volcanic Line (West Central Africa). Zoologica Scripta, 41, 327–345. Missoup, A.D., Nicolas, V., Eiseb, S., Chung, E.K. & Denys, C. (2016) Phylogenetic position of the endemic Mount Oku rat, Lamottemys okuensis (Rodentia: Muridae), based on molecular and morphological data. Zoological Journal of the Linnean Society, 177, 209–226. Monadjem, A., Taylor, P.J., Denys, C. & Cotterill, F.P.D. (2015) Rodents of sub-Saharan Africa. A biogeographic and taxonomic synthesis. Walter de Gruyter GmbH, Berlin/ Munich/Boston. Musser, G.G. & Carleton, M.D. (2005) Superfamily Muroidea. Mammal species of the world: a taxonomic and geographic reference (ed. by D.E. Wilson and D.M. Reeder), pp. 894–1531. Johns Hopkins University Press, Baltimore. Plana, V. (2004) Mechanisms and tempo of evolution in the African Guineo-Congolian rainforest. Philosophical Transactions of the Royal Society B: Biological Sciences, 359, 1585–1594. Plumptre, A.J., Davenport, T.R.B., Behangana, M., Kityo, R., Eilu, G., Ssegawa, P., Ewango, C., Meirte, D., Kahindo, C., Herremans, M., Kerbis Petterhans, J., Pilgrim, J.D., Wilson, M., Languy, M. & Moyer, D. (2007) The biodiversity of the Albertine Rift. Biological Conservation, 134, 178–194. Rambaut, A. & Drummond, A.J. (2007) Tracer v1.5. Available at: http://beast.bio.ed.ac.uk/Tracer. Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd Ree, R.H. & Smith, S.A. (2008) Maximum likelihood inference of geographic range evolution by dispersal, local extinction, and cladogenesis. Systematic Biology, 57, 4–14. Ronquist, F. & Huelsenbeck, J.P. (2003) MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics, 19, 1572–1574. Rovero, F., Menegon, M., Fjelds a, J., Collett, L., Doggart, N., Leonard, C., Norton, G., Owen, N., Perkin, A., Spitale, D., Ahrends, A. & Burgess, N.D. (2014) Targeted vertebrate surveys enhance the faunal importance and improve explanatory models within the Eastern Arc Mountains of Kenya and Tanzania. Diversity and Distributions, 20, 1438–1449. Scheet, P. & Stephens, M. (2006) A fast and flexible statistical model for large-scale population genotype data: applications to inferring missing genotypes and haplotypic phase. American Journal of Human Genetics, 78, 629–644. Stamatakis, A. (2014) RAxML Version 8: a tool for phylogenetic analysis and postanalysis of large phylogenies. Bioinformatics, 30, 1312–1313. Tamura, K., Stecher, G., Peterson, D., Filipski, A. & Kumar, S. (2013) MEGA6: molecular evolutionary genetics analysis version 6.0. Molecular Biology and Evolution, 30, 2725–2729. Tolley, K.A., Tilbury, C.R., Measey, G.J., Menegon, M., Branch, W.R. & Matthee, C.A. (2011) Ancient forest fragmentation or recent radiation? Testing refugial speciation models in chameleons within an African biodiversity hotspot. Journal of Biogeography, 38, 1748–1760. Tolley, K.A., Townsend, T.M. & Vences, M. (2013) Largescale phylogeny of chameleons suggests African origins and Eocene diversification. Proceedings of the Royal Society B: Biological Sciences, 280, e20130184. White, F. (1981) The history of the Afromontane archipelago and the scientific need for its conservation. African Journal of Ecology, 19, 33–54. White, F. (1983) The vegetation of Africa. UNESCO, Switzerland. Wiens, J.J. & Donoghue, M.J. (2004) Historical biogeography, ecology and species richness. Trends in Ecology and Evolution, 19, 639–644. Yu, Y., Harris, A.J., Blair, C. & He, X.J. (2015) RASP (Reconstruct Ancestral State in Phylogenies): a tool for historical biogeography. Molecular Phylogenetics and Evolution, 87, 46–49. SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article: Appendix S1 Collecting localities and genetic data. Appendix S2 Additions to phylogenetic analyses. Appendix S3 Detailed Bayesian phylogeny of mtDNA. DATA ACCESSIBILITY New sequences used in phylogenetic analyses are available in GenBank under accession numbers KU723898–KU724057 193 J. Bryja et al. and KU747156–KU747161 (CYTB), KU723674–KU723792 (16S), KU723651–KU723656 and KU723793–KU723897 (IRBP), KU723660–KU723673 (RAG1), KU723657–KU723659 (BRCA1) (see Appendix S1). Further details of specimens, including museum numbers, are specified in Appendix S1. BIOSKETCH Josef Bryja is head of the molecular ecology group at the Institute of Vertebrate Biology ASCR and has a general interest in factors affecting the evolution of vertebrate 194 populations. His specialities include phylogeography and speciation in Africa, conservation genetics and mechanisms of host–parasite co-evolution. Author contributions: J.B., R.
S., C.D. and E.V. conceived and designed the study; J.B., R.
S., J.K.P., C.D., V.N., T.A. and E.V. collected or provided important part of samples; T.A. and A.B. genotyped most samples; J.B., O.M. and T.A. analysed data; and J.B. wrote the first draft of the manuscript. All authors contributed to the final version of the paper. Editor: Judith Masters Journal of Biogeography 44, 182–194 ª 2016 John Wiley & Sons Ltd