ArticlePDF Available

Sky, sea, and forest islands: Diversification in the African leaf‐folding frog Afrixalus paradorsalis (Anura: Hyperoliidae) of the Lower Guineo‐Congolian rain forest



Aim To investigate how putative barriers, forest refugia, and ecological gradients across the lower Guineo‐Congolian rain forest shape genetic and phenotypic divergence in the leaf‐folding frog Afrixalus paradorsalis, and examine the role of adjacent land bridge and sky‐islands in diversification. Location The Lower Guineo‐Congolian Forest, the Cameroonian Volcanic Line (CVL), and Bioko Island, Central Africa. Taxon Afrixalus paradorsalis (Family: Hyperoliidae), an African leaf‐folding frog. Methods We used molecular and phenotypic data to investigate diversity and divergence among the A. paradorsalis species complex distributed across lowland rain forests, a land bridge island, and mountains in Central Africa. We examined the coincidence of population boundaries, landscape features, divergence times, and spatial patterns of connectivity and diversity, and subsequently performed demographic modelling using genome‐wide SNP variation to distinguish among divergence mechanisms in mainland (riverine barriers, forest refugia, ecological gradients) and land bridge island populations (vicariance, overwater dispersal). Results We detected four genetically distinct allopatric populations corresponding to Bioko Island, the CVL, and two lowland rain forest populations split by the Sanaga River. Although lowland populations are phenotypically indistinguishable, pronounced body size evolution occurs at high elevation, and the timing of the formation of the high elevation population coincides with mountain uplift in the CVL. Spatial analyses and demographic modelling revealed population divergence across mainland Lower Guinea is best explained by forest refugia rather than riverine barriers or ecological gradients, and that the Bioko Island population divergence is best explained by vicariance (marine incursion) rather than overseas dispersal. Main conclusions We provide growing support for the important role of forest refugia in driving intraspecific divergences in the Guineo‐Congolian rain forest. In A. paradorsalis, sky‐islands in the CVL have resulted in greater genetic and phenotypic divergences than marine incursions of the land bridge Bioko Island, highlighting important differences in patterns of island‐driven diversification in Lower Guinea.
Sky, sea, and forest islands: Diversification in the African
leaf-folding frog Afrixalus paradorsalis (Anura: Hyperoliidae)
of the Lower Guineo-Congolian rain forest
Kristin L. Charles
Rayna C. Bell
David C. Blackburn
Marius Burger
Matthew K. Fujita
aclav Gvo
Gregory F.M. Jongsma
Marcel Talla Kouete
Adam D. Leach
Daniel M. Portik
Department of Biology, University of Nevada, Reno, Nevada
Department of Vertebrate Zoology, National Museum of Natural History, Smithsonian Institution, Washington, District of Columbia
Museum of Vertebrate Zoology, University of California, Berkeley, California
Florida Museum of Natural History, University of Florida, Gainesville, Florida
African Amphibian Conservation Research Group, Unit for Environmental Sciences and Management, North-West University, Potchefstroom,South Africa
Flora Fauna & Man, Ecological Services Ltd., Tortola, British Virgin Island
Department of Biology, The University of Texas at Arlington, Arlington, Texas
Institute of Vertebrate Biology, Czech Academy of Sciences, Brno,Czech Republic
Department of Zoology, National Museum, Prague, Czech Republic
Department of Biology, University of Washington, Seattle, Washington
Burke Museum of Natural History and Culture, University of Washington, Seattle, Washington
Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona
Daniel M. Portik, Department of Ecology
and Evolutionary Biology, University of
Arizona, Tucson, AZ.
Funding information
Division of Environmental Biology, Grant/
Award Number: 1202609, 1309171,
1311006, 1457232; Grantov
a Agentura
e Republiky, Grant/Award Number: 15-
Editor: Jim Provan
Aim: To investigate how putative barriers, forest refugia, and ecological gradients
across the lower Guineo-Congolian rain forest shape genetic and phenotypic diver-
gence in the leaf-folding frog Afrixalus paradorsalis, and examine the role of adjacent
land bridge and sky-islands in diversification.
Location: The Lower Guineo-Congolian Forest, the Cameroonian Volcanic Line
(CVL), and Bioko Island, Central Africa.
Taxon: Afrixalus paradorsalis (Family: Hyperoliidae), an African leaf-folding frog.
Methods: We used molecular and phenotypic data to investigate diversity and
divergence among the A. paradorsalis species complex distributed across lowland
rain forests, a land bridge island, and mountains in Central Africa. We examined the
coincidence of population boundaries, landscape features, divergence times, and
spatial patterns of connectivity and diversity, and subsequently performed demo-
graphic modelling using genome-wide SNP variation to distinguish among diver-
gence mechanisms in mainland (riverine barriers, forest refugia, ecological gradients)
and land bridge island populations (vicariance, overwater dispersal).
Results: We detected four genetically distinct allopatric populations corresponding
to Bioko Island, the CVL, and two lowland rain forest populations split by the
Sanaga River. Although lowland populations are phenotypically indistinguishable,
DOI: 10.1111/jbi.13365
Journal of Biogeography. 2018;114. ©2018 John Wiley & Sons Ltd
pronounced body size evolution occurs at high elevation, and the timing of the
formation of the high elevation population coincides with mountain uplift in the
CVL. Spatial analyses and demographic modelling revealed population divergence
across mainland Lower Guinea is best explained by forest refugia rather than river-
ine barriers or ecological gradients, and that the Bioko Island population diver-
gence is best explained by vicariance (marine incursion) rather than overseas
Main conclusions: We provide growing support for the important role of forest
refugia in driving intraspecific divergences in the Guineo-Congolian rain forest. In A.
paradorsalis, sky-islands in the CVL have resulted in greater genetic and phenotypic
divergences than marine incursions of the land bridge Bioko Island, highlighting
important differences in patterns of island-driven diversification in Lower Guinea.
Africa, amphibian, historical demography, land bridge island, Lower Guinea, phylogeography
The Guineo-Congolian rain forests of West and Central Africa are
home to approximately one fifth of the worlds terrestrial biodiver-
sity (Myers, Mittermeier, Mittermeier, da Fonseca, & Kent, 2000),
with up to 80% of this diversity considered endemic (Plana, 2004).
The highest species richness in the Guineo-Congolian hotspot is
centered along the Cameroonian Volcanic Line (CVL), a chain of vol-
canoes that comprise the Cameroonian Highland mountain ranges
and the Gulf of Guinea archipelago, which includes a land bridge
island (Bioko) and three oceanic islands (Pr
ıncipe, S~
ao Tom
e and
on; Figure 1). These volcanic peaks of the CVL function as
sky islandswith steep elevational gradients that act as barriers to
dispersal between surrounding lowland forests, and this region is
characterized by high diversity of montane and submontane ende-
mics (Oates, Bergl, & Linder, 2004). The land bridge island and low-
land forest habitats of the Guineo-Congolian hotspot also harbour
exceptional species diversity (Jones, 1994; Stuart, Adams, & Jenkins,
1990). However, the mechanisms that mediate exchanges across
montane, lowland, and land bridge island habitats and how these
dynamics contribute to subsequent diversification are still poorly
Climatic refugia, riverine barriers, and ecological gradients all
potentially contribute to diversification in tropical forests (Haffer,
1997; Moritz, Patton, Schneider, & Smith, 2000; Plana, 2004; Smith,
Wayne, Girman, & Bruford, 1997). Using phylogeographical patterns
to infer which of these mechanisms contributes to population diver-
gence can be problematic when the relevant landscape features
have overlapping geographical arrangements. For example, several
large rivers appear to act as barriers to dispersal (the Sanaga, Congo
and Ogoou
e: Figure 1), yet these patterns may be confounded by
the historical presence of lowland rain forest climatic refugia on
either side of the rivers (reviewed in Portik et al., 2017). However,
population divergence due to refugia, rivers or ecological gradients
should result in different population demographic histories, and
these corresponding demographic predictions can be examined in a
model-testing framework to improve inferences about diversification
mechanisms (Portik et al., 2017). The Lower Guinea region also con-
tains several forest-specific ecotones, which can promote parapatric
speciation through disruptive selection across ecological gradients
(Moritz et al., 2000; Smith et al., 1997). The longitudinal E-W gradi-
ent in precipitation across Lower Guinea (Olson et al., 2001) and
the northsouth climate hinge (latitude 2°N; Leroux, 1983) both
coincide with population structuring in plants (reviewed in Hardy
et al., 2013; Heuertz, Duminil, Dauby, Savolainen, & Hardy, 2014).
Population boundaries along ecotones provide indirect evidence of
the ecological gradient hypothesis, though other landscape features
such as the historical presence of climatic refugia or riverine barriers
can confound these patterns. Consequently, methods that explicitly
quantify demographic changes in population size and gene flow may
better differentiate between climatic refugia, riverine barriers, and
ecological gradients as potential diversification mechanisms (Portik
et al., 2017).
Cycles of isolation and connectivity on land bridge islands can
also contribute to diversification in tropical forests of Lower Guinea
(Bell et al., 2017; Leach
e, Fujita, Minin, & Bouckaert, 2014). How-
ever, relying on phylogeographical patterns to infer the timing and
duration of vicariance during marine incursions can be challenging.
For example, Bioko Island is composed of three volcanic peaks on
the continental shelf that range in age from 13 Myr (Marzoli et al.,
2000) and is approximately 30 km from the present-day coast of
Cameroon (Figure 1). Rising and retreating sea levels have resulted
in multiple periods of isolation and connectivity, and patterns of
molecular divergence between populations on Bioko and the adja-
cent mainland are consistent with reduced gene flow across the
present-day marine barrier, particularly for amphibians and other
organisms that are poor overseas dispersers (Barej et al., 2014; Bell
et al., 2017; Leach
e et al., 2014). Yet amphibians from multiple
families colonized two oceanic islands in the Gulf of Guinea archi-
pelago via overseas dispersal (Bell, Drewes, Channing, et al., 2015;
Measey et al., 2007) indicating that these sweepstakesoverseas
dispersal events have occurred multiple times in the archipelago.
Thus, a model-testing framework that accounts for population
demographic signatures characteristic of population vicariance,
founding events, and intermittent dispersal may greatly improve
inferences about the evolutionary history of land bridge island
Here, we investigate how putative barriers, refugia, and ecologi-
cal gradients across the Guineo-Congolian rain forest shape genetic
and phenotypic divergence in the leaf-folding frog Afrixalus parador-
salis Perret, 1960, a tree frog species complex comprised of two sub-
species. The nominate form, A. p. paradorsalis, has a broad
distribution across lowland forests throughout Lower Guinea, includ-
ing Cameroon, Gabon, Equatorial Guinea and the Republic of Congo,
and also occurs on Bioko Island. A morphologically distinct popula-
tion in the submontane region surrounding Mt. Manengouba in the
CVL (Figure 1) was previously described as the subspecies A. p.
manengubensis Amiet, 2009. We use the A. paradorsalis species com-
plex to investigate the following questions about diversification
mechanisms in the Guineo-Congolian forests: (a) How many distinct
populations comprise A. paradorsalis? (b) How are these populations
related? (c) Do population boundaries coincide with geographic fea-
tures and effective migration? (d) What demographic mechanisms
have played a role in population diversity and divergence? (e) Does
phenotypic variation (body size and colour pattern) reflect the taxo-
nomic diversity detected by genetic data?
We obtained 140 genetic samples of Afrixalus paradorsalis (138 A. p.
paradorsalis and two A. p. manengubensis) from 29 localities across
Cameroon, Gabon, Equatorial Guinea, and Republic of the Congo
(Figure 1). Based on the current understanding of phylogenetic rela-
tionships within the genus Afrixalus (Portik, 2015), we included two
samples of A. osorioi as an outgroup in our mtDNA analyses.Tissue
samples (including liver, muscle, or toe clips) were preserved in 95%
ethanol or RNA later. Specimen vouchers are deposited in natural
history museum collections (Appendix S1).
Molecular data collection
We extracted DNA, collected 16S ribosomal RNA (16S) mitochon-
drial sequence data, and used the double-digest RADseq laboratory
protocol (ddRADseq; Peterson, Weber, Kay, Fisher, & Hoekstra,
2012) to collect genome-wide SNP data as described in
Appendix S3. We processed ddRADseq data using the process_rad-
tags module of the STACKS 1.35 workflow (Catchen, Amores, Hohen-
lohe, Cresko, & Postlethwait, 2011; Catchen, Hohenlohe, Bassham,
Amores, & Cresko, 2013), which demultiplexes pooled reads and per-
forms quality filtering. We then used USTACKS to align short reads and
assemble them into sets of loci, requiring a minimum depth of cover-
age of five reads and a maximum of two discrepancies. We gener-
ated a catalog of consensus loci using CSTACKS, and matched the loci
Volcanic Line
0 km 400 km
Mt Kupe
Mt Manengouba
Bakossi Mtns
Mt Cameroon
Bioko Island
FIGURE 1 Sampling localities of Afrixalus paradorsalis paradorsalis (circles) and A. p. manengubensis (triangles) in the Lower Guinean forests
of continental Central Africa and Bioko Island. Sampling localities are coloured according to the mitochondrial (mtDNA) haplotype groups and
distinct genetic lineages identified in analyses of the ddRADseq (nuDNA) dataset (Figure 2). Symbols with white borders reflect localities with
only mtDNA sequence data and symbols with no border reflect localities with only morphological data. The right panel depicts the locations of
key mountains along the Cameroon Volcanic Line
from USTACKS to this catalog using SSTACKS. The POPULATIONS module
was then used to generate alleles for loci present in 75% of all indi-
viduals, which resulted in a dataset of 3,968 loci. We further pruned
this dataset to retain only variant and biallelic loci. If a locus con-
tained multiple SNPs, a single site was chosen at random and
retained for all subsequent analyses. Our final unlinked SNP dataset
consisted of 3,917 loci for 50 individuals and is available on the
Open Science Framework (OSF) ( The mtDNA
sequences generated for this study are deposited in GenBank (acces-
sion numbers: MH378334-378405).
Phylogeographical structure and divergence
We conducted Bayesian divergence-dating analyses with our mtDNA
dataset (16S) using BEAST 1.8 (Drummond, Suchard, Xie, & Rambaut,
2012). We used MAFFT 5 (Katoh, Kuma, Toh, & Miyata, 2005) with
the E-INS-I algorithm to align 16S sequences. We performed analy-
ses using a GTR substitution model, and a constant size coalescent
tree prior with a strict molecular clock calibrated with an amphibian
rate of 2% per Myr rate of divergence (Crawford, 2003). Analyses
were run for 100 million generations with sampling every 5,000 gen-
erations, producing a total of 20,000 trees. We examined conver-
gence using TRACER 1.6 (Rambaut, Drummond, & Suchard, 2013),
discarded a burn-in of 25% and generated a maximum clade credibil-
ity tree from 15,000 trees.
For our SNP dataset, we performed a discriminant analysis of
principle components to identify genetic clusters of individuals
(DAPC; Jombart, Devillard, & Balloux, 2010) using the R package
adegenet1.8 (Jombart, 2008; Jombart & Ahmed, 2011) imple-
mented in RStudio (0.99.903). We determined the optimal number
of clusters by running a principle components analysis and calculat-
ing the Bayesian information criterion (BIC; Schwarz, 1978) for
sequential k-values after the retention of 100 principle components.
To minimize over-fitting, an initial DAPC was used to find the
a-score for each set of clusters and this value was used to select the
number of principal components to retain in a subsequent re-analysis
(Jombart, 2008; Jombart & Ahmed, 2011). We also performed hier-
archical population-clustering analyses using the maximum likelihood
approach of ADMIXTURE (Alexander, Novembre, & Lange, 2009). We
used five replicates to evaluate up to six discrete populations per
analysis, and subsequently determined the Kvalue with the lowest
cross-validation error. Preliminary results indicated the two samples
of A. p. manengubensis were highly distinct and they were removed
from hierarchical clustering for A. p. paradorsalis.
We investigated the evolutionary relationships among popula-
tions using the SNP data in a coalescent framework with SNAPP 1.3
(Bryant, Bouckaert, Felsenstein, Rosenberg, & RoyChoudhury, 2012)
implemented in BEAST2 2.4 (Bouckaert et al., 2014). The SNAPP model
is based on the coalescent process and therefore assumes that
shared polymorphisms among lineages are due to incomplete lineage
sorting and not gene flow (Bryant et al., 2012). We subsampled each
distinct population (determined by DAPC and clustering analyses) to
include 26 representatives, for a total of 19 samples, estimated
mutation rates (uand v) from the data (0.997 and 1.002, respec-
tively), and fixed the birth rate (k) of the Yule prior to 25. We per-
formed two independent runs with a chain length of one million
generations, sampling every thousand generations. We examined
convergence using TRACER 1.6 (Rambaut et al., 2013) and created a
maximum clade credibility tree from the post-burn-in sample. To
obtain a rough estimate of divergence dates we converted branch
lengths using a human mutation rate of 1 910
(Lynch, 2010), as
no estimate currently exists for amphibians.
Molecular diversity and spatial connectivity
To assess genetic diversity and population divergence in the mtDNA
and SNP data, we used ARLEQUIN 3.5.2 (Excoffier & Schneider, 2005)
to calculate nucleotide diversitynumber of segregating sites (h
and pairwise sequence comparisons (h
), and expected homozygosity
) for the distinct lineages. For the ddRADseq dataset we included
loci that were present in at least 75% of individuals in a lineage. We
also computed pairwise F
for both the mtDNA and SNP data using
group assignments supported by the 16S gene tree, population clus-
tering methods, and the DAPC.
We visualized spatial patterns of gene flow and genetic diversity
& Stephens, 2016), a method that uses genome-wide SNP variation
and locality information to highlight regions where genetic similarity
decays more quickly than expected under isolation by distance. The
number of migration routes and deme sizes is specified through a grid
size, and resistance distance is used to approximate the expected dis-
similarity between two samples. These estimates are interpolated
across geographical space to provide a visualization of levels of gene
flow and diversity across regions. We chose a deme size of 700 and
ran three independent analyses using RUNEEMS_SNPS, with a burn-in of
1,000,000 and MCMC length of 20,000,000. We combined the three
independent runs per deme size, assessed convergence, and generated
surfaces of effective diversity (q) and effective migration rates (m)
using the R EEMS PlotsR package (Petkova et al., 2016).
Demographic modelling
To investigate alternative divergence scenarios for lineages of A. p.
paradorsalis, we used the diffusion approximation method of dadito
analyse two-dimensional joint site frequency spectra (2D-JSFS;
Gutenkunst, Hernandez, Williamson, & Bustamante, 2009). We
examined several models representing forest refugia, riverine barrier
or ecological gradient divergence scenarios using an established 2D
analysis pipeline (Portik et al., 2017), and created a new set of mod-
els to explicitly investigate divergence events for mainland and island
lineages. Before creating our site frequency spectra, we further fil-
tered our ddRADseq data using a minimum minor allele frequency
threshold (0.05) in the POPULATIONS module of STACKS (Catchen et al.,
2011, 2013). For all analyses we projected down samples to reduce
missing data and maximize the number of segregating sites for
analyses, resulting in the following allele numbers: northern, 16 alle-
les; southern, 32 alleles; Bioko Island, 18 alleles. Our small sample
size for A. p. manengubensis prevented us from including this lineage
in the demographic modelling.
To determine whether the joint demographic history between
the northern and southern populations was best captured by diver-
gence resulting from forest refugia, riverine barriers or ecological
gradients, we carried out 2D analyses using a set of 21 demographic
models (Appendix S2). In addition to the 15 models described by
Portik et al. (2017), we defined four, three-epoch forest refugia mod-
els that consist of divergence in isolation, followed by secondary
contact with or without instantaneous size change, followed by iso-
lation. We also included two, two-epoch models of continuous gene
flow that allow different migration rates within each epoch. To
determine if pure vicariance or a founder event better explained the
demographic history of the Bioko Island lineage, we generated addi-
tional types of island demographic models (Appendix S2). The vicari-
ance models did not include changes in population size, whereas
founder event models enforced exponential growth in the island
population. For models in both the vicariance and founder event cat-
egories, we included a variable sthat defines the fraction of the
ancestral population (nuA) founding each daughter population, where
nuA*srepresents the island population and nuA*(1-s) represents the
mainland population. We enforced an upper limit of 0.5 for s, which
constrains the incipient island population at less than 50% of the
ancestral population. To determine if intervals of continuous migra-
tion or discrete admixture events explain additional features of the
2D-JSFS, we defined models that included discrete admixture events
for the vicariance and founder event models, in which a fraction fof
the mainland population is instantaneously present in the post-
admixture island population. These events were placed immediately
after the initial divergence, between two discrete time intervals
allowing for genetic drift, or at the end of a single drift interval.
Although we included a model of ancient migration and a model of
secondary contact for the vicariance scenario, these required inter-
vals of continuous migration, which are less plausible for island sys-
tems than discrete admixture events.
Using threefold perturbed random starting parameters, we per-
formed 50 optimization replicates using the NelderMead method (op-
timize_log_fmin) with a maximum of 20 iterations for each of the
models included in the 2D model population comparison sets. Each
optimized parameter set was used to simulate the 2D-JSFS, and the
log-likelihood of the 2D-JSFS given the model was estimated using
the multinomial approach. We used the best scoring replicate of each
model to select starting parameters for a second round of 50 replicate
optimizations involving twofold parameter perturbations, and the
parameter values from the best replicate were subsequently used to
generate onefold perturbed starting parameters for a final set of 100
replicate optimizations. All newly created models for our analyses are
available at: We compared
the results from our 2D modelling using the Akaike information crite-
rion (AIC), and the replicate with the highest likelihood for each model
was used to calculate AIC scores, DAIC scores, and Akaike weights (x
(Burnham & Anderson, 2002). We did not transform raw parameters
using a mutation rate because our primary aim was to perform model
selection, and parameter values should ideally be obtained through
bootstrapping to produce confidence intervals (Gutenkunst et al.,
2009). We provide the estimate of h(h=4N
lL, where L is the total
length of sequenced region SNPs were ascertained from) the effective
mutation rate of the reference population, which here corresponds to
the ancestral population.
Morphological and colour pattern variation
Regional differences in both body size and colour pattern are docu-
mented throughout the range of A. paradorsalis in Cameroon (Amiet,
2009). We collected and analyzed body size data and 14 linear mor-
phological measurements for adult specimens of A. p. paradorsalis
and A. p. manengubensis as described in Appendix S3. We also char-
acterized and examined the phylogenetic distribution of dorsal
colouration as described in Appendix S3, and image data are avail-
able from:
Phylogeographical structure and divergence
We recovered three distinct lineages in our mtDNA dataset that cor-
respond to A. p. manengubensis, northern A. p. paradorsalis (north of
the Sanaga River), and southern A. p. paradorsalis (south of the
Sanaga River; Figure 2a). The Bioko Island population forms a dis-
tinct group nested within northern A. p. paradorsalis and we find
strong support for A. p. manengubensis as sister to the northern and
Bioko Island clade. The main divergence event between the northern
and southern lineages is estimated to have occurred in the Pleis-
tocene at 2.22 Ma (95% highest posterior density region [HPD]
1.532.99 Ma), with subsequent divergence between the northern
lineage and A. p. manengubensis approximately 1.56 Ma (0.992.19,
95% HPD; Table 1). The estimated times at which lineages in the
Bioko Island population coalesce is approximately 100 ka (40
230 ka, 95% HPD; Table 1).
In our DAPC analyses of 3,917 unlinked SNPs, we detected four
genetically distinct populations that correspond to A. p. manenguben-
sis and three populations of A. p. paradorsalis (Figure 2c). We
inferred the same three populations of Afrixalus p. paradorsalis with
our hierarchical maximum likelihood population clustering analyses
(Figure 2d). These populations correspond to the northern and
southern lineages recovered in the mtDNA analyses and a distinct
Bioko Island lineage. Our Bayesian phylogenetic analysis of genome-
wide SNP variation using a coalescent model recovered the same lin-
eage relationships as the mtDNA analyses (Figure 2b), and with high
support for all nodes (>95% posterior probability). Based on a human
mutation rate, the divergence between the northern and southern
lineages is estimated at 2.68 Ma (2.203.15, 95% HPD), and A. p.
manengubensis diverged from the northern lineage approximately
0 Ma123
Pliocene Pleistocene
A. p. manengubensis
0 Ma123
Pliocene Pleistocene
A. p. manengubensi
2.68 0.42
North Bioko Id
A. p. manengubensis
North Bioko IslandSouth
0.45 0.500.55
K Value =
Cross-Validation Error
1234 5
0.30 0.35 0.40
K Value =
Cross-Validation Error
12 345
(c) (d)
A. p. paradorsalis A. p. paradorsalis
A. p. paradosalis
FIGURE 2 Chronograms of the Afrixalus paradorsalis complex from (a) the BEAST analysis of mtDNA (16s) sequence data calibrated with a
2% per Myr divergence rate, and (b) Bayesian coalescent analysis of 3,917 SNPs using SNAPP. Median ages are provided above nodes, with
error bars representing the 95% HPD shown, and nodes with high support are filled black (posterior probability >0.95). Population
identification and assignment results from analyses of 3,917 SNPs using (c) a discriminant analysis of principal components of 48 A. p
paradorsalis (11 northern, 23 southern and 14 Bioko) and two A. p. manengubensis and (d) maximum likelihood population clustering of 48 A. p.
paradorsalis (two samples of A. p. manengubensis not shown) using ADMIXTURE. The colour scheme matches the sampling localities depicted in
Figure 1
TABLE 1 Summary of divergence date estimates for key nodes in the mitochondrial phylogeography
Northern+Southern A. p. manengubensis TMRCA North North+Bioko TMRCA Bioko TMRCA South
2.22 (1.532.99) 1.56 (0.992.19) 0.41 (0.200.69) 0.22 (0.100.42) 0.10 (0.040.23) 0.77 (0.421.21)
Notes. The median date estimates for a node are given in millions of years, along with associated 95% highest posterior density region.
1.14 Ma (0.891.43, 95% HPD). Finally, the northern and Bioko lin-
eages are estimated to have shared a common ancestor 420 ka
(330540 ka, 95% HPD).
Molecular diversity and spatial connectivity
We found very little diversity in mtDNA or nuDNA in the A. p. manen-
gubensis lineage; however, this is likely due to our small sample size
for A. p. manengubensis (Table 2). By contrast, our sampling for the
Bioko Island population was on par with that of the northern and
southern populations of A. p. paradorsalis and yet our estimates of
genetic diversity were much lower in the island lineage relative to the
continental lineages (Table 2). Estimates of pairwise F
indicate that
the A. p. manengubensis lineage is highly divergent from the northern
lineage of A. p. paradorsalis on the basis of mitochondrial (F
and nuclear sequence data (F
=0.42; Table S3.1). Within the three
A. p. paradorsalis lineages, the northern and southern lineages are also
well differentiated (mtDNA F
=0.82, nuDNA F
=0.54), whereas
the Bioko lineage is only moderately differentiated from the northern
lineage (mtDNA F
=0.51, nuDNA F
=0.23; Table S3.1).
The EEMS analysis highlighted a clear migration barrier along the
Sanaga River, starting in northeast Cameroon and following the river
diagonally to its mouth at the Bay of Douala (Figure 3a). Likewise,
there is a large area of reduced migration in central Gabon that
roughly coincides with the Ogoou
e-Ivindo Rivers. We detected
exceptionally high genetic diversity along a substantial portion of the
CVL and in Gabon at the intersection of the Ogoou
e-Ivindo Rivers
(Figure 3b), and we found comparatively lower genetic diversity in
the Bioko Island lineage.
Demographic modelling
For the northern and southern pairwise comparison, we found
unambiguous support for two refugial models (99.9% of the total
model weight) that involve divergence in isolation followed by popu-
lation size expansion and secondary contact (Table S3.2), The models
differ only in whether they include one or two migration rates, and
we found stronger support for the model with a single migration rate
between the populations (DAIC =3.4, x
=0.844; Figure 4a). The
model with two migration rates estimated greater migration from
the northern to the southern population (Table S3.2). The initial sizes
of the northern and southern populations were comparable, but the
northern population became larger than the southern population fol-
lowing the period of demographic expansion. We found the three-
epoch refugial models, which involve divergence in isolation, sec-
ondary contact, and contemporary isolation, were generally a poor
fit (DAIC range =261.2285.2). These results strongly suggest that
after an initial divergence period in isolation, gene flow has resumed
between the northern and southern regions.
In our investigation of the northern and Bioko Island populations,
we found that models of pure vicariance provided a substantially
better fit to our data than the founder event models (Table S3.3).
The best overall model consists of vicariance and genetic drift, fol-
lowed by a late discrete admixture event (DAIC =3.0, x
=0.79; Fig-
ure 4b). Across all the vicariance models included in our analyses,
the point estimates for the proportion of the ancestral population
founding the Bioko Island population ranged from 27.736.6%
(Table S3.3) and our results demonstrate that Bioko Island likely
maintained a stable population size, rather than experiencing expo-
nential growth associated with founder events. In the top-ranked
model, we estimated that the fraction of the mainland (northern)
population present in the post-admixture island population (f) was
quite low (1.5%, Table S3.3). These results indicate that if a post-
vicariance exchange did occur between the mainland and Bioko, the
extent of this exchange was minimal. This notion is further sup-
ported by the second-best model, which consists of vicariance with-
out any migration or admixture events (x
Morphological and colour pattern variation
A summary of all measurements is provided in Table S3.4. For
both males and females, most of the total variance in morphology
was captured in the first two principal components (66.9% and
88.7% for males and females, respectively; Tables S3.5, S3.6). The
first principal component (PC1) loaded heavily on SUL indicating
that differences in body size were responsible for most of the
variance (Tables S3.5, S3.6). The second principal component
(PC2) loaded heavily on characters pertaining to leg morphology
(TL, THL, FL; Tables S3.5, S3.6). Both male and female A. p.
manengubensis are strongly differentiated from all populations of
A. p. paradorsalis in our principal components analysis of morpho-
logical variation and in the more extensive SUL dataset (Figure 5).
TABLE 2 Summary statistics for mitochondrial locus (16s) and nuclear SNPs (ddRADseq) collected from the Afrixalus paradorsalis complex
16s Nuclear SNPs
bp hshpNLoci N
A. p. paradorsalis south 37 17 495 0.0082 0.0108 23 3,310 1.572 0.576 0.0732 0.0789
A. p. paradorsalis Bioko 15 2 495 0.0000 0.0002 14 3,448 1.163 0.164 0.0477 0.0501
A. p. paradorsalis north 19 8 495 0.0058 0.0041 11 3,109 1.315 0.318 0.0759 0.0821
A. p. manengubensis 2 1 495 0.0000 0.0000 2 2,543 0.835 0.051 0.0211 0.0216
N: number of individuals sampled; N
: number of haplotypes; bp: sequence length in base pairs; hs: genetic diversity based on number of segregating
sites; hp: genetic diversity based on pairwise sequence comparisons; loci: number of loci with <25% missing data within the lineage; P: proportion of
polymorphic sites; N
: allelic richness; H
: expected heterozygosity; h
: genetic diversity based on expected homozygosity.
Effective Migration Surface Effective Diversity Surface
log(m) log(q)
FIGURE 3 Contour maps representing the posterior mean of (a) effective migration surface and (b) effective diversity surface, for 3,917
SNPs collected from 48 samples of Afrixalus. p. paradorsalis (A. p. manengubensis excluded). In (a), blue colours represent areas of high
migration, or dispersal corridors, whereas orange regions represent areas of low migration, or dispersal barriers. In (b), white colour indicates
areas of lower than expected genetic diversity, and dark purple colouration represents higher levels of genetic diversity. Sampling locality
symbols are scaled according to the number of samples in a merged locality
018Bioko Id
018Bioko Id
018Bioko Id
ResidualsDemographic Model
Bioko IdNorth
nu1 nu2
nuA*(1-s) (nuA*s)
nu1b nu2b
nu1a nu2a
FIGURE 4 Population genetic model comparisons for (a) north (n=8) and Bioko Island (n=9), and (b) north (n=8) and south (n=16)
Afrixalus. p. paradorsalis using the two-dimensional joint site frequency spectrum (2D-JSFS) and the filtered ddRADseq SNP dataset (2,205
SNPs). A simplified visual representation of the best-fit model is depicted, along with comparisons of the 2D-JSFS for the data, the model, and
resulting residuals
Among populations of A. p. paradorsalis, pattern type 1 is more
common in the Bioko lineage, with only one out of 15 individuals
exhibiting pattern type 2 (Figure S3.1). Pattern types 1 and 2 are
common in both the northern and southern populations of A. p.
paradorsalis, and occur in the same frequency in A. p. manen-
gubensis, and the distribution of pattern types across the phy-
logeny revealed no clear relationship with population identity or
Sky island diversification along the
Cameroonian Volcanic Line
We found that Afrixalus p. manengubensis, the submontane lineage of
the A. paradorsalis complex on Mount Manengouba and the adjacent
Bakossi Mountains, is differentiated from lowland rain forest
SUL (mm)
A. p. manengubensis Bioko North South
A. p. manengubensis Bioko North South
SUL (mm)
A. p. paradorsalis
(n=48) (n=32) (n=39)(n=4)
A. p. paradorsalis
(n=2) (n=11) (n=3)(n=1)
−0.4 −0.2 0.0 0.2
−0.25 0.00 0.25 0.50
Males Females
(a) (b)
FIGURE 5 Bivariate ordination of first two components from a principal components analysis of 15 linear measurements for male (a) and
female (b) Afrixalus p. manengubensis and Afrixalus p. paradorsalis. Variation in body size (snout-urostyle length) from the more extensive sample
size of male (c) and female (d) Afrixalus p. manengubensis and Afrixalus p. paradorsalis
populations of the complex (A. p. paradorsalis) in both our mtDNA
and nuDNA datasets. Mt. Manengouba formed ~1 Ma (Marzoli et al.,
2000), with the oldest geochemical samples from the mountain esti-
mated at 1.55 Ma (Fitton & Dunlop, 1985; Gouhier, Nougier, & Nou-
gier, 1974), and this period of mountain formation roughly coincides
with our estimate of 1.56 Ma (0.992.19 Ma) for divergence
between A. p. manengubensis and A. p. paradorsalis. Four additional
anuran species are thought to be endemic to Mt. Manengouba (Car-
dioglossa manengouba, C. trifasciata, Leptodactylodon erythrogaster,
and Phrynobatrachus manengoubensis), and several other amphibians
occur exclusively on Mt. Manengouba and adjacent mountains (Mt.
Kupe, Mt. Nlonako, Bakossi Mts.) (Amiet, 1975; Blackburn, 2008;
Herrmann et al., 2005; Portik et al., 2016; Schmitz, Euskirchen, &
ohme, 1999). Comparative evidence from other taxonomic groups
is limited, but elevated genetic diversity or patterns of divergence
across the CVL have been documented in plants (Budde, Gonz
ınez, Hardy, & Heuertz, 2013; Hardy et al., 2013), birds (Smith
et al., 2000), and chameleons (Barej et al., 2010). Together, this sug-
gests that endemic diversity in this mountainous region has accumu-
lated in a relatively short period of timea hypothesis that can be
further tested by estimating divergence times of other co-occurring
endemics. Divergence between submontane and lowland species
may reflect allopatric speciation following the formation of eleva-
tional barriers to dispersal during uplift of the Cameroon Volcanic
Line (CVL). Alternatively, disruptive selection mirroring environmental
variation along the elevational gradient may drive diversification
through ecological speciation (Zhen et al., 2017). Unfortunately, due
to our small sample size we were unable to include A. p. manen-
gubensis in our demographic modelling approach to differentiate
between allopatric and parapatric models of speciation. Dense sam-
pling along ecological and altitudinal gradients across the mountain
ranges coupled with demographic models may provide insights into
the role of gene flow and disruptive selection in driving divergence
between the submontane and lowland lineages of A. paradorsalis.
The significantly smaller mean SUL of both male and female A. p.
manengubensis individuals is characteristic of many submontane and
montane anurans, including the Mt. Manengouba/Bakossi Mts. ende-
mic P. manengubensis, which is also smaller than its nearest relatives
(Zimkus & Gvo
ık, 2013). Smaller body size may arise in montane
populations due to reduced resource availability and slower growth
rates in highland ecosystems (Berven, 1982; Licht, 1975); however,
trends in anuran body size with increasing elevation vary with
respect to life history and other taxon-specific factors (Ma, Tong, &
Lu, 2009). In anurans, differences in male advertisement calls are
considered one of the most important premating barriers to repro-
duction, and the dominant frequency of the call a male produces is
inversely correlated with body size (Gingras, Boeckle, Herbst, &
Fitch, 2013). Consequently, reduced body size in highland popula-
tions may reflect selection for higher frequency calls, potentially as a
response to elevation-induced ecological or environmental variation
such as differences in ambient noise (Hoskin, James, & Grigg, 2009).
Alternatively, because of the strong relationship between body size
and dominant frequency, differences in mating signal may arise as a
by-product of adaptive changes in body size and may ultimately lead
to reproductive isolation. Future studies characterizing the advertise-
ment calls of A. p. manengubensis and A. p. paradorsalis across habi-
tats (e.g., Kirschel et al., 2011), as well as differences in the acoustic
environment, may clarify whether these mechanisms are related to
body size differences between montane and lowland populations.
Strong support for climatic refugia underlying
lineage divergence in lowland tropical forests
The distribution of A. p. paradorsalis in Lower Guinea encompasses
several major landscape features, including the CVL, Sanaga River,
and Ogoou
e-Ivindo rivers, multiple proposed historical forest refugia
(Maley, 1996), and two major forest ecotones related to precipitation
and seasonal inversion. We recovered two genetically distinct allopa-
tric populations of A. p. paradorsalis in Lower Guinea through model-
based and nonparametric clustering methods (northern, southern;
Figure 2c,d), and found evidence for decreased effective migration
occurring along the putative population boundary that overlaps with
the Sanaga River (Figure 3a). We did not find any additional popula-
tion structuring coinciding with the CVL, Ogoou
e-Ivindo rivers or
forest-specific ecotones, suggesting that these landscape features
have not played a significant role in the diversification of A. p.
To determine whether divergence of the northern and southern
populations resulted from limited dispersal across the Sanaga River
or from climate-induced forest contractions, we assessed demo-
graphic models derived from expected population demographic
responses to riverine barriers and climatic refugia. Our demographic
model selection supported a refugial model involving divergence in
isolation, followed by size expansion in both populations accompa-
nied by gene flow (Figure 4a; Table S3.2). Given the results of con-
temporary gene flow across the Sanaga River, an isolation by river
scenario for these lineages is only feasible if major historical changes
occurred in course or flow of the Sanaga River surrounding the early
Pleistocene divergence time between the lineages. A combination of
evidence from offshore fluvial paleodrainage systems (Ngueutchoua
& Giresse, 2010) and divergence-dating estimates of endemic fresh-
water fishes (Day et al., 2013; Goodier, Cotterill, ORyan, Skelton, &
de Wit, 2011; Pinton, Agn
ese, Paugy, & Otero, 2013) indicates that
recent changes in the course of the Sanaga are unlikely, and there-
fore the initial isolation of A. p. paradorsalis populations is better
explained by shifts in forest cover that occurred in this time period
(Anhuf et al., 2006; Cowling et al., 2008; Maley, 1996; deMenocal,
2004). Our results parallel those of the sympatric Gaboon Forest
Frog (Scotobleps gabonicus), including temporal overlap in the forest-
refugia-driven divergence of the northern and southern populations
across these species (A. p. paradorsalis: 2.22 Ma, 1.532.99 Ma 95%
HPD; S. gabonicus: 2.97 Ma, 2.613.34 Ma 95% HPD; Portik et al.,
2017). If forest-refugia continue to emerge as a widespread mecha-
nism of divergence, investigating the synchronicity of divergence
times across taxa could statistically demonstrate whether species
display a shared response to key climate change events.
Marine incursions underlie vicariance and
population divergence on Bioko Island
The Bioko population of A. p. paradorsalis is moderately genetically
differentiated from the northern population of A. p. paradorsalis in
both our mtDNA and SNP datasets, indicating that the island popu-
lations are genetically isolated from their mainland counterparts. We
recovered a late Pleistocene origin for the monophyletic Bioko lin-
eage (TMRCA 40230 ka based on mtDNA). This divergence time
estimate overlaps with those obtained for the Bioko Island popula-
tions of Hyperolius ocellatus (30290, and 90460 ka) and H. tubercu-
latus (40210 ka), and the three sets of estimates coincide with
periods of land bridge connectivity between Bioko and the continent
(Bell et al., 2017). Although amphibians have dispersed overseas to
colonize the oceanic islands in the Gulf of Guinea archipelago (Bell,
Drewes, Channing, et al., 2015; Measey et al., 2007), relatively
recent divergence and moderate genetic diversity in Biokos amphib-
ian populations (relative to those on S~
ao Tom
e and Pr
ıncipe; Bell,
Drewes, & Zamudio, 2015) are more consistent with vicariance due
to marine incursions than founder events resulting from overseas
dispersal (Bell et al., 2017). Furthermore, the best-fit demographic
model strongly supports a scenario of vicariance by marine incursion,
rather than overseas dispersal (Figure 4b, Table S3.3). Our demo-
graphic modelling also revealed support for a discrete admixture
pulse, in which a small proportion of the mainland population
entered the Bioko Island population in the recent past (comprising
~1.5% of the Bioko population, Table S3.3). Although we found sup-
port for an exchange between the island and mainland, it is quite
limited in extent, and models involving more extensive islandmain-
land interactions (such as periods of continuous gene flow) were a
poor fit (Table S3.3). The monophyly of mtDNA haplotypes on Bioko
Island also supports a pattern of limited exchange, as multiple colo-
nization events would be expected to produce a pattern of para-
phyly with the mainland clade (Figure 2a).
Peripatric speciation may occur between island and mainland
populations of A. p. paradorsalis if the two lineages undergo substan-
tial divergence due to independent selective pressures or genetic
drift prior to the next period of connectivity, leading to reproductive
isolation even if sympatry is eventually restored. Alternatively, in the
absence of reproductive isolation, introgression may occur during
secondary contact, and genomic and phenotypic differentiation may
erode as a result. Though we did not find significant morphological
differences between island and mainland populations of A. p.
paradorsalis, the relative rarity of Pattern Type 2 on Bioko (Fig-
ure S3.1) may reflect a difference in allele frequency between these
two groups, particularly if dorsal colour pattern follows a single locus
mode of inheritance in which each pattern is encoded by a separate
allele, as demonstrated in Eleutherodactylus coqui (ONeill & Beard,
2010). In addition, although Bioko Island and northern A. p. parador-
salis males do not differ in body size and therefore likely produce
advertisement calls with similar dominant frequencies, a recent study
of advertisement call evolution in island reed frogs indicates that clo-
sely related, allopatric species may differ dramatically in the number
of pulses a male produces per call (Gilbert & Bell, 2018). Conse-
quently, future efforts to characterize advertisement calls across
island and mainland populations of A. p. paradorsalis may recover
early stages of divergence in this important component of species
recognition and mate choice.
The Cameroon Ministry of Forests and Wildlife (MINFOF) and Min-
istry of Scientific Research and Innovation (MINRESI) provided nec-
essary permits for conducting research and exportation to D.C.B.,
D.M.P., G.F.M.J., M.T.K., and V.G. Fieldwork in Cameroon was sup-
ported by National Science Foundation grant DEB # 1202609 to
D.C.B., and under the approval of the Institutional Animal Care and
Use Committee (2014-2) at the California Academy of Sciences. For
assistance with aspects of fieldwork in Cameroon, we thank B.
Evans, S. Menzepoh, L. Scheinberg, B. Freiermuth, W.P.T. Nkonme-
neck, D. Dornk, D. Fotibu, and M. Lebreton. For fieldwork in Gabon
we thank the CENAREST, ANPN, and the Direction de la Faune et
des A
eres Prot
ees for permits, the Wildlife Conservation Society
Gabon Program for logistical support, and B. Stuart, N. Emba-Yao, F.
Moiniyoko, B. Hylayre, E. Ekomy, A. Dibata, T. Ogombet, U. Eyagui,
P. Endazokou, for assistance in the field. For fieldwork in Equatorial
Guinea, supported by National Science Foundation grant DEB #
1309171 to R.C.B., we thank Universidad Nacional de Guinea Equa-
torial and Jose Manuel Esara Echube for permits, the Bioko Biodiver-
sity Protection Program, ExxonMobil Foundation, and Mobil
Equatorial Guinea Inc. for logistical support, and A. Fertig and P.
McLaughlin for assistance in the field. V.G. would like to thank M.
Dolinay, E.B. Fokam, and M. Jirk
u for assistance in the field, logistical
support and additional material. His research was supported by the
Czech Science Foundation (GACR # 15-13415Y), and Ministry of
Culture of the Czech Republic (DKRVO 2017/15, National Museum,
00023272). Fieldwork in the Republic of Congo was part of a rapid
biodiversity initiative, commissioned by Flora Fauna & Man, Ecologi-
cal Services Ltd (FFMES), with J. Gaugris of FFMES conducting the
study organization and design, and permits issued by the Groupe
dEtude et de Recherche sur la Diversit
e Biologique. We thank M.
McElroy for assistance with ddRADseq data collection, and D. Tar-
box, N. Knight, F. Tillack, and A.-G. Zassi-Boulou for contributing
SUL measurements and photographs of dorsal colouration. This pro-
ject was funded by a National Science Foundation DDIG (DEB:
1311006) awarded to D.M.P., a Gaige Award from the American
Society of Ichthyologists and Herpetologists awarded to K.L.C., and
National Science Foundation grants (DEB: 1457232; DEB: 1456098)
awarded to M.K.F. and A.D.L. This work used the Vincent J. Coates
Genomics Sequencing Laboratory at UC Berkeley, supported by NIH
S10 OD018174 Instrumentation Grant.
We developed a project page ( using the Open
Science Framework that includes our ddRADseq haplotypes format
files (, and the input data, analysis instructions, and
results files for BEAST (, EEMS (, SNAPP (
b7395), ADMIXTURE ( g), and dadi( All newly
created demographic models for our analyses are incorporated into
an updated version of the model-testing pipeline of Portik et al.
(2017), freely available at:
All mitochondrial sequences generated for this project are deposited
in GenBank (accession numbers: MH378334-378405).
Rayna C. Bell
David C. Blackburn
Adam D. Leach
Daniel M. Portik
Alexander, D. H., Novembre, J., & Lange, K. (2009). Fast model-based
estimation of ancestry in unrelated individuals. Genome Research,19,
Amiet, J.-L. (1975). Ecologie et distribution des Amphibiens Anoures de
la region de Nkongsamba (Cameroun). Annales de la Facult
e des
Sciences de Yaound
Amiet, J.-L. (2009). Observations sur les Afrixalus du Cameroun (Amphib-
ia, Anura, Hyperoliidae). Revue Suisse de Zoologie,116,5392.
Anhuf, D., Ledru, M.-P., Behling, H., Da Cruz Jr., F. W., Cordeiro, R. C.,
Van der Hammen, T., ... Da Silva Dias, P. L. (2006). Paleo-environ-
mental change in Amazonian and African rainforest during the LGM.
Palaeogeography, Palaeoclimatology, Palaeoecology,239, 510527.
Barej, M. F., Ineich, I., Gvozd
ık, V., Lhermitte-Vallarino, N., Gonwouo, N.
L., LeBreton, M., ... Schmitz, A. (2010). Insights into chameleons of
the genus Trioceros (Squamata: Chamaeleonidae) in Cameroon, with
the resurrection of Chamaeleon serratus Mertens, 1922. Bonn Zoologi-
cal Bulletin,57, 211229.
Barej, M. F., R
odel, M. O., Loader, S. P., Menegon, M., Gonwouo, N. L.,
Penner, J., ... Schmitz, A. (2014). Light shines through the spindrift
the phylogeny of African Torrent Frogs (Amphibia, Anura, Petropede-
tidae). Molecular Phylogenetics and Evolution,71, 261273. https://d
Bell, R. C., Drewes, R. C., Channing, A., Gvo
ık, V., Kielgast, J., L
S., ... Zamudio, K. R. (2015). Overseas dispersal of Hyperolius reed
frogs from Central Africa to the oceanic islands of S~
ao Tom
e and
ıncipe. Journal of Biogeography,42,6575.
Bell, R. C., Drewes, R. C., & Zamudio, K. R. (2015). Reed frog diversifica-
tion in the Gulf of Guinea: Overseas dispersal, the progression rule,
and in situ speciation. Evolution,64, 904915.
Bell, R. C., Parra, J. L., Badjedjea, G., Barej, M. F., Blackburn, D. C., Bur-
ger, M., ... Zamudio, K. R. (2017). Idiosyncratic responses to climate-
driven forest fragmentation and marine incursions in reed frogs from
Central Africa and the Gulf of Guinea Islands. Molecular Ecology,26,
Berven, K. A. (1982). The genetic basis of altitudinal variation in the wood
frog Rana sylvatica II. An experimental analysis of larval development.
Oecologia,52, 360369.
Blackburn, D. C. (2008). A new species of Cardioglossa (Amphibia: Anura:
Arthroleptidae) endemic to Mount Manengouba in the Republic of
Cameroon, with an analysis of morphological diversity in the genus.
Zoological Journal of the Linnean Society,154, 611630. https://doi.
Bouckaert, R., Heled, J., K
uhnert, D., Vaughan, T., Wu, C.-H., Xie, D., ...
Drummond, A. J. (2014). BEAST 2: A software platform for Bayesian
evolutionary analysis. PLOS Computational Biology,10, e1003537.
Bryant, D., Bouckaert, R., Felsenstein, J., Rosenberg, N. A., & RoyChoud-
hury, A. (2012). Inferring species trees directly from biallelic genetic
markers: Bypassing gene trees in a full coalescent analysis. Molecular
Biology and Evolution,29, 19171932.
Budde, K. B., Gonz
ınez, S. C., Hardy, O. J., & Heuertz, M.
(2013). The ancient tropical rainforest tree Symphonia globulifera L. f.
(Clusiaceae) was not restricted to postulated Pleistocene refugia in
Atlantic Equatorial Africa. Heredity,111,6676.
Burnham, K. P., & Anderson, D. R. (2002). Model selection and multimodel
inference: A practical information-theoretic approach. New York:
Catchen, J. M., Amores, A., Hohenlohe, P., Cresko, W., & Postlethwait, J.
H. (2011). Stacks: Building and genotyping loci de novo from short-
read sequences. G3: Genes, Genomes, Genetics,3, 171182.
Catchen, J. M., Hohenlohe, P. A., Bassham, S., Amores, A., & Cresko, W.
A. (2013). Stacks: An analysis tool set for population genomics.
Molecular Ecology,22, 31243140.
Cowling, S. A., Cox, P. M., Jones, C. D., Maslin, M. A., Peros, M., & Spall,
S. A. (2008). Simulated glacial and interglacial vegetation across
Africa: Implications for species phylogenies and trans-African migra-
tion of plants and animals. Global Change Biology,14, 827840.
Crawford, A. J. (2003). Relative rates of nucleotide substitution in frogs.
Journal of Molecular Evolution,57, 636641.
Day, J. J., Peart, C. R., Brown, K. J., Friel, J. P., Bills, R., & Moritz, T.
(2013). Continental diversification of an African catfish radiation
(Mochokidae: Synodontis). Systematic Biology,62, 351365. https://d
deMenocal, P. B. (2004). African climate change and faunal evolution
during the Pliocene-Pleistocene. Earth and Planetary Science Letters,
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, 19691973.
Excoffier, L. G. L., & Schneider, S. (2005). Arlequin ver. 3.0: An integrated
software package for population genetics data analysis. Evolutionary
Fitton, J. G., & Dunlop, H. M. (1985). The Cameroon Line, West Africa,
and its bearing on the origin of oceanic and continental alkalic basalt.
Earth and Planetary Science Letters,72,2338.
Gilbert, C. M., & Bell, R. C. (2018). Evolution of advertisement calls in an
island radiation of African reed frogs. Biological Journal of the Linnean
Gingras, B., Boeckle, M., Herbst, C. T., & Fitch, W. T. (2013). Call acous-
tics reflect body size across four clades of anurans. Journal of Zoology,
2, 143150.
Goodier, S. A. M., Cotterill, F. P. D., ORyan, C., Skelton, P. H., & de Wit,
M. J. (2011). Cryptic diversity of African Tigerfish (genus Hydrocynus)
reveals palaeogeographic signatures of linked Neogene geotectonic
events. PLoS ONE,6, e28775.
Gouhier, J., Nougier, J., & Nougier, D. (1974). Contribution a l
etude vol-
canologique du Cameroun (Ligne du CamerounAdamaoua).Annales
de la Facult
e des Sciences, Universit
e de Yaound
e, Cameroun,17,69
Gutenkunst, R. N., Hernandez, R. D., Williamson, S. H., & Bustamante, C.
D. (2009). Inferring the joint demographic history of multiple popula-
tions from multidimensional SNP frequency data. PLoS Genetics,5,
Haffer, J. (1997). Alternative models of vertebrate speciation in Amazo-
nia: An overview. Biodiversity and Conservation,6, 451477. https://d
Hardy, O. J., Born, C., Budde, K., Da
ınou, K., Dauby, G., Duminil, J., ...
Poncet, V. (2013). Comparative phylogeography of African rain forest
trees: A review of genetic signatures of vegetation history in the Gui-
neo-Congolian region. Comptes Rendus Geoscience,345, 284296.
Herrmann, H.-W., B
ohme, W., Herrmann, P. A., Plath, M., Schmitz, A., &
Solbach, M. (2005). African biodiversity hotspots: The amphibians of
Mt Nlonako, Cameroon. Salamandra,41,6181.
Heuertz, M., Duminil, J., Dauby, G., Savolainen, V., & Hardy, O. J.
(2014). Comparative phylogeography in rainforest trees from Lower
Guinea. PLoS ONE,9, e84307.
Hoskin, C. J., James, S., & Grigg, G. C. (2009). Ecology and taxonomy-dri-
ven deviations in the frog call-body size relationship across the
diverse Australian frog fauna. Journal of Zoology,278,3641.
Jombart, T. (2008). Adegenet: A R package for the multivariate analysis of
genetic markers. Bioinformatics,24, 14031405.
Jombart, T., & Ahmed, I. (2011). Adegenet 1.3-1: New tools for the analy-
sis of genome-wide SNP data. Bioinformatics,27, 30703071.
Jombart, T., Devillard, S., & Balloux, F. (2010). Discriminant analysis of
principal components: A new method for the analysis of genetically
structured populations. BMC Genetics,11, 94.
Jones, P. J. (1994). Biodiversity in the Gulf of Guinea: An overview. Biodi-
versity and Conservation,3, 772784.
Katoh, K., Kuma, K., Toh, H., & Miyata, T. (2005). MAFFT version 5:
Improvement in accuracy of multiple sequence alignment.
Nucleic Acids Research,33, 511518.
Kirschel, A. N. G., Slabbekoorn, H., Blumstein, D. T., Cohen, R. E., de
Kort, S. R., Buermann, W., & Smith, T. B. (2011). Testing alterna-
tive hypotheses for evolutionary diversification in an African
songbird: Rainforest refugia versus ecological gradients. Evolu-
tion,65, 31623174.
e, A. D., Fujita, M. K., Minin, V. N., & Bouckaert, R. R. (2014). Spe-
cies delimitation using genome-wide SNP data. Systematic Biology,
63, 534542.
Leroux, M. (1983). Le Climat de IAfrique Tropicale, Vol. 1, 2. Paris: Cham-
pion Slatkine.
Licht, L. E. (1975). Comparative life history features of the western spot-
ted frog, Rana pretiosa, from low- and high-elevation populations.
Canadian Journal of Zoology,53, 12541257.
Lynch, M. (2010). Rate, molecular spectrum, and consequences of human
mutation. Proceedings of the National Academy of Sciences USA,107,
Ma, X., Tong, L., & Lu, X. (2009). Variation of body size, age structure
and growth of a temperate frog, Rana chensinensis, over an
elevational gradient in Northern China. Amphibia-Reptilia,30, 111
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,
Marzoli, A., Piccirillo, E. M., Renne, P. R., Bellieni, G., Iacumin, M., Nyobe,
J. B., & Tongwa, A. T. (2000). The Cameroon Volcanic Line Revisited:
Petrogenesis of continental basaltic magmas from lithospheric and
asthenospheric mantle sources. Journal of Petrology,41,87109.
Measey, G. J., Vences, M., Drewes, R. C., Chiari, Y., Melo, M., &
Bourles, B. (2007). Freshwater paths across the ocean: Molecular
phylogeny of the frog Ptychadena newtoni gives insights into
amphibian colonization of oceanic islands. Journal of Biogeography,
Moritz, C., Patton, J. L., Schneider, C. J., & Smith, T. B. (2000). Diversifi-
cation of rainforest faunas: And integrated molecular approach.
Annual Review of Ecology and Systematics,31, 533563. https://doi.
Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. A. B., &
Kent, J. (2000). Biodiversity hotspots for conservation priorities. Nat-
ure,403, 853858.
Ngueutchoua, G., & Giresse, P. (2010). Sand bodies and incised
valleys within the Late Quaternary Sanaga-Nyong delta complex on
the middle continental shelf of Cameroon. Marine and Petroleum Geol-
ogy,27, 21732188.
Oates, J. F., Bergl, R. A., & Linder, J. M. (2004). Africas Gulf of Guinea
forests: Biodiversity patterns and conservation priorities. Advances in
Applied Biodiversity Science,6,190.
Olson, D. M., Dinerstein, E., Wikramanayake, E. D., Burgess, N. D., Pow-
ell, G. V. N., Underwood, E. C., ... Kassem, K. R. (2001). Terrestrial
ecoregions of the World: A new map of life on Earth. BioScience,51,
ONeill, E. M., & Beard, K. H. (2010). Genetic basis of a color pattern
polymorphism in the Coqui frog Eleutherodactylus coqui.Journal of
Heredity,101, 703709.
Peterson, B. K., Weber, J. N., Kay, E. H., Fisher, H. S., & Hoekstra, H.
E. (2012). Double digest RADseq: An inexpensive method for de
novo discovery and genotyping in model and non-model species.
PLoS ONE,7, e37135.
Petkova, D., Novembre, J., & Stephens, M. (2016). Visualizing spatial pop-
ulation structure with estimated effective migration surfaces. Nature
Pinton, A., Agn
ese, J.-F., Paugy, D., & Otero, O. (2013). A large-scale phy-
logeny of Synodontis (Mochokidae, Siluriformes) reveals the influence
of geological events on continental diversity during the Cenozoic.
Molecular Phylogenetics and Evolution,66, 10271040. https://doi.
Plana, V. (2004). Mechanisms and tempo of evolution in the African Gui-
neo-Congolian rainforest. Philosophical Transactions of the Royal Soci-
ety of London Series B, Biological Sciences,359, 15851594. https://d
Portik, D. M. (2015). Diversification of Afrobatrachian frogs and the her-
petofauna of the Arabian Peninsula. PhD Thesis. Berkeley: University
of California.
Portik, D. M., Jongsma, G. F., Kouete, M. T., Scheinberg, L. A., Freier-
muth, B., Nkonmeneck, W. P. T., & Blackburn, D. C. (2016). A survey
of amphibians and reptiles in the foothills of Mount Kupe, Cameroon.
Amphibian and Reptile Conservation,10(2) [Special Section], 3767
Portik, D. M., Leach
e, A. D., Rivera, D., Blackburn, D. C., R
odel, M. O.,
Barej, M. F., ... Fujita, M. K. (2017). Evaluating mechanisms of diver-
sification in a Guineo-Congolian tropical forest frog using demo-
graphic model selection. Molecular Ecology,26, 52455263. https://d
Rambaut, A., Drummond, A. J., & Suchard, M. (2013). Tracer v1.6.0.
Retrieved from
Schmitz, A., Euskirchen, O., & B
ohme, W. (1999). Zur Herpetofauna einer
montanen Regenwaldregion in SW-Kamerun (Mt. Kupe und Bakossi-
Bergland). I. Einleitung, Bufonidae, und Hyperoliidae. Herpetofauna
(Weinstadt),21(121), 517.
Schwarz, G. (1978). Estimating the dimension of a model. The Annals
of Statistics,6, 461464.
Smith, T. B., Holder, K., Girman, D., OKeefe, K., Larison, B., & Chan, Y.
(2000). Comparative avian phylogeography of Cameroon and Equato-
rial Guinea mountains: Implications for conservation. Molecular Ecol-
ogy,9, 15051516.
Smith, T. B., Wayne, R. K., Girman, D. J., & Bruford, M. W. (1997). A role
for ecotones in generating rainforest biodiversity. Science,276, 1855
Stuart, S. N., Adams, R. J., & Jenkins, M. (1990). Biodiversity in sub-
Saharan Africa and its islands: Conservation, management, and
sustainable use. Occasional Paper of the IUCN Species Survival Commis-
Zhen, Y., Harrigan, R. J., Ruegg, K. C., Anderson, E. C., Ng, T. C., Lao, S.,
... Smith, T. B. (2017). Genomic divergence across ecological gradi-
ents in the Central African rainforest songbird (Andropadus virens).
Molecular Ecology,26, 49664977.
Zimkus, B. M., & Gvo
ık, V. (2013). Sky Islands of the Cameroon
Volcanic Line: A diversification hot spot for puddle frogs
(Phrynobatrachidae: Phrynobatrachus). Zoologica Scripta,42,
Author contributions: D.M.P., K.L.C., and R.C.B conceived of the
study; D.M.P., R.C.B., D.C.B., M.B., V.G., G.F.M.J., and M.T.K. col-
lected field samples; K.L.C. performed lab work with assistance
and funding from A.D.L and M.K.F.; K.L.C., R.C.B, and D.M.P. per-
formed analyses and wrote the manuscript, with final approval
from co-authors.
Additional supporting information may be found online in the
Supporting Information section at the end of the article.
How to cite this article: Charles KL, Bell RC, Blackburn DC,
et al. Sky, sea, and forest islands: Diversification in the
African leaf-folding frog Afrixalus paradorsalis (Anura:
Hyperoliidae) of the Lower Guineo-Congolian rain forest. J
Biogeogr. 2018;00:114.
... To investigate alternative divergence scenarios of the three geographically adjacent groups of A. saxatilis revealed by the fastSTRUCTURE analysis (the Core Central Balkan Subgroup, the East Balkan-Carpathian Group without admixed populations, and the Central European Group, see Section "Results"), we used the diffusion approximation method of dadi to analyse two-dimensional (2D-JSFS) site frequency spectra (Gutenkunst et al., 2009). We used an established 2D analysis pipeline (Portik et al., 2017;Charles et al., 2018) and adapted publicly available python scripts 5 that define 2D models, perform model fitting, and execute plotting functions. As input data, we used a two-dimensional joint site frequency spectrum prepared via conversion of a vcf file to a folded SFS that included downprojection of the sampling size to maximise the number of sampled individuals while minimising levels of missing data for downstream multi-population comparisons and was performed using the easySFS tool 6 . ...
... Therefore, we chose to test simplified models, including either (i) no change in population size since the split, (ii) early stage of population expansion followed by constant population size, or (iii) continuously expanding population size (Supplementary Figure 2B). In a demographic context, we refer to genetic groups identified by the fastSTRUCTURE analyses as "populations" in accordance with recent studies (e.g., Portik et al., 2017;Charles et al., 2018). ...
... These eight models represent the three main hypotheses and modifications with ancestral or recent migration included to explain additional features of the 2D-JSFS (Supplementary Figure 3). Models representing an old vicariance and a recent founder event were taken from Charles et al. (2018), and models of old founder events were designed in Záveská et al. (2021). For models in both the vicariance and founder event categories, we followed Charles et al. (2018) and included a variable "s" that defines the fraction of the ancestral population (nuA) founding each daughter population, 5 ...
Full-text available
The Balkan Peninsula played an important role in the evolution of many Mediterranean plants and served as a major source for post-Pleistocene colonisation of central and northern Europe. Its complex geo-climatic history and environmental heterogeneity significantly influenced spatiotemporal diversification and resulted in intricate phylogeographic patterns. To explore the evolutionary dynamics and phylogeographic patterns within the widespread eastern Mediterranean and central European species Aurinia saxatilis, we used a combination of phylogenomic (restriction-site associated DNA sequencing, RADseq) and phylogenetic (sequences of the plastid marker ndhF) data as well as species distribution models generated for the present and the Last Glacial Maximum (LGM). The inferred phylogenies retrieved three main geographically distinct lineages. The southern lineage is restricted to the eastern Mediterranean, where it is distributed throughout the Aegean area, the southern Balkan Peninsula, and the southern Apennine Peninsula, and corresponds to the species main distribution area during the LGM. The eastern lineage extends from the eastern Balkan Peninsula over the Carpathians to central Europe, while the central lineage occupies the central Balkan Peninsula. Molecular dating places the divergence among all the three lineages to the early to middle Pleistocene, indicating their long-term independent evolutionary trajectories. Our data revealed an early divergence and stable in situ persistence of the southernmost, eastern Mediterranean lineage, whereas the mainland, south-east European lineages experienced more complex and turbulent evolutionary dynamics triggered by Pleistocene climatic oscillations. Our data also support the existence of multiple glacial refugia in southeast Europe and highlight the central Balkan Peninsula not only as a cradle of lineage diversifications but also as a source of lineage dispersal. Finally, the extant genetic variation within A. saxatilis is congruent with the taxonomic separation of peripatric A. saxatilis subsp. saxatilis and A. saxatilis subsp. orientalis, whereas the taxonomic status of A. saxatilis subsp. megalocarpa remains doubtful.
... Branch lengths were scaled to estimate relative divergence times using a generation time of one year and a genome-wide mutation rate estimated in humans (1 × 10 −8 ; Lynch, 2010); this rate was used because no genome-wide estimates exist for anurans and this approach has been applied in other studies of African frogs (e.g. Charles et al., 2018;Leaché et al., 2019;Portik et al., 2017). We examined convergence using TRACER 1.7 (Rambaut et al., 2018) and generated a maximum clade credibility tree from the post burn-in samples. ...
... Our divergence estimates amongst the three lineages of L. rufus range from the mid to late Pleistocene, and roughly coincide with estimates of intraspecific divergence for other anurans across this region that were obtained using similar approaches (i.e. ddRADseq and the same mutation rate; Charles et al., 2018;Portik et al., 2017). ...
... Phylogeographic studies of other large-bodied Bioko anurans including forest tree frogs , foam-nesting tree frogs (Leaché et al., 2019), white-lipped frogs , and clawed frogs (Evans et al., 2015) found very little genetic differentiation between island and continental populations. By contrast, phylogeographic studies of smallerbodied anurans including leaf-folding frogs (Charles et al., 2018) and reed frogs (Bell et al., 2017) found modest genetic differentiation. ...
Secondary sympatry among sister lineages is strongly associated with genetic and ecological divergence. This pattern suggests that for closely related species to coexist in secondary sympatry, they must accumulate differences in traits that mediate ecological and/or reproductive isolation. Here, we characterized inter‐ and intra‐specific divergence in three giant tree frog species whose distributions stretch across West and Central Africa. Using genome‐wide single‐nucleotide polymorphism data, we demonstrated that species‐level divergence coincides temporally and geographically with a period of large‐scale forest fragmentation during the late Pliocene. Our environmental niche models further supported a dynamic history of climatic suitability and stability, and indicated that all three species occupy distinct environmental niches. We found modest morphological differentiation among the species with significant divergence in tympanum diameter and male advertisement call. In addition, we confirmed that two species occur in secondary sympatry in Central Africa but found no evidence of hybridization. These patterns support the hypothesis that cycles of genetic exchange and isolation across West and Central Africa have contributed to globally significant biodiversity. Furthermore, divergence in both ecology and reproductive traits appear to have played important roles in maintaining distinct lineages. At the intraspecific level, we found that climatic refugia, precipitation gradients, marine incursions, and potentially riverine barriers generated phylogeographic structure throughout the Pleistocene and into the Holocene. Further studies examining phenotypic divergence and secondary contact among these geographically structured populations may demonstrate how smaller scale and more recent biogeographic barriers contribute to regional diversification.
... To investigate alternative divergence scenarios for the two main groups and two subgroups detected by STRUCTURE, we used the diffusion approximation method of dadi to analyze two-dimensional joint site frequency spectra (2D-JSFS, Gutenkunst et al., 2009). We used an established 2D analysis pipeline (Portik et al., 2017;Charles et al., 2018) and adapted publicly available python scripts 4 that define 2D models, perform model fitting, and execute plotting functions. ...
... Therefore, we chose to test the simplified models, including either (i) no change in population size since the split, (ii) an early stage of population expansion followed by constant population size, or (iii) a continuously expanding population size ( Figure 2B). In the context of demographic modeling we refer to the genetic groups identified by the STRUCTURE analyses as "populations" in accordance with recent studies (e.g., Portik et al., 2017;Charles et al., 2018). ...
... These eight models represent the three main hypotheses and modifications with ancestral or recent migration included to explain additional features of the 2D-JSFS (Supplementary Figure 1). While the models representing old vicariance and a recent founder event were taken from Charles et al. (2018), the models including an old founder event were newly designed here. For models in both the vicariance and founder event categories, we followed the strategy of Charles et al. (2018) and included a variable s that defines the fraction of the ancestral population (nuA) founding each daughter population, where nuA * s represents the island population and nuA * (1−s) represents the mainland population. ...
Full-text available
Glacial refugia of alpine and subnival biota have been intensively studied in the European Alps but the fate of forests and their understory species in that area remains largely unclear. In order to fill this gap, we aimed at disentangling the spatiotemporal diversification of disjunctly distributed black hellebore Helleborus niger (Ranunculaceae). We applied a set of phylogeographic analyses based on restriction-site associated DNA sequencing (RADseq) data and plastid DNA sequences to a range-wide sampling of populations. These analyses were supplemented with species distribution models generated for the present and the Last Glacial Maximum (LGM). We used exploratory analyses to delimit genomically coherent groups and then employed demographic modeling to reconstruct the history of these groups. We uncovered a deep split between two major genetic groups with western and eastern distribution within the Southern Limestone Alps, likely reflecting divergent evolution since the mid-Pleistocene in two glacial refugia situated along the unglaciated southern margin of the Alps. Long-term presence in the Southern Limestone Alps is also supported by high numbers of private alleles, elevated levels of nucleotide diversity and the species’ modeled distribution at the LGM. The deep genetic divergence, however, is not reflected in leaf shape variation, suggesting that the morphological discrimination of genetically divergent entities within H. niger is questionable. At a shallower level, populations from the Northern Limestone Alps are differentiated from those in the Southern Limestone Alps in both RADseq and plastid DNA data sets, reflecting the North-South disjunction within the Eastern Alps. The underlying split was dated to ca. 0.1 mya, which is well before the LGM. In the same line, explicit tests of demographic models consistently rejected the hypothesis that the partial distribution area in the Northern Limestone Alps is the result of postglacial colonization. Taken together, our results strongly support that forest understory species such as H. niger have survived the LGM in refugia situated along the southern, but also along the northern or northeastern periphery of the Alps. Being a slow migrator, the species has likely survived repeated glacial-interglacial circles in distributional stasis while the composition of the tree canopy changed in the meanwhile.
... Under this model, during cool humid glacial periods, suitable forested habitats expanded from high elevations into the intervening lowland Mbos Plain (a savanna of 300 km 2 ), through which montane taxa dispersed. Recent inter-and intraspecific phylogenetic studies of rodents and frogs supports this as a general pattern across the Cameroon Volcanic Line because many of the inferred divergences between montane populations or between montane and lowland populations occurred in the Pleistocene (Taylor et al., 2014;Bell et al., 2017;Charles et al., 2018;Dolinay et al., 2021). ...
... However, most of the divergences within the clade of four montane species occurred in the Plio-Pleistocene, thus providing support to Amiet's (1975) hypothesis that this period of climatic change drove the evolution of new montane species in Cardioglossa. Further, the mid-Pleistocene divergence between C. oreas oreas in the Bamiléké Plateau and C. oreas manengouba on Mount Manengouba is similar to other populations of frogs that colonized Mount Manengouba or the nearby Bakossi Mountains during the Pleistocene, including Afrixalus paradorsalis (Charles et al., 2018;A. manengubensis in Channing and Rödel, 2019), Hyperolius dintelmanni (Bell et al., 2017), Phrynobatrachus jimzimkusi (Dolinay et al., 2021), and P. manengoubensis (Gvoždík et al., 2020). ...
Full-text available
The African anuran genus Cardioglossa contains 19 described species, most of which are distinguished from one another by striking patterns and colors. We present a well-resolved phylogeny based on analyses of mitochondrial and nuclear loci for 18 species of Cardioglossa. This provides the basis for species-delimitation analyses and interpreting historical biogeography in the genus. Whereas much of the diversification within the genus occurred among Central African lineages during the Miocene following the origin of Cardioglossa in the latest Oligocene or earliest Miocene, most species-pairs in the genus diverged more recently during the Plio-Pleistocene. The two most geographically peripheral speciesC. cyaneospila in the Albertine Rift Mountains and C. occidentalis in the Upper Guinean Forests of West Africaboth diverged from other lineages during the mid-late Miocene. Because our analyses do not support C. manengouba and C. oreas as distinct species, we recognize these geographically separate and phenotypically distinct populations as subspecies of C. oreas that diverged subsequent to the origin of Mount Manengouba during the past 1.5 million years. In contrast, we find that C. leucomystax likely represents two species found in the Lower Guinean and Congolian forests, respectively. We find recent divergences between several allopatric lineages (either species or populations) that differ in coloration and pattern, including in C. nigromaculata which varies in color across its range in Central Africa and Bioko Island. These recent divergences among allopatric lineages with distinctive coloration and pattern raise new questions about the significance of these traits in this genus for which little is known of its natural history and biology.
... In maps (d)-(f), warmer colors represent areas of higher suitability over time that have remained stable compared with previous estimates of forest refugia (dotted black lines, Maley, 1996). Country borders (gray lines) and chimpanzee subspecies ranges (green lines, Humle et al., 2016) are also shown Mikula, Patzenhauerová, et al., 2014;Gaubert et al., 2016;Mizerovská et al., 2019;Nicolas et al., 2011), including other primates, (Anthony et al., 2014;Clifford et al., 2004;Gonder et al., 2011;Pozzi, 2016;Telfer et al., 2003), amphibians (Charles et al., 2018;Leaché et al., 2019;Portik et al., 2017), and plants (Faye et al., 2016;Hardy et al., 2013;Piñeiro et al., 2017Piñeiro et al., , 2019, albeit with some differences between species due to different ecological characteristics and idiosyncratic responses to climatic changes (Lowe et al., 2010). ...
... For example, habitat suitability fluc- Beyond correlative approaches such as those described above, mechanistic approaches would enable a deeper exploration of biogeographical patterns and processes that have affected chimpanzees, especially with the recent availability of comprehensive behavioral and molecular data for a large number of wild chimpanzee populations. Rapid developments in the generation and analysis of genome-wide molecular data over the past decade have revealed detailed demographic histories, enabling the identification of diversification mechanisms due to forest refugia, which are characterized by divergence, isolation, and secondary contact as refugial habitats fragment and reconnect with each other during glacial cycles (Barratt et al., 2018;Charles et al., 2018;Feng et al., in press;Leaché et al., 2019;Portik et al., 2017). The ability to distinguish signals of forest refugia from other diversification mechanisms such as landscape barriers, ecological gradients, and BARRATT ET AL | 13 of 18 anthropogenic habitat fragmentation, would represent a powerful approach for gaining a more mechanistic understanding of population diversification. ...
Full-text available
Paleoclimate reconstructions have enhanced our understanding of how past climates have shaped present-day biodiversity. We hypothesize that the geographic extent of Pleistocene forest refugia and suitable habitat fluctuated significantly in time during the late Quaternary for chimpanzees (Pan troglodytes). Using bioclimatic variables representing monthly temperature and precipitation estimates, past human population density data, and an extensive database of georeferenced presence points, we built a model of changing habitat suitability for chimpanzees at fine spatio-temporal scales dating back to the Last Interglacial (120,000 BP). Our models cover a spatial resolution of 0.0467° (approximately 5.19 km2 grid cells) and a temporal resolution of between 1000 and 4000 years. Using our model, we mapped habitat stability over time using three approaches, comparing our modeled stability estimates to existing knowledge of Afrotropical refugia, as well as contemporary patterns of major keystone tropical food resources used by chimpanzees, figs (Moraceae), and palms (Arecacae). Results show habitat stability congruent with known glacial refugia across Africa, suggesting their extents may have been underestimated for chimpanzees, with potentially up to approximately 60,000 km2 of previously unrecognized glacial refugia. The refugia we highlight coincide with higher species richness for figs and palms. Our results provide spatio-temporally explicit insights into the role of refugia across the chimpanzee range, forming the empirical foundation for developing and testing hypotheses about behavioral, ecological, and genetic diversity with additional data. This methodology can be applied to other species and geographic areas when sufficient data are available.
... The Cameroon Volcanic Line is known as an ancient refugial region serving as an endemism hotspot of animal and plant species with ecological affinities to both montane rainforest (Missoup et al., , 2016Zimkus and Gvoždík, 2013) and the lowland rainforest surrounding these mountains (Bohoussou et al., 2015;Charles et al., 2018;Leaché et al., 2019;Migliore et al., 2019). In some cases, the CVL was identified as a refugial source for long-distance dispersals (Demenou et al., 2020) or an important biogeographic barrier Portik et al., 2017) of lowland rainforest biota. ...
... As an example from the south, we can link the (sub)montane population of the otherwise lowland forest frog Afrixalus paradorsalis to P. jimzimkusi. The (sub)montane A. p. manengubensis (newly treated as full species; Channing and Rödel, 2019) from Mt. Manengouba was estimated to originate some 2-1 Mya (Charles et al., 2018), coinciding with the mountain uplift (Marzoli et al., 2000), and paralleled to the age of origin of P. jimzimkusi. ...
Puddle frogs of the Phrynobatrachus steindachneri species complex are a useful group for investigating speciation and phylogeography in Afromontane forests of the Cameroon Volcanic Line, western Central Africa. The species complex is represented by six morphologically relatively cryptic mitochondrial DNA lineages, only two of which are distinguished at the species level – southern P. jimzimkusi and Lake Oku endemic P. njiomock, leaving the remaining four lineages identified as ‘P. steindachneri’. In this study, the six mtDNA lineages are subjected to genomic sequence capture analyses and morphological examination to delimit species and to study biogeography. The nuclear DNA data (387 loci; 571,936 aligned base pairs) distinguished all six mtDNA lineages, but the topological pattern and divergence depths supported only four main clades: P. jimzimkusi, P. njiomock, and only two divergent evolutionary lineages within the four ‘P. steindachneri’ mtDNA lineages. One of the two lineages is herein described as a new species, P. amieti sp. nov. Reticulate evolution (hybridization) was detected within the species complex with morphologically intermediate hybrid individuals placed between the parental species in phylogenomic analyses, forming a ladder-like phylogenetic pattern. The presence of hybrids is undesirable in standard phylogenetic analyses but is essential and beneficial in the network multispecies coalescent. This latter approach provided insight into the reticulate evolutionary history of these endemic frogs. Introgressions likely occurred during the Middle and Late Pleistocene climatic oscillations, due to the cyclic connections (likely dominating during cold glacials) and separations (during warm interglacials) of montane forests. The genomic phylogeographic pattern supports the separation of the southern (Mt. Manengouba to Mt. Oku) and northern mountains at the onset of the Pleistocene. Further subdivisions occurred in the Early Pleistocene, separating populations from the northernmost (Tchabal Mbabo, Gotel Mts.) and middle mountains (Mt. Mbam, Mt. Oku, Mambilla Plateau), as well as the microendemic lineage restricted to Lake Oku (Mt. Oku). This unique model system is highly threatened as all the species within the complex have exhibited severe population declines in the past decade, placing them on the brink of extinction. In addition, Mount Oku is identified to be of particular conservation importance because it harbors three species of this complex. We, therefore, urge for conservation actions in the Cameroon Highlands to preserve their diversity before it is too late.
... poeppigii, we tested 2D-JSFS models incorporating differing migration levels at different time periods ( Figure S6). In addition to a model of (1) divergence with no migration, we tested the following models: (2) divergence with continuous symmetric migration, (3) divergence with continuous asymmetric migration, (4) divergence with continuous symmetric migration and a varying rate of migration across two epochs, (5) divergence with continuous asymmetric migration and a varying rate of migration across two epochs, (6) divergence in isolation, followed by symmetric secondary contact, (7) divergence in isolation, followed by asymmetric secondary contact, (8) ancient symmetric migration then subsequent isolation, (9) ancient asymmetric migration then subsequent isolation, (10) divergence in isolation followed by symmetric secondary contact with subsequent isolation, and (11) divergence in isolation followed by asymmetric secondary contact with subsequent isolation (Charles et al., 2018;Portik et al., 2017). ...
Full-text available
The effects of genetic introgression on species boundaries and how they affect species’ integrity and persistence over evolutionary time have received increased attention. The increasing availability of genomic data has revealed contrasting patterns of gene flow across genomic regions, which impose challenges to inferences of evolutionary relationships and of patterns of genetic admixture across lineages. By characterizing patterns of variation across thousands of genomic loci in a widespread complex of true toads (Rhinella), we assess the true extent of genetic introgression across species thought to hybridize to extreme degrees based on natural history observations and multi‐locus analyses. Comprehensive geographic sampling of five large‐ranged Neotropical taxa revealed multiple distinct evolutionary lineages that span large geographic areas and, at times, distinct biomes. The inferred major clades and genetic clusters largely correspond to currently recognized taxa; however, we also found evidence of cryptic diversity within taxa. While previous phylogenetic studies revealed extensive mito‐nuclear discordance, our genetic clustering analyses uncovered several admixed individuals within major genetic groups. Accordingly, historical demographic analyses supported that the evolutionary history of these toads involved cross‐taxon gene flow both at ancient and recent times. Lastly, ABBA‐BABA tests revealed widespread allele sharing across species boundaries, a pattern that can be confidently attributed to genetic introgression as opposed to incomplete lineage sorting. These results confirm previous assertions that the evolutionary history of Rhinella was characterized by various levels of hybridization even across environmentally heterogeneous regions, posing exciting questions about what factors prevent complete fusion of diverging yet highly interdependent evolutionary lineages.
... 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., 2011Penner et al., , 2019Wieczorek et al., 2000;Zimkus et al., 2010), but do not appear to represent an important barrier for many plant species Debout et al., 2011;Hardy et al., 2013;Ley et al., 2014; F I G U R E 1 Locations of major rivers and hypothesized refugia (labeled 1-10) in West and Central Africa, adapted from Maley (1996Maley ( ) et al., 2010. Refugia are suggested to have played an important role in the diversification of rodents (Bohoussou et al., 2015;Nicolas et al., 2011Nicolas et al., , 2012, primates (Clifford et al., 2004;Haus et al., 2013;Tosi, 2008), frogs (Bell et al., 2017;Jongsma et al., 2018), lizards (Allen et al., 2019;Leaché et al., 2017), birds (Fjeldså & Bowie, 2008), pangolins (Gaubert et al., 2016), and rainforest plants (Born et al., 2011;Budde et al., 2013;Daïnou et al., 2010;Dauby et al., 2010;Duminil et al., 2015;Faye et al., 2016;Gomez et al., 2009;Hardy et al., 2013;Ley et al., 2014Ley et al., , 2016Lowe et al., 2010). ...
Full-text available
The relative roles of rivers versus refugia in shaping the high levels of species diversity in tropical rainforests have been widely debated for decades. Only recently has it become possible to take an integrative approach to test predictions derived from these hypotheses using genomic sequencing and paleo-species distribution modeling. Herein, we tested the predictions of the classic river, refuge, and river-refuge hypotheses on diversification in the arboreal sub-Saharan African snake genus Toxicodryas. We used dated phylogeographic inferences, population clustering analyses, demographic model selection, and paleo-distribution modeling to conduct a phylogenomic and historical demographic analysis of this genus. Our results revealed significant population genetic structure within both Toxicodryas species, corresponding geographically to river barriers and divergence times from the mid-Miocene to Pliocene. Our demographic analyses supported the interpretation that rivers are indications of strong barriers to gene flow among populations since their divergence. Additionally, we found no support for a major contraction of suitable habitat during the last glacial maximum, allowing us to reject both the refuge and river-refuge hypotheses in favor of the river-barrier hypothesis. Based on conservative interpretations of our species delimitation analyses with the Sanger and ddRAD data sets, two new cryptic species are identified from east-central Africa. This study highlights the complexity of diversification dynamics in the African tropics and the advantages of integrative approaches to studying speciation in tropical regions.
Species' ecological traits influence their spatial genetic patterns. Bedrock preference strongly shapes the phylogeography of alpine plants, but its interactions with other ecological traits have rarely been disentangled. Here, we explore whether dispersal ability and degree of habitat specialization account for divergent postglacial expansion patterns of high‐elevation plants in spite of similar bedrock preference. The Pyrenees, southwestern Europe. Cirsium glabrum (Asteraceae), Salix pyrenaica (Salicaceae) and Silene borderei (Caryophyllaceae). Phylogenetic, genetic structure and demographic modelling analyses based on restriction‐site‐associated DNA sequencing (RADseq) data from a range‐wide populational sampling were conducted. Occurrence data and environmental variables were used to construct species distribution models, which were projected under current and Last Glacial Maximum conditions, and were combined with RADseq data to reconstruct the postglacial history of the study species. The degree of habitat specialization of each species was estimated based on the plant communities within which they occur, and their climatic niche breadth. Salix pyrenaica, which occupies a broad range of habitats, shows a high level of range filling, a blurred genetic structure and an admixture cline between the two main genetic groups, congruent with rapid postglacial expansion. The microsite specialists C. glabrum and S. borderei exhibit a strong genetic structure and low levels of range filling, indicative of slow postglacial expansion. The good disperser C. glabrum shows higher levels of admixture between genetic groups and weaker population differentiation than the poor disperser S. borderei. Factors other than bedrock preference have a strong impact on the postglacial range dynamics of high‐elevation species. Habitat specialization plays an important role, allowing species occupying a broad range of habitats to more rapidly expand their ranges after environmental change. The effect of dispersal ability is lower than expected for the study species.
In the central Congolian lowland forests we discovered for the first time Panaspis breviceps, a rarely found scincid lizard from the Central African riparian forests. Given that the Central African forests exhibit heterogeneity in the distribution of environmental characteristics and forms distinct ecoregions, the question arises as to how this newly discovered population compares with other populations in Central Africa and particularly in the Congolian lowland forests. We reviewed the distribution records of this species and examined and compared new and available genetic data (mitochondrial DNA). Maximum likelihood phylogenetic analysis revealed the existence of two evolutionary lineages differing by 2.0% in 16S rRNA. One lineage occurs in and around the southern Cameroon Highlands, but its distribution southwards is poorly documented. The other lineage includes the western, central and eastern populations of the Congo Basin, suggesting certain biogeographic connectivity across the Congolian forests. These results support the hypothesis of limited biogeographic barriers to the distribution of lizards in the Congolian lowland forests, but this remains to be tested using additional independent markers, denser sampling and multiple species.
Full-text available
Differences in mating signals among incipient species are an important mechanism driving reproductive isolation and speciation. Here, we investigate male advertisement call divergence across a radiation of reed frogs from the Gulf of Guinea archipelago and the most closely related species on the African continent (Hyperolius olivaceus). The two species endemic to the island of São Tomé (Hyperolius molleri and Hyperolius thomensis) differ in body size, coloration and breeding ecology, yet they hybridize where their habitats overlap. A third species, Hyperolius drewesi, is sister to H. molleri and endemic to Príncipe Island. We found significant differences in average dominant frequency and average number of pulses per call among the four species. The strong relationship between body size and dominant frequency irrespective of geographical history, breeding habitat or acoustic environment suggests that differences in this component of the mating signal may be a by-product of adaptive changes in body size among the four species. Hybrid males are intermediate in size between H. molleri and H. thomensis and produce calls at intermediate dominant frequencies. Future efforts to characterize courtship behaviour and breeding habitat preference in allopatric and sympatric populations will provide additional insights as to the potential for reinforcement in this system. © 2017 The Linnean Society of London, Biological Journal of the Linnean Society.
Full-text available
The accumulation of biodiversity in tropical forests can occur through multiple allopatric and parapatric models of diversification, including forest refugia, riverine barriers, and ecological gradients. Considerable debate surrounds the major diversification process, particularly in the West African Lower Guinea forests, which contain a complex geographic arrangement of topographic features and historical refugia. We used genomic data to investigate alternative mechanisms of diversification in the Gaboon forest frog, Scotobleps gabonicus, by first identifying population structure and then performing demographic model selection and spatially explicit analyses. We found that a majority of population divergences are best explained by allopatric models consistent with the forest refugia hypothesis, and involve divergence in isolation with subsequent expansion and gene flow. These population divergences occurred simultaneously and conform to predictions based on climatically stable regions inferred through ecological niche modeling. Though forest refugia played a prominent role in the intraspecific diversification of S. gabonicus, we also find evidence for potential interactions between landscape features and historical refugia, including major rivers and elevational barriers such as the Cameroonian Volcanic Line. We outline the advantages of using genome-wide variation in a model-testing framework to distinguish between alternative allopatric hypotheses, and the pitfalls of limited geographic and molecular sampling. Although phylogeographic patterns are often species-specific and related to life history traits, additional comparative studies incorporating genomic data are necessary for separating shared historical processes from idiosyncratic responses to environmental, climatic, and geological influences on diversification. This article is protected by copyright. All rights reserved.
Full-text available
We performed surveys at several lower elevation sites surrounding Mt. Kupe, a mountain at the southern edge of the Cameroonian Highlands. This work resulted in the sampling of 48 species, including 38 amphibian and 10 reptile species. By combining our data with prior survey results from higher elevation zones, we produce a checklist of 108 species for the greater Mt. Kupe region including 72 frog species, 21 lizard species, and 15 species of snakes. Our work adds 30 species of frogs at lower elevations, many of which are associated with breeding in pools or ponds that are absent from the slopes of Mt. Kupe. We provide taxonomic accounts, including museum specimen data and associated molecular data, for all species encountered. Finally, we compare the levels of biodiversity of Mt. Kupe to other regions, discuss biogeographic ties to other montane systems, and note current conservation threats.
Full-text available
This paper gives a historical overview of the African rain forest from its origins, towards, the end of the Cretaceous period. The areas around the Gulf of Guinea, in particular from Ivory Coast to Nigeria and especially Cameroon, Gabon and Congo, appear to have been already occupied at this time by wet tropical forest formations mainly composed of Angiosperms. In the course of the Tertiary period the combined effect of the equator being situated further north than now and the development of the Antarctic ice cap favoured the development of wet tropical conditions over a large part of North Africa. Towards the end of the Tertiary, the equator reached its present position and the northern hemisphere ice caps appeared, and these phenomena resulted in the disappearance of the forest formations spread across the north of Africa, and the concentration of these formations near the equatorial zone around the Gulf of Guinea and in the Congo-Zaire basin. From 800 000 years ago onwards the marked glacial variations at middle and high latitudes in both hemispheres, lowered temperatures in equatorial areas and brought arid climates at times of maximum glacial extension. -Author
Full-text available
Arlequin ver 3.0 is a software package integrating several basic and advanced methods for population genetics data analysis, like the computation of standard genetic diversity indices, the estimation of allele and haplotype frequencies, tests of departure from linkage equilibrium, departure from selective neutrality and demographic equilibrium, estimation or parameters from past population expansions, and thorough analyses of population subdivision under the AMOVA framework. Arlequin 3 introduces a completely new graphical interface written in C++, a more robust semantic analysis of input files, and two new methods: a Bayesian estimation of gametic phase from multi-locus genotypes, and an estimation of the parameters of an instantaneous spatial expansion from DNA sequence polymorphism. Arlequin can handle several data types like DNA sequences, microsatellite data, or standard multilocus genotypes. A Windows version of the software is freely available on
The package adegenet for the R software is dedicated to the multivariate analysis of genetic markers. It extends the ade4 package of multivariate methods by implementing formal classes and functions to manipulate and analyse genetic markers. Data can be imported from common population genetics software and exported to other software and R packages. adegenet also implements standard population genetics tools along with more original approaches for spatial genetics and hybridization. Availability: Stable version is available from CRAN: Development version is available from adegenet website: Both versions can be installed directly from R. adegenet is distributed under the GNU General Public Licence (v.2). Supplementary information:Supplementary data are available at Bioinformatics online.
The little greenbul, a common rainforest passerine from sub-Saharan Africa, has been the subject of long-term evolutionary studies to understand the mechanisms leading to rainforest speciation. Previous research found morphological and behavioral divergence across rainforest-savanna transition zones (ecotones), and a pattern of divergence with gene flow suggesting divergent natural selection has contributed to adaptive divergence and ecotones could be important areas for rainforests speciation. Recent advances in genomics and environmental modeling make it possible to examine patterns of genetic divergence in a more comprehensive fashion. To assess the extent to which natural selection may drive patterns of differentiation, here we investigate patterns of genomic differentiation among populations across environmental gradients and regions. We find compelling evidence that individuals form discrete genetic clusters corresponding to distinctive environmental characteristics and habitat types. Pairwise FST between populations in different habitats is significantly higher than within habitats, and this differentiation is greater than what is expected from geographic distance alone. Moreover, we identified 140 SNPs that showed extreme differentiation among populations through a genome-wide selection scan. These outliers were significantly enriched in exonic and coding regions, suggesting their functional importance. Environmental association analysis of SNP variation indicates that several environmental variables, including temperature and elevation, play important roles in driving the pattern of genomic diversification. Results lend important new genomic evidence for environmental gradients being important in population differentiation. This article is protected by copyright. All rights reserved.
Organismal traits interact with environmental variation to mediate how species respond to shared landscapes. Thus, differences in traits related to dispersal ability or physiological tolerance may result in phylogeographic discordance among co-distributed taxa, even when they are responding to common barriers. We quantified climatic suitability and stability, and phylogeographic divergence within three reed frog species complexes across the Guineo-Congolian forests and Gulf of Guinea archipelago of Central Africa to investigate how they responded to a shared climatic and geological history. Our species-specific estimates of climatic suitability through time are consistent with temporal and spatial heterogeneity in diversification among the species complexes, indicating that differences in ecological breadth may partly explain these idiosyncratic patterns. Likewise, we demonstrated that fluctuating sea levels periodically exposed a land bridge connecting Bioko Island with the mainland Guineo-Congolian forest and that habitats across the exposed land bridge likely enabled dispersal in some species, but not in others. We did not find evidence that rivers are biogeographic barriers across any of the species complexes. Despite marked differences in the geographic extent of stable climates and temporal estimates of divergence among the species complexes, we recovered a shared pattern of intermittent climatic suitability with recent population connectivity and demographic expansion across the Congo Basin. This pattern supports the hypothesis that genetic exchange across the Congo Basin during humid periods, followed by vicariance during arid periods, has shaped regional diversity. Finally, we identified many distinct lineages among our focal taxa, some of which may reflect incipient or unrecognized species. This article is protected by copyright. All rights reserved.