TYPE Original Research

PUBLISHED 05 June 2024

DOI 10.3389/fcosc.2024.1366248

Environmental variation predicts

OPEN ACCESS patterns of genomic variation in
EDITED BY
Jesus Eduardo Maldonado,
Smithsonian Conservation Biology Institute
an African tropical forest frog
(SI), United States

REVIEWED BY
Courtney A. Miller 1*, Geraud C. Tasse Taboue 2,3,
Katherine Andrea Solari, Eric B. Fokam 2, Katy Morgan 1, Ying Zhen 4, Ryan J. Harrigan 4,
Stanford University, United States
Kevin P. Mulder, Vinh Le Underwood 4, Kristen Ruegg 4, Paul R. Sesink Clee 5,
Ghent University, Belgium
Stephan Ntie 6, Patrick Mickala 6, Jean Francois Mboumba 6,
*CORRESPONDENCE
Courtney A. Miller
Trevon Fuller 4, Breda M. Zimkus 7, Thomas B. Smith 4,8
[email protected] and Nicola M. Anthony 1
RECEIVED 05 January 2024 1
Department of Biological Sciences, University of New Orleans, New Orleans, LA, United States,
ACCEPTED 18 April 2024 2
Department of Zoology and Animal Physiology, University of Buea, Buea, Cameroon, 3 Multipurpose
PUBLISHED 05 June 2024
Research Station, Institute of Agricultural Research for Development, Bangangte, Cameroon,
CITATION
4
Institute of Environment and Sustainability, University of California, Los Angeles, Los Angeles,
Miller CA, Tasse Taboue GC, Fokam EB, CA, United States, 5 Department of Biology, Drexel University, Philadelphia, PA, United States,
Morgan K, Zhen Y, Harrigan RJ, Le 6
Department of Biology, University of Science and Technology of Masuku, Franceville, Gabon,
Underwood V, Ruegg K, Clee PRS, Ntie S, 7
Museum of Comparative Zoology, Harvard University, Cambridge, MA, United States, 8 Department of
Mickala P, Mboumba JF, Fuller T, Zimkus BM, Ecology and Evolutionary Biology, University of California, Los Angeles, Los Angeles, CA, United States
Smith TB and Anthony NM (2024)
Environmental variation predicts patterns of
genomic variation in an African tropical
forest frog. Central African rainforests are predicted to be disproportionately affected by
Front. Conserv. Sci. 5:1366248.
doi: 10.3389/fcosc.2024.1366248
future climate change. How species will cope with these changes is unclear, but
rapid environmental changes will likely impose strong selection pressures. Here
COPYRIGHT
© 2024 Miller, Tasse Taboue, Fokam, Morgan, we examined environmental drivers of genomic variation in the central African
Zhen, Harrigan, Le Underwood, Ruegg, Clee, puddle frog (Phrynobatrachus auritus) to identify areas of elevated
Ntie, Mickala, Mboumba, Fuller, Zimkus, Smith
and Anthony. This is an open-access article
environmentally-associated turnover. We also compared current and future
distributed under the terms of the Creative climate models to pinpoint areas of high genomic vulnerability where allele
Commons Attribution License (CC BY). The frequencies will have to shift the most in order to keep pace with future climate
use, distribution or reproduction in other
forums is permitted, provided the original change. Neither physical landscape barriers nor the effects of past Pleistocene
author(s) and the copyright owner(s) are refugia influenced genomic differentiation. Alternatively, geographic distance
credited and that the original publication in
and seasonal aspects of precipitation are the most important drivers of SNP allele
this journal is cited, in accordance with
accepted academic practice. No use, frequency variation. Patterns of genomic differentiation coincided with key
distribution or reproduction is permitted ecological gradients across the forest-savanna ecotone, montane areas, and a
which does not comply with these terms.
coastal to interior rainfall gradient. Areas of greatest vulnerability were found in
the lower Sanaga basin, the southeastern region of Cameroon, and southwest
Gabon. In contrast with past conservation efforts that have focused on hotspots
of species richness or endemism, our findings highlight the importance of
maintaining environmentally heterogeneous landscapes to preserve genomic
variation and ongoing evolutionary processes in the face of climate change.

KEYWORDS

Central Africa, amphibians, RAD-seq, environmental gradients, genomic vulnerability,

climate change

Frontiers in Conservation Science 01 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

1 Introduction differences between sites, independent of geographic distance

(Shafer and Wolf, 2013; Wang and Bradburd, 2014). In contrast,
The tropical forests of the Congo Basin and Gulf of Guinea under a model of IBD, genetic differentiation is predicted to
represent one of the most biologically diverse regions in the world. increase as a function of geographic distance whereas genetic
This region ranks third in plant, mammal, bird, and amphibian differentiation under IBB is driven by landscape barriers to
species richness after the Amazon and New Guinea (Mittermeier animal or plant dispersal (Balkenhol et al., 2017; Cushman and
et al., 2003). With respect to amphibians, the Cameroon highlands Schwartz, 2006). Resistance distances due to barriers between
are recognized as one of the world’s most important biodiversity populations can be based on landscape features that may inhibit
hotspots (Herrmann et al., 2005; Stuart et al., 2008; Gvoždı́k et al., gene flow, including physical barriers such as rivers (Mitchell et al.,
2020). Several hypotheses have been advanced to explain the high 2015) or changes in elevation (IBB). Lastly, resistance matrices can
biodiversity in this region. The more arid conditions during the also be used to model the effects of past refugia (Ntie et al., 2017)
Pleistocene resulted in fragmented forest habitat, and previous (IBP), by hindcasting areas of suitable habitat during the last glacial
phylogeographic studies have shown that past Pleistocene refugia maximum (LGM) (Nogué s-Bravo, 2009). In many of these cases,
shaped population structure in several central African rainforest circuit theory is used to incorporate IBE, IBB, and IBP into models
species, providing support for the role of Pleistocene forest refugia of population connectivity and identify which variables are the most
as potential engines of diversification (Qué rouil et al., 2003; Plana, important predictors of gene flow (McRae and Beier, 2007).
2004; Anthony et al., 2007; Born et al., 2011; Nicolas et al., 2011; Central Africa faces a variety of threats from human activities
Murienne et al., 2013). Alternatively, the riverine barrier hypothesis and is especially vulnerable to climate change (Oates et al., 2004;
has argued that rivers could have led to the isolation and Laporte et al., 2007; Abernethy et al., 2013; James et al., 2013).
diversification of tropical forest species (Colyn et al., 1991). Temperatures are expected to rise along with potential shifts in
Support for this hypothesis has been found in primates (Telfer rainfall patterns, including more intense dry seasons that could
et al., 2003; Anthony et al., 2007; Mitchell et al., 2015), birds (Aleixo, result in forest retreat (James et al., 2013). Species in this region, if
2004) and rodents (Nicolas et al., 2011). Compared to these other they are to survive, would therefore be forced to respond to climate
taxa, far less attention has been paid to assessments of the impact of change either through dispersal, evolutionary adaptation or
physical landscape barriers or past Pleistocene refugia on gene flow, phenotypic plasticity (Holt, 1990; Davis et al., 2005). Given the
or the effects of environmental variation (i.e. precipitation, threat that climate change poses to many species, there is now an
temperature, and seasonality) on patterns of amphibian increasing need to identify current and historical drivers of
diversification in this region. evolutionary diversification and recognize key areas for future
Environmental variation can act as a strong agent of conservation where species capacity to adapt is greatest (Anthony
diversifying selection, particularly in areas of high environmental et al., 2015). Mapping landscape-level predictions of
heterogeneity (Endler, 1973), such as that observed across ecotones environmentally-associated genomic variation under both current
(Smith et al., 1997; Freedman et al., 2010; Termignoni-Garcı́a et al., and projected future environments can shed light on both the ability
2017) or across different levels of elevation (Thomassen et al., 2011). of populations to persist in their current state as well as their future
In central Africa, environmental variation has been shown to capacity to respond to change through evolutionary adaptation
explain patterns of genetic differentiation in olive sunbirds (Smith (Gunderson, 2000; Thrush et al., 2009; Sgrò et al., 2011). In this
et al., 2011), little greenbuls (Smith et al., 1997; Zhen et al., 2017), regard, the term “genomic vulnerability” (Bay et al., 2018; Ruegg
skinks (Freedman et al., 2010), chimpanzees (Mitchell et al., 2015), et al., 2018) has been used as a measure of the degree of “mismatch”,
forest antelope (Ntie et al., 2017), soft-furred mice (Morgan et al., or “offset”, between current and future projections of
2020), and reed frogs (Bell et al., 2017). These heterogeneous environmentally-associated genomic variation and can be used as
environments may capture ecological and evolutionary processes a proxy for population vulnerability to environmental change
that are fundamental to maintaining and generating biological (Ruegg et al., 2018). Genomic vulnerability estimates how much
diversity (Moritz et al., 2000). allele frequencies would have to change to keep track with the
One major challenge is being able to effectively partition the environmental changes predicted to occur at a certain location.
effects of isolation by environment (IBE) from other potential Thus, the locations with the greatest vulnerability are those that are
drivers of population differentiation, namely: isolation by distance predicted to have to undergo the greatest changes in allele
(IBD), isolation by resistance due to physical landscape barriers frequencies to keep pace with environmental change.
(IBB), and historical isolation due to past Pleistocene refugia (IBP). In the present study, we used a combination of statistical methods
Advances in landscape genomics can be used to simultaneously to determine the potential drivers of genomic diversification in the
assess the relative importance of competing ecological and historical widespread African puddle frog, Phrynobatrachus auritus across its
drivers on genomic differentiation (Manthey and Moyle, 2015; range in the west Central African countries of Cameroon, Equatorial
Termignoni-Garcı́a et al., 2017). Specifically, Fitzpatrick and Guinea, and Gabon. We used geospatial modeling to map patterns of
Keller (2015) have shown that Generalized Dissimilarity genomic turnover (i.e. the change in allele frequencies with
Modelling (GDM (Ferrier et al., 2007)) and Gradient Forests (GF geographic distance) and predict areas of elevated genomic
(Ellis et al., 2012)) can be powerful tools for analyzing gene– vulnerability. P. auritus serves as an ideal model for examining the
environment associations at the landscape level. Under a model effects of environmental heterogeneity because it occurs in a variety of
of IBE, genetic differentiation increases with environmental forest types (Zimkus and Schick, 2010) and occupies a wide range of

Frontiers in Conservation Science 02 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

environmental conditions. Findings from these methods were then differentiated from female frogs by the presence of a developed vocal
used to address the following: 1) Does IBE influence genomic sac, indicated by a dark throat and vocal folds. Frogs that could not be
differentiation more than IBB or IBP? 2) Are areas of greatest sexed by these secondary sexual characteristics were dissected and
environmentally-associated genomic turnover associated with sexed by either the type of reproductive organ or by the presence of
strong environmental gradients across the landscape? 3) Where is eggs. All animal handling procedures were carried out according to an
genomic vulnerability predicted to be highest across the study area in approved University of New Orleans Institutional Animal Care and
response to future climate change? Use Committee protocol 12-008. Specimens were deposited at the
Museum of Comparative Zoology, Harvard University.

2 Materials and methods

2.2 Environmental datasets
2.1 Field sampling
We used five environmental variables to predict the influence of
Frogs were sampled between the months of March and July in environmental variation on genomic differentiation given their
2013, 2014, and 2015. We collected a total of 191 P. auritus from relevance to amphibian ecology and biology as well as their
four sites in Cameroon (Campo Ma’an (CM), N = 29; Ebo forest (EF), relevance in previous studies (Ficetola and Maiorano, 2016;
N = 24; Ndikinimé ki (ND), N = 13; and Takamanda (TM), N = 15), Morgan et al., 2020). The five bioclimatic variables were: annual
and five sites in Gabon (Gamba Complex (GC), N = 24; Kessala (KS), temperature, temperature seasonality, annual precipitation,
N = 21; Lopé (LP), N = 18; Monts de Cristal (MC), N = 22; and precipitation seasonality, and precipitation of the coldest quarter
Minké bé (MK), N = 25) (Figure 1A). These sites encompassed a range from the WorldClim database (Fick and Hijmans, 2017)
of forest types, namely: lowland rainforest (CM, MK), sub-montane (www.worldclim.org). Precipitation of the coldest quarter was
rainforest (EF, TM), forest-savanna ecotone (KS, MC), coastal selected because it represents a unique characteristic of
rainforest (GC), and mixed lowland-agricultural forest (ND). After precipitation for this region. It effectively reflects the seasonal
frogs were euthanized with MS222 solution, muscle and kidney tissue inversion that occurs approximately across the equator such that
was placed in 95% ethanol prior to DNA extraction. Male frogs were the dry season in central Cameroon coincides with the rainy season

A B

FIGURE 1
Field sampling and inferred population structure of P. auritus. (A) Map of nine sampling locations: Takamanda (TM), Ndikiniméki (ND), Ebo Forest (EF),
Campo Ma’an (CM), Monts de Cristal (MC), Minkébé (MK), Lopé (LP), Kessala (KS), and Gamba Complex (GC). Each point is a sampling site and the
colors correspond to the assigned population. The green shading corresponds to forest cover. (B) PCA of 1631 SNPs. Each point presents a sample,
and samples are colored by site with similar color shades corresponding to their assigned populations. (C) FastStructure results of 1631 SNPs. Site
abbreviations are labeled within each population.

Frontiers in Conservation Science 03 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

in northern Gabon and vice versa (Heuertz et al., 2014; Morgan Rapid Run Mode (Illumina, San Diego, CA, USA) using paired-end
et al., 2020). We then tested for potential correlations between 100-bp sequence reads.
environmental variables to ensure no two variables had a Pearson’s
correlation coefficient > 0.8. All climate variables had a resolution of
30 arc-seconds (approximately 1 km2). Although estimates of 2.4 Bioinformatics analysis of RAD-
vegetation cover may also be important determinants of genomic seq data
differentiation, the lack of climate projection maps prevents the use
of these variables for modeling under future conditions and were We used the bioinformatics software pipeline, STACKS v.1.44
therefore not included in the current study. Future projections of (Catchen et al., 2011; Catchen et al., 2013) to process the restriction-
bioclimatic variables were taken from aggregated global climate site-associated DNA markers (RAD-tags) and generate single
models (Clee, 2017) for two representative concentration pathways nucleotide polymorphism (SNP) datasets. First, we executed the
(RCPs) 2.6 and 8.5, projected for year 2080 based on the “process_radtags” program in STACKS to demultiplex and trim
Intergovernmental Panel on Climate Change (IPCC) 5th sequence reads by the P1 barcodes and remove low quality reads
assessment report. Together, these two RCPs span the greatest (Phred quality score less than 20). After removing PCR duplicates with
range in predicted radiative forcing under climate change and the “clone_filter” script, the processed reads were used to generate RAD
represent “best” and “worst” case scenarios regarding global mean loci without a reference genome using “denovo_map.pl” (parameter
temperature increases. settings: m = 3 M = 5 n = 4). We empirically determined these
parameters to limit the impact of over-splitting loci (Ilut et al., 2014;
Harvey et al., 2015). This involved running the de novo assembly over a
2.3 RAD-seq data wide range of values of M (1–8) with “ustacks”. From these runs, we
selected a value of M = 5 since we observed that the percentage of
We extracted genomic DNA from either kidney or muscle tissue homozygous and heterozygous loci reached a stable value and thus
using Qiagen DNEasy Blood and Tissue kit (Qiagen, CA), following minimized over-splitting of alleles for the final SNP calling.
the manufacturer’s protocol. A total of 164 individuals had genomic Stacks calls SNPs (“sstacks”) within RAD loci using a
DNA of sufficient quantity (> 50 ng) and quality needed for multinomial-based likelihood model that estimates the likelihood of
restriction site-associated sequencing (RAD-seq) as determined the two most frequently observed genotypes at each site and performs
through gel electrophoresis (Davey et al., 2011). RAD-seq library a standard likelihood ratio test using a chi-square distribution
preparation followed the BestRad protocol for Illumina sequencing (Hohenlohe et al., 2010; Catchen et al., 2011). For SNP inference,
as described in Ali et al. (2016). Briefly, genomic DNA (100 ng) was we used the default alpha significance level of 0.05. Paralogous loci
digested with 4.8 units of SbfI-HF restriction enzyme (New England that stacked together were identified and removed by subsequent
Biolabs NEB, R3632L) at 37°C for 1 h in a 12 µl reaction volume. quality control steps built into STACKS (max number of stacks per
Samples were heated to 65°C for 20 min and 4 µl of the indexed loci (m) = 3 (Ilut et al., 2014; Harvey et al., 2015)). After the
BestRad SbfI P1 RAD adapter (10 nM) was added to each sample. preliminary assembly of catalog loci using “denovo_map.pl”, we
Ligation of inline barcoded P1 adaptors to digested genomic DNA ran the STACKS correction mode (rxstacks-cstacks-sstacks) using
was performed overnight at 20°C with 640 units of T4 DNA ligase the bounded SNP model with a 0.05 upper bound for the error rate.
(NEB, M0202M), then 65°C for 20 min. Following ligation, 10 µl of The “rxstacks” program made corrections to genotype and haplotype
each sample in each 48 well plate was pooled into a single tube and calls based on population information, rebuilt the catalog loci and
cleaned using 1x Agencourt AMPure XP beads (A63881; Beckman filtered out loci with average log likelihood ratio of < 8.0.
Coulter). Pooled DNA for each plate was then resuspended in 100 We used three additional filtering steps to generate a set of high-
µl low TE and sheared to an average fragment size of 500 base pairs quality RAD loci for downstream population genetic analysis. First,
using a Bioruptor NGS sonicator (Diagenode). Sheared DNA was we retained only RAD loci that were present in 80% of all samples.
then concentrated to 55.5 µl using Ampure XP beads and used as Second, we removed RAD loci that contained more than 40 SNPs,
the template in the NEBNext Ultra DNA Library Prep Kit for as these likely represented sequencing errors or over-clustering of
Illumina (NEB E7370L; v.1.2). The standard NEBNext protocol for paralogous loci. Lastly, we used the BLAT alignment algorithm
library preparation was followed except that we used custom P2 (Kent, 2002) to de novo align the RAD loci and remove those that
adaptors which were created by annealing a NEBNext Multiplex aligned to multiple positions. The final consensus set of RAD loci
Oligo for Illumina (NEB, E7335L) to the oligonucleotide comprised SNP data from a total of 139 individuals. Genotypes
GATCGGAAGAGCACACGTCTGAACTCC AGTCACIIIIII were called, filtered, and bi-allelic SNPs were exported in VCF
ATCAGAACA*A (where * represents a phosphorothioate DNA format using the STACKS “populations” program. SNPs from
base). In addition, instead of the USER® enzyme step, we used a the last seven bp of the RAD loci were removed as this part of
universal P1 RAD primer (AATGATACGGCGACCACC the locus is likely to contain sequence errors at the 3’ end of the
GAGATCTAC ACTCTTTCCCTACACGAC*G) and a universal reads. The SNP dataset was further filtered with VCFtools v.0.1.14
P2 RAD primer (CAAGCAGAAGACGGCATACG*A) during final (Danecek et al., 2011) to remove SNPs below a minor allele
amplification. The final RAD library was cleaned using AMPure XP frequency (MAF) of 0.05 cutoff to reduce artifacts of sequence
beads and sequenced at the UC Berkeley QB3 Vincent J Coates and assembly error. Finally, the dataset was filtered to include only
Genome Sequencing Laboratory (GSL) on an Illumina HiSeq2500: one random SNP per RAD locus for use in all downstream analyses

Frontiers in Conservation Science 04 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

in order to avoid linkage disequilibrium between SNPs within rivers, using the raster calculator available in QGIS v.2.18. We then
RAD loci. calculated pairwise resistance distances from these raster layers with
CIRCUITSCAPE 4.0 (McRae et al., 2013). Two IBB matrices were
generated, IBB1 and IBB2. For IBB1, resistance values increased with
2.5 Analyses of population increasing elevation and major rivers were treated as impenetrable.
genomic structure For IBB2, resistance increased with increasing elevation and also with
Strahler order, which reflects size and strength of perennial river
We performed a principal component analysis (PCA) using the systems. IBP represents the historical landscape based on modeled
Bioconductor package SNPRelate (Zheng et al., 2012) (https:// available habitat. For IBP, we first projected habitat suitability for
www.bioconductor.org/) to summarize population genomic P. auritus under climate conditions during the LGM using two global
structure. We used the program FastStructure (Raj et al., 2014) to climate models (CCSM and MIROC). We then created resistance
estimate the number of genetically distinct populations within the surfaces where resistance was considered to be inversely proportional
sampled P. auritus range. We tested a range of K values (where K to habitat suitability, and finally, calculated pairwise resistance
denotes the number of inferred populations) from 1 to 10. The distances from this raster layer with CIRCUITSCAPE. Further
script “chooseK.py” included in the FastStructure package was used details on how these predictor variables were generated and the
to determine the best estimate of K that maximizes the marginal resultant distance matrices can be found in Appendix 1. I-spline
likelihood. We also calculated pairwise estimates of FST (Weir and turnover functions describing the relationship between the biological
Cockerham, 1984) among sites and among K populations inferred response variable (pairwise dissimilarity in genomic composition)
from FastStructure using VCFtools. To test for an IBD effect, a and each of the predictor matrices were visualized using rug plots and
Mantel test was used to assess the correlation between pairwise FST their significance was tested using 1000 permutations.
values and geographic distance. Mantel tests were run with 999,999
permutations using the vegan package (Oksanen et al., 2022) in R
and are reported using both raw FST and transformed FST/(1-FST) 2.7 Mapping genomic turnover and
distances, as well as both raw Euclidian geographic distance and predicting patterns of genomic
log-transformed Euclidean distances (Slatkin, 1995; Rousset, 1997). vulnerability under future climate change

We evaluated the importance of environmental variables as

2.6 Quantifying the relative impact of IBE, predictors of environmentally-associated genomic turnover and
IBD, IBB, and IBP on genomic differentiation spatialized these patterns across the study region using GF
modeling within the gradientForest package (Ellis et al., 2012) in
We used GDM to compare the importance of IBE to isolation R (Appendix 1). Response variables were individual SNP minor
by landscape barriers (IBB) or Pleistocene refugia (IBP) on patterns allele frequencies within each population. Predictor variables were
of genomic differentiation. GDM is a matrix regression technique represented by the same environmental variables that were included
that evaluates the relationship between site-site dissimilarities in in the GDM along with latitude and longitude. GF uses a machine-
environmental or landscape ‘predictor’ variables and a biotic learning algorithm to divide the biological data into different bins
‘response’ variable (e.g. pairwise genetic distances). A major (i.e. different values of allele frequencies), with partitions occurring
advantage of GDM over other modeling methodologies is that it at several split values along each environmental variable. This
can fit non-linear relationships between environmental variables binning is performed for every SNP, weighting each SNP
and the biological response variable through the use of I-spline basis individually according to its fit to the model (i.e. R2) before
functions (Ferrier et al., 2007). This approach includes straight-line aggregating across all SNPs. GF determines the “split importance”
geographic distance as a predictor variable and can also incorporate by measuring the amount of biological variation explained by a
a range of environmental data layers, and matrices derived from given split value (e.g. between 26 and 27°C), which is then
resistance surfaces as different predictors. cumulatively summed along each gradient to construct turnover
Pairwise dissimilarity in genomic composition between sites was functions (Fitzpatrick and Keller, 2015). The top three
modeled using two measures: 1) pairwise FST values and 2) a pairwise environmental variables in modeling genomic turnover from a
Bray-Curtis dissimilarity index based on the presence or absence of a total of 2000 regression trees were used to predict and map
SNP at each locus referred to as the Bray-Curtis allele frequency environmentally-associated turnover across the study region using
difference or AFD (Sherwin, 2022). IBE was represented by the set of a random grid of 100,000 sample points. To ensure that our GF
five environmental variables described previously. In addition to model was performing better than random, we shuffled the
these environmental variables, a set of predictor variables were environmental-predictor matrix to generate 200 randomized
generated to model the effect of IBB and IBP under the Last Glacial datasets and compared the number of SNP loci with R2 positive
Maximum approximately 21,000 years ago (i.e. IBP). IBB represents values to the mean R2 value across SNP loci using GF models
physical barriers (elevation and rivers) to gene flow. Pairwise describing variation in the real versus randomized datasets.
resistance distances for IBB were generated by creating raster layers Lastly, we predicted future environmentally-associated genomic
of resistance surfaces based on landscape features, elevation and variation based on projecting GF models under climate change for

Frontiers in Conservation Science 05 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

the years 2050 and 2080 for RCPs 2.6 and 8.5, representing “best” five distinct populations: 1) EF, ND, and TM, 2) CM, 3) MC and
and “worst” cases, respectively. To map predicted changes of MK, 4) KS and LP, and 5) GC (Figure 1C, Appendix 3).
genomic variation associated with environment, we calculated the Pairwise FST values between sites ranged from 0 to 0.438 (mean =
Euclidean distance between models based on current and future 0.229; Appendix 4), indicating low to moderate levels of genomic
climate conditions. Areas where environmentally-associated differentiation between sites. We found a significant correlation
genomic variation changes the least are considered to have low between pairwise FST and geographic distances between the sites
genomic vulnerability whereas areas where they change the most (Mantel r = 0.6728; mantel simulated p-value = 0.001), suggesting a
are considered to have high genomic vulnerability. strong pattern of IBD.

3.3 Quantifying the relative impact of IBE,

3 Results IBD, IBB, and IBP on
genomic differentiation
3.1 SNP variation
The GDM based on pairwise FST values and AFD explained 80%
RAD-sequencing of 139 P. auritus samples generated a total of
and 81% of the variation in the data, respectively. Precipitation of
838,425,400 paired-end reads across both plates (308,255,191;
the coldest quarter was the only significant variable (p = 0.02) and
530,170,209) after filtering out low-quality samples and reads. The
the most important variable in the model with F ST values.
number of raw sequencing reads per sample ranged from 133,185 to
Geographic distance was the only significant variable in the
16 million. The mean coverage depth ranged from 5x to 26x across
model based on AFD (p = 0.01), however precipitation of the
individual samples (mean = 8x, median = 7x, Appendix 2). From
coldest quarter was the most important variable, followed by
these reads, we assembled 2,979 high-quality RAD loci and a total of
geographic distance (Table 1). The I-spline plot for precipitation
32,966 SNPs that were present in 80% or more samples. Using a
of the coldest quarter shows that observed compositional turnover
minor allele frequency cutoff of 5%, we retained 1631 RAD loci
in genomic differentiation increases with greater levels of
encompassing 3,092 SNPs, and from which we selected one random
precipitation (Supplementary Figure S1). Importantly, neither of
SNP per RAD locus.
the two resistance matrices (IBB1, IBB2) modeling the effects of
riverine barriers nor the two matrices modeling the distribution of
suitable habitat since the Pleistocene (IBP-CCSM, IBP-MIROC)
3.2 Population genomic structure had any significant effect on the model (i.e. all matrices had a
coefficient = 0 except IBP-MIROC).
The PCA identified significant population structure across the
sampled range of P. auritus. PC1 explained 25.89% of the variation
and separated the three northern sites (EF, ND, TM) from the 3.4 Landscape patterns of genomic
remaining six sites (CM, MC, MK, LP, KS, GC). PC2 explained turnover and genomic vulnerability
9.26% of variation and separated GC, the southernmost coastal site,
from all other sites (Figure 1B). FastStructure analyses also revealed We used a GF approach to determine associations between SNPs
a pattern of population structure that is organized latitudinally into and environmental variables and map environmentally-associated

TABLE 1 GDM results using genomic data and environmental variables across all models.

FST AFD
Variable Significance (p-values) Importance Significance (p-values)2 Importance2
Geographic distance 0.068 3.185 0.016 9.935

Temperature Seasonality 0.300 0.341 0.681 0.241

Annual Precipitation 0.217 0.114 0.509 0.083

Precipitation Seasonality 0.098 3.368 0.283 2.847

Precipitation of Coldest Quarter 0.024 18.118 0.157 10.671

IBB1 – – – –

IBB2 – – – –

IBP-MIROC – – 0.334 0.010

IBP-CCSM – – – –

P-values and relative importance of significant variables in modeling of both pairwise FST values and genetic dissimilarity based on AFD. The bolded values indicate significant results.

Frontiers in Conservation Science 06 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

genomic turnover across the total study area. A total of 458 SNPs pronounced genomic turnover throughout the Cameroonian
(28% of all SNPs) had R2 values > 0 (average = 0.28) (Figure 2A). highlands (green to orange), forest-savanna ecotone of south-
When testing model performance, the number of SNPs with R2 central Cameroon (orange to green), across the equator (green to
values > 0 for all of the randomized datasets fell below the number blue), and from the coast to the interior of Gabon (purple to
observed for the real data (Supplementary Figure S2) and the mean blue) (Figure 2C).
R2 value generated for the real dataset fell within the upper 95% Predictions of environmentally-associated genomic turnover
quartile of values generated for the randomized datasets under future climate change projections showed similar patterns of
(Supplementary Figure S2), both indicating that the GF model genotype-environment associations across the landscape relative to
shows a stronger association between environmental and genomic current predictions. When we subtracted the current prediction of
variation for our dataset relative to the set of randomized datasets. environmentally-associated genomic turnover from the future
Precipitation of the coldest quarter, latitude, and precipitation predictions under each of the four climate change projections (RCP
seasonality were the most important environmental predictors of 2.6 & RCP 8.5 for the years 2050 and 2080), we found a number of
genomic turnover (Figure 2B). Projected associations between allele areas with high genomic vulnerability across the landscape
frequencies and these three predictor variables revealed areas of (Figure 2D; Supplementary Figure S3). Areas with high genomic

A B

C D

FIGURE 2
GF results with maps of genomic variation and genomic vulnerability. (A) PC plot with labeled vectors indicating the direction and relative magnitude
of environmental variables with the greatest contribution to the predicted patterns of SNP allele frequency differentiation. Vectors for precipitation of
the coldest quarter and latitude are overlapping. Each point is a SNP and the color gradient corresponds to the map in panel C with circles indicating
sampling sites. (B) Environmental and geographic variables ranked by their importance in explaining SNP allele frequency variation. (C) Map of the GF
model of environmentally-associated SNPs for P. auritus. Larger color differences between any two areas in the landscape correspond to larger
genetic differences. Circles indicate sampling sites. (D) Genomic vulnerability under climate projection RCP 8.5 for the year 2080. Red indicates
greater changes in allele frequencies and higher genomic vulnerability, while blue indicates smaller changes in allele frequencies and less
genomic vulnerability.

Frontiers in Conservation Science 07 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

vulnerability (i.e. greater than 50% difference in environmentally- recognized as important drivers of genetic heterogeneity and, as is
associated allele frequencies) occur to the north and south of the the case here, are important for the conservation of
Cameroon highlands and throughout the southwest region of Gabon. evolutionary potential.
Considering all analyses, results show a strong role for both
distance and environment in shaping genomic differentiation in P.
4 Discussion auritus. The role of IBD was supported by findings from GDM,
Mantel tests, and the significance of latitude (but not longitude) in
We adopted a comprehensive statistical approach to disentangling predicting genomic turnover in GF analyses, whereas IBE was
the effects of geographic distance, environmental variation, landscape supported by both GDM and GF. GDM and GF identified
barriers, and Pleistocene refugia on patterns of genomic differentiation precipitation of the coldest quarter as a key driver in genomic
in the African puddle frog P. auritus. Overall, we found differentiation, and GF also identified precipitation seasonality as a
that environmental variation plays an important role in shaping top predictor variable. As further support for the role of environment
patterns of genomic differentiation. This is in addition to, as an important factor, we see those areas of elevated genomic
but independent of, geographic distance. In particular, seasonal turnover span regions of strong ecological transition, corresponding
patterns of precipitation appear to be key in driving patterns of primarily with patterns of seasonal variation in precipitation. This
diversification in this tropical region, in keeping with a recent meta- suggests the role of environmental gradients and ecotones in shaping
analysis conducted of environmentally-mediated selection across adaptive environmentally-associated diversification.
the tropics (Siepielski et al., 2017). Through future modeling Patterns of environmentally-associated genomic differentiation
approaches, we also find that heterogeneous landscapes overlap with reported here are consistent with previous investigations of gene-
patterns of high environmentally-associated genomic variation, environmental associations in this region. For example, precipitation
suggesting that they may play an important role in promoting and has been shown to be an important predictor of patterns of genetic
maintaining biodiversity. variation in central African lizards (Freedman et al., 2010),
First, we addressed whether IBE will influence genomic chimpanzees (Mitchell et al., 2015), birds (Smith et al., 2011), and
differentiation more than IBB or IBP with GDM. Both precipitation forest antelope (Ntie et al., 2017). In the present study, precipitation
of the coldest quarter (FST model) and geographic distance (AFD of the coldest quarter is highest in the Cameroon highlands and
model) are significant predictors of genomic differentiation. decreases progressively throughout central Cameroon and Gabon
Precipitation of the coldest quarter was also important as an (Supplementary Figure S4), mirroring shifts in genomic turnover
explanatory variable in the AFD model, but despite its importance observed in P. auritus. Conversely, precipitation seasonality is more
was not found to be significant. Contrary to many phylogeographic consistent across the study region with subtle increases in seasonality
studies that have been carried out previously in central Africa, we did moving from the Gabon-Cameroon border into northern Cameroon.
not find evidence for an effect of landscape barriers or Pleistocene There are relatively sharper shifts in seasonality with increasing
refugia on population genomic differentiation. These findings are in elevation in the Cameroon highlands. Both precipitation patterns
stark contrast to many previous studies that have placed emphasis on demonstrate shifts in genomic differentiation throughout the
the role of Pleistocene refugia and/or rivers (Eriksson et al., 2004; highlands, across the equator, and subtly from coastal to inland
Anthony et al., 2007; Nicolas et al., 2011; Bohoussou et al., 2015) with Gabon. Gradients in rainfall not only shape the distribution of forest
the exception of Bell et al. (Bell et al., 2017) where rivers were not cover but also present potentially strong selection pressures on the
important in reed frog diversification. phenology of P. auritus since the timing and duration of amphibian
Second, we investigated if areas of greatest environmentally- reproductive events are very sensitive to rainfall levels (Corn, 2005;
associated genomic turnover are associated with strong Ficetola and Maiorano, 2016).
environmental gradients across the landscape with GF. Areas of Precipitation of the coldest quarter is also indicative of seasonal
elevated genomic turnover in P. auritus appear to correspond to patterns in rainfall availability that are inverted across the Equator
known ecological gradients. Genomic turnover is predicted to be high separating Cameroon and Gabon. Rainforests on either side of the
throughout the forest-savanna ecotone region south of the montane equator have their own distinct seasonal patterns of rainfall
region in Cameroon where rainforest habitat in the south gradually (Heuertz et al., 2014) such that the dry season in central
transitions to savanna in the north. These findings are consistent with Cameroon coincides with the rainy season in northern Gabon
patterns of high intraspecific genomic diversity across this ecotonal and vice versa. This seasonal inversion could be responsible for
region in the rainforest bird Andropadus virens (Zhen et al., 2017) the shift in genomic variation observed in P. auritus across the
and soft-furred mouse Praomys misonnei (Morgan et al., 2020). There equator. It has been hypothesized that these contrasting patterns of
is also high genomic turnover in P. auritus across the Cameroon seasonal rainfall could lead to reproductive isolation and speciation
highlands, reflecting both elevation and distance from the coast. across this region (Heuertz et al., 2014). A life history study of P.
The Cameroon highlands are a known biodiversity hotspot, especially auritus found that females lay eggs several times in the year with
for amphibian richness and endemism (Herrmann et al., 2005; breeding peaking during the rainy season (Tasse Taboue and
Pauwels and Rodel, 2007; Stuart et al. 2008; Zimkus and Gvoždı́k, Fokam, 2016). Thus, if populations breed at different times either
2013) so that elevated genomic turnover in this region is to be side of the equator, this could result reproductive isolation and
expected. Mountain ranges and elevational gradients are often account for some of the patterns in genetic differentiation we find

Frontiers in Conservation Science 08 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

here. Future work should look more closely at the seasonal Ethics statement
inversion hypothesis and how heterogeneous annual patterns of
rainfall influence genomic differentiation in other rainforest species. The animal study was approved by University of New Orleans
Finally, we examined the range of genomic vulnerability across the Institutional Animal Care and Use Committee protocol 12-008. The
study region given predicted climate change. We identified multiple study was conducted in accordance with the local legislation and
areas of high genomic vulnerability where populations may be more institutional requirements.
susceptible to climate change under future projections. In Cameroon,
there are patches of high genomic vulnerability alongside the
Cameroon highlands and within the Sanaga basin. Southwest Gabon Author contributions
also encompasses a large area of elevated genomic vulnerability that
contains a matrix of forest and savanna ecosystems. Genomic CM: Conceptualization, Formal analysis, Investigation, Writing –
vulnerability may be an important metric to incorporate into original draft, Project administration. GT: Conceptualization,
conservation prioritization as it may also indicate areas where Methodology, Writing – review & editing. EF: Writing – review &
populations are already susceptible to present-day environmental editing, Project administration. KM: Conceptualization,
pressures. For example, Bay et al. (Bay et al., 2018) have recently Methodology, Writing – review & editing. YZ: Resources, Software,
shown that yellow warbler (Setophaga petechia) populations with the Writing – review & editing. RH: Methodology, Resources, Software,
highest genomic vulnerability were also experiencing the largest Writing – review & editing. VLU: Resources, Writing – review &
population declines. Therefore, areas of high genomic turnover and editing. KR: Resources, Software, Writing – review & editing. PS:
vulnerability may be important targets for future conservation efforts Methodology, Resources, Software, Writing – review & editing. SN:
since the former serves as centers of high adaptive potential whereas the Project administration, Writing – review & editing. PM: Funding
latter signal susceptibility to environmental change. acquisition, Project administration, Supervision, Writing – review &
Although we adopted a genome-wide approach in the present editing. JM: Supervision, Writing – review & editing. TF: Writing –
study, our SNP dataset is only likely to capture a fraction of the total review & editing. BZ: Writing – review & editing. TS: Funding
number of loci in the genome that constitute targets for selection and/ acquisition, Supervision, Writing – review & editing. NA:
or regions of the genome that may be linked loci under selection. Conceptualization, Funding acquisition, Project administration,
Further research should focus on linking genotypic variation to Supervision, Writing – review & editing.
phenotypic traits under selection to understand the evolutionary
significance of divergence more fully across ecological gradients as
well as examine the relative importance of genetic versus
environmental factors that may influence morphological variation. Funding
This could involve assembling and annotating a reference genome
for this species, and sequencing and SNP genotyping candidate genes The author(s) declare financial support was received for the
that could be targets of selection. research, authorship, and/or publication of this article. This work was
Understanding the ecological and historical processes involved in supported by National Science Foundation grant no. OISE 1243524.
diversification is important not only for increasing our knowledge of
evolutionary mechanisms, but also for making evolutionarily
informed conservation decisions to protect biodiversity and Acknowledgments
prioritize new areas for preservation in the light of rapid climate
change. By taking a robust statistical approach to disentangling We thank the Agence Nationale des Parcs Nationaux, ANPN
competing drivers of differentiation, we show that environmental (permit #AE130012), Centre National de la Recherche Scientifique
factors are largely responsible for patterns of genomic differentiation et Technologique, CENAREST (permit #AR0010/13, AR0024/14),
and genomic turnover in our study species. In contrast, landscape Ministère des Forêts et de la Faune, MINFOF (permit #153/AO/
barriers (rivers and elevation) and historical barriers (Pleistocene MINFOF/PNCM, 008/A/MINFOF/R), and Ministère de la
refugia) to gene flow have little influence on genomic differentiation. Recherche Scientifique et de l’Innovation, MINRESI, as well as all
These findings, therefore, highlight the importance of preserving our valuable field guides for helping organize field collections and
heterogeneous environments, such as environmental gradients, in processing samples for exportation. We also thank University of
maintaining species potential to respond to future environmental California, Berkeley’s Vincent J. Coates Genomic Sequencing
change and underline the importance of considering evolutionary Laboratory (GSL), for sequencing services.
processes in the design of future protected areas.

Data availability statement Conflict of interest

The original contributions presented in the study are publicly The authors declare that the research was conducted in the
available. This data can be found here: GitHub, https://github.com/ absence of any commercial or financial relationships that could be
cmiller504/p_auritus. construed as a potential conflict of interest.

Frontiers in Conservation Science 09 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

The author(s) declared that they were an editorial board constant. The slope of each function indicates the rate of compositional
turnover and how this rate varies along the gradient of the predictor variable.
member of Frontiers, at the time of submission. This had no
Variables with all coefficients=0 are not shown because they have no
impact on the peer review process and the final decision. relationship with the modeled biological pattern.

SUPPLEMENTARY FIGURE 2
GF model performance testing results from comparing association between
environmental and genomic variation for our dataset relative to the set of
Publisher’s note randomized datasets. (A) The number of SNPs with R2 values > 0 for all of the
randomized datasets fell below the number observed for the real data. (B) The
All claims expressed in this article are solely those of the authors mean R2 value generated for the real dataset fell within the upper 95% quartile
of values generated for the randomized datasets.
and do not necessarily represent those of their affiliated
organizations, or those of the publisher, the editors and the SUPPLEMENTARY FIGURE 3
Map of genomic vulnerability across the landscape for each of the four
reviewers. Any product that may be evaluated in this article, or
climate change projections (RCP 2.6 & RCP 8.5 for the years 2050 and
claim that may be made by its manufacturer, is not guaranteed or 2080). Red indicates greater changes in allele frequencies and thus, higher
endorsed by the publisher. genomic vulnerability, while blue indicates smaller changes in allele
frequencies and less genomic vulnerability.

SUPPLEMENTARY FIGURE 4
Map of precipitation of the coldest quarter across the study region. Darker
Supplementary material shades of green correspond to more precipitation, measured in millimeters (one
millimeter of rainfall is the equivalent of one liter of water per square meter).
The Supplementary Material for this article can be found online
APPENDIX 2
at: https://www.frontiersin.org/articles/10.3389/fcosc.2024.1366248/ Output from Stacks including raw sequencing reads and mean coverage
full#supplementary-material depth per sample.

SUPPLEMENTARY FIGURE 1 APPENDIX 3

GDM I-spline plots. Fitted functions of observed compositional turnover in FastStructure results including marginal likelihood for all values of K.
genomic differentiation for the five significant predictor variables. The
maximum height of each curve indicates the total amount of compositional APPENDIX 4
turnover associated with that variable, while holding all other variables Pairwise FST values and Mantel test results.

References
Abernethy, K. A., Coad, L., Taylor, G., Lee, M. E., and Maisels, F. (2013). Extent and Domain: evidence for the role of refugia in the intraspecific differentiation of
ecological consequences of hunting in Central African rainforests in the twenty-first Aucoumea klaineana. Mol. Ecol. 20 (1), 131–142. doi: 10.1111/mec.2010.20.issue-1
century. Philos. Trans. R. Soc. B: Biol. Sci. 368:20120303. doi: 10.1098/rstb.2012.0303
Catchen, J., Hohenlohe, P. A., Bassham, S., Amores, A., and Cresko, W. A. (2013).
Aleixo, A. (2004). Historical diversification of a terra-firme forest bird superspecies: Stacks: An analysis tool set for population genomics. Mol. Ecol. 22, 3124–3140.
A phylogeographic perspective on the role of different hypotheses of amazonian doi: 10.1111/mec.12354
diversification. Evol. (N Y) 58, 1303–1317. doi: 10.1111/j.00143820.2004.tb01709.x
Catchen, J. M., Amores, A., Hohenlohe, P., Cresko, W., and Postlethwait, J. H.
Ali, O. A., O’Rourke, S. M., Amish, S. J., Meek, M. H., Luikart, G., Jeffres, C., et al.
(2011). Stacks: Building and genotyping loci de novo from short-read sequences. G3:
(2016). RAD capture (Rapture): flexible and efficient sequence-based genotyping.
Genes Genomes Genet. 1, 171–182. doi: 10.1534/g3.111.000240
Genetics 202 (2), 389–400. doi: 10.1534/genetics.115.183665
Clee, P. R. S. (2017). Distributions, Drivers, and Risks of Wildlife Infectious Diseases
Anthony, N. M., Atteke, C., Bruford, M. W., Dallmeier, F., Freedman, A., Hardy, O.,
across Africa: using geospatial analyses to elucidate disease occurrence in biodiversity
et al. (2015). Evolution and conservation of central African biodiversity: priorities for
hotspots. Drexel University.
future research and education in the Congo basin and Gulf of Guinea. Biotropica 47 (1),
6–17. doi: 10.1111/btp.12188 Colyn, M., Gautier-Hion, A., and Verheyen, W. (1991). A Re-Appraisal of
Palaeoenvironmental History in Central Africa: Evidence for a Major Fluvial Refuge
Anthony, N. M., Johnson-Bawe, M., Jeffery, K., Clifford, S. L., Abernethy, K. A.,
in the Zaire basin. J. Biogeogr 18, 403–407. doi: 10.2307/2845482
Tutin, C. E., et al. (2007). The role of Pleistocene refugia and rivers in shaping gorilla
genetic diversity in central Africa. Proc. Natl. Acad. Sci. U.S.A. 104 (51), 20432–20436. Corn, P. S. (2005). Climate change and amphibians. Anim. Biodivers Conserv. 28, 59–
doi: 10.1073/pnas.0704816105 67. doi: 10.32800/abc
Balkenhol, N., Dudaniec, R. Y., Krutovsky, K. V., Johnson, J. S., Cairns, D. M., Cushman, S. A., McKelvey, K. S., Hayden, J., and Schwartz, M. K. (2006). Gene flow
Segelbacher, G., et al. (2017). Landscape genomics: understanding relationships in complex landscapes: testing multiple hypotheses with causal modeling. The
between environmental heterogeneity and genomic characteristics of populations. In: American Naturalist. 168 (4), 486–499. doi: 10.1086/506976
O. Rajora (eds) Population Genomics. Population Genomics. Springer, Cham., 261–322.
doi: 10.1007/13836_2017_2 Danecek, P., Auton, A., Abecasis, G., Albers, C. A., Banks, E., DePristo, M. A., et al.
(2011). The variant call format and VCFtools. Bioinformatics 27 (15), 2156–2158.
Bay, R. A., Harrigan, R. J., Underwood, V. L., Gibbs, H. L., Smith, T. B., and Ruegg, K. doi: 10.1093/bioinformatics/btr330
(2018). Genomic signals of selection predict climate-driven population declines in a
migratory bird. Sci. (1979) 361, 83–86. doi: 10.1126/science.aan4380 Davey, J. W., Hohenlohe, P. A., Etter, P. D., Boone, J. Q., Catchen, J. M., and Blaxter,
M. L.. (2011). Genome-wide genetic marker discovery and genotyping using next-
Bell, R. C., Parra, J. L., Badjedjea, G., Barej, M. F., Blackburn, D. C., Burger, M., et al. generation sequencing. Nat. Rev. Genet. 12 (7), 499–510. doi: 10.1038/nrg3012
(2017). Idiosyncratic responses to climate-driven forest fragmentation and marine
incursions in reed frogs from Central Africa and the Gulf of Guinea Islands. Mol. Ecol. Davis, M. B., Shaw, R. G., and Etterson, J. R. (2005). Evolutionary responses to
26 (19), 5223–5244. doi: 10.1111/mec.14260 changing climate. Ecology 86, 1704–1714. doi: 10.1890/03-0788
Ellis, N., Smith, S. J., and Roland Pitcher, C. (2012). Gradient forests: Calculating
Bohoussou, K. H., Cornette, R., Akpatou, B., Colyn, M., Kerbis Peterhans, J., Kennis, importance gradients on physical predictors. Ecology 93, 156–168. doi: 10.1890/11-0252.1
J., et al. (2015). The phylogeography of the rodent genus Malacomys suggests multiple
Endler, J. A. (1973). Gene flow and population differentiation. Sci. (1979) 179, 243–
Afrotropical Pleistocene lowland forest refugia. J. Biogeogr 42 (11), 2049–2061.
250. doi: 10.1126/science.179.4070.243
doi: 10.1111/jbi.12570
Eriksson, J., Hohmann, G., Boesch, C., and Vigilant, L. (2004). Rivers influence the
Born, C., Alvarez, N., McKey, D., Ossari, S., Wickings, E. J., Hossaert-McKey, M., population genetic structure of bonobos (Pan paniscus). Mol. Ecol. 13, 3425–3435.
et al. (2011). Insights into the biogeographical history of the Lower Guinea Forest doi: 10.1111/j.1365-294X.2004.02332.x

Frontiers in Conservation Science 10 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

Ferrier, S., Manion, G., Elith, J., and Richardson, K. (2007). Using generalized Praomys misonnei in tropical Africa. J. Biogeogr 38 (1), 191–207. doi: 10.1111/
dissimilarity modelling to analyse and predict patterns of beta diversity in regional jbi.2010.38.issue-1
biodiversity assessment. Divers. Distrib 13, 252–264. doi: 10.1111/j.1472- Nogué s-Bravo, D. (2009). Predicting the past distribution of species climatic niches.
4642.2007.00341.x Global Ecol. Biogeography 18, 521–531. doi: 10.1111/j.1466-8238.2009.00476.x
Ficetola, G. F., and Maiorano, L. (2016). Contrasting effects of temperature and Ntie, S., Davis, A. R., Hils, K., Mickala, P., Thomassen, H. A., Morgan, K., et al.
precipitation change on amphibian phenology, abundance and performance. Oecologia (2017). Evaluating the role of Pleistocene refugia, rivers and environmental variation in
181, 683–693. doi: 10.1007/s00442-016-3610-9 the diversification of central African duikers (genera Cephalophus and Philantomba).
BMC Evol. Biol. 17, 1-17. doi: 10.1186/s12862-017-1054-4
Fick, S. E., and Hijmans, R. J. (2017). WorldClim 2: new 1-km spatial resolution
climate surfaces for global land areas. Int. J. Climatology 37, 4302–4315. doi: 10.1002/ Oksanen, J., Simpson, G., Blanchet, F., Kindt, R., Legendre, P., Minchin, P., et al.
joc.5086 (2022). Vegan: community ecology package. R package version 2. 6–4. Available at:
https://CRAN.R-project.org/package=vegan.
Fitzpatrick, M. C., and Keller, S. R. (2015). Ecological genomics meets community-
Oates, J. F., Bergl, R. A., and Linder, J. M. (2004). Africa’s gulf of Guinea forests:
level modelling of biodiversity: Mapping the genomic landscape of current and future
biodiversity patterns and conservation priorities. Washington, DC: Conservation
environmental adaptation. Ecol. Lett. 18, 1–16. doi: 10.1111/ele.12376
International. 90.
Freedman, A. H., Thomassen, H., Buermann, W., and Smith, T. B. (2010). Genomic
Pauwels, O. S. G., and Rodel, M. (2007). Amphibians and National Parks in Gabon,
signals of diversification along ecological gradients in a tropical lizard. Mol. Ecol. 19,
western Central Africa Amphibien und Nationalparks in Gabun, westliches
3773–3788. doi: 10.1111/j.1365-294X.2010.04684.x
Zentralafrika. Herpetozoa 19, 135–148. Available at: https://pauwelsolivier.com/docs/
Gunderson, L. H. (2000). Ecological resilience-in theory and application. Annu. Rev. AmphibiansAndGabonNP_2007.pdf.
Ecol. Syst. 31, 425–439. doi: 10.1146/annurev.ecolsys.31.1.425 Plana, V. (2004). Mechanisms and tempo of evolution in the African Guineo-
Gvoždı́k, V., Nečas, T., Dolinay, M., Zimkus, B. M., Schmitz, A., and Fokam, E. B. Congolian rainforest. Philos. Trans. R. Soc. B: Biol. Sci. 359, 1585–1594. doi: 10.1098/
(2020). Evolutionary history of the Cameroon radiation of puddle frogs rstb.2004.1535
(Phrynobatrachidae: Phrynobatrachus), with descriptions of two critically Qué r ouil, S., Verheyen, E., Dillen, M., and Colyn, M. (2003). Patterns of
endangered new species from the northern Cameroon Volcanic Line. PeerJ 8, e8393. diversification in two African forest shrews: Sylvisorex johnstoni and Sylvisorex
doi: 10.7717/peerj.8393 ollula (Soricidae, Insectivora) in relation to paleo-environmental changes. Mol.
Harvey, M. G., Judy, C. D., Seeholzer, G. F., Maley, J. M., Graves, G. R., and Phylogenet Evol. 28, 24–37. doi: 10.1016/S1055-7903(03)00027-7
Brumfield, R.T. (2015). Similarity thresholds used in DNA sequence assembly Raj, A., Stephens, M., and Pritchard, J. K. (2014). FastSTRUCTURE: Variational
fromshort reads can reduce the comparability of population histories across species. inference of population structure in large SNP data sets. Genetics 197, 573–589.
PeerJ 3, e895. doi: 10.7717/peerj.895 doi: 10.1534/genetics.114.164350
Herrmann, H.-W., Böhme, W., Euskirchen, O., Hermann, P. A., and Schmitz, A. Rousset, F. (1997). Genetic differentiation and estimation of gene flow from F-statistics
(2005). African biodiversity hotspots: the amphibians of Mt Nlonako, Cameroon. under isolation by distance. Genetics 145, 1219–1228. doi: 10.1093/genetics/145.4.1219
Salamandra 41, 61–81.
Ruegg, K., Bay, R. A., Anderson, E. C., Saracco, J. F., Harrigan, R. J., Whitfield, M.,
Heuertz, M., Duminil, J., Dauby, G., Savolainen, V., and Hardy, O. J. (2014). et al. (2018). Ecological genomics predicts climate vulnerability in an endangered
Comparative phylogeography in rainforest trees from lower Guinea, Africa. PloS One southwestern songbird. Ecol. Lett. 21 (7), 1085–1096. doi: 10.1111/ele.12977
9 (1), e84307. doi: 10.1371/journal.pone.0084307
Sgrò, C. M., Hoffmann, A. A., and Lowe, A. J. (2011). Building evolutionary resilience
Hohenlohe, P. A., Bassham, S., Etter, P. D., Stiffler, N., Johnson, E. A., and Cresko, W. for conserving biodiversity under climate change. Evol. Appl. 4 (2), 326–337.
A. (2010). Population genomics of parallel adaptation in threespine stickleback using doi: 10.1111/j.1752-4571.2010.00157.x
sequenced RAD tags. PloS Genet. 6 (2), e1000862. doi: 10.1371/journal.pgen.1000862
Shafer, A. B. A., and Wolf, J. B. W. (2013). Widespread evidence for incipient
Holt, R. D. (1990). The microevolutionary consequences of climate change. Trends ecological speciation: A meta-analysis of isolation-by-ecology. Ecol. Lett. 16, 940–950.
Ecol. Evol. 5, 311–315. doi: 10.1016/0169-5347(90)90088-U doi: 10.1111/ele.12120
Ilut, D. C., Nydam, M. L., and Hare, M. P. (2014). Defining loci in restriction-based Sherwin, W. B. (2022). Bray-Curtis (AFD) differentiation in molecular ecology:
reduced representation genomic data from nonmodel species: Sources of bias and Forecasting, an adjustment (AA), and comparative performance in selection detection.
diagnostics for optimal clustering. BioMed. Res. Int. 2014, 9. doi: 10.1155/2014/675158 Mol. Ecol. 12 (9), e9176. doi: 10.1002/ece3.9176
James, R., Washington, R., and Rowell, D. P. (2013). Implications of global warming Siepielski, A. M., Morrissey, M. B., Buoro, M., Carlson, S. M., Caruso, C. M., Clegg, S.
for the climate of African rainforests. Philos. Trans. R. Soc. B: Biol. Sci. 368, 20120298. M., et al . (2017). Precipitation drives global variation in natural selection. Sci. (1979)
doi: 10.1098/rstb.2012.0298 355 (6328), 959–962. doi: 10.1126/science.aag2773
Kent, W. J. (2002). BLAT—The BLAST-like alignment tool. Genome Res. 12, 656–664. Slatkin, M. (1995). A measure of population subdivision based on microsatellite allele
http://www.genome.org/cgi/doi/10.1101/ frequencies. Genetics 139, 457–462. doi: 10.1093/genetics/139.1.457
Laporte, N. T., Stabach, J. A., Grosch, R., Lin, T. S., and Goetz, S. J. (2007). Expansion Smith, T. B., Thomassen, H. A., Freedman, A. H., Sehgal, R. N., Buermann, W.,
of industrial logging in Central Africa. Sci. (1979) 316, 1451. doi: 10.1126/ Saatchi, S., et al. (2011). Patterns of divergence in the olive sunbird Cyanomitra olivacea
science.1141057 (Aves: Nectariniidae) across the African rainforest-savanna ecotone. Biol. J. Linn. Soc.
103 (4), 821–835. doi: 10.1111/bij.2011.103.issue-4
Manthey, J. D., and Moyle, R. G. (2015). Isolation by environment in white-breasted
nuthatches (Sitta carolinensis) of the Madrean Archipelago sky islands: A landscape Smith, T. B., Wayne, R. K., Girman, D. J., and Bruford, M. W. (1997). A role for
genomics approach. Mol. Ecol. 24, 3628–3638. doi: 10.1111/mec.13258 ecotones in generating rainforest biodiversity. Sci. (1979) 276, 1855–1857. doi: 10.1126/
science.276.5320.1855
McRae, B. H., and Beier, P. (2007). Circuit theory predicts gene flow in plant and
Stuart, S. N., Hoffmann, M., Chanson, J. S., Cox, N. A., Berridge, R. J., Ramani, P.,
animal populations. Proc. Natl. Acad. Sci. U.S.A. 104, 19885–19890. doi: 10.1073/
et al. (eds.) (2008). Threatened amphibians of the world. Lynx Edicions with IUCN—The
pnas.0706568104
World Conservation Union, Conservation International and NatureServe. (Lynx
McRae, B. H., Shah, V. B., Mohapatra, T. K., and Anantharaman, R. (2013). Edicions, Barcelona, Spain: IUCN, Gland, Switzerland; and Conservation
Circuitscape 4 (Seattle WA: The Nature Conservancy). International, Arlington, Virginia, USA), 16.
Mitchell, M. W., Locatelli, S., Sesink Clee, P. R., Thomassen, H. A., and Gonder, M. Tasse Taboue, G. C., and Fokam, E. B. (2016). Life history of the golden puddle frog,
K. (2015). Environmental variation and rivers govern the structure of chimpanzee phrynobatrachus auritus boulenger 1900 (Anura: phrynobatrachidae). Int. J. Biol. 8, 77.
genetic diversity in a biodiversity hotspot. BMC Evol. Biol. 15, 1–13. doi: 10.1186/ doi: 10.5539/ijb.v8n3p77
s12862-014-0274-0 Telfer, P. T., Souquiere, S., Clifford, S. L., Abernethy, K. A., Bruford, M. W., Disotell,
Mittermeier, R. A., Mittermeier, C. G., Brooks, T. M., Pilgrim, J. D., Konstant, W. R., T. R., et al. (2003). Molecular evidence for deep phylogenetic divergence in Mandrillus
Da Fonseca, G. A., et al. (2003). Wilderness and biodiversity conservation. Proc. Natl. sphinx. Mol. Ecol. 12 (7), 2019–2024. doi: 10.1046/j.1365-294X.2003.01877.x
Acad. Sci. U.S.A. 100 (18), 10309–10313. doi: 10.1073/pnas.1732458100 Termignoni-Garcı́a, F., Jaramillo-Correa, J. P., Chablé -Santos, J., Liu, M., Shultz, A.
Morgan, K., Mboumba, J.-F., Ntie, S., Mickala, P., Miller, C. A., Zhen, Y., et al. (2020). J., Edwards, S. V., et al . (2017). Genomic footprints of adaptation in a cooperatively
Precipitation and vegetation shape patterns of genomic and craniometric variation in breeding tropical bird across a vegetation gradient. Mol. Ecol. 26 (17), 4483–4496.
the central African rodent Praomys misonnei. Proc. Biol. Sci. 287, 20200449. doi: 10.1111/mec.14224
doi: 10.1098/rspb.2020.0449 Thomassen, H. A., Fuller, T., Buermann, W., Mila, B., Kieswetter, C. M., Jarrín-V, P.,
et al. (2011). Mapping evolutionary process: a multi-taxa approach to conservation
Moritz, C., Patton, J. L., Schneider, C. J., and Smith, T. B. (2000). Diversification of prioritization. Evol. Appl. 4 (2), 397–413. doi: 10.1111/j.1752-4571.2010.00172.x
rainforest faunas: An integrated molecular approach. Annu. Rev. Ecol. Syst. 31, 533–
Thrush, S. F., Hewitt, J. E., Dayton, P. K., Coco, G., Lohrer, A. M., Norkko, A., et al.
563. doi: 10.1146/annurev.ecolsys.31.1.533
(2009). Forecasting the limits of resilience: Integrating empirical research with theory.
Murienne, J., Benavides, L. R., Prendini, L., Hormiga, G., and Giribet, G. (2013). Proc. R. Soc. B: Biol. Sci. 276 (1671), 3209–3217. doi: 10.1098/rspb.2009.0661
Forest refugia in Western and Central Africa as ‘museums’ of Mesozoic biodiversity. Wang, I. J., and Bradburd, G. S. (2014). Isolation by environment. Mol. Ecol. 23,
Biol. Lett. 9, 20120932. doi: 10.1098/rsbl.2012.0932 5649–5662. doi: 10.1111/mec.12938
Nicolas, V., Missoup, A. D., Denys, C., Kerbis Peterhans, J., Katuala, P., Couloux, A., Weir, B. S., and Cockerham, C. C. (1984). Estimating F-statistics for the analysis of
et al. (2011). The roles of rivers and Pleistocene refugia in shaping genetic diversity in population structure. Evolution 38 (6), 1358–1370. doi: 10.2307/2408641

Frontiers in Conservation Science 11 frontiersin.org

Miller et al. 10.3389/fcosc.2024.1366248

Zhen, Y., Harrigan, R. J., Ruegg, K. C., Anderson, E. C., Ng, T. C., Lao, S., et al. (2017). Zimkus, B. M., and Gvoždı́k, V. (2013). Sky Islands of the Cameroon Volcanic Line:
Genomic divergence across ecological gradients in the Central African rainforest A diversification hot spot for puddle frogs (Phrynobatrachidae: Phrynobatrachus). Zool
songbird (Andropadus virens). Mol. Ecol. 26 (19), 4966–4977. doi: 10.1111/mec.14270 Scr 42, 591–611. doi: 10.1111/zsc.12029
Zheng, X., Levine, D., Shen, J., Gogarten, S. M., Laurie, C., and Weir, B.S. (2012). A Zimkus, B. M., and Schick, S. (2010). Light at the end of the tunnel: insights into the
high-performance computing toolset for relatedness and principal component analysis molecular systematics of East African puddle frogs (Anura: Phrynobatrachidae). Syst.
of SNP data. Bioinformatics 28 (24), 3326–3328. doi: 10.1093/bioinformatics/bts606 Biodivers 8, 39–47. doi: 10.1080/14772000903543004

Frontiers in Conservation Science 12 frontiersin.org