The Gambian epauletted fruit bat shows increased genetic divergence in the Ethiopian highlands and in an area of rapid urbanization

Abstract The Gambian epauletted fruit bat (Epomophorus gambianus) is an abundant species that roosts in both urban and rural settings. The possible role of E. gambianus as a reservoir host of zoonotic diseases underlines the need to better understand the species movement patterns. So far, neither observational nor phylogenetic studies have identified the dispersal range or behavior of this species. Comparative analyses of mitochondrial and nuclear markers from 20 localities across the known distribution of E. gambianus showed population panmixia, except for the populations in Ethiopia and southern Ghana (Accra and Ve‐Golokwati). The Ethiopian population may be ancestral and is highly divergent to the species across the rest of its range, possibly reflecting isolation of an ancient colonization along an east–west axis. Mitochondrial haplotypes in the Accra population display a strong signature of a past bottleneck event; evidence of either an ancient or recent bottleneck using microsatellite data, however, was not detected. Demographic analyses identified population expansion in most of the colonies, except in the female line of descent in the Accra population. The molecular analyses of the colonies from Ethiopia and southern Ghana show gender dispersal bias, with the mitochondrial DNA fixation values over ten times those of the nuclear markers. These findings indicate free mixing of the species across great distances, which should inform future epidemiological studies.

Long-distance animal movements also can drive the transmission of pathogens within and between species, shaping epidemiological dynamics among wildlife populations. For example, Eidolon helvum, the most populous large fruit bat in sub-Saharan Africa that is often found in urban areas including megacities (DeFrees & Wilson, 1988;Hayman, McCrea, et al., 2012), has the largest panmictic population among terrestrial mammals, showing similar seroprevalences against henipaviruses and Lagos bat virus among disparate continental African countries (Peel et al., 2013). Animal dispersal may have important implications for public health, but the true role that these movement patterns play in pathogen transmission is still not well understood (Suzán et al., 2015). For example, although there is a generalized assumption that migratory animals increase pathogen dispersal (Figuerola & Green, 2000;Rappole, Derrickson, & Hubalek, 2000;Reed, Meece, Henkel, & Shukla, 2003), it has been suggested that in some circumstances, the opposite can be true, for example, migration can allow healthy hosts to escape infected habitats, reducing the impact of disease on a population (Altizer, Bartel, & Han, 2011;Hall, Altizer, & Bartel, 2014).
This highlights the need for accurate data and a better understanding of animal movement, particularly for potential reservoir species of zoonotic diseases.
Epomophorus gambianus (Figure 1), commonly known as the Gambian epauletted fruit bat, is a potential reservoir host of Ebola virus (Hayman, Yu, et al., 2012). Across its distribution (Figure 2), E. gambianus has been reported to roost in small colonies of up to 100 individuals (Boulay & Robbins, 1989). It is described as a lowland species usually found below 500 meters above sea level (m a.s.l.), apart from in Ethiopia, where it has been reported to occur up to nearly 2,000 m a.s.l. (Mickelburgh, Hutson, & Bergmans, 2008). Epomophorus gambianus is a medium-sized bat that has not been described previously as undergoing migration or long-distance dispersal (Boulay & Robbins, 1989;Mickelburgh et al., 2008). Bats which fly long distances have morphological characteristics (ecomorphology) that enable energy-efficient flight, such as a high aspect ratio (long, narrow wings), which favors aerodynamic efficiency and lower losses of energy in flight, and high wing loading (low wing area relative to body mass), which correlates with high speed flights but low maneuverability (Norberg & Rayner, 1987;Olival, 2012). Norberg and Rayner (1987) determined that E. gambianus has the characteristics of a fast, maneuverable and agile flyer (e.g., low aspect ratio, relatively short wingspan, high wing loading, and an average wingtip shape) that are not typical features for long-distance flight.
Population genetics has been increasingly used to elucidate wildlife movement, particularly for species that are difficult to track directly. Mitochondrial DNA (mtDNA) has historically been selected as a molecular marker for phylogeographic studies and has also been widely used for the study of speciation (Boattini et al., 2013;Song, Lan, & Kohn, 2014;Talbot, Vonhof, Broders, Fenton, & Keyghobadi, 2016). For example, sequencing of the mitochondrial cytochrome b gene (CYTB) revealed a polyphyletic relationship between E. gambianus and Micropteropus pusillus (species within the Epomophorini tribe). Using only a region of the mtDNA, however, did not robustly identify introgression between these species (Nesi, Nakoune, Cruaud, & Hassanin, 2011). To assess this, either complete mitochondrial genomes can be used (Riesle-Sbarbaro et al., 2016) or the analysis can be complemented using biparentally inherited markers such as microsatellites. Microsatellites are nuclear DNA (ncDNA) markers commonly used in phylogeographic studies of populations (Goldstein & Pollock, 1997;Hindley, Graham, Pulgarin, & Burg, 2018;Muriira, Muchugi, Yu, Xu, & Liu, 2018;Rossiter, Benda, Dietz, Zhang, & Jones, 2007). So far, the population-based phylogeography of E. gambianus has not been investigated. In this study, we aim to determine the genetic structure of this species across its range using both nuclear and mitochondrial markers, not only to increase the currently limited knowledge of the ecology and demographic history of this bat, but also to inform future epidemiological studies, by answering the following questions:

| Sample collection
Tissue samples or extracted DNA of 308 E. gambianus was collected from 20 localities from across the species geographical distribution (IUCN, 2016) along a linear east-west axis ( Figure 2; Table 1).
Eleven colonies were sampled within Ghana, 2013-2015, by the collection of 3-to 4-mm-diameter wing membrane biopsies of using a biopsy punch (Henry Schein, UK), while tissue samples from Nigeria and Ethiopia and extracted DNA samples from the Central African Republic were acquired from museum specimens (Table 1).

| Sequencing and genotyping
Genomic DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen Ltd., UK). Samples obtained from museum collections were extracted using the QIAamp DNA FFPE Tissue Kit (Qiagen Ltd., UK). Tissues sampled from the pectoral muscle were digested overnight (24 hr) using 180 μl of ATL buffer and 40 μl of proteinase K. A paired-end Illumina sequencing library was constructed, as previously described (Riesle-Sbarbaro et al., 2016), and the mitochondrial genome of E. gambianus was assembled and annotated. Primers for CYTB and D-loop regions were selected (see Supporting information, Table S1) for Sanger sequencing, and DNA was amplified in 10 μl of reaction mix containing 2 ng of template DNA, 10 μM of forward and reverse primers, and 5 μl of MegaMix-Gold master mix. Touchdown PCR settings used to amplify the CYTB fragments were as follows: 5 min at 95°C; followed by 12 cycles of 20 s at 94°C, 20 s at 66°C (decreasing one degree per cycle), and 20 s at 72°C; 30 cycles of 1 min at 94°C, 1 min at 54°C, and 1 min at 72°C; and a final extension of 7 min at 72°C; and for D-loop, the conditions used were as follows: 5 min at 95°C; 40 cycles of 1 min at 93°C, 90 s at 55°C and 2 min at 72°C; and a final extension of 7 min at 72°C. PCR products were screened using 2% agarose gel electrophoresis and purified using the Exonuclease I/ Shrimp Alkaline Phosphatase Method (ExoSAP-IT kit, Affymetrix Ltd., UK). PCR products were Sanger sequenced externally (Source Bioscience, UK, Ltd.), and Geneious v8.1 software was used for quality-trimming and alignment of the sequences. Both CYTB and D-loop sequences were realigned using the software Gblocks 0.91b (Castresana, 2000) under semistringent parameters. CYTB sequences were kept in two conserved alignment blocks retaining 99% of the original alignment (532 bp). Due to the high variability of D-loop sequences, some PCRs failed to generate unique products, so four alignment blocks were generated using only 75% of the original sequence information (414 bp).
Microsatellite primers for nuclear DNA (ncDNA) were developed using E. gambianus paired-end reads (that passed the Illumina filters) imported to the "SSR_Pipeline" (Miller, Knaus, Mullins, & Haig, 2013), which was programmed to identify simple microsatellite markers that presented at least 9 repeats of dinucleotide motifs or at least 5 repeats for tetranucleotide motifs (Supporting information, Table S2). From this output, an initial selection of 34 nuclear microsatellites was filtered to 20, according to the reproducibility of scoring and heterozygosity detected in 20 bats. Forward primers were extended with universal primers M13, T7, SP6, and T3 to their 5′ end to indirectly label them with fluorescent dyes (NED,VIC,and PET,respectively

| Characterization of molecular markers
Samples from 308 bats were initially obtained for genotyping, but due to poor quality or low yield of DNA (reflected as ambiguous peaks or no amplification of SSR), 277 bats (144 females and 133 males) were included in subsequent analyses. The statistical power of both mtDNA and microsatellite markers in the dataset to detect significant population differentiation was tested using the software POWSIM (Ryman & Palm, 2006). A thousand simulations were run using the empirical values of the ncDNA dataset, detecting F ST values from 0.001 to 0.01. To test the CYTB sequences, the dataset was adjusted for the organelle data (mtDNA) halving sample size (Larsson, Charlier, Laikre, & Ryman, 2008 or had a large proportion of missing peak calls (L23 and L15) were excluded from the analyses (Supporting information, Table S2), and HWE equilibrium of the colonies was checked afterward. Linkage disequilibrium (LD) was tested using the R package Adegenet and the software GENEPOP version 4.2 (Rousset, 2008). LD was tested using MCMC (100 replicates, 10,000 dememorization steps, and 10,000 randomizations). An additional verification of these results was performed using the software FSTAT version 2.9.3.2 (Goudet, 1995) and GenoDive version 2.0b27 (Meirmans & Van Tienderen, 2004).  (Piry, Luikart, & Cornuet, 1999) and M_P_Val (Garza & Williamson, 2001) were used to detect recent and past effective population size reduction from the ncDNA allele data of 17 loci. The BOTTLENECK analyses were calculated using as input values 12% of variance and 95% of single-step mutations, as suggested by Piry et al. (1999), for 5,000 replications. The M ratio analyses, which calculate the mean ratio of allele size per locus to the range of alleles sizes and compare it with simulated values excepted under equilibrium (significance assumed as <5% of the M ratios generated below the observed value), were calculated assuming
Class III statistics (mismatch distributions) were analyzed for both spatial and demographic expansion models using ARLEQUIN version 3.5 (Excoffier, Laval, & Schneider, 2005). The demographic expansion model did not converge for the AC, YE, and ET populations; therefore, these indices and statistics are not shown for any colony.
For this, we used a substitution model HKY, selected in jModelTest version 2.1.7 (Posada, 2008), using strict linked and unlinked clock models with the concatenated (CYTB and D-loop) mtDNA fragment.
The MCMC was run twice for 30 million generations, sampling every 10,000 generations, discarding 10% as burn-in and combined using the package LogCombiner 2.4.8. We compared models using TRACER 1.7.1 (Rambaut, Drummond, Xie, Baele, & Suchard, 2018) and selected the ESBP linked model as a better fit. The plot was done in R.

| Phylogenetic analyses
A best nucleotide substitution model for the CYTB and concatenated mtDNA alignments was calculated with the Akaike information criterion (AIC) and the Bayesian information criterion in jModelTest.
The model selected for the CYTB alignment was HKY + G, and for the concatenated fragment, the model selected was HKY + I + G. A gamma mixed model was also used for the latter. In order to quantify divergence times between the clades of E. gambianus, a CYTB substitution clock of 3%/My was initially used in Mega5 software (Hulva, Horacek, Strelkov, & Benda, 2004;Nabholz, Glemin, & Galtier, 2008;Tamura et al., 2011). The resulting divergence time, presented as millions of years ago (Mya), was checked and further calibrated against the best estimates of divergence time between the out-group species (Rousettus aegyptiacus and Epomops franqueti) and E. gambianus (Hedges, Dudley, & Kumar, 2006;Hedges, Marin, Suleski, Paymer, & Kumar, 2015). Haplotype alignments were generated using the software DNAsp v5.10 for the processed fragments. Graphical representations of the interspecific relationships of the individuals were generated using tree-like phylogenies and haplotype networks.
Maximum-likelihood trees (Supporting information, Figure S3B-D) were run with 1,000 bootstrap supports, and the Bayesian models were run with 6 simultaneous chains, sampled every 100 generations for 10 9 generations, or until the standard deviation split frequencies reach 0.01. The first 25% of the trees were discarded. The output files were processed with FIGTREE (https://tree.bio.ed.ac.
uk/software/figtree/). The haplotype networks were constructed with the software NETWORK version 5.0 (www.fluxus-engineering. com) using a Median-Joining algorithm (Bandelt, Forster, & Röhl, 1999). To decrease the complexity of the reticulations in the graph, mutations at a given nucleotide were weighted according to their frequency (Supporting information, Table S4), increasing number of mutations on a particular position were deemed less informative and down-weighted. Transversion changes were given three times the weight of transitions as the latter events are over 15 times more likely to occur in mammal mitochondria (Šnábel, 2012).

| Population structure
Gene flow disruption, evidenced as population structure, was assessed using pairwise F-statistics, hierarchical differentiation of the populations, and spatial-genetic distance correlations. Pairwise differentiation between populations was tested for both molecular analogues, using F ST and Φ ST statistics (Weir & Cockerham, 1984).
Pairwise exact tests were performed with 10,000 steps in the Markov chain, 10,000 dememorization steps, and 10,000 randomi-  (Earl & vonHoldt, 2011) was used to parse and format the replicated analyses. The best K that fitted the data based on the Evanno method (Evanno, Regnaut, & Goudet, 2005) was also explored with this software and verified using a discriminant analysis of principal components (DAPC, Jombart, Devillard, & Balloux, 2010) in the Adegenet package in R. The pipeline CLUMPAK version 1.1 (Kopelman, Mayzel, Jakobsson, Rosenberg, & Mayrose, 2015) was implemented, using a LargeGreedy algorithm, to align the samples of each K repetition and to graphically visualize it.

| Indices of genetic diversity
From the total number of CYTB sequences (n = 308) included in the analysis, 87 unique haplotypes (h) were found (

| Demographic statistics
Neutrality statistics that use the mutation frequency spectrum (Class I) show a strong signature of population expansion in Ghana ( shown for northern regions of the country, contrasting with the positive values of the AC and NG colonies (5.54 and 1.02, respectively).

The pairwise distance statistics (Class III mismatch distributions)
show that AC and NG have ragged (rg > 0.03) bimodal distributions, which reflects constant population size over a long period of time, in contrast to all other colonies that have a smooth (rg < 0.03) unimodal-shaped distribution (reflecting rapid population growth). These results were also corroborated with coalescent extended Bayesian skyline analyses. Epomophorus gambianus show a rapid exponential population growth, displaying over an eightfold increase in the population in the past 30 years (Figure 3). Ghanaian populations show a constant size after a rapid increase (17-fold) in the last 200 years, except for the colony from AC (Supporting information, Figure S5A, B) where a constant population size over time could not be rejected (NG was not analyzed due to sample size). However, the weakness of this method to small sample size and genetic structuring is reflected by the incongruence between the y axes of the Ghanaian versus the entire E. gambianus population ESBPs (Grant, Liu, Gao, & Yanagimoto, 2012).
Thus, accurate historical population sizes cannot be determined. The  Figure S6).
The results from the models run in BOTTLENECK (Table 3)   Note. Bottleneck events were tested using the software BOTTLENECK, viewing two models of mutation and showing p values for Wilcoxon test; and M_P_Val: assessing the M ratio (k/r), the critical M (Mc) and a combination of θ and Δg input values (noted in italics) to assess the percentage in 10,000 simulations that produced equilibrium M ratio values that were significantly lower than those expected under mutation-drift equilibrium. Values that are statistically significant (p < 0.05) are shown in bold.

| Isolation by distance
When analyzing the correlation between the logarithmic geographic distance and the genetic distance, either as linearized Φ ST for the CYTB (Supporting information Figure S9A-E) or linearized F ST for the microsatellite markers (Supporting information Figure S9F-J), a positive significant correlation was detected when all the populations of TA B L E 5 Hierarchical AMOVA analysis and population structure using: mtDNA CYTB (Φ-statistics) and ncDNA microsatellites (F-statistics)

| Exploring population structure using Bayesian clustering methods-ncDNA
All models and datasets that were tested with STRUCTURE clus- selected by Delta K, that best explains the data for both models was K = 4 (Supporting information Figure S11).

| Exploring population structure using intraspecific phylogeny-mtDNA
The haplotype network of the CYTB fragment of E. gambianus

| D ISCUSS I ON
While several species of bat have been shown previously to be panmictic (Peel et al., 2013;Russell et al., 2005;Sinclair, Webb, Marchant, & Tidemann, 1996;, genetic structuring within a species can vary due to various dispersal or migratory behaviors (Fleming, Murray, & Carstens, 2010;Moussy et al., 2013). Several studies exemplify male, but not female, panmixia in bats (Kerth, Mayer, & Petit, 2002;Rossiter et al., 2002;Rydell, 1989); however, female-biased bat dispersal has been reported (Nagy et al., 2007). By comparing the fixation indices between uniparental and biparental markers, it is possible to estimate the extent of sex-biased dispersal (Prugnolle & de Meeus, 2002). In the current study, we show that across most of the range of E. gambianus, there is congruent phylogeographic structuring between both maternally and biparentally inherited markers, which suggests long-range dispersal of both sexes. Epomophorus gambianus connectivity and free mixing were demonstrated across the colonies sampled throughout most of its distribution. Besides the peripheral Ethiopia, Accra and, in some degree, Ve-Golokwati colonies, there was no evidence of an IBD pattern of dispersal between the colonies. There is, however, a signal of genetic differentiation in the female line of descent, which is not explained by the expected conflict between hemizygous mtDNA and multiallelic ncDNA markers (Birky, Maruyama, & Fuerst, 1983), indicating that male dispersal occurs more frequently and/or over longer distances than female dispersal. It was also possible to infer past population expansions and bottleneck events.
The Ethiopian colony was consistently identified as a divergent population. These bats were surrounded by the Ethiopian highlands Even though in some studies, mountains can be considered a weak obstacle for flying taxa (Demont, Blanckenhorn, Hosken, & Garner, 2008;Moussy et al., 2015;Petit & Mayer, 1999;Xu et al., 2010), this distance combined with numerous water bodies seems an effective geographical barrier for this lowland bat species. In addition, the phylogeny of E. gambianus revealed that the Ethiopian bats diverged over ~1.6 Mya to the rest; however, the age of our samples is likely overestimating this divergence. It has been hypothesized that an Asian ancestor of the myonycterine-epomophorine clade colonized Africa through the forested corridors that linked Asia and Africa (prior to the rise of the mountains), with consequent evolutionary radiation (Juste et al., 1999). Therefore, the colonization of E. gambianus could have followed an east-west axis. However, other dispersal routes or vicariance events, that may explain this phylogenetic pattern, cannot be ruled out. Furthermore, the series of ice ages that occurred during the Pleistocene and upper Pliocene (~3.5 Mya-12,000 years ago) were associated with processes of forest contraction and habitat fragmentation that shaped the vegetation and forest systems present in ancient Africa (Hamilton & Taylor, 1991;Hewitt, 2000).
During this era, the fragmented patches of refugia drove widespread speciation and divergence during three progressive climate shifts that increased arid conditions (deMenocal, 1995). The second one The Volta River and Lake Volta, one of the biggest water bodies in West Africa (8,500 km 2 of surface), are situated in the Volta basin, which lies to the north and east of Accra and Ve-Golokwati. Even though E. gambianus is likely to be able to fly across the Volta river, F I G U R E 5 Epomophorus gambianus systematics. (a) Median-joining haplotype network. The circles represent unique haplotypes of CYTB sequences. Circle size is proportional to the frequency of specimens sharing that haplotype, and the color reflects the population of origin. The lines between two haplotypes show base substitutions, and its length is proportional to the number of point mutations. There is a clear spatial clustering between the Ethiopian colony (in black) and the rest of the African populations. (b) Bayesian phylogeny of E. gambianus CYTB haplotype alignment. Out-group species are labeled in red: Rousettus aegyptiacus (Hap 94) and Epomops franqueti (Hap 53-58). There are two distinct clades in the E. gambianus phylogeny, one monophyletic group generated by the Ethiopian population (black box) and the rest (subdivisions highlighted with colored boxes). Haplotypes 2 and 14 are typed in blue. Private haplotypes to Nigeria ( †) and to Central African Republic ( ‡) are denoted. Posterior probabilities are shown above the main nodes, and estimated divergence time (Mya) between clades is shown within brackets genetic divergence due to rivers (smaller in width than the bats' average foraging flight distance) has been reported for other bat species, including the insectivorous Eptesicus serotinus (Moussy et al., 2015) and the frugivorous Scotonycteris bergmansi (Hassanin et al., 2015).
The demographic history of E. gambianus shows both spatial and demographic expansion in most of its colonies, which is consistent with a sudden population growth after reduced population sizes (Grant, 1998) and previously recognized expansion signatures of bats after colonizations from vicariance periods (Juste et al., 1999;Petit & Mayer, 1999). An exception is the colony of Accra, where the strong genetic signature reflects a past female germ line bottleneck.
Here  (Luikart et al., 1998). Using the BOTTLENECK software, the "standardized differences test" and the I.A.M. model were excluded due to their low statistical power and unsuitability for microsatellite data. The most powerful test to analyze less than 20 loci is the "Wilcoxon signed-rank test," which detected a significant heterozygosity deficiency in this dataset, suggesting population expansion instead of decline (Cornuet & Luikart, 1996). However, the detected high frequency of a few alleles could suggest an ancient bottleneck and the heterozygosity deficiency could reflect inbreeding (Wright, 1921), nonrandom sampling of family members  or false expansion signals often detected in IBD structures (Leblois, Estoup, & Streiff, 2006). The M ratio was also analyzed, as it performs better at detecting past bottleneck events than the BOTTLENECK algorithm (Peery et al., 2012;Piry et al., 1999). However, no ancient bottleneck signature was identified with the parameters selected and the results obtained with a range of input values were inconsistent, likely due to the known sensibility of this model to the parameter assumptions and IBD (Leblois et al., 2006). This conflict between markers could be due to the fourfold difference in effective population size from the haploid mitochondria, generating bottleneck signatures in the mtDNA from smaller reductions of the population size and/or an ancient brief bottleneck event, which would not be evidenced otherwise by the ncDNA (Birky, 1991;Wilson et al., 1985). Also, as the nuclear microsatellite markers have higher mutation rates, a bottleneck signal could have already been erased from them (Cornuet & Luikart, 1996;Rogers, 1995). Furthermore, selective sweeps (Maruyama & Birky, 1991), founder events (Ashley & Wills, 1987), stochastic lineage extinctions (Avise, Neigel, & Arnold, 1984), or lack of power for detecting recent bottlenecks (Peery et al., 2012) cannot be ruled out.
Likely the habitat disturbance due to the rapidly expanding urbanization of Accra and/or the still unknown effects that social dynamics with conspecifics have to roost integrity (Kunz & Fenton, 2003) could have driven a decline (via migration, hunting, etc.) and genetic bottleneck in the Accra population. Furthermore, there is a clear association between the Accra and Ve-Golokwati colonies, perhaps reflecting selective connectedness due to the isolation with the other colonies due to an extensive forest cover loss (Supporting information Figure   S14; Hansen et al., 2013). Nevertheless, a past founder event from Ve-Golokwati or the continued connectedness between the colonies cannot be disregarded. Fossil evidence dating 21,000-12,000 years ago shows that vegetation zones in lowland Ghana (below the Volta basin) were depressed at least for several hundred meters of altitude and the area presented major forest reduction at glacial maximum (Hamilton & Taylor, 1991). This forest loss and fragmentation possibly shaped the genetic signatures of both the Accra and Ve-Golokwati colonies. The extremely limited haplotype richness in Accra despite a strong central continent connectivity (Supporting information Figure   S15), however, suggests a more recent bottleneck or founder effect.
As the time of divergence was assessed with general clock rates of mammalian mtDNA used at the speciation level and the substitution rates at the species level differ greatly to that of intraspecific divergence (Ho, Saarma, Barnett, Haile, & Shapiro, 2008), it was not possible to accurately evaluate the time of recent demographic events, particularly of nearby colonies within Ghana.

| CON CLUS IONS
The results presented in this study confirm connectivity and free gene flow of E. gambianus across much of its range. Panmixia was demonstrated throughout the Central African Republic to the northern and central regions of Ghana. Between these colonies, there was no evidence of population divergence due to geographical isolation or preferential breeding, although males seem to disperse longer distances or more frequently. In contrast, the Ethiopian colony of E. gambianus is genetically divergent from the rest of its population. Complementary studies using homochromous samples would greatly benefit the evaluation of the extent of this divergence. The Ghanaian lowland colonies sampled also show genetic differentiation from the rest of the other sampled colonies, with a strong genetic signature of a past bottleneck in the female line of descent in the Greater Accra colony.
Both the geographical landscape and the species ecology suggest that the geographical barriers (mountains and water bodies) and other environmental developments (e.g., urbanization of megacities) evaluated in this research are likely drivers of the regional genetic divergences.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
Silke Riesle-Sbarbaro designed and performed the research, analyzed the data, and wrote the first draft of the manuscript. Kofi Amponsah-Mensah contributed in sampling methodology and ac- reviewed and edited the manuscript; and contributed reagents. All authors improved the manuscript.

DATA ACCE SS I B I LIT Y
The mitochondrial DNA alignments used in this study, the nuclear microsatellite RAW data (.fsa) and the outputs of the STRUCTURE analysis performed using 277 bats, are included in Dryad Digital Repository: https://doi.org/10.5061/dryad.m8m6142.
Mitochondrial DNA alignments will be also included in GenBank.
Primers for both mitochondrial DNA and nuclear microsatellite markers are included in the Supplemental Information.