Male‐biased dispersal and the potential impact of human‐induced habitat modifications on the Neotropical bat Trachops cirrhosus

Abstract Gene flow, maintained through natal dispersal and subsequent mating events, is one of the most important processes in both ecology and population genetics. Among mammalian populations, gene flow is strongly affected by a variety of factors, including the species’ ability to disperse, and the composition of the environment which can limit dispersal. Information on dispersal patterns is thus crucial both for conservation management and for understanding the social system of a species. We used 16 polymorphic nuclear microsatellite loci in addition to mitochondrial DNA sequences (1.61 kbp) to analyse the population structure and the sex‐specific pattern of natal dispersal in the frog‐eating fringe‐lipped bat, Trachops cirrhosus, in Central Panama. Our study revealed that—unlike most of the few other investigated Neotropical bats—gene flow in this species is mostly male‐mediated. Nevertheless, distinct genetic clusters occur in both sexes. In particular, the presence of genetic differentiation in the dataset only consisting of the dispersing sex (males) indicates that gene flow is impeded within our study area. Our data are in line with the Panama Canal in connection with the widening of the Río Chagres during the canal construction acting as a recent barrier to gene flow. The sensitivity of T. cirrhosus to human‐induced habitat modifications is further indicated by an extremely low capture success in highly fragmented areas. Taken together, our genetic and capture data provide evidence for this species to be classified as less mobile and thus vulnerable to habitat change, information that is important for conservation management.

be maintained by dispersal and subsequent mating events (Freeland, Kirk, & Petersen, 2011b). Generally, life history, behavioral, morphological, and habitat-associated traits all contribute to a species' dispersal ability (Bohonak, 1999;Bonte et al., 2012). Therefore, dispersal can be influenced by a variety of factors, including the availability of suitable mating partners and resources, and the occurrence of pathogens, parasites, and predators (Freeland et al., 2011b).
Moreover, in addition to a species' general ability and propensity to move, dispersal and thus gene flow can also be impeded by the composition of the environment (Freeland et al., 2011b). In a landscape where physical barriers to dispersal occur, for example, rivers or mountain ranges, restricted connectivity of habitats may lead to population genetic differentiation (e.g., Manel, Schwartz, Luikart, & Taberlet, 2003). Another major factor that impedes dispersal and contributes to isolation and subdivision of populations is human-induced habitat fragmentation (e.g., Rudel, Defries, Asner, & Laurance, 2009). Forest fragments often are embedded in a matrix of inhospitable habitat types causing isolation and subdivision of animal populations (e.g., Watling & Donnelly, 2006). Consequently, genetic drift in combination with reduced gene flow in fragmented landscapes may result in an increase of genetic differentiation and a loss of genetic diversity (e.g., Fahrig, 2003).
Even though bats are generally considered to be able to cross open areas due to their high mobility, different species-specific reactions to habitat fragmentation have been reported (e.g., Avila-Cabadilla et al., 2012;Ferreira et al., 2017;Kerth & Melber, 2009).
Depending on the matrix the remaining habitat patches are located in, bats can be more tolerant to habitat modifications compared to other animals due to their capacity to fly and their ability to exploit resources that are patchy in time and space (e.g., Bernard & Fenton, 2007;O'Donnell, Richter, Dool, Monks, & Kerth, 2016). However, in fragmented areas with an unfavorable matrix, bats can be sensitive to the modified habitats or forest edges (e.g., Meyer, Kalko, & Kerth, 2009;Ripperger, Tschapka, Kalko, Rodriguez-Herrera, & Mayer, 2013). For many forest-dwelling bat species, open water seems to be one of the least favorable types of matrix, as it provides no protection from potential predators and offers limited resources (e.g., Albrecht, Meyer, & Kalko, 2007).
Among mammalian populations, gene flow is often strongly affected by sex-biased dispersal (Perrin & Mazalov, 2000). In particular, many species of colonially breeding animals exhibit sex-biased behaviors (Greenwood, 1980). While in mammals, stronger philopatry to their natal area is typically shown by females, whereas males often disperse when reaching maturity, the opposite pattern has been described for birds (Dobson, 1982;Greenwood, 1980). The main evolutionary forces suggested to shape sex-specific dispersal patterns include kin cooperation and the avoidance of inbreeding, local mate competition, and local resource competition (see Lawson Handley & Perrin, 2007 for a review).
As bats are small, highly mobile, and nocturnal, their dispersal can be challenging to monitor using radio telemetry and capturemark-recapture methods (e.g., Petit & Mayer, 1999). However, population genetics can shed light into mating and dispersal behavior of bats. Providing a complementary approach to traditional field techniques, genetic approaches allow us to estimate the degree of population structuring and, therefore, provide cost efficient, relatively noninvasive methods for surveying the spatial structure of mammalian populations (e.g., Frantz, Do Linh San, Pope, & Burke, 2010). As the degree of genetic differentiation between and within subpopulations is affected by dispersal, philopatry, and the mating system, understanding population structure can provide insights into the social organization of a species (e.g., Burland & Wilmer, 2001). Additionally, different markers can be used to assess different aspects of the population genetic composition of a population.
Differences in mutation rate between nuclear microsatellite loci and mitochondrial DNA (mtDNA) allow for a different resolution in terms of time scale. Whereas microsatellites provide excellent resolution to understand contemporary gene flow (e.g., Angers & Bernatchez, 1998), the much lower mutation rates of coding regions of mtDNA reflect rather historical signals (Avise et al., 1987). Moreover, as the variability of bi-parentally and uni-parentally inherited loci may be affected differently, for example, by the presence of sex-biased dispersal, it is often informative to use both mitochondrial and nuclear molecular markers (e.g., Castella, Ruedi, & Excoffier, 2001;Kerth, Mayer, & Petit, 2002).
The fringe-lipped bat, Trachops cirrhosus (Phyllostomidae), is a Neotropical animalivorous species that occurs from southern Mexico to southern Brazil (Cramer, Willig, & Jones, 2001). While it is widespread in lowland forest, this species is rather rare in agricultural areas and at higher elevations (Cramer et al., 2001). It roosts in caves, hollow trees, road culverts, and buildings in groups of up to 50 individuals (Hall & Dalquest, 1963), where both sexes can be encountered roosting together (Nowak, 1999). Trachops cirrhosus hunts frogs and various insects (Cramer et al., 2001;Tuttle & Ryan, 1981) and its relatively small foraging grounds (3-12 ha) are typically located 200 m to 1.6 km from its roost (Jones, Hamsch, Page, Kalko, & O'Mara, 2017;Kalko, Friemel, Handley, & Schnitzler, 1999). Although numerous studies have been conducted on the predatory preferences and the foraging behavior of T. cirrhosus (reviewed in Page & Jones, 2016), very little is known about dispersal and population dynamics in this species. Generally, the mating system of most leafnosed bats (Phyllostomidae) is not known and is likewise unknown for T. cirrhosus (McCracken & Wilkinson, 2000). Furthermore, information is scarce about the effects of environmental disturbances on the population genetic structure of insectivorous or carnivorous bats in the Neotropics in general (Cunto & Bernard, 2012;Fenton et al., 1992).
While little more than 100 years ago the only potential barrier to gene flow for T. cirrhosus in our study area in Central Panama might have been the Río Chagres (Figure 1), the construction of the Panama Canal has caused additional large-scale fragmentation to its habitat. Most lowland forest has been flooded through the damming of the Río Chagres between 1910 and 1914, which caused former hilltops to become isolated islands surrounded by a matrix of water (Albrecht et al., 2007). Over 200 such islands covered with semideciduous, lowland tropical moist forest exists within the Panama Canal varying in size and degree of isolation, the largest being Barro Colorado Island (BCI) with 1,560 ha (Leigh, 1999). In 1999, Kalko et al. predicted that habitat alterations, particularly fragmentation and isolation of forested areas, would negatively affect populations of T. cirrhosus. This study indicated that the relatively sedentary foraging behavior of T. cirrhosus, reflected in its wing morphology and its use of small foraging areas, makes this bat species vulnerable to habitat changes. Therefore, the aim of our study was to use population genetic tools to assess patterns of gene flow between distinct populations of T. cirrhosus in this highly fragmented landscape in Central Panama and to determine whether gene flow is driven by sex-specific natal dispersal in this species. Based on the available information of the species' biology, we hypothesized that populations of T. cirrhosus within our study area would be genetically differentiated and expected to find evidence for the impact of habitat fragmentation on the genetic structure of populations of this species. However, for the natal dispersal pattern of T. cirrhosus we could not make a clear prediction: even though female philopatry has been hypothesized to be the strategy most beneficial for mammals, bats in the Neotropics have been shown to display various patterns of natal dispersal (e.g., Dechmann, Kalko, & Kerth, 2007;Wilkinson, 1985).

| Studyareaandsamplecollection
Bats were captured over an eleven-year period, from 2005 to 2016, using mist nets (Avinet, Dryden NY, USA and Ecotone, Gdynia, Poland), set in the forest, along streams, near small ponds, and at known roosts. Wing tissue samples (4 mm ø wing punch) were collected and stored in >95% ethanol until DNA extraction.
Bats of both sexes were captured and juveniles were identified by the presence of epiphyseal gaps in the phalanges (Brunet-Rossini F I G U R E 1 The study area and sampling sites of Trachops cirrhosus in Central Panama. The filled triangular markings represent the different sampling sites Barro Colorado Island (BCI); Peña Blanca (PB), Bohio (BO), Gigante (GI), Culebra Cut (CUL), and Gamboa (GA). As GA and CUL both consist of various netting sites, those are indicated by open circles for GA and open triangles for CUL. The stars represent all islands were sampling efforts were undertaken (I) and the diamonds represent those netting sites summarized as A. The gray line roughly represents the former route of Río Chagres before the construction of the Panama Canal (Shepherd, 1911) & Wilkinson, 2009). All sampling protocols followed guidelines approved by the American Society of Mammalogists for capture, handling, and care of mammals (Sikes, 2016) and were conducted in accordance with the standards of the Smithsonian Tropical Research Institute (STRI) Institutional Animal Care and Use Committee (IACUC; 07113001, 04113002, 2007-14-06-15-07, 20100816-1012-16, 2014-0101-2017, 2013-0401-2016, 2016-0627-2019   The following abbreviations are used: the observed fragment length range (Size range), in base pair (bp), the multiplex in which each marker was included (Multiplex) and which fluorescent label was used (Label) and both primer sequences (F, forward; R, reverse). All sequences have been deposited in the GenBank under the accession numbers provided.
Central Panama (Figure 1). Some of these sampling sites pool many individual netting sites (see Figure 1 for details). Peña Blanca has been removed from the analyses of nuclear DNA (nucDNA) due to insufficient sample size; however, we nevertheless used it for the analyses of mtDNA. The maximum distance between our sampling sites was 27.3 km, between PB and CUL, and the minimum distance amounted to 2.2 km between the closest respective netting sites within GA and CUL.
Sampling efforts were also undertaken on various islands within the Panama Canal and in an area where forest patches are embedded in a matrix of agriculturally managed land close to the town El Giral (Figure 1) between 2013 and 2016 using an adapted standardized mist-netting approach (Meyer & Kalko, 2008a). Despite the fact that the mist netting effort during this time was comparable to that of the other sampling sites, only one individual was encountered in each of these two highly fragmented areas, respectively. This insufficient sample size led to the exclusion of these two individuals from population genetic analyses.
Each multiplex reaction contained 1× Qiagen Multiplex Master Mix and between 0.02 μmol/L and 1.10 μmol/L of each primer (Table S1).  Cycler (Applied Biosystems) using a touchdown program with an initial 15-min denaturation at 95°C, followed by 2 cycles at 95°C for 30 s, annealing at 60°C for 30 s and extension at 72°C for 2 min.
Every two cycles, the annealing temperature was reduced by 2°C (60-52°C) and then kept at 50°C for the remaining 30 cycles. Final extension occurred at 72°C for 5 min. Amplified products were puri-

| Sex-biaseddispersal
First, sex-biased dispersal was investigated using the assignment test implemented in the program fstat v.2.9.3 (Goudet, Perrin, & Waser, 2002). The method assumes a species with nonoverlapping generations where dispersal occurs at the juvenile stage. As further postdispersal sampling is assumed, this test was conducted only on the nucDNA datasets of adult bats assumed to be postdispersal (N F = 106, N M = 196) using the sampling site as substructure.
Expectations of this test are that the dispersing sex should show (1) greater variance in assignment (vAI C ), (2) weaker source-populationassignment (mAI C ), (3) higher within-group diversity (H S ), (4) a deficiency in heterozygotes due to samples representing a mixture of genetic populations (resulting in higher F IS -values, showing signs of a Wahlund effect) and (5) lower diversity measures among groups (F ST ) (Goudet et al., 2002). We used all the sets of tests mentioned above and conducted a one-sided test, thereby either setting females or males to be the philopatric sex, respectively, with 10,000 permutations each.
To confirm the result achieved with all five measures by the test mentioned above (that males are the dispersing sex in T. cirrhosus, see Results), we conducted an additional run using the same tool in fstat v.2.9.3 (Goudet et al., 2002) to act as a control. This run was performed using males of different age groups, thereby treating juvenile (before their dispersal period) male bats as residents and adult male bats as the dispersers (N ADULT = 196;  (Chaverri & Kunz, 2006). However, in contrast to a bimodal reproductive phenology in frugivores, reproduction in animalivorous bats has been described to be unimodal (Durant, Hall, Cisneros, Hyland, & Willig, 2013). Thus, it can be expected that dispersing juveniles would leave their natal group before their mother gives birth to the next offspring (within the first year of their lives). Adult males, on the other hand, should have dispersed before being sampled and should therefore be categorized as dispersers. Again, a one-sided test with 10,000 permutations was operated assuming juveniles to be residents (e.g., "philopatric").
Further assessment of dispersal patterns was conducted by performing spatial autocorrelation analyses, that is, analyses of genetic relatedness between pairs of individuals as a function of the natural logarithm of geographical distance using SPAGedi 1.2 (Hardy & Vekemans, 2002) for two datasets separately. The first dataset consisted of adult females (data F ) and the second one was comprised of all adult males (data M ). The degree of spatial genetic structuring can be measured by the slope of the relationship mentioned above (Hardy & Vekemans, 2002). To obtain a multi-allelic, multi-locus mean measure of spatial genetic structure per given distance, the kinship coefficient F ij presented in Loiselle, Sork, Nason, and Graham (1995) was estimated between all pairs of individuals. The kinship coefficient F ij (Loiselle et al., 1995)

| Assessmentofpopulationgenetic structuringonthebasisofnucDNA
Before conducting population genetic structure analyses using the program structure (Pritchard, Stephens, & Donnelly, 2000), closely  Table 2). The sampling site GI was excluded from structure analyses of data M due to limited sample size (only one adult male).
structure was run on each nuclear DNA dataset assuming admixture and correlated allele frequencies using the LOCPRIOR model that allows for the use of sample group information (here the sampling sites BCI, BO, GA, GI, and CUL) in the clustering process (Hubisz, Falush, Stephens, & Pritchard, 2009). The LOCPRIOR model has been shown to detect genetic structure at lower levels of divergence, or with less data, than previous structure models, but does not tend to find structure where none is present (Hubisz et al., 2009). Ten independent runs of K = 1-10 were conducted for each of the two datasets, respectively. All runs used 10 6 iterations after a burn-in period of 10 5 .
As uneven sampling can bias inferences on the number of clusters in the program structure (Puechmaille, 2016), efforts were made to have comparable number of individuals from each sampling site after the removal of closely related individuals. Therefore, subsampling was carried out by randomly choosing 30 different individuals from GA (10 subsampled datasets were analyzed). Moreover, we followed the procedure from Puechmaille (2016)  Moreover, the larger a threshold value for the mean membership coefficient, the larger the differentiation between two subpopulations needs to be for them to be considered to belong to different genetic clusters. Therefore, setting a threshold value too high might potentially underestimate the real number of clusters. Therefore, we analyzed the results of the program structure using the thresholds 0.6, 0.7, and 0.8, but mainly focused on the threshold of 0.6 for the interpretation of the results.
To assess the level of genetic diversity, the observed (H o ) and expected heterozygosity (H e ) for each locus as well as for each population inferred by structure for data F+JUV were calculated using Genetix 4.05.2 (Belkhir, Borsa, Chikhi, Raufaste, & Bonhomme, 1996-2004. The mean number of alleles (A) and the allelic richness (A R ), were calculated for each locus and each cluster using fstat v.2.9.3 (Table S1, supplementary material).

| Assessmentofpopulationgenetic structuringonthebasisofmtDNA
Regarding mtDNA, pairwise genetic distances were calculated between the 53 individual sequences using the Kimura 2-parameter model implemented in MeGa 6.0 (Kimura, 1980;Tamura, Stecher, Peterson, Filipski, & Kumar, 2013). Additionally, a median-joining (MJ) haplotype network was constructed using NETWORK v.5.0 (Bandelt, Forster, & Rohl, 1999) to graphically illustrate the relationships among the two different haplotypes found. Additionally, due to a lack of available sequences that span the complete fragment of the mitochondrial region that we investigated which also included noncoding and

| Assessmentofcapturerates
Additionally, capture rates were calculated for a subset of the individuals included in the genetic datasets mentioned above, as only those bats captured in accordance with the standardized mistnetting approach (Meyer & Kalko, 2008a) were suitable for the calculation of capture rates (N = 29). This sampling took place between 2013 and 2016 within a total of 5380 mist netting hours (mnh).
Capture rates were calculated for BCI, BO, GI, PB, all islands within the Panama Canal combined (I) and the forest fragments within the agriculturally managed land close to El Giral (A). Among others, the one individual sampled in A was not captured following the standardized mist-netting approach and is therefore not represented here.
We calculated the capture rate as a standardized measure of relative abundance by dividing the number of bats (recaptures included) captured at each sampling site by the number of mnh employed (1 mnh represents one 6 m mist net open for 1 hr). Relative abundance is then provided as number of bats per 100 mnh.

| Sex-biaseddispersal
For the sex-biased dispersal test between adult females and males assuming female philopatry, all five tests conducted in fstat were significant (p < .05) and clearly indicated male-biased dispersal ( Fij (mean ± SE) Pairwise spatial distances (m) When regressing geographic distance and relatedness between individuals belonging to respective datasets, weak but significant regression patterns of isolation-by-distance were detected (slope b F ± SE = −0.0047 ± 0.00097, p = .0005; slope b M ± SE = −0.0038 ± 0.00043, p < .0001; Figure 2). The increase in genetic differentiation among individuals with geographical distance is more pronounced in data F as indicated by a steeper slope compared with data M . However, this difference in slope was not statistically significant (p = .274).

| Assessmentofpopulationgeneticstructuring onthebasisofnucDNA
Using the 16 autosomal microsatellite loci, the Bayesian clustering method inferred three distinct genetic clusters for data F+JUV Analysing data F+JUV using the program fstat v.2.9.3.2 (Goudet, 1995), we detected significant pairwise genetic differentiation between the three genetic clusters identified by structure

| Assessmentofpopulationgeneticstructuring onthebasisofmtDNA
Regarding mtDNA, the 53 sequences were successfully aligned resulting in a 1,605 bp long fragment. Only two different haplotypes were found within the study area with a K2P genetic distance of 2.83% (GenBank accession numbers: MH102398, MH102399).
Fifty individuals represented haplotype 1, and three individuals, all sampled in CUL, displayed haplotype 2. Throughout the complete fragment of mtDNA analyzed, the two haplotypes differed from each other at 46 sites. Within the cytochrome b alone, the two discovered haplotypes still differed at 23 sites ( Figure 4).
Moreover, setting them in context with sequences derived from GenBank that originate from French Guyana, Venezuela, and Brazil shows that even though these two sequences are still most similar to each other, they fit well into context with samples from other geographical regions within Central/South America (Figure 4).
The time of divergence between the two mitochondrial haplotypes found in the course of this study was estimated to have occurred 1.3 million years ago. Highest posterior density for the time to the most recent common ancestor was estimated to be between 766,000 and 1,870,400 years ago.

| Assessmentofcapturerates
According to the capture rates obtained for individuals sampled during the standardized mist-netting approach, the relative abundance of T. cirrhosus was much lower in the two highly fragmented

| D ISCUSS I ON
Several lines of evidence clearly indicate that T. cirrhosus shows a pattern of male-biased dispersal. In terms of gene flow, male-biased dispersal has various consequences and can thus be identified by comparing the sex-specific population genetic patterns of a species (Freeland, Kirk, & Petersen, 2011a;Prugnolle & de Meeus, 2002).
First, it has been hypothesized that the dispersing sex should show lower levels of among-population differentiation than the philopatric sex (e.g., Mossman & Waser, 1999). Indeed, our results demonstrate that males show lower levels of population differentiation. This is indicated by only two genetic clusters in the male dataset compared to the three distinct genetic clusters identified in data F+JUV , and by lower F ST values between genetic clusters in data M (0.0247), compared to data F+JUV (0.0392-0.0440). Secondly, spatial autocorrelation, or the decrease in genetic relatedness among individuals with geographical distance (e.g., Frantz et al., 2010), in T. cirrhosus is more pronounced in data F as indicated by a steeper slope compared with data M . Even though these slopes of isolation-by-distance do not differ significantly between the sexes, a trend is visible. Third, assignment tests can be used to compare the number of individuals that are genetically assigned to a population different to the one where they were sampled in. Consistent with the assumption that the dispersing sex should show significantly more miss-assignments (Goudet et al., 2002), male T. cirrhosus showed both a significantly greater variance in assignment and a weaker source-population-assignment compared to females which clearly indicates male-biased dispersal.
Finally, in the case of male-biased dispersal, maternally inherited mtDNA markers should show higher levels of population differentiation compared to biparentally inherited markers. Unfortunately, within our study area, we only detected two mitochondrial haplotypes. Although these haplotypes did differ strongly from each other (genetic distance = 0.0283), one of the haplotypes was only found in three of six individuals from CUL and all other individuals showed the same mitochondrial haplotype. Therefore, no conclusions can be drawn here in terms of sex-specific dispersal.
The presence of only one dominant mitochondrial haplotype within the study area is quite surprising for tropical species (compare Ditchfield, 2000). In particular, as our mtDNA sequences comprise coding regions, such as the full cytochrome b, as well as noncoding parts of the D-loop region, which is well known for its high mutation rate (e.g., Wilkinson, Mayer, Kerth, & Petri, 1997). The fact that the full cytochrome b sections of our sequences did not harbor any stop codon suggests amplification of the true mtDNA rather than nuclear copies, known as Numts (Triant & Dewoody, 2007). This remarkably low level of mitochondrial diversity in T. cirrhosus could be related to the history of the species in the last million years, as the divergence of the two strongly differentiated mitochondrial haplotypes (approximately 3% differentiation) was estimated to have occurred in the early Pleistocene (between 766,000 and 1,870,400 years ago).
However, due to a lack of a more complete geographic sampling across the species range, we can only speculate about the origin of these distinct mtDNA haplotypes.
Critically, if our study had only focused on the analysis of mtDNA, the lack of spatial differentiation may have led to the tentative conclusion that females are the dispersing sex. Therefore, our study emphasizes that it is not always sufficiently conclusive to compare the results of uni-parentally and biparentally inherited markers to make inferences about sex-specific dispersal, but additionally to search for signs of sex-specific dispersal among the biparentally inherited genotypes (Goudet et al., 2002).
Male-biased dispersal is the typical pattern for temperate zone bat species (see Moussy et al., 2013 for a review). In Neotropical bats, however, various patterns of natal dispersal have been described with male-biased dispersal being comparatively rare (e.g., Dechmann et al., 2007;Nagy, Gunther, Knornschild, & Mayer, 2013;Wilkinson, 1985). Generally, philopatry is assumed to be the optimal strategy for female mammals due to the various benefits accrued by the females staying in or in close proximity to the natal area and/or social group (Greenwood, 1980). These benefits include familiarity with local resources, and improved fecundity and breeding success when associating with kin (e.g., Clutton-Brock & Lukas, 2012). Despite this general assumption, Nagy et al. (2013) hypothesized that a large number of Neotropical bat species have dispersal patterns that differ from those of the majority of mammals. This hypothesis was based on the fact that female breeding behavior in Neotropical bats (especially in some phyllostomid bats) has been shown to commence very early in life (e.g., in Dermanura watsoni sexual maturity of females was reached in as little as 3 months; Chaverri & Kunz, 2006).
Female dispersal allows females to avoid father-daughter inbreeding if the age of females at first conception falls below the tenure of males (Dechmann et al., 2007).
Female T. cirrhosus have been observed to be lactating within less than a year after being captured as a juvenile (V. Flores, unpublished data). We can therefore assume that this species reaches sexual maturity as early as other phyllostomids (Chaverri & Kunz, 2006).
Consequently, one would expect female dispersal in T. cirrhosus as well. However, this conclusion might be immature because Nagy and co-workers discovered a male-biased dispersal pattern within Balantiopteryx plicata (Nagy, Knornschild, Gunther, & Mayer, 2014), a bat species where the age of females at first conception fell below the tenure of males. They argue that the reason for this was that father-daughter inbreeding was circumvented by mating outside of the nursing roost.
In phyllostomid bats, natal dispersal has thus far only been investigated in four species. In Uroderma bilobatum habitat characteristics, specifically, the characteristics of their roosts, seem to determine whether females disperse or show philopatry (Sagot, Phillips, Baker, & Stevens, 2016). Both Lophostoma silvicolum (Dechmann et al., 2007) and Phyllostomus hastatus (McCracken & Bradbury, 1981) show all-offspring dispersal, which seems to have developed due to a mixture of inbreeding avoidance and avoidance of local mate competition and local resource competition. In Central Panama, female T. cirrhosus might profit from knowing their local habitat because this species' relatively small foraging grounds (Jones et al., 2017;Kalko et al., 1999) mainly consist of streams and ponds, which are patchily distributed and thus harder to find. Therefore, females might profit from staying in or at least close to their natal habitat. Note, that with our method, we cannot discriminate between strict philopatry where females stay in the natal area/group and philopatry that results from females moving much shorter distances than males.
However, unpublished capture-mark-recapture data suggest that at least some females remain in their natal colony.
In Desmodus rotundus, dispersal is male-biased (Streicker et al., 2016;Wilkinson, 1985). As cooperation both between related and unrelated females is frequently observed in this species (Carter & Wilkinson, 2013), (kin)cooperation seems to be the driving force for the observed female philopatry and male-biased dispersal. While In addition to the potential introduction of barriers to dispersal, human-induced habitat fragmentation may also critically limit the distribution of T. cirrhosus as a function of reduced population viability in smaller habitat fragments. This is supported by the observation that T. cirrhosus was not caught in the highly fragmented, agriculturally dominated landscape around El Giral (Figure 1) on the smaller islands within the Panama Canal (results consistent with former studies: Meyer & Kalko, 2008b;Meyer et al., 2009).
The capture rates determined in the course of this study clearly show that the relative abundance of T. cirrhosus was much lower in A (0.00) and I (0.04) compared to the other sites (BCI = 1.20; BO = 1.92; GI = 2.78; PB = 1.39). This further emphasizes the species' sensitivity and vulnerability to fragmentation (as predicted by Kalko et al., 1999).
Mobility has been suggested to be a good predictor of a species' vulnerability to fragmentation and altered population genetic structure. In bats, the ability and proclivity for long distance flight and dispersal is mostly reflected in the wing morphology of a species (Meyer et al., 2009). Trachops cirrhosus has short, broad wings which allow it to be highly maneuverable in obstacle-rich environments (Norberg & Rayner, 1987). This low wing loading is efficient for gleaning prey in the forest understory, but means that T. cirrhosus is a slow flyer. Both GPS (S. Greif, unpublished data) and telemetry tracking studies (Jones et al., 2017;Kalko et al., 1999) show that T. cirrhosus moves very little within the landscape, mostly stays within the forest understory, and tends to avoid open areas. In accordance with these findings, our own observations and genetic results confirm the classification of T. cirrhosus as a less mobile species (Meyer et al., 2009).

| CON CLUS IONS
In accordance with our hypothesis, our results indicate that habitat fragmentation seems to influence both the population genetic structure and the distribution of T. cirrhosus. Moreover, we were able to demonstrate that T. cirrhosus shows male-biased dispersal.
Further research is needed to investigate the driving force for the strong pattern of male-biased dispersal observed in T. cirrhosus.
Finally, the effect of the Panama Canal as a potential barrier to gene flow should be further investigated in T. cirrhosus as well as in other bat species within the same habitat. In this context, the phyllostomid family would be particularly interesting as species of this family are relatively closely related, but show high levels of ecological diversity through various differences in morphological, physiological and behavioral traits (Baker, Jones, & Carter, 1976;Datzmann, von Helversen, & Mayer, 2010). Therefore, the same fragmented habitat might appear to hinder movement for one species, thus impeding genetic connectivity, while for another closely related species with a different ecology, it may allow uninhibited movements resulting in no genetic structuring.
This is the first study to provide population genetic data for T. cirrhosus. Knowledge on sex-specific patterns of natal dispersal is important for understanding the social system of this otherwise well-studied species and can be used as a basis for future social and behavioral investigations. Moreover, the results from our study may be used to inform effective conservation measures in regions suffering from a high impact of human-induced habitat fragmentation and to gain additional knowledge on the effects of environmental disturbance on bats.

ACK N OWLED G M ENTS
This work was funded by the Deutsche Forschungsgemeinschaft (KE 746/7-1, 7-2; TS 81/7-1, 7-2) within the priority program "Ecology and species barriers in emerging viral diseases (SPP 1596)." We thank the Smithsonian Tropical Research Institute for logistical support, the Autoridad del Canal de Panamá for permission to work on the islands in Gatún Lake and all the members of the Gamboa and BCI bat laboratories for their work in the field. For general support during the laboratory work, we thank Ina Römer. We are further grateful to Pierre-Loup Jan and Lisa Lehnen for helping us with applying the R script from Frantz et al.
(2010) to our dataset. Additionally, we are thankful to Serena Dool for assisting with the construction of the mtDNA network.
Moreover, we thank Nicola Fischer, Lisa Lehnen, and Jaap van Schaik, and two anonymous referees for providing comments on the manuscript.

CO N FLI C TO FI NTE R E S T
The authors declare that they have no conflict of interest.

AUTH O RCO NTR I B UTI O N S
TH and GK designed the study with input from RP and SJP. RP, SB, MT, VF, and TH provided data. TH performed the laboratory work and the data analyses and drafted the manuscript with input from RP, SJP, and GK. SJP contributed further to analyses. SB conducted the analyses on the capture rates at the different sampling sites. All authors have commented on the manuscript and have read and approved the final version of the manuscript.