Urban landscapes increase dispersal, gene flow, and pathogen transmission potential in banded mongoose (Mungos mungo) in northern Botswana

Abstract Disease transmission can be strongly influenced by the manner in which conspecifics are connected across a landscape and the effects of land use upon these dynamics. In northern Botswana, the territorial and group‐living banded mongoose (Mungos mungo) lives across urban and natural landscapes and is infected with a novel Mycobacterium tuberculosis complex pathogen, M. mungi. Using microsatellite markers amplified from DNA derived from banded mongoose fecal and tissue samples (n = 168), we evaluated population genetic structure, individual dispersal, and gene flow for 12 troops. Genetic structure was detectable and moderately strong across groups (F ST = 0.086), with K = 7 being the best‐supported number of genetic clusters. Indications of admixture in certain troops suggest formation of new groups through recent fusion events. Differentiation was higher for troops inhabiting natural areas (F ST = 0.102) than for troops in urban landscapes (F ST = 0.081). While this suggests increased levels of gene flow between urban‐dwelling troops, the inclusion of a smaller number of study troops from natural land types may have influenced these findings. Of those individuals confirmed infected with M. mungi, the majority (73%, n = 11) were assigned to their natal group which is consistent with previous observations linking lower levels of dispersal with infection. Twenty‐one probable dispersing individuals were identified, with all suspected migrants originating from troops within the urban landscape. Findings suggest that urbanized landscapes may increase gene flow and dispersal behavior with a concomitant increase in the risk of pathogen spread. As urban landscapes expand, there is an increasing need to understand how land use and pathogen infection may change wildlife behavior and disease transmission potential.


| INTRODUC TI ON
Our ability to predict infectious disease dynamics remains limited, particularly in free-ranging wildlife populations. Pathogen transmission processes can be complex, variable, and strongly shaped by attributes of interactions among hosts, pathogens, and their environments. Across host species, sociality can have an important influence on pathogen transmission dynamics, determining the degree of connectivity within and between groups, which in turn can influence the potential for pathogen spread (Cremer et al., 2007;Galvani, 2003;Loehle, 1995;Reiner et al., 2013;Sanderson et al., 2014;Sattenspiel & Simon, 1988). A central feature shaping these connections is the manner in which individuals or groups of individuals disperse across a landscape, moving from their natal area to establish themselves in another area or habitat patch. Here, landscape type and structure can interact with species behavior, shaping these movements and connections within and between groups, modifying disease transmission potential, and epidemic dynamics. The influence of land type can be significant, as for example in human-modified landscapes where the dispersal potential of a species may be either hindered under certain circumstances (Forman et al., 2003;Hamer & McDonnell, 2008) or increased in others (McKinney, 2006). Natural landscape features also may limit animal movement and consequently transmission and spread of pathogens. For example, spread of rabies in raccoon (Procyon lotor) populations was inhibited by large river courses that blocked raccoon movements (Côté et al., 2012, Hirsch et al., 2013Rioux Paquette et al., 2014). Spatial and temporal landscape variation may significantly influence a species movement behavior across their range, limiting the accuracy of regional dispersal estimates and model-based predictions of pathogen spread (Bowler & Benton, 2005). The complexity of these host-pathogenlandscape interactions continues to challenge our ability to computationally characterize dispersal in predictive infectious disease models.
The banded mongoose, Mungos mungo (Figure 1), is a small, fossorial carnivore that lives in social groups that can number from 5 to 75 individuals . This territorial species has a predominantly egalitarian social system, with low reproductive skew (Cant, 2000). In northern Botswana, banded mongooses are infected with a novel tuberculosis pathogen Mycobacterium mungi, a member of the M. tuberculosis complex . This tuberculosis (TB) pathogen is transmitted primarily through infected scent marks and associated contact that arises from olfactory communication behaviors. This population lives across a mixed land use area including both natural ecosystems and urbanized areas, with movement behavior of mongooses varying according to land type and proximity to humans . Evidence from previous studies (Fairbanks, 2013) suggests that mongoose dispersal behavior is increased in this region, differing significantly from that of populations living in a protected area in Uganda, where mongoose dispersal is extremely limited (Gusset, 2007). Occurrence of TB disease in the Botswana systems appears also to affect dispersal behaviors, with clinically ill mongooses dispersing less frequently than healthy mongooses (Fairbanks et al., 2014). While more data are needed, there was also evidence to suggest that healthy individuals residing in troops with more infected mongooses may be more likely to disperse than individuals in troops with lower infection levels (Fairbanks et al., 2014). Bidirectional interactions between dispersal and disease may have critical implications for infectious disease dynamics, information central to understanding and predicting epidemic dynamics.
Given the apparent complexity of mongoose movement behaviors in this system, we used microsatellite DNA markers as a tool to characterize troop and local-scale population structure and to infer patterns of dispersal and genetically effective migration among troops. We predicted a priori that (1) human-dominated landscapes would increase mongoose dispersal above that observed in natural landscapes, and (2) the occurrence of infectious disease would have a negative association with dispersal behaviors in infected mongooses, with sick mongooses being assigned to their putative natal troop. Our aim was to use this model system to evaluate the influence of landscape and infection on dispersal dynamics and our understanding of pathogen transmission and persistence dynamics in transforming landscapes.

| Study area and species
Our long-term study site is located in the northern part of Botswana in Chobe District (Figure 2

| Sample collection
Mongoose troops were monitored weekly through the use of VHF radio collars that were fitted to one or two individuals in each study troop as previously described . Banded mongooses use latrine sites for defecation, providing a localized area to collect fecal samples from a majority of troop members in one event (Pesapane et al., 2013). The evening prior to a sampling event, a troop of interest was selected and tracked to their denning site using radio telemetry. Existing fecal boluses that could be detected were removed from the latrine site so that fresh samples could be collected the following morning with reduced risk of crosscontamination. To obtain samples that were fresh and with the least host DNA degradation, troops were observed as they emerged from the den shortly after sunrise. Mongooses that exited the den were counted and monitored to see where they moved and to observe signs of defecation (squatting, lifted tail, etc.). The latrine site was approached for sample collection at least 10 min after the last individual left to ensure that most members had the opportunity to defecate without disturbance. Even after leaving the communal latrine site, the troop of interest was followed in case individuals defecated in another location. Distinguishing fresh fecal samples from older samples was based on color, firmness, and moisture of the stool (Pesapane et al., 2013). Using a sterilized surgical blade, the outer surface of each fecal bolus was carefully removed, avoiding surfaces in contact with the soil, which can act as a PCR inhibitor (Pesapane et al., 2013). The fecal matter was transferred into a sterilized 1.5-ml microcentrifuge tube and placed in a portable cooler with ice packs.
Blood and tissue samples were also obtained from banded mongooses sampled during capture activities, opportunistic postmortems (individuals subject to hit by car, dog attack, or human persecution), or collected in association with other management activities. Each sample was assigned an animal identification number, and data were recorded regarding sex of the animal, age class (juvenile, sub-adult, adult), date of capture, and affiliated troop at the time of capture or carcass discovery. M. mungi infection status can only be determined postmortem, with infection status assigned based on the presence of clinical disease, postmortem TB lesions, and/or the detection of M. mungi DNA in tissues and/or secretions as previously described . Individuals genotyped by fecal samples were, therefore, not given an infection status. Similarly, while we can identify sick animals based on the above criteria, we cannot definitively identify animals as being free from infection. We, therefore, restricted our examination of dispersal and disease to those mongooses that had died and had their infection status specifically determined.
Fecal DNA was extracted using the PowerSoil DNA Isolation Kit (MoBio ® ). DNA samples were stored in a −20°C freezer. Extractions F I G U R E 2 Spatial distribution of sampled lartrine sites (colored circles) belonging to 12 troops of mongooses (at time of fecal sample collection) along the Chobe River in northern Botswana were done within 24-48 hr in an attempt to obtain the highest quality and yield of host DNA from each sample. For DNA originating from blood or tissue, Quick-DNA Plus Kits (Zymo ® ) were used for the extraction process. Both kits were used according to the manufacturer's specifications.

| Genotyping
Genotyping was performed at 20 microsatellite loci (Griffin et al., 2001Waldick et al., 2003; Table 1) derived from previous studies of banded mongooses and other members of Family Herpestidae. Individual samples were genotyped in three multiplexed polymerase chain reactions (PCR). To ensure strong amplification at each locus without allelic dropout or other molecular artifacts, amplification conditions for each multiplex were optimized by assessing the fluorescence of amplicons subjected to electrophoresis in an ethidium bromide-stained 3.5% agarose gel with a 100 base pair molecular weight ladder. Each reaction was carried out in a 10µl reaction consisting of 5 µl of Qiagen Multiplex PCR Kit buffer, 1 µl of a 10× primer mix, 1 µl of bovine serum albumin (BSA), 2 µl of water, and 1 µl of 5ng DNA template. The conditions for the PCR protocol were as follows: 1 cycle of initial activation for 15 min at 95°C, followed by 34 cycles of denaturation for 30 s at 95°C, annealing for 45 s at 57°C, and extension for 30 s at 72°C, and a final extension for 10 min at 72°C. Forward primers were fluorescently labeled with 6-FAM, NED, PET, or VIC dyes, ensuring that different primers with similar product lengths (within 10 base pairs) TA B L E 1 Sources, genetic variability metrics across all troops within the study area, and results of analysis of molecular variance (AMOVA) for polymorphic loci used for genetic analysis of banded mongoose troops in the Chobe district of northern Botswana: average observed heterozygosities (H O ) and expected heterozygosities (H E ), average number of alleles (A), and average allelic size range. Results of partitioning of genetic variation using AMOVA are displayed as % variation among troops, among individuals, and within individuals would exhibit different labeling. The PCR products were sent to the Cornell University Biotechnology Resource Center for amplification fragment size analysis using an ABI3730 Genetic Analyzer.
Given the small amounts of host DNA in fecal samples (Bellemain et al., 2005;Waits & Paetkau, 2005), a multiple-tube amplification approach (Navidi et al., 1992;Tab erlet et al., 1996;Watts et al., 2011) was implemented to ensure accurate and repeatable host genotyping from mongoose fecal samples. From results of an initial PCR, if an individual was scored as heterozygous at a given locus, it was recorded as heterozygous. Results for loci that were scored as homozygous were compared to results of a second PCR for the same individual; if the genotype was heterozygous in the second PCR, then the individual was recorded as heterozygous at that locus. Individuals for whom both PCRs indicated homozygosity for the same allele at the same locus were considered homozygous. If an individual was homozygous at one allele for a locus for the initial PCR and homozygous for a different allele at the same locus for the additional PCR, the individual was recorded as heterozygous for the observed alleles at that locus. We interpreted results showing three or more alleles at one locus across the two PCR amplifications as having been cross-contaminated with another sample, and those results were removed from subsequent analyses.

| Genetic analysis
Each individual's genotyping file was manually uploaded, and amplicon sizes were visualized and scored using GeneMarker ver 2.6.2 (Holland & Parson, 2011); peaks were scored automatically using the default software settings and a GS-500 size standard (Applied Biosystems).
Fluorescence peaks were manually scored on three separate occasions to ensure consistent allele calling. Since fecal samples were collected without knowing the individual donor, we screened for any duplicate samples using Microsatellite Toolkit (Park, 2001). If fewer than four alleles differentiated individuals, then they were considered duplicates.
Any duplicated multilocus genotypes were removed from the dataset.
MICROCHECKER (Van Oosterhout et al., 2004) was used to assess the possibility of genotyping errors attributed to null alleles, large allele dropout, and accidental scoring of stutter peaks and to estimate frequencies of any null alleles.
Exact tests of deviation from Hardy-Weinberg and linkage equilibria, and calculation of observed and expected heterozygosities were conducted using Arlequin ver. 3.5 (Excoffier & Lischer, 2010).
Microsatellite Toolkit (Park, 2001) and Arlequin provided allelic size ranges, and Arlequin calculated Garza and Williamson (2001) m indices for each locus and troop. Private alleles-those occurring in only one troop-were identified using GeneAlEx 6.5 (Peakall & Smouse, 2012). Estimates of fixation indices, genetic distances between populations, and analysis of molecular variance (AMOVA) were conducted using Arlequin. We tested for the effects of isolation-by- STRUCTURE (Pritchard et al., 2000) was used to assess population structuring across the study area by assigning multilocus genotypes of individual mongooses to given numbers of subpopulations using Bayesian clustering approaches. Posterior support for various numbers of clusters (K) from K = 1 to 15 was evaluated, with support for each K tested using 10 iterations. Each iteration was tested using 50,000 burn-ins and 500,000 Markov chain Monte Carlo (MCMC) repetitions. GeneClass2 (Piry et al., 2004)

| Study permissions
Methods for this study were approved by the Virginia Tech

| Genotyping and loci metrics
From the 167 individual fecal samples collected, 77% were successfully amplified at a minimum of 13 loci. DNA from 49 blood or tissue samples, each representing a unique individual with sex and ageclass data, was also genotyped and included in the dataset. After identification and removal of nine duplicate samples, multilocus data from 168 individuals across 12 troops were used in analyses.
Due to a lack of polymorphism or unreliable PCR amplification, data from five of the original 20 microsatellite loci (MmAAC5, Mm18-1, Mm7-5, Mon-67, and Mm19) were omitted from subsequent analyses, leaving data for 15 loci (Table 1) numbers of breeders within troops, mixing among troops, generational overlap), data from all loci were retained in the analysis. Using a Bonferroni-corrected criterion for significance, we could not detect linkage disequilibrium among loci. We estimate that we sampled 80% of the individuals across 12 banded mongoose troops as determined from troop counts, representing nearly 50% of the troops estimated to be in our study focal area (n = 25). Sampling intensity varied according to troop size, latrine location, and visibility of feces, as well as individual mongoose latrine behaviors. We were not able to monitor all troops in the study area given limitations on staff and resources. Gaps in sampled troops exist particularly in the national park between CGL and CCH, where radio collar loss is higher and collar deployment is more challenging given dense vegetation and dangerous wildlife species that limit vehicle access and prevent following animals on foot. We use these samples and associated data to estimate local population genetic structure and individual dispersal.
Genetic diversity metrics for the 12 banded mongoose troops that we screened for microsatellite DNA variation are presented in

| Population structure and differentiation
Using the Evanno et al. (2005) ad-hoc ∆K statistic, K = 2 was the bestsupported number of clusters; the Evanno et al. (2005) method, however, has a known tendency to support the choice of K = 2 clusters (Puechmaille, 2016), and the pattern of clustering and the resulting individual assignments offered no particular insights. At K = 4 (Figure 3

| Group relatedness
Results from analysis of relatedness (Table 4) showed considerably higher values within (mean = 0.127) than between troops F I G U R E 3 Genetic structure of banded mongoose troops in northern Botswana inferred using program STRUCTURE for K = 4, 7, or 12 multilocus genotypic clusters. Troops are arranged in sequential order based on geographic location from east to west. Each histogram bar shows the probability coefficients (q) for each individual reflecting individual assignment to seven inferred genetic clusters (K) using the LOCPRIOR model. Troop WA had merged with troop WP, and troop KUBU had joined with nearby troop KWA at the time of sampling. Asterisks above the bar diagram for K = 7 show inferred migrants (two asterisks) and offspring of migrants (one asterisk) (mean = −0.018, Table 4); these values, respectively, suggest some level of relatedness within troops and none between troops.
Although some troops such as CSL and SEF had intragroup relatedness values relatively close to zero, which indicates little relatedness, all within-troop values were considerably larger than pairwise comparisons among troops.

| Effective population sizes
Estimates of N e (Table 5)

| Dispersal, detection of first-generation migrants, and infection status
Individual assignment using STRUCTURE and detection of firstgeneration migrants using GeneClass2 led to the assignment of 148 individuals (88%) to the troop from which they were sampled. Five of the 12 troops in the study area included individuals that apparently originated from other troops (Table 6), with CSL contributing the most migrants (n = 8). There was no indication that troops with adjacent or overlapping home ranges exchanged more migrants than spatially distant troops. In fact, two troops on the eastern edge of the study area, SEF and KUBU-KWA (Figure 1), produced the highest numbers of detectable migrants, nine and five, respectively ( Table 6). The test for first-generation migrants identified 11 individuals likely to have emigrated away from their parental troops (Table 6). Each first-generation migrant identified using GeneClass2 corresponded with individuals identified from the STRUCTURE assignment test as having high q values for troops other than the one where they were sampled (Table 6, Figure 3). SEF had the largest number of first-generation immigrants with four, while KUBU-KWA had three. All the first-generation migrants detected in the study assigned back to a natal troop within the urban land type (100%, n = 21, p = 0.03).
Eleven of our genotyped mongooses were confirmed on postmortem examination as being infected with M. mungi. Results of GeneClass2 analyses assigned eight of these individuals with high probability to the troop from which they were sampled. However, three of these infected mongooses were identified as migrants (sampled in troop CCH) with a higher assignment probability to a neighboring troop, CSL.

| D ISCUSS I ON
Among factors contributing to the dynamics of disease transmission across a landscape, host movements can play a central role by connecting individuals across the landscape and affecting pathogen transmission potential and movement (Alexander, Carlson, et al., 2018;Altizer et al., 2015;Fèvre et al., 2006;Gaidet et al., 2010;Morse, 2001). Social behaviors of a species influence how individuals disperse through a given area, especially for species that are highly territorial (Craft et al., 2011;Cross et al., 2009

KUBU-KWA
habitat patch and social group. Landscape features also may affect the movement behaviors of a host. This is particularly true for wildlife species that inhabit urbanized environments where storage and disposal of human waste may affect how far individuals or groups of individuals will travel in search of resources and how they will congregate if the resource is abundant .
Data from this study suggest that land use may influence dispersal behaviors, population structure, and, ultimately, pathogen transmission potential among banded mongoose in Northern Botswana.

| Population structure
Considerable genetic structure was evident among sampled banded mongoose troops over a scale of 20 km along the Chobe River.
Among the 12 troops, highest support was provided for seven clusters of multilocus genotypes. Certain troops were apparent mixtures of individuals from two or more genetically distinct source groups.
Such troops could be the result of individuals or groups from different troops either voluntarily leaving or being forcibly evicted (Cant et al., 2001(Cant et al., , 2010  Anthropogenic environments may facilitate intertroop movement. Troops living in urbanized areas with abundant food resources (i.e., human garbage, with larger and more calorie-dense foods than insects) may be more receptive to immigrants than troops living in natural ecosystems where denning and natural food resources (restricted to insects and small vertebrates) are more limiting. The abundant denning and food resources that are often available in urban landscapes have been shown previously to facilitate elevated levels of congregation in this species and others that are able to adapt and thrive in anthropogenic habitats (Bateman & Fleming, 2012;Bradley & Altizer, 2007;DeStefano & DeGraaf, 2003;Hassell et al., 2017).
These findings are consistent with previous results from our study system demonstrating the importance of the urban landscape on species behavior and pathogen transmission potential. Mongoose troops living in these anthropogenic landscapes had smaller home ranges in the dry season and concentrated space use around buildings and human refuse . Urban areas also appeared to relax territorial behaviors, with den sharing occurring among troops living in these land areas (Nichols & Alexander, 2019).

| Dispersal
We found high levels of dispersal of individuals into established troops (12.5% of genotyped individuals), in contrast to the Uganda population where dispersal of this nature was infrequent to absent among study troops (Nichols et al., 2012). Overall, 21 apparent migrants or dispersing mongooses were detected among the 168 multilocus genotypes analyzed using GeneClass2 (Table 6). From this pool of individuals, 11 were identified as first-generation migrants, themselves having dispersed from their inferred natal troop. There was no indication, however, that troops with adjacent or overlapping home ranges exchanged more migrants than spatially distant troops; the correlation of pairwise genetic differentiation among given troops and geographic distance between them yielded an R 2 value of just 0.159.
Our results from GeneClass2 analysis assigned 73% (n = 11) of M. mungi-infected individuals to the troop from which they were sampled. This appears to be consistent with previous observational studies (Fairbanks et al., 2014) identifying that clinically infected banded mongooses tended not to disperse as frequently as putatively healthy individuals. In our dataset, there were three individuals originating from troop CCH that had a higher assignment probability to neighboring troop CSL. These mongooses could have been infected after dispersal or moved, while in the latent stages of infection, a period of time that is still uncertain for this pathogen.
Given TB latency and the low numbers of infected individuals evaluated in this study, our data do not provide a strong test of the effect of TB infection upon dispersal behavior.
Anthropogenic habitats also may influence gene flow by nonconventional modes of dispersal. Human-mediated movement of wildlife can have important impacts on gene flow (Banks et al., 2015;Capinha et al., 2015;Waterkeyn et al., 2010) influencing, in turn, the potential for pathogen movement. In our context, banded mongooses can be moved incidental to human activities, crossing distances that normal ecological conditions likely would not allow.
For example, genotyping and individual assignment test results obtained for an orphaned mongoose pup found by a Kazungula fisherman showed that it was most likely a member of MOGO troop near Kasane. The 8-km journey was an unrealistic geographic distance for pup movement outside of the natal area, and human-mediated transport appears likely.
With troops living in close proximity, some with overlapping home ranges Nichols & Alexander, 2019), a mean genetic differentiation metric F ST of 0.086 across all land uses was relatively high, especially given evidence of intertroop movements and the limited size of our study area. Movement between banded mongoose troops in Uganda was more limited, but mating between groups was reasonably common, with around 20% of pups being the product of extra-group mating. Gene flow between groups therefore does occur, but without the levels of migration observed in Botswana; while several of the groups in Uganda were in an area of human activity, with access to refuse and anthropogenic den sites, immigration into these groups was rare, so there may be other factors at play. In the future, it would be important to sample more troops within the protected area in Botswana to further advance our understanding of the influence of urbanization on behavior and troop dynamics in this species.
As noted above, the F ST between grouped "park" and "town" Previous studies of banded mongoose behavior (Gusset, 2007;Nichols et al., 2012) have shown that individuals are typically philopatric and often mate within their natal group. Findings from our troop relatedness analysis suggest that banded mongoose practice similar behaviors in our study site. Although detection of mixture and admixture ( Figure 2) and dispersal events (  , and in Kenya about two (Waser et al., 1995). Mean average annual adult survivorship is 0.67 in Kenya (Waser et al., 1995) and 0.78 to 0.86 in Uganda (Cant, 1998;. Among individuals surviving to one year in Uganda, males lived an average of 42 months and females 38 months (Cant et al., 2016).
Based on the results of parentage and relatedness analysis, Nichols et al. (2014) found that female banded mongoose in Uganda breed with close relatives, suggesting the cost of inbreeding avoidance may outweigh the benefits under certain circumstances.
However, where extra-group matings did occur, pups in the Uganda population were genetically more heterozygous, heavier, and had a greater likelihood of survival to independence , benefits that would appear to offset risks. In Botswana, extragroup matings have also been observed during agonistic troop encounters. Additionally, we have observed nocturnal excursions by lone mongoose and investigation of occupied den sites in other mongoose territories, the purpose of which is uncertain (Nichols & Alexander, 2018). These behaviors suggest that gene flow may also arise from other types of intra-troop contacts beyond dispersal events where breeding opportunities arise and are exploited.
Extra-group breeding behaviors are expected to be influenced by adaptive fitness advantages that would be predicted to shift in respect of troop history and heterozygosity. For example, in early establishment with higher levels of troop heterozygosity (e.g., troop fusion), extra-troop breeding would presumably be less adaptive and less common as compared to older troops where extended maintenance of the group as a semi-closed social unit would be predicted to reduce heterozygosity, increasing the fitness benefits of extragroup matings. In our study site, land use would appear to have an important influence on these dynamics. In this system, disease also has a potential to influence potential fitness advantages of extratroop breeding with agonistic interactions increasing the potential risk of injury and M. mungi disease transmission Flint et al., 2016).

| Urban landscapes, species behavior, and infectious disease transmission
Considerable attention has been directed at the manner in which landscape transformation and urbanization contribute to habitat loss, movement, and fragmentation, in particular, how these processes can decrease landscape connectivity and species movement (Bradley & Altizer, 2007;Liu et al., 2009;Lowry et al., 2013;Noël et al., 2007;Tremblay & St Clair, 2009). Our data suggest that urban landscapes may increase population connectivity for banded mongoose in northern Botswana, at least within this land type.
Heightened connectivity may, in turn, increase the potential for pathogen transmission in socially structured populations that are able to adapt to anthropogenic landscapes, an inference central to disease modeling efforts and intervention design. However, infection status itself may influence individual dispersal behavior, further complicating our ability to predict landscape-host-pathogen interactions and disease spread. As urban landscapes across the globe grow, there is increasing pressure to understand how these growing landscapes influence disease transmission and persistence, potentially escalating the risk of disease transmission in both human and animals.

ACK N OWLED G M ENTS
We are grateful to Dr. Mark Vandewalle for input on the sampling scheme and Dr. Claire Sanderson for help with optimizing PCR conditions. Special thanks are given to Dr. Lena Patiño Westermann for assistance with radio telemetry, sample collection, and DNA extraction. We would also like to thank the full-time staff and summer volunteers at CARACAL for their assistance in the field and laboratory in Botswana. Support for this project was provided by the National Science Foundation Ecology and Evolution of Infectious Disease program (grant #479367).

CO N FLI C T O F I NTE R E S T
The authors state that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in University Libraries, Virginia Tech at https://doi. org/10.7294/29KX-E267.