Genetic structure across urban and agricultural landscapes reveals evidence of resource specialization and philopatry in the Eastern carpenter bee, Xylocopa virginica L

Abstract

Human activity continues to impact global ecosystems, often by altering the habitat suitability, persistence, and movement of native species. It is thus critical to examine the population genetic structure of key ecosystemservice providers across human-altered landscapes to provide insight into the forces that limit wildlife persistence and movement across multiple spatial scales. While some studies have documented declines of bee pollinators as a result of human-mediated habitat alteration, others suggest that some bee species may benefit from altered land use due to increased food or nesting resource availability; however, detailed population and dispersal studies have been lacking. We investigated the population genetic structure of the Eastern carpenter bee, Xylocopa virginica, across 14 sites spanning more than 450 km, including dense urban areas and intensive agricultural habitat. X. virginica is a large bee which constructs nests in natural and human-associated wooden substrates, and is hypothesized to disperse broadly across urbanizing areas. Using 10 microsatellite loci, we detected significant genetic isolation by geographic distance and significant isolation by land use, where urban and cultivated landscapes were most conducive to gene flow. This is one of the first population genetic analyses to provide evidence of enhanced insect dispersal in human-altered areas as compared to semi-natural landscapes. We found moderate levels of regional-scale population structure across the study system (G'_ST = 0.146) and substantial co-ancestry between the sampling regions, where co-ancestry patterns align with major human transportation corridors, suggesting that human-mediated movement may be influencing regional dispersal processes. Additionally, we found a signature of strong site-level philopatry where our analyses revealed significant levels of high genetic relatedness at very fine scales (<1 km), surprising given X. virginica's large body size. These results provide unique genetic evidence that insects can simultaneously exhibit substantial regional dispersal as well as high local nesting fidelity in landscapes dominated by human activity.

Keywords: human‐altered landscapes; human‐mediated dispersal; isolation by resistance; landscape genetics; nest‐site fidelity; pollinators.

Figure 1

STRUCTURE results for Xylocopa virginica at all 14 collection sites. A subsample of random individuals were selected from transect locations and analyzed together as a single site for sites BC, CL, MA, SM, and WQ, each having a maximum of n = 30 individuals included in the analysis. (a) Population membership as computed by STRUCTURE for K = 1, K = 2, K = 3, and K = 4 from 10 microsatellite loci and visualized using CLUMPAK. Sites are arranged left to right by increasing latitude. Each vertical bar represents the probability that an individual genotype can be assigned to a particular population, indicated by different colors. Blue represents membership to genetic population 1, orange to genetic population 2, purple to genetic population 3, and green to genetic population 4. K = 4 was the most likely number of genetic populations (Evanno et al., 2005). (b) Map of collection sites displaying pie charts of mean population membership for K = 4 genetic populations. Segments are color‐coded by population membership as in the STRUCTURE plots (a)

Figure 2

(a) Isolation by distance (IBD), and (b, c) isolation by resistance (IBR) for two resistance maps (Hypothesis A and Hypothesis C, cultivated; Table 1) among 34 sampling locations. Each point represents the average pairwise genetic distance (Bruvo et al., 2004) between sample sites. ** indicates significance at the α < 0.01 level for the MRDM models

Figure 3

Spatial autocorrelation of genetic relatedness (Loiselle's F_ij) versus geographic distance. The solid black line represents the mean of all pairwise relatedness coefficients between individuals at different distance intervals. Dashed lines represent permuted 95% confidence intervals (CI) for the null hypothesis that there was no correlation between relatedness and distance (F_ij = 0). Upper and lower CI values at distance 0 are small but do not equal zero. Individuals at distances lower than 0.75 show significantly higher genetic relatedness than expected by random mating. * indicates significance at α < 0.05, and *** indicates significance at α < 0.001