The field cricket hybrid zone in Pennsylvania is a mosaic of genetically and morphologically distinct populations. The distribution of both genetic and morphological types is highly heterogeneous and cannot be explained as a function of distance across the hybrid zone; there is no clear clinal pattern of variation for any trait, but rather a patchwork of populations. This is similar to the patterns seen in other regions of the hybrid zone in Virginia (Harrison and Arnold 1982) and Connecticut (Harrison 1986; Harrison and Rand 1989; Rand and Harrison 1989) (Table 5). Mosaic hybrid zones occur when the ecological settings and/or geography in the area of overlap are heterogeneous or complex, and species distributions are determined by environmental selection (e.g., Harrison 1986; Howard 1986; Harrison and Rand 1989; Bridle et al. 2001; Ross and Harrison 2002; Bierne et al. 2003; Vines et al. 2003; Ross et al. 2008). In Pennsylvania, we find an association between species distribution and natural habitat; G. pennsylvanicus occupies natural habitat along forest edges and natural clearings, whereas G. firmus occupies more disturbed areas in agricultural and suburban environments. Hybridization and introgression occur across patch boundaries; there is evidence of substantial admixture both in morphological characters and mtDNA, over a broad geographic area.
Broad scale distribution of G. firmus and G. pennsylvanicus: morphology and mtDNA
Our sampling of G. firmus and G. pennsylvanicus revealed the same four major mtDNA haplotype groups that were found by Willett et al. (1997) and Maroja et al. (2009a). Gryllus pennsylvanicus consists of two major clades (northern and southern) and G. firmus has a distinct southern clade, and a northern group that is distinguishable from the other three well-supported clades (Fig. S1). Given that we find all four mtDNA haplotypes at three collecting localities in central Pennsylvania (BB, AR, BK), this region appears to represent the geographic divide between northern and southern groups of each species (Fig. 3). The hybrid zone appears to extend further north than previously described (Harrison and Arnold 1982; Harrison 1986; Maroja et al. 2009a). We found crickets in Rhode Island and Massachusetts that have introgressed mtDNA and the Massachusetts locality contains both parental types. To the north, we found only pure G. pennsylvanicus populations. Thus, Massachusetts may represent the northern range limit of G. firmus.
The relationship among the mtDNA groups has not been resolved with sequence data from the mtDNA COI gene. Willett et al. (1997) found five equally parsimonious tree topologies for these groups and in all cases, either northern or southern G. pennsylvanicus clades were the basal group, with the southern G. pennsylvanicus clade having the greatest haplotype diversity. Additional genotyping of seven mtDNA SNPs identified two nucleotide positions in the ATPase6 and COIII genes that are shared between northern and southern G. firmus (Fig. 3A). This suggests that an ancestral cricket lineage split into two daughter lineages, one that became either northern or southern G. pennsylvanicus and a second that split into the other G. pennsylvanicus clade and G. firmus. Gryllus firmus has subsequently diverged into northern and southern groups.
Patches of natural habitat contribute to the mosaicism of the Pennsylvania hybrid zone
Throughout their ranges, both G. pennsylvanicus and G. firmus can be found in disturbed habitats along roadsides, in fields and pastures and around human settlement. Yet, in the Pennsylvania hybrid zone, we see an association between species distribution and natural habitat; Gryllus pennsylvanicus occupies natural habitat along forest edges and clearings, while G. firmus occupies disturbed habitat near human settlement and agriculture (Table 4, Fig. 8). In the Pennsylvania study area, there is more natural habitat in the Appalachian Mountains to the north, which can explain why we also see a correlation between the distribution of G. pennsylvanicus and higher latitudes, greater vegetation density, lower temperatures, and more rainfall. However, we also find G. pennsylvanicus crickets in patches of natural habitat further south along the Blue Ridge Mountains (e.g., AJ, AK, AN, BI) and near rivers, lakes, and parks in the large gap between the Blue Ridge Mountains and the Reading Prong of the Northern Highlands (e.g., C, D, H, L, Z, Y). Likewise, G. firmus occurs further west into the Appalachian Mountains than earlier surveys of the hybrid zone suggested (Harrison and Arnold 1982; Maroja et al. 2009a).
We found a high proportion of crickets with both G. firmus morphology and mtDNA haplotypes in the Great Appalachian Valley and the small valleys within the Appalachians. Gryllus firmus likely expanded north through these corridors and crossed the steep mountain ridges along roadways through natural water and wind gaps (CG, CH, CD, CE, BT). In some areas, it appears that human disturbance has facilitated the persistence of G. firmus in otherwise heavily forested, natural habitats (CC and AZ).
Given that both cricket species seem well adapted to disturbed areas, it is unlikely that either performance in or preference for disturbed habitat restricts the distribution of G. pennsylvanicus. It is more likely that G. firmus is either less well suited for the habitat characteristic of G. pennsylvanicus' range or that G. firmus is particularly well suited for disturbed habitat and is a better colonizer. Typically, field crickets have short hind-wings and are incapable of flight. Daily movements such as feeding, reproduction, and predator avoidance are accomplished by walking. But in both species, individuals can be found with long hind-wings and fully developed flight muscles, capable of long-distance dispersal (Alexander 1968). The development of long-winged morphs is triggered by environmental variables such as temperature, population density, and resource availability (Harrison 1980). However, G. firmus and G. pennsylvanicus have both intra- and interspecific variation in their proportions of long-winged individuals that can be explained in part by genetic differences (Harrison 1979). Gryllus pennsylvanicus has on average of 4% long-winged individuals (Alexander 1968; Harrison 1979), whereas the frequency for G. firmus has been reported to be as high as 10–30% in some populations (Veazey et al. 1976; Harrison 1979). Wing dimorphism is thought to be particularly prevalent in G. firmus because its natural habitat is often highly disturbed and ephemeral (e.g., sand dunes, beach grass, and under shoreline debris) and may necessitate frequent dispersal among habitat patches (Roff 1990). Indeed, approximately 82% of the long-wing crickets identified in this study were G. firmus, whereas only 3% were G. pennsylvanicus (the remainder had fuzzy morphological membership coefficients) (Fig. 7). Wing dimorphism averaged 5–10% in G. firmus populations, but ranged as high as 30–75% at some collecting localities. In all cases, long-winged crickets were found at localities with high population densities and in disturbed habitats (Table 1). This suggests that G. firmus is both capable of thriving in disturbed habitats and may have a greater propensity for long-distance dispersal.
The association of G. pennsylvanicus with natural habitat and G. firmus with disturbed habitat is likely to be a major factor contributing to the mosaic structure of the hybrid zone in Pennsylvania. However, the environmental variables associated with hybrid zone structure are very different between Pennsylvania and other regions of the hybrid zone (Table 5). In Connecticut, soils vary over very short distances from loam to sand, and there is a clear association between species distributions and soil type; G. firmus occur on sandy soils and G. pennsylvanicus on loam (Harrison 1986; Rand and Harrison 1989; Ross and Harrison 2002). This may reflect adaptive differences in ovipositor length (G. firmus has a relatively longer ovipositor) for egg placement in different soil types (Masaki 1979; Bradford et al. 1993). In contrast, Pennsylvania soils are predominantly clay (≥20% clay) and we saw no correlation between soil properties and species distributions (Table 3). In Virginia, there is also no association between species distribution and soil type. Instead, elevation and temperature appear to contribute to hybrid zone structure; G. pennsylvanicus occupies high elevation sites in the Appalachian Mountains, while G. firmus is primarily in the lowlands. This is likely driven by differences in development time. Gryllus firmus from Virginia develop more slowly than G. pennsylvanicus (both in the field and in the lab) resulting in offset adult emergence (Harrison 1985). There are likely climatic life cycle shifts in G. firmus; southern crickets develop quickly and have multiple generations per year, but in mid-latitudes (Virginia), development may slow to accommodate only one generation per year and at even higher latitudes (Connecticut), shorter growing seasons may again favor faster development rate (Fulton 1952; Alexander 1968; Walker 1980; Harrison 1985). In both Connecticut and Pennsylvania, crickets appear to emerge synchronously, and there is no evidence that temporal isolation contributes to hybrid zone structure.
Patterns of admixture suggest strong pre-zygotic barriers
In mosaic hybrid zones, the patchy distribution of parental types results in extensive contact throughout the zone. Hybridization and introgression occur across patch boundaries or in intermediate habitats. In the Pennsylvania hybrid zone, there is a patchy distribution of natural and disturbed habitat and a corresponding distribution of G. pennsylvanicus and G. firmus. There are numerous opportunities for contact in areas where there are transitions in patch type: along mountains slopes, intersecting roadways, and near encroaching human development. Despite these opportunities for hybridization, we found that the majority of collecting sites are predominantly composed of a single parental type and a few individuals with intermediate morphologies that may be admixed (most likely backcrosses) (Fig. 6). Indeed, many of the crickets from sites with intermediate G. firmus morphologies had G. pennsylvanicus mtDNA haplotypes, suggesting that morphology is a good indicator of admixture. Each of these individual populations has an L-shaped distribution of morphological cluster membership (Fig. S2), but the combination of these predominantly G. firmus and G. pennsylvanicus populations results in an overall bimodal distribution within the hybrid zone (Fig. 5B). A few collecting localities contained both parental types, and a number of sites appeared to be mixed (containing both parental types and morphologically intermediate individuals).
The topographic complexity of the region may also explain why the hybrid zone appears broader across the central Appalachian Mountains than early surveys of the hybrid zone suggested (Harrison and Arnold 1982; Willett et al. 1997; Maroja et al. 2009a). The sharp transitions between forested mountains and populated valleys increase the patchiness of natural habitat and could increase the extent of hybridization. In addition, increased human disturbance as a result of suburban expansion, agriculture, and resource extraction is likely expanding the area of contact by increasing suitable habitat for G. firmus. Contact in some of these areas may even be very recent. For instance, the occurrence of G. firmus along the Pennsylvania turnpike (CG, CH, CD) in relatively discrete locations suggests that G. firmus may have only begun occupying these high elevation sites in recent decades.
Although we found evidence of substantial admixture both in morphological characters and mtDNA over a broad geographic area, the two species remain distinct. Most admixed individuals are morphologically like one or the other parental type (Fig. 5B), and there are few intermediate individuals. Given that F1 hybrids are viable and fertile in the lab, this suggests that strong pre-zygotic barriers are operating in this portion of the hybrid zone, a pattern consistent with characterizations of other regions of the hybrid zone in Virginia (Harrison and Arnold 1982) and Connecticut (Harrison 1986; Harrison and Bogdanowicz 1997). Barriers involved in behavioral isolation (Harrison 1986; Harrison and Rand 1989; Maroja et al. 2009b) and post-mating barriers that prevent fertilization (Harrison 1983; Larson et al. 2012) appear to be consistent across these different regions of the hybrid zone. In contrast, the ecological barriers that likely contribute to the hybrid zone's mosaic structure appear to vary between geographic regions (Table 5). In Connecticut, crickets are associated with different soil types, whereas in Virginia, crickets occur at different elevations and are temporally isolated due to differences in development time. In Pennsylvania, we found that the extent of natural habitat best explains the distribution of the two cricket species. This variation can have important consequences for patterns of introgression among different regions of the hybrid zone; genes involved in ecological isolation may vary in the extent of introgression in different environmental contexts (Harrison 1990; Payseur 2010). Interpreting patterns of variable introgression requires a clear understanding of the environmental context of species interactions (Nolte et al. 2009; Teeter et al. 2010; Macholan et al. 2011; Janousek et al. 2012). Species boundaries have been described as semipermeable, and this permeability varies not only across different genomic regions but also among different geographic areas and ecological contexts (Rand and Harrison 1989). Characterizing multiple regions within a hybrid zone is therefore critical for understanding hybrid zone dynamics, and gaining insights into the nature of species boundaries.