First regional evaluation of nuclear genetic diversity and population structure in northeastern coyotes (Canis latrans)

Introduction

Historically restricted to the open deserts and plains of central and western North America, the coyote (Canis latrans) has colonized almost the whole continent in the last 100 years, with few exceptions (e.g. Long Island)^1,2. The eastward range expansion was likely facilitated by widespread deforestation associated with agricultural development in the early 20^th century and by the near extirpation of eastern wolves (Canis lupus lycaon or Canis lycaon) and red wolves (Canis rufus)^3–5. The range expansion into northeastern North America advanced as two primary colonization fronts: the northern route through Ontario, where coyotes hybridized with resident eastern wolves, and the southern route through Ohio, where wolves were eradicated prior to coyote expansion^2,5,6. Along with its geographic range expansion, the coyote also experienced a niche expansion by rapidly colonizing whole new biomes, including eastern temperate and boreal forests¹. This complex scenario of colonization provides an interesting opportunity to explore the swift formation of population genetic structure following a rapid expansion in geographic distribution and ecological niche.

Recent analyses of population structure in northeastern coyotes have described a general lack of genetic differentiation among sampling localities, except at the coarsest geographic scales. Way et al.⁷ examined genetic variation and structure in a sample of coyotes from eastern Massachusetts using mitochondrial DNA (mtDNA) and eight microsatellite loci. They found no genetic structure in coyotes within Massachusetts or even within the broader region of northeastern North America. Instead, northeastern coyotes seemed to constitute one uniform population slightly differentiated from western coyotes. In another analysis of genetic variation in northeastern coyotes, Kays et al.⁵ identified three coarse phylogeographic areas: Ohio, the northeast zone, and a contact zone in western Pennsylvania and New York where the colonization front from Ohio has spread into the northeastern population. Although Kays et al. surveyed genetic variation in a dense geographic sample of 687 coyotes, they only used one genetic marker, the hypervariable mtDNA control region. vonHoldt et al.⁸ conducted a genome-wide analysis of North American Canis species and detected population structure in C. latrans, but only at the broadest continental scale. Although vonHoldt et al. genotyped tens of thousands of loci, they only sampled 14 northeastern coyotes, making detection of finer levels of population structure in the region very improbable. Thus, all the previous studies of population structure in northeastern coyotes have adopted one of two strategies: sampling many individuals using one (mtDNA) or very few molecular markers, or sampling very few individuals using thousands of genomic markers. A regional analysis of genetic diversity and population structure in northeastern coyotes using many samples and many nuclear molecular markers is currently lacking.

Genetic structure is a ubiquitous property of natural, domesticated, and human populations. Population genetic structure plays considerable roles in evolution, as both the basis and the consequence of local adaptation, the splitting of one species into two if the environments are markedly different, and the adaptability of a species as a whole across its range (i.e., transformation rather than speciation)⁹. The detection of genetic structure largely depends on the type and number of molecular markers examined, their variability in the target population, the number of individuals sampled, and the spatial sampling scheme^10,11. Single nucleotide polymorphisms (SNPs) have become a popular and inexpensive tool in the field of molecular population genetics. SNPs have properties that make them a superior alternative to other widely used genetic markers, such as microsatellites and mtDNA sequences, in evaluating genetic diversity and population structure^12–14. Furthermore, the sampling scheme can greatly influence the location and composition of genetic clusters, especially in species that are continuously distributed across a landscape¹¹, as are northeastern coyotes. Thus the conclusion of no population structure from previous analyses based on limited individual or genomic sampling may be imprecise.

The objective of this study was to test the hypothesis that fine-scale population structure in northeastern coyotes exists, but remains undetected due to the small number of individuals or the low resolution of the genetic markers previously analyzed. I hypothesized that population structure would be detectable, at finer levels than in previous analyses, by using an array of 16 high-heterozygosity nuclear SNPs ascertained from the dog genome and a spatially dense sample of 385 coyotes. This is the first comprehensive regional survey of genetic diversity and population structure in northeastern coyotes that uses a dense geographic sampling scheme and several SNPs. This regional analysis reveals a cryptic population structure and a geographic pattern of nuclear genetic diversity that is discrepant with previous mtDNA- and microsatellite-based surveys. More generally, this study adds to the growing body of recent literature that demonstrates the utility of SNPs discovered in a model organism for evolutionary inference in wild relatives, as long as ascertainment bias is explicitly evaluated.

Materials and methods

Study area and sampling

The study area was located in northeastern North America. Coyotes were sampled from New York (N = 174), Pennsylvania (N = 103), Vermont (N = 34), Ohio (N = 30), New Jersey (N = 14), New Hampshire (N = 11), Connecticut (N = 8), Massachusetts (N = 5), southern Quebec (N = 4), and Rhode Island (N = 2) (Figure 1). All samples (Total N = 385) used in this study are archived and vouchered in the New York State Museum, Albany, NY, where they were kept at -80°C. Specimens (tissues or combination of skin, skull, skeleton) were obtained primarily through donations from licensed local hunters and trappers since 1999. Six samples came from previous scat surveys in New York^15,16 (Data File). No Institutional Animal Care and Use Committee (IACUC) review was required for this study because the DNA samples came from scat or animals killed for reasons other than research.

Figure 1. Study area and sampling localities of 385 coyotes from northeastern United States and southeastern Canada.

Symbol shapes represent phylogeographic zones as in Kays et al.⁵: circle, Ohio; square, contact zone; triangle, northeast zone.

Results

I interrogated 16 SNP loci in 378 coyotes for a total of 6,048 expected genotypes (Data File). Genotyping efficacy varied by source of DNA, with fecal samples amplifying less effectively than tissue samples (Mann-Whitney test: U = 1771, P = 0.017). That is, the fecal samples had, on average, more missing or ambiguous genotypes than the tissue samples.

The overall sample of 385 eastern coyotes had lower genetic diversity than expected (Table 2), even though observed heterozygosity generally exceeded expected heterozygosity in the ascertainment panel of 17 northeastern and Ohio coyotes (Table 1). Ohio coyotes were the most genetically diverse in the region, but the eastward decay in genetic diversity observed with mtDNA was not replicated with nuclear SNPs. Although coyotes from the northeast zone had the lowest levels of mitochondrial genetic diversity, these same individuals had a level of nuclear genetic diversity comparable to Ohio coyotes (Table 2). The most pronounced differences between observed and expected heterozygosity occurred in the contact zone and in the overall regional analysis (Table 2). In the overall sample of 385 eastern coyotes, five loci were in Hardy-Weinberg equilibrium. The number of loci in Hardy-Weinberg equilibrium increased when each phylogeographic zone was analyzed separately: 13 in Ohio, nine in the contact zone, and nine in the northeast zone (Table 2).

Dogs appeared to be the most genetically diverse when the diversity of the five different canid groups was estimated using all 61,435 SNPs from the canine microarray. The genome-wide ascertainment bias was towards dogs: the expected heterozygosity of dogs was almost twice that of western coyotes. But the ascertainment bias reversed when heterozygosity was measured using only the 16 selected SNPs: coyotes appeared to be the most genetically diverse, with northeastern coyotes having a very high expected heterozygosity, whereas dogs appeared the least genetically diverse (Table 3).

In the region-wide population structure analysis of 247 individuals with little missing data, the values of K with the strongest statistical support were K = 2 and K = 3 (Figure 2A, B). Pairwise F_ST among the three genetic clusters varied from 0.08 to 0.10 and all were significant in the AMOVA framework (P = 0.001). This indicates that there are three primary genetic subdivisions in the broad sampling area. The three groups did not correspond to the Ohio, contact, and northeast zones previously inferred by mtDNA (Figure 2C). There is some geographic structuring, but the three groups overlap extensively in space. Although the red cluster in Figure 2B includes most of the Ohio coyotes, it is more cosmopolitan, also including many coyotes from the contact and northeast zones. The green cluster is mostly restricted to the contact zone, but extends slightly into eastern New York and Vermont. The PCA corroborated the results from STRUCTURE (Figure 3). The first two PCA axes explained 24.2% of the total variance and clearly separated the three STRUCTURE-inferred clusters. Most admixed individuals with no clear membership in any STRUCTURE-inferred cluster also showed no clear association with any PCA cluster.

Figure 2. Regional population genetic structure of eastern coyotes.

A) Estimation of the number of genetic populations using two criteria: ΔK (solid line) and Ln P(D) (dashed line). The most probable number of populations is the value of K with the maximal value of ΔK and Ln P(D). The data strongly support K = 2 and K = 3. B) STRUCTURE bar plots of N = 247 coyotes subdivided into K = 2 and K = 3 genetic populations; each individual is represented by a vertical bar partitioned into two or three colored segments indicating that individual’s proportional membership in each of two or three genetic clusters. C) Spatial distribution of K = 3 genetic populations. Symbol colors represent the genetic cluster with > 80% assignment from panel B, or black if highly admixed (i.e., no assignment > 80%). Symbol shapes represent phylogeographic zones as in Figure 1. Symbol locations are slightly jittered to display each individual and to reduce clutter.

Figure 3. Principal components analysis of 16 autosomal SNPs genotyped in 247 eastern coyotes.

Symbol colors correspond to K = 3 STRUCTURE-inferred clusters with > 80% assignment, or black if highly admixed (i.e., no assignment > 80%), as in Figure 2C. Inset shows scree plot of eigenvalues.

Ample population genetic structure was detected in the contact zone; the values of K with the strongest statistical support were K = 3 and K = 5 (Figure 4A). However, the K = 5 structure seems biologically unrealistic, characterized by highly admixed individuals of the various “populations” (Figure 4B). Alternatively, the K = 3 structure identified for the contact zone was virtually identical to the regional K = 3 structure (Figure 4C). In contrast to the contact zone, no fine-scale genetic structure was detected in Ohio or in the northeast zone. In Ohio, the value of K with the highest explanatory power was K = 1 (Figure 5A); in the northeast zone, the value of Ln P(D) does not increase beyond K = 2 and the change in Ln P(D) between K = 1 and K = 2 is minimal, indicating weak support for genetic structure (Figure 5B).

Figure 4. Population genetic structure of eastern coyotes in the contact zone.

A) Estimation of the number of genetic populations using two criteria: ΔK (solid line) and Ln P(D) (dashed line). The data strongly support K = 3 and K = 5. B) STRUCTURE bar plots of N = 118 coyotes subdivided into K = 3 and K = 5 genetic clusters. C) Spatial distribution of K = 3 genetic clusters. Symbol colors represent the genetic cluster with > 80% assignment from panel B, or black if highly admixed (i.e., no assignment > 80%).

Figure 5.

Absence of fine-scale population genetic structure in A) N = 16 coyotes from Ohio, and B) N = 113 coyotes from the northeast zone. See Figure 1 for zone designations.

Discussion

This study documents the presence of fine-scaled population genetic structure in eastern coyotes. Specifically, the contact zone exhibits a strong signal of population structure, the same signal detected in the regional analysis. This pattern may reflect the recent merging of two colonization fronts and the highly heterogeneous landscape of New York and Pennsylvania. A separate, spatially-explicit assessment with an independent set of SNPs demonstrated that northeastern coyotes exhibit a strong signal of population structure in Central New York¹. Furthermore, the same investigation showed that population structure is partially explained by ecological factors, such as deer density and human land use¹. Indeed, eastern coyotes living in areas of high deer density are genetically more wolf-like than those living in areas of low deer density³¹.

In contrast to the contact zone, coyotes in Ohio appear to make up a single panmictic population, as do coyotes in the northeast zone. Given the high mtDNA and nuclear diversity observed in Ohio by Kays et al.⁵ and in this study, it is unlikely that the lack of genetic structure stems from a founder effect. The failure to uncover more than one genetic population in Ohio may be due instead to its landscape homogeneity or to the lack of resolution afforded by 16 individuals to detect finer levels of structure. On the other hand, the failure to uncover more than one genetic population in the northeast zone is surprising given the vast geographic area with its ecological heterogeneity and the large number of individuals sampled. Coyotes in the northeast zone are likely the descendants of a few founders and therefore do not exhibit a strong signal of population genetic structure.

The perception of population structure, even in highly vagile animals where it was least expected, has been refined by steadily improving molecular data and geographic sampling. Initially, using mtDNA restriction site polymorphisms and nuclear microsatellites, no evidence of population structure or isolation by distance was found in coyotes, even at the continental scale^32,33. Various behavioral and historical explanations have been invoked to explain these early genetic patterns. But a more likely explanation is that the patterns of weak differentiation were artifacts of sparse geographic sampling or poor resolution due to the use of few molecular markers. More recent studies employing advanced analyses of spatial and genetic data have revealed strong differentiation among parapatric populations of coyotes and wolves, even in the absence of physical barriers to movement^1,34–38. These investigations used multiple loci and dense geographic sampling to uncover cryptic genetic subdivisions. Strong genetic differentiation between adjacent populations of coastal and inland wolves in British Columbia was shown with mtDNA³⁹, demonstrating that fine-scale genetic differentiation can be detected with denser sampling alone, even using a single molecular mtDNA marker. Similar cryptic subdivisions have been discovered in several highly mobile groups, such as Lynx^40,41, ungulates^42–44, cetaceans^45,46, and hawks⁴⁷. In all these cases, genetic subdivisions appear to emerge from ecological factors and local foraging adaptations. Future studies should focus on the ecological mechanisms underlying the cryptic genetic structure in northeastern coyotes, especially because they have only inhabited the region for the last 30–80 years². Confirming ecological determinants of population structure in the absence of obvious physical dispersal barriers would provide an interesting example of rapid ecological differentiation.

There are some important similarities and discrepancies between mtDNA and autosomal SNP patterns. The data indicate that coyotes in Ohio are the most genetically diverse in the region when surveyed with nuclear SNPs, as with mtDNA⁵. However, the gradual eastward decay in genetic diversity observed with mtDNA is not replicated with nuclear SNPs. The marked reduction of heterozygosity in the contact zone and in the overall region is very likely caused by population structure, i.e., the Wahlund effect. This interpretation is supported by the increase in the number of loci in Hardy-Weinberg equilibrium at smaller geographic scales and by the congruent signal of population structure in the overall region and in the contact zone. In addition, the three primary populations detected in this study do not correspond to the three subdivisions inferred with mtDNA⁵. Together, these results suggest that studies based solely on mtDNA should be interpreted cautiously. For example, mtDNA sequence similarity suggested that a small population of Scandinavian wolves was founded by individuals released from Swedish zoos, but nuclear polymorphic markers falsified the release hypothesis and instead supported a hypothesis of natural immigration or expansion from an unknown relict wolf pack⁴⁸. Discrepancies between patterns observed with mtDNA and nuclear DNA may be caused by true organismal processes, such as sex-biased dispersal⁴⁹. However, there is no evidence for sex-specific dispersal behaviors in eastern coyotes, consistent with their monogamous breeding system^50,51. Alternatively, discrepancies may be caused by marker-specific phenomena such as effective population size, lineage sorting, mutation rate, and coalescent times⁵², or the violation of certain assumptions of mtDNA inheritance, such as recombination, paternal leakage, and heteroplasmy⁵³. Future studies should further evaluate these sources of discrepancies.

The present study underscores three related methodological issues that are of broad interest, especially as SNPs continue to be in vogue in population and wildlife genomics. First, these data confirm that SNPs discovered in a model organism are an appropriate tool to address various questions regarding the ecology and evolution of non-model relatives. The sequencing of the dog genome²⁰ quickly enabled SNP-based investigations into wild members of the family Canidae, including coyotes, wolves, jackals, and foxes^8,54–57. Second, this study highlights the importance of evaluating the ascertainment bias of markers employed in a survey of genetic variation, especially in multi-species comparisons. Many SNP-based studies are not addressing the issue of ascertainment bias¹⁴. In the present study, the genome-wide analysis of variation is dog-biased because SNPs were ascertained primarily from comparisons of boxer and poodle genomes after the completion of the dog genome project⁸. On the other hand, the high-heterozygosity SNP analysis is coyote-biased because the 16 SNPs were chosen from an ascertainment panel of northeastern coyotes. The bias in the latter case is not problematic because the ascertainment panel from which the loci were selected is representative of the whole population of northeastern coyotes. However, the diversity measures reported in this study should not be compared to similar measures from other populations, unless the comparison corrects for ascertainment bias⁵⁸. Lastly, this study emphasizes the necessity of selecting SNPs very carefully to match the research question of interest. Here, a set of high-heterozygosity SNPs was interrogated in order to examine geographic patterns of genetic diversity and population structure. But other research questions may require polymorphic markers with other properties. For example, in order to better understand the complex hybrid ancestry of the northeastern coyote, Monzón et al.³¹ used species-diagnostic SNPs to quantify the relative contributions of its parental populations. In addition, recent advances in the molecular genetics of phenotypic traits in dogs allow the use of SNPs linked to genes of known function to address long-standing questions about morphological, physiological, and behavioral adaptations in northeastern coyotes and other wild canids^1,59.

Data availability

figshare: Data of genetic diversity in northeastern coyotes. http://dx.doi.org/10.6084/m9.figshare.943483⁶⁰