Samples and DNA extraction
We obtained blood (n = 63), tissue (n = 273), and hair (n = 4) samples from 340 coyotes collected from southeastern Ontario during 1974–1984 (n = 120) and 2005–2010 (n = 220) (Fig. 1). The coyote samples were predominantly submitted by hunters and trappers or obtained from road kills. We generally restricted our sampling to south of 45° latitude, except for four samples collected east of 75° longitude and three collected west of 81° longitude (Fig. 1), to reduce the potential of sampling eastern wolves in the vicinity of Algonquin Provincial Park. We obtained tissue (n = 16) and blood (n = 59) samples of 75 domestic dogs collected during 2006–2011, which were obtained predominantly from a local veterinarian in the Peterborough region of Ontario (Fig. 1) and comprised mostly mixed-breed individuals. Breeds represented in the sampled dogs included Beagle, Border Collie, German Shepherd, Golden Retriever, Poodle, Husky, Labrador Retriever, Great Dane, Akita, Mastiff, Rottweiler, Malamute, Rhodesian Ridgeback, Spaniel, Terrier, and Hounds.
Figure 1. Approximate locations of historical (1974–1984) and contemporary (2005–2010) coyote samples collected from southeastern Ontario, Canada; some samples may be overlapping. The black square in the inset map depicts the location of the study area in North America. Dog samples were collected in the Peterborough region.
Download figure to PowerPoint
We extracted DNA from samples using a DNeasy Blood and Tissue Kit (Qiagen, Mississauga, ON, Canada), and determined gender by amplification of Zfx/Sry primer pairs (P1-5EZ and P2-3EZ: Aasen and Medrano 1990; Y53-3C and Y53-3D: Fain and LeMay 1995) or Zfx/Zfy primers (LGL-331 and LGL-335: Shaw et al. 2003).
Autosomal and Y-chromosome microsatellite genotyping
For each sample we amplified 12 autosomal microsatellite loci in three multiplex reactions with published primers (cxx225, cxx2, cxx123, cxx377, cxx250, cxx204, cxx172, cxx109, cxx253, cxx442, cxx410, and cxx147: Ostrander et al. 1993, 1995) as in Wheeldon et al. (2010b). For male samples we amplified four Y-chromosome microsatellite loci with published primers (MS34A, MS34B, MS41A, and MS41B: Sundqvist et al. 2001), as in Wheeldon et al. (2010b). We purified amplified products through ethanol precipitation prior to genotyping on a MegaBACE 1000 (GE Healthcare, Baie d'Urfé, QC, Canada) or an AB3730 (Applied Biosystems, Burlington, ON, Canada). We accounted for allele shifts between instruments with multiple control samples and scored alleles in Genemarker (v1.7, Softgenetics LLC, State College, PA).
Mitochondrial DNA control region and Zfy intron amplification and sequencing
We amplified a 343–347 base pair (bp) fragment of the control region of the mtDNA with published primers (AB13279: Pilgrim et al. 1998; AB13280: Wilson et al. 2000) as in Wheeldon et al. (2010b). For some samples, we amplified the same fragment of the mtDNA control region with different published primers (ThrL, DL-Hcan: Leonard et al. 2002) under similar conditions to those in Wheeldon et al. (2010b). For a subsample of males we amplified a 658 bp fragment of the last intron of the Zfy gene with published primers (LGL-331: Shaw et al. 2003; Yint-2-335: Wilson et al. 2012) as in Wilson et al. (2012). We purified polymerase chain reaction (PCR) products using Exosap-IT (USB Corporation, Cleveland, OH), or Exonuclease I and Antarctic Phosphatase (New England BioLabs Inc., Ipswich, MA), prior to sequencing on a MegaBACE 1000 (GE Healthcare) or an AB3730 (Applied Biosystems). We edited and aligned sequences in Bioedit (v7.0.9, Hall 1999) or MEGA (v5, Tamura et al. 2011). Many of the coyote samples analyzed herein were previously analyzed at the mtDNA control region by Wheeldon et al. (2010a).
Genetic data analysis
We obtained 411 autosomal microsatellite genotypes based on 12 loci for the coyote and dog samples; four coyote samples (n = 3 historical; n = 1 contemporary) that amplified at less than six loci were excluded from subsequent analyses. Furthermore, we obtained genotypes of genetically assigned eastern wolves from Algonquin Provincial Park (n = 62: Rutledge et al. 2010) and gray wolves from northeastern Ontario (n = 62: Holloway 2009; Rutledge et al. 2010; Wheeldon and Patterson 2012) for inclusion in autosomal data analyses to account for possible admixture from these groups in the sampled coyotes and to identify potential noncoyote migrants sampled in southeastern Ontario. We acknowledge the mixed ancestry of these wolf groups, but we refer to them as eastern wolves and gray wolves for simplicity.
We analyzed the individual autosomal microsatellite genotypes in the Bayesian-clustering program Structure (v2.3, Pritchard et al. 2000; Hubisz et al. 2009) to assess admixture between coyotes and dogs. For all Structure analyses we inferred the parameter alpha and implemented the F-model (assumes correlated allele frequencies) and I-model (assumes independent allele frequencies) separately to compare results. We analyzed the historical and contemporary coyotes separately to avoid clustering problems associated with disparate sample sizes among groups. Based on prior findings (Rutledge et al. 2010; Benson et al. 2012) and accounting for the inclusion of dogs, we anticipated that K = 4 would be optimal given the data. To confirm this for the data set that included contemporary coyotes, eastern wolves, gray wolves, and dogs, we ran the admixture model of Structure for K = 1–7 with five repetitions of 106 iterations following a burn-in period of 105 iterations for each K. We calculated the mean posterior probability (ln P[D]) for each K by averaging across the five runs and confirmed that K = 4 was optimal for the data set based on quantitative criteria (Fig. S1; ln P[D], Pritchard et al. 2000; ∆K, Evanno et al. 2005) and consideration of the biological significance of clusters. Additionally, we performed a factorial correspondence analysis on the entire data set of individual autosomal microsatellite genotypes (n = 535) as implemented in GENETIX (Belkhir et al. 2004) and observed clustering patterns generally concordant with the results from Structure (Fig. S2). Notably, there was complete overlap of the historical and contemporary coyote samples (Fig. S2), indicating that K-determination in Structure was justifiably unnecessary for the data set that included historical coyotes. Subsequently, for both data sets, we ran the admixture model of Structure 10 times at K = 4 for 106 iterations following a burn-in period of 105 iterations and collected information on the 90% probability intervals of individual assignments (ANCESTDIST = 1). We obtained individual admixture proportions (Q-values) from the run with the highest posterior probability and lowest variance; we considered individuals to be admixed, if Q < 0.8 (e.g., Vähä and Primmer 2006). We compared individual assignments between the F-model and I-model and observed general concordance (Fig. S3). Specifically, we observed five cases of an individual being assigned as admixed under the F-model but not admixed under the I-model model, and five cases of the reverse scenario (Fig. S3). We suggest neither model was optimal for our data set considering the variable evolutionary relationships between Canis species, therefore we averaged Q-values between the F-model and I-model; hereafter references to individual assignments are based on the averaged Q-values, unless stated otherwise.
We generated 223–228 bp mtDNA sequences and distinguished those evolved in gray wolves and dogs (i.e., Old World origin) from those evolved in coyotes and eastern wolves (i.e., New World origin) based on a diagnostic indel (Pilgrim et al. 1998; Wilson et al. 2000).
We generated 400 bp Zfy intron sequences and observed three previously described sequences based on two variable sites: Yint-1 of coyote origin; Yint-2 of gray wolf (or dog) origin; and Yint-4 of eastern wolf origin (Wilson et al. 2012). We did not observe Yint-3 of coyote origin (Wilson et al. 2012) in our sample, which probably reflects a founder effect whereby eastward colonizing coyotes carried the Yint-1 sequence.
We generated Y-haplotypes based on the Y-chromosome microsatellite genotypes. We constructed a median-joining network (Bandelt et al. 1999) based on combined Y-chromosome microsatellite and Zfy intron data using the program Network (v184.108.40.206; available at www.fluxus-engineering.com/sharenet.htm) with default settings (ε = 0). The Y-chromosome microsatellite loci were weighted equally and the Zfy intron sequence variation was weighted twice as high as microsatellite loci.
We compared the Y-haplotypes that we observed in coyotes and dogs in this study with those of previously analyzed wolves, coyotes, and dogs to investigate Y-haplotype sharing among these species for the purpose of inferring introgression. We defined western and eastern coyotes as occurring west and east of the Mississippi River, respectively, and Great Lakes coyotes as occurring in the western Great Lakes states and western Ontario (Wheeldon et al. 2010b). We defined Great Lakes wolves as occurring in Manitoba, northern Ontario, southern Quebec, and the western Great Lakes states (Wheeldon 2009). Allele sizes for this study were calibrated with those in Rutledge et al. (2010), Wheeldon et al. (2010b), Wheeldon and Patterson (2012), and Wilson et al. (2012), the last of which calibrated allele sizes with those in Hailer and Leonard (2008). Furthermore, allele sizes were standardized among Musiani et al. (2007), Hailer and Leonard (2008), Sundqvist et al. (2001, 2006) (allele sizes obtained from the authors), and Fain et al. (2010).1 Thus, allele sizes were standardized for comparisons across studies. We generated unique Y-haplotype identifiers (Yhn) for each Y-chromosome microsatellite genotype because existing identifiers were not the same for matching Y-haplotypes across studies (Table S1). We performed a reduced-median analysis (r = 2; Bandelt et al. 1995) of the Y-haplotypes (loci weighted equally; frequency >1) and then constructed a median-joining network (ε = 10) with maximum parsimony postprocessing (Polzin and Daneshmand 2003).