Coyotes Demonstrate How Habitat Specialization by Individuals of a Generalist Species Can Diversify...

Coyotes Demonstrate How Habitat Specialization by Individuals of a GeneralistSpecies Can Diversify Populations in a Heterogeneous Ecoregion

Benjamin N. Sacks,*�� Danika L. Bannasch,� Bruno B. Chomel,� and Holly B. Ernest*�*Veterinary Genetics Laboratory, University of California, Davis; �Department of Population Health and Reproduction, School ofVeterinary Medicine, University of California, Davis; and �Department of Biological Sciences, California State University,Sacramento

The tendency for individuals to disperse into habitat similar to their natal habitat has been observed in a wide range ofspecies, although its population genetic consequences have received little study. Such behavior could lead to discretehabitat-specific population subdivisions even in the absence of physical dispersal barriers or habitat gaps. Previousstudies of coyotes have supported this hypothesis in a small region of California, but its evolutionary significanceultimately depends on the extent and magnitude of habitat-specific subdivision. Here, we investigated these questionsusing autosomal, Y chromosome, and mitochondrial markers and .2,000 coyotes from a broad region, including 2adjacent ecoregions with contrasting levels of habitat heterogeneity—the California Floristic Province (CFP)(heterogeneous landscape) and the Desert–Prairie ecoregion (DPE) (homogeneous landscape). Consistent withpredictions, we found a close correspondence between population genetic structure and habitat subdivisions throughoutthe CFP and virtual panmixia over the larger DPE. Conversely, although genetic diversity was similar in these 2ecoregions overall, it was lower within sites of the CFP, as would be the expected consequence of greater genetic driftwithin subregions. The magnitude of habitat-specific genetic subdivisions (i.e., genetic distance) in the CFP variedconsiderably, indicating complexity (e.g., asymmetric gene flow or extinction/recolonization), but, in general, was higherthan that due to geographic distance or recent human-related barriers. Because habitat-specific structure can enhancea species’ adaptive potential and resilience to changing environments, these findings suggest the CFP may constitute anevolutionarily important portion of the range for coyotes and sympatric species exhibiting habitat-specific populationstructure.

Introduction

Recent evolutionary diversification within taxa char-acterized by long-range movements presents challengesto traditional evolutionary models. In the absence of vicar-iant events, gene flow is expected to homogenize popula-tions, especially those with broad habitat affinities,countering local differences in drift or selection that mightotherwise drive populations apart. In lower mobility ani-mals, populations sometimes remain faithful to ancient geo-logical formations separated by contemporary landscapefeatures, such as small streams or mountain ranges, thatwould pose insignificant barriers to larger, more wide-rang-ing species (Calsbeek et al. 2003). Geographic distancealone also can provide a sufficient barrier for divergenceamong low-mobility populations, for example, evidencedby ring species (e.g., Irwin et al. 2001), but in the absenceof discrete barriers is rarely sufficient to enable significantdiversification in wide-ranging animals. Sympatric and par-apatric modes of diversification therefore could be espe-cially important in explaining Quaternary diversificationof wide-ranging habitat-generalist species. In this study,we examine the role of a behavioral phenomenon as a diver-sifying agent in a wide-ranging carnivore.

It is well known that carnivores typically disperse longdistances and often exhibit low levels of genetic divergencebetween distantly separated sampling sites (Lehman andWayne 1991; Harrison 1992; Roy et al. 1994; Schwartzet al. 2002). Although these observations, gleaned primarilyfrom coarse grain studies, clearly indicate that gene flowcan homogenize carnivore populations across large por-

tions of their range, they do not necessarily imply that car-nivores cannot also exhibit strong population structurewithin certain parts of their range. Such high-structure por-tions of ranges could be evolutionarily important, especiallyif subdivisions correspond to different habitat types withcorrespondingly different selective regimes. For example,effects of random genetic drift are expected to be greaterin subdivided populations, which can increase a species’adaptive potential by enabling individual populations to‘‘explore the adaptive landscape’’ thereby reaching noveland varied fitness peaks (Wright 1931, 1932; Wade andGoodnight 1998). Habitat-specific subdivision also maybe expected to facilitate the maintenance of functionalgenetic variability through differential local selection(Maynard-Smith and Hoekstra 1980; Hedrick 1986;Spichtig and Kawecki 2004). Such diversification amongpopulations, whether due to drift or selection, might beexpected, in turn, to lead species overall to be more resilientto environmental changes. But why expect highly mobileanimals to exhibit population structure in certain parts oftheir range, and why expect this to correspond to habitat?

Recent evidence from behavioral studies indicates thatmany animals exhibit dispersal preferences toward habitatresembling their natal habitat (reviewed by Davis andStamps 2004), a tendency, which, if strong enough, canbe expected to translate to discretely subdivided popula-tions along physically unobstructed habitat boundaries. Re-gions of high habitat heterogeneity, in turn, would beexpected to correspond to high-structure portions of animalranges. Examples of such habitat-specific subdivisions areaccumulating (Geffen et al. 2004; Sacks et al. 2004; Pilot etal. 2006), as is evidence for conservative habitat selectionbehavior as a proximate cause (Sacks et al. 2005). For ex-ample, on or near the boundary between mountain and val-ley habitats of radiocollared coyotes, no interpack (i.e.,between territory) relationships (i.e., relatedness .0.25)were observed across the boundary although several

Key words: Canis latrans, gene flow, genetic diversity, geneticstructure, habitat, isolation-by-distance.

E-mail: [email protected].

Mol. Biol. Evol. 25(7):1384–1394. 2008doi:10.1093/molbev/msn082Advance Access publication April 7, 2008

� The Author 2008. Published by Oxford University Press on behalf ofthe Society for Molecular Biology and Evolution. All rights reserved.For permissions, please e-mail: [email protected]

by guest on Decem

ber 27, 2015http://m

be.oxfordjournals.org/D

ownloaded from

http://mbe.oxfordjournals.org/

interpack relationships occurred within habitats, indicatingthat dispersal across the habitat boundary was rare (Sackset al. 2005). The ecological and evolutionary significanceof these small-scale studies, however, depends first onthe spatial extent of such habitat-specific subdivisionsand second on their strength relative to that of other diver-sifying factors, such as isolation-by-distance, which poten-tially swamp this locally important isolating mechanismover larger spatial extents. In this study, we investigated thisquestion using genomic profiles of .2,000 coyotes from 2broadly defined ecoregions of southwestern North Americacontrasting greatly in their levels of habitat heterogeneity: 1)the California Floristic Province (CFP) and 2) southwesterndeserts and adjacent short grass prairie within the ‘‘Dry do-main’’ (Bailey 1983), here referred to as the Desert–Prairieecoregion (DPE).

The CFP is one of the most biologically diverse ecor-egions in the world and, like other Mediterranean ecore-gions (i.e., in Chile, South Africa, Australia, around theMediterranean Sea), is known for its landscape heterogene-ity and correspondingly high phylogeographic structureamong low-mobility organisms (e.g., Calsbeek et al.2003). The DPE contrasts with the sharply dissected land-scape of the CFP most importantly in that it is composed ofcontinuous tracts of gradually intergrading flora and fauna.Thus, habitat in the DPE is relatively homogeneous (at thecoyote-relevant grain) within the typical dispersal radius ofa coyote and dispersal is therefore predicted to be relativelyunconstrained by natal-habitat–biased dispersal. Coyotes—as a species—are continuously distributed across both ofthese ecoregions in virtually every type of habitat (Youngand Jackson 1951).

Here, we investigate 3 predictions about the contem-porary population structure of coyotes stemming from thenatal-habitat–biased dispersal hypothesis: 1) coyotes sam-pled from the CFP exhibit population structure concordantwith habitat subregions, 2) coyotes from widely dispersedsampling sites within the DPE exhibit little or no structure,and 3) genetic diversity within CFP subregions is lowerthan in DPE subregions due to higher rates of random ge-netic drift in the former. We used 14 autosomal, biparen-tally inherited microsatellite loci and paternally inheritedY chromosome haplotypes (composed of 5 nonrecombin-ing microsatellite loci) to test these predictions. We alsoused mitochondrial DNA (mtDNA) data from different in-dividuals previously published by Lehman and Wayne(1991) as an independent check with respect to a maternallyinherited marker.

Materials and MethodsSampling

Coyote tissue samples, including muscle and driedblood of filter paper strips (Nobuto Blood Filter Strips, Ad-vantec Manufacturing, Dublin, CA), were collected andmapped as described previously (Sacks et al. 2004) fromthroughout California. An additional 30 coyotes (20 malesand 10 females) were sampled from blood clots donated byEric Gese from southeastern Colorado. The CFP samplingwas concentrated in 8 bioregions (hereafter, ‘‘subregions’’):

CFP-Northwestern, CFP-northern Great Valley, CFP-SierraNevada, CFP-Cascades, CFP-Central Western (i.e., centralCalifornia Coast ranges), CFP-southern Great Valley, CFP-Southwestern, and CFP-Transverse Ranges (Hickman1993). The samples from the DPE spanned 5 subregions:DPE-Modoc Plateau, DPE-Eastern Sierra-Nevada, DPE-Mojave Desert, and DPE-Sonoran Desert of California,and DPE-short grass prairie of southeastern Colorado.

Microsatellite Genotyping

Procedures for DNA extraction were described previ-ously (Sacks et al. 2004). Coyotes were genotyped at 14autosomal microsatellite loci used in previous studies(Sacks et al. 2004, 2005) and 5 Y chromosome loci(MS41A, MS41B, Sundqvist et al. 2001; 990-35, 650-79.2A, 650-79.2B, Bannasch et al. 2005). Two Y chromo-some loci (650-79.2A and 650-79.2B) were coamplified bythe same pair of primers. As has been done previously forcoamplified Y chromosome microsatellites (Malaspinaet al. 1998), we assumed that the smaller of the 2 coampli-fied alleles corresponded to the ‘‘A’’ locus, whereas thelarger of the 2 alleles corresponded to the ‘‘B’’ locus; wherea single peak was noted, each locus was assigned the sameallele. In this case, the procedure was likely as accurate as ifwe had amplified the loci separately because the overlap inallele size between the A and B loci involved a single allele.That is, the occurrence of single peaks only occurred at thissize and there were no instances of co-occurrence in thesame genotype of peaks both higher or both lower than thissize. Polymerase chain reaction (PCR) conditions were de-scribed for autosomal loci elsewhere (Sacks et al. 2004) aswere those for Y chromosome loci (Sundqvist et al. 2001;Bannasch et al. 2005). The PCR products were visualizedand alleles scored using an ABI 3730 capillary DNA ana-lyzer (Applied Biosystems, Foster City, CA) and STRand2.2.30 software (Veterinary Genetics Laboratory, Univer-sity of California, Davis, CA).

Autosomal Genotype-Based Analysis with ProgramStructure

The principle on which this method works is to clustermultilocus genotypes (i.e., representing individual coyotes)so as to minimize within-cluster deviations from Hardy–Weinberg and linkage equilibrium (Pritchard et al.2000). The number of clusters (K) was chosen after runningthe algorithm 10 times each at K 5 1–10 and plottingagainst K the average logarithm of the probability of thedata given K [log Pr (X|K)], the geographic index (the av-erage geographic distance between all pairs of coyoteswithin clusters divided by that between all coyotes in theentire sample; Sacks et al. 2004), and the consistency acrossruns (the percent of coyotes assigned to the same cluster inthe 2 runs with the highest log Pr (X|K)). These initial runsincluded 10,000 Markov Chain Monte Carlo (MCMC) runsfor burn-in (i.e., transition from starting conditions to sta-tionarity) and 10,000 additional postburn-in runs. A final runat the selected K value was conducted using 30,000-cycle

Selective Dispersal Structures Coyote Populations 1385

by guest on Decem



ownloaded from


burn-in and 1 million–cycle postburn-in and used for theanalysis. The admixture model was specified, allele frequen-cies were assumed to be correlated among genetic groupings(Falush et al. 2003), and default valueswere used for all otherparameters. The analyses were performed in Structure(V 2.0; Pritchard et al. 2000). Locations were displayed usingArcView GIS (v 3.2) software (ESRI, Redlands, CA) colorcoded with respect to their cluster assignment, defined as thecluster corresponding to the highest estimated proportion ofancestry.

Autosomal Genotype-Based Analysis with ProgramGeneland

This method is similar to that implemented in theprogram Structure except that spatial proximity is usedto weight genetic cluster assignments (Guillot, Estoup,et al. 2005; Guillot, Mortier, and Estoup 2005). Also, pro-gram Geneland enables the number of clusters (K) to betreated as a variable or to be fixed and, when varied, enablesdetermination of the modal (i.e., most likely) value. We per-formed a preliminary run whereKwas allowed to vary from1 to 30 to determine the modal number of clusters, whichwas 26.We then ran the analysis 5 times with the number ofclusters fixed to 26. The number of clusters with individualsassigned to them varied from 8 to 10, with the majority ofclusters having no genotypes assigned to them (normal forthis program; Guillot, Estoup, et al. 2005; Guillot, Mortier,and Estoup 2005). All runs were conducted using the spatialD-model for the priors in allele frequency and 200,000MCMC iterations, a maximum of 200 nuclei, and a10,000-iteration postprocessing burn-in consistent withrecommendations (Guillot, Estoup, et al. 2005; Guillot,Mortier, and Estoup 2005).

Allele Frequency-Based Analyses

These analyses were conducted using autosomal mi-crosatellites, Y chromosome haplotypes (composed of 5haploid microsatellite alleles), and the mtDNA haplotypesof Lehman and Wayne (1991). For autosomal and Y chro-mosome analyses, we grouped coyotes into 45 samples ac-cording to spatial proximity (i.e., sites), of which 39 wereused for autosomal analyses (n . 14 coyotes each, averagen 5 45) and 36 were used for Y chromosome analyses (.8male coyotes, average n 5 24). Fifteen similarly locatedsites identified by Lehman and Wayne (1991; table 1, sitesd–q, s; average n 5 12.4 coyotes) were used in analyseswith their mtDNA restriction fragment length polymor-phism data (table 3). Specifically, mtDNA data weregrouped as follows: sampling sites d, o, q, s (DPE); e, j, l(CFP-southern Great Valley); h (CFP-northern GreatValley); i, k (CFP-CentralWestern); f, g (CFP-Northwestern);and m, n, p (CFP-Southwestern).

To assess the presence of genetic isolation-by-distance, for each marker type, we tested for correlationsbetween genetic distance and geographic distance usingMantel tests (Mantel 1967) performed in Arlequin(v2.000; Schneider et al. 2000) with significance basedon 1,000 matrix permutations. The dependent variable

matrix was of genetic distance (expressed as F̂ST=ð1� F̂STÞ;Slatkin 1995) calculated in Arlequin (v2.000). The indepen-dent variable matrix was of geographic distance (kilometer)between sample sites (i.e., centroids), calculated as thesquare root of the sum of squared differences in centroidX-coordinates and centroid Y-coordinates. The centroidwas calculated as average coordinates among coyotes ineach site; coordinates were projected in Teale Albers beforecalculating centroids and distances using ArcView GISsoftware (v 3.2; ESRI). Geographic distances associatedwith some pairs of sampling sites were calculated fromthe sum of 2 or 3 of these straight-line distances to avoidcutting across the San Francisco Bay Estuary (includingSan Pablo Bay and delta), whichwas assumed to completelyobstruct dispersal (Sacks et al. 2004).

Nested hierarchical analysis of molecular variance(AMOVA; Excoffier et al. 1992) was conducted for eachmarker type to assess the proportion of between-site geneticvariance between subregions relative to that within sub-regions, using sampling sites as replicates. The AMOVAwas performed in Arlequin (v2.000) using 1,000 permutationsto test for significance. Because our prediction was that theDPE would not contain substructure and in this sense be anal-ogous to a subregion of the CFP, we treated this ecoregion onthe same hierarchical level as subregions for this analysis.

Within-Subregion Genetic Diversity

To compare genetic diversity between CFP and DPEsubregions, we used the sampling sites as the sample units,thereby minimizing the potential for confounding effects ofpopulation substructure. Because the number of coyotesvaried somewhat from site to site and could affect estimatesof genetic diversity, we included sample size as a covariatein the analysis. To maximize statistical power, we per-formed a single analysis using all marker types and in-cluded an interaction term for marker type and ecoregionto test for the possibility that one or both of the uniparen-tally inherited markers showed a larger difference, as mightbe expected due to their smaller effective population sizes.We used as our measure of genetic diversity

1 �XX

p2;

Table 1Nested Hierarchical AMOVA Results for the Assessment ofGenetic Differentiation between Subregions Relative toVariation among Sampling sites within Subregions Using3 Types of Molecular Genetic Marker

BetweenSubregions

(F̂CT)

Between-Siteswithin

Subregions(F̂SC)

BetweenSites(F̂ST)

Between-SiteVariance

Explained bySubregions (%)

Autosomal 0.017** 0.022** 0.039** 44**mtDNAa 0.136** 0.068* 0.195** 70**Y chromosome 0.092** 0.089** 0.172** 53**

NOTE.—Significance was based on 1,000 permutations. For this analysis, the

DPE was treated as a subregion.a Calculated from data in Lehman and Wayne (1991).

*P , 0.01, **P , 0.001.

1386 Sacks et al.

by guest on Decem



ownloaded from


where p is allele frequency, the first summation is acrossloci (one in the case of mtDNA and Y chromosomes),and the second summation is across alleles within loci(Weir 1996). For the autosomal microsatellite loci, thismeasure was equivalent to expected heterozygosity. Weemployed a general linear model in SYSTAT (v9.0; SPSSIncorporated, Chicago, IL) with the following variables:ecoregion (CFP, DPE), marker type (autosomal, mtDNA,and Y chromosome), ecoregion*marker type, and, as a co-variate, sample size (number of coyotes in site).

Hardy–Weinberg and Linkage Equilibrium

For the autosomalmarkers, we testedHardy–Weinbergequilibrium (Guo and Thompson 1992) in Arlequin(v2.000) and linkage disequilibrium in Genepop (v3.3;Raymond and Rousset 1995) within sampling sites. Weused a sequential Bonferroni test to determine significanceof deviations from Hardy–Weinberg and linkage equilib-rium (Rice 1989).

Results

A total of 301 alleles was identified among the 14 au-tosomal microsatellite loci, with the number of alleles rang-ing from 8 to 48 per locus. We identified a total of 73 Ychromosome haplotypes composed of the 5 microsatelliteloci. These loci contained 34 alleles in total, with numbersranging from 4 to 14 alleles per locus. A total of 32 mtDNAhaplotypes from the Lehman and Wayne (1991) study wasidentified in the subset of data used in the present study.

Structure Analysis

Our first approach, assigning individuals to geneticclusters based on the autosomal microsatellite genotypesirrespective of geography, clearly supported predictions 1(CFP population is structured according to habitat bound-aries) and 2 (DPE population is relatively unstructured)(fig. 1). We chose K 5 6 for the final run in this analysisbecause both the geographic index and log Pr(X|K) plateauhere (as minimum and maximum, respectively) and highervalues of K were unstable in terms of variance in logPr(X|K) and/or consistency of assignments among runs(fig. 2). The analysis revealed substantial subregion-spe-cific genetic structure within the CFP and none in theDPE despite the vastly greater spatial extent of the latter.These results were highly consistent across runs and robustto numbers of clusters assumed. For example, at K 5 8,DPE coyotes clustered together in a single cluster, similarlyto K 5 6, but the Central Western and Southwestern sub-regions were each divided into 2 clusters. Genetic subdivi-sions clearly corresponded to subregion boundaries andcould not be explained by physical barriers associated withrecent anthropogenic habitat modifications. For example,the entire Southwestern subregion clustered together de-spite a tremendously fragmented landscape and numeroushighways with high volume traffic that currently pose bar-riers to gene flow (Hunter et al. 2003; Riley et al. 2006).

Geneland Analysis

Similarly to the Structure analysis, several distinctclusters corresponding to subregions were identified withinthe CFP, whereas all coyotes from the DPE-Modoc Plateauand the DPE-Short Grass Prairie were clustered togetherdespite up to 1,500 km separating them. However, DPE-Mojave Desert and DPE-Sonoran Desert coyotes were clus-tered with these other DPE coyotes in only 3 of the 5 runs,with the other 2 runs grouping most DPE-Mojave Desertand DPE-Sonoran Desert coyotes with the CFP-southernGreat Valley cluster. Interestingly, the CFP-southern GreatValley subregion is the most similar of the CFP subregionsto the DPE-Mojave Desert and DPE-Sonoran Desert sub-regions with respect to vegetation and climate (Hickman1993). Several clear boundaries also were identified betweensubregions within the CFP in each of the runs (fig. 3), al-though the particular boundaries corresponding to geneticclusters varied somewhat across runs; this instability likelywas due to the nonrandom distribution of sampling locations(Guillot, Estoup, et al. 2005; Guillot, Mortier, and Estoup2005). Nevertheless, the Geneland runs consistently lumpedmost DPE coyotes together and identified separate clusterscorresponding to CFP subregions despite the method’s pre-disposition to cluster spatially proximate samples togetherand spatially distant samples apart.

Allele Frequency-Based Analyses

Genetic diversity was similar for CFP ecoregion andDPE as wholes (i.e., data pooled across sites and sub-regions) for autosomal microsatellites (0.80, 0.79, respec-tively), Y chromosome microsatellites (0.94, 0.91), andmtDNA markers (0.88, 0.87). As with Structure and Gene-land analyses, pairwise genetic distances based on allelefrequencies likewise supported predictions 1 and 2. Therewas a weak relationship between genetic distance and geo-graphic distance based on the autosomal (Mantel testr2 5 0.08; P 5 0.008) and Y chromosome (r2 5 0.04;P 5 0.047) markers and no significant relationship basedon mtDNA markers (r2 5 0.00; P 5 0.32). However,the AMOVA analysis indicated that any isolation-by-distance relationship was dwarfed by the amount of var-iance explained by subregions; whereas geographic distanceexplained 0–8% (depending on marker type) of thebetween-site genetic variance (indicated by r2 valuesabove), subregions explained 44–70% of between-site ge-netic variance (table 1). Genetic distances between sites indifferent subregions within the CFP were highly variableand higher on average than those in the DPE, which wereuniformly low despite their greater (on average) separationdistances (fig. 4). Within subregions of the CFP, geneticdistances were low similarly to those over the entireDPE. While sampling and estimation error contribute tothe observed variance, these sources of error cannot explainthe difference in variance in between- versus within-subregion (and the DPE) comparisons (fig. 4). Thus, thehigh variance in genetic distances between different CFPsubregions, even at short geographic distances, suggeststhat subregion boundaries varied considerably with respectto their permeability to gene flow.


by guest on Decem



ownloaded from


Within-Site Genetic Diversity

Also as predicted (prediction 3)—and in contrast toecoregion-widegeneticdiversity—within-sitegeneticdiver-sity was lower in the CFP than in the DPE (F1,71 5 7.0,P 5 0.01; fig. 5). Although differences were somewhatgreater in uniparentally inherited markers, consistent withtheir smaller effective population sizes, there was no signif-icant interaction between marker type and ecoregion

(F2,71 5 0.87, P 5 0.42). Number of coyotes was not asignificant covariate in themodel (F1,71 5 0.42,P 5 0.52).

Hardy–Weinberg and Linkage Equilibrium

Loci were significantly out of Hardy–Weinberg equi-librium in 66 instances (out of 616 population locus cases),but these were distributed widely among populations and

FIG. 1.—Coyote sampling locations (n 5 2,045) color coded according to assignment (in Structure) in 1 of 6 distinct genetic groupings in relationto 13 subregions (bounded by thin lines) of the California Floristic Province (CFP; bounded by thick line, labeled in yellow boxes) and the Desert-Prairie ecoregion (DPE, labeled in grey boxes) (upper left). Colored circles and asterisks indicate individuals with .80% and ,80%, respectively, oftheir ancestry estimated to be from a single genetic grouping. Blow ups illustrate sharp genetic divisions in coyote populations corresponding todramatic changes in the landscape with no significant physical barriers: a finger of Great Basin Desert habitat, the Carson Valley, extends into the SierraNevada Mountains (top right); Los Angeles Basin meets the Transverse Ranges and the Transverse Ranges meet the Mojave Desert (bottom left).Lower right: southwestern United States, illustrating sampling site centroids (black circles) for geographic groupings used in allele frequency analysesin relation to the CFP (gray) and DPE (white), with the southern Rocky Mountains indicated for reference (black). The spatially nonrandom distributionof samples reflects sample availability rather than coyote distribution, which was continuous throughout the entire study area (i.e., of the CFP and DPE).Subregion abbreviations are as follows: CFP-Northwestern (NW), CFP-northern Great Valley (GVN), CFP-Sierra Nevada (SN), CFP-Cascades (Cas),CFP-Central Western (CW), CFP-southern Great Valley (GVS), CFP-Southwestern (SW), CFP-Transverse Ranges (TR), DPE-Modoc Plateau (Mod),DPE-Eastern Sierra Nevada (ESN), DPE-Mojave (Moj), DPE-Sonoran (Son) deserts of California, and DPE-short grass prairie of southeasternColorado (SGP).

1388 Sacks et al.

by guest on Decem



ownloaded from


loci. There was, however, one systematic relationship,a positive correlation between proportion of loci signifi-cantly out of Hardy–Weinberg equilibrium and sample size(r 5 0.67), to be expected due to the greater sensitivity

(higher statistical power) of the tests employed on largersamples. The average FIS (based on the heterozygote defi-ciency) was 0.09, which is similar to previous estimates forcoyotes (Roy et al. 1994; Sacks et al. 2004). There were 198instances of locus pairs in linkage disequilibrium out of a to-tal 4,186 possible, but these were not clustered at any par-ticular locus pair. However, there were 2 sampling sites thathad much higher proportions of locus pairs in linkage dis-equilibrium (32%, 47%). Interestingly, these were bothpairs that straddled the Sierra Nevada and southern GreatValley subregions. Thus, although Structure and Genelandanalyses did not illuminate a genetic subdivision along thisboundary, the high degree of linkage disequilibrium ob-served in these 2 sites nonetheless suggests the possiblepresence of substructure along this boundary. Excludingthe 2 outlier populations, as with Hardy–Weinberg equilib-rium, there was a positive correlation between proportionof locus pairs significantly out of linkage equilibriumand sample size (r 5 0.55).

Discussion

We investigated 3 predictions of the natal-habitat–biased dispersal hypothesis with respect to the contemporarypopulation structure of coyotes in a Mediterranean ecore-gion: 1) coyotes sampled from the CFP exhibit populationstructure concordant with habitat subregions, 2) coyotesfromwidely dispersed sampling sites within the neighboringDPE exhibit little or no structure, and 3) genetic diversitywithin CFP sites is lower than in DPE sites due to higherrates of random genetic drift in the former. Our findingsclearly supported each of these predictions based on auto-somal, Y chromosome, and mtDNA markers. Although wehad previously observed, based on microsatellites, that coy-ote population subdivisions were concordant with habitatsin central California (Sacks et al. 2004, 2005), the presentstudy, based on a 4-fold increase in the number of coyotesand a much larger study area, demonstrated that the phe-nomenon was present throughout the CFP and, importantly,absent from the DPE. The occurrence of a few habitat-spe-cific subdivisions also had been reported for gray wolves(Canis lupus) across North America (Carmichael et al.2001; Geffen et al. 2004) and eastern Europe (Pilot et al.2006). However, the present study is the first of which weare aware to show extensive habitat-specific substructuringin a continuously distributed species throughout a heteroge-neous ecoregion and in direct contrast to low genetic diver-gence between distantly sampled sites in an adjacent,relatively homogeneous ecoregion.

Several aspects of this study serve to strengthen theinference that natal-habitat–biased dispersal was the pri-mary cause of the population structure observed in theCFP. First, the hypothesis and predictions were formulateda priori based on individual-level criteria and preliminarilytested on smaller scales (Sacks et al. 2004, 2005). Theexpansion to the entirety of the CFP, in particular the ad-ditional habitat subdivisions, in the current study consti-tuted replication. Finally, the DPE essentially served asa negative control, providing a reference in the case whereno sharply differentiated habitat types occurred. The high

FIG. 2.—Three criteria used to select the K to display graphically:average (±standard deviation) log Pr(X|K) across 10 runs with 10,000iteration burn-in and 10,000 postburn-in iterations, geographic index(calculated for each K from the run with the highest associated logPr(X|K)), and individual assignment consistency from one run to the next(using the 2 runs with the highest associated log Pr(X|K)). Log Pr(X|K)also shown for the final run at K 5 6, with a 30,000 iteration burn-in and106 postburn-in iterations.

FIG. 3.—One of 5 runs in Geneland, showing coyote samplinglocations color coded according to assignment in 1 of 8 distinct geneticgroupings in relation to 13 subregions (bounded by thin lines) of the CFP(bounded by thick line) and the DPE (all other samples). Samples in thecircle were from the DPE-short grass prairie subregion from southeasternColorado.


by guest on Decem



ownloaded from


genetic diversity within sites of this ecoregion indicatedthat the genetic homogeneity among sites of the DPEwas due to high contemporary rates of gene flow rather thana recent population expansion. Thus, the genetic homoge-neity associated with this vast ecoregion served to empha-size the genetic discontinuities within the heterogeneousCFP. In fact, the magnitude of genetic divergence at habitatboundaries in the latter ecoregion was, on average, consid-erably greater than that due to geographic distance overa substantial portion of the species range, suggesting thathabitat-specific population subdivisions could be of funda-mental significance.

Although it is difficult to explain the observed popu-lation genetic structure without invoking natal-habitat–biased dispersal as the primary cause within the CFP, a briefconsideration of additional factors is warranted. Otherprocesses undoubtedly contributed to various extents ondifferent spatiotemporal scales to the observed populationstructure, including historical patterns associated withPleistocene/Holocene climatic change (Hewitt 1996) aswell as contemporary processes such as habitat-specific de-mography, habitat selection asymmetries, and metapopula-tion dynamics. For example, despite the high within-sitegenetic diversity observed in the DPE, which indicates highlevels of contemporary gene flow, it remains possible thatthe population stems from a post-Pleistocene range expan-sion (but sufficiently in the past for gene flow and drift to

have equilibrated). Further, the subdivision between theCFP and DPE could reflect secondary contact between pre-viously isolated populations associated with inhospitablehabitat during much of the Pleistocene along the crest ofthe Sierra Nevada Mountains, as appears to be the casefor many other taxa (Remington 1968; Swenson andHoward 2004). Although coyotes currently occur through-out the Sierra Nevada range, this subregion may containa zone of post-Pleistocene admixture. Extensions of Great

FIG. 4.—Genetic distance plotted against geographic distance,illustrating uniformly low genetic distance between even widely spacedsampling sites of the DPE and contrastingly more variable, and generallyhigher, genetic distances between sampling sites in CFP regardless ofseparation distance. Graphs are shown for autosomal microsatellites (toprow), mtDNA haplotypes based on data from Lehman and Wayne (1991)(middle row), and Y chromosome haplotypes (bottom row). The highergenetic distances associated with the mtDNA and Y chromosome markersthan the autosomal markers are consistent with the lower effectivepopulation sizes associated with the uniparentally inherited markers.Average genetic distances (pooled across geographic distances) areshown and indicate greater subdivision between CFP subregions thanwithin CFP subregions or within the DPE for all 3 marker types.

FIG. 5.—Average (±standard error) gene diversity (calculated fromallele frequencies, p) within sampling sites of the CFP versus DPE basedon (top) autosomal microsatellites, (middle) mtDNA, and (bottom) Ychromosome markers.

1390 Sacks et al.

by guest on Decem



ownloaded from


Basin Desert habitat along the eastern Sierra Nevada cor-responded to the Desert-Prairie genetic cluster, whereascoyotes along the western Sierra Nevada clustered sepa-rately (with the Great Valley subregions), consistent witha subdivision along the crest (e.g., fig. 1). On the other hand,in South Lake Tahoe, our only eastern Sierra Nevada sam-pling site with typical high Sierra Nevada habitat (e.g., co-niferous forest and alpine meadow), coyotes clustered withtheir western Sierra Nevada counterparts, supporting a sharpsubdivision approximately on the eastern edge of this sub-region (e.g., corresponding to the meeting of Great Basinand Subalpine vegetation types), as opposed to the crestof the Sierra Nevada per se. Phylogenetic analyses of Ychromosome and rapidly mutating regions of the mtDNAgenome are needed to investigate and compare the histor-ical demography of coyote populations in both ecoregionsto assess these phylogeographic scenarios.

Regardless of the possible role of a Pleistocene barrierin the Sierra Nevada, a similar explanation seems unlikelyto account for the other habitat-specific population subdi-visions. No such barriers are known to have occurred alongborders of most subregions in the CFP, nor are they sug-gested by the presence of multispecies hybrid zones be-tween populations of late Pleistocene age (Remington1968; Swenson and Howard 2004). Genetic subdivisionscorresponded to nearly all subregional boundaries withinthe CFP, although the magnitude varied, apparently stron-gest between coastal and interior subregions and weakestbetween the Sierra Nevada and Great Valley subregions.Although we did not detect a clear subdivision separatingthe Sierra Nevada from the Great Valley in the presentstudy, previous analyses conducted on smaller spatialscales indicated clear subdivisions separating the westernSierra Nevada from the northern Great Valley (Sackset al. 2004, 2005). Further, the high level of linkage disequi-librium observed within sampling sites near the SierraNevada boundary with the south Great Valley in thisstudy suggests the possibility of genetic subdivision inthe southern portion of these subregions as well.

Our primary focus in this study was on identificationof boundaries between coyote subpopulations. However,a full understanding of the role of habitat boundaries in de-termining population dynamics will require an understand-ing of fine-scale social processes, which can reinforcesubdivisions, as well as asymmetries in demographic pro-cesses and habitat selection proclivities (apart from natalhabitat), which can be expected to affect the evolutionarytrajectory of the metapopulation. As might be expected,fine-grained studies indicate that relatedness among packs(i.e., family groups) in neighboring territories tends to beconsiderably greater on the same side than on opposite sidesof a habitat or physical subdivision (Sacks et al. 2005; Rileyet al. 2006). To the extent that dispersal is facilitated bysettlement adjacent to relatives (Temeles 1994), social fab-ric could further deter dispersal across genetic subdivisions(Sacks et al. 2005). Where boundaries are hard, an increasein density due to territory packing may have a similar effectirrespective of familial relationships among neighboringpacks (Riley et al. 2006). In addition to social factors, dif-ferential demography and asymmetric dispersal betweenhabitats could influence the particular patterns of subdivi-

sion (Stenseth et al. 2004). Habitats associated with subre-gions also may vary in quality with some serving as sourcepopulations and others as sinks (Pulliam 1988). Previously,Sacks et al. (2005) found gene flow to be considerably high-er from the CascadeMountains to the northern Great Valleythan in the opposite direction in one local study site, con-sistent with differential habitat quality, although it was un-known to what extent the asymmetry in gene flow wasdriven by demographic versus behavioral factors. In thepresent study, the high variability in genetic divergence be-tween nearby subregions of the CFP indicated that patternsof gene flow may be more complex than is indicated by thesimple pattern of subdivision apparent in from the Structureand Geneland analyses. These details, which could includeextinction/recolonization, source/sink, and other metapopu-lation dynamics, should be investigated in future studies.

Regardless of the mix of factors that ultimately deter-mine and refine patterns of population genetic structure, thepredominance of enduring habitat-specific population sub-divisions, likely due to natal-habitat–biased dispersal, couldhave important evolutionary implications on both micro-and macroevolutionary timescales. For example, one ofthe fundamental problems of microevolution pertains tothe maintenance of genetic diversity in natural populations.Drift-based hypotheses emphasize the role of populationgenetic subdivisions enabling small, partly isolated subpo-pulations to ‘‘explore’’ and find alternative fitness peaks onthe adaptive landscape by freeing the genomes from strongselection for universally adaptive phenotypes (e.g., Wright1931, 1932), whereas selection-based theories emphasizethe role of habitat heterogeneity in causing differential se-lection in different localities (Levene 1953; Gillespe 1978;Gillespe and Turelli 1989; Spichtig and Kawecki 2004).Despite their disparate roots in the literature, drift-basedand selection-based frameworks are not mutually exclusive.Indeed, population structure corresponding to habitat sub-divisions can be expected to increase the occurrence of bal-anced polymorphism because a tendency to mate within thehabitat in which an individual is born results in a correlationbetween the habitat individuals select and that in which theyare most fit, which should increase the influence of localselection relative to between-habitat selection (Taylor1976; Jones 1980; Jones and Probert 1980; Garcia-Dorado1986; Hedrick 1986). The level of subdivision observed inthe present study is in the range Wright (1931, 1932) rea-soned, and Wade and Goodnight (1991) showed experi-mentally, would facilitate drift-based evolution. Theobservation that genetic diversity within sites of the CFPwas less than in the DPE is also consistent with higher ef-fects of genetic drift within the more subdivided portion ofthe range. Moreover, although we did not investigate selec-tion in this study, habitat-specific differences in selectivepressures seem likely and would be expected to further pro-mote divergence. For example, a recent empirical study in-dicated that local habitat-specific adaptation was sustainedin the presence of as much as 13% immigration (Postma andvan Noordwijk 2005).

On the macroevolutionary end, natal-habitat–biaseddispersal conceivably provides a powerful reinforcingmechanism for reproductive isolation and parapatric speci-ation. The ability of populations to diverge evolutionarily in


by guest on Decem



ownloaded from


the presence of incomplete reproductive isolation is an ex-citing area of research that has received renewed interest inrecent years (Bolnick et al. 2003). The conditions typicallylisted for sympatric or parapatric divergence include nichediversification, involving some form of disruptive selectionand assortative mating. Species with broad niches (i.e., gen-eralists) may be ideal candidates for the study of sympatric/parapatric niche diversification, especially, when they arecomposedby individualswithnarrowniches (i.e., specialists),but which vary widely among individuals (Roughgarden1972; Bolnick et al. 2003). Potential for reproductive iso-lation would seem to be greatest where niche diversificationis concordant with spatial segregation, as is the case withnatal-habitat–biased dispersal.

Although we do not suggest that contemporary coyotesubpopulations in the CFP are incipient species, the phe-nomenon observed in this study could have contributedto speciation at other times or in other taxa in this and otherMediterranean ecoregions. The CFP, and Mediterraneanecoregions generally, are hotbeds of endemism and biodi-versity, characterized by high beta diversity. Although thispattern is thought to stem partly from rapid phylogeneticdiversification associated with past geologic and hydrologicisolating events (Myers et al. 2000; Latimer et al. 2005)—supported within the CFP by a high correspondence in thephylogeography and temporal distribution of low-mobilityspecies (e.g., Maldonado et al. 2001; Calsbeek et al. 2003;Spinks and Shaffer 2005)—the role of natal-habitat–biaseddispersal and, consequently, parapatric speciation alsocouldhaveplayed a role in generating contemporarypatternsof biodiversity. This seems especially likely given the rela-tively stable climate associated with the CFP throughout thePleistocene glacial cycles, providing not only an ice-free re-fugium forwestern fauna and flora but also long-termhabitatheterogeneity. Thus, it is intriguing to ponder the extent towhich natal-habitat–biased dispersal might have affectedthe evolution of vertebrates in this and similar regionsglobally. Given the high degree of population structuringobserved in the wide-ranging coyote, it seems likely thatnatal-habitat–biaseddispersal could be evenmore influentialfor less wide-ranging generalist species in the ecoregion.Indeed, behavioral studies indicate the presence of natal-habitat–biaseddispersal in a largenumberof small vertebratetaxa (Davis and Stamps 2004). At present, investigations ofthe population genetic consequences of this phenomenon insmaller vertebrates are lacking, providing an exciting andwide-open area in need of future research.

Conclusions

Our results, based on biparentally, paternally, and ma-ternally inherited markers, clearly demonstrate substantialpopulation genetic structure in coyotes of the CFP in con-trast to low genetic divergence among sampling sites span-ning a large area of the desert and prairie. Moreover, theassociated reduction in genetic diversity observed withinsites of the CFP indicates that structuring in that regionwas sufficient for genetic drift to leave a visible mark onallele frequencies. Although historical contingency andmetapopulation dynamics could contribute to the observed

patterns to some extent, these factors alone seem in-sufficient to explain them without also invoking natal-habitat–biased dispersal as a fundamental cause. Thus,although coyotes commonly disperse distances of 100 kmor more (Harrison 1992; Sacks et al. 2005) and, as a species,occupy virtually every habitat type (Young and Jackson1951), their propensity, as individuals, to disperse intohabitat similar to their natal habitats was apparently ofsufficient strength to structure their populations. Giventhe highly vagile nature and catholic habitat preferencesof our study species, these findings also are likely applica-ble to a broad spectrum of terrestrial vertebrates most ofwhich are less mobile and less continuously distributedto begin with. If so, heterogeneous landscapes such as thatof the CFP, which have also acted as refuges during glacialmaxima, could play an important evolutionary role by pro-viding high-structure portions of animal ranges, accelerat-ing evolutionary divergence, for example, during glacialcycles. Because habitat-specific structure can enhance aspecies’ adaptive potential and resilience to changingenvironments, these findings further suggest the CFPmay constitute an evolutionarily important portion of therange for coyotes and sympatric species exhibiting habitat-specific population structure.

Acknowledgments

We are grateful to J.Wiscomb, V. Kramer, K. Blejwas,B. R.Mitchell, L. Lyren, E. Boydston, S. Riley, D. Simpson,N. Tietje, E. Gese, and USDA/WS specialists, too manyto name, for contributing samples. We thank A. M. Irish,K. Records, E. Granzow, C. Williams, J. Well, G. Guillot,and S. K. Brown for technical assistance; J. C. C. Neale R.K.Wayne, and several anonymous reviewers for comments;and N. C. Pedersen for facilitating the project. Supportedfinancially by the University of California (UC) Davis,School of Veterinary Medicine, Veterinary GeneticsLaboratory; by grants from theUCDavis Genetic ResourcesConservation Program; and, logistically, by the USDA/An-imalPlantHealth InspectionService (APHIS)/WildlifeServ-ices, Santa Clara County Vector Control District, andCalifornia Department of Health Services.

Literature Cited

Bailey RG. 1983. Delineation of ecosystem regions. EnvironManage. 7:365–373.

Bannasch DL, Bannasch MJ, Ryun JR, Famula TR,Pedersen NC. 2005. Y chromosome haplotype analysis inpurebred dogs. Mamm Genome. 16:273–280.

Bolnick DI, Svanback R, Fordyce JA, Yang LH, Davis JM,Hulsey CD, Forister ML. 2003. The ecology of individuals:incidence and implications of individual specialization. AmNat. 161:1–28.

CalsbeekR, Thompson JN, Richardson JE. 2003. Patterns of molec-ular evolution and diversification in a biodiversity hotspot: theCalifornia Floristic Province. Mol Ecol. 12:1021–1029.

Carmichael LE, Nagy JA, Larter NC, Strobeck C. 2001. Preyspecialization may influence patterns of gene flow in wolvesof the Canadian Northwest. Mol Ecol. 10:2787–2798.

1392 Sacks et al.

by guest on Decem



ownloaded from


Davis JM, Stamps JA. 2004. The effect of natal experience onhabitat preferences. Trends Ecol Evol. 19:411–416.

Excoffier L, Smouse P, Quattro J. 1992. Analysis of molecularvariance inferred from metric distances among DNAhaplotypes application to human mitochondrial DNA re-striction data. Genetics. 131:479–491.

Falush D, Stephens M, Pritchard JK. 2003. Inference ofpopulation structure using multilocus genotype data: linkedloci and correlated allele frequencies. Genetics. 164:1567–1587.

Garcia-Dorado A. 1986. The effect of niche preference onpolymorphism protection in a heterogeneous environment.Evolution. 40:936–945.

Geffen E, Anderson MJ, Wayne RK. 2004. Climate and habitatbarriers to dispersal in the highly mobile grey wolf. Mol Ecol.13:2481–2490.

Gillespe JH. 1978. A general model to account for enzymevariation in natural populations. V. The SAS-CFF model.Theor Popul Biol. 14:1–45.

Gillespe JH, Turelli M. 1989. Genotype-environment interactionsand the maintenance of polygenic variation. Genetics. 121:129–138.

Guillot G, Estoup A, Mortier F, Cosson JF. 2005. A spatialstatistical model for landscape genetics. Genetics. 170:1261–1280.

Guillot G, Mortier F, Estoup A. 2005. GENELAND: a computerpackage for landscape genetics. Mol Ecol Notes. 5:712–715.

Guo SW, Thompson EA. 1992. Performing the exact test ofHardy-Weinberg proportion for multiple alleles. Biometrics.48:361–372.

Harrison DJ. 1992. Dispersal characteristics of juvenile coyotesin Maine. J Wildl Manage. 56:128–138.

Hedrick PW. 1986. Genetic polymorphism in heterogeneousenvironments: a decade later. Annu Rev Ecol Syst. 17:535–566.

Hewitt GM. 1996. Some genetic consequences of ice ages, andtheir role in divergence and speciation. Biol J Linn Soc.58:247–276.

Hickman JC. 1993. The Jepson manual: higher plants ofCalifornia. Berkeley (CA): University of California Press.

Hunter RD, Fisher RN, Crooks KR. 2003. Landscape-levelconnectivity in coastal southern California, USA, as assessedthrough carnivore habitat suitability. Nat Area J. 23:302–314.

Irwin DE, Irwin JH, Price TD. 2001. Ring species as bridgesbetween microevolution and speciation. Genetica. 112–113:223–243.

Jones JS. 1980. Can genes choose habitats? Nature. 286:757–758.

Jones JS, Probert RF. 1980. Habitat selection maintainsa deleterious allele in a heterogeneous environment. Nature.287:632–633.

Latimer AM, Silander JA, Cowling RM. 2005. Neutral ecologicaltheory reveals isolation and rapid speciation in a biodiversityhotspot. Science. 309:1722–1725.

Lehman N, Wayne RK. 1991. Analysis of coyote mitochondrialDNA genotype frequencies: estimation of the effectivenumber of alleles. Genetics. 128:405–416.

Levene H. 1953. Genetic equilibrium when more than oneecological niche is available. Am Nat. 87:331–333.

Malaspina P, Cruciani F, Ciminelli BM, et al. (24 co-authors).1998. Network analyses of Y-chromosomal types in Europe.North Africa: and Western Asia reveal specific patterns ofgeographic distribution. Am J Hum Genet. 63:847–860.

Maldonado JE, Vila C, Wayne RK. 2001. Tripartite geneticsubdivisions in the ornate shrew (Sorex ornatus). Mol Ecol.10:127–147.

Mantel N. 1967. The detection of disease clustering anda generalized regression approach. Cancer Res. 27:209–220.

Maynard-Smith J, Hoekstra R. 1980. Polymorphism in a variedenvironment: how robust are the models? Genet Res. 35:47–57.

Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GAB,Kent J. 2000. Biodiversity hotspots for conservation priorities.Nature. 403:853–858.

Pritchard JK, Stephens M, Donnelly P. 2000. Inference ofpopulation structure using multilocus genotype data. Genetics.155:945–959.

Pilot M, Jedrzejewski W, Branicki W, Sidorovich VE,Jedrzejewska B, Stachura K, Funk SM. 2006. Ecologicalfactors influence population genetic structure of Europeangray wolves. Molec Ecol. 15:4533–4553.

Postma E, van Noordwijk J. 2005. Gene flow maintains largegenetic difference in clutch size at a small spatial scale.Nature. 433:65–68.

Pulliam HR. 1988. Sources, sinks, and population regulation. AmNat. 132:653–661.

Raymond M, Rousset F. 1995. GENEPOP (Version 1.2):population genetics software for exact tests and ecumenicism.J Hered. 86:248–249.

Remington CL. 1968. Suture zones of hybrid interaction betweenrecently joined biotas. Evol Biol. 2:321–428.

Rice WR. 1989. Analyzing tables of statistical tests. Evolution.43:223–225.

Riley SPD, Pollinger JP, Sauvajot RM, York EC, Bromley C,Fuller TK, Wayne RK. 2006. A southern California freeway isa physical and social barrier to gene flow in carnivores. MolEcol. 15:1733–1741.

Roughgarden J. 1972. Evolution of niche width. Am Nat. 106:683–718.

Roy MS, Geffen E, Smith D, Ostrander EA, Wayne RK. 1994.Patterns of differentiation and hybridization in NorthAmerican wolflike canids, revealed by analysis of micro-satellite loci. Mol Biol Evol. 11:553–570.

Sacks BN, Brown SK, Ernest HB. 2004. Population structure ofCalifornia coyotes corresponds to habitat-specific breaks andilluminates species history. Mol Ecol. 13:1265–1275.

Sacks BN, Mitchell BR, Williams CL, Ernest HB. 2005. Coyotemovements and social structure along a cryptic populationgenetic subdivision. Mol Ecol. 14:1241–1249.

Schneider S, Roessli D, Excoffier L. 2000. Arlequin: a softwarefor population genetics data analysis. Geneva (Switzerland):Genetics of Biometry Laboratory. University of Geneva.

Schwartz MK, Mills LS, McKelvey KS, Ruggiero LF,Allendorf FW. 2002. DNA reveals high dispersal synchro-nizing the population dynamics of Canada lynx. Nature. 415:520–522.

Slatkin M. 1995. A measure of population subdivision based onmicrosatellite allele frequencies. Genetics. 139:457–462.

Spichtig M, Kawecki TJ. 2004. The maintenance (or not) ofpolygenic variation by soft selection in heterogeneousenvironments. Am Nat. 164:70–84.

Spinks PQ, Shaffer HB. 2005. Range-wide molecular analysis ofthe western pond turtle (Emys marmorata): cryptic variation,isolation by distance, and their conservation implications. MolEcol. 14:2047–2064.

Stenseth NC, Ehrich D, Rueness EK, et al. (10 co-authors). 2004.The effect of climatic forcing on population synchrony andgenetic structuring of the Canadian lynx. Proc Natl Acad SciUSA. 101:6056–6061.

Sundqvist AK, Ellegren H, Olivier M, Vila C. 2001. Y chro-mosome haplotyping in Scandanavian wolves (Canis lupus)based on microsatellite markers. Mol Ecol. 10:1959–1966.

Swenson NG, Howard DJ. 2004. Do suture zones exist?Evolution. 11:2391–2397.


by guest on Decem



ownloaded from


Taylor CE. 1976. Genetic variation in heterogeneous environ-

ments. Genetics. 83:887–894.Temeles EJ. 1994. The role of neighbors in territorial systems:

when are they ‘‘dear enemies. Anim Behav. 47:339–350.Wade MJ, Goodnight CJ. 1991. Wright’s shifting balance theory:

an experimental study. Science. 253:1015–1018.Wade MJ, Goodnight CJ. 1998. Perspective: the theories of Fisher

and Wright in the context of metapopulations: when nature

does many small experiments. Evolution. 52:1537–1553.Weir BS. 1996. Genetic data analysis, 2nd ed. Sunderland (MA):

Sinauer.

Wright S. 1931. Evolution in Mendelian populations. Genetics.16:97–159.

Wright S. 1932. The roles of mutation, inbreeding, cross-breeding, and selection in Evolution. Proc 6th Int CongrGenet. 1:356–366.

Young SP, Jackson SP. 1951. The clever coyote. Lincoln (NE):University of Nebraska Press.

Connie Mulligan, Associate Editor

Accepted March 31, 2008

1394 Sacks et al.

by guest on Decem



ownloaded from


Coyotes Demonstrate How Habitat Specialization by Individuals of a Generalist Species Can Diversify...

Documents

Transcript of Coyotes Demonstrate How Habitat Specialization by Individuals of a Generalist Species Can Diversify...