MBE Advance Access originally published online on April 7, 2008
Molecular Biology and Evolution 2008 25(7):1384-1394; doi:10.1093/molbev/msn082
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Research Articles |
Coyotes Demonstrate How Habitat Specialization by Individuals of a Generalist Species Can Diversify Populations in a Heterogeneous Ecoregion
,
* Veterinary Genetics Laboratory, University of California, Davis
Department of Population Health and Reproduction, School of Veterinary Medicine, University of California, Davis
Department of Biological Sciences, California State University, Sacramento
E-mail: bnsacks{at}ucdavis.edu.
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
The tendency for individuals to disperse into habitat similar to their natal habitat has been observed in a wide range of species, although its population genetic consequences have received little study. Such behavior could lead to discrete habitat-specific population subdivisions even in the absence of physical dispersal barriers or habitat gaps. Previous studies of coyotes have supported this hypothesis in a small region of California, but its evolutionary significance ultimately depends on the extent and magnitude of habitat-specific subdivision. Here, we investigated these questions using autosomal, Y chromosome, and mitochondrial markers and >2,000 coyotes from a broad region, including 2 adjacent ecoregions with contrasting levels of habitat heterogeneity—the California Floristic Province (CFP) (heterogeneous landscape) and the Desert–Prairie ecoregion (DPE) (homogeneous landscape). Consistent with predictions, we found a close correspondence between population genetic structure and habitat subdivisions throughout the CFP and virtual panmixia over the larger DPE. Conversely, although genetic diversity was similar in these 2 ecoregions overall, it was lower within sites of the CFP, as would be the expected consequence of greater genetic drift within subregions. The magnitude of habitat-specific genetic subdivisions (i.e., genetic distance) in the CFP varied considerably, indicating complexity (e.g., asymmetric gene flow or extinction/recolonization), but, in general, was higher than that due to geographic distance or recent human-related barriers. Because habitat-specific structure can enhance a species' adaptive potential and resilience to changing environments, these findings suggest the CFP may constitute an evolutionarily important portion of the range for coyotes and sympatric species exhibiting habitat-specific population structure.
Key Words: Canis latrans • gene flow • genetic diversity • genetic structure • habitat • isolation-by-distance
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Recent evolutionary diversification within taxa characterized by long-range movements presents challenges to traditional evolutionary models. In the absence of vicariant events, gene flow is expected to homogenize populations, especially those with broad habitat affinities, countering local differences in drift or selection that might otherwise drive populations apart. In lower mobility animals, populations sometimes remain faithful to ancient geological formations separated by contemporary landscape features, such as small streams or mountain ranges, that would pose insignificant barriers to larger, more wide-ranging species (Calsbeek et al. 2003
). Geographic distance alone also can provide a sufficient barrier for divergence among low-mobility populations, for example, evidenced by ring species (e.g., Irwin et al. 2001), but in the absence of discrete barriers is rarely sufficient to enable significant diversification in wide-ranging animals. Sympatric and parapatric modes of diversification therefore could be especially important in explaining Quaternary diversification of wide-ranging habitat-generalist species. In this study, we examine the role of a behavioral phenomenon as a diversifying agent in a wide-ranging carnivore.
It is well known that carnivores typically disperse long distances and often exhibit low levels of genetic divergence between distantly separated sampling sites (Lehman and Wayne 1991; Harrison 1992; Roy et al. 1994; Schwartz et al. 2002). Although these observations, gleaned primarily from coarse grain studies, clearly indicate that gene flow can homogenize carnivore populations across large portions of their range, they do not necessarily imply that carnivores cannot also exhibit strong population structure within certain parts of their range. Such high-structure portions of ranges could be evolutionarily important, especially if subdivisions correspond to different habitat types with correspondingly different selective regimes. For example, effects of random genetic drift are expected to be greater in subdivided populations, which can increase a species' adaptive potential by enabling individual populations to "explore the adaptive landscape" thereby reaching novel and varied fitness peaks (Wright 1931, 1932; Wade and Goodnight 1998). Habitat-specific subdivision also may be expected to facilitate the maintenance of functional genetic variability through differential local selection (Maynard-Smith and Hoekstra 1980; Hedrick 1986; Spichtig and Kawecki 2004). Such diversification among populations, whether due to drift or selection, might be expected, in turn, to lead species overall to be more resilient to environmental changes. But why expect highly mobile animals to exhibit population structure in certain parts of their range, and why expect this to correspond to habitat?
Recent evidence from behavioral studies indicates that many animals exhibit dispersal preferences toward habitat resembling their natal habitat (reviewed by Davis and Stamps 2004), a tendency, which, if strong enough, can be expected to translate to discretely subdivided populations along physically unobstructed habitat boundaries. Regions of high habitat heterogeneity, in turn, would be expected to correspond to high-structure portions of animal ranges. Examples of such habitat-specific subdivisions are accumulating (Geffen et al. 2004; Sacks et al. 2004; Pilot et al. 2006), as is evidence for conservative habitat selection behavior as a proximate cause (Sacks et al. 2005). For example, on or near the boundary between mountain and valley habitats of radiocollared coyotes, no interpack (i.e., between territory) relationships (i.e., relatedness >0.25) were observed across the boundary although several interpack relationships occurred within habitats, indicating that dispersal across the habitat boundary was rare (Sacks et al. 2005). The ecological and evolutionary significance of these small-scale studies, however, depends first on the spatial extent of such habitat-specific subdivisions and second on their strength relative to that of other diversifying factors, such as isolation-by-distance, which potentially swamp this locally important isolating mechanism over larger spatial extents. In this study, we investigated this question using genomic profiles of >2,000 coyotes from 2 broadly defined ecoregions of southwestern North America contrasting greatly in their levels of habitat heterogeneity: 1) the California Floristic Province (CFP) and 2) southwestern deserts and adjacent short grass prairie within the "Dry domain" (Bailey 1983), here referred to as the Desert–Prairie ecoregion (DPE).
The CFP is one of the most biologically diverse ecoregions in the world and, like other Mediterranean ecoregions (i.e., in Chile, South Africa, Australia, around the Mediterranean Sea), is known for its landscape heterogeneity and correspondingly high phylogeographic structure among low-mobility organisms (e.g., Calsbeek et al. 2003). The DPE contrasts with the sharply dissected landscape of the CFP most importantly in that it is composed of continuous tracts of gradually intergrading flora and fauna. Thus, habitat in the DPE is relatively homogeneous (at the coyote-relevant grain) within the typical dispersal radius of a coyote and dispersal is therefore predicted to be relatively unconstrained by natal-habitat–biased dispersal. Coyotes—as a species—are continuously distributed across both of these ecoregions in virtually every type of habitat (Young and Jackson 1951).
Here, we investigate 3 predictions about the contemporary population structure of coyotes stemming from the natal-habitat–biased dispersal hypothesis: 1) coyotes sampled from the CFP exhibit population structure concordant with habitat subregions, 2) coyotes from widely dispersed sampling sites within the DPE exhibit little or no structure, and 3) genetic diversity within CFP subregions is lower than in DPE subregions due to higher rates of random genetic drift in the former. We used 14 autosomal, biparentally inherited microsatellite loci and paternally inherited Y chromosome haplotypes (composed of 5 nonrecombining microsatellite loci) to test these predictions. We also used mitochondrial DNA (mtDNA) data from different individuals previously published by Lehman and Wayne (1991) as an independent check with respect to a maternally inherited marker.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Sampling
Coyote tissue samples, including muscle and dried blood of filter paper strips (Nobuto Blood Filter Strips, Advantec Manufacturing, Dublin, CA), were collected and mapped as described previously (Sacks et al. 2004
) from throughout California. An additional 30 coyotes (20 males and 10 females) were sampled from blood clots donated by Eric Gese from southeastern Colorado. The CFP sampling was concentrated in 8 bioregions (hereafter, "subregions"): CFP-Northwestern, CFP-northern Great Valley, CFP-Sierra Nevada, CFP-Cascades, CFP-Central Western (i.e., central California Coast ranges), CFP-southern Great Valley, CFP-Southwestern, and CFP-Transverse Ranges (Hickman 1993). 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 previously (Sacks et al. 2004). Coyotes were genotyped at 14 autosomal 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 chromosome loci (650-79.2A and 650-79.2B) were coamplified by the same pair of primers. As has been done previously for coamplified Y chromosome microsatellites (Malaspina et al. 1998), we assumed that the smaller of the 2 coamplified alleles corresponded to the "A" locus, whereas the larger of the 2 alleles corresponded to the "B" locus; where a single peak was noted, each locus was assigned the same allele. In this case, the procedure was likely as accurate as if we had amplified the loci separately because the overlap in allele size between the A and B loci involved a single allele. That is, the occurrence of single peaks only occurred at this size and there were no instances of co-occurrence in the same genotype of peaks both higher or both lower than this size. Polymerase chain reaction (PCR) conditions were described for autosomal loci elsewhere (Sacks et al. 2004) as were those for Y chromosome loci (Sundqvist et al. 2001; Bannasch et al. 2005). The PCR products were visualized and alleles scored using an ABI 3730 capillary DNA analyzer (Applied Biosystems, Foster City, CA) and STRand 2.2.30 software (Veterinary Genetics Laboratory, University of California, Davis, CA).
Autosomal Genotype-Based Analysis with Program Structure
The principle on which this method works is to cluster multilocus 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 running the algorithm 10 times each at K = 1–10 and plotting against K the average logarithm of the probability of the data given K [log Pr (X|K)], the geographic index (the average geographic distance between all pairs of coyotes within clusters divided by that between all coyotes in the entire sample; Sacks et al. 2004), and the consistency across runs (the percent of coyotes assigned to the same cluster in the 2 runs with the highest log Pr (X|K)). These initial runs included 10,000 Markov Chain Monte Carlo (MCMC) runs for burn-in (i.e., transition from starting conditions to stationarity) and 10,000 additional postburn-in runs. A final run at the selected K value was conducted using 30,000-cycle burn-in and 1 million–cycle postburn-in and used for the analysis. The admixture model was specified, allele frequencies were assumed to be correlated among genetic groupings (Falush et al. 2003), and default values were used for all other parameters. The analyses were performed in Structure (V 2.0; Pritchard et al. 2000). Locations were displayed using ArcView GIS (v 3.2) software (ESRI, Redlands, CA) color coded with respect to their cluster assignment, defined as the cluster corresponding to the highest estimated proportion of ancestry.
Autosomal Genotype-Based Analysis with Program Geneland
This method is similar to that implemented in the program Structure except that spatial proximity is used to weight genetic cluster assignments (Guillot, Estoup, et al. 2005; Guillot, Mortier, and Estoup 2005). Also, program Geneland enables the number of clusters (K) to be treated as a variable or to be fixed and, when varied, enables determination of the modal (i.e., most likely) value. We performed a preliminary run where K was allowed to vary from 1 to 30 to determine the modal number of clusters, which was 26. We then ran the analysis 5 times with the number of clusters fixed to 26. The number of clusters with individuals assigned to them varied from 8 to 10, with the majority of clusters having no genotypes assigned to them (normal for this program; Guillot, Estoup, et al. 2005; Guillot, Mortier, and Estoup 2005). All runs were conducted using the spatial D-model for the priors in allele frequency and 200,000 MCMC iterations, a maximum of 200 nuclei, and a 10,000-iteration postprocessing burn-in consistent with recommendations (Guillot, Estoup, et al. 2005; Guillot, Mortier, and Estoup 2005).
Allele Frequency-Based Analyses
These analyses were conducted using autosomal microsatellites, Y chromosome haplotypes (composed of 5 haploid microsatellite alleles), and the mtDNA haplotypes of Lehman and Wayne (1991). For autosomal and Y chromosome analyses, we grouped coyotes into 45 samples according to spatial proximity (i.e., sites), of which 39 were used for autosomal analyses (n > 14 coyotes each, average n = 45) and 36 were used for Y chromosome analyses (>8 male coyotes, average n = 24). Fifteen similarly located sites identified by Lehman and Wayne (1991; table 1, sites d–q, s; average n = 12.4 coyotes) were used in analyses with their mtDNA restriction fragment length polymorphism data (table 3). Specifically, mtDNA data were grouped as follows: sampling sites d, o, q, s (DPE); e, j, l (CFP-southern Great Valley); h (CFP-northern Great Valley); i, k (CFP-Central Western); 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 correlations between genetic distance and geographic distance using Mantel tests (Mantel 1967
) performed in Arlequin (v2.000; Schneider et al. 2000) with significance based on 1,000 matrix permutations. The dependent variable matrix was of genetic distance (expressed as 
Slatkin 1995) calculated in Arlequin (v2.000). The independent variable matrix was of geographic distance (kilometer) between sample sites (i.e., centroids), calculated as the square root of the sum of squared differences in centroid X-coordinates and centroid Y-coordinates. The centroid was calculated as average coordinates among coyotes in each site; coordinates were projected in Teale Albers before calculating centroids and distances using ArcView GIS software (v 3.2; ESRI). Geographic distances associated with some pairs of sampling sites were calculated from the sum of 2 or 3 of these straight-line distances to avoid cutting across the San Francisco Bay Estuary (including San Pablo Bay and delta), which was assumed to completely obstruct dispersal (Sacks et al. 2004).
Nested hierarchical analysis of molecular variance (AMOVA; Excoffier et al. 1992) was conducted for each marker type to assess the proportion of between-site genetic variance between subregions relative to that within subregions, using sampling sites as replicates. The AMOVA was performed in Arlequin (v2.000) using 1,000 permutations to test for significance. Because our prediction was that the DPE would not contain substructure and in this sense be analogous to a subregion of the CFP, we treated this ecoregion on the same hierarchical level as subregions for this analysis.
Within-Subregion Genetic Diversity
To compare genetic diversity between CFP and DPE subregions, we used the sampling sites as the sample units, thereby minimizing the potential for confounding effects of population substructure. Because the number of coyotes varied somewhat from site to site and could affect estimates of genetic diversity, we included sample size as a covariate in the analysis. To maximize statistical power, we performed a single analysis using all marker types and included an interaction term for marker type and ecoregion to test for the possibility that one or both of the uniparentally inherited markers showed a larger difference, as might be expected due to their smaller effective population sizes. We used as our measure of genetic diversity
 |
). For the autosomal microsatellite loci, this measure was equivalent to expected heterozygosity. We employed a general linear model in SYSTAT (v9.0; SPSS Incorporated, Chicago, IL) with the following variables: ecoregion (CFP, DPE), marker type (autosomal, mtDNA, and Y chromosome), ecoregion*marker type, and, as a covariate, sample size (number of coyotes in site).
Hardy–Weinberg and Linkage Equilibrium
For the autosomal markers, we tested Hardy–Weinberg equilibrium (Guo and Thompson 1992) in Arlequin (v2.000) and linkage disequilibrium in Genepop (v3.3; Raymond and Rousset 1995) within sampling sites. We used a sequential Bonferroni test to determine significance of deviations from Hardy–Weinberg and linkage equilibrium (Rice 1989).
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
A total of 301 alleles was identified among the 14 autosomal microsatellite loci, with the number of alleles ranging from 8 to 48 per locus. We identified a total of 73 Y chromosome haplotypes composed of the 5 microsatellite loci. These loci contained 34 alleles in total, with numbers ranging from 4 to 14 alleles per locus. A total of 32 mtDNA haplotypes from the Lehman and Wayne (1991)
study was identified in the subset of data used in the present study.
Structure Analysis
Our first approach, assigning individuals to genetic clusters based on the autosomal microsatellite genotypes irrespective of geography, clearly supported predictions 1 (CFP population is structured according to habitat boundaries) and 2 (DPE population is relatively unstructured) (fig. 1). We chose K = 6 for the final run in this analysis because both the geographic index and log Pr(X|K) plateau here (as minimum and maximum, respectively) and higher values of K were unstable in terms of variance in log Pr(X|K) and/or consistency of assignments among runs (fig. 2). The analysis revealed substantial subregion-specific genetic structure within the CFP and none in the DPE despite the vastly greater spatial extent of the latter. These results were highly consistent across runs and robust to numbers of clusters assumed. For example, at K = 8, DPE coyotes clustered together in a single cluster, similarly to K = 6, but the Central Western and Southwestern subregions were each divided into 2 clusters. Genetic subdivisions clearly corresponded to subregion boundaries and could not be explained by physical barriers associated with recent anthropogenic habitat modifications. For example, the entire Southwestern subregion clustered together despite a tremendously fragmented landscape and numerous highways with high volume traffic that currently pose barriers to gene flow (Hunter et al. 2003; Riley et al. 2006).
|
|
Geneland Analysis
Similarly to the Structure analysis, several distinct clusters corresponding to subregions were identified within the CFP, whereas all coyotes from the DPE-Modoc Plateau and the DPE-Short Grass Prairie were clustered together despite up to 1,500 km separating them. However, DPE-Mojave Desert and DPE-Sonoran Desert coyotes were clustered with these other DPE coyotes in only 3 of the 5 runs, with the other 2 runs grouping most DPE-Mojave Desert and DPE-Sonoran Desert coyotes with the CFP-southern Great Valley cluster. Interestingly, the CFP-southern Great Valley subregion is the most similar of the CFP subregions to the DPE-Mojave Desert and DPE-Sonoran Desert subregions with respect to vegetation and climate (Hickman 1993
). Several clear boundaries also were identified between subregions within the CFP in each of the runs (fig. 3), although the particular boundaries corresponding to genetic clusters varied somewhat across runs; this instability likely was due to the nonrandom distribution of sampling locations (Guillot, Estoup, et al. 2005; Guillot, Mortier, and Estoup 2005). Nevertheless, the Geneland runs consistently lumped most DPE coyotes together and identified separate clusters corresponding to CFP subregions despite the method's predisposition to cluster spatially proximate samples together and spatially distant samples apart.
|
Allele Frequency-Based Analyses
Genetic diversity was similar for CFP ecoregion and DPE as wholes (i.e., data pooled across sites and subregions) for autosomal microsatellites (0.80, 0.79, respectively), Y chromosome microsatellites (0.94, 0.91), and mtDNA markers (0.88, 0.87). As with Structure and Geneland analyses, pairwise genetic distances based on allele frequencies likewise supported predictions 1 and 2. There was a weak relationship between genetic distance and geographic distance based on the autosomal (Mantel test r2 = 0.08; P = 0.008) and Y chromosome (r2 = 0.04; P = 0.047) markers and no significant relationship based on mtDNA markers (r2 = 0.00; P = 0.32). However, the AMOVA analysis indicated that any isolation-by-distance relationship was dwarfed by the amount of variance explained by subregions; whereas geographic distance explained 0–8% (depending on marker type) of the between-site genetic variance (indicated by r2 values above), subregions explained 44–70% of between-site genetic variance (table 1). Genetic distances between sites in different subregions within the CFP were highly variable and higher on average than those in the DPE, which were uniformly low despite their greater (on average) separation distances (fig. 4). Within subregions of the CFP, genetic distances were low similarly to those over the entire DPE. While sampling and estimation error contribute to the observed variance, these sources of error cannot explain the difference in variance in between- versus within-subregion (and the DPE) comparisons (fig. 4). Thus, the high variance in genetic distances between different CFP subregions, even at short geographic distances, suggests that subregion boundaries varied considerably with respect to their permeability to gene flow.
|
Within-Site Genetic Diversity
Also as predicted (prediction 3)—and in contrast to ecoregion-wide genetic diversity—within-site genetic diversity was lower in the CFP than in the DPE (F1,71 = 7.0, P = 0.01; fig. 5). Although differences were somewhat greater in uniparentally inherited markers, consistent with their smaller effective population sizes, there was no significant interaction between marker type and ecoregion (F2,71 = 0.87, P = 0.42). Number of coyotes was not a significant covariate in the model (F1,71 = 0.42, P = 0.52).
|
Hardy–Weinberg and Linkage Equilibrium
Loci were significantly out of Hardy–Weinberg equilibrium in 66 instances (out of 616 population locus cases), but these were distributed widely among populations and loci. There was, however, one systematic relationship, a positive correlation between proportion of loci significantly out of Hardy–Weinberg equilibrium and sample size (r = 0.67), to be expected due to the greater sensitivity (higher statistical power) of the tests employed on larger samples. The average FIS (based on the heterozygote deficiency) was 0.09, which is similar to previous estimates for coyotes (Roy et al. 1994
; Sacks et al. 2004). There were 198 instances of locus pairs in linkage disequilibrium out of a total 4,186 possible, but these were not clustered at any particular locus pair. However, there were 2 sampling sites that had much higher proportions of locus pairs in linkage disequilibrium (32%, 47%). Interestingly, these were both pairs that straddled the Sierra Nevada and southern Great Valley subregions. Thus, although Structure and Geneland analyses did not illuminate a genetic subdivision along this boundary, the high degree of linkage disequilibrium observed in these 2 sites nonetheless suggests the possible presence of substructure along this boundary. Excluding the 2 outlier populations, as with Hardy–Weinberg equilibrium, there was a positive correlation between proportion of locus pairs significantly out of linkage equilibrium and sample size (r = 0.55). |
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
We investigated 3 predictions of the natal-habitat–biased dispersal hypothesis with respect to the contemporary population structure of coyotes in a Mediterranean ecoregion: 1) coyotes sampled from the CFP exhibit population structure concordant with habitat subregions, 2) coyotes from widely dispersed sampling sites within the neighboring DPE exhibit little or no structure, and 3) genetic diversity within CFP sites is lower than in DPE sites due to higher rates of random genetic drift in the former. Our findings clearly supported each of these predictions based on autosomal, Y chromosome, and mtDNA markers. Although we had previously observed, based on microsatellites, that coyote population subdivisions were concordant with habitats in central California (Sacks et al. 2004
, 2005), the present study, based on a 4-fold increase in the number of coyotes and a much larger study area, demonstrated that the phenomenon was present throughout the CFP and, importantly, absent from the DPE. The occurrence of a few habitat-specific 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 we are aware to show extensive habitat-specific substructuring in a continuously distributed species throughout a heterogeneous ecoregion and in direct contrast to low genetic divergence between distantly sampled sites in an adjacent, relatively homogeneous ecoregion.
Several aspects of this study serve to strengthen the inference that natal-habitat–biased dispersal was the primary cause of the population structure observed in the CFP. First, the hypothesis and predictions were formulated a priori based on individual-level criteria and preliminarily tested on smaller scales (Sacks et al. 2004, 2005). The expansion to the entirety of the CFP, in particular the additional habitat subdivisions, in the current study constituted replication. Finally, the DPE essentially served as a negative control, providing a reference in the case where no sharply differentiated habitat types occurred. The high genetic diversity within sites of this ecoregion indicated that the genetic homogeneity among sites of the DPE was due to high contemporary rates of gene flow rather than a recent population expansion. Thus, the genetic homogeneity associated with this vast ecoregion served to emphasize the genetic discontinuities within the heterogeneous CFP. In fact, the magnitude of genetic divergence at habitat boundaries in the latter ecoregion was, on average, considerably greater than that due to geographic distance over a substantial portion of the species range, suggesting that habitat-specific population subdivisions could be of fundamental significance.
Although it is difficult to explain the observed population genetic structure without invoking natal-habitat–biased dispersal as the primary cause within the CFP, a brief consideration of additional factors is warranted. Other processes undoubtedly contributed to various extents on different spatiotemporal scales to the observed population structure, including historical patterns associated with Pleistocene/Holocene climatic change (Hewitt 1996) as well as contemporary processes such as habitat-specific demography, habitat selection asymmetries, and metapopulation dynamics. For example, despite the high within-site genetic diversity observed in the DPE, which indicates high levels of contemporary gene flow, it remains possible that the population stems from a post-Pleistocene range expansion (but sufficiently in the past for gene flow and drift to have equilibrated). Further, the subdivision between the CFP and DPE could reflect secondary contact between previously isolated populations associated with inhospitable habitat during much of the Pleistocene along the crest of the Sierra Nevada Mountains, as appears to be the case for many other taxa (Remington 1968; Swenson and Howard 2004). Although coyotes currently occur throughout the Sierra Nevada range, this subregion may contain a zone of post-Pleistocene admixture. Extensions of Great Basin Desert habitat along the eastern Sierra Nevada corresponded to the Desert-Prairie genetic cluster, whereas coyotes along the western Sierra Nevada clustered separately (with the Great Valley subregions), consistent with a subdivision along the crest (e.g., fig. 1). On the other hand, in South Lake Tahoe, our only eastern Sierra Nevada sampling site with typical high Sierra Nevada habitat (e.g., coniferous forest and alpine meadow), coyotes clustered with their western Sierra Nevada counterparts, supporting a sharp subdivision approximately on the eastern edge of this subregion (e.g., corresponding to the meeting of Great Basin and Subalpine vegetation types), as opposed to the crest of the Sierra Nevada per se. Phylogenetic analyses of Y chromosome and rapidly mutating regions of the mtDNA genome are needed to investigate and compare the historical demography of coyote populations in both ecoregions to assess these phylogeographic scenarios.
Regardless of the possible role of a Pleistocene barrier in the Sierra Nevada, a similar explanation seems unlikely to account for the other habitat-specific population subdivisions. No such barriers are known to have occurred along borders of most subregions in the CFP, nor are they suggested by the presence of multispecies hybrid zones between populations of late Pleistocene age (Remington 1968; Swenson and Howard 2004). Genetic subdivisions corresponded to nearly all subregional boundaries within the CFP, although the magnitude varied, apparently strongest between coastal and interior subregions and weakest between the Sierra Nevada and Great Valley subregions. Although we did not detect a clear subdivision separating the Sierra Nevada from the Great Valley in the present study, previous analyses conducted on smaller spatial scales indicated clear subdivisions separating the western Sierra Nevada from the northern Great Valley (Sacks et al. 2004, 2005). Further, the high level of linkage disequilibrium observed within sampling sites near the Sierra Nevada boundary with the south Great Valley in this study suggests the possibility of genetic subdivision in the southern portion of these subregions as well.
Our primary focus in this study was on identification of boundaries between coyote subpopulations. However, a full understanding of the role of habitat boundaries in determining population dynamics will require an understanding of fine-scale social processes, which can reinforce subdivisions, as well as asymmetries in demographic processes and habitat selection proclivities (apart from natal habitat), which can be expected to affect the evolutionary trajectory of the metapopulation. As might be expected, fine-grained studies indicate that relatedness among packs (i.e., family groups) in neighboring territories tends to be considerably greater on the same side than on opposite sides of a habitat or physical subdivision (Sacks et al. 2005; Riley et al. 2006). To the extent that dispersal is facilitated by settlement adjacent to relatives (Temeles 1994), social fabric could further deter dispersal across genetic subdivisions (Sacks et al. 2005). Where boundaries are hard, an increase in density due to territory packing may have a similar effect irrespective of familial relationships among neighboring packs (Riley et al. 2006). In addition to social factors, differential demography and asymmetric dispersal between habitats could influence the particular patterns of subdivision (Stenseth et al. 2004). Habitats associated with subregions also may vary in quality with some serving as source populations and others as sinks (Pulliam 1988). Previously, Sacks et al. (2005) found gene flow to be considerably higher from the Cascade Mountains to the northern Great Valley than in the opposite direction in one local study site, consistent with differential habitat quality, although it was unknown to what extent the asymmetry in gene flow was driven by demographic versus behavioral factors. In the present study, the high variability in genetic divergence between nearby subregions of the CFP indicated that patterns of gene flow may be more complex than is indicated by the simple pattern of subdivision apparent in from the Structure and Geneland analyses. These details, which could include extinction/recolonization, source/sink, and other metapopulation dynamics, should be investigated in future studies.
Regardless of the mix of factors that ultimately determine and refine patterns of population genetic structure, the predominance of enduring habitat-specific population subdivisions, likely due to natal-habitat–biased dispersal, could have important evolutionary implications on both micro- and macroevolutionary timescales. For example, one of the fundamental problems of microevolution pertains to the maintenance of genetic diversity in natural populations. Drift-based hypotheses emphasize the role of population genetic subdivisions enabling small, partly isolated subpopulations to "explore" and find alternative fitness peaks on the adaptive landscape by freeing the genomes from strong selection for universally adaptive phenotypes (e.g., Wright 1931, 1932), whereas selection-based theories emphasize the role of habitat heterogeneity in causing differential selection in different localities (Levene 1953; Gillespe 1978; Gillespe and Turelli 1989; Spichtig and Kawecki 2004). Despite their disparate roots in the literature, drift-based and selection-based frameworks are not mutually exclusive. Indeed, population structure corresponding to habitat subdivisions can be expected to increase the occurrence of balanced polymorphism because a tendency to mate within the habitat in which an individual is born results in a correlation between the habitat individuals select and that in which they are most fit, which should increase the influence of local selection relative to between-habitat selection (Taylor 1976; Jones 1980; Jones and Probert 1980; Garcia-Dorado 1986; Hedrick 1986). The level of subdivision observed in the present study is in the range Wright (1931, 1932) reasoned, and Wade and Goodnight (1991) showed experimentally, would facilitate drift-based evolution. The observation that genetic diversity within sites of the CFP was less than in the DPE is also consistent with higher effects of genetic drift within the more subdivided portion of the range. Moreover, although we did not investigate selection in this study, habitat-specific differences in selective pressures seem likely and would be expected to further promote divergence. For example, a recent empirical study indicated that local habitat-specific adaptation was sustained in the presence of as much as 13% immigration (Postma and van Noordwijk 2005).
On the macroevolutionary end, natal-habitat–biased dispersal conceivably provides a powerful reinforcing mechanism for reproductive isolation and parapatric speciation. The ability of populations to diverge evolutionarily in the presence of incomplete reproductive isolation is an exciting area of research that has received renewed interest in recent years (Bolnick et al. 2003). The conditions typically listed for sympatric or parapatric divergence include niche diversification, involving some form of disruptive selection and assortative mating. Species with broad niches (i.e., generalists) may be ideal candidates for the study of sympatric/parapatric niche diversification, especially, when they are composed by individuals with narrow niches (i.e., specialists), but which vary widely among individuals (Roughgarden 1972; Bolnick et al. 2003). Potential for reproductive isolation would seem to be greatest where niche diversification is concordant with spatial segregation, as is the case with natal-habitat–biased dispersal.
Although we do not suggest that contemporary coyote subpopulations in the CFP are incipient species, the phenomenon observed in this study could have contributed to speciation at other times or in other taxa in this and other Mediterranean ecoregions. The CFP, and Mediterranean ecoregions generally, are hotbeds of endemism and biodiversity, characterized by high beta diversity. Although this pattern is thought to stem partly from rapid phylogenetic diversification associated with past geologic and hydrologic isolating events (Myers et al. 2000; Latimer et al. 2005)—supported within the CFP by a high correspondence in the phylogeography and temporal distribution of low-mobility species (e.g., Maldonado et al. 2001; Calsbeek et al. 2003; Spinks and Shaffer 2005)—the role of natal-habitat–biased dispersal and, consequently, parapatric speciation also could have played a role in generating contemporary patterns of biodiversity. This seems especially likely given the relatively stable climate associated with the CFP throughout the Pleistocene glacial cycles, providing not only an ice-free refugium for western fauna and flora but also long-term habitat heterogeneity. Thus, it is intriguing to ponder the extent to which natal-habitat–biased dispersal might have affected the evolution of vertebrates in this and similar regions globally. Given the high degree of population structuring observed in the wide-ranging coyote, it seems likely that natal-habitat–biased dispersal could be even more influential for less wide-ranging generalist species in the ecoregion. Indeed, behavioral studies indicate the presence of natal-habitat–biased dispersal in a large number of small vertebrate taxa (Davis and Stamps 2004). At present, investigations of the population genetic consequences of this phenomenon in smaller vertebrates are lacking, providing an exciting and wide-open area in need of future research.
|
Conclusions |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Our results, based on biparentally, paternally, and maternally inherited markers, clearly demonstrate substantial population genetic structure in coyotes of the CFP in contrast to low genetic divergence among sampling sites spanning a large area of the desert and prairie. Moreover, the associated reduction in genetic diversity observed within sites of the CFP indicates that structuring in that region was sufficient for genetic drift to leave a visible mark on allele frequencies. Although historical contingency and metapopulation dynamics could contribute to the observed patterns to some extent, these factors alone seem insufficient to explain them without also invoking natal-habitat–biased dispersal as a fundamental cause. Thus, although coyotes commonly disperse distances of 100 km or more (Harrison 1992
; Sacks et al. 2005) and, as a species, occupy virtually every habitat type (Young and Jackson 1951), their propensity, as individuals, to disperse into habitat similar to their natal habitats was apparently of sufficient strength to structure their populations. Given the highly vagile nature and catholic habitat preferences of our study species, these findings also are likely applicable to a broad spectrum of terrestrial vertebrates most of which are less mobile and less continuously distributed to begin with. If so, heterogeneous landscapes such as that of the CFP, which have also acted as refuges during glacial maxima, could play an important evolutionary role by providing high-structure portions of animal ranges, accelerating evolutionary divergence, for example, during glacial cycles. Because habitat-specific structure can enhance a species' adaptive potential and resilience to changing environments, these findings further suggest the CFP may constitute an evolutionarily important portion of the range for coyotes and sympatric species exhibiting habitat-specific population structure.
|
Acknowledgements |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
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 many to 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. Supported financially by the University of California (UC) Davis, School of Veterinary Medicine, Veterinary Genetics Laboratory; by grants from the UC Davis Genetic Resources Conservation Program; and, logistically, by the USDA/Animal Plant Health Inspection Service (APHIS)/Wildlife Services, Santa Clara County Vector Control District, and California Department of Health Services.
|
Footnotes |
|---|
Connie Mulligan, Associate Editor
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Bailey RG. Delineation of ecosystem regions. Environ Manage (1983) 7:365–373.
Bannasch DL, Bannasch MJ, Ryun JR, Famula TR, Pedersen NC. Y chromosome haplotype analysis in purebred dogs. Mamm Genome (2005) 16:273–280.[CrossRef][Web of Science][Medline]
Bolnick DI, Svanback R, Fordyce JA, Yang LH, Davis JM, Hulsey CD, Forister ML. The ecology of individuals: incidence and implications of individual specialization. Am Nat (2003) 161:1–28.[CrossRef][Web of Science][Medline]
Calsbeek R, Thompson JN, Richardson JE. Patterns of molecular evolution and diversification in a biodiversity hotspot: the California Floristic Province. Mol Ecol (2003) 12:1021–1029.[CrossRef][Medline]
Carmichael LE, Nagy JA, Larter NC, Strobeck C. Prey specialization may influence patterns of gene flow in wolves of the Canadian Northwest. Mol Ecol (2001) 10:2787–2798.[Medline]
Davis JM, Stamps JA. The effect of natal experience on habitat preferences. Trends Ecol Evol (2004) 19:411–416.[CrossRef][Medline]
Excoffier L, Smouse P, Quattro J. Analysis of molecular variance inferred from metric distances among DNA haplotypes application to human mitochondrial DNA restriction data. Genetics (1992) 131:479–491.[Abstract]
Falush D, Stephens M, Pritchard JK. Inference of population structure using multilocus genotype data: linked loci and correlated allele frequencies. Genetics (2003) 164:1567–1587.
Garcia-Dorado A. The effect of niche preference on polymorphism protection in a heterogeneous environment. Evolution (1986) 40:936–945.[CrossRef][Web of Science]
Geffen E, Anderson MJ, Wayne RK. Climate and habitat barriers to dispersal in the highly mobile grey wolf. Mol Ecol (2004) 13:2481–2490.[CrossRef][Medline]
Gillespe JH. A general model to account for enzyme variation in natural populations. V. The SAS-CFF model. Theor Popul Biol (1978) 14:1–45.[CrossRef][Web of Science][Medline]
Gillespe JH, Turelli M. Genotype-environment interactions and the maintenance of polygenic variation. Genetics (1989) 121:129–138.
Guillot G, Estoup A, Mortier F, Cosson JF. A spatial statistical model for landscape genetics. Genetics (2005) 170:1261–1280.
Guillot G, Mortier F, Estoup A. GENELAND: a computer package for landscape genetics. Mol Ecol Notes (2005) 5:712–715.[CrossRef][Web of Science]
Guo SW, Thompson EA. Performing the exact test of Hardy-Weinberg proportion for multiple alleles. Biometrics (1992) 48:361–372.[CrossRef][Web of Science][Medline]
Harrison DJ. Dispersal characteristics of juvenile coyotes in Maine. J Wildl Manage (1992) 56:128–138.[CrossRef]
Hedrick PW. Genetic polymorphism in heterogeneous environments: a decade later. Annu Rev Ecol Syst (1986) 17:535–566.[CrossRef][Web of Science]
Hewitt GM. Some genetic consequences of ice ages, and their role in divergence and speciation. Biol J Linn Soc (1996) 58:247–276.[CrossRef][Web of Science]
Hickman JC. The Jepson manual: higher plants of California (1993) Berkeley (CA): University of California Press.
Hunter RD, Fisher RN, Crooks KR. Landscape-level connectivity in coastal southern California, USA, as assessed through carnivore habitat suitability. Nat Area J (2003) 23:302–314.
Irwin DE, Irwin JH, Price TD. Ring species as bridges between microevolution and speciation. Genetica (2001) 112–113:223–243.
Jones JS. Can genes choose habitats? Nature (1980) 286:757–758.[CrossRef][Medline]
Jones JS, Probert RF. Habitat selection maintains a deleterious allele in a heterogeneous environment. Nature (1980) 287:632–633.[CrossRef]
Latimer AM, Silander JA, Cowling RM. Neutral ecological theory reveals isolation and rapid speciation in a biodiversity hotspot. Science (2005) 309:1722–1725.
Lehman N, Wayne RK. Analysis of coyote mitochondrial DNA genotype frequencies: estimation of the effective number of alleles. Genetics (1991) 128:405–416.[Abstract]
Levene H. Genetic equilibrium when more than one ecological niche is available. Am Nat (1953) 87:331–333.[CrossRef][Web of Science]
Malaspina P, Cruciani F, Ciminelli BM, et al. Network analyses of Y-chromosomal types in Europe. North Africa: and Western Asia reveal specific patterns of geographic distribution. Am J Hum Genet (1998) 63:847–860.[CrossRef][Web of Science][Medline]
Maldonado JE, Vila C, Wayne RK. Tripartite genetic subdivisions in the ornate shrew (Sorex ornatus). Mol Ecol (2001) 10:127–147.[CrossRef][Medline]
Mantel N. The detection of disease clustering and a generalized regression approach. Cancer Res (1967) 27:209–220.
Maynard-Smith J, Hoekstra R. Polymorphism in a varied environment: how robust are the models? Genet Res (1980) 35:47–57.
Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GAB, Kent J. Biodiversity hotspots for conservation priorities. Nature (2000) 403:853–858.[CrossRef][Medline]
Pritchard JK, Stephens M, Donnelly P. Inference of population structure using multilocus genotype data. Genetics (2000) 155:945–959.
Pilot M, Jedrzejewski W, Branicki W, Sidorovich VE, Jedrzejewska B, Stachura K, Funk SM. Ecological factors influence population genetic structure of European gray wolves. Molec Ecol (2006) 15:4533–4553.[CrossRef][Medline]
Postma E, van Noordwijk J. Gene flow maintains large genetic difference in clutch size at a small spatial scale. Nature (2005) 433:65–68.[CrossRef][Medline]
Pulliam HR. Sources, sinks, and population regulation. Am Nat (1988) 132:653–661.
Raymond M, Rousset F. GENEPOP (Version 1.2): population genetics software for exact tests and ecumenicism. J Hered (1995) 86:248–249.
Remington CL. Suture zones of hybrid interaction between recently joined biotas. Evol Biol (1968) 2:321–428.
Rice WR. Analyzing tables of statistical tests. Evolution (1989) 43:223–225.[CrossRef][Web of Science]
Riley SPD, Pollinger JP, Sauvajot RM, York EC, Bromley C, Fuller TK, Wayne RK. A southern California freeway is a physical and social barrier to gene flow in carnivores. Mol Ecol (2006) 15:1733–1741.[CrossRef][Medline]
Roughgarden J. Evolution of niche width. Am Nat (1972) 106:683–718.[CrossRef][Web of Science]
Roy MS, Geffen E, Smith D, Ostrander EA, Wayne RK. Patterns of differentiation and hybridization in North American wolflike canids, revealed by analysis of microsatellite loci. Mol Biol Evol (1994) 11:553–570.[Abstract]
Sacks BN, Brown SK, Ernest HB. Population structure of California coyotes corresponds to habitat-specific breaks and illuminates species history. Mol Ecol (2004) 13:1265–1275.[CrossRef][Medline]
Sacks BN, Mitchell BR, Williams CL, Ernest HB. Coyote movements and social structure along a cryptic population genetic subdivision. Mol Ecol (2005) 14:1241–1249.[CrossRef][Medline]
Schneider S, Roessli D, Excoffier L. Arlequin: a software for population genetics data analysis (2000) Geneva (Switzerland): Genetics of Biometry Laboratory. University of Geneva.
Schwartz MK, Mills LS, McKelvey KS, Ruggiero LF, Allendorf FW. DNA reveals high dispersal synchronizing the population dynamics of Canada lynx. Nature (2002) 415:520–522.[CrossRef][Medline]
Slatkin M. A measure of population subdivision based on microsatellite allele frequencies. Genetics (1995) 139:457–462.[Web of Science][Medline]
Spichtig M, Kawecki TJ. The maintenance (or not) of polygenic variation by soft selection in heterogeneous environments. Am Nat (2004) 164:70–84.[CrossRef][Web of Science][Medline]
Spinks PQ, Shaffer HB. Range-wide molecular analysis of the western pond turtle (Emys marmorata): cryptic variation, isolation by distance, and their conservation implications. Mol Ecol (2005) 14:2047–2064.[CrossRef][Medline]
Stenseth NC, Ehrich D, Rueness EK, et al. The effect of climatic forcing on population synchrony and genetic structuring of the Canadian lynx. Proc Natl Acad Sci USA (2004) 101:6056–6061.
Sundqvist AK, Ellegren H, Olivier M, Vila C. Y chromosome haplotyping in Scandanavian wolves (Canis lupus) based on microsatellite markers. Mol Ecol (2001) 10:1959–1966.[CrossRef][Medline]
Swenson NG, Howard DJ. Do suture zones exist? Evolution (2004) 11:2391–2397.
Taylor CE. Genetic variation in heterogeneous environments. Genetics (1976) 83:887–894.
Temeles EJ. The role of neighbors in territorial systems: when are they "dear enemies. Anim Behav (1994) 47:339–350.[CrossRef][Web of Science]
Wade MJ, Goodnight CJ. Wright's shifting balance theory: an experimental study. Science (1991) 253:1015–1018.
Wade MJ, Goodnight CJ. Perspective: the theories of Fisher and Wright in the context of metapopulations: when nature does many small experiments. Evolution (1998) 52:1537–1553.[CrossRef][Web of Science]
Weir BS. Genetic data analysis (1996) 2nd ed. Sunderland (MA): Sinauer.
Wright S. Evolution in Mendelian populations. Genetics (1931) 16:97–159.
Wright S. The roles of mutation, inbreeding, cross-breeding, and selection in Evolution. Proc 6th Int Congr Genet (1932) 1:356–366.
Young SP, Jackson SP. The clever coyote (1951) Lincoln (NE): University of Nebraska Press.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||