Accurate delineation of lineage diversity is increasingly important, as species distributions are becoming more reduced and threatened. During the last century, the subspecies category was often used to denote phenotypic variation within a species range and to provide a framework for understanding lineage differentiation, often considered incipient speciation. While this category has largely fallen into disuse, previously recognized subspecies often serve as important units for conservation policy and management when other information is lacking. In this study, we evaluated phenotypic subspecies hypotheses within shovel-nosed snakes on the basis of genetic data and considered how evolutionary processes such as gene flow influenced possible incongruence between phenotypic and genetic patterns. We used both traditional phylogenetic and Bayesian clustering analyses to infer range-wide genetic structure and spatially explicit analyses to detect possible boundary locations of lineage contact. Multilocus analyses supported three historically isolated groups with low to moderate levels of contemporary gene exchange. Genetic data did not support phenotypic subspecies as exclusive groups, and we detected patterns of discordance in areas where three subspecies are presumed to be in contact. Based on genetic and phenotypic evidence, we suggested that species-level diversity is underestimated in this group and we proposed that two species be recognized, Chionactis occipitalis and C. annulata. In addition, we recommend retention of two subspecific designations within C. annulata (C. a. annulata and C. a. klauberi) that reflect regional shifts in both genetic and phenotypic variation within the species. Our results highlight the difficultly in validating taxonomic boundaries within lineages that are evolving under a time-dependent, continuous process.
Citation: Wood DA, Fisher RN, Vandergast AG (2014) Fuzzy Boundaries: Color and Gene Flow Patterns among Parapatric Lineages of the Western Shovel-Nosed Snake and Taxonomic Implication. PLoS ONE 9(5): e97494. https://doi.org/10.1371/journal.pone.0097494
Editor: Wolfgang Arthofer, University of Innsbruck, Austria
Received: January 16, 2014; Accepted: April 16, 2014; Published: May 21, 2014
This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Funding: This project was funded by the Science Support Partnership between USGS and USFWS, Arizona Game and Fish Department, and by the US Geological Survey, Western Ecological Research Center. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Although scientific debates over species concepts persist, the notion that species exist as separately evolving lineages, presumably under various stages of speciation, is central to discussions related to intraspecific diversity. While most contemporary species concepts agree conceptually that species are separately evolving “metapopulation lineages” (i.e., the general lineage concept of a species; de Queiroz ) the challenge lies in determining where a species should be recognized within this continuum. Intraspecific diversity garnered early interest in the field of evolutionary biology, as Darwin  makes clear: “those forms which possess in some considerable degree the character of species, but which are so closely similar to some other forms, or are so closely linked to them by intermediate gradations, that naturalists do not like to rank them as distinct species, are in several respects the most important for us.” This sentiment extended through the 20th century, where researchers often used the subspecific rank to fulfill two roles: (1) to denote the phenotypic variation within a species range, and (2) to provide a framework for understanding the heirarchical levels of lineage differentiation, often considered incipient speciation –. During the late 20th century this either-or approach to subspecies delimitation unleashed confusion on the concept of species and sparked critisism of the subspecies rank. As a result, the subspecies rank has largely fallen into disuse. However, within the field of conservation biology, historically described subspecies often serve as important units for conservation policy and management when other information is lacking –. Therefore, accurate delineation of lineage diversity and population genetic structure remains an important task, as species distributions are becoming ever more reduced and threatened –.
Criticism of the subspecies rank itself has helped to bring modern statistical and genetic techniques to bear on the issue of whether recognized subspecies form natural groups. Phylogenetic methods and molecular characters are often used to identify genetic lineages and evaluate subspecies, often through the use of single locus markers (e.g. mitochondrial DNA). While numerous studies have demonstrated that phenotypic subspecies often fail to correspond to monophyletic genetic lineages, the generality of this pattern across taxa remains controversial , , –. Furthermore, several authors have suggested that the presence or absence of monophyly is too stringent of a criterion –, because introgression and incomplete lineage sorting between diverging populations can lead to incongruent gene trees at the species-subspecies boundary , . Therefore, using criteria that only emphasize historical isolation based on gene trees from single, neutrally evolving loci ignores other factors contributing to phenotypic differences that may be important for conservation. For example, local adaptation may contribute to genetic divergence, but the divergence may be so recent that insufficient time has passed to achieve reciprocally monophyletic gene trees , .
Multi-locus genetic surveys that examine larger portions of the genome have garnered broad support as a preferred strategy for studying lineage divergence and genetic differentiation , , . Clustering-based approaches that define genetic clusters using multi-locus datasets can be used to evaluate units at the species-subspecies boundary , . These methods incorporate a Bayesian framework to probabilistically cluster individuals into the most likely number of groups given the data, which can then be evaluated against how well they correspond to phenotypic subspecies circumscriptions.
In this study, we investigate the species-subspecies boundaries of Western shovel-nosed snake (Chionactis occipitalis) and evaluate whether the phenotypic subspecies correspond to genetic groups on the basis of multiple selectively neutral loci. Chionactis occipitalis is a small colubrid snake occupying the arid valley floors and bajadas of the Mojave and Sonoran deserts in southwestern North America. Four subspecies have been recognized on the basis of variation in banding pattern and coloration throughout the distribution – and are recognized as follows: C. o. occipitalis (Mojave shovel-nosed snake), C. o. annulata (Colorado shovel-nosed snake), C. o. talpina (Nevada shovel-nosed snake), and C. o. klauberi (Tucson shovel-nosed snake). Chionactis occipitalis klauberi is of particular conservation interest. This subspecies occurs along the eastern edge of the species range in south central Arizona and has experienced declines in population size and range over the past several decades due to alteration and loss of habitat through agricultural and urban development. These factors prompted a petition to list the subspecies as threatened or endangered under the U.S. Endangered Species Act . At present, C. o. klauberi has been elevated to “candidate status” (Priority 3) under the ESA –. Even so, existing taxonomic uncertainty within C. occipitalis may hinder further management actions . For example, C. o. klauberi was described despite the assumption that it formed a broad zone of intergradation with C. o. annulata across Arizona (Fig.1; ). Specimens taken from within the intergrade zone are often difficult to unambiguously assign to either subspecies because (i) they contain phenotypic characters consistent with C. o. klauberi but are found well outside the presumed geographic distribution, or (ii) specimens display intermediate morphological characteristics between the two subspecies. As a result, various authors have treated the range of C. o. klauberi differently , . Therefore, one problem in determining subspecies validity lies in the fact that the distribution of C. o. klauberi was never clearly delineated.
The presumed range of the morphological intergrade zone at the interface of C. o. annulata and C. o. klauberi is depicted by the dashed line.
Previous research by Wood et al.  investigated the genetic structuring of C. occipitalis using mitochondrial DNA (mtDNA) sequence data from 81 snakes and variation in 14 morphological characters (excluding color pattern) from 1543 snakes from across the range of the species. Sequence data revealed a deep phylogenetic split within the species with substantial genetic structuring within the two main clades. Most relevant to this study, mtDNA haplotypes of individuals identified as C. o. klauberi were intermixed with haplotypes belonging to C. o. annulata. Broad morphological overlap was detected between the four subspecies, and only minor geographic structure was evident for C. o. klauberi. Taken together, Wood et al.  suggested that C. o. klauberi may represent a “morphological endpoint of clinal variation without concordant phylogenetic distinction”, but encouraged further hypothesis testing with nuclear loci to evaluate the mtDNA gene tree patterns and possible ecological/adaptive differences that may exist despite non-exclusive mtDNA genetic variation.
Here, we build on this research using 11 nuclear microsatellite loci and additional mtDNA sequence data to further evaluate the genetic groups within C. occipitalis, placing emphasis on evaluating the genetic distinctiveness between klauberi and the neighboring annulata phenotypes across the Colorado and Sonoran deserts. Our objectives addressed three gaps in the current knowledge. First, we tested whether the nuclear genome shows patterns of structure that are consistent with the ranges of the four recognized phenotypic subspecies. Second, we tested for intergradation between annulata and klauberi, and identified the geographic location of a contact zone. Third, we quantified the extent to which the nuclear genetic clusters within C. occipitalis are isolated. Finally, we provide information from multiple genetic and phenotypic datasets to suggest taxonomic recommendations consistent with lineage evolution within this group.
Materials and Methods
Sampling and DNA extraction
We obtained 269 tissue samples of Chionactis occipitalis from localities throughout California, Arizona, Nevada, Baja California and Sonora, Mexico. Sampled localities represent the full range of the species and include all formally recognized subspecies (Fig.1). A list of all tissues and their associated data (e.g. museum and/or field number, collection site, and GPS coordinate information) is given in Table S1 (Dryad Digital Respository, doi:10.5061/dryad.77rf2). All necessary permits were obtained for our field efforts in California, Arizona, and Nevada: (1) Arizona permits issued by the Arizona Game and Fish Department (SP613877, SP802036, SP572402, SP755971, SP586491, SP711106), California permits issued by the California Department of Fish and Wildlife (SCP000838, SCP003850, SCP000297, SCP006488, SCP003696, SCP004186), Nevada permit issued by the Nevada Department of Wildlife (S33762), and permits issued by the National Park Service (JOTR 25A0 9–07, JOTR−2005−SCI−0024). All other tissues were obtained from museum and university loans, and the following abbreviations were used for these collections: Arizona State University (ASU); Royal Ontario Museum (ROM); San Diego Natural History Museum (SDSNH); San Diego Natural History Museum Tissue Series (SDField); University of Arizona, (UAZ); and University of Texas, Arlington (UTA). This study did not involve threatened, endangered, or protected species. This study was approved by the Western Ecological Research Center Animal Care and Use Committee in association with the University of California, Davis. Non-destructive tissue sampling techniques (i.e., drawn blood, tail-tips, and salvaged specimens) were used to obtain tissue for DNA extraction. Whole genomic DNA was extracted from tissues using Qiagen DNeasy extraction kits (Qiagen Inc., Valencia, CA, USA).
mtDNA data collection
We sequenced portions of the mitochondrial 16S rRNA and ND1 genes (total of 1094 bp) for 140 individuals of C. occipitalis and three Sonoran shovel-nosed snakes (C. palarostris) for use as an outgroup taxon in phylogenetic analyses. We also included the 80 mtDNA sequences generated from our previous study (, Genbank numbers: EU280331 – EU280410). The total number of sequences used for mtDNA analyses was 223. Primer sequences and PCR protocols are described in . We purified and directly sequenced PCR products on an ABI 3100 capillary system. We edited and aligned sequences using Sequencher 5.0.
Microsatellite data collection
Microsatellite library construction was performed in the Sequencing and Genotyping Facility at Cornell University Core Laboratory Center (CLC) using standard development techniques modified from . We used the program msatcommander  to scan the FASTA file generated from the 454-automated DNA sequencer (Roche) run for all dimeric, trimeric, tetrameric, and pentameric microsatellites.
We tested 32 loci for variability using six individuals from Arizona and California collection sites. Of the 32 loci, we selected 11 that were variable, amplified consistently, and yielded reliable genotyping scores. We divided these loci into three groups and simultaneously amplified 3–4 loci within each group using a Qiagen multiplex PCR kit in 10 µL reactions containing 5 µL of Qiagen multiplex PCR Master Mix, 1 µL primer mix (containing 2 µM of each primer), 1 µL Q-solution and 2 µL of RNase-free water. Amplified products were run on an ABI 3100 and performed in the CSUPERB Microchemical Core Facility at San Diego State University. We used GENE-MARKER v1.90 (SoftGenetics) to edit the raw allelic data and score allele sizes. We used MICROCHECKER  to test for the presence of null alleles and scoring errors. We genotyped 258 individuals and reanalyzed 10% of individuals at all loci.
Mitochondrial lineage estimates
We used BEAST v1.8.0  to simultaneously estimate phylogenetic relationships and divergence time estimates among mtDNA sequences from a total of 223 individual snakes (nexus file, Dryad Digital Respository, doi:10.5061/dryad.77rf2). We partitioned the data and used the best-fitting models of sequence evolution following Wood et al. . Because adequate fossil evidence was lacking for Chionactis or closely related sister species, we did not use fossil evidence to calibrate our clock-based analyses. We instead employed fixed substitution rates for the ND1 and 16S rRNA flanking regions that were estimated for other squamates (ND1: 1.39 × 10−2 substitutions/site/million years; flanking region: 4.92 × 10−3 substitutions/site/million years)  to provide lineage age estimates within shovel-nosed snakes. Substitution rates were set in BEAUti v1.8.0  using normal priors, with standard deviations reflecting uncertainty in the estimate. Preliminary analyses were conducted to test the assumption of clock-like evolution and were confirmed using the Bayes Factor test  and implemented in Tracer 1.5 . We conducted four independent runs in BEAST using a strict clock model and a coalescent constant size tree prior. We obtained posterior distributions of model parameters and genealogies by sampling from the Markov chain Monte Carlo (MCMC) posterior distribution using four chains every 1000th generation for a total of 20 million generations (first 40% of samples were discarded as burnin). We assessed convergence by visually inspecting the cumulative posterior probabilities of split frequencies using the program awty . We used the maximum clade credibility tree and posterior probabilities (Pp ≥ 0.95) to infer mtDNA relationships and evaluate support for phylogenetic lineages within C. occipitalis.
Microsatellite diversity and cluster estimates
Genetic diversity was assessed at each microsatellite locus and genetic cluster by estimating the observed heterozygosity (HO), expected heterozygosity (HE), number of alleles (NA), and allelic richness (AR). These genetic indices were calculated using GENALEX v6.41 . We tested for linkage disequilibrium between all pairs of loci using Fisher's exact tests implemented in GENEPOP on the web , , and the level of statistical significance (α = 0.05) was adjusted using the Bonferroni correction .
We used two Bayesian genotypic clustering analyses to infer discrete genetic clusters within the dataset at two different spatial scales (microsatellite data file, Dryad Digital Respository, doi:10.5061/dryad.77rf2). First, we conducted a range-wide analysis of population structure using structure version 2.3.2 , a nonspatial algorithm that is routinely used for identifying the number of genetic clusters in microsatellite datasets. Second, a fine-scale analysis was conducted on collection sites sampled across Arizona to detect spatial patterns where C. o. annulata and C. o. klauberi supposedly intergrade. For this analysis, we used the clustering algorithm implemented in geneland version 3.3  because it can take spatial information into account to produce inferences of genetic structure and boundary locations based on the geographic distribution of individuals. Both programs group individuals into the most likely number of clusters (K) that maximizes the within-cluster Hardy-Weinberg and linkage equilibria.
For the structure analyses, the number of clusters (K) that putatively best explains the dataset is inferred from the posterior probability distribution of the data given the range of K specified by the user, P(X|K). We ran the admixture model with an uncorrelated frequency model. We inferred the number of clusters (K) by comparing the results from the mean lnP(D|K) score against the Kmax (i.e. where the lnP(D|K) curve plateaus) and the Δ K criterion . We also hierarchically tested for substructuring within clusters by performing subsequent structure analyses within each inferred cluster . For all structure analyses, we estimated the probability of one through 10 clusters (K) using 1,000,000 iterations of the MCMC algorithm following a burn-in of 500,000 iterations, with each run replicated ten times to check for the consistency across the Markov chains.
For geneland analyses, all parameters (including the number of clusters, K) are co-estimated simultaneously by the MCMC algorithm. Once completed, the most probable number of clusters and their geographic boundaries are inferred from the highest average posterior probability density (PPD) of genetic clusters. We followed the recommendations given by Guillot et al.  that data analyses begin with the uncorrelated frequency model. Although the correlated model can be more powerful at detecting subtle differentiation of genetic structure, it also can be more prone to depart from model assumptions (e.g., presence of isolation-by-distance) that are present in our dataset. We performed five independent runs of K = 1–5 with 10,000,000 MCMC iterations and a thinning of 250. Coordinates from each individual were used to run the spatial model and the potential error for spatial coordinates was set at 3 m. Post-processing of the MCMC outputs with the highest PPD followed a burn-in of 10,000 in order to obtain posterior probabilities of cluster membership for each individual.
We used FST estimates and isolation-by-distance (IBD) analyses as secondary methods to summarize microsatellite variation and to identify significantly differentiated genetic clusters inferred from the assignment tests. Pairwise genetic distances among clusters were based on allele frequency differences using FST following  and were implemented in GENALEX v6.41. To test for genetic isolation by distance, we compared pairwise matrices of geographic and genetic distance using Mantel tests for matrix correlation , with significance assessed by 5,000 randomizations of the genetic distance matrix. We calculated the individual-based genetic distance â developed by Rousset  between all pairs of individuals within each of the three clusters using the program GENEPOP on the web. In each analysis, we examined untransformed and log transformed axes, and report the combination that exhibited the highest correlation coefficient. IBD was examined for all sites and across three clusters identified with geneland (see results) that span the geographic range of C. o. annulata and C. o. klauberi. Because clustering approaches can sometimes mistake IBD for hierarchical population structure , we also performed a series of partial Mantel tests to assess whether the clusters identified by geneland could be explained by IBD. We accomplished this by testing the association between genetic distances and cluster membership with geographic distances treated as a covariate. All IBD analyses were performed in IBDWS 3.21 .
Microsatellite genetic isolation estimates
We used the population migration rate (2Nem), which is the effective rate at which genes enter a population, as a criterion for defining the relative strength of genetic isolation between the range-wide clusters identified using structure . We used the isolation-with-migration model implemented in the program IMa2 ,  to estimate effective population migration rates. We specified the topology of the population tree that is required for IMa2 analysis based on the mtDNA tree [Mojave, (Colorado, Sonoran)]. We used all 11 loci to estimate demographic parameters and applied the stepwise mutation model and an inheritance scalar of 1.0 to each locus. We assumed a generation time of 3 years for C. occipitalis to obtain a measure of migration on a scale of generations, which was based on demographic observations of a similar fossorial species, Chilomeniscus straminus (P. C. Rosen, personal communication). We used the state of the Markov chain from a long run using final priors to seed three separate runs (10,000 genealogies saved from each), after a series of initial runs to determine the most appropriate search parameters that maximized mixing.
For each analysis, we used the following settings for the prior distributions on population parameters: 160 chains with a geometric heating scheme (g1 = 0.99 and g2 = 0.5), maximum scalars for theta (q0 = 40; q1 = 40; q2 = 40; q3–4 = 160), maximum migration prior value (m = 3), and maximum time of population splitting (t = 15). A total of 100,000 steps were retained with a step length of 20 and the first 150,000 steps discarded as burn-in. We ordered the relative strength of the 2Nem estimates following recommendations of  and : strong isolation when 2Nem ≤ 1, moderate isolation when 1 < 2Nem ≤ 5 and weak when 5 < 2Nem ≤ 25.
The mtDNA Bayesian analyses yielded a well-resolved phylogeny. Most notably, C. occipitalis was composed of three geographically distinct mtDNA lineages that were similar to Wood et al. , although the increased geographic sampling in this study provided added resolution of clade boundaries and increased posterior probability support for a third lineage (Fig. 2; Fig. S1-3). The northernmost lineage, hereafter called the ‘Mojave lineage’, was composed of two well supported clades. Individuals belonging to clade A occupied the range of C. o. occipitalis and C. o. talpina throughout the Mojave Desert, with the exception of samples taken from eastern Mojave Desert in California and Arizona (see below). Individuals of clade B occurred farther south in the range of C. o. occipitalis and were restricted to the Coachella Valley in California. A second lineage, hereafter called the ‘Sonoran lineage,’ was composed of two geographically disjunct clades (clades C and D). Individuals of clade C were primarily located in the Sonoran Desert of central Arizona encompassing the range of C. o. klauberi and the northeastern range of C. o. annulata. Clade D encompassed the range of C. o. occipitalis within the southeastern portion of the Mojave Desert in California and Arizona. A third clade (clade E) was routinely placed as sister to the Sonoran lineage, but posterior probabilities supporting this relationship were weak (Pp < 0.50). This clade was located in the Sonoran Desert of western Arizona, partially encompassed the ranges of C. o. annulata and annulata-klauberi ‘intergrades’, and was geographically nested in between clade C and D. Finally, the southernmost lineage, hereafter called the ‘Colorado lineage,’ occupied the southwestern portion of the range of C. o. annulata and was distributed in the lower Colorado River subdivision of the Sonoran Desert in California and Baja California (clade F), and southwestern Arizona and Sonora, Mexico (clade G).
The geographic distribution of clades within each lineage are outlined with a dashed line, clade E (yellow clade and dots) was routinely placed as sister to the Sonoran lineage, but posterior probabilities supporting this relationship were weak (Pp < 0.50). Black circles at nodes represent Bayesian posterior probabilities of ≥ 0.95. See Figures S1-3 for more detail within each mtDNA lineage.
Our Bayesian clock estimates of divergence indicate that diversification of the major lineages within C. occipitalis occurred in the Pliocene to early Pleistocene (Table 1), which are consistent with molecular clock estimates reported by Wood et al. . The Mojave lineage diverged from the Colorado and Sonoran lineages in the Pliocene (3.3 Ma, 95% HPD 3–4 Ma), followed by further divergence of Colorado and Sonoran lineages in the late Pliocene and early Pleistocene (3.0 and 1.8 Ma respectively).
Genetic structure inferred from microsatellites
Allelic richness for each microsatellite locus genotyped across the 258 individuals ranged between 6.1 and 15.1 (9.2 on average), and observed heterozygosity for each locus ranged from 0.419 and 0.812, with an average value of 0.665 for all loci (Table 2). There were no pairs of loci with significant linkage disequilibrium after Bonferroni correction (P = 0.0009).
At the range-wide spatial scale, structure analyses strongly indicated a sharp plateau at Kmax = 3 as an approximate estimate for K under the admixture model. This result was corroborated by the Δ K criterion (Fig. S4). On the basis of Kmax = 3, cluster assignments for each sample and the geographic distribution of each cluster, hereafter referred to as the Mojave, Colorado, and Sonoran clusters, are depicted in Figures 3 and 4.
a) Each individual is represented by a single column with membership assignment probabilities for each of the three clusters (K) noted as the relative proportion of each color (blue, green, and red represent the Mojave, Colorado, and Sonoran clusters, respectively). The vertical black bars represent the a priori subspecies assignment for each individual that was used in the analysis.
Circles on the map are colored proportional to their posterior probability assignment to each of the three clusters.
The Mojave cluster encompassed the range of C. o. occipitalis in California and Arizona as well as the range of C. o. talpina, and the proportion of each individual's genome assigned to this cluster was fairly exclusive (i.e., ≥ 0.95). Unlike the mtDNA patterns, C. o. occipitalis from the eastern Mojave Desert in California and Arizona were grouped with other geographically proximate individuals of C. o. occipitalis rather than clustering with individuals of C. o. annulata in the Sonoran Desert as inferred using the mtDNA data (see mtDNA clade D). The Colorado cluster encompassed the range of C. o. annulata in California, Arizona, and Mexico and portions of the previously reported intergrade zone. Finally, the Sonoran cluster occupied the presumed range of C. o. klauberi, portions of the previously reported intergrade zone, and a small portion of the range of C. o. annulata in west-central Arizona. Admixture among the Colorado and Sonoran clusters was apparent in the vicinity of the reported intergrade zone. Individual snakes sampled throughout this region tended to exhibit lower probability assignments (0.5 ≤ P ≤ 0.7; see Fig 4), a pattern consistent with contemporary gene flow between C. o. annulata and C. o. klauberi. Within each of the three clusters, we conducted additional structure runs to look for evidence of further genetic structure, but additional clusters were not supported.
We restricted our geneland analyses to collection sites sampled across Arizona to detect fine-scale spatial patterns and boundary locations where C. o. annulata and C. o. klauberi intergrade. Taking into account the spatial location of genotyped samples across Arizona, the MCMC analysis of genetic structure using geneland placed greater than 99% of the posterior density on K = 3 (Fig. S5). The spatial depictions of cluster membership and posterior probability contours are presented in Figure 5. The assignments across Arizona were geographically similar with the results obtained using structure. However, an additional cluster was identified and boundary locations for each were marked by sharp posterior probability contours. The most notable feature of the geneland analysis was the spatial distribution of cluster B. This cluster was primarily composed of individuals that were of admixed assignment according to the structure analysis and encompassed the reported intergrade zone between C. o. annulata and C. o. klauberi. The spatial distribution of cluster C included the same localities as the Sonoran cluster identified by the structure analysis (Pinal, Maricopa, and northern La Paz Counties; Fig. 5). However, in northwestern Maricopa County, several individuals with admixed structure assignments were assigned to cluster A and one to B, all with high posterior probabilities. Other similarities between the cluster analyses included snakes sampled along the west-to-east transect in Maricopa and Pinal Counties. In both analyses these samples exhibited increasing posterior probability assignments from one cluster to another that coincides with the presumed transition from C. o. annulata to C. o. klauberi.
a) Map of Arizona and the geographic distribution of sampled individuals colored according to the probability of assignment within each cluster identified using geneland (green, orange, and red circles represent individuals assigned to Clusters A, B, and C, respectively). The shading indicates elevations below 300m (light grey), between 300 and 1000m (grey) and above 1000m (dark grey). b) Geneland probability contour maps for the three clusters across Arizona. The highest posterior probabilities are in white (p ≥ 0.9) and the lowest are in red (p ≤ 0.1) – contour lines within each map depict the spatial change in population assignment probability.
Genetic diversity and differentiation using microsatellites
Measures of genetic diversity did not vary considerably across the three structure clusters. We provide general summary statistics, including number of samples, average number of alleles, heterozygosity, and fixation index in Table 3.
We used pairwise FST estimates as an additional method to test the significance of genetic differences between the identified clusters. Allele frequencies were significantly different between the clusters identified by structure (Table 4). The highest among cluster pairwise FST estimates were found between the Mojave and Sonoran clusters (FST = 0.141) and lowest pairwise comparison was between the Mojave and Colorado clusters (FST = 0.054). We detected a significant isolation-by-distance pattern for all collection sites combined (r = 0.346, P ≤ 0.001) and for the three geneland clusters across Arizona (r = 0.385, P ≤ 0.001). However, partial Mantel tests revealed a significant association between clusters and genetic distance (r = 0.116, P ≤ 0.001) after correcting for the geographical distances. This result suggests allele frequency differences between clusters are not due to geographic distance alone.
The IMa2 genetic isolation estimates detected asymmetric rates of gene migration between clusters obtained from structure analyses (Table 5). Moderate values of genetic isolation were detected between Mojave and Sonoran clusters, with greater gene migration (2Nem) occurring from the Sonoran cluster to the Mojave cluster compared to the opposite direction. Strong to moderate genetic isolation was detected between the Colorado and Sonoran clusters, and higher gene migration was inferred from the Sonoran cluster into the Colorado cluster. We detected negligible variation among 2Nem values estimated between the Mojave and Colorado clusters, where gene migration was strongly restricted in both directions.
Phenotypic subspecies correspondence with genetic structure
In this study, we used both mitochondrial DNA and 11 microsatellite loci to further assess whether patterns of population genetic structure are concordant with the spatial structuring of phenotypic variation that originally led to the subspecies descriptions within Chionactis occipitalis. This significantly expanded geographic sampling and the addition of nuclear loci, allowed us to better characterize the genetic structure within C. occipitalis. Our study did not find support for C. occipitalis subspecies as exclusive genetic groups. Patterns of discordance between the phenotypic and genotypic boundaries were detected in areas where C. o. occipitalis, C. o. annulata, and C. o. klauberi are presumed to be in contact. The only exception to this pattern was C. o. talpina, which was entirely nested within C. o. occipitalis on the basis of both mtDNA and microsatellite datasets.
Differences in the spatial distribution of genetic and phenotypic variation are not unexpected when the boundaries of divergent lineages are parapatric. For instance, even though lineages may have diverged to the point of having some diagnosable differences (both phenotypic and genetic), these lineages may not be sufficiently reproductively isolated from each other to prevent gene flow in secondary contact. Although gene flow is often a homogenizing force , , instances of population (or species) divergence despite gene flow exist in snakes and other taxa –. Such admixture can create fuzzy boundaries along these zones of contact. Therefore, criteria that demand perfect overlap between phenotypic and genetic divergence patterns may not be realistic for parapatric entities, whether subspecies or species.
These results highlight the difficultly in delimiting taxonomic boundaries within lineages that are evolving under a time-dependent, continuous process (e.g., species – subspecies boundary) . Although we showed a lack of genetic exclusivity among the subspecies of C. occipitalis, there were a large proportion of individuals within C. o. occipitalis, C. o. annulata, and C. o. klauberi that were assigned to the Mojave, Colorado, and Sonoran clusters (Fig. 4), respectively. Contact between C. o. occipitalis and C. o. annulata is presumed to occur along the Mojave and Sonoran Desert ecotone, with putative intergrades following the crests of the Little San Bernardino and Chocolate Mountains in eastern Riverside and Imperial Counties in California . These boundaries are largely consistent with our multi-locus genetic data (Fig. 4). Klauber  noted the only exception to this division lies within the Coachella Valley in the northern Colorado Desert of California, the southernmost region occupied by C. o. occipitalis. He suggested that C. o. occipitalis may have accessed the Coachella Valley through the low valley passes within the Little San Bernardino Mountains where it eventually meets (according to phenotypic patterns) with C. o. annulata along the northwestern region of the Salton Sea at the Imperial-Riverside county line. Assignment probabilities of C. o. annulata individuals within the northern Colorado Desert are consistent with some genetic admixture in this region. Although our gene migration estimates indicate strong isolation between Mojave (occipitalis) and Colorado (annulata) clusters, these estimates are greater than zero suggesting that complete isolation between them has not been achieved.
We also found that C. o. klauberi and C. o. annulata were not genetically exclusive. The cluster containing all samples of C. o. klauberi (Sonoran cluster & cluster C according to structure & geneland, respectively) also included samples within the putative intergrade zone and a few C. o. annulata from northern La Paz County in Arizona. The key morphological characteristic used to diagnose C. o. klauberi, dark maculations infused within the red secondary bands that contact dorsally , are not present in all individuals forming this cluster, but this character likely increases in frequency from west to east throughout this cluster. Although phenotypic variation was not examined rigorously here, the presence of dark maculations were observed in most individuals assigned to the Sonoran cluster, but were primarily present laterally rather than connecting along the dorsum. However, Klauber  observed that the dark maculations only contact dorsally in about 50% of C. o. klauberi. He further noted several localities with “klauberi influences”, in particular “the Wickenburg area” and “east of Aguila” in northwestern Maricopa County, that corroborate well with the inferred cluster boundaries (Fig. 4 & Fig. 5a), which suggests that the western boundary of C. o. klauberi may be more extensive than previously assumed.
Although previous interpretations of phenotypic variation had presumed a broad intergrade zone between C. o. annulata and C. o. klauberi , the genetic data identified specific locations of genetic admixture that are primarily restricted within central Arizona (Fig 5, cluster B). From our sampling scheme, two regions of gene migration are apparent and merit further investigation: (i) within northwestern Maricopa County along the Gila River valley through Buckeye valley to the Hassayampa River valley, and (ii) within southwestern Maricopa County along the Gila River valley west of the Maricopa Mountains (Fig 5). Additional research quantifying both phenotypic and genetic variation should help to better understand whether the zone of contact results in uniformly admixed populations or whether divergence with gene flow is maintained between two distinctive phenotypes.
Previous authors  speculated whether the subspecies within C. occipitalis provided any meaningful understanding of the evolutionary history of this species (i.e., whether they serve as useful taxonomic surrogates for evolutionary lineages). The multilocus perspective presented here (mtDNA & microsatellites), provided evidence for at least three divergent groups that have broadly concordant boundaries with three of the phenotypically recognized subspecies (occipitalis, annulata, and klauberi). Because taxonomy should be consistent with the evolutionary history of a group, it is possible that these genetic groups represent independent evolutionary lineages and therefore constitute distinct species (following criteria outlined in the evolutionary and general lineages species concepts) , , . Among these three, klauberi exhibited the most discordance between the spatial distribution of genetic and phenotypic patterns, but this discordance may simply reflect the recency of its divergence (compared to divergence estimates between occipitalis and annulata), asymmetric gene flow along the zone of contact, and ambiguous delineation of the western phenotypic boundary. Therefore, some discordance within this geographic region would be expected. Nonetheless, we feel that the current phenotypic and genetic data available for klauberi at this time allows neither definitive acceptance nor rejection of species status for the lineage. Therefore, we propose the most advantageous and conservative strategy is to recognize two species in the group that conform to the divergent and phenotypically distinct Mojave and Colorado-Sonoran clusters: the Mojave shovel-nosed snake, Chionactis occipitalis (type locality, San Bernardino County, California)  and the Colorado Desert shovel-nosed snake, C. annulata (type locality, Imperial County, California) , . Previously reported color pattern variation along with the genetic variation can be used to diagnose these two species hypotheses (Table 6). In general, C. occipitalis possess more brown primary crossbands and no secondary red crossbands, while C. annulata have fewer black primary crossbands and possess red-to-orange secondary crossbands . We also recommend retaining annulata and klauberi as subspecific designations within C. annulata in keeping with the genetic structure that was recovered across Arizona, while placing talpina in synonomy with C. occipitalis. In this framework, maintenance of the subspecific nomenclature within C. annulata provides putative species hypotheses that can be examined in more detail, and provides important taxonomic handles for conservation policy and wildlife management agencies.
mtDNA vs. microsatellite variation
Because the genetic structure results differed slightly between marker types (mtDNA vs microsatellites), it is important to note where they differ and what these data reveal about subdivision within this species complex. Phylogenetic analysis of the mtDNA sequencing data revealed strong support for regionally distinctive lineages that sorted by desert basins (i.e. Mojave, Colorado, and Sonoran Deserts), with some overlap of lineages extending across desert boundaries. Likewise, our range-wide assignment tests based on microsatellite data confirmed the presence of at least three geographically cohesive clusters. The spatial discordance among gene tree lineage and cluster boundaries was narrow and occurred along two zones of contact. First, C. occipitalis mtDNA haplotypes from the eastern Mojave Desert (mtDNA clade D) were nested within the mtDNA Sonoran lineage, but the microsatellite data assigned these same samples with high probability to all other C. occipitalis samples within the Mojave Desert. In the same way, haplotypes of C. a. annulata from west-central Arizona (clade E) formed a well-supported mtDNA clade, but could not be grouped with confidence to either the Colorado or Sonoran lineages. However, nuclear genotypes of these same samples were assigned to the genetic cluster containing the majority of all other C. a. annulata samples. Thus, it appears that phylogenetic relationships based on mtDNA for clades D and E are misleading and suffer from introgression at lineage boundaries.
Wood et al.  noted the discordance between mtDNA and phenotypic patterns in these two regions, and suggested incomplete lineage sorting or contemporary gene flow as potential causes for the lack of agreement. Based on the results of this study, the likely cause of discordance is recent gene flow. Our mitochondrial divergence date estimates suggest that the recovered genetic groups may be quite old, having diverged approximately 1.8–3.3 Ma. Considering the age of diversification and geographic cohesiveness of the mitochondrial lineages, past mitochondrial gene flow, if it occurred at all, did not spread far beyond the contact zone. Second, results from our IMa2 analysis confirmed contemporary gene flow between the three clusters (where higher migration rates were observed from the Sonoran cluster into both Mojave and Colorado clusters). Because the sequencing analyses were conducted on a single mitochondrial locus, whereas the microsatellite analyses incorporated data from 11 different loci, our mtDNA patterns are likely more susceptible to locus-specific stochastic effects of gene flow at clade boundaries than the microsatellite data.
The southwestern deserts of North America are marked by phylogeographical breaks that result in distinctive genetic groups within species occupying this region. By comparing the phylogeographical signature of co-distributed taxa, Wood et al.  were able to identify broad spatial divisions that were associated with geographical barriers such as the Colorado River and the Mojave and Sonoran Desert ecotone. Division of lineages was associated with diversification events within the late Pliocene or older, but some discontinuities within species also corresponded to more recent events within the Pleistocene. With regard to our study, the recovered lineages have apparently existed in the southwestern deserts for a long period of time (≥ 3Ma; Table 2), only the Sonoran lineage had estimates that encompassed the Pleistocene. Such deep genetic structuring may indicate lineage diversification due to long standing barriers to gene flow that have since relaxed, allowing for possible secondary contact. Alternatively, the observed genetic structuring in this group may also indicate regional diversification driven by an adaptive response to physical and biotic factors that differ across the range. The boundary between C. occipitalis and C. annulata is defined by an abrupt transition in elevation, temperature, total rainfall, vegetation and terrain features known as the Mojave and Sonoran ecotone , . Such environmental gradients, if strong enough, may restrict gene flow , , . Several other aspects of our analyses are consistent with this hypothesis: (i) genetic differentiation among clusters remained significant, even after accounting for geographic distance among sampling locations, suggesting that some other mechanism besides geographic distance is driving genetic divergence, (ii) previous analyses found that the strongest predictors of genetic divergence among mitochondrial lineages were elevation, temperature and desert basin assignment , and (iii) we found asymmetric rates of gene flow between some clusters. While the spatial genetic patterns detailed in this study remain consistent with historical biogeographic reconstructions of the southwestern deserts (reviewed in ), the factors mentioned above lead us to believe there may also be more contemporary ecological and landscape barriers at play, equally sufficient to maintain genetic differentiation of the recovered lineages that may be locally adapted to specific desert basin conditions.
We have used data from a combination of mtDNA sequences and 11 microsatellite loci to study the genetic structure and phylogeography of Western shovel-nosed snakes. We evaluated whether phenotypic subspecies definitions were supported by genetic data, and did not find exact spatial congruence with the original subspecies hypotheses. Despite these results, both mitochondrial and nuclear data sets corroborated similar divergence patterns for at least three groups that are in differential contact along common distributional boundaries with low to moderate levels of gene exchange. Based on the available evidence, we suggested that species-level diversity is underestimated in this group. We proposed that at least two species be recognized, Chionactis occipitalis and C. annulata. For now, we recommend retention of the names annulata and klauberi as subspecific designations within C. annulata that conform to the patterns of genetic structure (i.e., Colorado and Sonoran clusters, respectively), while placing C. o. talpina in synomony with C. occipitalis. Whether additional species should be recognized within C. annulata will depend on future characterization of morphological and genetic variation within the zone of contact.
Because species require genetic diversity to respond to changing environmental conditions, conservation strategies aimed at protecting regional genetic groups can maximize the species' evolutionary potential and resilience , . Given the broad (but not exclusive) geographic cohesion found between the phenotypic and genetic variation in C. a. annulata and C. a. klauberi, continued recognition of each may assist in preserving the potential for future evolutionary change within this species. Further, these results may be particularly valuable for regional conservation planning in south-central Arizona. In addition to recovering unique genetic variation primarily within the range of C. a. klauberi, we were also able to better quantify the geographic extent of genetic intergradation with C. a. annulata. Instead of existing as a gradual cline over a broad area, the zone of intergradation was narrower and existed as steep shifts in allele frequencies. Gene flow between divergent lineages can create novel variation that can facilitate adaptive evolution –. Thus, protection of such diversity may best be achieved by maintaining viable populations within these heterogeneous areas – both ‘pure’ and the hybrid populations , , .
Mojave lineage pruned from the full mtDNA phylogeny, shown in the upper left. Numbers at nodes represent posterior probability support values for individual clades.
Sonoran lineage pruned from the full mtDNA phylogeny, shown in the upper left. Numbers at nodes represent posterior probability support values for individual clades.
Colorado lineage pruned from the full mtDNA phylogeny, shown in the upper left. Numbers at nodes represent posterior probability support values for individual clades.
Results from the mean lnP(D|K) score against the Kmax and the Δ K criterion of Evanno et al. .
Posterior density of the number of clusters (K) from the MCMC analysis of genetic structure across Arizona using geneland.
We thank Randy Babb, Bill Burger, Andrew Holycross, Thomas Jones, Dean Leavitt, Jonathan Richmond, and Phil Rosen for their assistance with this project and providing many helpful discussions. For critical review of the manuscript we thank Thomas Jones, Jonathan Richmond, Phil Rosen, and three anonymous reviewers. We also thank Marit Alanen and Brian Wooldridge (US Fish and Wildlife Service) both of whom were instrumental in supporting all aspects of this research. We thank Steve Bogdanowicz (EGCF, Cornell University) for assistance with microsatellite library development. We are grateful for the field efforts of the following people: Dennis Abbate, Christina Akins, Robery Bezy, Thomas Brennan, Valerie Boyarski, Chris Brown, Richard Cerka, Denise Clark, Todd Esque, David Grandmaison, Abigail King, Lori Hargrove, Bradford Hollingsworth, Robert Hansen, J. Lauerman, Christine LiWanPo, Aaron Lotz, Jerry McClusky, Thomas Meyers, Jonathan Miller, Audrey Owens, Lin Piest, Jimmy Rabbers, Carlton Rochester, Jim Rorabaugh, Abigail Rosenberg, Peter Scott, Angela Schwendiman, Richard Seigel, Justin & Tanzy Sullivan, Bruce Taubert, Michael Wall, and Rocco Wood. The use of trade, product, or firm names in this publication does not imply endorsement by the U.S. Government.
Conceived and designed the experiments: AGV DAW RNF. Performed the experiments: DAW. Analyzed the data: AGV DAW. Contributed reagents/materials/analysis tools: AGV DAW RNF. Wrote the paper: AGV DAW.
- 1. de Queiroz K (1998) The general lineage concept of species, species criteria, and the process of speciation. In: Howard DJ, Berlocher SH, editors. Endless Forms: Species and Speciation. Oxford, MA: Oxford University Press. pp. 57–75.
- 2. Darwin C (1859) On the origin of species by means of natural selection, or the preservation of favoured races in the struggle for life. London: John Murray.
- 3. Cracraft J (1983) Species concepts and speciation analysis. In: Johnston R, editor. Current Ornithology: Springer US. pp. 159–187.
- 4. Hawlitschek O, Nagy ZT, Glaw F (2012) Island evolution and systematic revision of comoran snakes: why and when subspecies still make sense. PLoS ONE 7: e42970.
- 5. Mayr E (1942) Systematics and the origin of species. New York: Columbia University Press.
- 6. Mayr E (1982) Of what use are subspecies? Auk 99: 593–595.
- 7. Wilson EO, Brown JWL (1953) The subspecies concept and its taxonomic application. Systematic Zoology 2: 97–111.
- 8. Avise JC (1994) Molecular markers, natural history, and evolution. London, UK: Chapman & Hall.
- 9. Bowen BW (1999) Preserving genes, species, or ecosystems? Healing the fractured foundations of conservation policy. Molecular Ecology 8: S5–S10.
- 10. Moritz C (2002) Strategies to protect biological diversity and the evolutionary processes that sustain it. Systematic Biology 51: 238–254.
- 11. Waples RS, Gaggiotti O (2006) What is a population? An empirical evaluation of some genetic methods for identifying the number of gene pools and their degree of connectivity. Molecular Ecology 15: 1419–1439.
- 12. Haig SM, D'Elia J (2010) Chapter 2: Avian subspecies and the U.S. Endangered Species Act. Ornithological Monographs 67: 24–34.
- 13. Mulcahy DG (2008) Phylogeography and species boundaries of the western North American Nightsnake (Hypsiglena torquata): Revisiting the subspecies concept. Molecular Phylogenetics and Evolution 46: 1095–1115.
- 14. O'Brien SJ, Mayr E (1991) Bureaucratic mischief: recognizing endangered species and subspeceis. Science 251: 1187–1188.
- 15. Phillimore AB, Owens IPF (2006) Are subspecies useful in evolutionary and conservation biology? Proceedings of the Royal Society B: Biological Sciences 273: 1049–1053.
- 16. Avise JC (1989) A role for molecular genetics in the recognition and conservation of endangered species. Trends in ecology & evolution (Personal edition) 4: 279–281.
- 17. Ball RM Jr, Avise JC (1992) Mitochondrial DNA phylogeographic differentiation among avian populations and the evolutionary significance of subspecies. The Auk 109: 626–636.
- 18. Burbrink FT, Lawson R, Slowinski JB (2000) Mitochondrial DNA phylogeography of the polytypic North American rat snake (Elaphe obsoleta): a critique of the subspecies concept. Evolution 54: 2107–2118.
- 19. Janzen FJ, Krenz JG, Haselkorn TS, Brodie ED (2002) Molecular phylogeography of common garter snakes (Thamnophis sirtalis) in western North America: implications for regional historical forces. Molecular Ecology 11: 1739–1751.
- 20. Kodandaramaiah U, Weingartner E, Janz N, Leski M, Slove J, et al. (2012) Investigating concordance among genetic data, subspecies circumscriptions and hostplant use in the nymphalid butterfly Polygonia faunus. PLoS ONE 7: e41058.
- 21. Zink RM (2004) The role of subspecies in obscuring avian biological diversity and misleading conservation policy. Proceedings of the Royal Society of London Series B: Biological Sciences 271: 561–564.
- 22. Haig SM, Beever EA, Chambers SM, Draheim HM, Dugger BD, et al. (2006) Taxonomic considerations in listing subspecies under the U.S. Endangered Species Act. Conservation Biology 20: 1584–1594.
- 23. King TL, Switzer JF, Morrison CL, Eackles MS, Young CC, et al. (2006) Comprehensive genetic analyses reveal evolutionary distinction of a mouse (Zapus hudsonius preblei) proposed for delisting from the U.S. Endangered Species Act. Molecular Ecology 15: 4331–4359.
- 24. Patten MA (2010) Chapter 3: Null expectations in subspecies diagnosis. Ornithological Monographs 67: 35–41.
- 25. Funk DJ, Omland KE (2003) Species-level paraphyly and polyphyly: frequency, causes, and consequences, with insights from animal mitochondrial DNA. Annual Review of Ecology, Evolution, and Systematics 34: 397–423.
- 26. Ballentine B, Greenberg R (2010) Common garden experiment reveals genetic control of phenotypic divergence between Swamp Sparrow subspecies that lack divergence in neutral genotypes. PLoS ONE 5: e10229.
- 27. Ballard JWO, Whitlock MC (2004) The incomplete natural history of mitochondria. Molecular Ecology 13: 729–744.
- 28. Emerson KJ, Merz CR, Catchen JM, Hohenlohe PA, Cresko WA, et al. (2010) Resolving postglacial phylogeography using high-throughput sequencing. Proceedings of the National Academy of Sciences 107: 16196–16200.
- 29. Klauber L (1951) The shovel-nosed snake, Chionactis with descriptions of two new subspecies. Transactions of the San Diego Society of Natural History 11: 141–204.
- 30. Stickel W (1941) The subspecies of the spade-nosed snake, Sonora occipitalis. Bulletin of the Chicago Academy of Sciences 6: 135–140.
- 31. Stickel W (1943) The Mexican snakes of the genera Sonora and Chionactis with notes on the status of other colubrid genera. Proceedings of the Biological Society of Washington 56: 109–128.
- 32. Center for Biological Diversity (2004) Petition to list the Tucson shovel-nosed snake (Chionactis occipitalis klauberi) as an endangered species. http://www.biologicaldiversity.org/species/reptiles/Tucson_shovel-nosed_snake/
- 33. U.S. Fish and Wildlife Service (2010) Endangered and threatened wildlife and plants; 12-month finding on a petition to list the Tucson shovel-nosed snake (Chionactis occipitalis klauberi) as threatened or endangered with critical habitat. Federal Register 75: 16050–16065.
- 34. U.S. Fish and Wildlife Service (2011) Endangered and threatened wildlife and plants; review of native species that are candidates for listing as endangered or threatened; annual notice of findings on resubmitted petitions; annual description of progress on listing actions. Federal Register 76: 66396.
- 35. U.S. Fish and Wildlife Service (2012) Endangered and threatened wildlife and plants; review of native species that are candaditates for listing as endangered or threatened; annual notice of findings on resubmitted petitions; annual description of progress on listing actions. Federal Register 77: 70018.
- 36. Mahrdt C, Beaman K, Rosen P, Holm P (2001b) Chionactis occipitalis. Catalogue of American Amphibians and Reptiles 731: 1–12.
- 37. Wood DA, Meik JM, Holycross AT, Fisher RN, Vandergast AG (2008) Molecular and phenotypic diversity in the Western Shovel-nosed snake, with emphasis on the status of the Tucson Shovel-nosed snake (Chionactis occipitalis klauberi). Conservation Genetics 9: 1489–1507.
- 38. Hamilton MB, Pincus EL, Di Fiore A, Fleischer RC (1999) Universal linker and ligation proceedures for construction of genomic DNA libraries enriched for microsatellites. BioTechniques 27: 500–507.
- 39. Faircloth BC (2008) msatcommander: detection of microsatellite repeat arrays and automated, locus-specific primer design. Molecular Ecology Resources 8: 92–94.
- 40. Van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) micro-checker: software for identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes 4: 535–538.
- 41. Drummond AJ, Suchard MA, Xie D, Rambaut A (2012) Bayesian phylogenetics with BEAUti and the BEAST 1.7. Molecular Biology and Evolution.
- 42. Chan LM, Choi D, Raselimanana AP, Rakotondravony HA, Yoder AD (2012) Defining spatial and temporal patterns of phylogeographic structure in Madagascar's iguanid lizards (genus Oplurus). Molecular Ecology 21: 3839–3851.
- 43. Suchard MA, Weiss RE, Sinsheimer JS (2001) Bayesian selection of continuous-time markov chain evolutionary models. Molecular Biology and Evolution 18: 1001–1013.
- 44. Rambaut A, Drummond AJ (2007) TRACER v1.5. <http://beast.bio.ed.ac.uk/Tracer>.
- 45. Nylander J, Wilgenbusch J, Warren D, Swofford D (2008) AWTY (are we there yet?): a system for graphical exploration of MCMC convergence in Bayesian phylogenetics. Bioinformatics 24: 581–583.
- 46. Peakall R, Smouse P (2012) GenAlEx 6.5: Genetic analysis in Excel. Population genetic software for teaching and research – an update. Bioinformatics.
- 47. Raymond M, Rousset F (1995) GENEPOP (Version 1.2): Population genetics software for exact tests and ecumenicism. Journal of Heredity 86: 248–249.
- 48. Rousset F (2008) genepop'007: a complete re-implementation of the genepop software for Windows and Linux. Molecular Ecology Resources 8: 103–106.
- 49. Rice WR (1989) Analyzing tables of statistical tests. Evolution 43: 223–225.
- 50. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155: 945–959.
- 51. Guillot G, Estoup A, Mortier F, Cosson JF (2005) A spatial statistical model for landscape genetics. Genetics 170: 1261–1280.
- 52. Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software structure: a simulation study. Molecular Ecology 14: 2611–2620.
- 53. Degner JF, Silva DM, Hether TD, Daza JM, Hoffman EA (2010) Fat frogs, mobile genes: unexpected phylogeographic patterns for the ornate chorus frog (Pseudacris ornata). Molecular Ecology 19: 2501–2515.
- 54. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution 38: 1358–1370.
- 55. Mantel N (1967) The detection of disease clustering and a generalized regression approach. Cancer Research: 209–220.
- 56. Rousset F (2000) Genetic differentiation between individuals. Journal of Evolutionary Biology 13: 58–62.
- 57. Meirmans PG (2012) The trouble with isolation by distance. Molecular Ecology 21: 2839–2846.
- 58. Jensen JL, Bohonak AJ, Kelley ST (2005) Isolation by distance, web service. BMC Genetics 6: 13.
- 59. Hey J (2010) Isolation with migration models for more than two populations. Molecular Biology and Evolution 27: 905–920.
- 60. Hey J, Nielsen R (2004) Multilocus methods for estimating population sizes, migration rates and divergence time, with applications to the divergence of Drosophila pseudoobscura and D. persimilis. Genetics 167: 747–760.
- 61. Reilly SB, Marks SB, Jennings WB (2012) Defining evolutionary boundaries across parapatric ecomorphs of black salamanders (Aneides flavipunctatus) with conservation implications. Molecular Ecology: n/a-n/a.
- 62. Wood DA, Vandergast AG, Barr KR, Inman RD, Esque TC, et al. (2013) Comparative phylogeography reveals deep lineages and regional evolutionary hotspots in the Mojave and Sonoran Deserts. Diversity and Distributions 19: 722–737.
- 63. Wright S (1931) Evolution in Mendelian populations. Genetics 16: 97–159.
- 64. Mayr E (1954) Change of genetic environment and evolution. In: Huxley JS, Hardy AC, Ford EB, editors. Evolution as a Process. London: Allen and Unwin. pp. 157–180.
- 65. Gavrilets S, Li H, Vose MD (2000) Patterns of parapatric speciation. Evolution 54: 1126–1134.
- 66. Garant D, Kruuk LEB, Wilkin TA, McCleery RH, Sheldon BC (2005) Evolution driven by differential dispersal within a wild bird population. Nature 433: 60–65.
- 67. Fitzpatrick BM (2008) Distinctiveness in the face of gene flow: hybridization between specialist and generalist gartersnakes. Molecular Ecology 17: 4107–4117.
- 68. Nadachowska K, Babik W (2009) Divergence in the face of gene flow: the case of two newts (Amphibia: Salamandridae). Molecular Biology and Evolution 26: 829–841.
- 69. Vandewege MW, Rodriguez D, Weaver JP, Hibbitts TD, Forstner MRJ, et al. (2012) Evidence of hybridization between Elaphe bairdi and Elaphe obsoleta lindheimeri including comparative population genetics inferred from microsatellites and mitochondrial DNA. Journal of Herpetology 46: 56–63.
- 70. Wiley EO (1978) The evolutionary species concept reconsidered. Systematic Zoology 27: 17–26.
- 71. Frost DR, Hillis DM (1990) Species in concept and practice: herpetological applications. Herpetologica 46: 87–104.
- 72. Hallowell E (1854) Descriptions of reptiles from California. Proceedings of the Academy of Natural Sciences of Philadelphia 7: 91–97.
- 73. Baird SF (1859) Reptiles of the boundary. In: Emory, editor. United States and Mexican Boundary Survey, Reptiles of the Boundary, vol 2. pp. 1–35.
- 74. Smith HM, Taylor EH (1950) Type localities of Mexican reptiles and amphibians. University of Kansas Science Bulletin 33: 313–380.
- 75. Bailey RG (1995) Description of the Ecoregions of the United States, 2nd edition revised and expanded (1st edition 1980). In: US Department of Agriculture, editor. 2nd ed. Washington, DC: USDA Forest Service. pp. 108.
- 76. McNab WH, Cleland DT, Freeouf JA, Keys JE Jr, Nowacki GJ, et al.. (2005) Description of ecological subregions: sections of the conterminous United States [CD-ROM]. In: U.S. Department of Agriculture FS, editor. Washington, DC. pp. 80.
- 77. Bull CM (1991) Ecology of parapatric distributions. Annual Review of Ecology and Systematics 22: 19–36.
- 78. Endler JA (1977) Geographic Variation, Speciation and Clines. Princeton, NJ: Princeton University Press.
- 79. Carvalho SB, Brito JC, Crespo EJ, Possingham HP (2011) Incorporating evolutionary processes into conservation planning using species distribution data: a case study with the western Mediterranean herpetofauna. Diversity and Distributions 17: 408–421.
- 80. Santamaría L, Méndez PF (2012) Evolution in biodiversity policy – current gaps and future needs. Evolutionary Applications 5: 202–218.
- 81. Barton NH, Hewitt GM (1985) Analysis of hybrid zones. Annual Review of Ecology and Systematics 16: 113–148.
- 82. Ebert D, Haag C, Kirkpatrick M, Riek M, Hottinger JW, et al. (2002) A selective advantage to immigrant genes in a Daphnia metapopulation. Science 295: 485–488.
- 83. Lavergne S, Molofsky J (2007) Increased genetic variation and evolutionary potential drive the success of an invasive grass. Proceedings of the National Academy of Sciences 104: 3883–3888.
- 84. Tallmon DA, Luikart G, Waples RS (2004) The alluring simplicity and complex reality of genetic rescue. Trends in Ecology and Evolution 19: 489–496.
- 85. Crandall KA, Bininda-Emonds ORP, Mace GM, Wayne RK (2000) Considering evolutionary processes in conservation biology. Trends in Ecology and Evolution 15: 290–295.