Model-based analysis supports interglacial refugia over long-dispersal events in the diversification of two South American cactus species.

Pilosocereus machrisii and P. aurisetus are cactus species within the P. aurisetus complex, a group of eight cacti that are restricted to rocky habitats within the Neotropical savannas of eastern South America. Previous studies have suggested that diversification within this complex was driven by distributional fragmentation, isolation leading to allopatric differentiation, and secondary contact among divergent lineages. These events have been associated with Quaternary climatic cycles, leading to the hypothesis that the xerophytic vegetation patches which presently harbor these populations operate as refugia during the current interglacial. However, owing to limitations of the standard phylogeographic approaches used in these studies, this hypothesis was not explicitly tested. Here we use Approximate Bayesian Computation to refine the previous inferences and test the role of different events in the diversification of two species within P. aurisetus group. We used molecular data from chloroplast DNA and simple sequence repeats loci of P. machrisii and P. aurisetus, the two species with broadest distribution in the complex, in order to test if the diversification in each species was driven mostly by vicariance or by long-dispersal events. We found that both species were affected primarily by vicariance, with a refuge model as the most likely scenario for P. aurisetus and a soft vicariance scenario most probable for P. machrisii. These results emphasize the importance of distributional fragmentation in these species, and add support to the hypothesis of long-term isolation in interglacial refugia previously proposed for the P. aurisetus species complex diversification.

Phylogeography's past, present, and future: 10 years after Avise, 2000.

Approximately 20 years ago, Avise and colleagues proposed the integration of phylogenetics and population genetics for investigating the connection between micro- and macroevolutionary phenomena. The new field was termed phylogeography. Since the naming of the field, the statistical rigor of phylogeography has increased, in large part due to concurrent advances in coalescent theory which enabled model-based parameter estimation and hypothesis testing. The next phase will involve phylogeography increasingly becoming the integrative and comparative multi-taxon endeavor that it was originally conceived to be. This exciting convergence will likely involve combining spatially-explicit multiple taxon coalescent models, genomic studies of natural selection, ecological niche modeling, studies of ecological speciation, community assembly and functional trait evolution. This ambitious synthesis will allow us to determine the causal links between geography, climate change, ecological interactions and the evolution and composition of taxa across whole communities and assemblages. Although such integration presents analytical and computational challenges that will only be intensified by the growth of genomic data in non-model taxa, the rapid development of "likelihood-free" approximate Bayesian methods should permit parameter estimation and hypotheses testing using complex evolutionary demographic models and genomic phylogeographic data. We first review the conceptual beginnings of phylogeography and its accomplishments and then illustrate how it evolved into a statistically rigorous enterprise with the concurrent rise of coalescent theory. Subsequently, we discuss ways in which model-based phylogeography can interface with various subfields to become one of the most integrative fields in all of ecology and evolutionary biology.

Testing population genetic structure using parametric bootstrapping and MIGRATE-N.

We present a method for investigating genetic population structure using sequence data. Our hypothesis states that the parameters most responsible for the formation of genetic structure among different populations are the relative rates of mutation (micro) and migration (M). The evolution of genetic structure among different populations requires rates of M << p because this allows population-specific mutation to accumulate. Rates of micro << M will result in populations that are effectively panmictic because genetic differentiation will not develop among demes. Our test is implemented by using a parametric bootstrap to create the null distribution of the likelihood of the data having been produced under an appropriate model of sequence evolution and a migration rate sufficient to approximate panmixia. We describe this test, then apply it to mtDNA data from 243 plethodontid salamanders. We are able to reject the null hypothesis of no population structure on all but smallest geographic scales, a result consistent with the apparent lack of migration in Plethodon idahoensis. This approach represents a new method of investigating population structure with haploid DNA, and as such may be particularly useful for preliminary investigation of non-model organisms in which multi-locus nuclear data are not available.

Accounting for coalescent stochasticity in testing phylogeographical hypotheses: modelling Pleistocene population structure in the Idaho giant salamander Dicamptodon aterrimus.

Several theoretical studies have demonstrated the importance of accounting for coalescent stochasticity in phylogeographical studies, however, there are few empirical examples that do so in the context of explicit hypothesis testing. Here, we provide an example from the Idaho giant salamander (Dicamptodon aterrimus) using 118 mtDNA sequences, nearly 2 kb in length. This species is endemic to mesic forests in northern and central Idaho, and several a priori hypotheses have been erected based both on palaeoclimatic grounds and from phylogeographical studies of codistributed amphibians. Phylogenetic analysis of the D. aterrimus data suggests an expansion from a single refugium south of the Salmon River, whereas the inference from nested clade analysis is one of expansion from a single refugium in the Clearwater drainage. Explicit testing of these hypotheses, using geographically structured coalescent simulations to erect null distributions, indicates we can reject expansion from the Clearwater drainage (pCLW = 0.089), but not expansion from the South Fork of the Salmon drainage (pSAL = 0.329). Furthermore, data from codistributed amphibians suggest that there may have been two refugia, and an amova shows that most of the molecular variance partitioned between the Clearwater and the Salmon drainages (54.40%; P < 0.001) and within drainages (43.61%; P < 0.001). As a result, we also tested three a priori hypotheses which predicted that both the Clearwater and Salmon drainages functioned as refugia during the late Pleistocene; we could reject (PCORD = 0.019) divergence dates during the Cordilleran glacial maxima [c. 20 000 years before present (ybp)], during the Sangamon interglacial (c. 35 000 ybp; pSANG = 0.032), as well as pre-Pleistocene divergence (c. 1.7 Ma; ppP < 0.001). Mismatch distributions and Tajima's D within the individual drainages provide further support to recent population expansion. This work demonstrates coalescent stochasticity is an important phenomenon to consider in testing phylogeographical hypotheses, and suggests that analytical methods which fail to sufficiently quantify this uncertainty can lead to false confidence in the conclusions drawn from these methods.

Exploring population genetic structure in three species of Lesser Antillean bats.

We explore population genetic structure in phyllostomid bats (Ardops nichollsi, Brachyphylla cavernarum and Artibeus jamaicensis) from the northern Lesser Antilles by investigating the degree to which island populations are genetically differentiated. Our hypothesis, that the island populations are genetically distinct because of a combination of founding events, limited migration and genetic drift exacerbated by catastrophe-induced fluctuations in population size, is derived from a priori hypotheses erected in the literature. The first prediction of this hypothesis, that within each species island populations are monophyletic, was tested using a parametric bootstrap approach. Island monophyly could not be rejected in Ardops nichollsi (P = 0.718), but could be rejected in B. cavernarum (P < 0.001) and Artibeus jamaicensis (P < 0.001). A second prediction, that molecular variance is partitioned among islands, was tested using an amova and was rejected in each species [Ardops nichollsi (P = 0.697); B. cavernarum (P = 0.598); Artibeus jamaicensis (P = 0.763)]. In B. cavernarum and Artibeus jamaicensis, the admixture in mitochondrial haplotypes from islands separated by > 100 km of ocean can be explained either by interisland migration or by incomplete lineage sorting of ancestral polymorphism in the source population. As an a posteriori test of lineage sorting, we used simulations of gene trees within a population tree to suggest that lineage sorting is an unlikely explanation for the observed pattern of nonmonophyly in Artibeus jamaicensis (PW < 0.01; PSE = 0.04), but cannot be rejected in B. cavernarum (PW = 0.81; PSE = 0.79). A conservative interpretation of the molecular data is that island populations of Artibeus jamaicensis, although isolated geographically, are not isolated genetically.