Phylogeny of Prunellidae and taxonomic implications
The topologies of our locus specific and combined dataset trees showed a great deal of similarity suggesting an overall lack of conflict between phylogenetic signal in the mitochondrial ND2 gene and the Z chromosome specific ACO1I9. The topologies supported, (1) the monophyly of Prunellidae, (2) two major clades within the family, (3) strongly supported relationships between several species, and (4) strongly supported divergence between subspecies of both collaris and atrogularis (Fig. 3; Appendixes S3, S4).
The two major clades of Prunellidae separated the larger, alpine species (collaris and himalayana) from smaller accentors associated with shrubs or scrub habitats (Hatchwell 2005). Due to the level of divergence between these clades, as well as the distinct ecological and morphological differences between them, recognition of Laiscopus and Prunella as distinct genera may be warranted.
All topologies also suggested that immaculata and rubeculoides are distantly related to other species in the strongly supported clade of small Prunella accentors. The relationship of these two species relative to each other and to remaining species in the clade of small accentors was well supported only in the species tree topology, which suggests that immaculata was the first to diverge from the common ancestor of the small accentors, with rubeculoides diverging more recently.
Among the remaining species within the clade of smaller accentors not all relationships were strongly supported. The sister relationship of rubida and montanella, of fulvescens and koslowi, and of ocularis and fagani were well supported as were the sister relationships of the subspecies pairs of collaris and atrogularis and the close relationship of atrogularis to ocularis + fagani (Fig. 3). Deeper divergences in this clade were poorly supported although the combined analysis produced geographically coherent topology eliminating the need to invoke chance long-distance dispersal events. We speculate that the likely explanation for such a lack of resolution is the rapid radiation of these lineages stemming from their colonization of new areas of the Palearctic.
Inclusion of two subspecies of both collaris and atrogularis into our study revealed inconsistencies of their taxonomic treatment relative to other Prunellids. The divergence between collaris subspecies was greater than divergence among atrogularis, fagani, and ocularis, and between fulvescens and koslowi (Fig. 3). The amount of divergence between these two subspecies suggests the need for a study of genetic variation across the range of collaris that could result in elevating some collaris subspecies to specific status.
Although the divergence between atrogularis subspecies was much lower than the divergence between the two subspecies of collaris, it was very similar to the divergence between fagani and ocularis, which are treated as distinct species. The treatment of the former pair as subspecies and the latter as species may not be justified given that both pairs are allopatric and have equally minor phenotypic differentiation. Given their geographic isolation from one another, and the lineage sorting apparent in our species tree, we suggest that it might be reasonable to elevate these atrogularis subspecies to specific status.
Historical biogeography and geographic mode of speciation
Given the scarce paleorecord for Prunellidae that predates the late Pleistocene (Tyrberg 1998) we used the evolutionary rates reported by (Lerner et al. 2011) for Hawaiian honeycreepers (Fringillidae) to date divergence events in our phylogeny of Prunellidae. However, two records allow us to compare fossil dates with the divergence ages on our species tree. The first fossil date is for modularis from Mallorca Is., dated to 1.64–1.80 Ma, and the second is for collaris from continental Spain, dated to 0.8 Ma (Tyrberg 1998). In our species tree, the node separating modularis from its sister clade is dated to 1.81 (95% HPD 1.29–2.36) Ma, and the split between collaris subspecies is dated to 0.93 (95% HPD 0.52–1.38) Ma. Although our molecular divergence dates are slightly older than these paleorecords, the latter are well within the 95% HPD intervals of the former. Furthermore, paleorecords represent minimum ages. Thus, we feel that our molecular dating of the Prunellidae phylogeny is reliable.
Our reconstruction of the biogeographic history of Prunellidae suggests that the origin of family, divergence of the two subgenera (Laiscopus and Prunella), and initial diversification events within subgenus Prunella happened within the Himalayan region between 14.8 Ma and 3.69 Ma (Fig. 3). Subgenus Prunella dispersed out of the Himalayan region and across the Palearctic from the mid- to late-Pliocene between 3.69 Ma and 2.1 Ma (Fig. 3). This colonization of the Palearctic was followed by a rapid radiation of accentors suggesting the importance of colonizing new biogeographic regions and vicariant events resulting from Pleistocene glacial retreats in their speciation history. The most recent diversification events in both subgenera, occurred in the early to middle Pleistocene, and happened within or between Palearctic regions (except strophiata; Fig. 3).
Regardless of whether we calculated sympatry between sister clades (Barraclough and Vogler 2000) or sympatry among species ranges (Fitzpatrick and Turelli 2006), the relationship between sympatry and divergence age was best described by the logistic sigmoidal curve (Figs. 4, 5). The reason for the nonlinear relationship between sympatry and age is that all clades and taxa that diverged <1.5 Ma are allopatric. The only exception is the sister pair of koslowi and fulvescens that diverged 0.91 Ma and have a range overlap of 49%. However, while these species have a substantial geographic overlap in range, there is a striking difference in their habitat preferences: koslowi lives in thin scrub, semi desert habitat (which is unusual for accentors), whereas fulvescens prefers shrubs near timberline in the mountains (which is typical of small accentors; Hatchwell 2005).
The degree of sympatry among clades and taxa rapidly increased between 1.5 Ma and 3 Ma since divergence. Although there are some divergent taxa that currently have nonoverlapping ranges with other species, the node sympatry, and both maximum and mean species sympatry remain relatively high (Figs. 4, 5). This pattern suggests that allopatric speciation was the predominant geographic mode of speciation in Prunellidae.
Despite the similarity of the pattern of relationships between sympatry and age, the clade sympatry fits the logistic sigmoidal curve much better than the maximum and, especially, the mean species sympatry. The use of individual species ranges greatly increases the variance and reduces the goodness of fit due to the stochasticity of individual range sizes and their location. Consider fagani whose range does not overlap with any other species because it is located at the far extreme of the family distribution (Fig. 2). Such a range provides no information about the relationship between the sympatry and age because its sympatry value (0) is determined by a historic accident – the probability of colonizing a small, remote habitat island rather than by interactions with other species. However, the presence of this range has a dramatic effect on the mean species sympatry calculation. Furthermore, inconsistencies in the taxonomic treatment of a group can also have a strong effect on calculations of maximum and mean species sympatry. This is because changes in taxonomic treatment (e.g., recognizing divergent collaris races as species) will result in changes in the sizes of species ranges and thus the degree of their overlap with other species, as well as in the number of comparison pairs. Therefore, the more species that are recognized and the smaller ranges that they have, the less likely they will overlap, reducing the probability of discovering a significant relationship between species sympatry and age.
By comparison, the use of node ranges eliminates the effects of historic accidents and inconsistent taxonomic treatments because individual ranges of species forming a clade are joined into a single, combined range where each individual species range has little effect on sympatry calculation. To illustrate this, consider the effect of taxonomic treatment on calculation of sympatry between modularis and the clade consisting of atrogularis, fagani, and ocularis. If both subspecies of atrogularis are treated as distinct taxa, the sympatry between modularis and P. a. atrogularis is the maximum sympatry for their shared basal node (0.945). The mean sympatry for this node is 0.269, when P. a. atrogularis is recognized as distinct. If, despite their allopatric ranges and genetic differentiation, the subspecies of atrogularis are treated as a single species, these values change to 0.061 and 0.013, respectively – a drop greater than an order of magnitude. For the clade sympatry calculation, however, the difference in the taxonomic treatment has no effect.
Although peripatric speciation appears to be the most common geographic mode of speciation in animals (Barraclough and Vogler 2000), we found no relationship between clade range symmetry and age as expected for peripatric speciation. In contrast, this relationship was positive and indicative of peripatric speciation in an avian subfamily with a large Palearctic distribution (grouse, Tetraoninae; Drovetski 2003). In grouse, the combination of widely distributed boreal species and their glacial relict sister taxa in southern montane regions result in the positive relationship between age and range symmetry. Regardless of region, all accentor species are closely associated with mountains where habitats are not continuous, but rather have a patchy distribution. The size of the habitat patches available to accentors is determined by the physical and geographic characteristics of each individual mountain range. Therefore, a peripatric mode of speciation may not be a widespread driver of divergences in taxa restricted to habitat islands.
Despite abundant overlap in species distributions suggesting possible speciation in sympatry, our results for Prunellidae are consistent with other studies in identifying allopatric speciation as a primary mode of speciation in Palearctic bird lineages (Voelker 1999; Drovetski 2003; Drovetski et al. 2004b, 2010; Voelker 2010). In comparison with those other studies, one significant difference here is the recent and rapid colonization of Palearctic regions, mostly in the late Pliocene to early Pleistocene. This late colonization of the Palearctic and following rapid succession of divergence events suggest a prompt response to glacial retreat, but not necessarily an effect of glacial expansion and contraction on speciation patterns in Prunellidae.