Phylogeny of the Dactyloa Clade of Anolis Lizards: New Insights from Combining Morphological and Molecular Data

Abstract We present a phylogenetic analysis of the Dactyloa clade of Anolis lizards, based on morphological (66 characters of external morphology and osteology) and molecular (?4,700 bases of mitochondrial and nuclear DNA) data. Our set of morphological characters includes some that exhibit continuous variation and others that exhibit polymorphism within species; we explored different coding methods for these classes of characters. We performed parsimony and Bayesian analyses on morphology-only and combined data sets. Additionally, we explicitly tested hypotheses of monophyly of: 1) Dactyloa including Phenacosaurus, 2) Dactyloa excluding Phenacosaurus (as traditionally circumscribed), 3) taxa previously ranked as series or species groups described based on morphological characters, and 4) clades inferred from molecular data. The morphological data alone did not yield Dactyloa or any of the previously recognized series described based on morphological characters; only the Phenacosaurus clade (as delimited based on molecular data) was inferred with the morphological data, and only in the parsimony analysis. In contrast, Dactyloa was inferred as monophyletic with the combined data set, although topology tests failed to reject the hypothesis of non-monophyly. Additionally, five clades inferred based on molecular data (eastern, latifrons, Phenacosaurus, roquet, and western) were inferred with the combined data sets with variable support and including additional species for which molecular data were not available and which have geographic distributions that conform to those of the clades in which they were included. Of the previously recognized taxa based on morphological characters, only the roquet series, which corresponds in species composition to the roquet clade, was inferred with the combined data. Topology tests with the combined data set rejected the monophyly of the aequatorialis, latifrons (as traditionally circumscribed), and punctatus series but not that of the tigrinus series and Phenacosaurus (as traditionally circumscribed). Our phylogenetic analyses and topology tests indicate that a new taxonomy for Dactyloa is warranted; we therefore present a revised taxonomy based on the results our phylogenetic analyses and employing phylogenetic definitions of taxon names. NOTE ADDED IN PROOF Shortly after our paper was accepted, Nicholson and colleagues published a phylogenetic analysis of anoles and a proposal to divide Anolis into eight genera (Nicholson, K. E., B. I. Crother, C. Guyer, and J. M. Savage. 2012. It is time for a new classification of anoles (Squamata: Dactyloidae). Zootaxa 3477: 1?108). Here, we comment briefly on their study as it pertains to the phylogeny and taxonomy of the Dactyloa clade. Despite not inferring Dactyloa to be monophyletic in the tree used for their proposed taxonomy (i.e., the consensus tree from the combined morphological and molecular parsimony analysis; their fig. 5A, note positions of Anolis bonairensis, A. chloris, A. peraccae, and A. apollinaris), Nicholson et al. (2012) recognized Dactyloa as one of their eight genera without making reference to this inconsistency (although Dactyloa was inferred to be monophyletic in their molecular tree, fig. 4A). By contrast, our combined data set supported the monophyly of Dactyloa (Figs.?3, 4), and we have chosen to treat Dactyloa as a subclade of Anolis rather than as a separate genus in the interest of avoiding disruptive and unnecessary name changes. Some of our informally named series correspond, with some differences in species composition, to the species groups proposed by Nicholson et al. (2012). We describe the differences below. Our latifrons series corresponds to their latifrons species group, except that in the tree purportedly used for their taxonomy (fig. 5A), A. aequatorialis and A. ventrimaculatus were inferred to be part of this species group (both species are absent from their molecular tree, fig. 4A), although their classification (appendix III) places both species in their punctata species group with no explanation for this inconsistency. We inferred these two species with strong support to be part of a monophyletic aequatorialis series that is mutually exclusive with respect to both the latifrons and punctatus series. Additionally, we have tentatively placed A. mirus and A. parilis in the aequatorialis series based on their previous inclusion in the traditional aequatorialis series (Williams, 1975; Ayala-Varela and Velasco, 2010); the tentative assignment reflects the current absence of these species from explicit phylogenetic analyses. By contrast, Nicholson et al. (2012) assigned A. mirus and A. parilis, neither of which was included in any of their analyses, to their latifrons species group without explanation. Finally, we placed A. propinquus in the latifrons series based on its hypothesized close relationship to A. apollinaris (Williams, 1988). By contrast, Nicholson et al. (2012) placed this species, which was not included in any of their phylogenetic analyses, in their punctata species group without explanation. The combination of our aequatorialis and punctatus series corresponds roughly to the punctata species group in the classification of Nicholson et al. (2012, appendix III). We inferred these two series to be mutually exclusive clades (results further supported by molecular data alone; Castañeda and de Queiroz, 2011). Contradicting their own taxonomy, the tree of Nicholson et al. (2012, fig. 5A) supports the separation of the aequatorialis series, in that A. aequatorialis, A. ventrimaculatus, A. chloris, and A. peraccae are not inferred to be part of their punctata species group, despite being referred to that group in their classification (appendix III). Their tree does place A. fasciatus in their punctata species group, whereas our results indicate that this species is part of the aequatorialis series. We have treated A. calimae and A. cuscoensis as incertae sedis within Dactyloa based on conflicting results for A. calimae (also found by Castañeda and de Queiroz, 2011) and the inferred inclusion of A. cuscoensis by Poe et al. (2008) in clades not inferred in our study. By contrast, Nicholson et al. (2012) referred these two species to their punctata species group, although neither species was included in any of their phylogenetic analyses. Similarly, we have treated A. laevis and A. phyllorhinus, species formerly placed in the laevis series, as incertae sedis based on their current absence from explicit phylogenetic analyses (although we consider it likely that A. phyllorhinus belongs to the punctatus series). By contrast, Nicholson et al. (2012) assigned both of these species to the punctata species group, although neither was included in any of their phylogenetic analyses. Our Phenacosaurus and our heterodermus series both correspond approximately to the heterodema species group of Nicholson et al. (2012), with the exception that they included A. carlostoddi, A. bellipeniculus, and A. neblininus. We consider these three species as incertae sedis within Dactyloa based on conflicting results in our analyses for A. carlostoddi and A. neblininus and the absence from explicit phylogenetic analyses of A. bellipeniculus, as well as its previously inferred close relationship to A. neblininus (Myers and Donnelly, 1996). Our roquet series corresponds approximately to their roquet species group. However, in the tree purportedly used for their taxonomy (their fig. 5A), their roquet species group is not monophyletic: A. bonairensis is inferred as sister to A. occultus outside of Dactyloa (A. bonairensis is not included in their molecular-only tree; fig. 4A). By contrast, we inferred A. bonairensis to be part of a monophyletic roquet series (Figs.?3, 4). Our combined analyses are based on a sample of 60 of the 83 currently recognized species in the Dactyloa clade, 40 of which were sampled for molecular data, whereas the combined analysis of Nicholson et al. (2012) is based on a sample of 31 Dactyloa species, 16 of which were sampled for molecular data (three others were sampled for molecular data only). Additionally, our molecular data consists of ?4,950 base positions representing three gene regions and both mitochondrial and nuclear DNA, whereas theirs consists of ?1,500 base positions representing one of the two mitochondrial gene regions used in our study. Because our results are based on larger samples of Dactyloa species (for both molecular and morphological data), as well as larger samples of molecular data (with respect to both numbers of bases and numbers of gene fragments, and including both mitochondrial and nuclear genes), and because many of their taxonomic conclusions that differ from ours are either contradicted by their own results or unsubstantiated, we do not consider any of the differences between our phylogenetic results and taxonomic conclusions compared with those in the study by Nicholson et al. (2012) to warrant changes to our proposed taxonomy. In contrast to Nicholson et al. (2012), we refrain from assigning some species to series and treat some taxonomic assignments as tentative because of contradictory results or poorly supported inferences, and we present justifications for all taxonomic decisions pertaining to species not included in our analyses.