Phylogeny and biogeography of a large radiation of Andean lizards (Iguania, Stenocercus)

Abstract

With 61 species occurring mostly in the Andes and adjacent lowland areas, Stenocercus lizards represent one of the most widespread and well‐represented Andean vertebrate groups. Phylogenetic relationships among species of Stenocercus are inferred using different datasets based on mitochondrial DNA sequence data of 35 species and morphological data of 59 species. Among morphological data, polymorphic and meristic/morphometric characters are coded under the frequency parsimony and gap‐weighting methods, respectively, and the accuracy of these methods is tested. When both types of characters are included, the resulting tree topology is more similar to the topologies obtained from analyses of DNA sequence data than those topologies obtained after exclusion of one or both types of characters. The phylogenetic hypotheses inferred including 59 species of Stenocercus (dataset 1) and excluding those species for which DNA data were not available (dataset 2) are generally congruent with each other, as well as with previously published hypotheses. The most parsimonious tree obtained from analysis of dataset 2 is used in a dispersal‐vicariance analysis to infer ancestral areas and major biogeographical events. Species of Stenocercus are divided into two major clades. Clade A has diversified mostly in the central Andes, with a few species in the northern Andes and one species in the southern Andes. Clade B is more widespread, with species in the northern, central, and southern Andes, as well as in the Atlantic lowlands and Amazon basin. The most recent common ancestor of Stenocercus is inferred to have occurred in the eastern cordillera of the central Andes. Given morphological similarity and altitudinal distribution of some species nested in a northern‐Andes clade, as well as the relatively recent uplift of this Andean region, it is possible that species in this clade have diverged as recently as the mid‐Pliocene.

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

“…Of the methods used to code continuous characters, Torres-Carvajal's (2007) version of the gap-weighting method resulted in the most left-skewed distributions (g 1 values most negative; g g 1 5 20.301 6 0.004), indicating that this method yielded characters that contain the most phylogenetic information. The method of Thiele (1993), with 101 character states, followed (g g 1 5 20.269 6 0.003), and the same method with 6 character states resulted in the least skewed distributions (g g 1 5 20.253 6 0.003).…”

“…Given that the default cost of character state transformation in PAUP* is 1, this approach requires weighting non-continuous characters by 1,000 to maintain equal weights among characters. Torres-Carvajal (2007) suggested a modification of Wiens' approach in which the term 1,000 in Wiens' equation is replaced by 1; this practice results in step matrices containing scores between 0 and 1 (rather than 0 and 1,000) and does not involve reweighting the non-continuous characters.…”

“…Step matrices, in which transformation costs between states are differences between these scores, were generated in PAUP* as described by Torres-Carvajal (2007).…”

“…Polymorphic characters (including presence/absence and meristic) were coded using two different approaches: the MANOB approximation (Manhattan distance, observed frequency arrays) of the frequency parsimony method described by Berlocher and Swofford (1997), and the majority or modal condition. Berlocher and Swofford's (1997) method was originally described for allele frequency data (see also Swofford and Berlocher, 1987) but has been applied to polymorphic morphological characters (Wiens, 2000;Brandley and de Queiroz, 2004;Torres-Carvajal, 2007). Under this approach, each taxon with a unique combination of allele (character state) frequencies is assigned a different character state, and changes between states are assigned costs equal to the Manhattan distances between those states using step matrices, which are analyzed under the parsimony criterion.…”

“…We did this by directly comparing g 1 values as estimates of the amount of hierarchical information recorded by each coding method. Values were compared between methods for coding continuous characters (Thiele's [1993] gap-weighting method with 101 character states versus Torres-Carvajal's [2007] version of the gap-weighting method versus Thiele's [1993] gap-weighting method with 6 character states) and between methods for coding polymorphic characters (frequency arrays using Manhattan distance step matrices versus using modal conditions). The g 1 values were estimated in PAUP* from data sets containing only those characters coded with the method being tested; g 1 values for each method were estimated from 10 samples of 500,000 random trees, and differences between the g 1 values were evaluated using t tests.…”