Material and methods
Taxonomic sampling and molecular datanext section
Species sampling was made according to the nomenclature described by Bisby et al. (2009). Authorities for each species are indicated in On-line supplementary table S1. We did not include taxa for which there was evidence for a hybrid origin, such as Mauremys iversoni, Mauremys pritchardi, Ocadia glyphistoma, Ocadia philippeni and Sacalia pseudocellata (Parham et al., 2001; Wink et al., 2001; Spinks et al., 2004; Stuart and Parham, 2007). To obtain a phylogeny as robust as possible, we included both mitochondrial (mtDNA) and nuclear (nuDNA) genes. We compiled all complete mitochondrial sequences available in October 2009 from GenBank (On-line supplementary table S2). We also compiled the sequences of five mitochondrial genes (12S, 16S, COI, NAD4, cytB) and four nuclear genes (R35, c-mos, RAG1 and RAG2). These particular genes were those for which the available number of sequences was highest. We combined the different mtDNA and nuDNA sequences obtained for each species into a single matrix. Only regions of straightforward alignment were taken into account. The length of the final alignment was 20,000 nucleotides (available via: http://purl.org/phylo/treebase/phylows/study/TB2:S12290).
To root the phylogenetic tree, we used total mtDNA for four outgroup species (Crocodylus porosus for crocodiles, Pycnonotus sinensis for birds, Lacerta viridis for squamates, and Sphenodon punctatus for Rhynchocephalia). Because nuDNA sequences were not always available for the same taxa, we hereafter refer to outgroup species as ‘crocodiles’, ‘birds’, ‘squamates’ and ‘sphenodons’ (references for all outgroup sequences are shown in Table S2). We root the tree either as sister-group of a clade that includes ‘birds’, ‘crocodiles’, ‘squamates’, and ‘sphenodons’ (Reisz and Laurin, 1991; Laurin and Reisz, 1995; Lee 1997, 2001; Lyson et al., 2010) or as sister-group of ‘birds’ and ‘crocodiles’ (Zardoya and Meyer 1998, 2001). It does not change the topology within the turtle clade.
We used the maximum likelihood algorithm of PhyML (Guindon and Gascuel, 2003) to conduct the phylogenetic analysis, starting with a parsimony tree. Parameterization of PhyML was performed using jModelTest 0.1 (Posada, 2008) to select a model of nucleotide substitution. To quantify branch support, we report confidence values (cv) as the result of an approximate likelihood-ratio test performed by PhyML (Anisimova and Gascuel, 2006). Nodes with cv < 0.5 have been considered as non-resolved and are polytomized.