3.1. Mitochondrial genome organization and structural features The complete nucleotide sequences of the L-strand of the mt genomes of three squamates were determined anew. The total lengths of the new squamate mt genomes were 17.035 bp forB. cinereus, 16.593 bp forT. mauritanicaand 17.479 bp forA. fragilis. These lengths are within the range reported for other squamate mt genomes (Kumazawa, 2007). All three mt genomes encoded for two rRNA, 22 tRNA and 13 protein-coding genes. The gene organization of the three newly determined mt genomes conformed to the vertebrate consensus mt gene arrangement (Boore, 1999; Jameson et al., 2003).
However, in the mt genome ofB. cinereus, the tRNA Pro gene was separated from the tRNA Thr by a tandem repeat.
The differences in length among the three mt genomes determined in this study were mainly due to tandem repetitions located in the control region (as described in other vertebrate mt genomes; see e.g.
San Mauro et al., 2004a). The number, length, and motif of tandem repeats in the control region differed across taxa.B. cinereusshowed seven repeats of 59 bp each at the 5′ end of the control region; A. fragilisexhibited 10 repeats of 54 bp each, andfive complete (60 bp each) plus one incomplete repeats at the 5′and 3′ends of the control region, respectively; T. mauritanicashowed seven repeats of 75 bp each and 11 complete (66 bp each) plus one incomplete repeats at the 5′ and 3′ends of the control region, respectively.
has been associated with the adjacent presence of
a nonfunctional origin of replication of the light strand (OL)in amphisbaenian bipedids (Macey et al., 2004) and acrodonts (Macey et al., 2000; Amer and Kumazawa, 2005). In contrast to these observations, the OL folds perfectly into a stem-loop secondary structure in B. cinereus, as in amphisbaenian trogonophids and amphisbaenids (Macey et al., 1997, 2004).
3.2. Squamate phylogenetic relationships
Phylogenetic relationships among squamate main lineages were inferred based on the allmt dataset that was analyzed at the amino acid level. The recovered BI tree (−lnL=94,708.37) using best-fit models for the four partitions (mtREV+Γ+I in all cases) is shown inFig. 2.ML analyses based on the same data set under the mtREV+Γ+I model arrived at the same topology. Mammals were used as outgroup, and turtles were recovered as a sister group to Archosauria (crocodiles + birds;Zardoya and Meyer, 1998; Hedges and Poling, 1999; Kumazawa and Nishida, 1999; Fig. 2). Within Lepidosauria, the tuatara was recovered as a sister group to Squamata (Rest et al., 2003; Townsend et al., 2004;Fig. 2) with high statistical support (ML, 87%; BI, 100%).
Within Squamata, the monophyly of Iguanidae, Anguimorpha, Amphisbaenia, Gekkota, Serpentes, and Acrodonta received high statistical support both with BI and ML (Fig. 2). Scincomorpha (including Scincoidea and Lacertoidea) was supported with BI but not with ML. Thus far, the monophyly of Scincomorpha had received support only from morphology (Estes et al., 1988; but seeLee, 1998) whereas no previous molecular phylogenetic analysis (Townsend et al., 2004; Vidal and Hedges, 2005; Kumazawa, 2007) was able to recover it. However, the recovered monophyly of Scincomorpha still remains tentative since it is not robust to changes in the taxon sampling of the outgroup (see below). Phylogenetic relationships among the main squamate lineages could not be resolved with ML but received strong support with BI (above 95%). However, it is important to note that Bayesian posterior probabilities are well known to be significantly higher than corresponding non-parametric bootstrap frequencies leading to overcredibility of the recovered nodes (Suzuki et al., 2002; Erixon et al., 2003). According to the BI tree (Fig. 2), Acrodonta and Serpentes form a clade (see also the NADH2 tree of Townsend et al. (2004), which is the sister group of the remaining squamate lineages. Within these, Gekkota were thefirst branching out, followed by Amphisbaenia, and a clade including Anguimorpha as sister group of Scincomorpha + Iguanidae (Fig. 2).
As reported in previous molecular studies (Vidal and Hedges, 2005; Böhme et al., 2007; Kumazawa, 2007), the newly reconstructed phylogeny of squamates does not support the Iguania–Scleroglossa split (Estes et al., 1988). According to overall molecular evidence, thus, it is rather arguable that changes in prey capture constitute a major unique evolutionary shift in squamates (Vitt et al., 2003). The sister group relationship of Serpentes + Acrodonta (Böhme et al., 2007; Douglas et al., 2006) and the relative basal position of both taxa within Squamata is in clear disagreement with most previous molecular phylogenetic studies, which placed Acrodonta in a more derived position in the squamate tree as sister group of Iguanidae based on the analyses of both nuclear (Townsend et al., 2004) and mt (but without including Serpentes in the phylogenetic analyses;Kumazawa, 2007) sequence data. Such sister group relationship would be in agreement with the morphology-based Iguania hypothesis (Estes et al., 1988). On the other hand, Serpentes are placed as sister group of either Anguimorpha (Townsend et al., 2004) or Anguimorpha + Iguanidae (Vidal and Hedges, 2005) based on nuclear evidence. Both Acrodonta and Serpentes exhibit relatively long branches in mt-based phylogenies, and their sister group relationship could be due to a long-16 E.M. Albert et al. / Gene 441 (2009) 12-21 branch attraction artifact (Felsenstein, 1978; see below). Phylogenetic relationships among the remaining squamate lineages as shown in Fig. 2strongly differed from those recovered by previous studies based on both nuclear (Townsend et al., 2004; Vidal and Hedges, 2005) and mt genome (Böhme et al., 2007; Douglas et al., 2006; Kumazawa, 2007) datasets, particularly regarding the relative phylogenetic position of Lacertidae and Amphisbaenia. In fact, the phylogenetic position of Amphisbaenia is not fully resolved, and varies depending on the study. Morphology-based phylogenetic studies support Amphisbaenia affinities with Dibamidae (Lee, 1998, 2000)or Scincomorpha (Schwenk, 1988). Most molecular analyses recover Lacertidae as the sister group of Amphisbaenia (Townsend et al., 2004; Vidal and Hedges, 2005; Kumazawa, 2007), but in some analyses based on complete mt genome sequence data, Amphisbaenia is grouped with either Gekkota (Zhou et al., 2006) or with Serpentes + Acrodonta (Douglas et al., 2006). More recent analyses based on multiple nuclear loci also failed to concludefirmly on the phylogenetic position of Lacertidae and Amphisbaenians within Squamata (Townsend et al., 2008).
Results of AU, SH, and KH tests of alternative tree topologies (Table 1; Supplementary Material) rejected the Scleroglossa (Estes et al., 1988) and Toxicofera (Vidal and Hedges, 2005; Fry et al., 2006) hypotheses. More generally, any other hypothesis that implied breaking up the Serpentes–Acrodonta clade was strongly rejected (not shown). This may be a consequence of an underlying long-branch attraction effect between Acrodonta and Serpentes (see below), and is in agreement with both the short length of the internodes connecting squamate main lineages, and the generally moderate bootstrap support of basal squamate phylogenetic relationships. The tests also indicated that alternative hypotheses placing Amphisbaenia as a sister group of either Gekkota or Serpentes + Acrodonta were not significantly different from the unconstrained hypothesis whereas putative sister group relationships of Amphisbaenia with either Lacertidae or Scincomorpha were rejected. As indicated by previous studies (San Mauro et al., 2004b), p-values from AU and KH are markedly correlated, whereas the SH test is always more conservative.
3.3. Effects of molecular marker choice, outgroup selection, and taxon coverage
Our best phylogenetic hypothesis for squamate relationships (Fig. 2) clearly differed from those previously reported based on different molecular markers, outgroups, and/or taxon samplings. In order to disentangle the relative contribution of these variables to the observed discrepancies, we performed further phylogenetic analyses.
Incorporating X. laevisinto the analyses tested the influence on the recovered topology of using a non-amniote species as outgroup (mtXenopusdataset). The resulting tree was similar to the one shown in Fig. 2, regarding the basal position of Acrodonta + Serpentes, and the subsequent branching out of Gekkota (Fig. 3A). However, the main phylogenetic relationships among squamates based on complete mt genomes. In order to circumvent this shortcoming, it would be desirable tofind snake and acrodont species exhibiting slower mt evolutionary rates, in order to incorporate them into the phylogenetic analyses. It would be also important to ensure that all main squamate lineages (e.g. dibamids) are represented in the mitogenomic phylogenetic analyses, and that each lineage encompasses thorough taxon coverage of their diversity (particularly for those lineages such as e.g.
Scincomorpha that may not be monophyletic). Given that the
phylogeny of Squamata shows relatively short internodes, phylogenetic accuracy could be improved by analyzing more genes (e.g., derived from ongoing nuclear genomic sequencing initiatives;Shedlock et al., 2007) with appropriate evolutionary rates that maximize the number of informative sites.
3.4. Squamate divergence dates
Although no fossil record of Squamata has been found before the early Jurassic,Evans (2003)suggested that the presence of crowngroup Rhynchocephalia in the Late Triassic (Sues and Olsen, 1990; Benton and Donoghue, 2007) could provide some evidence for a Triassic (250–206 Mya) origin of Squamata. Molecular clock BI analyses based on mt genome sequence data (four calibration points, Kumazawa, 2007) and nuclear sequence data (five calibration points, Vidal and Hedges, 2005) suggested a Permian origin of Squamata, and dated the radiation of major squamate lineages back to the Triassic– Jurassic times.
Our analyses based on the allmt dataset and nine calibration points rendered similar datings, and estimated that divergence of the main squamate lineages took place in a short window of time of less than 60 Myr around 200 Mya (Fig. 5). This rapid radiation pattern likely causes that internal branches of the Squamata tree are relatively short compared to the (much longer) terminal branches, thus rendering phylogenetic reconstruction of internal relationships particularly challenging. As with most other molecular dating studies (Benton and Ayala, 2003; Reisz and Müller, 2004), our Bayesian time estimates appear to be older than the ages deduced from the fossil record (Evans, 2003). Interestingly, the confidence intervals of the estimates based on PL were significantly shorter than those based on Bayesian analyses (one-way ANOVA F1,64
=98.81;pb0.001; Fig. 5), but the mean
estimates based on both methods were not significantly different (one-way ANOVAF1,64
=0.750;p=0.389;Fig. 5). A strong correlation
was detected between PL and Bayesian estimates (r=0.99; see Supplementary material). For most nodes, PL estimates were 10– 25 my older than Bayesian estimates. However, for the Acrodonta + Serpentes, Acrodonta, and Serpentes nodes, the PL estimate were about 40 my older than Bayesian estimates (Fig. 5). Given the use of similar input topologies and calibration constraints, the discrepancy between the two methods may be related to their different assumptions about rate change, and different implementations of models of evolution, branch length and confidence interval estimation, and use of prior information (Welch and Bromham, 2005). The difference in dating between PL and Bayesian methods is exacerbated at the nodes leading to long branches. However, it is not possible based on our results to conclude which, the PL or the Bayesian approach, is better suited to deal with extreme cases of long branches. Moreover, in a recent study (Smith et al., 2006), PL estimates are consistently younger than Bayesian estimates for the same equinoderm dataset. Hence, it is not easy to discern a clear trend on which method provides older or younger ages.
According to the estimated dates, and in agreement with previous studies (Vidal and Hedges, 2005; Kumazawa, 2007), we can conclude that the formation of the major squamate lineages predated the breakup of Pangaea. A similar pattern was found recently for amphibians (San Mauro et al., 2005), and contrasts with the more recent diversification of mammals, which was greatly influenced by continental fragmentation of the Pangea supercontinent during the Cretaceous (Wildman et al., 2007).