Evolution of the mitochondrial genome in snakes: Gene rearrangements and phylogenetic relationships

Characteristics of the snake mitochondrial genomes

By comparing the gene arrangements in 14 known snake mitochondrial genomes, six types of gene organization were summarized and shown in Figure 1. Type I and II represent gene organizations of two scoleophidian snakes. They both have similar organization to that of typical vertebrate except the absence of an identifiable origin of light strand replication (O_L) in the scolecophidian mtDNAs. Moreover, in the mitochondrial genomeof Leptotyphlops dulcis (Type II), trnQ was translocated from the I Q M tRNA gene cluster to the W ANCY cluster (the tRNA genes are abbreviated by single letters representing the amino acids to be decoded, and the sense strand of the underlined tRNA genes is the heavy strand), giving rise to the IM plus W QANCY organizations [11]. The common features of the remaining four arrangements (type III to VI) are the duplication of the control region and translocation of trnL, which have been noted in previous studies [9, 10]. Six alethinophidian snakes from six families shared type III, and type IV represented an arrangement shared between Colubridae and Homalopsidae. In these two types, the functional trnP is located next to CRI (location in typical vertebrate mtDNA), with or without a pseudogene (P*) close to CR II (there are two control regions in most snake mitogenomes except scolecophidian snakes shown as type I and II). In several previous studies, the viperid snakes appear to have another gene rearrangement in which a functional trnP moved from the end of CR I to that of CR II [9, 10, 13]. In this study, the positional switch of trnP is found for another viperid taxon Deinagkistrodon acutus, and is not found for the nonviperid snakes (Fig. 1). Gene arrangements in viperids can be summarize as type V and VI, with only a difference in the pseudogene (P*).

Phylogenetic analyses

The results of the phylogenetic reconstructions are displayed in Figure 2. Our phylogenetic estimates strongly support Ramphotyphlops braminus being the sister lineage to all remaining species sampled, and that the scolecophidian snakes do not form a monophyletic grouping. All of the analyses were congruent in receiving the monophyly of Alethinophidia (c in Figure 2). Within this monophyletic clade, Henophidia and Caenophidia were clustered as monophyletic sister groups (k and d, respectively). Our estimates of relationships among the henophidians are similar to that of Dong and Kumazawa [9] in the rejection of the traditional expectations of a sister-group relationship between boids and pythonids (m in Fig. 2) and the deep divergence of non-macrostomatan cylindrophiids (l in Fig 2). Consistent with recent molecular phylogenies of caenophidian snakes [2, 3, 5], we find support for the Viperidae as the deepest diverging lineage within the Colubroidea, sister to a clade containing colubrids, elapids, and homalopsids. The position of Enhydris plumbea outside of the (Colubridae + Elapidae) cluster seems strongly supported by all phylogenetic methods (f in Fig. 2). Traditional placement of Enhydris plumbea within Colubridae based on morphology was not retrieved by our analyses. Node d, corresponding to the position of Acrochordus granulatus, is the only one not reconstructed in all tree-building methods. In Bayesian and ML tree based on nucleotide data set and Bayesian tree on amino acid data set, the placement of Acrochordidae to the sister group to Colubroidea was well supported.

To further substantiate the above premises, we performed the SH statistical test [14] on both amino acid and nucleotide data sets that allows comparison of alternative phylogenetic hypotheses. Results of the SH test strongly rejected the placement of Enhydris plumbea within the colubrids cluster (P < 0.001). The monophyly of scolecophidian snakes was not rejected using both amino acid and nucleotide acid data sets (P > 0.05).

Evolution of snake mitochondrial genomes

Based on the phylogenetic relationships among the tested snakes and the comparisons of their gene organizations (Fig. 1, 2), we estimated the processes of evolutionary events occurred in snake mitochondrial genomes. In early snake lineages (type I and II), gene arrangements are similar to that of typical vertebrate, but O_L was lost within the W ANCY tRNA gene cluster. Incompatible with the commonly accepted view on monophyly of scolecophidian snakes [15–18], our phylogenetic estimates strongly supports Ramphotyphlops braminus being the sister lineage to all remaining species sampled. Given that the monophyly of scolecophidian snakes was not rejected in SH statistical test, loss of O_L may occur in two different scenarios, independently (if nonmonophyly) or descend from a common ancestry (if monophyly). After the divergence of the Ramphotyphlops lineage, changes involving the I Q M cluster took place. First in type II, trnQ underwent a long distance translocation (~1.2 kb) from one gene cluster to another (Fig. 1, 2). Subsequently, in the early alethinophidian lineage, the control region was duplicated and trnL relocated to the I Q M cluster, giving rise to type III which is present in most alethinophidian snakes (including henophidians, Acrochordus and Naja). New types emerged during the split in Caenophidia. Type IV is found in two branches, Dinodon semicarinatus, Pantherophis slowinskii (Colubridae), and Enhydris plumbea (Homalopsidae), and characteristic changes (P*) likely appeared ahead of node f, which then disappeared in Elapidae. It is also conceivable that the present of P* was resulted from independent evolution in Colubridae and Homalopsidae. Distinct arrangements (type V and VI) were found in viperids, suggesting that trnP was translocated in early stage of the viperid radiations [9]. Type VI, with no pseudogenes close to CRI, was found in two paraphyletic taxa, suggesting that P* could have been independently eliminated.

Familial rank of Homalopsidae

The Homalopsinae have been generally recognized as a valid monophyletic clade within the Colubridae [19, 20] and assigned a subfamilial rank, despite they being assigned familial [1] or tribal [21] status historically. Recent molecular studies placed the Homalopsinae as the sister group to most other members of the Colubroidea [2, 3, 5], and a familial status has been reassigned accordingly [2–4].

In this study, the placement of Enhydris plumbea, a representative of the Homalopsidae, as the sister lineage to the Colubridae + Elapidae clade was strongly supported by all phylogenetic methods. Moreover, SH test strongly rejected the hypothesis that Enhydris plumbea falls within the colubrids cluster (P < 0.001). The familial rank of Homalopsidae is therefore considered well-supported. Our work for the first time establishes the monophyly and distinctiveness of this family with phylogenetic evidence derived from complete mitochondrial genome sequences.

Samples, DNA amplification, and sequencing

Snakes from three alethinophidian families and one scoleophidian family were sampled (Table 1). Total DNA was extracted from a small quantity (20 mg) of tissues by DNeasy Tissue Kit (Qiagen). Several short mtDNA fragments were amplified using Ex-Taq DNA polymerase (Takara) and sequenced in order to design taxon-specific primers. PCRs were performed in a MJ PTC-200 thermal cycler under the profile: 5 min at 95°C followed by 35 cycles of 95°C for 30 s, 50–55°C for 30 s, and 72°C for 90 s. PCR products of 1~2.5 kb were purified and then sequenced employing an ABI 310 or 3700 system with bi-directional and several internal primers. Short fragments were assembled into a continuous sequence. In the mtDNA sequences thus obtained, 37 individual genes were identified based on corresponding homologues from other vertebrates. Identification of tRNA genes was based on their secondary structures using software DNASIS 2.5 (Hitachi Engineering, Tokyo, Japan), whereas boundaries of rRNA genes and control regions were tentatively defined by the boundaries of adjacent coding genes. The mtDNA sequences, with annotations, have been deposited at GenBank (DQ343647–DQ343650).

Taxa, alignment and phylogenetic analyses

We assembled 14 serpent ingroups with complete mitochondrial genomes available, and chose 4 taxa from 4 saurian families (Amphisbaena schmidti [22]; Eumeces egregius [23];Iguana iguana [24]; and Varanus komodoensis [25]) as outgroups (Table 1). Two data sets were prepared for concatenated amino acid sequences and for concatenated light-strand nucleotide sequences of the 12 protein genes. Nad6, the only protein gene encoded by the light strand, has been excluded for increased proportion of T and G in all codon positions due to the strand-specific base composition bias of mtDNAs. Multiple alignments were analyzed with the Gblocks program [26] to select conserved amino acid residues, which was later used as a backbone to align the corresponding nucleotide sequences.

The level of saturation in the whole codons, and at the first, second, and the third codon positions was independently analyzed using scatter plot graphics, by comparing the uncorrected p-distance with the distance calculated by applying the best-fit evolutionary model (GTR + I + G) selected by the Modeltest 3.7 [27]. The third positions of the protein genes were removed from the nucleotide data set because of high substitutional rates and consequent saturation as a source of noise in phylogenetic analyses. Thus a final alignment of 6566 bases was obtained.

Phylogenetic analyses were carried out using maximum likelihood (ML), Bayesian (BI), maximum parsimony (MP) and neighbor-joining (NJ) methods. The ML analyses with the nucleotide data set were conducted with PAUP*4.0b10 by a heuristic search with TBR branch swapping with 10 random taxon additions. The general reversible model (GTR + I + G) and parameters optimized by Modeltest 3.7 were used. Bayesian phylogenetic analyses of the nucleotide sequences were performed with MrBayes 3.1 [28] using a GTR + I + G model. The Markov chain Monte Carlo process was set to run four chains simultaneously. Posterior probabilities were calculated from the majority-rule consensus trees constructed after excluding the burn-in.

ML analyses with the amino acid data set were conducted using PUZZLE 5.2 [29] with the mtREV24 substitution matrix and amino acid frequency estimated from the data set. The Bayesian analyses of the amino acid data were conducted with MrBayes 3.0 using the mtREV24 + I + G model and an empirical amino acid frequency. The Bayesian tree and posterior probability values were obtained using the same procedures described above.