Phylogenetic Analysis

Individual 12SrRNA, 16SrRNA and COI equences were aligned with Clustal and then refined by eye using the graphical interface of MacClade4.08 (Maddsion and Maddison 2005) to correct obviously homologous regions that Clustal failed to recognize. These include cases which required a simple shift in the position of a gap to avoid a stop codon not present in other sequences or to complete a codon whose members flanked the Clustal inferred gap. We refined the rRNAs alignments with Rcoffee (Notredame et al. 2000) to account for secondary structure. Because Rcoffee runs are restricted to 50 taxa and 1000 base pairs, the Clustal alignments were divided into smaller subsets, aligned with Rcoffee and then concatenated for further analysis. Alignments within Rcoffee are guided by secondary structure characteristics (Notredame et al. 2000). The color-coded CORE indices were used to identify the best among alternative alignments that include the Clustal and Rcoffee alignments and hand refined versions of each.

Sequences were concatenated with MacClade4 (Maddsion and Maddison 2005) and then optimized using MrBayes (Huelsenbeck and Ronquist, 2001) with GTR+I+G maximum likelihood model parameters estimated separately for each gene. Each analysis comprised two simultaneous runs with four chains each. Two million generations reduced the average standard deviation of the split frequencies below 0.01. Plots of the number of generations against the maximum likelihood scores indicated equilibrium. Further diagnostics included the potential scale reduction factor (PSRF) that measures the fit of branch length and all parameters. Trees and parameters from the first 25% of the generations were discarded (the burn in) after completion of the MCMC (Markov Chain Monte Carlo) search.

For the maximum likelihood analyses of the individual genes, we estimated parameters for GTR+I+G models of sequence evolution and optimized the tree and using Phylml (Guindon and Gascuel, 2003). Our analysis of 1000 bootstrap replicates provided measures of support on the clades of the maximum likelihood trees. For the concatenated sequences of several genes we used RAxML (Stamatakis, et al. 2008) to optimise the maximum likelihood tree using the GTR+I+G model with parameters estimated separately for each gene and to analyse the bootstrap replicates

Finally we tested various alternative topologies (e.g. monophyly of the Turris species) using Shimodaira-Hasegawa tests with GTR+I+G optimised parameters in PAUP4b10 (Swofford, 2002) to determine whether any differed significantly from the optimal tree. Rather than simply moving the two intact Turris clades together, we optimized with maximum likelihood the relationships among all the Turris species simultaneously within the monophyly constraint. To ensure that these were tests of topology alone, we constrained the topologies but not the branch length estimates.

Molecular analysis

The trees inferred from the three individual genes, COI, 12S and 16S are poorly resolved (Supplementary Figure 1). On the other hand, the tree inferred from the concatenated sequence of the three genes (Figure 5) is comparatively well resolved. Contrary to traditional taxonomy, the tree suggests that the genus Turris is not monophyletic. This tree comprises a clade, hereafter referred to as Annulaturris, that includes Turris annulata, Turris cristata, Turris cryptorrhaphe, Turris nadaensis and Turris undosa to the exclusion of a clade comprising Lophiotoma and Gemmula species with the other Turris species (T. babylonia, T. dollyae, T. grandis, T. normandavidsoni, T. spectabilis T. assyria). The hypothesis that the Turris species are monophyletic is rejected (Shimodaira-Hasegawa test, ln likelihood difference=62.3, p≤0.0001).

Open in a separate window

Figure 5

Optimal tree inferred using gene-partitioned Bayesian methods from the concatenated alignments of 12SrRNA, 16S rRNAs and COI genes. The phylogeny describes the evolutionary relationships among Turris species and selected Lophiotoma and Gemmula species. Turridrupa bijubata is the out-group species. Some Turris taxa are labeled to indicate radulae and central tooth type relative to the optimal molecular tree. Branches are labeled with posterior probabilities greater than 50% from the Bayesian analysis of the three genes. Maximum likelihood bootstrap values over 50% appear to the right of the posterior probabilities. Clade A – Annulaturris, provisionally given subgeneric status. Clade B – Turris s.s., provisionally given subgeneric status.

Reconciling molecular and radular evidence

Our findings are consistent with our optimal molecular tree (Figure 5) supporting separation of the Turris species into two distinct clades within the Turrinae. The molecular tree supports a clade including Turris annulata, T. cristata, T. cryptorrhaphe, T. nadaensis and T. undosa (94% posterior probability, 63% ML bootsrap support) as a distinct clade within the Turris. A provisional taxonomic solution is to use a name for this clade previously provided by Powell, Annulaturris (Powell, 1966). Together with topology tests (Kishino-Hasegawa and Shimodaira-Hasegawa) that reject monophyly of the Turris species, it is clear that the proposed Annulaturris species should be considered to be well seaparated from the other Turris species.

The tree is a striking example of how misleading shell morphology can be; Turris cristata, with similar color pattern to the forms assigned to the Turris nadaensis complex is often mistaken for Turris nadaensis, while Turris cryptorrhaphe is appears unmistakably different. Nevertheless, based on all three genes, it is clear that Turris cryptorrhaphe is more closely related to the various forms in the nadaensis/undosa clade than to Turris cristata.

Based on shell morphology, three specimens conventionally assigned to Turris babylonia fall into two divergent branches within the larger clade comprising T. babylonia, T. dollyae, T. grandis, T. normandavidsoni, T. spectabilis. Powell (1964) regarded Turris babylonia 1008 and 1012 as the “typical form”. The specimen 1010 belongs to a separate species, which has been regarded as a distinct form of T. babylonia (“niniveh form”, Olivera,, 1999). The type material for Turris babylonia, stored in the Linnaean society in London, consists of two syntypes. One is conspecific with our specimens 1008 and 1012 while the other is rather similar to our Turris assyria (or babylonia, specimen 1010). The fact that the type appears to be a mixture of two species is taxonomically problematic with resolution remaining for future studies. And yet while the true identity of the type specimen for Turris babylonia remains controversial (R. Kilburn, personal communication), the two morphospecies both widely assigned to Turris babylonia clearly belong in the same clade (Turris s.s. and not Annulaturris).

Nevertheless, further molecular, anatomical and shell morphological analyses are indicated. For example, the type species of Annulaturris, T. amicta (E.A. Smith, 1877) was not available for this analysis. Although supported by morphological data, the inference that T. amicta species falls in the clade (Figure 5) that includes Turris cristata and Turris annulata is speculative. The relationships of subtropical species, such as Turris orthopleura and Turris ruthae, also needs to be assessed. As discussed by Kilburn (1983), these species have shell morphological and radular characters that suggest that they are not in Turris s.s.

Compared to the relatively unreliable shell morphology, the combination of differences in radular morphology and multiple molecular markers provides a firmer definition for the type species of the family Turridae, and the clade to which the type species belong, defined here as Turris, excluding the Annulaturris ssp. We predict that these new insights into the phylogeny of Turris and its close relatives will be reflected in the analysis of the gene products expressed in their venoms, an analysis that we have begun. A phylogenetic framework has been an extremely useful guide to the discovery of unique pharmacological agents from Conus venoms. Similarly, an understanding of evolutionary history, summarized by our phylogeny of the Turridae,, will extend future venom research.