Exhippolysmata, , AND

Baeza, J. Antonio & Fuentes, M. Soledad, 2013, Exploring phylogenetic informativeness and nuclear copies of mitochondrial DNA (numts) in three commonly used mitochondrial genes: mitochondrial phylogeny of peppermint, cleaner, and semi-terrestrial shrimps (Caridea: Lysmata, Exhippolysmata, and Merguia), Zoological Journal of the Linnean Society 168 (4), pp. 699-722 : 715-717

EXHIPPOLYSMATA, AND MERGUIA

The first two phylogenetic analyses conducted with peppermint, cleaner, and semiterrestrial shrimps demonstrated that: (1) semiterrestrial shrimps from the genus Merguia represent the sister group to a second natural clade composed by shrimps from the genera Lysmata and Exhippolysmata ; (2) the genus Lysmata is paraphyletic, and includes the genus Exhippolysmata ; (3) the monophyletic clade composed of species of Lysmata and Exhippolysmata can be divided into four well-supported subclades (Neotropical, Cleaner, Cosmopolitan, and Morphovariable; Baeza et al., 2009a, b; Baeza, 2010a). A third analysis conducted by Fiedler et al. (2010), which used many of the sequences generated in the first two studies, confirmed the inferences from Baeza et al. (2009a) and Baeza (2010a). Importantly, a preliminary exploration by Fiedler et al. (2010) of the relationship among 12 species of Lysmata , one species of Exhippolysmata, and one species of Merguia using the nuclear gene 28S suggested a close relationship between Exhippolysmata and shrimps from the genus Parhippolyte , with the latter from another family (Barbouridae), rather than between Exhippolysmata and Lysmata , as suggested by phylogenetic reconstructions using mtDNA.

The present study, using a total-evidence approach that included three different mtDNA gene fragments, 34 species, and a concatenated data set of 1403 characters, confirms most of the previous inferences about the intra- and intergeneric relationships of Lysmata and Exhippolysmata . For instance, the monophylies of the four currently accepted subclades within Lysmata + Exhippolysmata are confirmed with high support. These analyses additionally support previous results indicating that the genus Lysmata is paraphyletic, as it includes the genus Exhippolysmata ( Baeza et al., 2009a, b; Baeza, 2010a; Fiedler et al., 2010). This last observation, however, disagrees with preliminary phylogenetic inferences based on one nuclear gene ( Fiedler et al., 2010). Finally, most of the relationships previously found within monophyletic clades, including those between geminate species, were confirmed with high support during the present analyses.

On the other hand, this phylogeny failed to support the sister relationship previously suggested to exist between semiterrestrial shrimps from the genus Merguia and the Lysmata + Exhippolysmata monophyletic clade (see Baeza et al., 2009a, b; Baeza, 2010a). Importantly, the rates of evolution herein detected for the three mitochondrial genes studied and their PI profiles (see below for a detailed discussion about PI profiles) suggest that these mtDNA markers do not have enough phylogenetic information to resolve deep nodes such as those previously suggested among Lysmata and Merguia , and the different subclades within Lysmata + Exhippolysmata . We argue in favour of additional studies examining the relationships among peppermint, cleaner, and semiterrestrial shrimps using not only mitochondrial but also nuclear genes. The PI of nuclear genes is assumed, or, in some cases, has been demonstrated to peak at greater sequence divergence compared with more recent sequence divergence in mitochondrial genes ( Colgan et al., 1998; Mahon & Neigel, 2008; Klopfstein, Kropf & Quicke, 2010). Therefore, these nuclear genes should successfully resolve nodes reflecting deep phylogenetic relationships among the genera above. Of particular interest is the exploration of phylogenetic relationships among the genera Lysmata , Exhippolysmata , Merguia , and the allied genera Barbouria , Calliasmata , Lysmatella , and Parhippolyte . Disentangling the phylogenetic relationships among these genera will aid in revealing the evolutionary origin of protandric simultaneous hermaphroditism, the most remarkable sex expression pattern yet reported in the Caridea ( Bauer & Holt, 1998; Fiedler, 1998; Baeza, 2009).

Protandric simultaneous hermaphroditism has been demonstrated in at least 19 out of 40 recognized species of shrimps from the genus Lysmata and Exhippolysmata (see Baeza, 2009, and references therein). In these sequentially simultaneous hermaphroditic species, juveniles invariably mature as functional males first, and later they become functional simultaneous hermaphrodites capable of reproducing both in the male and female role ( Bauer & Holt, 1998; Fiedler, 1998; Baeza, 2009). At present, the information available on the phylogenetic affinities of Lysmata , Exhippolysmata , Parhippolyte , and other closely related genera does not allow us to infer with certainty the actual number of times that protandric simultaneous hermaphroditism has evolved in carideans, and the conditions favouring this remarkable sexual system. A new phylogeny of peppermint, cleaner, and semiterrestrial shrimps, and allied genera, using nuclear markers will also allow the examination of the evolutionary stability of mixed sexual systems (see Weeks et al., 2006), and their effect in the net diversification rate ( Schluter, 2000).

PHYLOGENETIC INFORMATIVENESS OF MTDNA MARKERS

Herein, a comprehensive set of tools was used to investigate the phylogenetic utility and information content of three widely used mitochondrial genes for phylogenetic reconstruction in marine arthropods. First, using PI profiles, the phylogenetic informativeness of COI, 12S, and 16S mtDNA fragments was explored across historical time in peppermint, cleaner, and semiterrestrial shrimps. PI profiles were found to vary markedly among the three mitochondrial gene fragments studied. Importantly, the profile of the COI gene fragment featured a relatively sharp peak at a shallow relative time, whereas the profiles for the 16S and 12S genes were relatively levelled over a broad range of sequence divergence, and did peak, but not abruptly, at similar, relatively moderate to deep, relative times. Thus, compared with 16S and 12S, COI has almost three times more predicted power to solve shallow nodes and two times more predicted power to solve deep nodes in the studied phylogeny.

Considering these marker-specific PI profiles, COI is expected to have the highest information content and greatest capability for resolving both shallow and deep nodes (closer to and far away from the tips, respectively) of the phylogenetic tree studied (see Townsend, 2007). Nonetheless, the analysis of boostrap values for nodes in trees generated using the 16S and 12S gene fragments separately were equally likely, or only slightly inferior (in the case of 12S), to resolve shallow nodes when compared with the COI gene fragment. Also, 16S and 12S more frequently resolved deep nodes in the phylogenetic tree studied when compared with the COI gene fragment. At first glance, this unexpected discrepancy between the predicted power and the actual utility of the different gene markers challenge the notion that PI profiles represent a tool for quantifying the phylogenetic information content in molecular markers.

The discrepancy between the expected power (as predicted by PI profiles) and the actual utility of the gene fragments studied can be further explored by assessing the effect of the COI gene fragment on the magnitude of the bootstrap support values. Given that the COI marker is predicted to have the highest information content (compared with 12S and 16S) for resolving shallow (and deep) nodes in the phylogeny studied, the removal of the COI marker from the total-evidence analyses is expected to diminish (considerably) the ML bootstrap and BI support values for many nodes in the resulting new phylogeny (using only the 16S and 12S markers). In disagreement with the prediction above, support values for recent nodes in the phylogenies did not decrease considerably in phylogenies reconstructed using the combined 16S + 12S data set (analysis not shown here). This analysis reveals the similar accuracy of the three genes in resolving shallow nodes in the phylogenetic tree studied.

Discrepancies between informativeness (as predicted by PI profiles) and the observed performance of genes has been previously noted by other authors that have compared the utility of COI and one nuclear gene (arginine kinase, AK) in brachyuran crabs ( Mahon & Neigel, 2008), and COI and two different nuclear gene segments (the D2 and D3 regions of 28S) in parasitoid wasps ( Klopfstein et al., 2010). Various reasons might explain the discrepancy between the expected power and the observed utility of the genes observed herein, as well as in previous studies (e.g. Mahon & Neigel, 2008; Klopfstein et al., 2010). For instance, PI analysis requires the forcing of a single optimal tree during the analysis, and the tree is assumed to depict the true evolutionary relationship among the species studied ( Townsend, 2007). If the tree is inaccurate, the analysis of PI profiles is expected to be flawed. Various strategies were used in this study to tackle this issue. First, we used two robust trees (ML and BI analyses) that were generated using a total evidence analysis (in which the three molecular markers were used together). Second, spectral analysis of support and conflict for splits in the data set was employed, as it permits the exploration of phylogenetic signal in markers independently of any phylogenetic reconstruction ( Hendy & Penny, 1993; Hendy et al., 1994; Lento et al., 1995). Finally, neighbour-nets were used to visualize and further study phylogenetic signal and conflict or the distribution of information in the data set.

Importantly, an ANOVA on support values for internal splits derived from the split decomposition analysis indicated no significant differences in phylogenetic signal among the three molecular markers studied. Spectral analysis also indicated that support was consistently greater and conflict was consistently lower for various internal splits in the 16S and 12S mtDNA markers, compared with those in the COI gene. Finally, the two-dimensional graphical display of neighbour-nets clearly indicated more conflicting signals in the COI data set than in the 12S and 16S data sets; neighbour-nets from the 16S and 12S markers were more tree-like compared with that of the COI marker, and the 16S neighbour-net was the best resolved (in relative terms), and showed segregation of species in all four clades previously detected with the ML and BI phylogenetic analyses. It could be argued that neighbour-nets should naturally become more star-like when an increasing number of taxa are used for their construction (see Kennedy et al., 2005). That the 16S neighbour-net, containing 34 species, is more resolved and less star-like than the 12S and COI neighbour-nets, constructed with a smaller number of species (26 and 25 species, respectively), represents clear evidence of the greater phylogenetic utility of the 16S gene compared with the other two gene fragments studied. Overall, the results from the split-decomposition analysis and neighbour-nets disagree with the results from the PI analysis. Analyses of Lento plots and neighbour-nets suggest that 16S and 12S gene fragments are equally, or even more, powerful than the COI fragment in resolving phylogenetic relationships, and in resolving nodes at all levels in the phylogeny of the genera Lysmata and Merguia .

Reasons other than phylogenetic uncertainty that might explain the discrepancy between the expected power and the observed utility of the genes studied include deviations from a strict molecular clock and tree-building method ( Townsend, 2007; Klopfstein et al., 2010). However, recent sensitivity analyses have demonstrated that the performance of PI profiles (as a tool for predicting gene utility) is robust to deviations from a strict molecular clock and tree-building method ( Klopfstein et al., 2010). Thus, the issues above are not considered to be a major problem in this study, and do not appear to explain (at least, to a great extent) the discrepancy between the expected power and the actual utility of the gene fragments studied. Importantly, the contrast between the expected power and the observed utility of the three genes studied might be interpreted as statistical problems in the PI profiling methodology proposed by Townsend (2007). Indeed, that the gene fragment with the greatest predicted power is the one that has evolved the most quickly (COI) is in agreement with previous studies indicating that PI profiles systematically overrate the performance of quickly versus slowly evolving markers ( Klopfstein et al., 2010).

Overall, the combined analysis of support values, split-decomposition analyses, and neighbour-nets provided insight, and were useful to explore the overall utility of the genes. Altogether, these analyses suggest that in shrimps from the genera Lysmata , Exhippolysmata, and Merguia , all the mtDNA fragments studied are equally useful for resolving phylogenetic trees. Importantly, this study also suggests that the shape of PI profiles are not useful to estimate the overall utility of genes: sharp PI curves with abrupt peaks do not necessarily characterize the genes most useful for resolving phylogenetic relationships on narrow timescales.