A Transcriptomic-Phylogenomic Analysis of the Evolutionary Relationships of Flatworms

1We assembled coding sequence data from 55 animal species, including 27 species of platyhelminth. We identified 1,348 orthologous genes and produced a large (>107,000 positions) and taxonomically broad phylogenomic dataset (27 flatworm species from 11 orders) for the analysis of the phylogeny of this important and diverse group of animals. The dataset contains very few missing data (average 72% complete, measured as the percentage of positions with data present within the total alignment), especially in the case of the newly sequenced taxa (average 82% complete, all but two >68% complete). We used site-heterogeneous Bayesian tree reconstruction (PhyloBayes CAT+GTR+G4 model, which has site-specific equilibrium frequency profiles; Figure 1) and site-homogenous maximum-likelihood (ML) approaches (PhyML LG+G4 and RAxML CATGTR, which have homogenous equilibrium frequency profiles; Figures S1 and S2) to reconstruct the phylogeny based on these data. Most relationships within Platyhelminthes are robustly resolved as shown by concordance between different analyses, Bayesian posterior probabilities (Figure 1), jackknife resampling (Figure 2), and phylogenetic signal dissection (Figure 3).

2Our tree supports the now canonical view of Platyhelminthes as members of Lophotrochozoa, which was first shown using 18S rDNA data and has subsequently received strong support from multigene phylogenies (e.g.,). Of perhaps greater interest is the finding of a strongly supported sister group relationship between the two species representing the order Catenulida and the remaining Platyhelminthes: the Rhabditophora (Figures 1, 2, 3, and 4). Rhabditophora share the convincing molecular synapomorphy of two changes in mitochondrial genetic code, and we provide phylogenomic confirmation of the monophyly of Platyhelminthes (Catenulida+Rhabditophora). Surprisingly, a convincing phenotypic synapomorphy of Platyhelminthes is still lacking. We have not considered the xenacoelomorphs, originally part of Platyhelminthes, as they have been shown by various means not to be part of the protostomes.

3While our ML tree supports Platyzoa ((Platyhelminthes, Gnathifera)(Gastrotricha)) (Figure S1), with the rotiferans representing the larger group of Gnathifera, our Bayesian analyses, in common with two recent well-sampled phylogenomic studies of lophotrochozoan relationships, show largely consistent support for Gastrotricha and Platyhelminthes being grouped together with Nemertea, Annelida, and Mollusca; Rotifera are outside of this clade (Figure 1). In our CAT+GTR+G4 analysis, Nemertea are sister group of Platyhelminthes+Gastrotricha. The exception to this finding is in our jackknife analysis, where the position of gastrotrichs relative to platyhelminths and other lophotrochozoan phyla is unresolved (Figure 2). In the signal dissection experiment, the fastest-evolving quartile of the data (most susceptible to long-branch attraction [LBA] or LBA artifact) supports Platyzoa (Figure 3, Q4), and this gives credence to the view of Platyzoa as arising from such a systematic error. Adopting measures to counter this problem with selected slowly evolving genes and well-fitted models (CAT+GTR+G4) rejects Platyzoa (Figure 3, Q1 and Q2).

4To date, the identity of the basalmost branching group of Rhabditophora has not been settled, with Macrostomorpha and Polycladida vying for this position. Members of both of these groups possess the likely primitive character of spiral cleavage (absent in many more derived groups, see Figure 4) and also have entolecithal eggs, again a likely primitive character. Polycladida have a larval stage (present in both major clades of polyclads) that some consider homologous to the trochophore seen in several other lophotrochozoan phyla. Macrostomorpha have aflagellate sperm; this contrasts with the remaining Rhabditophora, including Polycladida, which typically have a biflagellate sperm with a 9 × 2 + “1” pattern of microtubules and on this basis have been grouped as Trepaxonemata. Our data strongly support Macrostomorpha as sister group of all other rhabditophoran orders. Macrostomorpha are excluded from the monophyletic Trepaxonemata with posterior probability of 1.0 (Figure 1), jackknife support of 1.0 (Figure 3), and PhyML “SH-like” support of 1.0 (Figure S1) as well as being found with pp = 1.0 in all four quartiles of the signal dissection experiment (Figure 3).

5Apart from Macrostomorpha and Polycladida, all rhabditophoran groups, including Lecithoepitheliata, are distinguished by ectolecithal eggs (yolk not incorporated into the embryonic blastomeres) and the associated characteristic (absent in Lecithoepitheliata) of an ovary structured into separate germary and vitellary areas. This assemblage of Rhabditophora with ectolecithal eggs is generally considered to constitute a clade called Neoophora. Lecithoepitheliata have been reconstructed as sister group of other Neoophora based on morphological characters and limited marker molecular data. Lecithoepitheliata are split into freshwater-dwelling Prorhynchida and marine Gnosonesimida and may in fact be para- or polyphyletic. In the only molecular study involving members of both taxa, they are presented as being grouped with other ectolecithal Platyhelminthes (i.e., members of the Neoophora), but Prorhynchida were found to be sister group of all other Neoophora, and Gnosonesimida as sister group of all other Neoophora except Prorhynchida. This topology led these authors to support the monophyly of Neoophora and the single origin of ectolecithality.

6Our study includes two members of Prorhynchida and, in striking contrast to most previous studies, places them not in Neoophora but as sister group of Polycladida, in accordance with. The monophyly of Polycladida and Lecithoepitheliata/Prorhynchida is given maximum support in all analyses (Figures 1, 2, 3, S1, and S2).

7This result has important implications for our understanding of the evolution of ectolecithality within flatworms: eggs with extraembryonic yolk cells would have evolved at least twice independently, once in Lecithoepitheliata (at least Prorhynchida) and once in the common ancestor of the remaining Neoophora. Neoophora excluding Lecithoepitheliata were named Euneoophora. In further support of the monophyly of Euneoophora, we found that the parahox gene Caudal/Cdx was detectable in the transcriptomes of the lophotrochozoan phyla we sampled and also in the different orders of archoophorans, i.e., catenulids, macrostomorphans, polyclads, and lecithoepitheliates, but was undetectable in the transcriptomes of all included euneoophorans (Figure 4).

8The least confidently resolved part of the flatworm portion of our tree involves the relative positions of proseriates, rhabdocoels, and the remaining euneoophorans. In our CAT+GTR+G4 phylogeny of our complete dataset, Rhabdocoela are sister group of all other Euneoophora (pp = 1.0) (Figure 1), but with low jackknife support of 0.6 (Figure 2). Other analyses (ML) instead support Proseriata in this position (Figures S1 and S2), and this is in common with most previous analyses involving one or a few genes.

9On balance, we suggest that the basal Rhabdocoela solution is the most likely for two reasons. The first reason is the support it receives from the typically better-performing CAT+GTR+G4 model analysis over the PhyML analysis. The site-heterogeneous CAT model has been repeatedly shown to fit real data better than simpler models such as the site-homogenous model used in the ML analyses, and to be better able to overcome systematic error. The second reason is that we observed stronger support for basal Rhabdocoela when analyzing the slowly evolving genes (Q1 and Q2); the more rapidly evolving genes support an association of Rhabdocoela and Tricladida/Bothrioplanida/Neodermata (Q3) or Rhabdocoela and Tricladida/Neodermata (Q4) (Figure 3). The support that this particular grouping receives in the analyses of more rapidly evolving genes seems likely to be due to an LBA artifact that leads to an incorrect association between the rhabdocoels and neodermatans, both of which have long branches. LBA is exacerbated by rapidly evolving genes. In the fastest quartile of data (Q4; a priori most susceptible to LBA) the long-branched rhabdocoels move even closer to the long-branched Tricladida/Neodermata than the short-branched Bothrioplana (Figure 3).

10More reliably resolved is the position of Tricladida, which has strong support for a position closer to Neodermata and Bothrioplanida than either Rhabdocoela or Proseriata. None of the triclads, bothrioplanids, or neodermatans show any sign of spiral cleavage in their early embryogenesis, and the loss of this trait is a persuasive morphological character uniting this group (Figure 4). Recent studies have noted that genes including SPD-2/Cep192, Nek2, and CCCAP, which have an evolutionarily conserved role in centrosome formation across Metazoa, were missing from the planarian Schmidtea mediterranea as well as from the neodermatan Schistosoma mansoni yet were present in the macrostomorphan Macrostomum lignano. This gene loss correlates with the loss of the centrosome in Schmidtea and possibly also in Schistosoma, and it was suggested that this loss of centrosomal genes is also implicated in the loss of the highly regulated spiral cleavage. Thanks to our taxonomically broad sample of transcriptomes, we have been able to extend this analysis and show that three genes associated with centrioles, SPD-2/Cep192, Nek2, and CCCAP, are at least partly present in most of the more basally branching platyhelminth taxa for which we have transcriptomes but are undetectable in any of the Tricladida, Bothrioplanida, or Neodermata (Figure 4). The evidence for absence of a gene based on inevitably partial transcriptomes must not be overinterpreted, however, and we note that none of these three genes are found in the transcriptomes that we have produced for two lecithoepitheliates (Figure 4), which show a rather conserved spiral cleavage pattern.

11The monophyly of Neodermata with well-characterized apomorphies such as a secondary unciliated syncytial epidermis is undisputed. It has long been clear that “Turbellaria” is a paraphyletic group and that the wholly parasitic Neodermata emerged from among free-living forms. That said, the identity of the closest free-living relative of Neodermata has proven elusive. In early morphological phylogenies, Rhabdocoela (or members of Rhabdocoela) were considered to be sister group of Neodermata. This relationship was not supported in subsequent molecular phylogenies using one or a few genes, in which a bewildering selection of higher flatworm taxa, e.g., Fecampiida, Prolecithophora, and Tricladida or Rhabdocoela, Fecampiida, Prolecithophora, and Tricladida, were proposed as sister group of Neodermata. In a recent study using four genes and many taxa, and now in our own study using 1,347 genes, Bothrioplanida, previously considered close to or part of the Proseriata, are shown to be sister group of Neodermata (Figure 1).

12Neodermata comprise Monogenea, Cestoda, and Trematoda. The interrelationships of these taxa has been debated, with Cestoda being considered sister group of either Monogenea ( = Cercomeromorpha) or Trematoda. The Cercomeromorpha hypothesis was rejected by phylogenetic analyses using 18S and 28S sequences, and the alternative sister group relationship between Trematoda and Cestoda was supported by studies employing whole mitochondrial gene phylogenies, by a microRNA study, by a multigene phylogeny using 312 gene models, and now by our own study. Surprisingly, a recent phylogenetic study using four genes and a large number of flatworm species supports the Cercomeromorpha hypothesis.

13With the sister group relationship between Polycladida and Lecithoepitheliata/Prorhynchida demonstrated by our phylogenetic analysis (Figure 1), the taxon Neoophora, defined as encompassing all flatworms with ectolecithal eggs, has become polyphyletic and should therefore be noted with quotation marks, “Neoophora.” “Neoophora,” excluding Prorhynchida, are monophyletic in our analyses, and this clade has previously been named Euneoophora, characterized by the presence of ectolecithal eggs and by germaria and vitellaria as spatially separated organs.

14We propose the name Amplimatricata new taxon for Polycladida+Lecithoepitheliata, based on the tendency in both groups for possession of an ample extracellular matrix. Taking into account the remaining uncertainty over the monophyly of Lecithoepitheliata, Amplimatricata encompasses at least Polycladida+Prorhynchida. Acentrosomata new taxon is a clade consisting of Tricladida and its closely related taxa Prolecithophora and Fecampiida (all three taxa making up Adiaphanida), Bothrioplanida, and Neodermata (Figure 4). The name is based on the implied absence of centrosomes in all of these taxa (Figure 4). Lacking strong similarities to serve as a clade-defining synapomorphy between Bothrioplanida and Neodermata, we use the name Bothrioneodermata new taxon to identify this monophyletic group (Figure 4).

15We have presented new transcriptomic data from 22 new species and produced a large and taxonomically complete dataset for assessing the relationships of Platyhelminthes. The majority of our conclusions are robust and are supported by different methods of analysis, high Bayesian posterior probabilities, and high jackknife support. The two instances of lower support concern the position of Platyhelminthes relative to other lophotrochozoan phyla and the early-branching position of rhabdocoels relative to other Euneoophora. The evidence against Platyzoa and support for early-branching Rhabdocoela by site-heterogeneous analyses (Figure 1) and by the slowest-evolving quartiles of the total dataset (Figure 3, Q1 and Q2) suggest that the alternatives, which are supported by the less-well-fitting site-homogenous analyses (Figures S1 and S2) and the faster-evolving quartiles of the data (Figure 3, Q3 and Q4), are the result of LBA. The suggested monophyly of Proseriata+Acentrosomata (Figure 4) might be tested further by the addition of the two additional members of Adiaphanida, Fecampiida and Prolecithophora, as well as the second clade of Lecithoepitheliata, Gnosonesimida, as there has been evidence that Lecithoepitheliata may be paraphyletic.

16B.E., C.B.C., C.N., E.B., F.L., K.A.R., J.G., M.A.T., and M.G. collected samples for RNA extraction. B.E., F.L., J.G., and K.A.R. extracted RNA. B.E. and M.J.T. assembled transcriptomes. B.T., C.D., N.S., and S.M. produced the OMA analyses. M.J.T. assembled the sets of orthologs and conducted phylogenetic analyses. B.E. and M.J.T. wrote the manuscript and designed the figures. All authors read and approved the final manuscript.