Near intron pairs and the metazoan tree

Abstract

Gene structure data can substantially advance our understanding of metazoan evolution and deliver an independent approach to resolve conflicts among existing hypotheses. Here, we used changes of spliceosomal intron positions as novel phylogenetic marker to reconstruct the animal tree. This kind of data is inferred from orthologous genes containing mutually exclusive introns at pairs of sequence positions in close proximity, so-called near intron pairs (NIPs). NIP data were collected for 48 species and utilized as binary genome-level characters in maximum parsimony (MP) analyses to reconstruct deep metazoan phylogeny. All groupings that were obtained with more than 80% bootstrap support are consistent with currently supported phylogenetic hypotheses. This includes monophyletic Chordata, Vertebrata, Nematoda, Platyhelminthes and Trochozoa. Several other clades such as Deuterostomia, Protostomia, Arthropoda, Ecdysozoa, Spiralia, and Eumetazoa, however, failed to be recovered due to a few problematic taxa such as the mite Ixodes and the warty comb jelly Mnemiopsis. The corresponding unexpected branchings can be explained by the paucity of synapomorphic changes of intron positions shared between some genomes, by the sensitivity of MP analyses to long-branch attraction (LBA), and by the very unequal evolutionary rates of intron loss and intron gain during evolution of the different subclades of metazoans. In addition, we obtained an assemblage of Cnidaria, Porifera, and Placozoa as sister group of Bilateria + Ctenophora with medium support, a disputable, but remarkable result. We conclude that NIPs can be used as phylogenetic characters also within a broader phylogenetic context, given that they have emerged regularly during evolution irrespective of the large variation of intron density across metazoan genomes.

Introduction

The evolutionary relationships of metazoan phyla still constitute a challenge for both morphological and molecular-based analyses. The traditional view arranges bilaterian metazoans into acoelomates, pseudocoelomates, and coelomates. Starting with the work of Aguinaldo et al. (1997), sequence data, initially mostly rDNA, have been used to establish a “new animal phylogeny” with far-reaching consequences: (1) the protostomes were divided into ecdysozoans (Aguinaldo et al., 1997) and lophotrochozoans (Halanych et al., 1995), and (2) several phyla representing apparently lower grades of complexity (Platyhelminthes, Nemertea, and Nematoda) were relocated amongst the coelomate groups at the crown of the tree. In contrast, some studies employing genomic datasets containing only a few taxa (e.g. Wolf et al., 2004) supported the monophyly of coelomates. Later studies, however, have shown that these results are likely artefacts, misled by a faster evolution of some genomes, such as that of Caenorhabditis elegans (Philippe et al., 2005). These are thus often excluded from phylogenetic datasets. In addition, conflicting signals are often obtained from mitochondrial, nuclear rRNA, phylogenomic and also morphological data (Trautwein et al., 2012). Despite a plethora of studies based on both molecular and morphological data, a consensus on the phylogenetic tree of metazoan phyla is still not in sight (Edgecombe et al., 2011). This concerns in particular the non-bilaterians (Dunn et al., 2008, Schierwater et al., 2009, Pick et al., 2010) and the Lophotrochozoa (Hejnol, 2010).

Characters resulting from structural changes of the genomic sequence, so-called rare genomic changes (RGCs), such as coding insertions/deletions (indels) (Belinky et al., 2010), spliceosomal intron positions (Irimia and Roy, 2008), and positions of mobile genetic elements (Kriegs et al., 2006, Kriegs et al., 2010), are expected to be less prone to homoplasy than substitution patterns of sequence data and hence provide valuable additional information to resolve conflicts in phylogenetic tree reconstruction. For holometabolic insects, novel phylogenetic hypotheses have been introduced on the basis of such characters, for example the basal position of Hymenoptera (Krauss et al., 2005). Later, this proposal received strong support by sequence-based analyses of single-copy nuclear genes and additional intron position data (Savard et al., 2006, Zdobnov et al., 2007, Krauss et al., 2008, Wiegmann et al., 2009). Another study used retrotransposon insertions to improve our knowledge about the basal branching order of rodents (Churakov et al., 2010). Earlier attempts to reconstruct the radiation of rodents are well-known to have suffered from long-branch attraction (LBA) artefacts.

The present study utilizes the conservation of positions of spliceosomal introns among orthologous coding sequences (CDS). Intron positions have already been used by several authors to resolve problematic branches of the metazoan tree (for review see Irimia and Roy, 2008). For instance, an intense debate emerged about the concepts of Ecdysozoa (Roy and Gilbert, 2005) and Coelomata (Zheng et al., 2007) using intron position data. Roy and Gilbert (2005) supported the taxon Ecdysozoa using a pattern of intron conservation. This was criticized by Zheng et al. (2007) by showing that intron loss rates within specific branches are strongly correlated. These authors argued that high rates of independent intron losses within the used nematode and arthropod species had misled the former study. However, in turn, Roy and Irimia (2008) identified several weaknesses of the latter analysis, among them biases in the procedure used to differentiate between intron gain and loss. Pointing to both large intron loss and gain rate variations, Roy and Irimia (2008) avoided a clearcut conclusion about the Ecdysozoa/Coelomata problem.

In order to reduce the impact of homoplastic characters due to parallel intron gains or losses, we specifically consider pairs of nearby introns. More precisely, a near intron pair (NIP) consists of two intron positions in an alignment of two or more orthologous genes that are separated by a small number of nucleotides. Exons smaller than about 50 nt are relatively rare (Saeys et al., 2007) and in general functionally detrimental (Weir et al., 2006). The two nearby intron positions are thus very unlikely to have coexisted. Under the assumption that parallel intron gain is very rare, a NIP can be used to parsimoniously infer an edge of the phylogenetic tree along which both intron loss and gain must have occurred, separating the species sharing one of the positions from those that share the other.

In previous work, we found some evidence that NIPs arise not only from uncoupled, successive processes of intron loss and intron gain, but also from intron sliding (Krauss et al., 2005, Krauss et al., 2008, Lehmann et al., 2010). For Drosophila we could show that some of the younger NIPs were indeed caused by shifts of splice donor and acceptor sites in relation to conserved CDS (Lehmann et al., 2010). In the same study, we used NIPs for a systematic investigation of intron gain mechanisms in Drosophila.

Encouraged by the successful application of NIPs to the phylogeny of holometabolan insects (Krauss et al., 2008, Niehuis et al., 2012), we here try to resolve the phylogenetic tree of animals based exclusively on NIP data from 45 metazoan and 3 outgroup taxa. Our results demonstrate the usefulness of NIPs as phylogenetic marker also for deep metazoan phylogeny. In particular, we evaluate the Ecdysozoa hypothesis, as well as the general agreement of our tree reconstructions with current proposals of metazoan phylogeny.