Introduction

A decade since the publication of the first complete mitochondrial (mt) poriferan genomes (Lavrov et al. 2005), our understanding of animal mtDNA evolution is still incomplete. These genomes revealed major structural differences to their bilaterian counterparts. Subsequently, published mt-genomes of placozoans (Dellaporta et al. 2006), ctenophores (Pett et al. 2011), and calcareous sponges (Lavrov et al. 2013) displayed even more dramatic differences to the relatively homogeneous bilaterian mt-genomes. Several differences can directly be attributed to ancestral traits, successively lost in Bilateria. Additionally, there are several mt-intragenic elements only present in sponges and other non-bilaterian mtDNA, such as repetitive hairpin elements (see e.g., Erpenbeck et al. 2009). Likewise mitochondrial introns, frequently self-splicing and mobile elements, are known from choanoflagellates, ichthyosporeans, and other close relatives of Metazoa (Burger et al. 2003; Burt and Koufopanou 2004), furthermore in Porifera, Cnidaria, and Placozoa (e.g., Beagley et al. 1996; Dellaporta et al. 2006; Rot et al. 2006). Bilateria, however, (mostly) lack mitochondrial introns (see Vallès et al. 2008).

Mitochondrial introns in sponges are rare. Only three different types are known, all in the cytochrome oxidase subunit 1 (CO1) gene of tetillid Demospongiae (Cinachyrella and Tetilla), but also in plakinid Homoscleromorpha (Plakina and Plakinastrella) (Gazave et al. 2010; Rot et al. 2006; Szitenberg et al. 2010) (Fig. 1). All these introns harbor an ORF for a putative LAGLIDADG-type homing endonuclease. The high similarity of the introns (>85–99 %, Szitenberg et al. 2010) and their identical insertion site in CO1 are in strong contrast to the distant relationship of their hosts as Plakinidae and Tetillidae are members of two distant poriferan classes (see e.g., Wörheide et al. 2012). Insertions of LAGLIDADG-ORF introns into genes are sequence-tolerant, but site-specific (Edgell 2009), which complicate our understanding whether their patchy distribution derives from vertical transmission among ancestral sponge lineages with subsequent loss (Wang and Lavrov 2008), or from horizontal transmission (Haugen et al. 2005; Rot et al. 2006), or from a combination of both (see also Fukami et al. 2007; Szitenberg et al. 2010). We now detected a CO1 intron in a second demosponge lineage, Verongida, which corroborates the theory of combined vertical and horizontal transmission in sponges, but impede the use of popular CO1 primers.

Fig. 1
figure 1

Schematic view of the CO1 intron position in sponges. Three types of introns are known. They differ by their sequence and insertion points and are subsequently called “714”, “723”, and “870” following their distinct position relative to Amphimedon queenslandica COI (cf. Szitenberg et al. 2010). P. onkodes is the only sponge taxon with two introns, its “714” intron is short and lacks an ORF

Full size image

Materials and Methods

DNA of Aplysinella rhax (De Laubenfels 1954) (Queensland Museum voucher QMG313039) was extracted, and CO1 sequenced following previously published methods (Erpenbeck et al. 2012) using the reverse primer COX1-R1 (Rot et al. 2006) and 1:30 min PCR extension time (Genbank Acc. No: KP026315). Sequence annotation and alignments were performed with MacClade 4.08 (Maddison et al. 1999) and Geneious v6.1.6 (Biomatters). LAGLIDADG-type homing endonuclease protein sequences from published alignments (Rot et al. 2006; Szitenberg et al. 2010) and Aplysinella rhax were realigned with MAFFT v7.149b (G-INS-i Katoh and Standley 2013), and end-trimmed prior to phylogenetic reconstruction with RAxML v8.0.26 (Stamatakis 2006) (see Fig. 2 for details).

Fig. 2
figure 2

Maximum-likelihood tree of LAGLIDADG-type homing endonuclease protein sequences from sponges, cnidarians, and non-metazoans from the published alignments (Rot et al. 2006; Szitenberg et al. 2010), and the partial sequences yielded from Aplysinella rhax. The data set is based on 80 protein characters from the 3′ region of A. rhax LAGLIDADG. The tree is reconstructed with RAxML with 100 rapid bootstraps and subsequent ML search under the PROTGAMMA model with MTREV. Numbers on the branches indicate bootstrap support (>65). Numbers preceding taxon names are Genbank accession numbers. Scale bar depicts substitutions per site

Full size image

Results

CO1 of Aplysinella rhax possesses an intron at a similar position to Cinachyrella alloclada (Uliczka 1929), C. levantinensis Vacelet et al. 2007 and Plakinastrella onkodes (Uliczka 1929) (Fig. 1), and resembles their “723”-type LAGLIDADG homing endonuclease (cf. Szitenberg et al. 2010) with ≥80 % nucleotide similarity. Interestingly, p-distances of Cinachyrella sequences to the homoscleromorph P. onkodes are smaller (0.118) than to its fellow Demospongiae A. rhax (0.188). Similarly, the 3′intron region between the LAGLIDADG-ORF stop codon and the CO1 exon 3 (113 bp) of Cinachyrella spp. is 70 % identical to P. onkodes, but barely alignable to A. rhax.

The Maximum-likelihood analysis recovered all metazoan “723”-type CO1 introns as monophyletic (Fig. 2). The sponge sequences, however, are paraphyletic with A. rhax as sister group to all other Metazoa. The cnidarian sequences are nested within the remaining sponges as sister group to the Tetillidae.

Discussion

The specificity of a mobile intron for insertion into homologous sites leads to rapid insertions at identical positions (see Haugen et al. 2005). Unlike e.g., SINE insertions, which may constitute potent phylogenetic markers (see also Miyamoto 1999), evolutionary tracking of group I introns remains more challenging. The initial interpretation of mt-introns in sponges as the result of a horizontal transfer from fungi (Rot et al. 2006) was subsequently doubted after homoscleromorph sponge (Wang and Lavrov 2008), and coral (Fukami et al. 2007) introns from the same sites were discovered. Instead, the patchy distribution of introns among poriferans in two families of different classes was interpreted with multiple losses of an ancestral intron (Wang and Lavrov 2008), as known among Bilateria (Cho et al. 2004). However, the A. rhax intron phenetic discrepancies in coding and noncoding regions and the resulting phylogenetic position of its protein are strongly incongruent with the currently accepted sponge systematics, and therefore oppose this scenario (Redmond et al. 2013; Wörheide et al. 2012). Instead, our discovery of a second intron-bearing lineage in demosponges supports a combination of vertical and horizontal transmissions (cf. Szitenberg et al. 2010). So far, complete co-speciation between the introns and their hosts, a prerequisite for the multiple-loss hypothesis could never be shown, therefore some lineages must have acquired introns independently (Fukami et al. 2007; Szitenberg et al. 2010). Group I introns with homing endonuclease likewise can invade each other or invade a “naked” intron without endonuclease ORF (Haugen et al. 2005). The horizontal transmission (see e.g., Vallès et al. 2008) with fixation will remain a rare event in metazoans in comparison to other opisthokonts as metazoan germ lines are relatively segregated, which inhibits the easy incorporation of DNA fragments (Burt and Koufopanou 2004).

The poriferan mitochondrial introns are inserted in the 5′region of CO1, a popular marker for phylogenetic analyses (e.g., Erpenbeck et al. 2007), and therefore hamper experiments by disrupting annealing sites of popular universal CO1 (barcoding) primers (e.g., Rot et al. 2006). Our findings that demosponge mt-introns are not restricted to Tetillidae (one of 97 demosponge families, Van Soest et al. 2012), but may be distributed among other demosponge lineages, must subsequently be considered for standardized mt-studies (e.g., DNA barcoding). The ~100 demosponge mt-genomes sequenced so far still fail to provide sufficient overview on intron distribution among the >8,500 sponge species for the identification of intron “hotspots” (Szitenberg et al. 2010) such as Tetillidae and Plakinidae.