SSU rRNA length differences and base composition
To avoid biases due to missing data, we analyzed base composition and sequence length for a fragment of SSU rRNA that covers about 95 % of the gene (see Methods). Base composition and fragment length differed considerably among the 123 poriferan sequences (Fig.
1). The GC content varied between 45.5 and 56.3 %.
| Figure 1GC content against SSU rRNA fragment length. (Fragment corresponds to A. queenslandica positions 48–1896). A ca. 95% -fragment of SSU rRNA was used for analysis and only sequences with sequence information over the whole range of this fragment (more ...) |
Calcarea posses the lowest GC contents with a modest variation from 45.5 to 46.8 %. In this aspect they are clearly separated from demosponges, which display significantly higher GC contents, since the lowest demosponge value (47.1%) still exceeds the highest GC content (46.8%) of Calcarea (Fig. 1). Most demosponge SSU rRNAs show modest length variations in a range comparable to those of Calcarea. Notable exceptions are the extraordinary large rRNA molecules found in several marine haplosclerids. The highest GC contents of Porifera are also found within this group (with a maximum of 56.3% in Amphimedon queenslandica). The high GC pattern is independent of the presence of insertions in these large molecules, since members of marine Haplosclerida with smaller rRNA molecules also possess similar GC contents (Fig. 1). SSU rRNAs of the few available hexactinellid sponges are approximately equal in length to large molecules of several haplosclerids (with the exception of Farrea occa [GenBank: AF159623], see below). In contrast to haplosclerids, hexactinellid sponges have lower GC contents, with base compositions in the range of those of Demospongiae and Calcarea.
Secondary structure
Porifera have the typical eukaryotic core SSU rRNA structure (see Figs.
2,
3,
4). The moderate length variation between Calcarea and most demosponges is primarily caused by insertions in unpaired regions or by elongation of helices 10, E10_1 and 43 (Table
1). In Hexactinellida, on average, these three helices are largely elongated compared to Calcarea and Demospongiae (Fig.
3), but the lengths of the E10_1 helices of some demosponge sequences fall into the same range.
| Figure 2SSU rRNA secondary structure for Calcarea. Sequence is given as 90% consensus with variable positions in black boxes. Lower case indicates deletions at the site for some sequences, according to the consensus level. Differences in helices between Calcaronea (more ...) |
| Figure 3SSU rRNA secondary structure of Acanthascus dawsoni [GenBank: AF100949] (Lyssacinosida, Rossellidae). Hexactinellid-specific helical insertions within E23_1 are shown in a box. Inset: Prediction of secondary structure insertions in E23_1 within other (more ...) |
| Figure 4SSU rRNA secondary structure of the demosponge Amphimedon queenslandica (Haplosclerida). Nucleotides conserved in Demospongiae at the 90% level are shown in black, other nucleotides are in grey. Nucleotides at positions that are present in demosponges (more ...) |
| Table 1 Mean and range of the length of the most variable helices within the three sponge classes |
In addition, we observed extra-helical insertions in Hexactinellida and in several marine haplosclerid demosponges that are not part of the eukaryote core structure. In marine haplosclerids, these extra sequences were inserted within helices E23_1, E23_14 and 43, and in Hexactinellida, the insertions only occurred within helix E23_1 at a different position than in marine haplosclerids (Figs. 3, 4). All of the helices where sequence elongations and/or insertions occur belong to regions that are known to be highly variable within eukaryotes [39,50].
Calcarea
The SSU rRNA of this sponge class comprises all of the typical eukaryote helices and lacks unusual structural features. A calcarean SSU rRNA consensus sequence and structure is shown in Fig.
2. Several synapomorphies for the two Calcinea and Calcaronea subclasses were detected in the secondary structure. In Calcinea, helices 10 and E23_1 are shorter by at least one base pair when compared to Calcaronea (Fig.
2, insets I & III). In helix E10_1, Calcaronea typically have three pairs at the helix end, whereas Calcaronea dominantly possess four pairs (Fig.
2, inset II). However, independent elongations of this helix can be found in both subclasses (Fig.
2, inset IIa: Calcaronea:
Plectroninia neocaledoniense; Calcinea:
Soleneiscus radovani). These elongations are homoplasies as is evident when considering the subclass-specific compensatory base change (Calcaronea: A-U; Calcinea: G-C) at the beginning of inset II (Fig.
2): The A-U pair in the corresponding structure of the calcaronean
Leucascandra caveolata (Borojevic & Klautau, 2000) supports a secondary loss of a pair compared to other Calcaronea. Differences in helix nucleotides between both subclasses occur in helices 11, E23_7, E23_14 and 29 (Fig.
2, and insets IV and V). Most of these changes maintain the helix-relevant pairings (e.g., in 11 or E23_7), but a few cause mismatches in at least some sequences (in E23_7, E23_14, 29 and 49). Base changes and insertions in unpaired regions are also specific for the Calcinea-Calcaronea split. This is indicated in Fig.
2 for three bases in the bulge between helices 8 and 9, one base within the loop of E23_12, and a calcaronean-specific insertion of one adenosine between helix 9 and 10.
Hexactinellida
The SSU secondary structure of
Acanthascus dawsoni is presented in Fig.
3 representatively for Hexactinellida. In all hexactinellid sequences, specific insertions were observed (Fig.
3 inset). As mentioned previously, hexactinellid SSU rRNA sequences are considerably longer than in other poriferans, except in some marine haplosclerid demosponges (Fig.
1). The additional nucleotides occur in extensive elongations of common helices (10, E10_1, and 43; Table
1, Fig.
3), and a helical insertion in helix E23_1. The insertions in helix E23_1 occur at a unique position among sponges and may form two helices (assigned the names E23_1b.1 and E23_1b.2 in Fig.
3) in all the studied SSU rRNA molecules of subclass Hexasterophora. Helix E23_1b.2 contains 10 doublets and is much more conserved within Hexasterophora than Helix E23_1b.1, which varies in length from 37 to 55 bp. In contrast to Hexasterophora,
Semperella schulzei (Subclass Amphidiscophora) has a helical insertion of 107 bp within E23_1, which is predicted to form a single helix E23_1b.1 (Fig.
3 inset).
Within the sequence of Farrea occa [GenBank: AF159623], we found deletions in conserved regions. Helices 13 and 15 are missing completely, as are the 3' strand of helix 7, parts of helix 43 and the 5' strand of helix 15 (compare Fig. 3). Such complete or partial deletion of conserved helices has been shown to be typical for non-functional rRNA pseudogenes [51]. Potential paralogs like this one are not necessarily subject to concerted evolution, and are therefore not suitable for phylogenetic inference. In this context, the consideration of secondary structure is crucial for identification of such non-functional sequences, and prevents biases in phylogenetic reconstruction due to potentially misleading data. Nonetheless, predictions of insertions for this sequence are displayed in Fig. 3 (inset), since no suspicious modifications were found within this part of the molecule and no other sequence of Farreidae was available. However, the results for this species should be treated with caution.
Demospongiae
Most demosponges possess a SSU rRNA molecule with the common metazoan secondary structure. Remarkable exceptions are only found within the marine Haplosclerida (Figs.
4,
5,
6), which possess insertions that are long enough to be predicted to form additional helices. Those helices are found within known variable regions for eukaryotes and appear in the 5' strand of Helix E23_1/2, the 5'strand within Helix E23_14 and the 3' strand of helix 43.
| Figure 5Phylogeny inferred with PHASE. Nodes that differ from the topology published by Redmond et al. [41] are encircled. The boxed clades correspond to the excerpt displayed in Fig. 6. Support values are given at, or close to the corresponding node (values (more ...) |
| Figure 6Relationships of marine Haplosclerida (excerpt from larger phylogenetic analyses shown in Fig. 5) and evolution of extension regions. Sketches of predicted secondary structures for extensions and conserved flanking regions correspond to outlined boxes (more ...) |
Phylogenetic analyses
We inferred the phylogeny of marine haplosclerids to compare the evolutionary history of helical insertions found in this group of Demospongiae (see section "Successive evolution of additional helices in marine haplosclerids"). Results from the PHASE- and MrBayes analyses of 78 taxa are shown in Fig.
5. Although more general 7-state models have been shown to result in higher likelihood values for phylogenies than less parameter-rich models for real rRNA data [
47], our analyses with PHASE with the RNA7A model and the less complex RNA7D model yielded identical tree topologies (with almost identical support values). Independent runs in PHASE and MrBayes resulted in similar, almost identical topologies, and differences in demosponge relationships were only observed in the positions of clades with weak support values. Namely these are the relationship of Dictyoceratida to the Myxospongiae (sensu Borchiellini et al. 2004 [= clade Verongida +
Chondrosia reniformis]), the position of
Scopalina ruetzleri and relationships within freshwater sponges (where branch lengths were short, Fig.
5). Additionally, differences were observed in Calcarea and Cnidaria.
The order Haplosclerida was not resolved as monophyletic, since Suborder Spongillina (freshwater sponges) fell into other distantly related demosponge clades, rather than into marine Haplosclerida. The two suborders, Haplosclerina (families Callyspongiidae, Chalinidae and Niphatidae) and Petrosina (represented here by the families Petrosiidae and Phloeodictyidae) were not supported as monophyla (Fig. 6). These results are congruent with results from former analyses of SSU rRNA, 28S rRNA and cytochrome oxidase subunit I [20,22,52].
According to our analysis, Amphimedon queenslandica (Family Niphatidae) is most closely related to Oceanapia sp. (Family Phloeodictyidae), with Xestospongia muta (Family Petrosiidae) as sister taxon, and both nodes in the tree are very highly supported by posterior probability (PP) values in both Bayesian analyses. Other species of the family Niphatidae (Niphates sp. and Dasychalina fragilis) are not closely related to each other or to Amphimedon queenslandica (Fig. 6). In addition, other members of families Petrosiidae and Phloeodictyidae are not found in a closer relationship to the three species clade. Our results were mostly concordant with Redmond et al. [41], but with higher support values in several clades. We did not find any monophyletic haplosclerid families or genera in our taxon set. Differences between our results and the previous study are highlighted at the nodes in Fig. 5. We could not recover monophyletic Petrosiidae in clade I, and relationships of several clade III taxa differed. Furthermore, Xestospongia muta and Oceanapia sp. cluster in one clade (IV) (including Amphimedon queenslandica).
Successive evolution of additional helices in marine haplosclerids
Within haplosclerids, the evolution of additional helices can be reconstructed by plotting structures to the well-supported phylogenetic backbone (Fig.
6). Primary sequences of these motifs were not included in the tree construction (Fig.
5) due to ambiguous alignment, but can be regarded as additional phylogenetic characters. The helical insertions apparently evolved in at least two steps, which fits the findings of the SSU rRNA gene tree strikingly well. The relationships within marine haplosclerids can be described as four well-supported (PP > 0.97) nested clades I-IV that display different stages of secondary structure evolution (see above and Figs.
5,
6). Clade I contains all marine Haplosclerida. The basal diverging taxa lack any large insertions that are typical for other marine Haplosclerida. However, the predicted structure within helix 43 differs from the standard structure in this region found in other Porifera (compare outgroup in Fig.
6) and displays a larger bulge of unpaired bases at the insertion point of the larger helical structures found within all taxa in Clade II. This bulge may be the precursor for the extensions at this position observed in Clade II. Within basal diverging taxa of Clade II (i.e. Clade II without Clade III), a similar bulge is found for
Xestospongia muta in helix E23_14 at the insertion-site of subsequent extensions in Clade III, but not the other sequences lacking E23_14.1.
Larger insertions appeared in helices E23_1 and 43 'simultaneously' (according to phylogenetic resolution recovered by our analyses) in the common ancestor of Clade II taxa. The three taxa of Clade IV according to our minimum free energy calculations share an additional helix 43b.2 as synapomorphy (Figs. 4, 5, 6).
After the introduction of helical insertions in helices E23_1 and 43, a long extension evolved within helix E23_14 as a synapomorphy in Clade III. An autapomorphy for Dasychalina fragilis is an additional helix formed b insertions within E23_1. Within the complete taxon of marine haplosclerids, no loss of formerly gained additional helical insertions has been documented, therefore, no SSU rRNA molecule from a descendent of a taxon with extraordinary features has returned to the ancestral basic metazoan core structure.