Contribution to Zoology, 76 (2) – 2007
Affinities of the family Sollasellidae (Porifera, Demospongiae).II. Molecular evidence
Dirk Erpenbeck1,2,3, John N.A. Hooper1, Sue E. List-Armitage1, Bernard M. Degnan3, Gert Wörheide2, Rob W.M. van Soest4
Keywords: sponges, classification, Raspailiidae, Sollasella, Raspailopsis, 28S rDNA, molecular systematics
This is the second part of a revision and re-classification of the demosponge family Sollasellidae, and an example of a successful use of combined morphological and molecular data. Sollasella had been a poorly known, long forgotten taxon, placed incertae sedis in the order Hadromerida in the last major revision of the demosponges. It has recently been suggested to belong to Raspailiidae in the order Poecilosclerida due to striking morphological similarities. The present analysis verified this re-classification using molecular markers. Comparing 28S rDNA fragments of Sollasella cervicornis, a newly described species S. moretonensis and a representative set of raspailiid and hadromerid samples. In our analyses Sollasella clearly clusters inside the Raspailiidae clade, and distantly from hadromerid taxa. Supporting morphological hypothesis of Van Soest et al. (2006), that Sollasella is a raspailiid sponge.
A morphological re-examination of material of the demosponge family Sollasellidae has shed a new light on its classification (Van Soest et al., 2006). This monospecific, previously poorly known family, with the type locality in East Australia (Sollasella digitata Lendenfeld, 1888), had been nearly completely forgotten since Hallmann (1914), and consequently assigned more recently as incertae sedis in the order Hadromerida based on its cortex and radiating skeleton (Hooper and Van Soest, 2002). However, the recent collection and examination of a morphologically similar sponge from Oman identified as Raspailopsis cervicornis Burton, 1959 (=Ras-pailia (Parasyringella) cervicornis sensu Hooper, 2002b), raised evidence for a placement of Sollasella in the family Raspailiidae (order Poecilosclerida, Van Soest et al., 2006). After further morphological analyses, R. cervicornis was merged with Sollasella formingthe twentieth valid genus in the Raspailiidae (Van Soest et al., 2006). In the same publication Van Soest and colleagues describe a similar species from subtropical Australia, Sollasella moretonensis, which shares the characteristic polygonal perforation-like surface pattern of S. cervicornis and S. digitata, in addition to other features. Thesepolygonal surface structures remain a distinguishing feature for the genus.
The aim of our analysis is to test the hypothesis that Sollasella is more closely related to poecilosclerid taxa as concluded recently based on morphologic data in the family Raspaillidae than to hadromerid sponges (Van Soest et al., 2006). Among candidate hadromerid taxa are Polymastiidae, with which Sollasella shares the presence of a cortex (particularly the genus Pseudotrachya, with which Sollasella shares the combination of choanosomal styles and ectosomal oxeas), and Suberitidae, with which Sollasella shares the stalked habit and axially condensed skeleton (genera Homaxinella , Plicatellopsis , Rhizaxinella, see Van Soest, 2002a). There is also an apparent slight affinity with Stylocor-dylidae (Van Soest, 2002b).
Reliability of morphological systematics in sponges is hampered by the paucity of complex characters, many of which are also prone to homoplasies. The family Raspailiidae, as in many other demosponge taxa, does not possess unique autapomorphies, and is defined by a combination of characters that may each be found also in some other demosponge orders and families (see Hooper, 2002b). DNA sequence data should provide independent evidence to test morphological hypotheses. However, sponge molecular systematics has pitfalls of its own such that phylogenies arising from these analyses need to be cautiously interpreted pending better insight into demosponge molecular evolution. Molecular data occasionally generate hypotheses that conflict dramatically with phylogenies based on morphological characters (e.g., McCormack et al., 2002; Erpenbeck et al., 2006). Improved algorithms may aid to filter the correct phylogenetic signal from random noise and provide explanations for the occasionally ‘odd’ phylogenies. Nevertheless, sponge molecular systematics has repeatedly been shown to outcompete morphological approaches (e.g. Erpenbeck et al., 2006).
In our present approach we generate and analyse 28S rDNA data from specimens of Sollasella cervicornis (sensu Van Soest et al., 2006) from Oman, Sollasella moretonensis from northeastern Australia, and several species of other Raspailiidae. The sequences are compared with data sets from Erpenbeck et al. (2005) and partially from Nichols (2005), which is currently the most comprehensive 28S rDNA data set with a strong representation of hadromerid sequences.
Material and methods
The list of specimens used in this analysis is given in Table 1. We chose the 28S D3-D5 fragment for which a comprehensive taxon set has been sequenced. However, no amplifiable DNA could be obtained from material of the type specimen of the family Sollasellidae, Sollasella digitata , from the Australian Museum in Sydney (AM G9107). Sequences of Sollasella cervicornis, S. moretonensis and a representative set of Raspailiidae have been recently generated and are deposited in GenBank (see table 1). Their DNA was extracted with Qiagen DNeasy kits. PCR primers employed were taken from McCormack et al. (2002) (RD3A: GACCCGTCTTGAAACACGA and RD5B2: ACACACTCCTTAGCGGA, temperature regime: 94°C 2 min, 35 cycles at 94°C 30 sec; 47°C 20 sec; 65°C 30 sec, followed by 65°C 10 min). PCR amplifications contained 11.25 µl ddH2 O, 4.15 µl dNTP (10 mM), 3.25 µl MgCl2 (25 mM), 2.5 µl 10× HotMaster PCR Buffer, 2.5 µl BSA (100 mM, Sigma), 0.5 µl primer (10 mM) and 4 µl HotMaster polymerase (Eppendorf). The BSA was used only on the Sollasella moretonensis DNA extracts and replaced with ddH2 O for all other samples. Cycle sequencing of both strands was performed with BigDye terminator v1.1 (ABI) and a capillary sequencer (ABI). MacClade 4.06 (Maddison and Maddison, 1992) was used for sequence management and manual alignment. The sequences were incorporated into two modified alignments based on secondary structure information. They consisted of the full length D3-D5 data set as published in Erpenbeck et al. (2005) and a second, shorter data set with additional overlapping sequences of Nichols (2005). As this overlap data set is comparatively short - it resulted in an alignment with 441 characters - we restricted ourselves to incorporate only those sequences of Nichols (2005) with a complete 5’ region (see Table 1) and used it for comparative purposes only.
Phylogenetic relationships were reconstructed with Bayesian inference methods (BI) using MrBayes 3.1 (Ronquist and Huelsenbeck, 2003). Potential overparameterization does not influence the correctness of BI reconstructions in MrBayes (Huelsenbeck et al., 2001). Therefore, the GTR+G+I model was chosen for all BI analyses of the D3-D5 data set and non-pairing sites in secondary structure specific analyses with the Nichols’ (2005) data set (SH for paring sites, see Erpenbeck et al. 2007 for details). Results were compared with Minimum-evolution (ME) reconstructions under Maximum-likelihood (ML) distances. Modeltest 3.06 (Posada and Crandall, 1998) estimated the relatively best-fitting ML-model. Bayesian inference analyses consisted of two runs of four Markov chains each for maximal 10,000,000 generations. Runs were stopped automatically when the average standard deviation of split frequencies dropped below 0.01. All other phylogenetic analyses were performed with PAUP*b10 (Swofford, 2002). The ME bootstrap values were calculated on 1000 replicates on maximum likelihood distances. Different phylogenetic hypotheses were tested using the COUNSEL 0.942 package (Shimodaira and Hasegawa, 2001) under default settings.
The different phylogenetic reconstruction methods of both data sets displayed identical results on the position of the Sollasella cervicornis sequence. In the longer data set this sequence clusters inside a clade comprising all the other Raspailiidae sequences including Raspailia (Raspailia ), R. (Raspaxilla ) and Eurypon (Fig. 1). This pattern is supported by high posterior and bootstrap probabilities (BI: 100, ME: 94). Similarly, sequences of the newly described species Sollasella moretonensis cluster in the same clade as the Raspailiidae, but the genus Sollasella is not recovered as monophyletic in those
Table 1. Sample list with the accession numbers for the particular data sets. (New sequences are highlighted in bold.)
reconstructions using molecular data, as its members form an unresolved polytomic clade with Eurypon sp. Nevertheless, the position of Sollasella in these reconstructions is clearly more distant from the Ha-dromerida representatives of this particular data set, viz. the Clionaidae Spheciospongia vagabunda and the Suberitidae Suberites spp. and Aaptos suberitoides .
A more representative Hadromerida taxon set was obtained after merging the data with the sequences of Nichols (2005), with sequences from multiple hadromerid families. Although this data set is considerably shorter, and therefore more poorly resolved due to having fewer informative characters, there is no indication of a closer relationship of Sollasella with any Hadromerida (Fig. 2). Regarding the position of Sollasella the resulting phylogenetic topology is congruent with the previous results using the smaller taxon set. The Sollasella cervicornis and S. moretonensis cluster is well-supported within the Raspailiidae, which forms a strongly supported clade. This pattern has been verified using an
Fig. 2. Bayesian inference consensus tree of the overlap with selected taxa from Nichols (2005). The numbers in regular font refer to Bayesian posterior probabilities. Numbers in italics are Minimum-evolution bootstrap support values of corresponding clades. Values lower than 75 are omitted from both methods. Numbers behind taxon names are QM collection numbers (G....) or sample numbers as given in Nichols (2005) respectively (UCMPW...). This tree is for comparison only due to the shortness of its data set. Analyses under secondary structure specific models support the branch combining Sollasella with Raspailiidae (and the Heteroxyidae) also by a posterior probability of 100.
additional analysis under application of secondary-structure specific models in the Bayesian analyses (see Erpenbeck et al. 2007). The use of secondary structure specific models in phylogenetic analyses has been shown to significantly improve gene trees (e.g., Dohrmann et al., 2006) and is suggested as a standard for phylogenetic analyses (Erpenbeck et al. 2007).
Nevertheless, the Raspailiidae do not appear to be monophyletic in the resulting topologies. The sequences of Myrmekioderma granulata and the closely related Didiscus oxeata (both Heteroxyidae (= formerly Des-moxyidae), order Halichondrida) cluster within the Raspailiidae resulting in a paraphyletic assemblage. This pattern is consistently independent from the size of the data set.
Two different phylogenetic topologies were tested statistically: Sollasella clustering with the Raspailiidae (as suggested by the phylogenetic analysis); and Sollasella clustering with hadromerid taxa (as the previous classification suggested). Table 2 displays the result of the CONSEL analysis. A relationship of Sollasella with Raspailiidae is significantly more likely under the given data than a constellation of Sollasella clustering with Hadromerida.
Table 2. Output of COUNSEL 0.942b on the support of two different hypotheses (Sollasella + Raspailiidae (Sol+Ras) against Sollasella + Hadromerida (Sol+Had). See Shimodaira and Hasegawa (2001) for further details and references.
The molecular analysis of the 28SrDNA provides a clear picture of the phylogenetic position of Sollasella in the classification of demosponges. These 28S sequence data fully support the morphological hypothesis of Van Soest et al. (2006) that Sollasella should be classified within the Raspailiidae of the order Poecilosclerida and not as a hadromerid sponge. The family Sollasellidae Lendenfeld, 1887, which was previously assigned as incertae sedis in the Hadromerida (Van Soest, 2002), has therefore been abandoned.
Our data provides an example of the successful application of molecular tools to sponge phylogeny without contradicting current morphological hypotheses and posing additional questions (e.g., McCormack et al., 2002; McCormack and Kelly 2002; Nichols, 2005; Erpenbeck et al., 2005). Conversely, the resulting phylogenies provide us with another pattern that needs further explanation: the clustering of the Heteroxyidae Myrmekioderma granulata and Didiscus oxeata with the Raspailiidae. Although the coherence of the family Heteroxyidae is unverified and molecular data could not unambiguously assign them to the Halichondrida (Erpenbeck et al., 2005), a relationship with Raspailiidae appears unlikely based on morphometric characteristics. The largely confused arrangement of oxeote megascleres with ectosomal microxeas in Heteroxyidae differs fundamentally from the structured raspailiid skeleton. Although the polygonal grooves found in Sollasella are remarkably similar to those seen in Didiscus and Myrmekioderma , these have probably have been acquired independently. In Didiscus the surface is “strongly grooved with angular striations forming polygonal plates; plates contractile with oscula in between”. In Myrmekioderma the surface is “convoluted with large conules or rounded or polygonal plates, each separated by shallow but distinct grooves, excavated channels containing large oscula” (Hooper, 2002a). This is morphologically and functionally different from Sollasella, whose surface is “provided with a characteristic polygonal pattern of lines of round shallow depressions presumed to be inhalant openings” (Van Soest, 2002, from Hallmann, 1914). Such similarity of 28S rDNA sequences is comparable with other instances in which 28S rDNA resulted in “odd” phylogenies (e.g. McCormack et al. 2002). More intensive studies on the molecular evolution of 28S rRNA genes in demosponges are required to explain such phenomena. Nevertheless, the present analysis has demonstrated that the clustering of Sollasella with the Raspailiidae is clearly no such ‘28S rDNA’ artefact because an alternative data set (morphology) provides independent evidence for this scenario.
The authors thank Penny Berents, Australian Museum Sydney for providing us with type material of S. digitata, Jessica Worthington-Wilmer and the Molecular Identities Laboratory (M.I.L.) of the Queensland Museum for assistance in the lab and furthermore the supercomputing facility of the University of Göttingen (GWDG). Further thanks to Elly Beglinger, Zoological Museum Amsterdam. DE acknowledges financial support of the European Union under a Marie-Curie outgoing fellowship (MOIF-CT-2004 Contract No 2882). GW acknowledges funding from the German Research Foundation (DFG: Projects Wo896/3, 6).
Dohrmann M, Voigt O, Erpenbeck D, Wörheide G. 2006. Non-monophyly of most supraspecific taxa of calcareous sponges (Porifera, Calcarea) revealed by increased taxon sampling and partitioned Bayesian analysis of ribosomal DNA. Molecular Phylogenetics and Evolution 40: 830-843.
Erpenbeck D, Breeuwer JAJ, Parra FJ, van Soest RWM. 2006. Speculation with spiculation? – Three independent gene fragments and biochemical characters versus morphology in demosponge higher classification. Molecular Phylogenetics and Evolution 38: 293-305.
Erpenbeck D, Breeuwer JAJ, van Soest RWM. 2005. Implications from a 28S rRNA gene fragment for the phylogenetic relationships of halichondrid sponges (Porifera: Demospon-giae). Journal of Zoological Systematics and Evolutionary Research 43: 93-99.
Erpenbeck D, Nichols SA, Voigt O, Dohrmann M, Degnan BM, Hooper JNA, Wörheide G. 2007. Phylogenetic analyses under secondary structure-specific substitution models outperform traditional approaches - case studies with diploblast LSU. Journal of Molecular Evolution 54: 543-557.
Hallmann EF. 1914. A revision of the monaxonid species described as new in Lendenfeld’s ‘Catalogue of the Sponges in the Australian Museum’. Part I, II, III. Proceedings of the Linnean Society of New South Wales 39: 263-315, 327-376, 398-446, pls XV-XXIV.
Hooper JNA. 2002a. Family Desmoxyidae Hallmann, 1916. In: Systema Porifera. Guide to the classification of sponges (Ed. by Hooper JNA and van Soest RWM), pp. 755-772. New York, Boston, Dordrecht, London, Moscow: Kluwer Academic/ Plenum Publishers.
Hooper JNA. 2002b. Family Raspailiidae Hentschel, 1923. In: Systema Porifera. Guide to the classification of sponges (Ed. by Hooper JNA and van Soest RWM), pp. 469-510. New York, Boston, Dordrecht, London, Moscow: Kluwer Academic/ Plenum Publishers.
Hooper JNA, van Soest RWM. 2002. Systema Porifera, A guide to the classification of sponges. New York: Kluwer Academic/Plenum Publishers.
Huelsenbeck JP, Ronquist F, Nielsen R, Bollback JP. 2001. Evolution - Bayesian inference of phylogeny and its impact on evolutionary biology. Science 294: 2310-2314.
McCormack GP, Erpenbeck D, van Soest RWM. 2002. Major discrepancy between phylogenetic hypotheses based on molecular and morphological criteria within the Order Haplosclerida (Phylum Porifera : Class Demospongiae). Journal of Zoological Systematics and Evolutionary Research 40: 237-240.
Nichols SA. 2005. An evaluation of support for order-level monophyly and interrelationships within the class Demo-spongiae using partial data from the large subunit rDNA and cytochrome oxidase subunit I. Molecular Phylogenetics and Evolution 34: 81-96.
Posada D, Crandall KA. 1998. MODELTEST: testing the model of DNA substitution. Bioinformatics 14: 817-818.
Ronquist F, Huelsenbeck JP. 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19: 1572-1574.
Shimodaira H, Hasegawa M. 2001. CONSEL: for assessing the confidence of phylogenetic tree selection. Bioinformatics 17: 1246-1247.
Van Soest RWM. 2002a. Family Suberitidae Schmidt, 1870. In: Systema Porifera. A guide to the classification of sponges (Ed. by Hooper JNA and van Soest RWM), pp. 227-244. New York, Boston, Dordrecht, London, Moscow: Kluwer Academic/ Plenum Publishers.
Van Soest RWM. 2002b. Family Sollasellidae Lendenfeld, 1887. In: Systema Porifera. A guide to the classification of sponges (Ed. by Hooper JNA and van Soest RWM), pp. 279-280. New York, Boston, Dordrecht, London, Moscow: Kluwer Academic/ Plenum Publishers.
Van Soest RWM, Hooper JNA, Beglinger E, Erpenbeck D. 2006. Affinities of the family Sollasellidae (Porifera, Demospongiae). I. Morphological evidence. Contributions to Zoology 75: 133-144.
Swofford DL. 2002. PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). Sunderland, Massachusetts: Sinauer Associates.