The Astrophorida (Porifera, Demospongiaep) is geographically and bathymetrically widely distributed. Systema Porifera currently includes five families in this order: Ancorinidae, Calthropellidae, Geodiidae, Pachastrellidae and Thrombidae. To date, molecular phylogenetic studies including Astrophorida species are scarce and offer limited sampling. Phylogenetic relationships within this order are therefore for the most part unknown and hypotheses based on morphology largely untested. Astrophorida taxa have very diverse spicule sets that make them a model of choice to investigate spicule evolution.
With a sampling of 153 specimens (9 families, 29 genera, 89 species) covering the deep- and shallow-waters worldwide, this work presents the first comprehensive molecular phylogeny of the Astrophorida, using a cytochrome c oxidase subunit I (COI) gene partial sequence and the 5′ end terminal part of the 28S rDNA gene (C1-D2 domains). The resulting tree suggested that i) the Astrophorida included some lithistid families and some Alectonidae species, ii) the sub-orders Euastrophorida and Streptosclerophorida were both polyphyletic, iii) the Geodiidae, the Ancorinidae and the Pachastrellidae were not monophyletic, iv) the Calthropellidae was part of the Geodiidae clade (Calthropella at least), and finally that v) many genera were polyphyletic (Ecionemia, Erylus, Poecillastra, Penares, Rhabdastrella, Stelletta and Vulcanella).
The Astrophorida is a larger order than previously considered, comprising ca. 820 species. Based on these results, we propose new classifications for the Astrophorida using both the classical rank-based nomenclature (i.e., Linnaean classification) and the phylogenetic nomenclature following the PhyloCode, independent of taxonomic rank. A key to the Astrophorida families, sub-families and genera incertae sedis is also included. Incongruences between our molecular tree and the current classification can be explained by the banality of convergent evolution and secondary loss in spicule evolution. These processes have taken place many times, in all the major clades, for megascleres and microscleres.
Citation: Cárdenas P, Xavier JR, Reveillaud J, Schander C, Rapp HT (2011) Molecular Phylogeny of the Astrophorida (Porifera, Demospongiaep) Reveals an Unexpected High Level of Spicule Homoplasy. PLoS ONE 6(4): e18318. https://doi.org/10.1371/journal.pone.0018318
Editor: Dirk Steinke, Biodiversity Insitute of Ontario - University of Guelph, Canada
Received: November 3, 2010; Accepted: March 3, 2011; Published: April 8, 2011
Copyright: © 2011 Cárdenas et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Participation of P. Cárdenas at the ‘Molecular Evolution Workshop’ (Woods Hole, 2009) was made possible by the Bergen Forskningsstiftelse (“Mohn Grant”, to C. Schander). Visit to the Zoological Museum in Amsterdam (P. Cárdenas) received support from the SYNTHESYS Project which is financed by European Community Research Infrastructure Action under the FP6 “Structuring the European Research Area” Program. J.R. Xavier is funded by the Science and Technology Foundation (FCT-Portugal, grant no. SFRH/BPD/62946/2009). J. Reveillaud has received funding from the European Community's Seventh Framework Program (FP7/2007-2013) under the HERMIONE project, grant agreement no. 226354. This work was supported by a grant from the ‘Norwegian Deep Sea Program’ and by the Norwegian Research Council (project number 158923/S40, to H.T. Rapp). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Demospongiaep Sollas, 1885 [Borchiellini et al., 2004] make up 85% of all living sponges, and is today subdivided in 13 extant orders. Based on molecular results, Demospongiaep are subdivided in four clades: G1/Keratosap [Borchiellini et al., 2004], G2/Myxospongiaep [Borchiellini et al., 2004], G3/Haplosclerida and G4/Democlavia , . The Astrophorida Sollas, 1888 are found within the Democlavia clade and represent one of the few sponge orders to have been consistently and with strong support, shown to be monophyletic , , , . The Astrophorida is geographically and bathymetrically widely distributed around the world, and represent around 660 extant species (van Soest et al. 2010; this study). In tropical and parts of warm temperate waters Astrophorida species are common at quite shallow depths, while in boreal/antiboreal and Arctic/Antarctic waters they are usually deep-water species. Astrophorida species have colonized hard- as well as soft-bottoms from various depths. In gravely hard-bottom habitats on the outer shelf and upper slope, Astrophorida can dominate ecosystems in terms of abundance and biomass forming sponge grounds , . Astrophorida species display a wide array of external morphologies (massive to thin encrusting, subspherical-, fan-, cup- or irregularly-shaped) and external colors (Fig. 1a–d), and they range in size from a few millimeters to more than a meter in diameter. There is no single morphological synapomorphy of the Astrophorida. They are nonetheless well characterized by the simultaneous presence of star-shaped microscleres (small spicules called ‘asters’) and tetractinal megascleres (large spicules called ‘triaenes’) (Fig. 1e–g). Star-shaped microscleres may be euasters — asters in which the rays radiate from a central point (e.g. oxyasters, strongylasters, spherasters, sterrasters) or streptasters — asters in which the rays proceed from an axis that can be straight (amphiasters) or spiral (e.g. spirasters, metasters, plesiasters). According to the latest major revision of the Astrophorida , five families are included: Ancorinidae Schmidt, 1870, Calthropellidae Lendenfeld, 1907, Geodiidae Gray, 1867, Pachastrellidae Carter, 1875, and Thrombidae, Sollas, 1888. Thirty-eight genera and two subgenera are currently distributed in those families. In an effort to incorporate some lithistids in the Astrophorida, the sub-orders Euastrophorida Reid, 1963 (Astrophorida with euasters) and Streptosclerophorida Dendy, 1924 (Astrophorida/lithistids with streptasters) were erected , , but in spite of molecular evidence confirming their incorporation within the Astrophorida , , , lithistids have been kept apart in the Systema Porifera . Other taxa such as the boring sponges Alectona and Neamphius have also been suggested to be derived Astrophorida species, based on morphological , molecular  and larval data , , but they are still considered to belong to the order Hadromerida in the Systema Porifera .
(a–d) A few Astrophorida species. (a): Geodia phlegraei (Geodiidae) collected in the Denmark Strait. Uniporal oscules are on the top surface. (b): Cross-section of a Stelletta raphidiophora (Ancorinidae) collected on the ‘Schultz Massive’ seamount (Greenland Sea) (ZMBN 85223). The grayish thick cortex is clearly visible. Specimen is 13 cm in diameter. (c): Calthropella geodioides (Calthropellidae) collected South of the Azores (ZMAPOR 21659). (d): Thenea valdiviae (Pachastrellidae) collected on the Norwegian coast. (e): Characteristic Astrophorida microscleres. ox – oxyaster of Geodia papyracea (diameter: 23 µm); st – sterraster of Geodia barretti (diameter: 80 µm); as – aspidaster of Erylus expletus (length: 330 µm); mi – microrhabd of Pachymatisma normani (length: 20 µm); pl – plesiaster of Poecillastra compressa (diameter: 37 µm); sa – sanidaster of Stryphnus raratriaenus; am – amphiaster of Characella pachastrelloides (length: 18 µm); sp – spiraster of Thenea levis (length: 23 µm). (f): cross-section of the cortex of Geodia barretti showing the skeleton organization. ec – ectocortex made of a thin layer of strongylaster and microxeas. en – endocortex made of a thick layer of sterrasters. ch – choanosome. tr – triaene supporting the cortex. Scale: 1 mm. (g): Characteristic Astrophorida megascleres. cal – calthrop of Pachastrella sp. from Norway (actine length: 100 µm); tr – long-shafted triaene of Stelletta sp. from Panama (rhabdome length: 850 µm); dicho – dichotriaene of Characella pachastrelloides (rhabdome length: 500 µm); ana – cladome of anatriaene of Geodia tumulosa from Panama (clad length: 24 µm); disco – discotriaene of Discodermia polymorpha (disc diameter: 180 µm) (photo: A. Pisera); phyllo – phyllotriaene of Theonella sp. (cladome: 730 µm) (photo: A. Pisera).
The Astrophorida is an order with one of the most diverse spicule repertoire among the Demospongiaep. For example, Geodia barretti (Geodiidae, Astrophorida) has up to ten different spicule types while Halichondria panicea (Halichondriidae, Halichondrida) has only one. This spicule diversity within the Astrophorida is ideal to trace spicule evolution and thereby evaluate the importance of homoplasy in this group. Homoplasy (convergent evolution and secondary loss) has always been acknowledged by sponge taxonomists and phylogeneticists but few studies have been able to show to what extent these evolutionary processes occur in sponges, due to the paucity of spicule types and morphological characters. Secondary loss has been particularly difficult to reveal in morphological studies and molecular studies of species with too few spicule types. Meanwhile, the paraphyly and polyphyly of many sponge orders in Demospongiaep and Calcarea (e.g. Haplosclerida, Halichondrida, Clathrinida, Murrayonida) in molecular phylogenetic studies clearly suggest that the evolution of spicules may be more intricate than currently thought , , , , , .
To date, the most complete molecular phylogenetic study focusing on the Astrophorida is based on ten species belonging to six families, including two species of lithistids . Other Demospongiaep molecular phylogenies include only three to six species of Astrophorida , . With over 660 species of Astrophorida described worldwide , needless to say that phylogenetic relationships within this order are for the most part unknown and hypotheses based on morphology largely untested. And, since Astrophorida families might not be monophyletic , any Astrophorida phylogenetic study needs to have the broadest sampling as possible, from the five Astrophorida families as well as from putative Astrophorida (lithistids, Alectona, Neamphius). With a sampling of 153 specimens (9 families, 89 species) covering the deep- and shallow-waters of the Atlantic, Pacific, Indian, and Southern Ocean, the overall aim of this work was to present the first comprehensive molecular phylogeny of the Astrophorida. More specifically, the first aim of this study was to test the monophyly i) of the Euastrophorida/Streptosclerophorida sub-orders and ii) of the Astrophorida families and genera. Our second aim was to revise the taxonomy of this order using both the classical rank-based nomenclature (i.e. Linnaean classification) and the phylogenetic nomenclature following the PhyloCode, independent of taxonomic rank. To be clear, names established under the PhyloCode are always in italics and will be identified with the symbol ‘p’ (e.g. Demospongiaep). Authors of PhyloCode names are between square brackets (e.g. Demospongiaep Sollas, 1885 [Borchiellini et al., 2004]). Finally, our third aim was to investigate the evolution of Astrophorida megascleres and microscleres in order to evaluate the importance of homoplastic spicule characters in this order.
Materials and Methods
This study has been approved by the University of Bergen through the acceptance of a Ph.D. project proposal.
Most of our collecting was done in the Northeast Atlantic. Sampling in the Korsfjord (60°10′N, 05°10′E), Langenuen (59°53′N, 05°31′E) and the Hjeltefjord (60°24′N, 05°05′E) (Western Norway, south of Bergen) were carried out using a triangular dredge and a bottom trawl between 40 and 500 meters (between the years 2005 and 2009). Southern Norway samples (58°13′N, 08°35′E) were dredged during the BIOSKAG 2006 cruise. Northern Norway samples were collected during the Polarstern ARK-XXII/1a 2007 cruise with large boxcores and the Jago manned-submersible. Localities sampled were Sotbakken (70°45′N, 18°40′E), Røst reef (67°30′N, 9°24′E) and Trænadjupet (66°58′N, 11°7′E). Greenland Sea samples were collected on the “The Schultz Massive” seamount (73°47′N, 07°40′E) during the BIODEEP 2007 and H2DEEP 2008 cruises using the ROV Bathysaurus XL. Samples from Bocas del Toro (9°20′N, 82°15′E, Panama, Atlantic), Berlengas Islands (39°24′N, 09°30′W, Portugal) and the Azores Islands were collected by snorkeling/diving. The Gorringe Bank (36°31′N, 11°34′W) specimens were collected by diving during Luso Expedição 2006 . Samples from deep-water coral reefs off Cape Santa Maria di Leuca (Ionian Sea, Apulian Plateau, 39°33′N, 18°26′E) were collected with the ROV Victor and an Usnel core during the ‘Ifremer MEDECO 2007’ cruise. Samples of the seamounts Southern of the Azores were collected in the course of the campaigns EMEPC-G3-2007/2008 of the Task Group for the Extension of the Continental Shelf (EMEPC, Portugal) employing the ROV Luso. Other samples were kindly provided by different institutions and scientists (cf. Acknowledgments). Hologenophores — a sample or preparation of the same individual organism as the study organism  — were preserved in 95% ethanol and stored at room temperature at the Bergen Museum. Species, voucher numbers, Genbank accession numbers and collecting localities are given in Table S1.
Outgroups belong to the Spirophorida since all previous Demospongiaep molecular phylogenetic studies place them in a strongly supported sister-order relationship with the Astrophorida , , , ,  (see also the comprehensive COI, 18S and 28S phylogenetic Demospongiaep trees on the Sponge Genetree Server, www.spongegenetrees.org/, accessed on the 15th of October 2010).
Specimens collected were identified to the genus and species level by P. Cárdenas, H. T. Rapp and J. R. Xavier. Identifications of specimens donated by other institutions were also checked. Astrophorida vouchers from previous studies , , , ,  were re-examined by us or by others ,  and in some cases, given new identifications (Table S2. Some of the voucher specimens sequenced have been morphologically described previously: Pachymatisma species  and all specimens collected in Panama . The Norwegian Pachastrellidae specimens will be described and reviewed in a separate paper.
Isops and Sidonops are synonyms of Geodia ; Isops and Sidonops species of this study were therefore all transferred to Geodia. Geodia neptuni Sollas, 1886 has been synonymized with Geodia vosmaeri Sollas, 1886 . Erylus euastrum has been transferred to the genus Penares, owing to molecular and morphological results . The lithistid Exsuperantia sp. corresponds to Racodiscula clava sensu Topsent, 1892 from the Azores  which had been re-identified as Rimella sp. , later found to be a preoccupied genus . Because Thrombus abyssi can have variable spicule morphologies , it is important to note that our specimens have amphiasters and trichotriaenes with an extension of the rhabdome.
DNA extraction, amplification and sequencing
Two independent genes were used for this study: the Folmer fragment of the mitochondrial cytochrome c oxidase subunit 1 (COI) and the 5′ end terminal part of the nuclear 28S rRNA gene. These have previously been shown to give robust and congruent results for Geodiidae relationships . DNA extraction from choanosome samples was performed using the Tissue Genomic DNA extraction kit (Viogene, Sunnyvale, CA, U.S.A.) in accordance with the manufacturer's instructions. A single centrifugation step was added just before pipeting the mixture into the columns in order to remove the spicules. For some species (Pachastrella sp. and Stryphnus raratriaenus), polymerase chain reactions (PCR) worked only when the DNA was extracted following a standard chloroform protocol extraction. The 5′ end region of COI (659 bp.) was amplified using LCO1490 and HCO2198  (5 min/94°C; 5 cycles [30 s/94°C, 1 min30 s/45°C, 1 min/72°C]; 30–35 cycles [30 s/94°C, 1 min30 s/50°C, 1 min/72°C]; 7 min/72°C). C1′ASTR (5′–ACC CGC TGA ACT TAA GCA T–3′)  and the D2 (5′–TCC GTG TTT CAA GAC GGG–3′)  reverse universal primer were used to amplify a 768–832 bp. region of 28S comprising part of the C1 domain, and the total of the D1, C2 and D2 domains  (1 cycle [4 min/95°C, 2 min/59–60°C, 2 min/72°C]; 35 cycles [1 min/94°C, 45 s/59°C, 1 min/72°C]; 7 min/72°C). In some cases, C1′ASTR did not work and we used an intermediate primer instead: Ep1a' (5′–GGC AGA GGC GGR TGC ACC–3′) . Sequences were then shorter, ca 690 pb (1 cycle [4 min/95°C, 2 min/59°C, 2 min/72°C]; 35 cycles [1 min/94°C, 45 s/59°C, 1 min/72°C]; 7 min/72°C). PCR products were purified using the ExoSAP-IT® kit (USB Europe, Staufen, Germany) or gel purified using a Gel-M™ Gel Extraction System (Viogene). Cycle sequencing was performed using a dye-labeled dideoxy terminator (Big Dye® Terminator v3.1, Applied Biosystems, Foster city, CA, U.S.A.). Products were analyzed using an ABI Prism 3700 DNA Analyzer (Applied Biosystems). The Astrophorida origin of the sequences was checked by BLAST searches (http://blast.ncbi.nlm.nih.gov).
Sequence alignments and phylogenetic analyses
The COI data matrix includes 118 sequences (with outgroups) of which 86 are new. 245/660 characters are parsimony informative. The 28S data matrix includes 108 sequences of which 80 are new and 9 are lengthened since Cárdenas et al. . 381/864 characters are parsimony informative. COI sequences were manually aligned in Se-Al v2.0a11 . 28S sequences were first automatically aligned using MAFFT v.6.705  with default parameters, implemented in SeaView v.4.1 . Four insertion-deletion regions (4–20 bp long) in the D2 domain were ambiguous to align and regional realignments using the MAFFT's ENSI strategy were computed on these four regions. The alignment was subsequently improved visually using Se-Al.
Altogether, maximum likelihood (ML) analyses were conducted on four datasets: COI, COI amino-acids, 28S and 28S+COI. 28S (D1-D2) and COI have been shown to evolve at similar rates , so the two datasets were concatenated in a single matrix containing a total of 148 Astrophorida specimens (29 genera, 2 sub-genera, 89 species) and 1,527 characters, of which 811 are constant, 110 are parsimony uninformative and 606 parsimony informative. For some species we had both markers, but in different specimens from the same region (e.g. Stelletta normani from Western Norway, Geodia megastrella from the Hebrides Islands, Pachastrella ovisternata from the NEA). The sequences of these specimens were concatenated in the final matrix. Overall, we had a sequence for both genes for 67 specimens and 59 species of Astrophorida. ModelTest 3.7  and ProtTest 2.4  were used to find the most appropriate models of evolution respectively for the nucleotide datasets and the amino-acid dataset. For COI, COI amino-acids, 28S and COI+28S, the models were respectively (according the Akaike Information Criterion): HKY+I+G, metREV+G, TrN+I+G and GTR+I+G. For ML runs and bootstrap analyses we used GARLI v.0.96  and Grid computing  through The Lattice Project , which includes clusters and desktops in one encompassing system . A Grid service for GARLI was developed using a special programming library and associated tools . Following the model of Cummings et al. , who used an earlier Grid computing system , the Astrophorida data matrix was distributed among hundreds of computers, where the analyses were then conducted asynchronously in parallel. 100 ML search replicates were run for each dataset. Each replicate was run with a random starting topology and for 5,000,000 generations. Lscores of the 100 best trees from each replicate were re-estimated in PAUP* 4.0b10  and trees were compared using the Symmetric Difference (Robinson-Foulds) tree distance metric, essentially to make sure the best trees collected had similar topologies. 2,000 bootstrap replicates were conducted for each of these four datasets.
To investigate spicule evolution, we reconstructed the microscleres and megascleres states at ancestral nodes on the molecular tree using likelihood reconstruction methods under the Mk1 model , with the help of Mesquite 2.74  and a morphological matrix with 13 characters combined from our observations and from species descriptions in the literature (Table S3).
Astrophorida species can be found at various depths. To investigate a possible relationship between depth, evolution of spicules and/or phylogeny, we have color-coded shallow and deep-water species (>100 m) in the character states reconstructions. Shallow submerged cave environments are prone to harbor deep-water sponge species , , so specimens collected in shallow Mediterranean caves were considered as deep-water species if records outside caves were in deep-water: this concerns Penares euastrum, Calthropella pathologica, Discodermia polymorpha and Neophrissospongia nolitangere. Stelletta lactea and Penares helleri were the only species to appear in both shallow and deep waters.
Phylogenetic classification of the Astrophorida
Following our effort to revise sponge classification as we construct new molecular phylogenies , we followed the principles of phylogenetic nomenclature under the rules of the PhyloCode v.4c (http://www.ohiou.edu/PhyloCode/) to build a phylogenetic classification based on our results. Phylogenetic nomenclature provides the opportunity to propose taxonomical changes while waiting for independent evidence to confirm them, and before implementing those changes to the more widely used rank-based Linnaean classification. This is particularly important to reduce the phylogeny/classification gap. It is also very useful for intra-genera relationships (e.g. in Geodia) where the rank-based classifications are insufficient to name and describe all the clades present . We named clades that have a bootstrap higher than 70 in the 28S+COI analysis. For the use and establishment of clade names, including species names, we will follow Cárdenas et al. .
The best tree resulting from the COI amino-acids analyses is poorly resolved with very few supported clades (Fig. S1). The best trees from the COI analyses (Fig. S2) and the 28S analyses (Fig. S3) are well resolved and congruent except for a few deep poorly-supported nodes. The main topology differences between the COI and 28S trees are: i) Alectona clusters with the Spirophorida outgroups (28S) or with the rest of the Astrophorida (COI); ii) Thenea and Poecillastra+Vulcanella form a monophyletic group (28S) or not (COI); iii) Geodinaep Sollas, 1888 [Cárdenas et al., 2010] cluster either with the Erylinaep Sollas, 1888 [Cárdenas et al., 2010] (COI) or with some Ancorinidae (28S).
The best tree from the 28S+COI analyses (Fig. 2) is fairly close to the COI tree except for the poorly-supported positions of Pachastrella, Poecillastra and Vulcanella (Vulcanella). From now on, we will present the results of the best tree obtained with the 28S+COI dataset (Fig. 2), unless significant topology differences were observed in the analyses of the other datasets. Parameters estimated by GARLI for the best 28S+COI tree were (lnL = −19335.557146; A = 0.191611; C 0.247736; G = 0.290797; T = 0.269856; R-matrix = (1.137933 3.456486 1.476993 0.844493 4.787326); pinv = 0.367474; α = 0.557592). Out of the 100 best trees (each obtained from a different ML replicate), the first 66 trees (19335.56<PAUP* lnL<19336.10) had only minor topology differences, essentially within the Geodinaep and the Erylinaep. The best tree presented and discussed here is the one with the highest score (−lnL = 19335.56); it is also representative of more than half of the trees found.
Bootstrap nodal support values >50 are given at the nodes (2,000 replicates). Species names (according to the Linnaean classification) and sampling localities are given in Table S1. Names established under the PhyloCode are in italics and identified with the symbol ‘p’.
Geodiidae, Calthropellidae and Ancorinidae
Astrophorida (including lithistids, Alectona and Neamphius) was monophyletic in all analyses except for the 28S analyses, were Alectona was within the Spirophorida outgroups. Out of the 100 best trees retrieved from the 28S+COI analyses, the first 76 trees suggested identical topologies concerning the relationships between the Geodiidae, Calthropellidae and Ancorinidae. The Geodiidae and the Ancorinidae were not monophyletic, while the Calthropellidae was monophyletic (but with only one genus sampled: Calthropella). Some Ancorinidae genera were distributed within the Geodiidae while the rest clustered in the Ancorinidae sensu stricto. Furthermore, some of the Ancorinidae genera appeared polyphyletic: i) within Geodinaep (Ecionemia and Rhabdastrella), or ii) distributed between Geodinaep and Ancorinidaep (Stelletta). Melophlus sp., another Ancorinidae, clustered with Caminus vulcani in the Erylinaep.
Geodiidaep Gray, 1867 [Cárdenas et al., 2010] is poorly supported, but retrieved in the COI analyses (Fig. S2) and in the first 76 best trees of the 28S+COI analyses (Fig. 2). The 77th best tree offers a new topology: ((Geodinaep+Ancorinidae s.s.) Erylinaep). When we go from tree 76 to tree 77 we go from lnL = −19937.93 to lnL = −19939.79, a significant jump in likelihood when compared with the lnL very slow decrease from tree 51 to tree 76. We therefore also ran constrained analyses on the 28S+COI dataset (100 ML replicates) forcing the Geodinaep and Ancorinidae s.s. together. The best constrained tree scored a lnL = −19339.79 (same as our tree number 77). An Approximately Unbiased (AU) test using CONSEL v.0.1j  showed that the best constrained and unconstrained trees were not significantly different (P-value = 0.395), so both topologies are plausible according to our molecular data. We should also note that the ((Geodinaep+Ancorinidae s.s.) Erylinaep) topology is also retrieved in the 28S analyses (Fig. S3). Geodinaep and Erylinaep were both strongly supported (bootstraps of 96). Erylus and Penares were both found polyphyletic, with most Erylinaep internodes poorly supported. Within Geodinaep, Depressiogeodiap [Cárdenas et al., 2010] and Geodiap Lamarck, 1815 [Cárdenas et al., 2010] were strongly supported (boostraps of 99), while Cydoniump Fleming, 1828 [Cárdenas et al., 2010] was moderately supported (boostrap of 86). All species for which we had sampled more than one specimen were found monophyletic except for Geodia cydonium (the British specimens were clearly separated from the Mediterranean/Portuguese specimens, K2P distance = 0,04606), Geodia gibberosa (paraphyletic) and Penares helleri (paraphyletic). Geodia simplicissima and Geodia barretti had identical COI sequences.
A Calthropella+Geodia intuta clade appeared as sister-group to Erylinaep. This topology was poorly supported (bootstraps of 66 and 72) but retrieved in all ML replicates.
Ancorinidae sensu stricto
The Ancorinidae s.s. have the most recent common ancestors with lithistids, Characella pachastrelloides (Pachastrellidae) and Neamphius huxleyi (Alectonidae). The Ancorinidae s.s. included Asteropus, Stryphnus, Ancorina and some Stelletta (henceforth called Stelletta sensu stricto). Stryphnus and Stelletta s.s. appeared paraphyletic, the first one because of the placement of Asteropus sp., the second because of Ancorina sp.. Dercitus bucklandi (Pachastrellidae) was found basal to the Stryphnus+Asteropus clade. As detailed above, a few 28S+COI trees (with lower likelihoods) and the 28S analyses suggested that the Ancorinidae s.s. was sister-group to Geodinaep.
Pachastrellidae and lithistids
The Pachastrellidae appeared as a polyphyletic group distributed in four clades: clade 1) Characella pachastrelloides, clade 2) Pachastrella+Poecillastra amygdaloides+Triptolemma intextum, clade 3) Poecillastra compressa+Vulcanella(Vulcanella) and clade 4) Thenea/Vulcanella(Annulastrella). As a result, Thenea and Pachastrella were monophyletic while Poecillastra and Vulcanella were polyphyletic. C. pachastrelloides is grouping next to the lithistids. Clade 2 was found to be sister group to the Geodiidaep clade but this was very poorly supported (boostrap<50). Clade 2 moved closer to the Erylinaep and Calthropella in the COI and 28S analyses. Clade 3 and 4, both very well-supported, appeared closer to the base of the Astrophorida clade, but the nodes were moderately to poorly supported (bootstraps of 68 and 53). In the 28S analyses, Clade 3 and 4 form a poorly-supported monophyletic clade. In the COI analyses, Clade 3 is sister-group to the Geodiidaep, the branch is very short and poorly-supported.
The lithistids were here limited to three families two of which (Corallistidae and Phymaraphiniidae) were only represented by a single species. N. nolitangere and Exsuperantia sp. were found close to C. pachastrelloides but this was poorly supported (bootstraps<50). With three species sampled, the Theonellidae was found monophyletic (bootstrap of 100).
Thrombidae and Alectonidae
With two species sampled, the Alectonidae was found polyphyletic. Alectona millari branched between the Thrombidae and the rest of the Astrophorida. In the 28S analyses, Alectona was placed between the Cinachyrella and Craniella outgroups. Neamphius huxleyi was sister-group to the Ancorinidae s.s. but this association was not supported (bootstrap<50). In the COI analyses, N. huxleyi branched with the lithistids, but not far away from the Ancorinidae s.s.; this position was not supported either. Thrombus abyssi is the most basal Astrophorida, branching before A. millari.
Maximum likelihood reconstruction of ancestral states
Mapping of the 13 characters on the molecular tree gave us 13 trees, each with relative probabilities for every character state for every node in the tree. We have summarized these results for megascleres (Fig. 3) and microscleres (Fig. 4) by only showing character states with 0.65>p>0.95, and p>0.95. Numerous cases of spicule convergent evolution and secondary losses are revealed.
The ML reconstructions of the ancestral conditions at the nodes were estimated using Mesquite 2.74. For the readers' convenience, species clades have been reduced to one sample (except in cases of para- or polyphyletic species). Species names in blue represent deep-water species. Species names in black represent shallow-water species. For the color-codes of the Astrophorida families sensu Systema Porifera, see Figure 2.
The ML reconstructions of the ancestral conditions at the nodes were estimated using Mesquite 2.74. For the readers' convenience, species clades have been reduced to one sample (except in cases of para- or polyphyletic species). Species names in blue represent deep-water species. Species names in black represent shallow-water species. For the color-codes of the Astrophorida families sensu Systema Porifera, see Figure 2.
On a total of 89 species sampled, we found 43 to be shallow and 46 to be deep-sea species. If we consider secondary losses of megascleres with p>0.95, we found 9 losses in shallow-species vs. 2 losses in deep-sea species (Fig. 3). We note there are no losses of triaenes in deep-sea species. If we consider secondary losses of microscleres with p>0.95, we found 14 losses in shallow-species vs. 5 losses in deep-sea species (Fig. 4).
Convergent evolution can be difficult to identify since we often have low probabilities for all character states in deep ancestors. With such an uncertain ancestor separating two clades, we cannot be sure that a spicule appearing in a clade is homologous to the same spicule type in the other clade, or not (e.g. microxeas, amphiasters). We nonetheless notice that convergent evolution is also quite frequent and concerns nearly all types of microscleres (amphiasters, toxas, sanidasters, euasters, aspidasters, microrhabds and possibly microxeas) and megascleres (short- and long-shafted triaenes, discotriaenes, phyllotriaenes, anatriaenes, calthrops). Desmas may have also appeared independently three times.
Astrophorida and phylogenetic classification
A phylogenetic classification of the Astrophorida, henceforth named Astrophoridap, is presented in File S1 and summarized in Figure 5. Names have been given to the well-supported clades (boostraps >70). Rank-based names have also been given to clades for which no names existed in the Linnaean classification. Moreover, new definitions of families and genera were also required. The revised Astrophorida Linnaean classification is presented in File S2.
Species names are given according to the PhyloCode (Article 21.5). Bootstrap nodal support values of clades defined by the PhyloCode are given (2,000 replicates). For the color-codes of the Astrophorida families sensu Systema Porifera, see Figure 2.
Very early on, sponge taxonomists subdivided the Astrophorida between those that possessed streptaster and those that possessed euasters : Streptosclerophorida and Euastrophorida respectively. Chombard et al.  previously found the Euastrophorida monophyletic and the Streptosclerophorida paraphyletic because they had mainly sampled Geodiidaep species, except for Stryphnus mucronatus that they had classified as a Streptosclerophorida (on the basis that its sanidasters were homologous to streptasters). However, our study suggests that both sub-orders are polyphyletic (irrespective of the nature of the sanidasters of Stryphnus). Therefore, we propose to formally abandon the two suborders Euastrophorida and Streptosclerophorida.
Geodiidaep and reallocated Ancorinidae
Since the last molecular phylogeny of Geodiidaep , we lengthened the 28S sequences and increased the sampling from 24 to 38 Geodiidae species and from 24 to 62 Geodiidae specimens. We also added species from phylogenetically close families (Ancorinidae and Calthropellidae). Clearly, Geodiidaep is poorly supported in our 28S+COI best tree (Fig. 2), but morphological data  and a majority of our 28S+COI best trees support the Erylinaep+Geodinaep grouping. This is therefore the topology we will discuss in this paper. However, as we stated earlier (cf. Results), the alternative topology Erylinaep(Geodinaep+Ancorinidae) found in a few 28S+COI searches and 28S analyses could not be rejected on statistical grounds. The contentious Geodiidaep node should therefore be investigated further with additional molecular markers.
The Geodiidae is here redefined: it appears as a much larger family than expected since it includes genera from the Calthropellidae and Ancorinidae. This is surprising for a group whose monophyly and morphological synapomorphies appear quite clearly . To understand this, we must consider the morphology of the unexpected groups. The Ancorinidae is partly composed of species which have the same set of spicules as the Geodiidae except for the presence of sterrasters (ball-shaped euasters, Fig. 1e). Consequently, these Ancorinidae may have never had sterrasters or they may have secondarily lost them. In the second case, these species should be reallocated within the Geodiidae.
Penares is one of these former Ancorinidae genera reallocated to the Geodiidae based on morphological, molecular and biochemical data , . To understand this reallocation, it was hypothesized that Penares helleri had secondarily lost its sterrasters . Our study confirms this reallocation by adding two other species of Penares. Furthermore, the latter double the frequency of the secondary loss of sterrasters since our results suggest that Penares is polyphyletic, just like its counterpart Erylus. Secondary loss of sterrasters therefore happened at least twice in two different newly named clades: Penaresp (P.p euastrum, P.p helleri and P.p sclerobesa) and Erylusp (E.p discophorus, E.p mamillaris, E.p deficiens, E.p sp., E.p granularis and E.p candidata) (Fig. 4, Fig. 5). If it happened twice, it could have happened more, and this is what the placement of Erylusp sp. (an Erylus with no aspidasters) and other genera of Ancorinidae within the Geodiidaep suggest: Melophlus sp., Rhabdastrella, Ecionemia, and some Stelletta would have also lost their sterrasters (Fig. 4). As in the example of Penaresp, this is fairly easy to conceive since these Ancorinidae species share i) spicule repertoires identical to the Geodiidaep except for the presence of sterrasters, and often ii) a similar external morphology (e.g. oscule organization). Despite these similarities, the placement of the polyphyletic Rhabdastrella and Ecionemia within the Geodinaep is not straightforward.
Based on the possession of microrhabds in the cortex, Chombard et al.  wondered if Ecionemia should be reallocated to the Erylinaep. Our analysis suggests that the three Ecionemia species sampled belong to the Geodinaep, and are distributed in two groups. The two Australian Ecionemia group with some Stelletta — thus forming the new clade Geostellettap — while Ecionemia megastylifera from the Caribbean is branching at the base of Cydoniump. These three species of Ecionemia all share large spiny microrhabds in the cortex along with euasters. Since microrhabds are absent from all the other Geodinaep of this study, the origin of these microrhabds is uncertain at this point and may represent yet another case of morphological spicule convergence in sponges (Fig. 4). Other species of Ecionemia, with small sanidaster-like microrhabds (e.g. E. acervus, type species of the genus, E. demera, E. walkeri), might instead be linked to sanidaster-bearing Ancorinidaep as previously suggested , , . In our opinion, the genus Ecionemia should therefore be kept valid for the remaining species of Ecionemia whose phylogenetic positions remain to be tested.
Based on its spicules and skeleton organization, Rhabdastrella has previously been suspected to be close to the Geodiidae  or even part of the Geodiidae . Biochemical data also concurs with this result: isomalabaricane triterpenes have been found in R. globostelletta and Geodia japonica , . Rhabdastrella species from our study are distributed in three groups: 1) R. globostelletta and Rhabdastrella sp. form a clade of uncertain position within the Geodinaep, 2) R. cordata from Australia forms a strongly supported group with Geodiap pachydermata and Geodiap sp. 2, both from the Atlantic/Mediterranean area, and 3) R. intermedia forms a strongly supported clade with Geodiap phlegraei. Rhabdastrella species are characterized by sterrospherasters in the cortex. Sterrospherasters is a general ambiguous term that includes two main types of large euasters: i) very large spherasters with smooth conical rays, filling the whole cortex (e.g. R. globostelletta and Rhabdastrella sp.) or ii) sterrasters, sometimes with incompletely fused actines (e.g. R. rowi, R. aurora, R. cordata), placed in the endocortex. These morphological observations coupled with our results suggest that these sterrospherasters might actually be, in the first case, true spherasters — they resemble the ones found in the phylogenetically close G.p phlegraei and G.p angulata — and are, in the second case, true sterrasters. Rhabdastrella with true spherasters may therefore have secondarily lost their sterrasters (and these have been replaced by the large spherasters). In light of these results we expect all Rhabdastrella species to be redistributed in Geodinaep. The genus Rhabdastrella is therefore not valid and should be synonymized with Geodia. As a consequence of the polyphyly of Rhabdastrella, the confusing spicule term ‘sterrospheraster’ should be once and for all rejected, as suggested before .
We should not be surprised to find Ancorinidae species with microrhabds such as Melophlus sp. grouping with Caminus vulcani (an Erylinaep with spherules) since it has been argued that spherules may have evolved from microrhabds . Furthermore, like the rest of the Erylinaep, Melophlus sp. has no ana/protriaenes. The phylogenetic position of Melophlus sp. among the Erylinaep may be further supported by biochemical data: sarasinoside M, a triterpenoidal saponin isolated from Melophlus sarassinorum, has strong similarities with the framework of Eryloside L, isolated in Erylus lendenfeldi .
To conclude, the reallocation of numerous Ancorinidae species in the Geodiidae calls for new definitions for these families (File S2).
Most of the clades found in this study are identical to those found previously with fewer species and a shorter 28S fragment . Geodiap, Cydoniump and Depressiogeodiap were still strongly supported groups. The Depressiogeodiap+Cydoniump clade, poorly supported in Cárdenas et al. , was better supported here (bootstrap of 77), it exclusively grouped Atlantic species. In the following paragraphs, we will go through these clades and discuss new taxonomical results that have arisen due to the addition of new species since Cárdenas et al. .
The addition of Geodiap corticostylifera from Brazil confirmed that the Geodiap include species from North and South America, from the Atlantic and Pacific sides. Different clades of Geodiap vosmaeri (former G.p neptuni) appeared, two from Florida, another from Belize+Bahamas suggesting i) a strong geographical structure and that ii) the molecular markers used may be suited for future intra-specific studies. Our results confirmed that Geodiap gibberosa represented a species complex, as previously hypothesized with morphological observations . We propose that G.p tumulosa Bowerbank, 1872 (a synonym of G.p gibberosa) should be resurrected for the mangrove specimen from Panama. Its tumulose shape is clearly different from the barrel-shape of our reef specimens from Belize and Mexico, more similar to the shape of the holotype of G.p gibberosa (specimen MNHN DT-608).
Geodiap conchilega and E. megastylifera are part of Cydoniump so this clade still gathers Atlanto-Mediterranean species. The polyphyly of Geodiap cydonium calls for a revision of this species whose taxonomical history is old and complex.
Geodiap megastrella is part of the Depressiogeodiap. This clade thus remained a Northeast Atlantic deep-water species group. The inclusion of G.p megastrella in the Depressiogeodiap also confirmed a suggested morphological synapomorphy of the group: a deep preoscule lacking sterrasters in its cortex . It should be noted that the G.p megastrella ZMBN 85208 (Scotland) and ZMAPOR 21654 (Azores) both had a distinct large deletion (35 bp long) in their 28S D2 domain while ZMAPOR 21231 (Morocco) appeared to have a slightly different sequence, notably without the deletion. This specimen's morphology needs to be further investigated as G.p megastrella may represent a species complex.
The two deep-water Geodiap species from New Caledonia grouped together but this is poorly supported. The most basal Geodinaep was a strongly supported clade named Synopsp grouping G.p pachydermata, Geodiap sp. 2 and R. cordata. The surprising phylogenetic position of Geodia intuta with Calthropella will be discussed below. The positions of other Geodinaep species (e.g. G.p phlegraei, G.p angulata) were poorly supported and uncertain (different positions in different trees) so we cannot discuss their taxonomy at this point.
Erylinaep was a very strongly supported group (boostrap of 96). The monophyly of Erylus has been previously challenged by morphological and molecular data , . Our results suggested that it was a polyphyletic genus, mixed with Penares, Caminus, Melophlus and Pachymatismap species. Erylus species were distributed in three clades: Erylusp (‘nomen cladi conversum’ because it holds the type species of Erylus: E.p mamillaris), Penaresp (‘nomen cladi conversum’ because it holds the species type of Penares: P.p helleri) and Erylus1 (temporary name for the clade including E. aleuticus+E. expletus+E. topsenti, poorly supported). If Erylus is polyphyletic, the most parsimonious scenario is that flattened sterrasters ( = aspidasters) have appeared independently at least three times; this is also suggested by our character reconstruction using ML methods (Fig. 4). Our study has not revealed the identity of Erylusp sp. collected in the Gorringe Bank . Erylusp sp. which has lost its aspidasters was part of the E.p mamillaris/discophorus complex, but more rapidly evolving markers are required to fully understand this group.
Calthropellidae and Geodia intuta
The association of calthrops and euasters essentially characterizes the Calthropellidae. According to some morphologists, the Calthropellidae do not really have characters of their own and should be within the Ancorinidae , , . However, the first molecular evidence suggested a sister-group relationship between the Calthropellidae and the Erylinae . Although the Erylinaep(G. intuta+Calthropellap) association was weakly supported (bootstrap of 66) it was present in all our trees obtained from the 100 ML searches. Furthermore, the external morphology of Calthropellap geodioides and some basal Erylinaep species (e.g. E. expletus) is quite similar: they are massive sub-spherical sponges with numerous white uniporal oscules on the top surface. We propose to reallocate the Calthropellidae to the Geodiidae by downgrading them to a sub-family: the Calthropellinae. Paxataxa and Corticellopsis are the other genera of the Calthropellinae since Chelotropella has been reallocated to the Ancorinidae . Sequences of Pachataxa and Corticellopsis are therefore needed to confirm the monophyly and the position of this group.
The clustering of Geodia intuta with Calthropellap was surprising, but less so when reconsidering its external and spicule morphologies. Like Erylusp and Penaresp, G. intuta is a massive sub-hemispherical sponge with a smooth cortex, it is easily compressible, and has a rather confused skeleton organization. It was originally described as an Isops because of its uniporal oscule and pores. According to our observations, the oscule actually leads to a branching atrium, similar to the ones found in Erylusp, Penaresp or Caminus. This prompted von Lendenfeld  to describe it in a new genus, as Caminella loricata, before it was synonymized with Geodia intuta . Moreover, it has long-shafted triaenes (as in the Geodinaep) but no ana/pro/mesotriaenes (as in the Erylinaep). It has spherasters in the ectocortex and globular sterrasters in the endocortex. Globular sterrasters are also present in many Erylinaep (e.g. Caminus, Pachymatismap, E. topsenti). As for spherasters, they resemble the spherules found in C. vulcani (an Erylinaep) or Calthropellap durissima. All in all, although G. intuta shares many characters with some Erylinaep (Erylusp, Penaresp, Caminus), the presence of long-shafted triaenes and the absence of microrhabds suggest that it is not an Erylinaep. Therefore, we decided to resurrect the Geodiidae genus Caminella von Lendenfeld, 1894 to welcome this species. On the other hand, we will wait for further data to confirm its phylogenetic position and name the G. intuta+Calthropellap clade.
Ancorinidae sensu stricto form a well-supported clade henceforth named Ancorinidaep. Stelletta species were distributed in three Ancorinidaep clades: clade 1) (Ancorina sp.+Stelletta sp. 1)+Stelletta clarella, clade 2) (Stelletta normani+Stelletta raphidiophora)+Stelletta lactea and clade 3) (Stelletta grubii+Stelletta carolinensis)+Stelletta dorsigera. Clade 1 was poorly supported (bootstrap<50). Clade 2 clustered three Northeast Atlantic species; it was very well supported by our data (bootstrap of 98) and by the synapomorphy of trichodragmas (raphides in bundles) (Fig. 4): it was therefore named Dragmastrap. Clade 3 held the type species of the genus (S. grubii) so it was named Stellettap. It should be noted that S. dorsigera does not group with S. grubii in the 28S analyses (Fig. S3). The unstable position of S. dorsigera may be due to the fact that the Stelletta COI sampling is quite poor with respect to the Stelletta 28S sampling. The grouping of clade 1+Dragmastrap is poorly supported or absent (28S analyses) but we nonetheless note that all of these species have dichotriaenes, except for Ancorina sp.. Conversely, species in the Stellettap clade do not possess dichotriaenes. Instead, 28S analyses fully support a Dragmastrap+Stellettap clade (Fig. S3).
Since Ancorina and Stryphnus share similar spicule repertoires , notably the presence of sanidasters (Fig. 4), we were expecting them phylogenetically closer to each other than here observed. But the grouping of Ancorina sp. with two Stelletta species was poorly supported and may be due to the poor sampling of these speciose genera.
The close relationship between Asteropus and Stryphnus has often been discussed , , , , , . Both genera have similar spicules, except for triaenes that Asteropus would have secondarily lost (Fig. 3). For the first time, the synonymy of Asteropus with Stryphnus is confirmed by molecular results. Therefore, we formally propose that Asteropus becomes a junior synonym of Stryphnus and name this clade Stryphnusp.
The presence of Dercitus bucklandi — a Pachastrellidae with calthrops, sanidasters and toxas — within the Ancorinidaep is once more supported by morphological data. Dercitus (Stoeba included) and Stryphnus notably share sanidasters, large spherulous cells, and a similar aquiferous system , , . But other authors had considered that the origin of the toxas being ambiguous, emphasis should instead be placed on the presence of calthrops, which had brought Dercitus closer to the Pachastrellidae , , , . D. bucklandi as an Ancorinidaep suggests that toxas would have originated from asters, as previously hypothesized . The modification of oxyasters into toxa-like spicules is actually quite common in the Astrophoridap (e.g. Erylus nummulifer, Erylus expletus, Geodia apiarium, Erylus papulifer, Rhabdastrella oxytoxa and Stelletta toxiastra). The difference between the latter and D. bucklandi, which troubled morphologists, is that toxas in D. bucklandi have completely lost trace of the original euaster centrum. The position of D. bucklandi also shows that its sanidasters are homologous to those of Stryphnusp (Fig. 4). Unfortunately, we did not get 28S sequences for D. bucklandi and the strongly supported Stryphnusp+D. bucklandi clade needs to be confirmed before resurrecting the Sanidasterinae Sollas, 1888, characterized by the possession of sanidasters. Furthermore, Stoeba (not sampled here) having been synonymized with Dercitus , we can be confident that Stoeba species should also be reallocated to the Ancorinidaep.
The polyphyletic Alectonidae
The Alectonidae Rosell, 1996 (Hadromerida) are excavating sponges recently separated from the rest of the Clionaidae d'Orbigny, 1851 notably due to the possession of amphiasters or microrhabds, and absence of tylostyles. Alectona are known to produce a unique type of larva in the Porifera: an armored planktonic larva ( = hoplitomella larva) with discotriaenes , . These are then lost by the adult, which settles and bores into biogenic substrata such as calcareous rocks or coral. The association of triaenes and amphiasters suggest that Alectona should be placed near or within the Tetractinellidap [Borchiellini et al., 2004] , . A 28S (D1-C2) phylogenetic study then showed that the Alectonidae sensu Rützler  is polyphyletic and that Alectona millari belonged to the Tetractinellidap . Our data not only confirmed this but also suggested that the Alectonidae genera Alectona and Neamphius belonged to the Astrophoridap. In the 28S+COI analyses, A. millari branched after Thrombus abyssi, an acknowledged Astrophoridap. In the 28S analyses, Alectona appeared within the Spirophorida outgroups branching between Cinachyrella and Craniella (Fig. S3), but the node between A. millari and Craniella sp. is not supported, and the branch is short. This result may be due to the fact that the Alectona 28S sequence is significantly shorter (409 bp.: D1-C2 domains) than the others sequences from this study. The ambiguous position of Alectona certainly deserves further investigation as it may represent a pivotal evolutionary step between Astrophorida and Spirophorida.
Having amphiasters but no triaenes, Neamphius huxleyi (the single species of its genus) has also been suspected to be an Astrophorida by morphologists . According to our results it may be close to Characella and the lithistids. This is further supported by biochemical data showing that N. huxleyi and Astrophorida lithistids (Callipelta sp., Theonella mirabilis and Theonella swinhoei) share cyclic peptides and depsipeptides with cytotoxic and antiviral effects, notably with HIV-inhibitory activity , . However, the position of N. huxleyi being equivocal and poorly supported, we propose to temporarily consider it as incertae sedis.
Our results also have consequences for the rest of the Alectonidae genera. Following Borchiellini et al. , we advocate the reallocation of Thoosa along with Alectona. Delectona might also join them since it shares amphiasters and toxas with Thoosa. These three genera (representing ca 29 species) would group in the Thoosidae Rosell and Uriz, 1997, here resurrected. The position of the rest of the Alectonidae (Spiroxya, Dotona and Scolopes) is at the moment uncertain although Spiroxya and Dotona are suspected to be phylogenetically close to each other . On the Sponge Gene Tree Server (www.spongegenetrees.org , accessed on the 15th of October 2010), a phylogenetic 28S (B9-B21) tree of the Demospongiae suggested that Spiroxya levispira should remain close to the Placospongidae and the Trachycladidae (Hadromerida).
Since Lévi , the puzzling Thrombidae have been linked to the Astrophorida, based on their unique amphiasters and trichotriaenes. With the discovery of Yucatania sphaerocladoides, it appeared clear that Thrombus species had secondarily lost their triaenes , which confirmed that they belonged to the Tetractinellidap. Our study showed that Thrombus abyssi is alone, at the base of the Astrophoridap tree which suggests, as for Alectona, the key role of this group in understanding how and when the Astrophoridap originated.
The Pachastrellidae and the lithistids
The latest revision of the Pachastrellidae includes 12 genera  which share streptasters (rays proceeding from an axis that can be straight or spiral, Fig. 1e) and do not have euasters (rays radiating from a central point, Fig. 1e). Topsent  suggested that the Pachastrellidae could be subdivided between those that share a diverse set of streptasters (Thenea, Vulcanella, Poecillastra, some Corallistidae) and those whose streptasters are mainly restricted to amphiasters (rays radiating from both ends of a straight shaft, Fig. 1e) (Pachastrella, Characella, most Astrophorida lithistids). However, in our study, none of these groups were monophyletic (Fig. 2). We sampled six Pachastrellidae genera and they were distributed in five different clades: clade 1) Dercitus was reallocated to the Ancorinidaep (cf. above); clade 2) Characella appeared at the base of the Ancorinidaep along with lithistids and Neamphius; clade 3) Poecillastra amygdaloides+Pachastrella+Triptolemma was the sister clade of the Geodiidaep. Although the positions of Characella and clade 3 were poorly supported and unstable depending on the dataset (Fig. S2, S3), they were clearly separated from the other Pachastrellidae genera branching further down in the tree: clade 4) Poecillastra+Vulcanella(Vucanella) and clade 5) Thenea+Vulcanella(Annulastrella). Clearly the Pachastrellidae were built on a plesiomorphy (the streptasters) and the family must be revised.
Characella is defined by amphiasters and at least two categories of monaxonic spicules (microxeas, microstyles, microstrongyloxeas) while Poecillastra is defined by a diverse set of streptasters (spirasters, metasters and plesiasters) and microxeas in a single category . As Characella has been occasionally difficult to characterize with respect to Poecillastra, morphologists have questioned their validity , . Their definitions may overlap and many species are found to be “intermediate”, with characters of both genera (e.g. Poecillastra saxicola). According to our results, Characella was clearly separated from Poecillastra and phylogenetically closer to amphiaster-bearing lithistids. The definitions of Characella and Poecillastra should therefore prioritize the nature of streptasters and consider the number of categories of microxeas as a less reliable character, since these can be more ambiguous to characterize (cf. new definitions in File S2). Due to a lack of robustness, we propose to have Characella as incertae sedis at the moment, although we suspect that it could be allocated to a lithistid family in the future.
According to the ICZN and our results, the Pachastrellidae name should be kept for the Pachastrella+Triptolemma clade, henceforth named Pachastrellap. Until further molecular data, we propose to include Poecillastra amygdaloides in this newly defined Pachastrellidae (File S2), although its position was poorly supported. P. amygdaloides has calthrops: this species and its synonym Poecillastra debilis had therefore originally been described as Pachastrella . But P. amygdaloides was moved to Poecillastra because of its atypical triactinal calthrops, with a reduced fourth actine, later considered to be a modified triaene , . Its sister-group position with Pachastrellap is supported by its spicule characters which seem intermediate between the Poecillastra+Vulcanella(Vulcanella) clade and Pachastrellap: i) plesiasters (most of them are amphiaster-like) and ii) no microstrongyles. Other species (not sampled here) share the triactinal calthrops with P. amygdaloides: Poecillastra nana, Poecillastra connectens and Characella capitolii. We propose to resurrect Nethea Sollas, 1888 (originally defined as resembling Poecillastra but with triaenes with an underdeveloped rhabdome) to welcome these species. Triptolemma are cryptic excavating species penetrating the tissue of other sponges or coral. Many morphological characters support the Pachastrellap clade claimed by Topsent . Triptolemma are characterized by short-shafted mesotriaenes of all sizes, which can be also produced by some Pachastrella species (e.g. P. ovisternata). Microscleres of Triptolemma are streptasters (from only amphiasters to a diverse set), microstrongyles and even microrhabdose streptasters . These last two microscleres are apomorphies shared with Pachastrella. Brachiaster (not sampled here) surely belongs to this clade since it also produces short-shafted mesotriaenes, microstrongyles and amphiasters .
Thenea, Vulcanella, and Poecillastra share a diverse set of streptasters . Poecillastra+Vulcanella(Vulcanella) further share i) an oscule area surrounded by cloacal oxeas (in Poecillastra compressa this area has expanded over a whole side of the sponge but the cloacal oxeas are still there), ii) an abundance of spiny microxeas, iii) a reduction of the triaenes to short-shafted triaenes or calthrops (even if long-shafted triaene species also exist) and iv) an absence of pro/anatriaenes (except in Poecillastra rudiastra). In order to welcome this very well supported clade named Vulcanellidaep, we created the Vulcanellidae fam. nov. (File S2). On the other hand, the Thenea+Vulcanella(Annulastrella) clade was poorly supported (bootstrap<50). And yet, these two genera share i) large plesiasters and ii) absence of microxeas. For the time being, the Theneidae Carter, 1883 is resurrected to welcome these two genera. Also, Vulcanella(Annulastrella) needs to be upgraded to genus since it was clearly separated from Vulcanella(Vulcanella). The Thenea clade, here named Theneap, is very well supported (boostrap of 93) and also one of the few clades supported by the COI amino acid analyses (tree not shown). It groups species that share i) a characteristic external morphology (massive, hispid mushroom shape, Fig. 1d), with ii) a typical poral area, iii) long-shafted dichotriaenes (never calthrops), iv) an abundance of pro/anatriaenes and v) a system of roots to grow on muddy bottoms. Based on morphology, Cladothenea (not sampled here) should belong to this clade . The Theneidae and the Vulcanellidae fam. nov. may i) form a poorly-supported clade (28S analyses, Fig. S3), ii) have a paraphyletic relationship (28S+COI tree, Fig. 2) or iii) be further apart (COI analyses, Fig. S2). All of these poorly supported topologies emphasize that relationships between these two families remain to be investigated.
As previously suggested by morphological , , , , ,  and molecular data , , , our phylogeny confirmed that some lithistids belong to the Astrophoridap. The Discodermia+Theonella clade named Theonellidaep was strongly supported (bootstrap of 100). According to morphology and a previous 18S phylogenetic study, Racodiscula may also be part of the Theonellidaep . We note that Discodermia has microxeas and microrhabds while Characella (phylogenetically close to Discodermia in our tree) has two sizes of microxeas. The microrhabds of the Theonellidaep might therefore be homologous to the small microxeas of Characella. We also notice that the microrhabds of Discodermia are similar to the ones found in Pachastrella (e.g. Discodermia proliferans): these might also be homologous. Exsuperantia sp. (Phymaraphiniidae) is morphologically very close to the Theonellidaep, but it has trider desmas instead of tetraclone desmas. Exsuperantia sp. either groups with Characella (28S+COI and 28S dataset), or with N. huxleyi (COI analyses). In both cases, the support was low. Morphological  and molecular  data suggest that the Corallistidae is a sister-group to the Theonellidae. Because of the low supported nodes between our lithistids this cannot be excluded: the position of Neophrissospongia nolitangere (Corallistidae) is unsure but certainly close to the other lithistids. Our results also hint that desmas have appeared independently in different Astrophorida lithistid groups (at least four times, if we would consider Brachiaster, not sampled here) (Fig. 3). This would not come as a surprise since desmas have appeared independently in other sponge orders as well . It should be emphasized that, in our opinion, 8 out of the 13 extant lithistid families are of Astrophorida affinities (Corallistidae, Isoraphiniidae, Macandrewiidae, Neopeltidae, Phymaraphiniidae, Phymatellidae, Pleromidae, Theonellidae) representing ca 128 species . A majority of them possess amphiaster streptasters while the remaining groups have additional spirasters (Corallistidae, Pleroma) or no asters (Macandrewiidae, Discodermia, Theonella). Therefore, although Astrophoridap lithistids do not seem to form a natural group, we can be certain that they all radiated along with amphiaster-bearing Astrophoridap (Characella, Pachastrella, Triptolemma, Brachiaster, and Neamphius). If they have a closest common ancestor with the Ancorinidaep, the Geodiidaep, or both, is still unclear at this point.
The node following that of the Vulcanellidae may be of importance since it supports, albeit moderately, a clade comprising amphiaster- and euaster-bearing Astrophoridap (Fig. 4), temporarily called ‘clade A’ (Fig. 2). Our study thus reveals for the first time the importance of amphiasters in Astrophoridap aster evolution, as an intermediate step between spirasters and euasters. The shortening of the amphiaster central shaft may represent an essential and preliminary stage to the appearance of euasters. Clade A includes all the Astrophoridap except for the Vulcanellidae, the Theneidae, Alectona and Thrombus, but since the position of the Vulcanellidae is unstable, so is the content of clade A. We thus refrain from formally naming clade A and wait for confirmation from other molecular markers. Lamellomorpha strongylata Bergquist, 1968 incertae sedis (not sampled) lacks triaenes and possesses only two types of microscleres: spiny microstrongyles and amphiaster-like streptasters. This species could therefore belong to the amphiaster/euaster-bearing clade, and may be phylogenetically close to Characella or to Pachastrellap, both of which have small ectosomal monoaxial spicules.
Evolution of Megascleres in the Astrophoridap (Fig. 3)
Astrophoridap species are well characterized by the simultaneous presence of asters (microscleres) and triaenes (megascleres) (Fig. 1e–g). Therefore, the classification of this order has essentially been based on variants of these two spicule types. The triaene is a synapomorphy of the Tetractinellidap so it appeared in the common ancestor of Spirophorida and Astrophoridap. Since then, it has evolved in different directions giving rise to numerous descriptive terms with respect to the cladome orientation (ortho/plagio/pro/meso/anatriaenes), cladome branching (phyllo/disco/dichotriaene) or the rhabdome length (long-shafted/short-shafted/pseudocalthrops/calthrops). According to our data, the presence of triaenes or anatriaenes is not likely in the common ancestor of Astrophoridap (Fig. 3). This is probably due to the presence of Alectona and Thrombus at the base of the tree, both without triaenes. Long-shafted triaenes possibly appear (p = 0.68) in the ancestor of the Theneidae and the rest of the Astrophoridap. Since then, they have evolved into short-shafted triaenes or calthrops. Calthrops have appeared independently many times (Calthropellap, Pachastrellap, Dercitus, some Vulcanella), and so have mesocalthrops and mesodichotriaenes (Calthropellap, some Pachastrellap). Concerning anatriaenes, our analyses (Fig. 3) suggest that they have appeared independently many times (in Theneap, Characella, some Stelletta, Geodinaep). Discotriaenes have appeared independently in some lithistid Astrophoridap (e.g. Discodermia) and in the larvae of Alectona, although we cannot rule out the possibility that they are present in other Astrophoridap larvae (never observed to date). Phyllotriaenes are only known in some lithistid families, but may have appeared independently at least twice (Phymaraphiniidae and Theonella). To conclude, most variants of triaenes are clearly the product of convergent evolution and thus homoplasic characters that cannot be used for Astrophoridap classification. On the other hand, they may still represent apomorphies at lower ranks.
Before going further, we should clarify the term ‘secondary loss’. An ‘absence’ state can be optimized as a plesiomorphy (true absence), a homoplasy (independent secondary losses which appeared through convergent evolution) or a synapomorphy (unique secondary loss shared by a single clade) . In this last case, ‘absence’ states may also potentially bring phylogenetic information. Furthermore, a spicule secondary loss can be i) a ‘true’ loss when nothing replaces the spicule lost (e.g. loss of sterrasters) or ii) a ‘semantic’ loss by modification of a spicule into another (e.g. sterrasters becoming aspidasters). It may not always be possible to discriminate a ‘true’ loss from a ‘semantic’ loss. For example, secondary loss of triaenes is ambiguous because some species may have retained megascleres derived from triaenes, such as styles while others may have really lost their triaenes. We therefore considered that when styles were present, it was a semantic loss, because when only oxeas remained it had a higher chance of being a true loss of triaenes.
Our study shows that triaenes have been secondarily lost (with p>0.65) independently at least four times in our sampling (e.g. Melophlus, Asteropus, Vulcanella (Annulastrella), Neamphius) and morphology suggests that it may have happened in even more Astrophorida taxa, not all sampled here (Thrombus, Lamellomorpha, Holoxea, Jaspis, some Stelletta, some Rhabdastrella, some Erylus, some Geodia) , , , . We observe similar results for anatriaenes which may have been lost eight times independently. It is also worth mentioning that anatriaenes do not seem to have been lost in the Erylinaep as suggested before . According to our results (Fig. 3), the common ancestor of the Geodiidaep did not have anatriaenes, they only seem to appear in the Geodinaep. Their absence should therefore not be considered as a synapomorphy of the Erylinaep  but as a plesiomorphy.
Our results clearly demonstrate how common secondary loss of a megasclere is, even when this megasclere has a clear function: providing support of the cortex, organization of the choanosome or even defending against predators. Secondary loss of triaene is a homoplasic character for the Astrophoridap, but it may become synapomorphic in more restricted clades (e.g. Vulcanella (Annulastrella), Melophlus). Also, we remind that loss of triaenes can be “partial” if it takes place during the development (e.g. Alectona) so increasing our knowledge in Astrophoridap larvae may shed some light on the classification and the evolution of triaenes.
Evolution of Microscleres in the Astrophoridap (Fig. 4)
Thrombidae species have a unique type of amphiaster with recurved spines at each end, not found anywhere else in the Astrophoridap. It has been secondarily lost in some species of Thrombus. It is unclear if their amphiasters are homologous to the more typical amphiasters observed in A. millari. Thrombidae also have trichotriaenes, not found anywhere else in the Astrophoridap. Since trichotriaenes are fairly small (compared to true triaenes) and coexist with true triaenes in Yucatania; they may be derived from a large microsclere, and are certainly not triaenes per se. Seemingly, in Theneap and Vulcanella (Annulastrella) large plesiasters have occasionally been considered as megascleres . Trichotriaenes could therefore have originated from a form of plesiaster. The characteristic large diactines in Alectona are also thought to be derived from large asters . Supporting this hypothesis are the large triactines found in some Alectona and the oxyasters found in Thoosa. However, according to the position of A. millari in our tree, and if we are right about the reallocation of Thoosa with Alectona, these oxyasters are not homologous to the ones that appeared later in the Ancorinidaep and the Geodiidaep. As for the fusiform amphiasters found in Alectona, their origin remains unknown. Meanwhile, the diversified streptaster set (spirasters, metasters, plesiasters) that developed in the Theneidae and Vulnellidaep may have been reduced to amphiasters in the ancestor of Clade A. On one side, the Ancorinidaep share a close common ancestor with the lithistids/Characella/Neamphius. On the other side, the Geodiidaep share a close common ancestor with the newly defined Pachastrellidae. In both cases, we can hypothesize that a shortening and disappearance of the shaft and/or compression of amphiaster, spirasters or even sanidasters could have easily led to the appearance of euasters. Indeed, such ‘intermediate’ forms of asters can be observed in Characella, Pachastrella , Dercitus  or Neophrissospongia . Two independent appearances of euasters in the Astrophoridap are not surprising in comparison with their independent appearance in Thoosa, some Hadromerida and in Chondrilla (Chondrosida). The reversed evolution is also known: amphiasters are derived from euasters in the case of Erylus amphiastera from Colombia (not sampled). According to our data, sterrasters have appeared once (p>0.65) in the ancestor of the Geodiidaep. Evolution of spherules seem to be possible from microrhabds (as in Caminus ) or from asters (as in some Calthropella ). The sanidasters may have evolved from amphiasters and/or microrhabds but our spicule reconstructions do not support this at the moment (Fig. 4). We have nonetheless observed sanidaster-like amphiasters (in Pachastrella abyssi) and sanidaster-like microrhabds (in some Pachymatismap normani). We must stress that the intermediate nodes leading to the Ancorinidaep and the Geodiidaep are poorly supported so these hypotheses need to be tested with additional molecular markers. The origin of microrhabds is seemingly contentious. The limit between microxeas, sanidasters and microrhabds is ambiguous and probably reflects their multiple appearances. They have independently appeared in (some) Ecionemia, Pachastrellap, the Erylinaep, some lithistids and Characella (if we consider that small microxeas present in the cortex are microrhabds). In some cases, such as in the Erylinaep, they might be derived from asters . The appearance of microxeas in the ancestor of the Vulcanellidaep might also be linked to asters. In the Theneidae, plesiasters reduced to two actines are common: they look like microxeas and are usually larger than the rest of the plesiasters. This is well documented in Vulcanella (Annulastrella) ,  and Theneap , , , , so we suggest that the microxeas found in the Vulcanellidaep (and maybe later in the lithistids, Pachastrellap and Characella) may have originated from large plesiasters reduced to two actines.
Sterrasters have been secondarily lost at least nine times independently (p>0.95) (Fig. 4): in Penaresp, Erylusp sp., Erylusp candidata, Melophlus sp., Geostellettap, Calthropellap, E. megastylifera, R. globostelleta+Rhabdastrella sp. and R. intermedia. This clearly demonstrates how common secondary loss of a microsclere is, even when it has a clear function (sterrasters form a strong barrier protecting the sponge). Interestingly, most of the secondary losses of sterrasters have occurred in shallow-water species, living in tropical or temperate — never boreal or arctic — waters (Fig. 4). Actually, our results suggest that secondary loss of megascleres and microscleres are more common in shallow-water species. It is therefore tempting to propose that secondary loss of spicules has been favored in tropical to temperate shallow-waters. This further suggests that environmental parameters such as lower pressure, higher water temperature and/or lower silica concentration could be responsible for the loss of these sterrasters. Such parameters are already known for their effect on spicule morphology , , , , especially silica concentration that appears to have played an important role in sponge evolution , . But since there is insufficient evidence for our hypotheses, we refrain from further speculation along these lines.
This study is the first comprehensive molecular phylogenetic study of the Astrophorida. We obtained a well-resolved tree that suggested phylogenetic relationships between 89 species of Astrophorida from nine families of sponges. Most incongruences found between the current classification (Systema Porifera) and our molecular tree systematically made sense in the light of morphology (e.g. reallocated Ancorinidae, G. intuta, D. bucklandi, C. pachastrelloides), scattered biochemical data and homoplasic processes (convergent evolution and secondary loss). The taxonomic translation of this tree was a revision of the Astrophorida for which we proposed new classifications: the Linnaean classification includes all extant taxa belonging to the Astrophorida (File S2) while the phylogenetic classification includes at the moment only clades supported by molecular data and morphological data (File S1, Fig. 5). We propose in File S3 a key to all the Astrophorida families, sub-families and genera incertae sedis. And Table S4 summarizes the nomenclatural changes resulting from our study with respect to the name of Astrophorida species. With addition of the eight families of lithistids as well as the Thoosidae and Neamphius huxleyi, the Astrophorida became a larger order than previously considered, comprising ca 820 species . However, the phylogenetic position of a few Astrophorida genera not sampled here is still pending (File S2). The polyphyly of some genera (Ecionemia, Rhabdastrella, Erylus, Stelletta) suggest that they should be tested on a species to species basis. Finally, other contentious groups need to be tested as potential members of the Astrophoridap: some may have been confused with aster-bearing Hadromerida (e.g. Jaspis vs. Hemiasterella) while others may have lost all their asters and triaenes and are mixed in polyphyletic orders such as the Halichondrida or Haplosclerida.
Our study is far from being the first study to show the potential misleading nature of spicules and to question their utility in sponge taxonomy , , , , especially with the numerous studies on the phenotypical plasticity of spicules (e.g. ) and the recent outburst of cryptic species identification , , . But this is certainly the first study to show how widespread convergent evolution and secondary loss can be in spicule evolution: they have taken place many times, in all taxa, in megascleres and microscleres, even when these seem to be adaptative and under selective pressures. Our results show for the first time the banality of spicule secondary loss (especially for microscleres) and its potential as a synapomorphy (e.g. in Geostellettap). With a sponge classification depending so much on spicules, secondary loss of spicules should from now on be taken more into account in future research on sponge taxonomy and phylogeny.
Molecular phylogeny of the Astrophorida obtained with maximum likelihood analyses (metREV+G model) of the COI amino-acid dataset. Bootstrap values >50 are given at the nodes (2,000 ML replicates).
Molecular phylogeny of the Astrophorida obtained with maximum likelihood analyses (HKY+I+G model) of the COI nucleotide dataset. Bootstrap values >50 are given at the nodes (2,000 ML replicates).
Molecular phylogeny of the Astrophorida obtained with maximum likelihood analyses (GTR+I+G model) of the 28S (C1-D2) dataset. Bootstrap values >50 are given at the nodes (2,000 ML replicates).
Locality of collection, museum voucher numbers and Genbank accession numbers for the sponge specimens used in this study.
Sponge identification modifications after re-examination of Astrophorida specimens from previous molecular phylogenetic and biochemistry studies.
Morphological matrix of the Astrophorida species sampled in this study.
Nomenclatural changes in the Linnaean and phylogenetic classification as a result of our study.
Definition of new clades defined in this study (following the rules of the PhyloCode v.4c).
Proposal for a new Linnaean classification of the Astrophorida.
The authors wish to thank the shipboard parties and the crews of the R/V Hans Brattström (Marine Biological Station of the University of Bergen), the R/V Polarstern (ARK-XXII/1a), the R/V G. M. Dannevig (Marine Biological Station of Flødevigen), the R/V Pourquoi Pas? (IFREMER), the R/V Kommandor Jack (EMEPC 2007) and the R/V Almirante Gago Coutinho (EMEPC 2008). The authors wish to thank people that contributed to the sampling of this study: Bernard Picton and Claire Goodwin (National Museums, Northern Ireland), Isabelle Domart-Coulon (Muséum National d'Histoire Naturelle, Paris), Mary Kay Harper (University of Utah, Salt Lake City) in collaboration with the NCDDG program and with G. Concepcion, Rob van Soest and Elly Beglinger (Zoölogisch Museum van de Universiteit van Amsterdam), Joseph R. Pawlik and Tse–Lynn Loh (University of North Carolina, Wilmington), Scott Nichols (University of California, USA), Marcel Jaspars (University of Aberdeen), IRD of Nouméa (New Caledonia), Shirley Sorokin (SARDI Aquatic Sciences, Adelaide) and Thierry Laperousaz (South Australian Museum, Adelaide), Thierry Perez (Centre d'Océanologie de Marseille, France), Kyle Sweeney and Grace McCormack (National University of Ireland, Galway, Ireland), Magnus Tornes (Norway), Chris Proctor (Devon), Marina Cunha (University of Aveira, Portugal) and Jeroen Ingels (University of Ghent). Geodia corticostylifera was collected by M. Klautau (permission number in IBAMA: 1822255). We thank Nicole Boury-Esnault for valuable comments on an earlier draft of this paper.
Conceived and designed the experiments: PC HTR. Performed the experiments: PC. Analyzed the data: PC. Contributed reagents/materials/analysis tools: HTR CS JX JR. Wrote the paper: PC. Collection of specimens: PC HTR CS JX JR. Identification of specimens: PC HTR JX. Phylogenetic analyses: PC. Contributed to the paper: HTR CS JX JR.
- 1. Borchiellini C, Chombard C, Manuel M, Alivon E, Vacelet J, et al. (2004) Molecular phylogeny of Demospongiae: implications for classification and scenarios of character evolution. Molecular Phylogenetics and Evolution 32: 823–837.
- 2. Sperling EA, Peterson KJ, Pisani D (2009) Phylogenetic-Signal Dissection of Nuclear Housekeeping Genes Supports the Paraphyly of Sponges and the Monophyly of Eumetazoa. Molecular Biology and Evolution 26: 2261–2274.
- 3. Erpenbeck D, Duran S, Rützler K, Paul V, Hooper JNA, et al. (2007) Towards a DNA taxonomy of Caribbean demosponges: a gene tree reconstructed from partial mitochondrial CO1 gene sequences supports previous rDNA phylogenies and provides a new perspective on the systematics of Demospongiae. Journal of the Marine Biological Association of the United Kingdom 87: 1563–1570.
- 4. Nichols SA (2005) An evaluation of support for order-level monophyly and interrelationships within the class Demospongiae using partial data from the large subunit rDNA and cytochrome oxidase subunit I. Molecular Phylogenetics and Evolution 34: 81–96.
- 5. Chombard C, Boury-Esnault N, Tillier S (1998) Reassessment of homology of morphological characters in Tetractinellid sponges based on molecular data. Systematic Biology 47: 351–366.
- 6. van Soest RWM, Boury-Esnault N, Hooper JNA, Rützler K, de Voogd NJ, et al. (2010) World Porifera database. Available: http://www.marinespecies.org/porifera.
- 7. Klitgaard AB, Tendal OS (2004) Distribution and species composition of mass occurrences of large-sized sponges in the northeast Atlantic. Progress in Oceanography 61: 57–98.
- 8. Maldonado M, Young CM (1996) Bathymetric patterns of sponge distribution on the Bahamian slope. Deep-Sea Research Part I-Oceanographic Research Papers 43: 897–915.
- 9. Hooper JNA, van Soest RWM (2002) Order Astrophorida Sollas, 1888. In: Hooper JNA, van Soest RWM, editors. Systema Porifera A Guide to the classification of Sponges. New York: Kluwer Academic/Plenum Publishers. pp. 105–107.
- 10. Lévi C (1991) Lithistid sponges from the Norfolk Rise. Recent and Mesozoic Genera. In: Reitner J, Keupp H, editors. Fossil and Recent Sponges. Berlin Heidelberg: Springer-Verlag. pp. 72–82.
- 11. Reid REH (1963) Preliminary notice of a classification of the Demospongia. The Irish Naturalists Journal 14: 90–94.
- 12. McInerney JO, Adams CL, Kelly M (1999) Phylogenetic resolution potential of 18S and 28S rRNA genes within the Lithistid Astrophorida. Memoirs of the Queensland Museum 44: 343–351.
- 13. Kelly-Borges M, Pomponi SA (1994) Phylogeny and classification of lithistid sponges (Porifera: Demospongiae): a preliminary assessment using ribosomal DNA sequence comparisons. Molecular Marine Biology and Biotechnology 3: 87–103.
- 14. Hooper JNA, van Soest RWM (2002) Systema Porifera. A guide to the classification of sponges;. In: Hooper JNA, van Soest RWM, editors. New York: Kluwer Academic/Plenum Publishers. 1708 p.
- 15. Sollas WJ (1888) Report on the Tetractinellida collected by H.M.S. Challenger, during the years 1873–1876. Report on the Scientific Results of the Voyage of H.M.S. Challenger, 1873–1876. Zoology 25: 1–458, pls I–XLIV, 451 map.
- 16. Borchiellini C, Alivon E, Vacelet J (2004) The systematic position of Alectona (Porifera, Demospongiae): a Tetractinellid sponge. Bollettino dei musei e degli istituti biologici dell' Universita di Genova 68: 209–217.
- 17. Topsent E (1920) Caractères et affinités des Thoosa Hanc. et des Alectona Cart. Considérations sur leurs germes à armure. Bulletin de la Société zoologique de France XLV: 88–97.
- 18. Vacelet J (1999) Planktonic armoured propagules of the excavating sponge Alectona (Porifera: Demospongiae) are larvae: evidence from Alectona wallichii and A. mesatlantica sp. nov. Memoirs of the Queensland Museum 44: 627–642.
- 19. Rützler K (2002) Family Alectonidae Rosell, 1996. In: Hooper JNA, van Soest RWM, editors. Systema Porifera A Guide to the classification of Sponges. New York: Kluwer Academic/Plenum Publishers. pp. 281–290.
- 20. 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.
- 21. Lavrov DV, Wang X, Kelly M (2008) Reconstructing ordinal relationships in the Demospongiae using mitochondrial genomic data. Molecular Phylogenetics and Evolution 49: 111–124.
- 22. Erpenbeck D, Breeuwer JAJ, Parra-Velandia 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.
- 23. Rossi AL, Russo CAM, Solé-Cava AM, Rapp HT, Klautau M (in press) Phylogenetic signal in the evolution of body colour and spicule skeleton in calcareous sponges. Zoological Journal of the Linnean Society.
- 24. Chombard C (1998) Les Demospongiae à asters: phylogénie moléculaire et homologie morphologique [Ph.D]. Paris: Muséum National d'Histoire Naturelle. 182 p.
- 25. Xavier J, van Soest R (2007) Demosponge fauna of Ormonde and Gettysburg Seamounts (Gorringe Bank, north-east Atlantic): diversity and zoogeographical affinities. Journal of the Marine Biological Association of the United Kingdom 87: 1643–1653.
- 26. Pleijel F, Jondelius U, Norlinder E, Nygren A, Oxelman B, et al. (2008) Phylogenies without roots? A plea for the use of vouchers in molecular phylogenetic studies. Molecular Phylogenetics and Evolution 48: 369–371.
- 27. Voigt O, Erpenbeck D, Worheide G (2008) Molecular evolution of rDNA in early diverging Metazoa: First comparative analysis and phylogenetic application of complete SSU rRNA secondary structures in Porifera. BMC Evolutionary Biology 8:
- 28. Wilcox T, Hill M, DeMeo K (2002) Observations on a new two-sponge symbiosis from the Florida Keys. Coral Reefs 21: 198–204.
- 29. Sorokin SJ, Lapeyrousaz TCD, Collings GJ (2008) Investigator group expedition 2006: sponges (Porifera). Transactions of the Royal Society of South Australia 132: 163–172.
- 30. Tabudravu JN, Jaspars M (2001) Stelliferin Riboside, a Triterpene Monosaccharide Isolated from the Fijian Sponge Geodia globostellifera. Journal of Natural Products 64: 813–815.
- 31. Tasdemir D, Mangalindan GC, Concepcion GP, Verbitski SM, Rabindran S, et al. (2002) Bioactive Isomalabaricane Triterpenes from the Marine Sponge Rhabdastrella globostellata. Journal of Natural Products 65: 210–214.
- 32. Pisera A, Vacelet J (2011) Lithistid sponges from submarine caves in the Mediterranean: taxonomy and affinities. Scientia Marina 75: 17–40.
- 33. Cárdenas P, Xavier J, Tendal OS, Schander C, Rapp HT (2007) Redescription and resurrection of Pachymatisma normani (Demospongiae, Geodiidae), with remarks on the genus Pachymatisma. Journal of the Marine Biological Association of the United Kingdom 87: 1511–1525.
- 34. Cárdenas P, Menegola C, Rapp HT, Díaz MC (2009) Morphological description and DNA barcodes of shallow-water Tetractinellida (Porifera: Demospongiae) from Bocas del Toro, Panama, with description of a new species. Zootaxa 2276: 1–39.
- 35. Cárdenas P, Rapp HT, Schander C, Tendal OS (2010) Molecular taxonomy and phylogeny of the Geodiidae (Porifera, Demospongiae, Astrophorida) — combining phylogenetic and Linnaean classification. Zoologica Scripta 39: 89–106.
- 36. da Silva CMM (2002) Revisão das espécies de Geodia Lamarck, 1815 (Porifera, Astrophorida, Geodiidae) do Atlântico Ocidental e Pacífico Oriental [Ph.D]. São Paulo.: Instituto de Biociências da Universidade de São Paulo. 305 p.
- 37. Topsent E (1892) Contribution à l'étude des Spongiaires de l'Atlantique Nord (Golfe de Gascogne, Terre-Neuve, Açores). Résultats des campagnes scientifiques accomplies par le Prince Albert I Monaco 2: 1–165, pls I–XI.
- 38. Pisera A, Lévi C (2002) Family Phymaraphiniidae Schrammen, 1910. In: Hooper JNA, van Soest RWM, editors. Systema Porifera A Guide to the classification of Sponges. New York: Kluwer Academic/Plenum Publishers. pp. 380–383.
- 39. Özdikmen H (2009) Substitute names for eight sponge genus group names (Porifera). Munis Entomolgy & Zoology 4: 212–218.
- 40. Topsent E (1904) Spongiaires des Açores. Résultats des campagnes scientifiques accomplies par le Prince Albert I Monaco 25: 1–280, pls 281–218.
- 41. Folmer O, Black M, Hoeh W, Lutz R, Vrijenhoek R (1994) DNA primers for amplification of mitochondrial cytochrome c oxidase subunit I from diverse metazoan invertebrates. Molecular Marine Biology and Biotechnology 3: 294–299.
- 42. Lê HLV, Lecointre G, Perasso R (1993) A 28S rRNA-based phylogeny of the Gnathostomes: first steps in the analysis of conflict and congruence with morphologically based cladograms. Molecular Phylogenetics and Evolution 2: 31–51.
- 43. Rambaut A (1996) Se-Al. Sequence Alignment Editor. v2.0a11. ed. Oxford, U.K.: University of Oxford.
- 44. Katoh K, Misawa K, Kuma K, Miyata T (2002) MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. Nucleic Acids Research 30: 3059–3066.
- 45. Galtier N, Gouy M, Gautier C (1996) SeaView and Phylo_win, two graphic tools for sequence alignment and molecular phylogeny. Computer Applications in the Biosciences 12: 543–548.
- 46. Posada D, Crandall KA (1998) Modeltest: testing the model of DNA substitution. Bioinformatics 14: 817–818.
- 47. Abascal F, Zardoya R, Posada D (2005) ProtTest: Selection of best-fit models of protein evolution. Bioinformatics 21: 2104–2105.
- 48. Zwickl DJ (2006) Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion. [Ph.D.]. Austin, TX: University of Texas. 115 p.
- 49. Cummings MP, Huskamp JC (2005) Grid computing. Educause Review 40: 116–117.
- 50. Bazinet AL, Cummings MP (2008) The Lattice Project: a grid research and production environment combining multiple grid computing models. In: Weber MHW, editor. Distributed & Grid Computing - Science Made Transparent for Everyone: Rechenkraft.net.
- 51. Myers DS, Bazinet AL, Cummings MP (2008) Expanding the reach of Grid computing: combining Globus- and BOINC-based systems. In: Talbi E-G, Zomaya A, editors. Grids for Bioinformatics and Computational Biology. New York: John Wiley & Sons. pp. 71–85.
- 52. Bazinet AL, Myers S, Fuetsch J, Cummings MP (2007) Grid services base library: a high-level, procedural application program interface for writing Globus-based grid services. Future Generation Computer Systems 23: 517–522.
- 53. Cummings MP, Handley SA, Myers DS, Reed DL, Rokas A, et al. (2003) Comparing bootstrap and posterior probability values in the four-taxon case. Systematic Biology 52: 477–487.
- 54. Myers DS, Cummings MP (2003) Necessity is the mother of invention: a simple grid computing system using commodity tools. Journal of Parallel and Distributed Computing 63: 578–589.
- 55. Swofford DL (2002) PAUP*. Phylogenetic Analysis Using Parsimony (* and other methods). 4.0. beta 10 ed. Sunderland, MA: Sinauer Associates.
- 56. Lewis PO (2001) A Likelihood Approach to Estimating Phylogeny from Discrete Morphological Character Data. Systematic Biology 50: 913–925.
- 57. Maddison WP, Maddison DR (2010) Mesquite: a modular system for evolutionary analysis. Version 2.73. Available at http://mesquiteproject.org.
- 58. Bakran-Petricioli T, Vacelet J, Zibrowius H, Petricioli D, Chevaldonné P, et al. (2007) New data on the distribution of the ‘deep-sea’ sponges Asbestopluma hypogea and Oopsacas minuta in the Mediterranean Sea. Marine Ecology 28: 10–23.
- 59. Pouliquen L (1972) Les spongiaires des grottes sous-marines de la région de Marseille: écologie et systématique. Tethys 3(4): 717–758.
- 60. Shimodaira H, Hasegawa M (2001) CONSEL: for assessing the confidence of phylogenetic tree selection. Bioinformatics 17: 1246–1247.
- 61. Wiedenmayer F (1989) Demospongiae (Porifera) from northern Bass Strait, southern Australia. Memoirs Museum Victoria 50: 1–242, pls 241–238.
- 62. Bergquist PR (1968) The Marine Fauna of New Zealand: Porifera, Demospongiae, Part 1. (Tetractinomorpha and Lithistida). New Zealand Department of Scientific and Industrial Research Bulletin [New Zealand Oceanographic Institute Memoir 37] 188: 1–105.
- 63. Dendy A (1924) Porifera. Part I. Non-Antarctic sponges. Natural History Report. British Antarctic (Terra Nova) Expedition, 1910 (Zoology) 6: 269–392, pls I–XV.
- 64. Dendy A (1916) Report on the Homosclerophora and Astrotetraxonida collected by H.M.S. ‘Sealark’ in the Indian Ocean. Reports of the Percy Sladen Trust Expedition to the Indian Ocean in 1905, volume 6. Transactions of the Linnean Society of London (2) 17: 225–271,pls 244–248.
- 65. Bergquist PR (1978) Sponges. Berkeley & Los Angeles & London: University of California Press & Hutchinson. 268 p.
- 66. Cheung FWK, Li CM, Che CT, Liu BPL, Wang LJ, et al. (2010) Geoditin A Induces Oxidative Stress and Apoptosis on Human Colon HT29 Cells. Marine Drugs 8: 80–90.
- 67. Clement JA, Li M, Hecht SM, Kingston DGI (2006) Bioactive Isomalabaricane Triterpenoids from Rhabdastrella globostellata that Stabilize the Binding of DNA Polymerase β to DNA. Journal of Natural Products 69: 373–376.
- 68. Boury-Esnault N, Rützler K (1997) Thesaurus of sponge morphology. Smithsonian Contributions to Zoology 596: 1–55.
- 69. Dai HF, Edrada RA, Ebel R, Nimtz M, Wray V, et al. (2005) Norlanostane triterpenoidal saponins from the marine sponge Melophlus sarassinorum. Journal of Natural Products 68: 1231–1237.
- 70. Topsent E (1928) Spongiaires de l'Atlantique et de la Méditerranée provenant des croisières du Prince Albert ler de Monaco. Résultats des campagnes scientifiques accomplies par le Prince Albert I Monaco 74: 1–376, pls I–XI.
- 71. Lévi C, Lévi P (1983) Éponges Tétractinellides et Lithistides bathyales de Nouvelle-Calédonie. Bulletin du Muséum national d'Histoire naturelle, 4e série, section A 5: 101–168.
- 72. Wilson HV (1925) Silicious and horny sponges collected by the U.S. Fisheries Steamer ‘Albatross’ during the Philippine Expedition, 1907–10. In: Contributions to the biology of the Philippine Archipelago and adjacent regions. Bulletin of the United States National Museum, 100 2: 273–532, pls 237–252.
- 73. van Soest RWM, Beglinger EJ, de Voogd NJ (2010) Skeletons in confusion: a review of astrophorid sponges with (dicho-)calthrops as structural megascleres (Porifera: Demospongiae: Astrophorida). Zookeys 68: 1–88.
- 74. von Lendenfeld R (1894) Die Tetractinelliden der Adria. (Mit einem Anhange über die Lithistiden). Denkschriften der Kaiserlichen Akademieder WissenschaftenWien Mathematisch-Naturwissenschaften Klasse 61: 91–204, pls I–VIII.
- 75. Topsent E (1895) Etude monographique des spongiaires de France II. Carnosa. Archives de Zoologie Expérimentale et Générale 3: 493–500, pls. XXI–XXIII.
- 76. van Soest RWM, Stentoft N (1988) Barbados Deep-Water Sponges. In: Hummelinck PW, Van der Steen LJ, editors. Studies on the Fauna of Curaçao and other Caribbean Islands. Amsterdam: Uitgaven van de Natuurwetenschappelijke Studiekring voor Suriname en de Nederlandse Antillen, No. 122. pp. 1–175.
- 77. Hajdu E, van Soest RWM (1992) A revision of Atlantic Asteropus Sollas, 1888 (Demospongiae), including a description of three new species, and with a review of the family Coppatiidae Topsent, 1898. Bijdragen tot de Dierkunde 62: 3–19.
- 78. Uriz MJ (2002) Family Ancorinidae Schmidt, 1870. In: Hooper JNA, van Soest RWM, editors. Systema Porifera A Guide to the classification of Sponges. New York: Kluwer Academic/Plenum Publishers. pp. 134–140.
- 79. Carvalho MS (2008) Revisão taxonômica de Astrophorida e Halichondrida do Chile (Demospongiae, Porifera), filogenia e biogeografia de Asteropus (Ancorinidae, Astrophorida) com uma discussão acerca das rotas trans-Pacíficas [Ph.D]. Rio de Janeiro: Museu Nacional, Universidade Federal do Rio de Janeiro. 282 p.
- 80. Topsent E (1902) Les Asterostreptidae. Bulletin de la Société Scientifique et Médicale de l'Ouest 11: 335–350.
- 81. Maldonado M (2002) Family Pachastrellidae Carter, 1875. In: Hooper JNA, van Soest RWM, editors. Systema Porifera A Guide to the classification of Sponges. New York: Kluwer Academic/Plenum Publishers. pp. 141–162.
- 82. Lévi C (1973) Systématique de la classe des Demospongiaria (Démosponges). In: Grassé PP, editor. Traité de Zoologie Spongiaires. Paris: Masson & Co. pp. 577–632.
- 83. von Lendenfeld R (1903) Porifera. Tetraxonia. In: Schulze FE, editor. Das Tierreich. Friedländer: Berlin. pp. vi–xv, 1–168.
- 84. Alander H (1942) Sponges from the Swedish West-Coast and adjacent waters. Göteborg.
- 85. Ratnayake AS, Bugni TS, Feng XD, Harper MK, Skalicky JJ, et al. (2006) Theopapuamide, a cyclic depsipeptide from a Papua New Guinea lithistid sponge Theonella swinhoei. Journal of Natural Products 69: 1582–1586.
- 86. Oku N, Gustafson KR, Cartner LK, Wilson JA, Shigematsu N, et al. (2004) Neamphamide A, a new HIV-Inhibitory depsipeptide from the Papua New Guinea marine sponge Neamphius huxleyi. Journal of Natural Products 67: 1407–1411.
- 87. Erpenbeck D, Voigt O, Gultas M, Wörheide G (2008) The sponge genetree server-providing a phylogenetic backbone for poriferan evolutionary studies. Zootaxa 58–60.
- 88. Hartman WD, Hubbard R (1999) A new species of Thrombus (Porifera : Demospongiae : Astrophorida) from Trinidad, West Indies. Bulletin of marine science 64: 1–8.
- 89. Pulitzer-Finali G (1986) A collection of West Indian Demospongiae (Porifera). In appendix, a list of the Demospongiae hitherto recorded from the West Indies. Annali del Museo civico di storia naturale Giacomo Doria 86: 65–216.
- 90. Carter HJ (1876) Descriptions and Figures of Deep-Sea Sponges and their Spicules, from the Atlantic Ocean, dredged up on board H.M.S.‘Porcupine’, chiefly in 1869 (concluded). Annals and Magazine of Natural History 4: (105): 226–240; (106): 307–324; (107): 388–410;(108): 458–479, pls XII–XVI.
- 91. van Soest RWM (2009) New sciophilous sponges from the Caribbean (Porifera: Demospongiae). Zootaxa 2107: 1–40.
- 92. Lévi C, Lévi P Forest J, editor. (1989) Spongiaires (MUSORSTOM 1 & 2). Résultats des Campagnes MUSORSTOM, Volume 4. Mémoires du Muséum national d'Histoire naturelle (A, Zoologie) 143: 1–260.
- 93. Reid REH (1970) Tetraxons and demosponge phylogeny. The Biology of the Porifera Symp zool Soc Lond Vol 25. London, New York: Academic Press. pp. 63–89.
- 94. Burton M (1929) Description of South African sponges collected in the South African Marine Survey. Part II. The ‘Lithistidae’, with critical survey of the desma-forming sponges. Fisheries Bulletin Fisheries and Marine Biological Survey Division, Union of South Africa Report 7 Special Report 2: 1–12.
- 95. de Laubenfels MW (1936) A Discussion of the Sponge Fauna of the Dry Tortugas in Particular and the West Indies in General, with Material for a Revision of the Families and Orders of the Porifera. Papers from Tortugas Laboratory 30: 1–225, pls 221–222.
- 96. Pisera A, Lévi C (2002) ‘Lithistid’ Demospongiae. In: Hooper JNA, van Soest RWM, editors. Systema Porifera A Guide to the classification of Sponges. New York: Kluwer Academic/Plenum Publishers. pp. 299–301.
- 97. Jenner RA (2004) When molecules and morphology clash: reconciling conflicting phylogenies of the Metazoa by considering secondary character loss. Evolution & Development 6: 372–378.
- 98. Adams CL, Hooper JNA (2001) A revision of Australian Erylus (Porifera : Demospongiae : Astrophorida : Geodiidae) with a tabular review of worldwide species. Invertebrate Systematics 15: 319–340.
- 99. Cárdenas P (2010) Phylogeny, Taxonomy and Evolution of the Astrophorida (Porifera, Demospongiae). Bergen: University of Bergen. 290 p.
- 100. Babiç K (1915) Zur Kenntnis der Theneen. Zoologische Jahrbücher Abteilung für Systematik, Geographie und Biologie der Tiere 40: 389–408, pl.316–318.
- 101. Koltun VM (1966) Four-rayed sponges of Northern and Far Eastern seas of the USSR (order Tetraxonida). Opredeliti Faunei SSSR 90 (Zoological Institute of the Academy of Sciences of the USSR: Moscow, Leningrad) 1–112, pls I–XXXVIII.
- 102. von Lendenfeld R (1907) Die Tetraxonia. Wissenschaftliche Ergebnisse der Deutschen Tiefsee-Expedition auf der Dampfer Valdivia 1898–1899 11: i–iv, 59–374, pls IX–XLVI.
- 103. Cruz T (2002) Esponjas marinas de Canarias; Canarias CdPTyMAdGd, editor. S/C Tenerife: Consejería de Política Territorial y Medio Ambiente del Gobierno de Canarias. 260 p.
- 104. Frøhlich H, Barthel D (1997) Silica uptake of the marine sponge Halichondria panicea in Kiel Bight. Marine Biology 128: 115–125.
- 105. Stone AR (1970) Seasonal variations of spicule size in Hymeniacidon perleve. Journal of the Marine Biological Association of the United Kingdom 50: 343–348.
- 106. Mercurio M, Corriero G, Scalera-Liaci L, Gaino E (2000) Silica content and spicule size variations in Pellina semitubulosa (Porifera: Demospongiae). Marine Biology 137: 87–92.
- 107. Bavestrello G, Bonito M, Sarà M (1993) Influence of depth on the size of sponge spicules. Scientia Marina 57: 415–420.
- 108. Maldonado M, Carmona MC, Uriz MJ, Cruzado A (1999) Decline in Mesozoic reef-building sponges explained by silicon limitation. Nature 401: 785–788.
- 109. Maldonado M (2009) Embryonic development of verongid demosponges supports the independent acquisition of spongin skeletons as an alternative to the siliceous skeleton of sponges. Biological Journal of the Linnean Society 97: 427–447.
- 110. Solé Cava AM, Boury-Esnault N, Vacelet J, Thorpe JP (1992) Biochemical genetic divergence and systematics in sponges of the genera Corticium and Oscarella (Demospongiae: Homoscleromorpha) in the Mediterranean Sea. Marine Biology 113: 299–304.
- 111. Klautau M, Solé-Cava AM, Borojevic R (1994) Biochemical systematics of sibling sympatric species of Clathrina (Porifera: Calcarea). Biochemical Systematics and Ecology 22: 367–375.
- 112. Schönberg CHL, Barthel D (1998) Unreliability of Demosponge skeletal characters: the exemple of Halichondria panicea. In: Watanabe Y, Fusetani N, editors. Sponge Sciences Multidisciplinary Perspectives. Tokyo: Springer. pp. 41–53.
- 113. Cavalcanti FF, Zilberberg C, Klautau M (2007) Seasonal variation of morphological characters of Chondrilla aff. nucula (Porifera: Demospongiae) from the south-east coast of Brazil. Journal of the Marine Biological Association of the UK 87: 1727–1732.
- 114. Xavier JR, Rachello-Dolmen PG, Parra-Velandia F, Schönberg CHL, Breeuwer JAJ, et al. (2010) Cryptic diversity in a “cosmopolitan” excavating sponge: the Cliona aff. celata complex (Porifera, Clionaidae). Molecular Phylogenetics and Evolution 56: 13–20.
- 115. Klautau M, Russo CAM, Lazoski C, Boury-Esnault N, Thorpe JP, et al. (1999) Does cosmopolitanism result from overconservative systematics? A case study using the marine sponge Chondrilla nucula. Evolution 53: 1414–1422.
- 116. Rapp HT (2006) Calcareous sponges of the genera Clathrina and Guancha (Calcinea, Calcarea, Porifera) of Norway (north-east Atlantic) with the description of five new species. Zoological Journal of the Linnean Society 147: 331–365.