Deceptive Desmas: Molecular Phylogenetics Suggests a New Classification and Uncovers Convergent Evolution of Lithistid Demosponges

Reconciling the fossil record with molecular phylogenies to enhance the understanding of animal evolution is a challenging task, especially for taxa with a mostly poor fossil record, such as sponges (Porifera). ‘Lithistida’, a polyphyletic group of recent and fossil sponges, are an exception as they provide the richest fossil record among demosponges. Lithistids, currently encompassing 13 families, 41 genera and >300 recent species, are defined by the common possession of peculiar siliceous spicules (desmas) that characteristically form rigid articulated skeletons. Their phylogenetic relationships are to a large extent unresolved and there has been no (taxonomically) comprehensive analysis to formally reallocate lithistid taxa to their closest relatives. This study, based on the most comprehensive molecular and morphological investigation of ‘lithistid’ demosponges to date, corroborates some previous weakly-supported hypotheses, and provides novel insights into the evolutionary relationships of the previous ‘order Lithistida’. Based on molecular data (partial mtDNA CO1 and 28S rDNA sequences), we show that 8 out of 13 ‘Lithistida’ families belong to the order Astrophorida, whereas Scleritodermidae and Siphonidiidae form a separate monophyletic clade within Tetractinellida. Most lithistid astrophorids are dispersed between different clades of the Astrophorida and we propose to formally reallocate them, respectively. Corallistidae, Theonellidae and Phymatellidae are monophyletic, whereas the families Pleromidae and Scleritodermidae are polyphyletic. Family Desmanthidae is polyphyletic and groups within Halichondriidae – we formally propose a reallocation. The sister group relationship of the family Vetulinidae to Spongillida is confirmed and we propose here for the first time to include Vetulina into a new Order Sphaerocladina. Megascleres and microscleres possibly evolved and/or were lost several times independently in different ‘lithistid’ taxa, and microscleres might at least be four times more likely lost than megascleres. Desma spicules occasionally may have undergone secondary losses too. Our study provides a framework for further detailed investigations of this important demosponge group.


Introduction
Background Demospongiae Sollas, 1885 [1], with more than 85% of all living species, represents the largest and morphologically most diverse group of the phylum Porifera [2]. Today demosponges encompass 15 orders, and more than 8,500 accepted extant species [3]. Recent molecular evidence was pivotal in the classification of Demospongiae into four major clades: Keratosa Grant, 1861 [4], Verongimorpha Erpenbeck et al., 2012 [5], Haploscleromorpha Cárdenas et al., 2012 [6] and Heteroscleromorpha Cárdenas et al., 2012 [6,7] -the latter representing the largest and evolutionary most important group within demosponges [6]. 'Lithistida ' Schmidt, 1870 [8], on the other hand, until now remained a highly problematic and likely polyphyletic group of living and fossil sponges, and indeed provides the richest fossil records of all Porifera. Lithistid sponges differ from other demosponges by the unique possession of choanosomal spicules called desmas. These have been defined as ''articulating choanosomal megascleres of various geometry and usually complex morphology, often secondarily modified and very irregular'' [9]. Most living and fossil desma-bearing demosponges have a solid, rigid, heavily silicified skeleton -an important feature used in the morphological-based classification [9] -but a much fewer number of species have sparse, disarticulated desmas scattered throughout the mesohyl of their otherwise compressible choanosomal skeleton [10,11].
Compared to the lithistid fossil record (34 families, .300 genera [12]), the diversity of Recent species is comparatively poor (13 families and 41 genera, including five poorly known and of uncertain status) [9]. However, the Recent diversity of lithistids might extend back to the late Mesozoic in Europe (AP, unpublished results), suggesting that Recent lithistids are severely understudied [13]. 'Lithistida' inhabit tropical, subtropical and temperate regions from shallow waters to the deep sea, where they usually form faunal assemblages with other demosponges and, in the deep sea, also with hexactinellid sponges. Frequently, lithistid sponges occur on marine seamounts, their vertical slopes, on margins of continental shelves [14], and are common in submarine caves, e.g. in the Mediterranean [15,16] and shallow lava tubes in French Polynesia. [17]. Furthermore, some lithistids such as e.g. Theonella swinhoei, Discodermia polydiscus, Discodermia dissoluta, produce a wide range of bioactive compounds [18,19] and therefore are of special interest to the biomedical industry.
Historic taxonomic overview on lithistid demosponges Sollas (1888) [20] undertook the first comprehensive taxonomic study of lithistid sponges, based mainly on the presence or absence of ectosomal spicules and microscleres. He created two suborders Hoplophora and Anoplia (see Fig. 1), and considered that desmas occurred as a single evolutionary event. Lithistids were suggested to form a monophyletic group together with the Choristida (5Astrophorida), with the Anoplia considered to be the end lineage with the loss of all ectosomal spicules and microscleres. Dendy (1905) [21], Schrammen (1910) [22] and Wilson (1925) [23], however, suggested 'Lithistida' were polyphyletic and criticized Sollas' classification for excluding microscleres within the concept of his suborder Anoplia. Burton (1929) [24] was the first who attempted to reallocate many lithistid genera to their closest non-lithistid families based on alleged morphological characters of the Theneidae, Pachastrellidae, Stellettidae (Choristida, 5 now Astrophorida), Tetillidae (Spirophorida), ''Myxilleae'' (Poecilosclerida), Axinellidae (Halichondrida), Spirastrellidae and Polymastiidae (Hadromerida) sensu Burton (1929). This classification was refined by de Laubenfels (1936) [25], who established two new families: Kaliapsidae for the genera belonging to Choristida and Gastrophanellidae for genera showing affinities to the order Hadromerida. Although both classifications show conflicting results within some genera (e.g. in Microscleroderma), several hypotheses were similar regarding the reallocation of many lithistid taxa (see Fig. 1). Bergquist (1978) [26] subsequently argued that Burton's and de Laubenfels' hypotheses were based on weak assumptions and more material and detailed descriptions would be needed to unequivocally allocate those lithistid sponges to their closest relatives. Lévi (1973) [27] followed Burton's and de Laubenfels' assumptions and stated that all lithistid genera belonging to the families Theonellidae, Corallistidae and Pleromidae should be placed within the Choristida under the name Desmophorida. Nevertheless, he also emphasized the uncertainty of relationships between the remaining 'non-choristid' lithistids [27,28].
Despite their long acknowledged polyphyly, and all these attempts to reallocate lithistids to alleged sister-taxa in the past, lithistid Demospongiae were maintained in a single 'order' of demosponges within the most recent comprehensive taxonomic revision of Porifera, the Systema Porifera [9]. This was primarily due to the many still-unresolved or contested phylogenetic hypotheses throughout the families of 'Lithistida' and incomplete independent (e.g. molecular) evidence to support or refute particular hypotheses across the 'order'.

Morphological spicule arrangements of lithistids and spicule evolution within demosponges
Lithistid demosponges also present a high diversity of desma morphologies, megascleres, microscleres and skeletal structures. For example, Neoschrammeniella norfolki Schlacher-Hoenlinger, Pisera & Hooper, 2005 (Family Corallistidae) can have up to six different spicule types including megascleres and microscleres. Hence, this broad spicule diversity within lithistids and other astrophorids can be used as an appropriate tracer to study spicule evolution within demosponges. The importance of spicule homoplasy (convergent evolution and secondary losses) within demosponges is well known from several studies based on morphological and molecular characters in a variety of different sponge taxa: such as Crambe crambe [40,41], in the order Astrophorida [36,42] and many other Heteroscleromorpha [7]. It is also well known that secondary character losses in phylogenetic studies can have a fundamental influence in understanding conflicting molecular and morphological datasets [43]. However, due to low spicule diversity and few morphological characters in most non-lithistid demosponges, except for those belonging to Tetractinellida (Astrophorida + Spirophorida) [44], only little is known on how frequent secondary losses have occurred throughout Demospongiae. Recent molecular and morphological analyses of the Astrophorida [36] emphasized the repeated occurrence of  Sollas (1888) to the hypotheses of polyphyly of modern authors, it shows the attempts to reallocate most genera of the order 'Lithistida' to their closest relatives. secondary losses of both spicule types, megacleres and microscleres, and concluded that this evolutionary process is more common in demosponges than previously thought.

State of knowledge on the molecular phylogeny of lithistid sponges
The first molecular investigations focusing on lithistid sponges were based on a small fragment of the 18S rDNA gene (550 bp), comprising nine species representing seven different families [14]. However, the final outcome of this study was hampered due to the low variation within the selected gene region and the small taxa-set [38]. While there is a growing number of molecular phylogenies of Demospongiae using different molecular markers only few species of lithistids were included [7,36,38,42,45,46,47,48]. The broadest molecular dataset for lithistid sponges was assembled during the Porifera Tree of Life project, based on a nearly complete small-subunit ribosomal 18S rDNA gene, and it included 29 specimens from 12 different genera and six families [39]. Table 1 summarizes the current molecular data available in NCBI GenBank (www.ncbi.nlm.nih.gov/ genbank) for lithistid sponges (16 genera from 9 families), together with their suggested reallocation to their closest non-lithistid relatives. However, this sample size is still very small compared to the currently approx. 300 Recent described Integrative Taxonomy of Desma-Bearing 'Lithistid' Demosponges 'valid' species in the World Porifera Database [44], from 41 genera (plus five genera of uncertain status) and 13 families. In summary phylogenetic relationships of lithistid demosponges with non-lithistid species based on morphological  Integrative Taxonomy of Desma-Bearing 'Lithistid' Demosponges data remains mostly speculative and untested by more substantial independent molecular evidence.

Aims of this study
This study examines the molecular signatures of 68 lithistid specimens belonging to 12 of the 13 lithistid families, and 21 of the 46 known genera based on new material from different localities worldwide. The study aims to (1) establish a robust molecular phylogeny of lithistids based on independent mitochondrial protein coding (CO1, ''Folmer fragment'') and nuclear ribosomal (28S rDNA, partition C1-D2) markers; (2) formally propose the reallocation of all but one lithistid family to their closest relatives, and integrate both molecular and morphological evidence; (3) study the complexity of spicule evolution within lithistid and astrophorid sponges to assess the importance of homoplasy in megascleres and microscleres through a newly constructed morphological character data matrix.

Taxonomy and Sample datasets
Most of the newly sequenced material (

DNA extraction, amplification and sequencing
Genomic DNA was extracted using a modified [52] PALL-plate based extraction method [53] with an increased amount of tissue and twice the amount of lysis mix. In order to avoid any clogging of the membrane an additional centrifugation step was added just before transferring the lysate to the PALL-plates. For some specimens, where only little tissue was available, DNA was extracted using the NucleoSpinTissue Kit (Macherey-Nagel, Düren, Germany), following the standard protocol with an additional centrifugation step before pipetting the lysate to the Spin Column. To quantify the amount of isolated genomic DNA, a NanoDrop 1000 Spectrophotometer (Thermo Scientific) was used. The following two unlinked genes were amplified for this study: The standard DNA barcoding fragment (cytochrome oxidase subunit 1, partial; 659 bp) using the primers dgLCO1490 and dgHCO2198 [54] and following the protocol: 95uC, 3 minutes; (95uC, 30 seconds; 40-43uC, 20 seconds; 72uC, 1 minute) 634 cycles; 72uC, 5 minutes. The 28S rDNA (partition C1-D2, 768-832 bp) was studied using the forward C1'ASTR [55] and the reverse universal D2 primers [56], with the following PCR settings of 95uC, 3 minutes; (95uC, 30 seconds; 56-59uC, 45

Family Alectonidae
Integrative Taxonomy of Desma-Bearing 'Lithistid' Demosponges   Newly generated sequences as well as downloaded GenBank sequences of the CO1 and 28S gene were separately aligned with Muscle (v.3.6) [58] as incorporated in SeaView [59]. Alignments were subsequently controlled by eye. Saturation of both markers (CO1 and 28S) was evaluated using Xia's test [60] as implemented in DAMBE v5.1.5 [61]. This entrophy-based index estimates a substitution saturation index (Iss) and compares it to a critical substitution saturation index (Iss.c). As both datasets (CO1 and 28S) were too different from each other with respect to taxon sampling and sequencing success of the CO1 gene region to be merged, analyses were done separately for each gene region. As 'Lithistida' is a polyphyletic group, a wide range of sequences from GenBank from Heteroscleromorpha families were added to the CO1 dataset to yield a representative adequate taxon set from the latest classification of Demospongiae according to Morrow et al. (2012) [49]. For the CO1 dataset, sequences of the subclasses Verongimorpha and Keratosa were chosen as outgroups, as these subclasses have shown shorter branch lengths than Haplosclerida, the sister group of Heteroscleromorpha [62]. Axinella damicornis and Halichondria panicea were chosen as outgroups for our 28S rDNA (C1-D2 partition). The alignments used in this study, as well as the morphological data matrix (see below) in Nexus format, are freely available at OpenDataLMU (http://dx.doi.org/10.5282/ubm/ data.66).

Phylogenetic analyses
For Bayesian phylogenetic analyses we used the parallel version of MrBayes v.3.1.2 [63] on a Linux cluster under the most general GTR+G+I model, as possible overparameterization does not appear to have a negative effect on the results [64]. Analyses were run in two concurrent runs of four Metropolis-coupled Markovchains (MCMC) for 100,000,000 generations or stopped when the average standard deviation of split frequencies decreased below 0.01. The first 25% of the sampled trees were discarded for further analysis as burn-in. In both datasets, Maximum Likelihood (ML) bootstrap analyses (1,000 replicates) were also performed under the GTRGAMMAI nucleotide evolution model using raxmlGUI v.1.3 [65]. Tree topologies from Bayesian and ML analyses were compared and visualized using TreeGraph2 [66].

Morphological Analyses
In order to investigate spicule evolution of megascleres and microscleres within the Astrophorida/lithistids we used Mesquite v2.75 [67]. We designed a new   character data matrix for lithistid sponges from our own observed data, and amended it with carefully selected data from another study (Cárdenas et al. 2011), representing the smaller part of the whole matrix. In total the final matrix consists of 69 taxa and 35 characters coded as 1 for present or 0 for absent (see S1 Table and S1 File). For tracing characters over the imported molecular Bayesian tree and testing the homoplasy within lithistids and astrophorids the parsimony ancestral state reconstruction method was used under the unordered state assumption.

Comparison of both gene trees
Both molecular markers were not significantly saturated, as the Iss.c (0.801) was significantly higher than the observed Iss (0.286), therefore, both markers are suitable for conducting phylogenetic analyses with lithistid demosponges. The CO1 gene tree (Fig. 5) was used to resolve the classification of lithistids sequenced here with respect to other major demosponge groups. The resulting data matrix from the CO1 gene comprises 121 taxa. From 121 taxa, 31 are lithistids, from which 29 are represented by de novo-generated sequences and two (Theonella swinhoei and Exsuperantia sp.) obtained from GenBank. The 28S rDNA data matrix included 94 taxa, of which 48 are lithistids (43 de novo-generated sequences and five sequences from GenBank). An overview of sequencing success is given in Table 3. Bayesian inference and Maximum Likelihood topologies are congruent in both analyses.

Intra-family relationships of lithistid sponges in relation to other demosponges
Based on both CO1 (Fig. 5) and 28S rDNA (Fig. 6)

Family Corallistidae
The family Corallistidae is monophyletic in both CO1 (Fig. 5) and 28S rDNA (Fig. 6) gene trees. The genus Herengeria is polyphyletic, Isabella is not monophyletic, and Neophrissospongia is monophyletic. All these clades are strongly supported. The 28S gene analysis does not resolve the genus Neoschrammeniella as monophyletic. However, both gene trees indicate a sister group relationship of Neoschrammeniella castrum and Neoschrammeniella norfolki to the Herengeria/Isabella clade, with high support values (bootstrap of 82% and posterior probabilities of 0.99). Species of Herengeria auriculata are sister to Isabella mirabilis, which is also highly supported in both CO1 and 28S rDNA gene trees. The genus Neophrissospongia represents a sister clade to the genera Herengeria, Neoschrammeniella and Isabella.

Family Pleromidae
The family Pleromidae is polyphyletic and is represented by the genera Pleroma and Anaderma. The genus Anaderma seems to be related to Characella pachastrelloides incertae sedis, sensu Cárdenas et al. (2011,2012) [36,68], however this is not supported by posterior probabilities or bootstrap values. In contrast the genus Pleroma is recovered as the sister to the Corallistidae, with strong support (bootstrap of 84% and posterior probability of 1.00).

Family Theonellidae
The family Theonellidae is monophyletic and contains the genera Theonella and Discodermia. Both genera are monophyletic and form a sister group to each other. All nodes are highly supported. The exact position within the Astrophorida however, remains unclear due to low resolution within the gene trees.

Family Scleritodermidae and Siphonidiidae
The monophyly of Tetillidae as suggested by Szitzenberg et al. 2013 [69] could not be corroborated in any of our analyses, independently of whether the lithistid families Scleritodermidae and Siphonidiidae were included (28S rDNA gene tree) or not (CO1 gene tree). Scleritodermidae is monophyletic and is represented in the 28S rDNA gene tree with the genera Microscleroderma, Scleritoderma and Aciculites. The genera Microscleroderma and Scleritoderma are sister groups, while Aciculites (Scleritodermidae) group together with Siphonidium sp. (Siphonidiidae). All these nodes are highly supported.

Parsimony reconstruction of ancestral states
A summary of the parsimony reconstruction of possible ancestral states for megascleres and microscleres composed of homologous characters is given in Fig. 7. These were derived from precise morphological descriptions from the literature and new observations in this study. For some taxa, such as Isabella mirabilis and Neoschrammeniella norfolki, however, there could be difficulties interpreting whether character states of the various streptaster microscleres (spiraster/amphiaster/plesiaster) [51] were homologs or analogs. In this case we followed the definition of streptasters sensu Sollas (1888), where amphiasters bear some analogy to spirasters as they are only differentiated in the shaft, which could be either straight (5amphiaster, see Fig. 4G) or spiral (5spiraster see Fig. 4I,J). In addition, definitions of other streptasters, like plesiasters and metasters were used as described in the study of Cárdenas et al. 2012 [68]. This produced a total number of 13 megascleres (five different desmas and eight different triaenes) and 13 microscleres (Fig. 7). Our results show possible multiple convergences of megascleres and microscleres. These data indicate that megaclone desmas could have evolved two times independently in Pleroma and Anaderma and tetraclone desmas could have developed twice independently in the families Theonellidae and Phymatellidae. By comparison, dicranoclone desmas possibly evolved only once in the family Corallistidae, and desmas of triaenose crepis (Exsuperantia sp.) and trider-like desmas (Macandrewia rigida) also possibly developed only once. Dichotriaenes could have evolved three times independently in the families Phymatellidae and Corallistidae as well as in the genus Anaderma. Mesotriaenes (Triptolemma intextum) and discotriaenes (Theonellidae) probably only appeared once. Phyllotriaenes could have evolved at least three times independently in the genus Theonella, Macandrewia and Exsuperantia. Anatriaenes may have evolved four times independently in different astrophorid and lithistid groups and were lost in some taxa (e.g. Stelletta lactea). Long and short-shafted triaenes were probably lost several times independently in many different astrophorid genera. Calthrops could have appeared at least twice independently according to this dataset (Pachastrella nodulosa and Calthropella geodioides). Our analysis also indicates a high potential of convergent spicule evolution and numerous secondary losses within most microscleres including amphiasters, spirasters, plesiasters, microxeas, euasters, sterrasters and microrhabds. This mapping indicates that secondary losses are four times more frequent in microscleres than in megascleres.

Phylogeny of lithistids compared with previous molecular and morphological studies
From molecular phylogenetic analysis of 68 lithistid demosponges (the largest lithistid taxon sampling to date), our study has showed the complexity of spicule evolution within the polyphyletic 'order Lithistida'. Previously Burton (1929) and de Laubenfels (1936) had suggested the affiliation of triaene-bearing lithistids to the Astrophorida, subsequently affirmed by Lévi (1973) and accepted by Pisera & Lévi (2002) (see Tab.1). Similarly, previous data from different gene regions suggested that the four families Corallistidae, Neopeltidae, Phymaraphiniidae and Theonellidae, including the eight genera (Corallistes, Neophrissospongia, Callipelta, Homophymia, Exsuperantia, Discodermia, Manihinea, and Theonella) all belonged to the Astrophorida (for references see Tab.1). Our study corroborates all these findings and additionally provides evidence that Herengeria, Isabella and Neoschrammeniella (Corallistidae) should also be included in Astrophorida. Previously, the assignment of the four triaene-bearing lithistid families Isoraphiniidae, Macandrewiidae, Phymatellidae and Pleromidae to the Astrophorida was based only on morphological observations [36,39], and is now confirmed based with molecular data. Molecular analyses undertaken in the present study corroborated these hypotheses for the first time, affirming the relationship of Pleromidae (Pleroma, Anaderma), Phymatellidae (Neoaulaxinia, Reidispongia), Isoraphiniidae (Costifer) and Macandrewiidae (Macandrewia) to the Astrophorida. Based on spicule morphology Burton (1929) suggested a close relationship of the lithistid genera Microscleroderma and Scleritoderma (Scleritodermidae), both characterized by rhizoclone desmas, to the spirophorid family Tetillidae due to the possession of similar microscleres (sigmaspires). He also included the rhizoclone desma-bearing genus Leiodermatium (Azoricidae) in this group, which lacks sigmaspires. Later morphological observations by de Laubenfels (1936) assigned Microscleroderma to the Poecilosclerida. In Systema Porifera [70,71], adopting a conservative taxonomy approach, rhizomorine lithistids were divided into two families: Azoricidae (Leiodermatium and Jereicopsis) and Scleritodermidae (Aciculites, Amphibleptula, Microscleroderma, Scleritoderma and Setidium), based on the presence or absence of certain microscleres [70]. Previous molecular phylogenies had suggested a close relationship of the genera Leiodermatium (Azoricidae) [35], Aciculites, Microscleroderma and Scleritoderma (Scleritodermidae) [7,14,39,46,47,72] to the Tetractinellida. However, the exact relationships to either Astrophorida or Spirophorida remained uncertain. Our 28S rDNA results revealed a highly supported monophyletic clade of Scleritodermidae+Siphonidiidae, which was also partly observed by   [39]. Conversely, the monophyly of Tetillidae [69] could not be confirmed from any of our analyses, independently of whether or not Scleritodermidae and/or Siphonidiidae were included. This result is also similar to the findings of Redmond et al. (2013) [39]. The clade Scleritodermidae+Siphonidiidae neither belongs to Astrophorida nor to spirophorids, but instead it shows a sister group relationship to the Astrophorida+Spirophorida clade. However, it should be mentioned here that other families of the order Spirophorida (Samidae and Spirasigmidae) are missing in our analysis and thus the exact classification of this clade is still in need of further investigations. Further, the homology and/or convergence of sigmatose microscleres still remains unclear and need further investigations. Our molecular findings that Aciculites orientalis (Scleritodermidae) is the sister to Siponidium sp. (Siphonidiidae) is supported by morphological observations, where Aciculites has relatively more tuberculate rhizoclone desmas and Siphonidium has thorny and spined rhizoclone desmas, suggesting that both rhizoclone desmas are analogs and probably belong to different desma categories.
Family Vetulinidae Lendenfeld, 1903 Vetulinidae is represented by one genus and species (Vetulina stalactites). It is only known from the Caribbean (Barbados) and morphologically characterized by sphaeroclone desmas and the absence of ectosomal spicules and microscleres [73]. Based on morphological observations Van Soest & Stentoft (1988) [74] and Gruber (1993) [75] suggested a close relationship of Vetulina to the genera Siphonidium (Siphonidiidae) and Leiodermatium (Azoricidae). However,   [73] indicated these were weak assumptions and noted the occurrence of uniaxial or polyaxial sphaeroclone or astroclone-like desmas -not observed in any other lithistid or non-lithistid demosponges. Molecular investigations using different markers and fragments (18S and 28S rDNA, see also Tab.1) indicated with strong support the sister group relationship of Vetulina to freshwater sponges (Spongillida). Our study strongly confirms and corroborates these findings, for the first time using the mitochondrial CO1 gene. Morphologically, Spongillida differ from Vetulina by the presence of microscleres, megascleres and gemmoscleres and absence of sphaeroclone desmas [76]. One explanation could be that Spongillida lost its possession of sphaeroclone desmas, as this process seems more phylogenetically parsimonious than the evolution of new desmas. This discrepancy of morphological versus molecular data remains unresolved at present and needs further attention. As 'Lithistida' is no longer an accepted ordinal taxon, and the genus Vetulina cannot be assigned to any other existing order of the Demospongiae (as well as their morphological differences to the Spongillida), lead us to the taxonomic action to resurrect Sphaerocladina for Vetulina, based on the existing paleontological concept of Sphaerocladina. Firstly, Vetulina has probably been separated from the order Spongillida for a long evolutionary time. Secondly, Vetulina has an unequivocally long and continuous history dating back to the Middle Jurassic to the present through the known fossil record [77]. Thirdly, in this particular case there is no reason to create a new higher taxon for a Recent genus when the taxonomic concept is otherwise identical to the continuous palaeontological concept of Sphaerocladina. The taxon Sphaerocladina Schrammen 1924 was first used as a Suborder to include fossil sponges with sphaeroclonar desmas, like those in Vetulina.
Family Desmanthidae Topsent, 1894 The family Desmanthidae comprises four genera: Paradesmanthus Pisera & Lévi, 2002, Sulcastrella Schmidt, 1879, Desmanthus Topsent, 1894 and Petromica Topsent, 1898. They are encrusting sponges with branching monocrepidial desmas. Ectosomal microscleres (sanidaster-like) are only found in the genus Paradesmanthus. Burton (1929) and de Laubenfels (1936) had already noted the similarity of these characters to other non-lithistid demosponges, and assumed a close relationship to the Halichondrida. Morphologically, desmas of Petromica are different from those found in other genera of this family, which would support the polyphyly of this family. Morphological descriptions of Lithobubaris (5Sulcastrella) confirm the close relationship of Desmanthus, Sulcastrella and Paradesmanthus to the bubarid genera Monocrepidium and Bubaris [78].   [79] acknowledged the resemblance of all these genera to halichondrids. However, their precise placement is not possible based solely on morphological characters. Only a few previous molecular studies had included some species of this family. In the dataset of Morrow et al. (2012) [49] two partitions of the 28S rDNA gene highly supported the grouping of Desmanthus within Dictyonellidae sensu Morrow et al. (2012) [49]. This result was also supported by Redmond et al. (2013) [39] based on the analysis of the 18S rDNA. Here, we add for the first time an unlinked molecular marker, from the mtDNA CO1 gene, and support the assignment of Desmanthus incrustans to Dictyonellidae, and further, provide moderate support of a sister group relationship to the species Dictyonella sp. Based on morphological character analysis, Van Soest & Hajdu (2000) [80] suggested resurrecting the family Desmanthidae Topsent, 1893 within the 'Lithistida' demosponges for the genera Desmanthus and Lithobubaris (5Sulcastrella) by excluding Petromica.   [39] already formally reallocated the genera Desmanthus, Sulcastrella and Paradesmanthus to Bubaridae. In contrast, our molecular data, based on the mtDNA CO1 gene, strongly recommend the reallocation of Desmanthus to Dictyonellidae, as proposed by Cárdenas et al. (2012) [6]. Since no molecular data for any species of the genera Sulcastrella and Paradesmanthus exists yet, for the time being we support their reallocation to the Bubaridae, as proposed by Redmond et al. (2013) [39]. Molecular data based on the 18S rDNA gene of the genus Petromica showed a close relationship to Halichondriidae sensu Morrow et al. (2012) [49]. Our analysis of mitochondrial CO1 sequences is consistent with their hypothesis. Additionally, our results display a strongly supported clade of the genus Petromica together with Topsentia ophiraphidites (Halichondriidae). This confirms earlier morphological findings of Van Soest & Zea (1986) [81]. Muricy et al. (2001) [82] amended the monophyly of Petromica, which is acknowledged in our molecular results, and showed support for the affinity with the Halichondriidae sensu Morrow et al. (2012). We therefore formally recommend reallocating Petromica close to halichondriids.

Family Corallistidae
Our molecular results (28S rDNA, C1-D2 partition) concerning the relationships within 'lithistids' provide strong evidence that the monophyletic family Corallistidae is closely related to Pleroma of the family Pleromidae. This outcome was expected from morphological observations, due to the similarity of desma structures. Megaclone desmas of Pleroma and dicranoclone desmas of Corallistidae might have originated in the same way and only final stages differ in these desmas. Additionally, dichotriaenes occur as ectosomal spicules in both families. Interestingly, no other astrophorids group with this clade, affirming the persistent occurrence of dicranoclone and megaclone desmas since the Paleozoic. Our molecular data further indicate that Herengeria auriculata is the sister-taxon to Isabella mirabilis. This relationship is morphologically supported with the main differences being the possession of euaster-like microscleres in Isabella mirabilis from the Norfolk Ridge [51]. Additionally, we confirmed the non-monophyly (CO1 gene tree) of the genus Isabella as also shown in the recent study of Carvalho et al. 2014 [83]. The sister group relationship of the species Herengeria vasiformis to a clade containing H. auriculata, Isabella mirabilis, Neoschrammeniella castrum and N. norfolki is highly supported. The polyphyly of the genus Herengeria could be explained by evidence of differing gross morphologies between the two species, indicative of taxonomic divergence. Herengeria vasiformis is vase-shaped and H. auriculata is much more massive; and H. vasiformis has thicker microxeas and more massive and less regularly developed rhabd-like spirasters, as well as smaller spirasters, as described by Schlacher-Hoenlinger et al. (2005) [51]. Neoschrammeniella norfolki differs from N. castrum and other genera of the family Corallistidae by the presence of plesiasters and absence of microxeas. When the genus Corallistes was included in the analyses, monophyly of the family Corallistidae was not supported in the study of Redmond et al. (2013) [39]. However, a ''Corallistes sp. (AY737636)'' formed a clade with Neophrissospongia microstylifera, while a ''Corallistes sp. (AJ224646)'' was not found to be related to Theonellidae. This is likely a consequence of misidentification of this taxon and/or an inexact alignment compared to other sequences of Corallistidae. Considering all these aspects, the family Corallistidae should also be reallocated to Astrophorida.

Family Pleromidae
The family Pleromidae was recovered as polyphyletic, with Pleroma menoui closely related to Corallistidae and Anaderma rancureli to Characella pachastrelloides (Pachastrellidae). This is in agreement with our morphological character analysis, which also indicated its likely polyphyly. Pleroma lacks anatriaenes in contrast to Anaderma, which unequivocally includes them. Even though the relationship between Anaderma and Characella is not supported in our 28S rDNA gene tree, it might be conceivable based on the presence of similar morphological characters (e.g. anatriaenes) [68].

Family Macandrewiidae
The status of Macandrewia (Macandrewiidae) has been revised many times in the past, changing from affinities to Corallistidae [84] to belonging to Callipelta [32]. The possession of phyllotriaenes and desmas with triaenose crepis, however, supports a close relationship to other astrophorids. Due to the low variation within the 28S rDNA gene, it was not possible to determine the exact relationships with other lithistids or to astrophorid clades. Therefore, escalated taxon sampling, as well as gene sampling, needs to be improved in future to clarify the phylogenetic position of the family Macandrewiidae.

Family Phymaraphiniidae
The family Phymaraphiniidae contains three genera: Exsuperantia Ö zdikmen, 2009 [85], Kaliapsis Bowerbank, 1869 [86] and Lepidothenea de Laubenfels, 1936. Burton (1929) suggested a close relationship of Exsuperantia to Stellettidae due to its possession of phyllotriaenes. The original placement of Exsuperantia was with Theonellidae, due to similar ectosomal phyllotriaenes and microscleres as found in the genus Racodiscula (Theonellidae). However, the sculpture of the trider-like desmas (Fig. 3 G-H) clearly differentiate those two genera and families [87]. The only previous molecular analyses of Exsuperantia sp. did not support its close relationship with the tetraclone-bearing family Theonellidae [36]. Our results group Exsuperantia sp. as a sister to the astrophorid families Ancorinidae and Pachastrellidae, and the lithistid species Anaderma rancureli. However, neither BI nor ML values support this suggestion and so for the moment we allocate Exsuperantia to Astrophorida until further data is available. The phylogenetic position of the other two genera Kaliapsis and Lepidothenea will be the matter of further investigations.

Family Theonellidae
The family Theonellidae contains five genera: Discodermia du Bocage, 1869, Manihinea Pulitzer-Finali, 1993, Racodiscula Zittel, 1878, Siliquariaspongia Hoshino, 1981 and Theonella Gray, 1868. Theonellidae is characterized by ectosomal spicules ranging from phyllotriaenes to discotriaenes, choanosomal tetraclone desmas and microscleres as acanthorhabds, microxeas, streptasters and amphiasters. Due to the possession of triaenes Theonellidae was usually considered to group with astrophorid sponges [24,25,27]. More recently there has been increased interest in bioactive compounds from theonellids [18], with the genera Discodermia and Theonella receiving special attention and resulting in the amplification of four different gene regions for Discodermia and three for Theonella (see Tab.1). Previous phylogenetic reconstructions based on mtDNA CO1 and 28S rDNA have shown that Theonellidae is monophyletic [36]. This result was in contrast to those observed from the 18S rDNA analysis [39]. Our present molecular analyses of both gene regions (mtDNA CO1 and 28S rDNA) strongly support the monophyly of Theonellidae, and additionally the sister group relationship of Theonella to Discodermia, supporting the conclusions of Cárdenas et al. (2011) [36]. A sister group relationship of Theonellidae and Corallistidae as proposed by earlier morphological [27] and molecular analyses [14], is not supported by any of our gene trees. Tetraclone desmas and dichotriaenes are the characteristic megascleres for the family, while the three genera are differentiated by the possession of different microscleres. Until the present study no molecular data existed for this group, and so its precise placement among the astrophorids remained uncertain. Here we show for the first time the monophyly of the family and its genera, and suggest a close relationship with the astrophorid family Pachastrellidae. Similar triaenes found in both families would support this moderately supported molecular sister group relationship. We therefore propose reallocating the family Phymatellidae to the Astrophorida.

Evolution of megascleres and microscleres in lithistid sponges
Our results suggest that desmas have evolved several times independently in different lithistid demosponge groups within the order Astrophorida. Furthermore, and conversely, secondary loss of desmas may have also occurred several times independently. However, the silica concentration of seawater has been shown to influence the development of spicules in demosponges [41], providing the possibility that if the silica concentration in seawater is low, desmas dis-articulate. So, if megascleres lose their function (e.g. as structural support for the cortex or as defense against predators), a secondary loss of megascleres is feasible. Microscleres have been lost frequently in the past within Tetractinellidae [36,69].