Integrating morphological and molecular taxonomy with the revised concept of Stelligeridae (Porifera: Demospongiae)

This study reinforces and extends the findings of previous molecular studies that showed that there is a close relationship between species assigned to Halicnemia , Higginsia , Paratimea and Stelligera a nd that the family Heteroxyidae is polyphyletic. In the present study the re-examination of a large number of specimens and species of Halicnemia, Higginsia and Paratimea has resulted in the description of one new species of Halicnemia and six new species of Paratimea ; H. caledoniensis sp. nov., P. aurantiaca sp. nov., P. dentata sp. nov., P. hoffmannae sp. nov., P. lalori sp. nov., P. mosambicensis sp. nov. and P. rosacea sp. nov. respectively; the resurrection of Halicnemia gallica and a better understanding of the characters uniting Stelligeridae. A new species of Heteroxya, H . beauforti sp. nov. is also described. We demonstrate that many of the taxa assigned to Heteroxyidae (on the basis of the possession of smooth or acanthose microxeas) are more closely related to other families and we propose several changes to the classification of Heteroscleromorpha. Desmoxyidae is resurrected from synonymy and transferred to Poecilosclerida; Higginsia anfractuosa is transferred to Hymedesmiidae and a new genus Hooperia gen. nov. is erected for its reception; Higginsia durissima is returned to Bubaris (Bubaridae); Higginsia fragilis is transferred to Spanioplon (Hymedesmiidae); Hemiasterella camelus is transferred to Paratimea ; Raspailia (Parasyringella) australiensis and Ceratopsion axiferum are transferred to Adreus (Hemiasterellidae).


INTRODUCTION
Sponges are one of the most ancient groups of multicellular organisms with a fossil history dating back to the Cambrian (Botting & Muir, 2017) and with steroid biomarkers in the Neoproterozoic some 635-717 Myr (Love et al., 2009).Recent molecular studies confirm that they are one of the earliest diverging metazoans (Pisani et al., 2016;Simion et al., 2017;Dohrmann & Wörheide, 2017).Sponges occupy a wide range of aquatic, benthic environments, from temporary freshwater pools to abyssal depths.They are a very successful group with over 9000 described species recorded in the World Porifera Database [herein referred to as WPD (van Soest et al., 2018)], however Hooper & van Soest (2002) state that there is likely to be a similar number of undescribed taxa and Appeltans et al. (2012) estimate that there may be as many as 18000 undescribed species.Traditional taxonomy and classification of sponges are based on morphology, cytology and reproductive traits but particularly spicule morphology and skeletal architecture (Hooper & van Soest, 2002).The combination of very ancient lineages, a high proportion of undiscovered taxa and relatively few morphological characters make understanding evolutionary relationships among sponge taxa difficult and has resulted in many orders, families and genera composed of polyphyletic assemblages (Boury-Esnault, 2006).Fortunately, a growing number of molecular phylogenetic studies, many of which integrate observations on morphology, are improving our understanding of the phylogenetic relationships among sponges; these were reviewed by Cárdenas et al. (2012).
The current study expands on the preliminary work of Morrow et al. (2012) on Stelligeridae using a combination of molecule-based phylogenetic hypotheses derived from 18S and 28S rRNA and CO1 barcoding fragments together with a careful re-examination of skeletal morphology to try and resolve the phylogenetic relationships within this group of sponges.It is not a monographic revision, but rather a contribution towards a better understanding of the systematics of Stelligeridae and, more broadly, Heteroscleromorpha Cárdenas, Pérez & Boury-Esnault, 2012 in general.
The class Demospongiae Sollas, 1885 represents almost 80% of all known sponges.Based on molecular results (Morrow & Cárdenas, 2015) Demospongiae is subdivided into three subclasses; Keratosa Grant, 1861; Verongimorpha Erpenbeck, Sutcliffe, de Cook, Dietzel, Maldonado, van Soest, Hooper & Wörheide, 2012 and Heteroscleromorpha.Heteroscleromorpha is by far the largest of all the subclasses with approximately 6773 valid species.Morrow et al. (2012) used data derived from partial 28S rRNA and CO1 barcoding fragments to propose a new classification for Heteroscleromorpha.They demonstrated that there was a close relationship between Halicnemia Bowerbank, 1864, Paratimea Hallmann, 1917and Stelligera Gray, 1867and resurrected Stelligeridae Lendenfeld, 1898 for this well supported clade.Erpenbeck et al. (2012b) using CO1 barcoding sequences confirmed these results and also showed that Higginsia, a genus which is morphologically similar to Halicnemia, also clustered within Stelligeridae.
The taxonomic interpretation of Halicnemia and Paratimea using traditional morphological characters has resulted in very different classifications.Historically Paratimea and Stelligera have always been considered to be closely related.The spicules in both genera are styles or tylostyles, oxeas (often centrotylote) and oxyaster microscleres.However, their relationship with Halicnemia is more controversial.In Halicnemia the megascleres are also styles/tylostyles and centrotylote oxeas however the microscleres are centrangulate spined microxeas (acanthoxeas) instead of asters.Table 1 summarizes the various relationships espoused by the main authors who have written on this subject.Below we discuss the confused taxonomic history of Halicnemia and Paratimea.Topsent (1897) reasoned that Halicnemia patera Bowerbank, 1864, Hymeraphia verticillata Bowerbank, 1866 and Bubaris constellata Topsent, 1893 belonged in a single genus (Halicnemia) on the basis of the shared possession of tylostyles and distinctive, centrotylote oxeas forming the ectosomal skeleton.He considered that the microscleres could be either oxyasters or acanthoxeas.He also speculated that the microxeas in Halicnemia could have been derived from asters; however, Dendy (1922) regarded it as more likely that the asters found in Halicnemia constellata were pseudasters derived from the spined microxeas.Carter (1875) was the first to notice the similarities in spicule composition between Halicnemia patera and Hymeraphia verticillata.
The monotypic genus, Desmoxya Hallmann, 1917 was created for Higginsia lunata Carter, 1885. Hooper & Lévi (1993) synonymised Desmoxya and Dendropsis Ridley & Dendy, 1886 with Higginsia Higgin, 1877 based on the shared apomorphy of spined microxeas, however they highlighted that there were major skeletal differences and that Desmoxya may need to be resurrected to accommodate Higginsia-like species that lack any evidence of axial and extra-axial skeletons e.g.Higginsia lunata and their new species H. anfractuosa.The discovery of another species, Desmoxya pelagiae van Soest & Hooper, 2005, from the cold water coral reefs of the North Atlantic led to the resurrection of Desmoxya.Van Soest & Hooper, 2005 considered Desmoxyidae synonymous with Heteroxyidae and changed the family to which Halicnemia, Higginsia, Desmoxya etc. were assigned from Desmoxyidae to Heteroxyidae.

COLLECTION OF MATERIAL
This study used a combination of voucher specimens from various institutions together with freshly collected material for DNA analysis.Shallow-water specimens were collected either by SCUBA diving or by shore collecting.Deep-water specimens were collected during the cruise CE10004 of RV Celtic Explorer, using the deep-water Remotely Operated Vehicle Holland I; the ARK-XXII/1a 2007 expedition on board the RV Polarstern to Northern Norway, using the manned-submersible JAGO; the PAMELA-MOZ01 2014 expedition (IFREMER) and the BIOMAGLO 2017 expedition (MNHN/IFREMER) to the Mozambique Channel, both using a Warrén dredge (DW).The sponges were photographed in situ, then tissue samples approximately 1 cm 3 were collected.The specimens were fixed in 96 % ethanol normally within 1 hour of collection.After 24 hours the ethanol was changed to prevent dilution by seawater.We have attempted to analyse as many species and genera that are currently assigned to Heteroxyidae and Hemiasterellidae as we could obtain.The taxa used in the analyses together with their catalogue numbers and, where relevant, their GenBank accession codes are listed in Supporting information, Data S.1.
The following abbreviations are used for the institutions from which we have examined material:

PREPARATION AND EXAMINATION OF MATERIAL
Nitric acid spicule preparations and thick section mounts were made following the protocols described by Picton & Goodwin (2007).Photomicrographs of the spicules and sections were made using Nikon Eclipse 80i light microscope equipped with a Jenoptic ProgRes CT3 digital camera and software.Spicules were measured from nitric acid histological preparations using this equipment.Unless otherwise stated (n: number of spicules measured) 20 spicules of each category were measured for each specimen.Spicule dimensions are either given as a range, or as a minimum-mean-maximum throughout the manuscript.
At NUI Galway, nitric acid preparations of spicules were made directly onto cut microscope slides, air dried then mounted on scanning electron microscope (SEM) discs and coated with gold then viewed in the HBB SEM.At the Netherlands Centre for Biodiversity Naturalis, spicules were air dried directly onto the SEM studs, sputter-coated with gold and examined and photographed using a JEOL SEM.At the Queensland Museum, tissue was dissolved in 12.5% sodium hypochlorite, neutralized in distilled water, rinsed twice in 70% ethanol, then in 98% ethanol and then air dried.SEM preparations were sputter coated in gold to improve resolution.The scanning electron micrograph photos were taken using a Hitachi TM-1000 SEM.All plates were assembled in Adobe Photoshop.

PHYLOGENETIC ANALYSES
The 18S rRNA sequences were obtained from GenBank and were primarily generated by Redmond et al. (2013).Many of these sequences were from the same specimens and DNA extracts used by Morrow et al. (2012Morrow et al. ( , 2013)).The 28S rRNA and CO1 sequences are a combination of previously published sequences available on GenBank (primarily from Morrow et al., 2012;2013), and sequences newly generated for this study.
Sequences were managed in Geneious R10 (http://www.geneious.com,Kearse et al., 2012).Forward and reverse reads were assembled into contigs using the assembly function of the software and checked for inconsistencies.Where inconsistencies arose Geneious used the better quality of the two reads or introduced IUPAC ambiguity codes into the consensus sequence.Sequences were aligned with MAFTT (Katoh et al. 2002) and trimmed in Geneious.Complete or nearly complete sequences of 18S rRNA, 28S rRNA (D1-D8) and CO1 Folmer fragments were concatenated for Figure 1.The best fitting model for each of the three partitions was separately selected using JModelTest (Darriba et al., 2012).The GTR+G+I model was identified as the best-fit model of molecular evolution for all datasets.

DISCUSSION
Our combined hypothesis (Fig. 1) and single-gene trees (Supporting Information, Figs S2-S5) are congruent and show that there is strong molecular evidence for a close relationship between some former 'hemiasterellid' taxa (Paratimea and Stelligera) and taxa that were previously assigned to Heteroxyidae (Halicnemia, Higginsia and Acanthoclada).Our results strongly support the resurrection of Stelligeridae sensu Morrow et al. (2012) for this clade.The recently established Plenaster Lim & Wiklund, 2017 was tenatively assigned to Stelligeridae on the basis that "Stelligeridae is the only family in the order Axinellida that has members bearing styles and euasters like Plenaster."However, their molecular trees (28S rRNA D1-D2 region and CO1 Folmer fragment) clearly show that Plenaster is not a stelligerid; the taxonomic affinities of this genus will be part of a future manuscript.Hooper (1986) suggested Paratimea might be an encrusting form of Stelligera given that their respective type species had similar aster morphology, whereas Voultsiadou-Koukoura & van Soest, 1991 considered Paratimea a valid genus.On our 28S D1-D2 tree (Supporting Information, Fig. S5) Paratimea appears polyphyletic, P. loennbergi and P. aurantiaca sp.nov.are paraphyletic with respect to Stelligera spp.and P. oxeata, P. hoffmannae sp.nov.and P. rosacea sp.nov.form a sister clade to Stelligera spp.+ Paratimea pars.It may be that in the future encrusting Paratimea spp.such as P. loennbergi, P. aurantiaca sp.nov., P. dentata sp.nov.and P. constellata (the type species) are transferred to Stelligera which has priority over Paratimea, and that a new genus is established for P. oxeata, P. hoffmannae sp.nov., P. rosacea sp.nov.and other massive, mostly deep-water 'Paratimea' species that have large oxeas as their principal spicules and relatively large, often asymmetric, asters.Until we have more sequence data to test the validitiy of this, we retain Paratimea.
There are several morphological characters that unite Stelligeridae, in spicule morphology, surface architecture, cell types, reproduction, and these are discussed below.
Spicules: Megascleres; the principal spicules in Paratimea and Halicnemia are long, slender tylostyles, styles or oxeas whereas in Stelligera and Higginsia they are only styles.Acanthoclada is unusual in that it has styles and rhabdostyles.Some species of Paratimea and Halicnemia have distinctive short, club-like tylostyles (Figs 2F, P. loennbergi; 11B, Halicnemia patera; 15B, H. caledoniensis sp.nov.) which may be a synapomorphy for the clade.Alander (1942), in his description of P. loennbergi (as Halicnemia loennbergi) remarked that the genus Halicnemia comprised species of two categories.One group to which H. constellata Topsent, 1897 belonged has one type of tylostyle plus oxyasters.The other group with H. patera has two types of tylostyles plus spined microxeas.In common with H. patera his new species H. loennbergi had two types of tylostyles but the microscleres were oxyasters.He considered that his new species united this group.These short, stout club-like tylostyles have also been observed in Halicnemia caledoniensis sp.nov.
Secondary spicules are small ectosomal styles, oxeas or anisoxeas: These spicules often form distinctive bouquets around the protruding larger styles, tylostyles or oxeas.In Paratimea and Halicnemia they are often centrotylote (Fig. 24A, D & G).In Halicnemia gallica there are usually two or three equidistant tylote swellings.The centrotylote oxea of H. verticillata are very distinctive, the ends of the oxea being fissurate (Fig. 24E).It shares this character with Higginsia bidentifera (Ridley & Dendy, 1886) and Higginsia petrosioides Dendy, 1922.In Paratimea dentata sp.nov., included in our 18S tree (Supporting Information Fig. S2), the centrotylote oxeas are also fissurate (Fig. 24C).This is the first time this distinctive spicule has been found in Paratimea, it has not been reported elsewhere within Demospongiae and is further evidence for Paratimea and Halicnemia being closely related.
Microscleres; in Stelligera and Paratimea the microscleres are smooth rayed oxyasters (Fig. 5B) whereas in Halicnemia and Higginsia they are acanthoxeas (Fig. 5F & I).Topsent (1897) speculated that the acanthoxea were derived from asters and that the two structures were homologous, however Dendy (1922) thought that it was more probable that the oxyasters were merely pseudasters derived from the acanthoxea.Topsent (1928) went further and suggested that the centrotylote oxeas could also have their origin in asters.Sollas (1882) considered the microrhabds in Pachymatisma Bowerbank in Johnston, 1842 to be homologous with the asters in Geodia Lamarck, 1815 as both share a similar position within the ectocortex.By examining the ontogeny of the microscleres he reported that it was possible to trace the development of the microrhabd from the adult form which is cylindrical with rounded ends and a roughened surface, to a smooth fusiform spicule with a central globular enlargement and pointed ends, which he regarded as a biradiate aster.In contrast, the developing asters though progressively smaller, remained multiradiate.Sollas described the number of rays in the asters as very variable with a reduction to four, three or even two rays frequent.The two rayed asters have a central enlargement and closely resemble developing microrhabds.Sollas considered this as evidence that asters descended from microxea.Conversely, Cárdenas et al. (2010) noted that microrhabds are frequently centrotylote and that the tylote swelling may represent the ancestral centre of the aster.We consider the asters and acanthoxea in Stelligeridae to be homologous, if we consider the acanthoxea in Halicnemia caledoniensis sp.nov., they occupy a similar position within the ectosome as the asters in Paratimea and Stelligera.Unfortunately, our molecular trees do not give any clues as to whether the aster or acanthoxea is the ancestral state in Stelligeridae.
In terms of morphology, Acanthoclada is quite different to other stelligerids, in addition to the unusual megascleres (rhabdostyles), it is the only member that has acanthose cladotoxas and birotules, however the 28S D3-D5 tree (Supporting Information, Fig. S3) strongly supports its status as a stelligerid.
Surface architecture: At least some species of Halicnemia, Higginsia, Paratimea and Stelligera share a strikingly similar surface architecture to typical raspailiid species, with large robust megascleres 2-3 mm long protruding from the surface surrounded by a bouquet of thin spicules which are variously described as styles, anisoxea or oxea (Morrow et al., 2013 p. 443).We consider the possession of a raspailiid surface architecture to be an apomorphy for the clade containing Raspailiidae and Stelligeridae.Desqueyroux-Faúndez & van Soest (1997) described a new species Halicnemia diazae from the Galapagos that they reported as "having characters that are typical of both Halicnemia and Higginsia and between the halichondrid family Desmoxyidae and the poecilosclerid families Raspailiidae and Rhabderemiidae".They suggested that further revisions of these genera might result in the merging of Desmoxyidae and Raspailiidae.Although this ectosomal surface architecture appears to be confined to Raspailiidae and Stelligeridae and is strong morphological support for a close relationship between these two families it is not ubiquitous for all the taxa.This highlights the difficulties in defining higher taxonomic groups on the basis of one or few morphological characters.

Cells with inclusions:
The presence of cells characterised by granules or vesicles appears to be constant in Stelligera, Paratimea and Halicnemia (Fig. 25A-C).Although not ubiquitous, these cells are also widespread in Raspailiidae and Axinellidae.These cells are also abundant in Heteroxya beauforti (Fig. 26D) which clusters within Axinellida.Topsent (1891) referred to these cells as 'spherulous cells,' and speculated that they could be responsible for the excretion of slime in Halicnemia (Halicnemia, Paratimea, Stelligera and Acanthoclada all produce copious amounts of slime on collection).As yet no indisputable function has been ascribed to these cells and the same term has been used to describe cells in different Demospongiae orders although it has not been demonstrated that the structure and function are equivalent.Thompson, Barrow & Faulkner (1983) demonstrated that in Aplysina fistularis (Pallas, 1766) the spherulous cells contain secondary metabolites that have an antibacterial function.They speculated that these secondary metabolites may prevent the growth of biofouling organisms, control bacterial communities or deter predators.
Oviparity: Stelligera, Paratimea, Halicnemia and Acanthoclada all appear to be oviparous, Fig. 25A-C shows the oocytes are often surrounded by cells with granular inclusions.The interaction between the oocytes and these cells is unknown.The observation that Stelligeridae is oviparous is further support for Axinellida since Raspailiidae and Axinellidae are oviparous (Alvarez & Hooper, 2002).
Using a combination of molecular phylogenies and careful re-examination of morphological characters we return to a classification system which is very similar to the earlier classifications of Topsent (1928) and Dendy (1922) who recognised the morphological similarities between Halicnemia, Higginsia, Paratimea and Stelligera.One important difference between our results and the classification of Topsent (1928) is the position of Hemiasterella.Topsent included Hemiasterella, Halicnemia (including Paratimea), Higginsia and Vibulinus (= Stelligera) in Astraxinellidae (see Table1), whereas in our molecular genetrees (Fig. 1; Supporting Information Figs S2-S5), Hemiasterellidae (represented by Adreus and Axos) groups with Tethyidae and Timeidae and not with Stelligeridae.In Stelligera and Paratimea the asters are smooth rayed whereas in Hemiasterella, Adreus and Axos the asters are often microspined and come in a variety of size classes.In Hemiasterella typus Carter, 1879 (type taxon) the megascleres are exclusively stylote whereas Hemiasterella s.l.contains a diverse group of species in which the megascleres can be stylote, oxeote or a combination of both.In order to resolve the higher taxonomic placing of Hemiasterella and hence Hemiasterellidae comparable DNA sequences from Hemiasterella species and in particular from H. typus are needed.
The surprising discovery that D. pelagiae clustered close to Tedaniidae (Poecilosclerida) and not with Halicnemia and Higginsia in our molecular trees (Fig. 1; Supporting Information Figs S2-S5) caused us to return to the specimens of D. pelagiae and indeed to the type material of D. lunata and re-examine the morphology more carefully.Using SEM we found morphological characters such as the presence of onychaetes and acanthostyles, that unite Desmoxya pelagiae with Tedania and gained an insight into the homoplasious nature of the acanthoxea that wrongly led to it being classified with Halicnemia and Higginsia.Whilst we are lacking any molecular data from the type species D. lunata, on the basis of the structure of the skeleton and the presence of onychaetes, we resurrect Desmoxyidae for Desmoxya only, and transfer it to Poecilosclerida.Hooper & Lévi (1993), in their description of their new species Higginsia anfractuosa from a lagoon in New Caledonia, stated that morphologically it was most similar to Higginsia lunata Carter, 1885 in growth form, papillose surface features and skeletal architecture and that both species are atypical of other Higginsia.Hooper (2002a) speculated that Desmoxya may need to be resurrected for Higginsia-like species that lack any evidence of axial compression (citing H. lunata and H. anfractuosa), having instead a halichondroid, meandering reticulation of choanosomal tracts.Our 18S and 28S trees (Supporting Information, Figs S2,  S3) contain a specimen from Tanzania, Hymedesmiidae OCDN 3725J, (previously identified as H. anfractuosa) in Redmond et al. (2013); Morrow et al. (2013) and Thacker et al. (2013).Our CO1 tree (Supporting Information Fig. S3) has sequence data from the type specimen of H. anfractuosa (QM G300723) which clusters with Hymedesmiidae G 304373 (identified on GenBank as Crella sp.), at the base of a clade containing species of Hemimycale and not with Higginsia species or indeed Desmoxya pelagiae.We have erected the genus Hooperia gen nov.for Higginsia anfractuosa comb.nov.; we consider it closely related to Hemimycale but easily distinguished from it by the presence of rugose oxea.Although we have no supporting molecular data, we transfer Higginsia fragilis Lévi, 1961 to Spanioplon Topsent, 1890 (Hymedesmiidae) as morphologically it is much closer to Spanioplon than to Higginsia.Erpenbeck, Breeuwer & van Soest (2005)  The remaining genera that are classified in Heteroxyidae by van Soest et al. (2018) and for which we have no molecular data (Alloscleria Topsent, 1927;Alveospongia Santos, Pinheiro, Hajdu & Van Soest, 2016;Julavis Laubenfels, 1936;Microxistyla Topsent, 1928;Negombo Dendy, 1905 andParahigginsia Dendy, 1924) are retained in Heteroxyidae.

Summary of Taxonomic Changes:
Desmoxyidae is resurrected from synonymy with Heteroxyidae and transferred to Poecilosclerida; a new genus, Hooperia Morrow gen.nov. is established for Higginsia anfractuosa; Higginsia durissima is returned to Bubaris; Higginsia fragilis is transferred to Spanioplon; Hemiasterella camelus is transferred to Paratimea; Halicnemia gallica is resurrected from synonymy with Halicnemia patera; Raspailia (Parasyringella) australiensis and Ceratopsion axiferum are transferred to Adreus.-Esnault & Solé-Cava (2004) and Cárdenas et al. (2012) pointed out that when morphological and molecular trees are not congruent we need to sample additional genes but above all to reassess very carefully the morphological characters without any preconceived ideas.Then the molecular trees can be reinterpreted in light of reconsidered morphological data.Using this approach has helped us gain a better understanding of the phylogenetic relationships within Stelligeridae and Heteroscleromorpha in general.Our study shows strong morphological and molecular support for the family Stelligeridae sensu Morrow et al. (2012).In Figure 1, which represents our largest dataset in terms of number of nucleotides but with fewer taxa, Paratimea is closest to Stelligera and Halicnemia and Higginsia form a sister relationship.The inclusion of more species and more molecular markers, especially from the type species are needed to determine whether Halicnemia and Higginsia are monophyletic.Our molecular trees do not support the monophyly of Paratimea, the 28S tree (Supporting Information Fig. S5) shows several of our new species of Paratimea (P.hoffmannae sp.nov., P. lalori sp.nov.and P. rosacea sp.nov.) clustering separately to other Paratimea spp.These new species are all from deep-water habitats, share a massive growth form, have oxeas as their principal spicules and have relatively large asters, often with unequal lengthed rays.

Boury
It is clear that within Stelligeridae microscleres can be either asters or microxeas (including acanthose cladotoxas and birotules) and the argument for the separation of taxa into separate families and orders based on whether they possess asters or microxeas can no longer be supported.This illustrates well the difficulties for sponge systematics, deciding which characters are diagnostic and which are variable or homoplasious.The molecular trees generated in this study have helped us to distinguish characters that are homologous e.g. a raspaillid surface architecture and those that are homoplasious e.g. the occurrence of acanthoxeas.
The inclusion in this study of species that are new to science e.g.Paratimea dentata sp.nov.and Halicnemia caledoniensis sp.nov., has enhanced our understanding of evolutionary relationships within Heteroscleromorpha.Emended Diagnosis: Axinellida in which the choanaosomal skeleton can be composed of styles, tylostyles, oxeas or rhabdostyles.Ectosomal region often with protruding megascleres surrounded by bouquets of smaller, slender accessory oxeas or styles.Ectosomal crust heavily reinforced with microscleres.Accessory oxeas often with centrotylote swellings, occasionally with fissurate terminations.Microscleres can be smoothrayed euasters; spined microxeas often bent or centrangulate or acanthose cladotoxas and birotules.Where known, reproduction is oviparous.Presence of cells with granular inclusions in most species.All produce slime on collection.

although see remarks below).
Remarks: A detailed description of the morphological characters that unite Stelligeridae is given in the discussion.The molecular trees of Lim et al. (2017) clearly show that Plenaster is not a stelligerid, the taxonomic affinities of this genus will be part of a future manuscript.

STELLIGERA GRAY, 1867
Diagnosis (modified from Hooper, 2002b): Stelligeridae with erect branching growth form; choanosome composed of axial region of styles and oxea and extra-axial region of long projecting styles perpendicular to axis and styles/oxea forming an irregular reticulation; ectosomal skeleton composed of bouquets of slender styles surrounding protruding extraaxial styles; smooth-rayed euaster microscleres form an ectosomal crust.Produces copious amounts of slime on collection.Remarks: Hooper (2002b) speculated that S. rigida (Montagu, 1818) from the British Isles and S. nux Lendenfeld, 1898 from the Mediterranean might be synonyms of S. stuposa although the types were not examined.From extensive collecting of S. stuposa and S. rigida from areas close or adjacent to the type locality we can verify that differences in external morphology are consistent with small differences in aster morphology.There were also small differences between the two taxa using complete 18S rRNA (Supporting Information, Fig. S2) and partial fragments of 28S rRNA (Supporting Information, Figs S3, S5) however the CO1 Folmer fragment (Supporting Information, Fig. S4) was identical for the two species.

PARATIMEA HALLMANN, 1917
Emended diagnosis: Stelligeridae with encrusting or massive growth form.Choanosomal skeleton lax, encrusting species have hymedesmioid skeletal architecture consisting of erect tylostyles and paratangential tracts of centrotylote or polytylote ectosomal oxea.Massive species have oxeote megascleres, arranged without order.Oxea are also scattered throughout the choanosome in dragmata.Microscleres are smooth rayed euasters, most abundant in surface layer.Produce slime on collection.
Description: Megascleres long, slender tylostyles 2500-3000 x 13-14 µm (these measurements are taken from the original description.In the slide from the holotype that we examined only one tylostyle was intact, it measured 2 mm), sub-trilobate head 17 µm (Fig. 2A).
Remarks: This species is very similar to P. loennbergi with the exception that in P. loennbergi there are two categories of tylostyles: long, slender tylostyles with a globular to sub-trilobate head (>2 mm x 32 µm) similar to those of P. constellata and, in addition, distinctive short tylostyles that are stout and club-like with a pear-shaped, annulated head.These are only found in the base of the sponge where it is in contact with the substratum.It is possible that these two taxa are conspecific but that the microscope preparations made from P. constellata did not include the basal layer with the distinctive short tylostyles.Unfortunately, the type material of P. constellata is missing and it is not possible to ascertain whether the basal layer, if present, contained the short, club-like tylostyles.Therefore, we retain P. constellata and P. loennbergi as two separate species until fresh material from the type locality can eventually be examined and sequenced.
We have examined specimens from the Ulster Museum collection identified as P. constellata.The spiculation and skeletal architecture is identical to that of P. loennbergi and the material has been re-identified accordingly.Records of P. constellata from the coasts of Britain & Ireland (Picton, Morrow & van Soest, 2007) should be attributed to Paratimea loennbergi.The GenBank sequences for Paratimea constellata listed in Morrow et al. (2012) (HQ379218; HQ379397; HQ379284; HQ379352; HQ379419) have been changed to P. loennbergi.
Remarks: The CO1 genetree (Supporting Information, Fig. S4) shows P. camelus clustering closely with P. oxeata, inside Stelligeridae, the pairwise identity matrix (Supporting Information, Fig. S6) shows 99.85% similarity between the two species.A re-examination of the holotype found a ‚raspailiid surface architecture', typical of many stelligerids.The smaller oxyspherasters reported by van Soest (2017; fig.109e) are considered to be contaminants as we did not find any in our spicule preparations or tissue section.These asters are very different to those found in other Stelligeridae and are likely to be contamination from a tethyid sponge.
Colour Topsent (1928) describes the colour as greyish in alcohol.
Choanosomal skeleton principal spicules have a disordered arrangement, spongin lacking, making the skeleton lax and friable.Asters abundant throughout skeleton.
Remarks: Topsent notes that P. duplex is unusual in having a mix of oxeas, styles and tylostyles as the principal megascleres.The oxea are much more common and also larger than the styles or tylostyles.

Colour pale yellow.
Choanosomal skeleton comprised of long tylostyles with their heads embedded in a basal layer of spongin, the shafts project through the surface.Smaller, club-like tylostyles also present in the basal layer.Cells with granular inclusions abundant throughout choanosome.
Ectosomal skeleton slender accessory oxea form bouquets around the projecting tylostyles.Asters form a dense layer at surface.
Reproduction the presence of oocytes has been observed in several specimens collected at depths between 20-30 m during June to August.
Slime produces copious amounts of slime on collection.Remarks: This species may be synonymous with P. constellata (see notes on P. constellata above).The main difference is the presence of short, stout tylostyles in P. loennbergi, however these are relatively scarce and might have been missed by Topsent.In P. loennbergi the larger tylostyles frequently have a second or third tylote swelling near the base (Fig. 2E, F).This has not been observed in P. constellata but does occur in Halicnemia patera.
Colour pale yellow-cream.
Choanosomal skeleton composed of irregularly arranged large oxeas and asters.
Ectosomal skeleton typical 'raspailiid surface architecture' whereby bundles of smaller oxeas surround large oxea, these in turn support a thick layer of asters.
Microscleres are oxyasters without a centrum, with 4-12 tapering rays, rays have slight annulations.Asters are typically 20-40 µm however where the rays are reduced in number they are generally much larger, up to 60 µm (Fig. 4E).
Reproduction the presence of oocytes were noted in one specimen collected from Cap Morgiou, Mediterranean Sea, 25.05.1992(J.Vacelet pers.comm.).
Slime produces copious amounts of slime on collection.Description: Outer morphology thinly encrusting with a hispid surface (Fig. 5E).
Ectosomal skeleton bundles of centrotylote oxeas penetrate the surface giving it its hispid appearance, oxyasters are common in the surface layer (Fig. 5F).

Slime produces copious amounts of slime on collection.
Habitat: Vertical to overhanging sublittoral rocky reefs with strong tidal currents.
Etymology: Aurantiaca L. = orange-coloured, refers to the yellow-orange colour of this species.Description: Outer morphology thinly encrusting with a hispid surface (Fig. 6E).
Choanosomal skeleton hymedesmoid arrangement consisiting of erect, long tylostyles and ascending bundles of centrotylote oxeas scattered throughout the skeleton (Fig. 6F).Cells with granular inclusions are abundant throughout the choanosome.
Ectosomal skeleton bundles of centrotylote oxeas penetrate the surface giving it its hispid appearance, oxyasters are common in the surface layer.

Slime Produces slime on collection.
Habitat: Vertical to overhanging sublittoral rocky reef with strong tidal currents.
Etymology: From the Latin for toothed dentate, refers to the 'toothed' ends of the oxea.

DNA sequences:
The 18S sequence on GenBank, accession no.KC902076 if from the holotype.Description: Outer morphology massive, subspherical, holotype is approximately 7 cm in diameter.Paratype ZMBN 125736 (Fig. 7A) was approximately 10 x 15 cm, however only a fragment of this specimen was collected by the manned-submersible.The surface is covered in large conules, 1-4 mm in height (Fig. 7F).
Ectosomal skeleton tufts of smaller oxeas are present in the surface layer.Surface conules are dense with asters (Fig. 7F & G).
Accessory oxeas are rare, bent (usually off-centre) and occasionally centrotyle (Fig. 7D).They are scattered in loose bundles close to the surface but do not appear to form surface bouquets around the principal oxeas.They measure 353-446-520 x 3-4-5 µm.

Slime produces copious amounts of slime on collection.
Etymology: This species is named in honour of the sponge biologist and microbiologist Friederike Hoffmann, for her role in making the investigation of the sponge fauna, a vital part of the Polarstern ARK-XXII expedition in 2007.Remarks: This species is similar in spicule morphology to P. duplex and P. lalori (see remarks above regarding P. lalori sp.nov.), but can be distinguished from it by the absence of tylostyles, stylote spicules are rare, the presence of large, surface conules and by it's massive growth form.The ectosomal skeleton consists of ascending bundles of large oxea and scattered oxea whereas in P. duplex and P. lalori sp.nov. the oxea are without order.In P. hoffmannae sp.nov. the accessory oxea do not appear to form bouquets around the larger oxea.
(Fig. 8 A-G Description: Outer morphology the type specimen was growing on dead Desmophyllum pertusum, it is spherical in shape, approximately 2.5 cm in diameter with oscules arranged in a cluster on the upper surface of the sponge.The surface is slightly uneven due to the presence of minute conules (Fig. 8A).
Choanosomal skeleton very little spongin present giving a lax, friable texture, megascleres distributed without order, asters distributed throughout choanosome.
Ectosomal skeleton ectosomal oxeas form bouquets around central large oxea but do do not protrude much beyong the surface.(Fig. 8B).Asters from a dense ectosomal layer.
Colour pale yellow to cream in life and in alcohol.
Slime produces copious amounts of slime on collection.
Habitat: On a branch of dead Desmophyllum pertusum, 1500 m depth.
DNA sequences: From the holotype we sequenced CO1 Folmer fragment GenBank accession XXXX Etymology: This species is named in recognition of the kind assistance of Pierce Lalor of the Centre for Microscopy and Imaging, Department of Anatomy, NUIG.
Remarks: Paratimea lalori sp.nov. is similar to P. duplex and P. hoffmannae sp.nov., all three species have large oxeas as their principal spicules and large asters with rays of inequal lengths.Externally the three species are quite different, Topsent (1928) describes P. duplex as a 3 mm thick, greyish coloured cushion, P. hoffmannae sp.nov. is cream-white in colour, massive-subspherical and covered in large conules whilst P. lalori sp.nov. is pale yellow-cream in colour, spherical in shape with only very small conules.In P. duplex the accessory oxea are longer and more slender (360-770 µm x 7-9 µm) than those in P. lalori sp.nov.(278-335-422 x 9-10.5-12µm), and form obvious bouquets surrounding large oxea.In P. lalori sp.nov. the accessory oxea are also arranged in bouquets around the principal oxea however this is not so obvious as in P. duplex, since they only project a very short distance beyond the ectosomal surface (Fig. 8B).The accessory oxea in P. hoffmannae sp.nov.are long and much more slender than those of P. lalori sp.nov., they are usually bent in the middle region and sometimes there is a centrotylote swelling.They measure 353-446-520 x 3-4-5 µm.P. lalori sp.nov.and P. hoffmannae sp.nov.also differ from P. duplex in the make up of their principal spicules, in addition to large oxea, P. duplex also has tylostyles and styles whereas P. lalori sp.nov.and P. hoffmannae sp.nov.have only occasional stylote spicules.
PARATIMEA MOSAMBICENSIS  Description: Outer morphology thick cushion, approximately 12 x 10 cm across by 5 cm thick.The oscules are arranged on the upper surface.The surface is slightly hispid (Fig. 9C).
Choanosomal skeleton a disordered arrangement of large oxea and large asters (Fig. 9D).
Accessory oxeas are rare and are very variable in length, sometimes they have a centrotylote swelling although this is also variable.They are relatively long and thin measuring 666-900-1200 x 10-12-15 µm (Fig. 9E).
Microscleres are huge asters with markedly inequal lengthed rays.The rays are very long relative the centrum, the asters are 73-84-100 µm across and the centrums measure 10-14-18 µm.The number of rays is very variable, in the smaller asters the rays are more numerous (Fig. 9B).
Etymology: Mosambicensis L. = from Mozambique, this specimen was collected from the Mozambique Channel.
DNA sequences: From the holotype we sequenced 28S D3-D5 GenBank accession XXXX and CO1 Folmer fragment GenBank accession XXXX.
Remarks: This species can be distinguished from other cushion shaped to massive Paratimea species by the absence of surface conules; its distinctive asters which at 73-100 µm, are amongst the largest found so far in Paratimea (see Supporting Information, Figure S7 for comparison of asters in Paratimea).The texture is much firmer relative to other Paratimea species which tend to have a very lax skeleton.
There is only 1 base pair difference of the CO1 Folmer fragment between the holotype of P. mosambicensis sp.nov.and that of P. hoffmannae sp.nov., however the external morphology, colour and aster size and morphology are all different.

DESCRIPTION
Description: Outer morphology globular, approximately 4 cm in diameter with oscules and radiating oscular channels on the upper surface (Fig. 10A).
Choanosomal skeleton large oxeas scattered without order, brown pigment bodies are common throughout choanosome.Asters are abundant.The skeleton is lax, with only small quantities of spongin.
Ectosomal skeleton composed of a dense layer of asters, accessory oxeas surround principal oxea however which penetrate the surface however this is not very obvious (Fig. 10B).

Slime produces copious amounts of slime on collection.
Etymology: Rosacea L. = rose-coloured, refers to the pink colour of this species.
Remarks: This species is distinguished from other massive Paratimea species by its pink colouration and asters with more or less equal lengthed rays.Although the asters are similar to those found in encrusting Paratimea species (e.g.P. constellata, P. loennbergi, P. aurantiaca sp.nov.) it can be distinguished from them by its globular growth form; having oxeas as the principal spicules as opposed to tylostyles; choanosomal spicules with a disordered arrangement rather than hymedesmoid and relatively large asters.
Accessory oxeas centrotylote (Fig. 11D), usually with one central swelling although two equidistant of centre swellings is also common.Occasional oxeas have 3 swellings and some have no swellings, only a V-shaped bend in the middle, reminiscent of a hairpin (Fig. 11E).They measure 1350-1620-1930 x 6-10.4-12µm.

Description: (BELUM Mc5427)
Outer morphology the neotype is a small encrusting sponge, approximately 3 x 4 cm in area by 2-3 mm in thickness.The surface is covered in conules (Fig. 12A).
Colour yellowish-orange in colour.
Reproduction the presence of oocytes has been observed in several specimens collected between 20-30 m during June to August.
Remarks: Topsent (1897) synonymised H. gallica with H. patera, however our reexamination of the type material of H. patera showed significant differences to the description given by Topsent (1893) for H. gallica.The type specimen of H. patera is discshaped and free-living (Fig. 11A) and was dredged from deep-water (330 m) at Shetland.By contrast Topsent's description of H. gallica was based on thickly encrusting specimens collected from shallow-water from the Roscoff area of the Celtic Sea and the Banyuls area of the Mediterranean Sea.In addition to the differences in habitat and growth form, Topsent (1897) also noted that the spiculation of the specimens from the French coast was less robust than that of the Shetland specimen and that there were also some differences in spicule morphology and skeletal architecture between the samples.Topsent (1897) considered the disc-shaped specimens from Shetland as a local form which was perhaps linked to the nature of the sea-bed.It is possible that Topsent's description of H. gallica was based on two different species.Topsent (1897) mentions how the colour of the specimens varies from yellow to orange-red.He linked the colour to the spherulous cells, specimens with large, uncoloured spherulous cells were yellow, specimens with smaller dark redcoloured spherulous cells were orange-red.Topsent also mentions that among the tylostyles there are short tylostyles similar to those of the Shetland specimens but that their presence is not constant.We have collected two species of Halicnemia from around the southwest coasts of Britain and Ireland, one is orange coloured and the second, Halicnemia Mc4307 sp.nov. is bright yellow and has short tylostyles in the base that are similar to those in H. patera from Shetland.Our orange Halicnemia matches the original description of H. gallica.Having compared numerous specimens of this orange encrusting Halicnemia with the type material of H. patera we consider the two sufficiently different and resurrect H. gallica for this species.All type material of H. gallica is believed to have been discarded from the Biological Oceanography Laboratory of Banyuls and no material remains in the porifera collection at the MNHN (Paris).In the absence of type material for H. gallica and the confusion in Topsent's description mentioned above, to define the nominal taxon objectively we designate a neotype, BELUM Mc5427: Huw's Reef, North Pembrokeshire: 51°57.8449'N 5°07.5460'W, depth: 17.4 m, 04.08.2009 coll.B. E. Picton (Fig. 12A) from the Porifera collection at the Ulster Museum, Belfast.Records of H. patera by Descatoire (1966), van Soest (1987), Ackers, Moss & Picton (1992), Picton et al. (2007), Morrow et al. (2012) should be reattributed to H. gallica.Higginsia coralloides var.arcuata Higgins, 1877 from Ireland may be synonymous with Halicnemia gallica however the type material, which had been deposited at the Museum of Liverpool, was destroyed in the 1941 Blitz.The description and illustrations are not sufficient for confirmation.
Microscleres the acanthoxea are different to the acanthoxea in other Halicnemia species.Dendy (1922) described them as 'trichites' they measure 49-80-119 µm (Fig. 13B).It is not possible to say with any certainty whether this species belongs in Halicnemia or elsewhere in the classification, for now we retain it in Halicnemia.
Reproduction the presence of oocytes were observed in BELUM Mc6786, collected in june from the channel Islands, 36 m.
Slime produces copious amounts of slime on collection.Remarks: On Hymeraphia verticillata, Bowerbank (1866, p.146) states; "this species differs from other British Hymeraphia species in having the primary skeleton spicules surrounded by fascicules of secondary skeleton spicules."He also states that it differs in the spicules of the dermal and interstitial membranes which have verticillate spines which is not known to occur in any other British sponge.Bowerbank does not justify his allocation of verticillata to Hymeraphia.
Topsent (1897) attempts to explain the confused taxonomic affinities between Hymeraphia verticillata and Halicnemia patera and comments on how Bowerbank (1866) overlooked the similarities between the two species described by him.Morrow et al. (2012) reassigned H. verticillata to Halicnemia.Using 28S rRNA they showed that verticillata was more closely related to the genus Halicnemia than to Hymeraphia which was also in their trees.BMNH 1406.70.5.3.21(slide) was examined, it was identified as Hymeraphia verticillata from Marquesas, Florida (Schmidt, 1870).The megascleres are oxeas with a double bend with fine microspining at the tips, they measure 515 x 5 µm (Fig. 14E & F).Centrotylote oxeas were observed but unmeasurable.Acanthoxeas (Fig. 14E & F) were very large and robust 558 x 40 µm, very acanthose, spines not arranged in verticils as in H. verticillata.The oxeas in this specimen (with microspined tips and a double bend) are similar to those of Heteroxya corticata Topsent, 1898.The spicules of this specimen are very different to those of Halicnemia verticillata and this specimen should be considered as an undescribed species, perhaps of Heteroxya.
In addition to tylostyles, verticillate acanthoxea and centrotylote oxea with fissurate ends, Topsent (1928;pl VI fig. 16, station 3144, Azores, 919 m), also illustrates oxyaster microscleres.Topsent considered the presence of asters in verticillata as further evidence of a close relationship with Paratimea.In the specimens collected by SCUBA diving from relatively shallow-water (38 m) the asters were never found, however they were present in the deep-water specimen (BELUM Mc2018.4) from the Whittard Canyon area.On the 28S D1-D2 tree (Supporting Information, Fig. S5) we can see that the shallow-water specimen which lacks asters is identical to the deep-water specimen with asters.This marker contains a region that usually shows variation between closely related species (see Morrow et al. 2012), therefore it seems likely that we are dealing with a single species.Uriz & Maldonado (1995) and Maldonado et al. (1999) demonstrated experimentally that there was a link between the spicule content of the sponge Crambe crambe and the silica concentration in seawater, while Cárdenas & Rapp (2013) observed this with Geodiidae in the environment.It is possible that the asterose microscleres in H. verticillata are only expressed in deeper water where the concentration of silica is relatively high.
Choanosomal skeleton hymedesmoid arrangement consisiting of erect, long tylostyles and bundles of long, slender, centrotylote oxeas scattered throughout the skeleton.Smaller, club-like tylostyles present in basal layer.Cells with granular content are abundant throughout the choanaosomal tissue (Fig. 15).
Ectosomal skeleton large tylostyles penetrate the surface, surrounded by supporting bundles of centrotylote oxeas.Acanthoxeas form a dense paratangential layer beneath the surface (Fig. 15F).
Reproduction the presence of oocytes has been observed in several specimens collected between June and August.
Slime produces copious amounts of slime on collection.
Habitat: Vertical to overhanging sublittoral rocky reef with strong tidal currents.
Remarks: Van Soest (2002) states that the spined microscleres can be curved and that raphides or trichodragmata may occur.He was referring to Higginsia lunata Carter, 1885. Van Soest & Hooper (2005) subsequently resurrected the genus Desmoxya for this species (see remarks on Desmoxya).
In this species the centrotylote swelling on the oxea is only slightly noticeable.The oxea are often bifurcate at one or both ends (Fig. 13C).Flanged ends to the oxea are also found in Higginsia bidentifera, Halicnemia verticillata and Paratimea dentata sp.nov.(BELUM Mc 6884).
Remarks: This species is characterised by the possession of relatively short and thick styles and acanthoxea (Fig. 13D).The styles are approximately 740 x 36 µm, occasionally short thick oxea of similar proportions are present.The acanthoxea are 42-58 µm.
Remarks: MNHN DT884 was not mentioned in the original description.The acanthoxea in this specimen are shorter and more robust, with sparser spination than the type material.
Remarks: Bergquist (1970) proposed Acanthoclada for sponges with a Higginsia-like skeleton but with the addition of echinating rhabdostyles.In addition to centrotrangulate oxea, Acanthoclada also has cladotoxa and curved birotule microscleres.Bergquist had misgivings regarding the allocation of Acanthoclada to Desmoxyidae due to the absence of oxeote microscleres.Hooper (2002a) retained Acanthoclada in Desmoxyidae with reservation, stating, "being most similar to Higginsia based largely on their affinities in skeletal structure, whereas this assignment is still not certain."Our 28S genetree (Supporting Information, Fig. S3) strongly supports a close relationship between Acanthoclada, Halicnemia and Higginsia.

HETEROXYIDAE DENDY, 1905
Diagnosis: Encrusting to massive growth forms.Surface hispid, sinuous or straight canals or grooves may be present.Choanosome consisting of (acanth)oxea either loosely scattered or forming a confused reticulation.Ectosomal skeleton consisting of dense brushes of (acanth)oxea perpendicular to surface.Megascleres, two size classes of smooth or spined oxea, some of the oxea have a characteristic double flex, occasionally styles present.Microscleres when present consist of raphides in trichodragmata in one or more size categories, larger raphides sinuous or curved.

Remarks:
The family Heteroxyidae was originally proposed for Heteroxya and Acanthoxifer Dendy, 1905[= Myrmekioderma Bergquist (1965) compared the type species of Myrmekioderma with the type species of Acanthoxifer and concluded that they were conspecific].Erpenbeck, Breeuwer & van Soest (2005) using partial 28S rRNA sequences showed Myrmekioderma granulatum (Esper, 1794) (type species of Myrmekioderma) clustering with Didiscus spp. in Raspailiidae.Redmond et al. (2013) using 18S rRNA showed M. granulatum clustering with Raspailiidae but M. rea clustered with Axinellidae suggesting that the genus is polyphyletic.Erpenbeck et al. (2012) using CO1 barcoding sequences showed M. granulatum and M. gyroderma clustering with Axinellidae and not with Didiscus in Raspailiidae.In our CO1 tree (Supporting Information, Fig. S4) we have used the Myrmekioderma sequences from Erpenbeck et al. (2012) and they cluster with Heteroxya corticata Topsent, 1898 and H. beauforti sp.nov., close to Axinellidae.Pending further investigation of the possible polyphyly of Myrmekioderma we retain the genus in Heteroxyidae.
Figure 1 provides strong molecular evidence for the exclusion of Didiscus and Desmoxya from Heteroxyidae.Whilst previous molecular studies have consistently shown Didiscus clustering within Raspailiidae this is the first study that provides molecular and morphological support for the allocation of Desmoxya (and Desmoxyidae) to Poecilosclerida.
HETEROXYA TOPSENT, 1898 (P. 231) Diagnosis emended: Encrusting growth form; surface highly hispid; choanosome with a condensed basal layer of spongin lying on the substrate, containing (acanth)oxeas distributed without appreciable order on basal spongin and strewn throughout the mesohyl; subectosomal skeleton consists of oxeas arranged perpendicular to the ectosome, protruding through the surface, but not embedded in basal spongin; ectosomal skeleton with a perpendicular palisade of smaller (acanth)oxeas, through which the larger subectosomal oxeas protrude; very long styles present in one species.Two categories of oxea, smooth or spined; larger oxea sometimes with a double flex; microscleres absent.DNA sequences: The CO1 Folmer fragment on GenBank, accession KP939318 is from the holotype.
Remarks: Fig. 14A is a photograph of an ethanol preserved specimen from the type lot.It is dirty-beige in colour with a hispid surface and is growing on coral.Fig. 16B is a photomicrograph showing the large oxea that often have a characteristic double flex.The large oxea measure 1600-1700-2000 x 26-32-37 µm.The ends of the oxea occasionally have some microspination (Fig. 16C).The small oxea (Fig. 16D) measure 235-310-420 x 12-23 µm.The spination is more pronounced in the smaller oxea, particularly towards the ends of the oxea but some spination may also be present over the entire surface of the spicule.

Colour dirty white in alcohol.
Ectosomal skeleton composed of a layer of large and small oxea arranged in tufts, perpendicular to the surface, the larger oxea penetrate the surface for most of their length giving surface its hispid appearance (Fig. 17B).
Choanosomal skeleton relatively aspicular, consists of scattered oxea and very long styles up to 6 mm, arranged perpendicular to the base of the sponge.The styles penetrate the surface for most of their length and are responsible for the hirsute appearance of the surface (Fig. 17B-C).Cells with granular content (Fig. 17B) are particularly abundant in the basal region of the choanosome.
Megascleres consist of two size classes of oxea (Fig. 17D) and very long styles (Fig. 17E).The larger oxea (Fig. 17D) are slightly bent in the middle region and taper to a smooth point at each end.They measure 622-1030-1385 x 10-16-21 µm.The smaller oxea (Fig. 17F) frequently have a double flex and taper to a fine point at each end, they measure 207-280-370 x 11-14-16 µm.Both categories of oxea are entirely smooth.The styles (Fig. 17E) are very long and slender and are parallel sided for most of their length.They measure 5000-5650-6300 x 23-25-27 µm.
Reproduction: Oocytes were observed in the holotype which was collected in May.
Habitat: Encrusting dead coral in the deep cold water coral reefs off the west coast of Ireland between depths of approximately 600-1500 m.
Etymology: Named after the Beaufort Marine Biodiscovery Research Award which provided funding for part of this research and in honour of the Irish hydrographer Francis Beaufort.
DNA sequences: We sequenced the CO1 Folmer fragment from the holotype, GenBank accession no.KF017197.
Remarks: Topsent (1898) established the genus Heteroxya for his new species H. corticata from deepwater (approximately 1200 m) off the Azores.Van Soest et al. (2007) recorded H. corticata from SE Rockall Bank (off the west coast of Ireland) at a depth of 629 m.We have re-examined the specimen from Rockall (ZMA POR20081) and it appears to agree with our new species H. beauforti.The new species differs from H. corticata in the following characteristics; both the large and small oxeas in H. beauforti are smooth whereas in H. corticata the smaller oxeas (Fig. 16D) are covered in spines and some of the large oxeas have scattered spines near the apices (Fig. 16C).The large oxeas in H. corticata measuring 1600-1700-2000 x 26-32-37 µm are longer and more robust than the large oxeas in H. beauforti (622-1030-1385 x 10-16-21 µm).In addition to oxeas the new species also has very long slender styles that measure up to 6.3 mm.The CO1 Folmer fragment from the holotype of H. corticata (GenBank KP939318.1) is identical to the sequence from H. beauforti.Whilst this marker is frequently used in species barcoding (Herbert et al., 2003), it has been demonstrated that it is highly conserved in sponges and cnidarians and not always suitable for species delimitation in these groups.Our CO1 tree (Supporting Information, Fig. S4) shows Heteroxya clustering close to Axinellidae.
Remarks: Although this species was transferred to Higginsia (van Soest et al., 2018 in WPD), we do not know of good morphological evidence to support placing it here.This species has shared affinities with Axinellidae, Hymerhabdiidae and Dictyonellidae.On spicule and skeletal morphology these three families are apparently indistinguishable.There is some evidence that there are chemical characters that characterise Hymerhabdiidae such as the presence of isocyanids (Braekman et al. 1992) and using DNA sequences they are easily distinguishable.Given that it is not possible to say with confidence whether durissima should be assigned to Axinellidae, Dictyonellidae or Hymerhabdiidae, and in the absence of molecular data, we return it to its original genus (Bubaris) the most conservative option.
Type species: Desmoxya lunata (Carter, 1885) Included species: D. lunata (Carter, 1885); D. pelagiae van Soest & Hooper, 2005Hallmann, 1917 speculated that the microscleres in Desmoxya lunata, which are terminally spined, crescent shaped or sigmoidal microxea, may have been derived from sigmata.He questioned whether genera with spined microxea such as Higginsia, Halicnemia and Desmoxya should be included in Axinellidae.He further speculated that the acanthoxea in Desmoxya and Halicnemia were similar to those of Acanthoxa Hentschel, 1914 and could be homologous with the acanthoscleres of myxillids.Our 18S tree (Supporting Information, Fig. S2) shows that there is molecular support for Hallmann's hypothesis (in part) as Desmoxya pelagiae clusters in Poecilosclerida; however Halicnemia and Higginsia group with Stelligera and Paratimea in Stelligeridae: Axinellida.
Remarks: A re-examination of the type specimen of D. lunata using SEM revealed that the raphides are acanthose (Fig. 19D-G), suggesting that it may have affinities with Tedaniidae.The raphides are unusual, they appear to occur in at least three length categories and are 'compound'.They consist of several slightly overlapping raphides glued together, whereas in D. pelagiae the raphides are spined mostly at the ends and are in trichodragmas with individual raphides single, not compound.The acanthoxea in D. lunata are diactinal (Fig. 19C) whereas in D. pelagiae they are monactinal.In other respects the spicules and skeletal arrangement are closest to those of D. pelagiae.We therefore take D. pelagiae as representative of Desmoxya and propose the transfer of Desmoxya from Heteroxyidae: Axinellida to Desmoxyidae: Poecilosclerida.VAN SOEST & HOOPER, 2005 (Fig. 20ª-H) Desmoxya pelagiae van Soest & Hooper, 2005(pp. 1368-1370, fig. 2ª-H).
Remarks: A re-examination of the type specimen of D. pelagiae found acanthostyle spicules and unidirectional spines on the raphides (similar to onychaetes) that were overlooked in the original description (Fig. 20 is an SEM of the spicules from the type specimen).The acanthostyles are relatively scarce and confined to the basal region of the sponge.They measure 288-314-353 x 9-10 µm n = 6 (measurements taken at middle portion of shaft), the head of the acanthostyles are approximately 12 µm across.
Remarks: This genus is erected for Higginsia anfractuosa.In assigning this species to Higginsia Hooper & Lévi (1993) emphasized the presence of acanthoxea and raphides of two size classes.They stated that morphologically it was most similar to Higginsia lunata Carter, 1885 in growth form, papillose surface features and skeletal architecture and that both species are atypical of other Higginsia.Re-examination of the holotypes of D. lunata and D. pelagiae using SEM (Figs. 19 & 20 respectively) showed that the 'raphides' were spined, similar to the onychaetes in Tedaniidae.Re-examination of the holotype of H. anfractuosa showed that the 'acanthoxea' described by Hooper & Lévi (1993) were slender rugose oxea unlike the crescent-shaped acanthoxea in Desmoxya.Thin spicules resembling raphides were present but they were approximately the same sizes as the strongyles and oxea and we would interpret them as developing spicules rather than raphides.
Morphologically we consider Hooperia anfractuosa comb.nov.most similar to Hemimycale.Both Hemimycale and Hooperia gen.nov.share conspicuous, raised areolated porefields supported by parallel columns of erect strongyles (Fig. 21C & F respectively).The choanosomal skeleton in Hemimycale and Hooperia consists of bundles of strongyles that branch and anastomose infrequently (Fig. 21A & D respectively).Hooperia gen.nov.can be distinguised from other hymedesmiid genera by the presence of a surface layer of rugose oxea (Fig. 22C) that are morphologically distinct from the acanthoxea in Crellidae.
Etymology: This genus is named in honour of Dr John Hooper, Head of Biodiversity & Geosciences Program, Queensland Museum, for his major contribution to sponge taxonomy.
Remarks: We examined the Tanzanian specimen (OCDN3725-J) that is included in our molecular trees (Supporting Information Figs S2, S3).The spicule morphology, particularly that of the acanthoxeas is substantially different to H. anfractuosa comb nov.and we do not consider it to be the same species.
Ectosomal skeleton surface layer of straight acanthoxeas arranged tangentially.
Remarks: This species appears to be a poecilosclerid, morphologically it is similar to Spanioplon armaturum (Fig. 23A & B), both share style/tylostyle megascleres and tornotes with basal ends rounded and slightly dilated.The 'acanthoxeas' in Higginsia fragilis are also similar to those of S. armaturum (which vary from acanthoxeas to acanthostyles) and are unlike the acanthoxeas in other Higginsia and Halicnemia species.On the basis of the morphology we propose the transfer of Higginsia fragilis to Spanioplon (Hymedesmiidae: Poecilosclerida).

ADREUS GRAY, 1867
Emended diagnosis: Arborescent growth form; choanosomal skeleton with strongly compressed axis composed of long smooth tylostyles in bundles running longitudinally through branches, poorly developed extra-axial skeleton composed of sparse plumose brushes of smaller smooth styles ascending to periphery; euasters with curved or sinuous, smooth or spined, strongylote or tylote rays often branched, mainly confined to the ectosomal region, absent in some species.

CAPTIONS
Table 1.A summary of the different classifications espoused by the main authors who wrote on Halicnemia and Paratimea.With the exception of Hooper & van Soest (2002), these studies were not complete taxonomic reviews but rather faunas of a particular region and therefore they only include the genera that were represented in their chosen area.* indicates supporting molecular data, **molecular data does not support the allocation of Plenaster to Stelligeridae (taxonomic affinities will be discussed in a separate manuscript).
using partial 28S rRNA were the first authors to show Didiscus spp.clustering with Myrmekioderma granulatum within Raspailiidae.Erpenbeck et al. (2012a) using CO1 barcoding sequences again showed Didiscus grouping within Raspailiidae, however their Myrmekioderma granulatum sequences clustered with Axinellidae and not with Didiscus.Morphologically Myrmekioderma and Didiscus share similar skeletal architectures, and a distinctive surface crust with sculptured grooves and subdermal drainage canals.The results of Redmond et al. (2013), using complete 18S rRNA, indicate that Myrmekioderma is polyphyletic with M. granulatum clustering within Raspailiidae and M. rea with Axinellidae.From the molecular data it is clear that Didiscus belongs in Raspailiidae however the potential polyphyly of Myrmekioderma needs further investigation.
DNA sequences: We sequenced 28S D1-D2 from the holotype and from BELUM Mc 2018.3,GenBank accessions XXXX and XXXX respectively and CO1 Folmer fragment GenBank accessions KC869429 and XXXX respectively.

Figure 3 .
Figure 3. A, drawing of the spicules in Paratimea duplex, reproduced from Topsent, 1928 Pl. 6 Fig. 21; B, photomicrograph of spicules from holotype showing combination of oxea and styles; C, photomicrograph showing large oxea with centrotylote swelling and large asters with inequal length rays; D, photomicrograph of transverse section of sponge showing accessory oxeas arranged in bouquets around large oxea and dense surface layer of asters.

Figure 5 .
Figure 5. Paratimea aurantiaca sp.nov.holotype, A-D, SEM's of spicules: A, long tylostyle; B, accessory oxea; C, centrotlyote swelling on accessory oxea; D, aster; E, in situ photograph; F, photomicrograph of transverse section of sponge showing hymedesmoid skeleton consisting of erect, long tylostyles and ascending bundles of centrotylote oxea scattered throughout the skeleton.Bundles of centrotylote oxea penetrate the surface, oxyasters are common in the surface layer.Cells with a granular content are abundant.

Figure 7 .
Figure 7. Paratimea hoffmannae sp.nov.A, Paratype ZMBN 125736 in situ, screenshot of footage taken with the manned-submersible JAGO; B, photomicrograph of transverse section of holotype showing ascending bundles of large oxeas and scattered oxeas and asters.Surface conules are dense with asters; C-E, SEM's of spicules: C, principal oxea; D, accessory oxeas; E, asters with inequal length rays; F, elongate conule dense with asters; F & G, tufts of smaller oxea penetrate the surface layer of the conules.

Figure 25
Figure 25.A, photomicrograph of transverse section of Stelligera stuposa BELUM Mc4330 showing oocytes surrounded by cells with granular inclusions (collected 30.06.2008, 25 m, Loch Sunart).B, photomicrograph of transverse section of Paratimea loennbergi BELUM Mc4323 showing mature oocytes (eggs) surrounded by cells with granular inclusions (collected 20.06.2008, 25 m, Firth of Lorn, Scotland).C, photomicrograph of transverse section of Halicnemia gallica BELUM Mc6677 showing an oocyte surrounded by cells with granular inclusions (collected 23.06.2010, 23 m, Guilleaumesse, Channel Isles).D, photomicrograph of transverse section of Heteroxya beauforti sp.nov.BELUM Mc7794 showing an oocyte and cells with granular inclusions (cgi) (collected 31.05.2010,1300 m, S. off Ireland).Supporting Information Data S1.The taxa used in the analyses together with their catalogue numbers, collecting locality and corresponding GenBank accession codes.

Figure S2 .
Figure S2.Best tree output from RaxML analysis of full length 18S rRNA.Figures at nodes correspond to bootstrap support >70 followed by posterior probabilities >0.7 from the Bayesian analysis.

Figure S3 .
Figure S3.Best tree output from RaxML analysis of 28S rRNA (D3-D5 region).Figures at nodes correspond to bootstrap support >70 followed by posterior probabilities >0.7 from the Bayesian analysis.

Figure S4 .
Figure S4.Best tree output from RaxML analysis of mitochondrial CO1 barcoding fragment.Figures at nodes correspond to bootstrap support >70 followed by posterior probabilities >0.7 from the Bayesian analysis.

Figure S7 .
Figure S7.Comparison of aster morphologies and dimensions in Paratimea species.