Affi nities of the family Sollasellidae ( Porifera , Demospongiae ) . I . Morphological evidence

Comparison of Sollasella digitata Lendenfeld, 1888, up until the present assigned to its own family Sollasellidae Lendenfeld, 1887 in the order Hadromerida, and Raspailopsis cervicornis Burton, 1959, assigned to Raspailiidae Nardo, 1833 in the order Poecilosclerida, leads to the conclusion that both should be considered congeneric and are best assigned to a single genus Sollasella. This conclusion is based on examination of habit and skeletal characters of the type material of S. digitata and both type and freshly collected material of S. cervicornis. The conclusion is strengthened by the discovery of a new species, Sollasella moretonensis n.sp. collected in North Australia (primarily in the northeastern coast, but also an isolated record from the northwestern Australian coast), which possesses in addition to the characteristic surface pattern and skeletal structure, genuine echinating acanthostyles. The redefi ned genus Sollasella shares axial / extra-axial arrangement of the skeleton, special surface brushes of oxeas surrounding a single protruding style, and vestigial occurrence of acanthostyles with many Raspailia s.l. Nevertheless, it is retained as a separate genus, on account of its peculiar polygonal arrangement of surface pores. The distribution of the genus is disjunctive including both (southeast, northeast and northwest) Australian and Western Indian Ocean localities, but so far no intermediate records. Based on this morphological evidence, it is proposed – pending publication of corroborating molecular evidence to be presented in a follow-up study – to reassign Sollasella and the family Sollasellidae to the poecilosclerid family Raspailiidae.


Introduction
The recent update of the sponge classifi cation 'Systema Porifera' (Hooper and van Soest, 2002a) typifi ed the family Sollasellidae and its sole representative the East Australian species Sollasella digitata Lendenfeld, 1888 as 'incertae sedis' and 'poorly known ' (p. 170), and commented that 'its true affi nities remain to be established ' (p. 280).Nevertheless, it was assigned to the order Hadromerida on the grounds that it possessed a cortex and strongly radiating skeletal architecture.The species is only known from its original description and from an excellent re-description of the type material by Hallmann (1914).The species is arborescent and hispid.It has a strongly axial-radiate architecture, with a thick axial column made up of oxeas and styles, long protruding extra-axial styles and a dense ectosomal palisade of oxeas.Between this 'cortical' palisade and the axial column there is a region relatively lightly spiculated.The surface of S. digitata has a very characteristic polygonal pattern of 'perforations' visible to the naked eye.Previous authors attempting to classify this enigmatic species arrived at divergent conclusions: Axinellidae, order Halichondrida (cf.Hallmann, 1914), Coppatiidae, order Astrophorida, (cf.Hooper and Wiedenmayer, 1994), whereas Polymastiidae and Stylocordylidae, both order Hadromerida were named as closely related families by van Soest (2002).
Recently, we obtained freshly collected specimens of a branching hispid sponge from Oman waters showing the same characteristic surface perforation patterns and essentially a similar skeletal architecture and spiculation.This material was initially identifi ed as Raspailopsis cervicornis Burton, 1959 -a species that was previously described from Oman watersbased on the characters presented in the very brief description.Recent examination of type material in the Natural History Museum, London, confi rmed that Burton's species was identical to our Oman material, but also that R. cervicornis is indeed very similar to Sollasella digitata from southeast Australia.These observations led to the discovery of a further similar species from subtropical Australia, which shared the surface perforations but possessed in addition abundant echinating acanthostyles, confi rming the raspailiid affi nity.This paper is intended to present morphological evidence for the close similarities of Sollasella and Raspailia, and to discuss the classification of the family Sollasellidae whose position in Hadromerida is proposed to be untenable.We will follow this up with further corroborating evidence from molecular sequence data (Erpenbeck et al. submitted).

Material and methods
Material was examined from the collections of the Natural History Museum, London (BMNH), the Australian Museum Sydney (AM), the Queensland Museum Brisbane (QM) and the Zoological Museum of the University of Amsterdam (ZMA).Details of the specimens are given below with the treatment of the species.Thick sections and spicule mounts for light and SEM microscopy were made following the usual methodology (cf.de Voogd and van Soest, 2002).

Ordinal and familial affi liations of Sollasella digitata
Hadromerida are defi ned (cf.Hooper and van Soest, 2002b) as comprising sponges with peripherally radiating skeleton built from tylostyles, styles, or occasionally oxeas, usually with a smaller category in an ectosomal palisade.The radiating architecture becomes confused or plumose in the choanosome, where the larger spicule categories predominate.Microscleres (euasters, spirasters, amphiasters) occur in many families, usually as an ectosomal cover, but are absent in several families.Family Sollasellidae and Sollasella digitata were hesitatingly assigned (cf.van Soest, 2002) to Hadromerida for want of a better placement.Shared characters with several Hadromerida are the arborescent shape, ectosomal palisade, confused choanosomal skeleton, robust oxeas and styles with faint tyle, but these features are widely distributed throughout the Demospongiae, not comprising clear synapomorphies with any particular order or family.The polygonally arranged surface pores are slightly reminiscent of the polygonal surface plates separated by grooves found in the family Placospongiidae, but they appear distinctly unique.Finally, the hispidity is shared with many Hadromerid genera, but can hardly be suffi cient for ordinal identity.
Compared to Hadromerida, affi nities of Sollasella digitata with the order Poecilosclerida also seem tenuous.This order is defi ned (Hooper and van Soest, 2002c) as possessing a skeleton of spicules and spongin, with distinct regionalization into ectosomal and choanosomal components.Among the spicules feature meniscoid microscleres (sigmas, chelae, toxas) and acanthose styles as characteristic.However, some families lack all or some of these.Sollasella digitata -in retrospect -shares several features with Poecilosclerid representatives, such as a distinct regionalization into ectosomal, extra-axial and axial skeletal arrangement, surface brushes of smaller megascleres surrounding single protruding long megascleres ('raspailid ectosome') and confused axial-longitudinal skeletal arrangement.Arborescent shapes with distinct hispidity like Sollasella digitata are very common features in microcionid Poecilosclerida.
The recent rediscovery of fresh material of South Arabian Raspailopsis cervicornis Burton, 1959 (see above and below), with its essentially similar surface pattern and structure, and an undescribed species from subtropical northeastern Australia 'Sp.1245' (see below) with the same surface pattern but in addition clear possession of acanthostyles, swung the balance defi nitely towards Poecilosclerida -Raspailiidae.Table 1 summarizes the different character states of Sollasella and representative families from Hadromerida and Poecilosclerida discussed above.Heuristic search with PAUP 3.1 (Swofford, 1993) using the characters unordered confi rmed the morphological affi nity of Sollasella with Raspailiidae Contributions to Zoology, 75 (3/4) -2006 Definition (emended).Raspailiidae with strong axial column of confusedly aligned oxeas and styles, and with extra-axial columns of short oxeas and long and short styles positioned at right angles to the axial column.At the surface there is a characteristic ornamentation of polygonally arranged inhalant (?) pores and the oxeas form a continuous palisade of brushes of oxeas pierced by long styles.A low proportion of short acanthostyles may be present.Lendenfeld, 1888Figs 1A-F, 2A Sollasella digitata Lendenfeld, 1888: 56;Hallmann, 1914: 287, pl.XV fi gs 1-2, text-fi g. 1.
Habit: Stalked ramose sponge (Figs.1A, B, 2A) of up to 14 cm long, with rounded short branches.Consistency very fi rm, tough.Surface hispid due to protruding spicules, and provided with a characteristic polygonal pattern of round pores (Fig. 1A), presumed to be inhalant openings by Hallmann.
(11 shared characters in Table 1), rather than with Suberitidae (7 shared characters) (tree not shown).A full phylogenetic reconstruction will be provided in a second contribution on this topic, incorporating molecular evidence.In the molecular sequence analyses Sollasella clearly clusters inside the Raspailiidae clade, and distant from hadromerid taxa (Erpenbeck et al, submitted).

Systematic descriptions
The three species discussed here are assigned to a redefi ned genus Sollasella, provisionally assigned to Raspailiidae, pending further support from molecular studies.They will all be diagnosed below and morphological evidence for their congeneric status will be illustrated with habit and microscopical images.

Phylum Porifera Class Demospongiae
Order Poecilosclerida Family Raspailiidae Nardo, 1833Subfamily Raspailiinae Nardo, 1833 Genus Sollasella Lendenfeld, 1887 Synonym.Raspailopsis Burton, 1959 Type species.Sollasella digitata Lendenfeld, 1888 (by monotypy) Larger openings, presumably oscules up to 2 mm in diameter, few in number.Skeleton: In cross section the branches (Fig. 1B) show three distinct regions: an axially condensed region, an open subcortical region of low spicular density, and an outer cortical region.The axial region consists of confusedly strewn spicules and anastomosing spicule bundles bound by considerable amount of spongin, the bundles consist of a mixture of styles and oxeas.The subcortical region is traversed by spicule bundles with orientation perpendicular to the axial core, and these consist of a single central style and a mass of surrounding oxeas.The styles protrude to some extent beyond the surface.The cortical skeleton is supported by the brushed endings of the subcortical spicule bundles.Choanocyte chambers about 25 μm in diameter, confi ned to the axial core.Cortex densely fi brous.
Distribution and ecology.Port Jackson, Manly Beach, E coast of Australia, no further data.

Description
Habit (Figs. 2B-F): Thickly arborescent sponges, branching dichotomously.Branches have a tendency to coalesce and the Burton material contains one 'caliculate' specimen (from Zanzibar, BMNH 1936.3.4.466).Surface hispid through protruding spicules.Colour dark brown in life (Fig. 2C), paler grey in alcohol (Figs.2B,F).Tough consistency, easily broken.Branch endings rounded and slightly swollen.Holdfast broader than stem, evenly rounded, spreading out equally.Height up to 12 cm, branch diameter 1-1.2 cm, specimens retained as ZMA 17450 are typically 5 or 6 cm high.A striking feature of all branches is the polygonal pattern of surface pores (Figs.2B, D, F).Upon collection, when lifted out of the water and with slight pressure, these openings emitted tiny water jets, indicating they may be exhalant contrary to what is assumed here.Polygonal areas are elongate, on average 4 × 2 mm (Fig. 2D), and uniformly spread over the branches (Fig. 2F).No separate larger oscules have been observed.
Skeleton (Figs. 2E, 3A): In cross section the branches contain three distinct regions recognized by eye (Fig. 2E) and differently coloured in life: a thick axial column, reddish orange in life, pale orange when preserved, a relatively fi brous extra axial region, and a beige coloured cortical layer at the sur-face.Long single spicules protrude through the cortex.The axial column is made up of a mixture of all spicules, but predominantly of small styles and oxeas.The extra-axial region (Fig. 3A) consists of thick cone-shaped columns of 2 mm × 1-3 mm positioned at right angles to the axial region and fanning out to the surface region.In between these columns there are fi brous areas, 1-3 mm in size with low spicular density.The cortical region is approx.1-1.5 mm thick and consists of a dense palisade of oxeas arranged in tight brushes.Burton described this cortex as typical Raspailia-like brushes surrounding a central long style, but this was not clearly observed by us.The cortical palisade is indeed pierced from the extra-axial region by long styles which protrude considerably beyond the surface at regular distances making the sponge quite hispid, but a clear association with brushes surrounding a single style is not evident.
Remarks.This species differs from S. digitata mainly in the abundant presence of short styles (which are very rare and probably oxea-derivates in S. digitata) next to long styles and oxeas.The short styles may have spines (rarely), which fi ts the family Raspailiidae.The oxeas are signifi cantly thicker in S. cervicornis (up to 29 μm vs.only up to 16 μm in S. digitata).4A-D): Vasiform, lobate, fan-shaped, cup-shaped or subvasiform habits, 11-28 cm high, 8-27 cm wide at apex, with thick lamellae 0.5-2 cm thick, bearing rounded 'lumpy' or convoluted margins.Specimens are usually attached to the substrate by a short cylindrical woody holdfast, ranging from 3-10 cm long, 1.5-3 cm diameter.Texture: Harsh (slightly hispid), fi rm, stiff but compressible in life, more rigid when preserved.Surface: Dead specimens are covered with a polygonal plate-like pattern formed by shallow pits surrounding low conules, 3-5 mm in diameter, 1-2 mm high, with some (but not all) perforated by an oscule 2-5 mm in diameter sitting at the apex of the conule.When preserved most oscules and pits are collapsed, leaving only the polygonal pattern of conules reminiscent of typical Polymastia or massive Cliona surfaces (Fig. 4B-D), and conules appear to be distributed equally on both the inner and outer surfaces of the lamellae, extend-ing across all lamellae to the upper part of the holdfast.In life, however, it is apparent (Fig. 4A) that oscules occur mostly on the inner surface of lamellae, less commonly on the upper margins of the outer surface, are about 4 mm in diameter, and are surrounded by a large raised membranous lip, pigmented darker brown than the surrounding area.The outer surface of living specimens is highly pitted and rendered convoluted by pointed conules surrounding large pits, interconnected by prominent ridges of fl esh forming stellate patterns reminiscent of dictyodendrillid sponges.This external sculpturing of pits and conules presumably represents the inhalant aquiferous system.Colour: In life the colour is bright orange to orange-brown, fading to darkish brown upon preservation.
Skeleton (Figs. 5A-E): In cross section the skeleton appears to be differentiated into three distinct regions: an open reticulate very fi brous core, a more plumo-reticulate extra-axial region, and a distinct radial ectosomal skeleton.There is no obvious axial compression, but in the core of the lamellae there are heavily collagenous fi bres and multispicular bundles of large styles and oxeas, forming an open but highly reticulated skeleton.Fibres are moderately large, 180-260 μm in diameter, forming open meshes in some places over 1 mm in diameter.Bundles of oxeas and styles appear to run predominantly longitudinally through lamellae.At the core of the skeleton fi bres are sparsely echinated by acanthostyles.In the extra-axial region the skeleton becomes more plumoreticulate, with thicker, more densely compacted and heavy collagenous fi bres together with radial bundles of large choanosomal styles and oxeas,with the latter becoming more radial towards the periphery.The ectosomal skeleton has only occasional large oxeas or styles protruding from the subectosomal regions, with the remainder consisting of a nearly continuous, erect palisade of ectosomal oxeas.In some cases  these bundles of ectosomal oxeas appear to be surrounding the larger protruding subectosomal spicules, an apomorphy of the Raspailiidae, but this is rare.The mesohyl is sparse at the core of the skeleton but is more heavily invested in collagen towards the periphery.Spicules (Table 2, Figs.5F-H): The longer (subectosomal) spicules are predominantly styles, occasionally oxeas or anisoxeas, are not common, and appear to extend from the core of the skeleton to the surface, or close to it.They are completely smooth, more-or-less straight, and have either evenly rounded bases or abruptly pointed bases, and either sharp tapering points or taper to very fi ne slightly telescoped points, approximately 2500->4000 × 20-55 μm).The (choanosomal) oxeas are far more abundant and comprise most of the skeletal spiculation.They are usually slightly curved at the centre, sometimes straight, with sharp, tapering points, 660-940 × 14-20 μm).Ectosomal oxeas are straight, slightly curved at the centre or asymmetrically curved, with slightly tapering sharp points, 360-515 × 5-8 μm).Echinating acanthostyles are straight or slightly curved at the centre, have slightly subtylote bases and are slightly rounded or blunt at the apex, and have a light, evenly distributed spination along the length of the spicule, slightly heavier spines at the tip, and spines are short and conical, (95-165 × 3-5 μm).
Etymology.This species is named for its predominantly subtropical distribution in the vicinity of the Moreton Bay region, SE Queensland.Distribution and ecology.Moderately prevalent in southern Queensland inshore waters (Brisbane and Sunshine Coast regions), associated with muddy substrata at the base of coral reefs.The two isolated records in Far North Queensland (Shelburne Bay) and NW of Western Australia (Cape Jaubert) were from commercial and scientific trawls made in deeper muddy substrates, and suggest that the species may have a wider distribution than presently known, or a discontiguous distribution in tropical Australia given that there has been intensive sampling of areas in between without trace of this species.
Remarks.The new species S. moretonensis shares the polygonal arrangement of surface pores, possession of three distinct skeletal regions and geometry of spiculation with the other two species, but differs in possessing echinating acanthostyles, a differentiated category of ectosomal oxeas, a vasiform growth form, orange live colouration and specifi c dimensions of spicules.The discovery of this new species confi rms the allocation of Sollasella to the Raspailiidae in possessing of the important raspailiid apomorphies -viz.echinating acanthostyles and bundles of ectosomal oxeas surrounding protruding subectosomal styles/oxeas (albeit rare), in conjunction with a continuous ectosomal palisade common to the other two species.

Discussion
In the latest revision of the Raspailiidae (Hooper, 2002), Raspailopsis was assigned to the synonymy of Raspailia (Parasyringella).Since three distinct species with disjunctive distributions share a number of unique features not shared with other Raspailia species, there is suffi cient justifi cation to abandon Hooper's synonymy decision and unite them in a separate genus.Accordingly we propose to retain Sollasella as a distinct genus in Raspailiidae, differing from all other genera in this family by the polygonally arranged surface pores and the continuous cortical palisade of long, robust oxeas.
The family Sollasellidae Lendenfeld, 1887 was erected a year before the description of the type species, Sollasella digitata Lendenfeld, 1888.Despite this slight deviation from the nomenclatural rules, it was accepted as a valid family name by subsequent revisors (Hallmann, 1914;de Laubenfels, 1936;van Soest, 2002).In view of the fact that the type species Sollasella digitata is now proposed to be assigned to Raspailiidae, the family name Sollasellidae falls into synonymy with that family name.The 'Systema Porifera' (Hooper and van Soest, 2002) gives Hentschel, 1923 as the author and date of Raspailiidae, which would potentially jeopardize the well-established name Raspailiidae.However, this authorship and date is predated by Nardo, 1833 who already employed a taxon name Raspeliae of the family group level (corrected to Raspailiae in Nardo, 1847).Thus, authorship of Raspailiidae should be attributed to Nardo, 1833 and this name remains the oldest available name.Sollasellidae Lendenfeld, 1887 is a junior synonym of Raspaillidae Nardo, 1833.
The shared characters are more numerous: -polygonal arrangement of pores is a striking feature of both and a strong synapomorphy.There are sev-eral other sponge genera showing a polygonal surface pattern (e.g.Myrmekioderma and Didiscus in Halichondrida: Heteroxyidae, and Tethya and Pla-

Table 2 .
Comparison of spicule sizes of Sollasella species.