Demosponges from the Northern Mid-Atlantic Ridge shed more light on the diversity and biogeography of North Atlantic deep-sea sponges

In July–August 2004, the Mid-Atlantic Ridge Ecosystems (MAR-Eco) expedition collected Demospongiae (Porifera) from the Northern Mid-Atlantic Ridge (MAR) north of the Azores, between 41°N and 61°N. Demosponges were found at 10 stations, at depths ranging from 753 to 3046 m. Twenty-two species were identified: 17 Tetractinellida, one Polymastiida, one Suberitida, two Poecilosclerida and one Dendroceratida. The study of this material is an opportunity to revise the taxonomy and the North Atlantic distribution of each of these deep-sea species. Some species are particularly rare and poorly known (Tetilla longipilis, Tetilla sandalina, Craniella azorica, Polymastia corticata) and two are new to science: Forcepia (Forcepia) toxafera sp. nov. and Iotroata paravaridens sp. nov. This study suggests that the MAR is not a longitudinal barrier for the dispersal of deep-sea demosponges while on the contrary, the Charlie-Gibbs Fracture Zone (CGFZ) may be a latitudinal border for the dispersal of deep-sea demosponges, due to great depths and currents.


I N T R O D U C T I O N
The Mid-Atlantic Ridge (MAR) between Iceland (638N) and the Azores (36 -398N) spans across 3000 km. About mid-way between Iceland and the Azores (around 528N), the Charlie-Gibbs Fracture Zone (CGFZ) offsets the MAR about 300 km, with depths reaching 4500 m (Felley et al., 2008), thereby opening the deepest connection between North-East and North-West Atlantic waters ( Figure 1). Furthermore, a northern branch of the North Atlantic Current (NAC) crosses the Mid-Atlantic Ridge from west to east just over the CGFZ while being under the influence of colder subpolar waters on its northern border. The CGFZ area is thus considered a major latitudinal transition zone in the environment and biodiversity of the MAR (Vecchione et al., 2010) and is now part of the world's first network of marine protected areas (MPAs) by both OSPAR (Oslo-Paris Convention) and the NEAFC (North-East Atlantic Fisheries Commission) (O'Leary et al., 2012). Therefore, the poorly explored MAR and CGFZ in particular are of considerable interest for deep-sea biodiversity and biogeography studies. The Mid-Atlantic Ridge Ecosystems (MAR-Eco) project (2001 -2010, www.mar-eco.no), part of the 'Census of Marine Life ' (www.coml.org), was aimed at studying the patterns and processes of the ecosystems of the northern Mid-Atlantic. In July -August 2004, the Norwegian RV 'G.O. Sars' collected benthic samples on the MAR between the Azores and the southern region of the Reykjanes Ridge. Most of the material has already been identified and published, notably in special issues of 'Deep-Sea Research Part II' and 'Marine Biology Research' in 2008 (http:// www.tandfonline.com/toc/smar20/4/1-2, http://www.sciencedirect.com/science/journal/09670645/55/1, both accessed 15 April 2014). Taxa examined and studied in these special issues include fish, fish parasites, Cetacea, corals, copepods, planktonic cnidarians and ctenophores, Holothuroidea, pourtalesiid sea urchins, Isopoda (Aegidae), Ophiuroidea, Anthozoa, Asteroidea, Echiura, Sipuncula, Brachiopoda and Hexactinellida. Additional taxa checklists of benthic organisms, including a preliminary demosponge species list of our identifications, were added and discussed by Gebruk et al. (2010). In the present paper, we update and refine this checklist while thoroughly describing and discussing the taxonomy of each species. Indeed, the study of this material is an opportunity to revise some poorly known Atlantic deep-sea species. We also compiled distribution maps for most of these species. They will form the basis of a discussion on the biogeography of deep-sea demosponges in the North Atlantic.

Sponge sampling
Sponge specimens were collected by bottom trawl on board the RV 'G.O. Sars' during Leg 2 (4 July -5 August) of the 2004 MAR-Eco expedition (cf. cruise report at http://www. mar-eco.no/sci/cruises/expedition_report_rv_g.o.sars_2004, accessed 8 March 2014). The trawl was a Campelen 1800 shrimp trawl with a 22 mm mesh size cod-end liner and a 12-17 m by 4.5 m opening at 50 m doorspread. Environmental data associated with the trawls were nearbottom measurements obtained with a CTD sensor (Søiland et al., 2008). For a detailed station list with coordinates, depths and environmental data, and for a full account of the approach and methods for sampling of the benthic fauna during this expedition, see Bergstad & Gebruk (2008). Demosponges were found at 10 stations, at depths ranging from 753 to 3046 m (Table 1, Figure 1): two stations were on the north-west of the CGFZ (stations 72 and 70), two shallower stations (,1000 m) were on MAR seamounts (stations 65 and 53), and finally six stations were north of the Azores (stations 40, 42, 44, 46, 50 and 52). Specimens were preserved in 4% buffered formaldehyde on board and later transferred to 70% ethanol. Specimens are stored at room temperature in the Bergen Museum (ZMBN).

Morphology studies
To collect the spicules, sponge tissue was digested in nitric acid on a microscope glass slide. Spicules were then washed with water and with ethanol 96% and mounted in Euparal; 25 spicules per spicule type were measured, unless otherwise stated. Measurements of all spicules were made with a light microscope. Width of triaene rhabdomes was measured right under the cladomes. Width of the oxeas was measured in the middle. For Thenea species, we only measured microscleres because megascleres have been shown to be similar in North Atlantic Thenea (Cárdenas & Rapp, 2012). Some of these spicules were placed on a stub, coated with a gold/palladium mix and observed with the ZEISS Supra 55VP and JEOL JSM-840A scanning electron microscopes (SEM) at the Laboratory of Electron Microscopy, University of Bergen. Thick sections (100-800 mm) were made with a diamond wafering blade and a low speed saw using an Agar Low Viscosity Resin kit (# Agar Scientific) in accordance with the manufacturer's mixing instructions to make a hard embedding medium. Digital pictures of these sections were taken with a Nikon camera fixed to a stereomicroscope (Leica M216 A). Thick sections, spicule preparations and SEM stubs are stored in the Bergen Museum. With respect to taxonomy, our study will follow the demosponge classification proposed by Morrow & Cárdenas (2015).

Distribution maps of species
All the records obtained from our identifications and the literature were compiled and mapped with GeoMapApp v. 3.3.9 (http://www.geomapapp.org), using the North Polar base map projection and the default Global Multi-Resolution Topography Synthesis (Ryan et al., 2009). When the latitude/longitude information was missing but the locality was given, we reconstructed the geographic coordinates using Google Earth. Distribution maps of boreo-arctic Geodia species have been updated from , by adding the MAR-Eco records and records from the Reykjanes Ridge (Copley et al., 1996). Additional records were also added for (i) Geodia atlantica from the Kerry Head Reefs cruise CV13012 (August 2013, chief scientist: Louise Allcock, material sorted by C. Morrow, identifications by PC) and for (ii) Geodia macandrewii from the Porcupine Bank (CE13008 campaign, June 2013, RV 'Celtic Explorer' using the ROV 'Holland I', chief scientist: Louise Allcock, material sorted by C. Morrow, identifications by PC). Maps of boreo-arctic Thenea species and Poecillastra compressa have also been updated from Steenstrup & Tendal (1982) (Søiland et al., 2008).  Table 1 lists the 22 demosponge species identified per station: 14 species belonged to the Astrophorina and three to the Spirophorina so that a total of 17 species belonged to the order Tetractinellida. The other five species belonged to the orders Polymastiida (one), Suberitida (one), Poecilosclerida (two) and Dendroceratida (one). An advantage we had when studying this collection is that most of the Astrophorina identified in this collection had been revised using morphological (Cárdenas & Rapp, 2012; and molecular data (Cárdenas et al., 2011) so we will often refer to these revisions for additional information. Specimens cited in the aforementioned papers have been used as comparative material.

discussion
Spicule measurements and morphologies fit with the description of this species . But the asters are clearly more strongly spined than in the NEA specimens (including the type). Also, the sterrasters are spherical, like in NWA specimens, whereas they were more 1478 paco ca ' rdenas and hans tore rapp elongated in NEA specimens. Although dichotriaenes are usually rare in this species, they are particularly common in specimens 105633 and 105637. The external morphology is a bit different from more northern specimens as well: darker colour and 'deflated' surface appearance. To conclude, as suggested before  there could be a southern morphotype of G. atlantica, to which the MAR-Eco specimens belong.

discussion
Spicule measurements and external morphologies are perfectly in accordance with the description of the species . A large category of oxyaster is present, as in other G. barretti which have been collected at depths deeper than 1000 m . The radial crystalline structures, thought to be calcareous, are common in many Tetillidae (e.g. Cárdenas et al., 2009)

discussion
We identified the MAR-Eco specimen as G. hentscheli and not G. barretti because it has no oxyasters I, sterrasters with 'cauliflower' surface ( Figure 6D, G), and many promesotriaenes. But this specimen is also slightly different from more northern specimens. The observed brown cells were not found in comparative material. The MAR-Eco specimen has elongated sterrasters which is unusual compared with more northern specimens. There are no irregular sterrasters, commonly found in Arctic specimens. The sterrasters are also fairly large for this species, but similar sizes were found in Davis Strait specimens. The microxeas are thinner: average of 5.5 vs averages of 8-16 mm in comparative material from the NEA and NWA    1482 paco ca ' rdenas and hans tore rapp found but these are usually found in the fur of G. hentscheli and here no fur was observed, maybe due to the trawling collecting method. The sterrasters are spherical, never irregular or with a 'cauliflower' aspect (a common feature in G. hentscheli). The dark brown colour seems to be the rule for this southern population whereas G. hentscheli is more often whitish or greyish; brown cells were observed only in the sections of G. hentscheli (105680), never in the G. northern mid-atlantic demosponges hentscheli specimens revised by . No budding was observed (vs occasional budding in G. hentscheli) but again, we may have seen too few specimens. The oxyasters are also fairly large but these sizes are also found in G. hentscheli . Likewise for the thicker cortex (1 mm vs 0.5 mm usually), it has been found in some G. hentscheli from Davis Strait . SEM observations of the strongylasters show that they also have hook-like spines, such as the ones observed in 105680. All in all, this southern morphotype of G. hentscheli may represent a separate southern species, which diverged from its Arctic counterpart. But they seem closer morphologically to the MAR-Eco G. hentcheli 105680. Instead of creating a new species, we prefer to wait for genetic data to take taxonomic action. We have found two other specimens belonging to this southern morphotype in the Bergen Museum (ZMBN 25668) and in Naturalis (RMNH 1458): both were collected in the Azores area at .2400 m depth. outer morphology and skeleton organization ( figure 8a, b ) Massive subspherical, whitish specimens with smooth surfaces. Specimens 105661, 105666 and 105669 are respectively around 12/4/1.5 cm in diameter. Cortex of 105661 is 1 mm thick. Thick sections of 105661 were made, skeleton organization is similar to that observed in other specimens  except that anatriaenes are more abundant below the cortex, associated with the orthotriaene bundles. Some sub-circular crystalline structures were observed, about 130 -240 mm in diameter with radial fibrous organization (very 'bushy' and confused appearance), and unclear borders. They are similar but larger than the ones observed in G. barretti, and are present below the cortex and in the choanosome.    northern mid-atlantic demosponges 1485 discussion The large size of the oxyasters is typical of individuals living deeper than 1000 m . The sterrasters measured in 105661 are the smallest sized ones ever found for this species and specimens 105666 and 105669 also had sterrasters in similar size ranges: G. macandrewii sterrasters are usually more than 200 mm in diameter . We reexamined thick sections of ZMBN 77924 (G. macandrewii from Korsfjord, Norway, fixed in ethanol 70%) but could not find any crystalline structures. outer morphology and skeleton organization ( figure 10a, b ) Subspherical specimen, 2 cm in diameter. Colour in ethanol is brown. A single small preoscule on the top surface, cribriporal pores all over the rest of the surface. The cortex is 1.2 -1.7 mm thick, very tough, and supported by triaenes and oxeas which form a 3.5 mm thick layer; below, oxeas occur in confusion. Oxyasters I can be found in high abundance in the choanosome. Microxeas can be found in the ectocortex and the choanosome. Large granulated cells about 25 mm in size can be found in the upper part of the endocortex, they contain many brown vacuoles inside. Some crystalline structures were observed in the choanosome, about 30 -87 mm in diameter with clear radial fibrous organization and 'hairy' border. They are of similar size and shape to the ones observed in G. hentscheli (105680) but the fibres seem to be thinner and less confused.   bathymetric range 200 -2600 m (Topsent, 1911; this study).

discussion
The external morphology of the MAR-Eco specimen from Station 50 (north of the Azores) is very similar to the specimen described and illustrated by Topsent (1928) from Madeira at 2380 m, and the specimen identified as 'Sidonops sp.?' by Arnesen (1920) from the Ibero-Moroccan Gulf at 1215 m (ZMBN 25652, re-examined for this study). A still photograph from a video by the manned-submersible MIR 1 above 1700 m depth in the CGFZ (Felley et al., 2008, Supplementary material) shows globular sponges with a single preoscule that look very much like G. megastrella, so this species may also be present in the CGFZ (dive coordinates 52858 ′ N 35801 ′ W). No distribution map was made for this species found between Ireland and the Azores in the NEA and between the New England seamounts and Florida in the NWA (P. Cárdenas, unpublished results) since we suspect it to be a species complex (Cárdenas et al., 2011) which needs to be properly revised before anything can be said about its biogeography. This is the first time that microxeas are observed to be occasionally centrotylote in G. megastrella. Occasional centrotylote microxeas may be a synapomorphy of the species belonging to the Depressiogeodia clade (G. barretti, G. hentscheli and G. megastrella complex) even though it has also been observed in G. macandrewii, albeit more rarely .  Massive spherical specimen (6 cm in diameter), white in ethanol, fairly hispid on one side, not compressible. Regular surface with no obvious large openings but small cribriporal pores and oscules are present. Cortex is 0.5 -0.6 mm thick and fairly easy to cut. The positions of the different euasters are clear on the thick sections. Ectocortex is 200 -250 mm thick, with sub-radial microxeas, numerous strongylasters (especially packed in the ectosome) and few spherasters. The fibrous endocortex is 500 -600 mm and packed with sterrasters; it is supported by large dichotriaenes. Protriaenes are crossing the cortex with their cladomes beyond the surface. In the choanosome, there are sparse oxyasters. Some irregular crystalline structures were observed in the choanosome, about 87-175 mm in length with radial fibrous organization. They are similar to the ones observed in G. macandrewii but more irregular in shape and less 'dirty'.  bathymetric range 98 -2600 m (Topsent, 1904;this study).

discussion
This study is an opportunity to revise the morphology of this poorly known, albeit fairly common, Lusitanian species. Many specimens photographed just after collection during the NEREIDA campaign (courtesy of F. J. Murillo) and the BANGAL cruise (courtesy of P. Rios) show that the natural external colour of this species is whitish to light brown. This species always has dichotriaenes, which often have 'wavy' deuteroclads. As previously observed by Topsent (1904) and Stephens (1915), the euasters have variable morphologies. The ectcortical strongylasters often have an inflated centrum (spherostrongylaster) and are then difficult to separate from the smallest spherasters; in other specimens they also sometimes have long actines with inflated tips as in tylasters (MNHN-DT1090, NEREIDA DR07-043b). The spherasters are most often fairly spherical but, sometimes, they are less inflated; their actine tips are also variable with sometimes blunt truncated tips, and sometimes sharp conical tips. As for the choanosomal oxyasters they sometimes have thin long actines, and sometimes wider conical actines. For instance, in the specimen from the Azores (Station 198) studied by Topsent (1892Topsent ( , 1904 and in the Balgim specimens, we found both morphologies of oxyasters. Stephens (1915) states that the spherasters are placed just below the cortex but our sections of the MAR-Eco specimen and MOM 04-0118 clearly show that spherasters are present in the ectocortex as well (i.e. above the sterraster layer). Actually, in the MAR-Eco specimen, they are only present in the ectocortex. Topsent (1904) notices that spherasters can become rare. We also observed this, and even found specimens where the characteristic spherasters were missing (e.g. BANGAL PC581, PC579, ZMBN 25660). In these specimens without spherasters, we did find many oxyspherasters but it is unclear if they are deflated spherasters or inflated oxyasters (since spherasters and oxyasters essentially have similar sizes).
The size of the oxyasters is also very variable, they can vary from 20-24 mm (Stephens, 1915) to 30-40 mm (MNHN-DT844) as in the MAR-Eco specimen and even up to 72 mm (BANGAL PC579, ZMBN 25660). The continuum of sizes makes it impossible to delimitate two size categories. The important variation in size of the choanosomal oxyasters has already been observed in all Atlantic boreo-arctic Geodia, and may be related to the depth and/or the silica concentration . We should stress here that specimens without spherasters and/or large oxyasters were confirmed to be G. nodastrella with external morphology, other spicules, as well as with molecular data (P. Cárdenas, unpublished results). We also noticed that the MAR-Eco specimen has spherical sterrasters, whereas they can be ellipsoid in the comparative material (e.g. BANGAL PC581, MNHN-DT846). The size of the sterrasters varied between 68 and 115 mm which is in the same range as most boreo-arctic Geodia, except for the very large sterrasters of G. macandrewii . Finally, another variation concerned the microxeas found in the ectocortex: the MAR-Eco specimen has fairly longer microxeas (260-500 mm) than in previous measurements (167-350 mm) ( Table 2).
With all these variations in mind, we re-examined the holotype of Geodia barretti divaricans from Madeira (MOM-INV-0022282, wet specimen and MNHN DT-1299, spicule preparation). The wet specimen is a small hispid fragment attached to coral, it is the only specimen known of this species. We made thick sections from the type and measured euasters ( Table 2). It occurred to us that the spicule repertoire was very close to that of G. nodastrella. The oxyasters can be found in various sizes, the smaller ones (15-22.5 mm) can be found just below the cortex and the larger ones (27-70 mm) are very numerous throughout the choanosome. Topsent (1928) surprisingly overlooked the small oxyaster sizes, even though we found some on Topsent's slide (MNHN-DT1299). The small oxyasters with an inflated centrum and spiny actines look similar to the ones we observed in specimens of G. nodastrella from Galicia (BANGAL). The length of the microxeas in G. divaricans (210-525 mm) are actually closer to the ones from our MAR-Eco specimen (260-500 mm). Geodia divaricans was also characterized by inflated rhabdomes of the dichotriaenes and rare flattened anatriaenes (Topsent, 1928). However, slight rhabdome swellings were also observed in the dichotriaenes of G. nodastrella from Galicia and more or less flattened anatriaenes (although not as flattened as in G. divaricans) were also observed in the Irish specimens (Stephens, 1915). Finally, the absence of spherasters as discussed previously is possible in G. nodastrella. So, in our opinion, no specific cortical or spicule characters really remain to keep the valid status of G. divaricans. Therefore, we formally propose that G. divaricans Topsent, 1928 is a junior synonym of G. nodastrella Carter, 1876. The G. nodastrella recorded by Burton (1934) in Greenland has been re-identified as G. hentscheli . We also re-examined the two specimens (ZMBN 25660) identified as 'Sidonops sp.?' by Arnesen (1920); these were collected quite near the Balgim CP63 station where G. nodastrella was reported. Their external morphology (large spherical sponges with cribriporal pores/ oscules) clearly matches that of G. nodastrella and their spicules match those of G. divaricans (no spherasters, very large oxyasters). The other sponge identified as 'Sidonops sp.?' (ZMBN 25652) is in fact G. megastrella (see above). outer morphology and skeleton organization ( figure 12a--e ) Specimens are spherical to subspherical sponges between 1.5 and 6 cm in diameter, with remains of hispidity on the sides, which are purple (possibly coloured by the encrusting sponge Hexadella dedritifera Topsent, 1913). 105678 has an Hexactinellida growing on it ( Figure 12B). Oscules and pores are uniporal. Thick sections of 105675 were made. Cortex is northern mid-atlantic demosponges moderately thick: 1.1-1.2 mm. Skeleton organization is in accordance with previous descriptions .

discussion
We identified these specimens as G. phlegraei and not G. parva -its sister species from the Arctic ) -based on the external morphology (thick cortex, regular smooth surface) and fairly large sterrasters. But we note however that these sterrasters are spherical as in some NWA specimens, and not elongated as in NEA specimens . We re-examined sections of G. phlegraei ZMBN 77929 (Korsfjord, Norway) for crystalline structures, and we found many: small (27-37 mm) dirty sub-circular ones were very abundant below the cortex, larger ones (similar to the ones observed in the MAR-Eco specimen) were observed deeper in the choanosome. note We place this Stelletta in the Geodiidae based on molecular phylogeny results from Cárdenas et al. (2011). It seems that, like Stelletta tuberosa, many species of Stelletta are actually Geodia species that have lost their sterrasters, they group in a Geostelletta p clade (named according to the rules of the PhyloCode). Before reallocating these Stelletta species in a new Geodiinae genus, we are waiting for more sequences of Stelletta species to have a better morphological understanding of this Geostelletta p clade. 1490 paco ca ' rdenas and hans tore rapp     Topsent (1892Topsent ( , 1904Topsent ( , 1928 from the Azores, and some we examined from Newfoundland, the actines of oxyasters usually have a tiny inflated tip. The MAR-Eco specimens, the Ingolf Expedition specimens, some Newfoundland specimens (UPSZMC 78269) and the specimen from Bay of Biscay do not have such a clear inflated tip. The former may be because the actines are thicker; on the other hand, in the Bay of Biscay specimen, it may be because the actines are so thin. The dichotriaenes of 105670 and 105679, MNHN-DCL4066 and some specimens described by Topsent (1904) from the Azores and Spain do not have a swelling on the rhabdome.
The rhabdomes of Newfoundland specimens are swollen (UPSZMC 78302) (although not as much as in the southern MAR-Eco specimens) or not (#474, UPSZMC 78301, 78269). So at this point, it is important to emphasize that the inflated tip of actines and the swelling of rhabdomes are not diagnostic characters of this species, they may be absent. Anatriaenes (in 105670) and pro/mesotriaenes (in 105679) were observed. The Bay of Biscay specimen (MNHN-DCL4066) and the one of the South Azores Seamounts (ZMAPOR 21665) had larger oxyasters (up to 96 mm) with thinner and pointier actines.
Family ANCORINIDAE Schmidt, 1870 Genus Stelletta Schmidt, 1862 Stelletta rhaphidiophora Hentschel, 1929 (Figures 13B & 15) material   outer morphology and skeleton arrangement ( figure 15a, b ) Massive subspherical, very hispid. Diameter of the specimens range between 6 cm (105664), 2 cm (105668) and 4 cm (105682). A small Geodia barretti (105665) is growing on 105664 ( Figure 15A). Surface colour in ethanol is whitish (sometimes dirty brown due to the sediments trapped in the hispid layer), choanosome colour is light brown. Oscules and pores not visible. Thick sections of 105682 were made. Cortex is 2-2.5 mm thick. Vitreous cortex, light greyish with two distinct layers of equal thickness: (i) the upper layer is covered with a very thin layer (50 mm or less) composed of a dense accumulation of strongylasters; below, there are small canals surrounded by trichodragmas and asters, (ii) the lower layer is fibrous with less trichodragmas and some asters. Radial bundles of dichotriaenes spread out like a fan in the cortex, their clads are essentially present in the upper layer of the cortex. Between those bundles, we find large sub-cortical canals. Below these bundles of dichotriaenes, large oxeas are present in no particular orientation.

discussion
In the original description of S. rhaphidiophora, there are two categories of anatriaenes, characteristic flattened ones and more usual ones (Hentschel, 1929). Although we did not find the flattened anatriaenes in any of our specimens, we are sure of our identification since the rest of the spicule morphologies and measurements perfectly match the original description and our comparative material. We did not find the common anatriaenes in 105682 but we found two in 105664: rhabdome length .4100 mm, width: 30 mm, clad: 130 -160 mm. This is the first time SEM observations are made for this species. It shows that the strongylasters have spined actines ( Figure 15F), which are not visible with the optical microscope, and thus not reported in the original description. As observed in the boreo-arctic Geodia species , there is some variation in the maximum size choanosomal oxyasters can reach. Type material has oxyasters that reach 40 mm (Hentschel, 1929), but in the MAR-Eco specimens the oxyasters only reach 27.5 mm. In ZMBN 85222 from Iceland (604 m) and UPSZMC 78297 -78298 from Davis Strait ( 850 m), oxyasters respectively reach 65 and 56 mm and have an inflated centre. On the other hand, specimens from the Greenland Sea, near the Schultz Massive seamount and the Arctic Mid-Atlantic ridge (ZMBN 85223, 1600 -1760 m) have smaller oxyasters (up to 25 mm) like in the MAR-Eco specimens; but they are in very low numbers, unlike the MAR-Eco specimens. ZMBN 85223 also has very few trichodragmas, rare and small triaenes. It is interesting to note that Geodia hentscheli (ZMBN 77925) collected from the same locality at similar depth has a similar phenotype: low number of oxyasters, rare and smaller triaenes , Figure 12D, Table 3) so in our opinion environmental parameters around the Schultz Massive seamount clearly influence these spicule morphologies, abundances and sizes. The MAR-Eco specimens considerably extend the range of this Arctic species southwards. Furthermore, our comparative material from the Davis Strait extends its range to the West. Stelletta rhaphidiophora is an amphi-Atlantic Arctic species while its sister-species, Stelletta normani Sollas, 1880 is typically boreal (P. Cárdenas, unpublished data). In the field, these species can usually be distinguished by looking at their colour and a cross-section in their cortex with the naked eye. Stelletta normani is usually hairy brown while S. rhaphidiophora is hairy white. Stelletta normani has a clear double-layered cortex with (i) a bright white layer (packed with trichodragmas) and (ii) a vitreous grey layer (fibrous layer). Stelletta rhaphidiophora has more of a vitreous grey single-layered cortex since the upper white layer is less obvious and bright (because it never has as much trichodragmas). outer morphology and skeleton organization ( figure 16a, b ) The single irregular specimen about 7 cm wide 'glues' together three G. atlantica (105639, 105641, 105642) and one Stelletta tuberosa (105648) (Figures 3A & 16A). Colour is dark brown in ethanol. Surface is rough. Radial organization of the skeleton at the surface, more confused in the choanosome. The cladomes of short-shafted dichotriaenes and plagiotriaenes reach a thin fibrous ectosome and sometimes slightly cross it. We could not find sanidasters in the ectosome. Below this ectosome and between the bundles of triaene rhabdomes there are many subectosomal canals. The ectosome and the subectosomal canals make a light layer 1 mm thick. A few anatriaenes and mesoanatriaenes were also occasionally found in this layer but usually crossed the ectosome (on sections #1 and #2). In the choanosome, there is a dense accumulation of oxyasters and oxeas in a confused arrangement. Large granular cells are present in the ectosome but are not obvious (they are not coloured).  northern mid-atlantic demosponges 1495 bathymetric range 157 -2600 m (Klitgaard & Tendal, 2004; this study).

discussion
Historically, the microscleres in Stryphnus fortis were called 'amphiasters' (Sollas, 1888) because they are somewhat symmetrical but we prefer to use 'sanidaster' and 'amphisanidaster' (the latter term was coined by Kelly & Smith (2012)) to emphasize that they are homologous to the ones found in Asteropus and Ancorina.
The morphology of the MAR-Eco specimen broadly agrees with the description of S. fortis but a few differences with the comparative material (including the type) were noted. The sanidasters in the type are slightly longer (12-12.8-15, N ¼ 10) and the proportion of amphisanidasters (vs sanidasters) is higher; indeed, the sanidasters in the MAR-Eco specimen often have additional actines on the shaft. We confirm that the type has only plagiotriaenes, as noted by Vosmaer (1885). Although most of the S. fortis comparative material we examined (from Norway, Sweden and Flemish Cap) had both plagiotriaenes and dichotriaenes, it was always with a majority of plagiotriaenes. Our specimen has on the contrary a higher proportion of dichotriaenes, which seems to be a characteristic of the southern population of S. fortis (Topsent, 1904;Boury-Esnault et al., 1994). Vosmaer (1885) states in his original description that oxyasters come in two sizes. However, this appears to be incorrect as we found only one size when we examined the holotype (ZMAPOR 02189). The oxyasters in the type are fairly large (42 -60.9 -75 mm, N ¼ 10) compared with the ones observed in the MAR-Eco specimen ( 17-30 mm), with many actines and a large centrum. More unexpected is the presence of anatriaenes, a spicule never observed before in this species, and usually never found in the genus Stryphnus; anatriaenes have only been found in one atypical New Zealand species rightly called Stryphnus atypicus Kelly & Smith (2012). Finally, the cortical arrangement in the type and the Norwegian specimen ZMBN 82977 is much more confused than in the MAR-Eco specimen. In the type and ZMBN 82977 some triaenes are somewhat radially disposed but they usually cross the ectosome; many other triaenes can be found in all other directions, including in the choanosome. To conclude, the MAR-Eco specimen is slightly different from typical boreal S. fortis but the examination of more southern specimens is required to see if these differences are consistent with a separate population or even species.
According to our observations and comparison with specimen Mc3395 of S. ponderosus, we confirm that S. fortis is a valid species, and not a synonym of S. ponderosus as suggested by some (Burton, 1926;Koltun, 1966). Stryphnus ponderosus has smaller oxyasters (usually never larger than 25 mm in diameter), it is a shallower species (0 -200 m) that lives in temperate waters of the North-East Atlantic (British Isles, Ireland, France, Spain) and Mediterranean Sea, it is often covered by the sponge Desmacella annexa Schmidt, 1870 (Topsent, 1928, Lévi, 1950, which is the case of Mc3395. On the other hand, S. fortis is a deep-sea amphi-Atlantic species (200-2598 m) found from the boreo-arctic region to the Azores, it is often covered by the deep-sea yellow sponge Hexadella detritifera Topsent, 1913. Since S. ponderosus and S. fortis mainly differ by the size of the oxyasters and megascleres, we wonder whether they could be conspecific, their spicule differences being a direct consequence of depth and silica concentration in the environment, as shown in other Astrophorina . However, molecular data shows that the two species are genetically different (Cárdenas et al., 2011), there is a 2 bp. difference in the Folmer COI fragment, and 1 bp. in the 28S (C1 -D2) fragment. So, in our opinion, S. fortis and S. ponderosus are sisterspecies that diverged recently from a common ancestor that colonized shallow waters from the deep-sea (or the reverse). This new environment may have influenced the size of the oxyasters. Today the shallow and deep-sea populations may not be genetically connected anymore (this should be tested with a wider sampling) and have retained their respective morphologies.
After examining material collected during the deep-sea NEREIDA 2009 campaign off Newfoundland, we concluded that the S. ponderosus recorded off Newfoundland is in fact S. fortis (Murillo et al., 2012;Kenchington et al., 2013;Kutti et al., 2013). For example, UPSZMC 78303 from the Flemish Cap has oxyasters which are 30-44.5-75 mm (N ¼ 10) in diameter. Likewise, S. ponderosus recorded from the Atlantic boreo-arctic region, as a major component of boreal sponge grounds, is in fact S. fortis (Hougaard et al., 1991;Klitgaard, 1995;Klitgaard & Tendal, 2004). Vosmaer (1885) gave no measurements for the spicules of S. fortis, so it was originally difficult to identify. Stryphnus rudis Sollas, 1888 was described as a new species based on the fact that it had dichotriaenes and plagiotriaenes (vs only plagiotriaenes in the type of S. fortis). It was collected in the Korsfjord in Norway, where we have collected extensively. The Stryphnus specimens that we found there were similar to the paratype of S. fortis so we confirm, as suggested by Topsent (1904), that S. rudis is a junior synonym of S. fortis. Likewise, it was initially thought that S. ponderosus only had dichotriaenes so when specimens were collected with dichotriaenes and plagiotriaenes, they were referred to as S. ponderosus var. rudis Sollas, 1888 (Topsent, 1894, Lévi, 1950Uriz, 1981). We now know that the proportion of dicho/plagiotriaenes can vary in S. ponderosus, so S. ponderosus var. rudis becomes a synonym of S. ponderosus (except for the records of Alander (1942) in Sweden which belong to S. fortis). bathymetric range 0-1740 m (Topsent, 1928, Sarà, 1964.

discussion
The presence of strongyles in our specimen is intriguing; they have never been recorded before in this species (Cárdenas & Rapp, 2012), the rest of the spicules agree well with those of P. compressa. More specimens and molecular data are needed to eventually test the status of these Mid-Atlantic ridge populations. Despite numerous records, this species has never been recorded to this day beyond the MAR, in the NWA. outer morphology ( figure 18a ) 105653 -105654 are the only large specimens with the typical elongated shape, they are respectively 5 and 3 cm long. The

discussion
No protriaenes were observed. In 105654 plesiasters are sometimes reduced to two actines, a feature not observed in Norwegian specimens but fairly common in our NWA specimens (UPSZMC 78289, 155197, 155199). Plesiasters of the MAR-Eco specimens are smaller than the ones measured in a shallower Norwegian specimen (actine length: 30 -83.2 -145 mm, ZMBN 85230, off Korsfjord, 300 m). We are currently missing spicule characters to properly discriminate T. levis from the other common North Atlantic Thenea. The SEM observations of the MAR-Eco specimens confirmed that large plesiasters were more minutely spined in T. levis (Cárdenas & Rapp, 2012) but this character is impossible to see without a SEM. We made two new observations: (1) the plesiaster actines in T. levis are 'fat' or bullet-shaped (in the MAR-Eco and comparative material), which is not the case of the plesiasters in T. muricata/schmidti/valdiviae; (2) anatriaenes in T. levis are on average more 'open' and with shorter clads than in T. muricata/schmidti/valdiviae. If confirmed these two spicule characters may be used to discriminate T. levis from the other North Atlantic Thenea. We extend the western distribution of T. levis by reporting its first presence off Newfoundland (Flemish Cap) at depths of 1079 m to 1462 m.  bathymetric range 349 -4020 m (Topsent, 1904). 1498 paco ca ' rdenas and hans tore rapp discussion 42 -368/44 -369 specimens and 70 -385 (105671) come from distant stations and the morphology of the spicules are slightly different and we decided to present both in Figure 19D -K. Plesiasters reduced to two actines were not recorded in this species before (Sollas, 1888;Cárdenas & Rapp, 2012) but they are common in the 42 -368/44-369 specimens, except for 105671. This is a tentative identification since without genetic data, it is still difficult to decide whether to call this species T. muricata or T. schmidti (its southern sister-species). Cárdenas & Rapp (2012) have shown that there are genetic differences between the two species (using the C1 -D2 fragment of 28S), a result which still needs to be confirmed with additional specimens and independent molecular markers. They have also shown that clear morphological characters are still missing to distinguish both species. We decided to call our specimens T. cf. schmidti because the morphology of the MAR-Eco specimens agrees more with the original description of this species (Sollas, 1888). All specimens had abundant large plesiasters. Plesiasters have a much wider range than in T. valdiviae. It is difficult to separate the largest metasters/spiraters and the smallest plesiasters, there is a continuum (whereas in T. levis, the spiraster category is quite distinct). outer morphology and skeleton organization ( figure 20a--d ) Sub-globular, slightly flattened, sometimes with a triangular shape. Fairly hispid surface. One to several oscules with large meshed sieves ( Figure 20B). Equatorial poral area (also sieved). Specimens all have a dirty colour in ethanol. Thick sections of 105631 were made. Large dichotriaenes are found at the surface, with their cladomes supporting the ectosome in a very regular fashion ( Figure 20D). Bundles of oxeas discussion Steenstrup & Tendal (1982) consider the species to be dimorphic with a spherical arctic form and a more flattened hemispherical boreal form. Cárdenas & Rapp (2012) later showed that there are two COI (Folmer fragment) haplotypes for this species: the Arctic COI haplotype is identical to the COI of T. muricata while the boreal haplotype has 1 bp. difference with the COI of T. muricata. With respect to their external morphologies, the MAR-Eco specimens are similar to the boreal form of T. valdiviae, also collected in western Norway (Cárdenas & Rapp, 2012, Figure 24C). Spicule measurements are also in the range of more northern specimens (Cárdenas & Rapp, 2012). Protriaenes were rarely observed and anatriaenes were not found; however, both of these megascleres do seem to be rare in this species (Cárdenas & Rapp, 2012). We further noted significant spicule differences with boreo-arctic comparative material. Plesiasters are much more common than in boreo-arctic specimens (from the Barents Sea, the Greenland Sea and the NWA) (Cárdenas & Rapp, 2012). These plesiasters have on average more actines (which are 1500 paco ca ' rdenas and hans tore rapp thinner) than boreo-arctic specimens. Above all, the most noticeable difference is the presence of amphiasters (instead of spirasters), which have never been observed in this species before (including in the type from the Faroe Islands). Metasters were absent in the MAR-Eco specimens, although they are present, albeit in small numbers, in boreo-arctic specimens. For these reasons (plesiaster abundance and morphology, amphiasters, metaster absence), we decided to cautiously identify these specimens under the name T. cf. valdiviae. The MAR-Eco specimens would be the deepest T. valdiviae ever collected (3046 m). These would also be the southernmost record for this species, usually considered to be boreo-arctic. While reviewing the distribution of this species we examined pictures of specimens from Stations 960 (394 m) and 1052 (440 m) collected in the Norwegian Sea and Barents Sea and identified by Topsent (1913) as T. muricata (pictures courtesy of M. Bruni, MOM). All these specimens are clearly T. valdiviae with sieved oscules. Specimens from Station 922 could not be retrieved (M. Bruni, personal communication) but since Topsent treated them also as T. muricata we assume they were also T. valdiviae. We also report here the first records of this species off Newfoundland (in the Flemish Cap). Actually, the specimen from the Flemish Cap identified as T. muricata by Murillo et al. (2012, Figure 6J) is also T. valdiviae: the oscule clearly has a sieve.
Suborder SPIROPHORINA Bergquist & Hogg, 1969 Family TETILLIDAE Sollas, 1886 Genus Craniella Schmidt, 1870 Craniella azorica (Topsent, 1913) (  Both specimens are small spherical hispid sponges, 1 cm and 1.5 cm in diameter. Surface colour is light greyish in ethanol. Cortex colour is light grey, choanosome colour is brown. Cortex is 0.7 -1 mm thick. 105673 has a triangular fringe of spicules on its surface which could be a closed oscule (such as the ones we observed in a boreal species: Craniella zetlandica (Carter, 1872)). Thick sections of 105674 were made. Oxeas I (and few pro/anatriaenes) form large radial tight bundles that cross the ectosome. Many protriaenes, and fewer anatriaenes appear in the extension of these bundles, at the surface of the sponge, thus forming the hispidity. We also observe a double-layered cortex: (i) an upper-layer with a thin ectosome ( 50 mm) made of large cells, and many sub-ectosomal canals, no spicules, (ii) a fibrous layer filled with a crisscross pattern of paratangential oxeas II. These oxeas II are absent in the choanosome. No embryos were observed in the thick sections. bathymetric range 599 -1650 m (Topsent, 1913; this study).

discussion
Our specimens agree well with the type material. The cortical oxeas II of the MAR-Eco specimens are slightly shorter and wider than in the lectotype (520 -809.6 -1010 × 18 -23.8 -32 mm). The only noticeable difference is that oxeas II in our specimens are slightly bent (or double bent in 105674) whereas they are very straight in both types. Topsent (1913) had noted that protriaenes could have irregular clads, we observed that these clads can even subdivide, as in the Balgim specimens . As for the anatriaenes, their cladomes have an umbrella shape, similar to the ones found in the paralectotype (and not as open as in the 1502 paco ca ' rdenas and hans tore rapp lectotype). We observed spermatic cysts in the choanosome (sections of 105674, collected on 26 July 2004), but we did not observe embryos (in either specimen). The MAR-Eco specimens greatly extend the northern distribution range for this species, until now only recorded from the Azores (Topsent, 1913(Topsent, , 1928 and the Ibero-Moroccan gulf (Balgim campaign) . The MAR-Eco specimens were found in the same depth range as the Balgim specimens (1510 m).
Pores not visible. Choanosome is light brownish in ethanol. Thick sections of 105616 ( Figure 23D, E) and 105617 were done, revealing identical skeleton organizations. Oxeas I (and few pro/anatriaenes) form large radial tight bundles that cross the ectosome to form the typical 'fur'. Protriaenes and anatriaenes are very abundant in the 'fur', which is full of sediments. Just below the surface where sigmaspires tend to accumulate, there is a fibrous layer 120 -200 mm thick, where some anatriaenes can also be observed. Below, the choanosome is full of oxeas II sometimes paratangential to the surface (they do not form bundles), and sometimes without a particular orientation. Sigmaspires are also very common there.  Koltun (1966) was wrong in synonymizing T. longipilis with T. polyura Schmidt, 1870; they have clearly different external morphologies and spicule differences, so they are both valid species. Furthermore, T. longipilis is Lusitanian and boreal, T. polyura is an arctic species. We considerably extend the northern range of T. longipilis since we report it for the first time in boreal waters, thanks to specimens we identified in the Ingolf Expedition and BIOICE. The external morphology and spicule measurements of the MAR-Eco specimens agree well with the type series described from the Azores. Our thick sections of 105616 and 105617 ( Figure 23D, E) were compared with the ones we made from the lectotype ( Figure 23F), collected in the Azores at 1846 m depth. The skeleton arrangements are similar but below the fibrous layer (300-400 mm thick), the oxeas II in the lectotype are more regularly arranged and concentrated in a paratangential manner making a loose layer about 300 mm thick. This layer is made of shorter oxeas II (around 600 mm long) than the other oxeas II in the choanosome (around 1000 mm long). Further observations of cross-sections in many specimens suggest that this layer of oxeas II, which makes a thin whitish layer visible to the naked eye, is more or less present depending on where the section is made. This layer is not characteristic of a Tetilla and reminds the microxea layer of Craniella species ( Figure 23F). Another character that reminds of Craniella is the presence of embryos in the type series; however, they are bigger and less abundant than in Craniella species (Topsent, 1904). We did not find them in the MAR-Eco specimens, nor in the Ingolf and BIOICE material that we examined. Based on the presence of cortical oxeas, Lehnert & Stone (2011) suggest the transfer of this species to Craniella but the thick sections made on the lectotype and the MAR-Eco specimens clearly show that even though T. longipilis has a denser arrangement of oxeas II below its fibrous layer, it has no true double-layered cortex, as in other Craniella species sensu Sollas (1888) (cf. description of Craniella azorica). So in our opinion, although this species might represent an evolutionary intermediate step between Tetilla and Craniella, we propose to keep it in Tetilla until additional data are obtained, and a revision of Tetilla and Craniella is made.

outer morphology ( figure 24a )
This is a small fragment of a larger sponge. No cortex is visible. A large spicule tuft is present on the surface which could represent an oscule.

discussion
We do not have a complete specimen but it is still enough to confirm the validity of this species. Indeed, both Topsent (1923) and Koltun (1966) suggest that T. sandalina is a synonym of T. polyura but after examining both species, we conclude that, although they are obviously morphologically close, they are clearly distinct. They both share the typical '2 short clads + 1 long clad' protriaene cladome and the raphide-like oxeas (called 'trichodal oxeas' by Sollas (1888)) but in T. sandalina, the sigmaspires are not centrotylote as in T. polyura. Furthermore, as in the original description, we could not find any anatriaenes (whereas they are very commonly found in T. polyura). One difference with the original description is that Sollas (1888) reports that one or both of the shortest clad can disappear; we have not observed this in our specimen. This is only the second record of this species, originally collected at a similar depth (1818 m) in the Azores during the Challenger expedition. We examined the ZMUC specimen of T. sandalina collected off Labrador (314 m depth) from Brøndsted (1933) and re-identified it as T. polyura (it has centrotylote sigmaspires, smaller than in T. sandalina: 12-15 mm); this is not surprising since T. polyura has been recorded in the Flemish Cap at similar depths (F. J. Murillo, personal communication outer morphology and skeleton organization ( figure 26a, b ) Massive cushion-shaped specimen. Surface is nearly smooth, with exhalant and inhalant papillae of different sizes, irregularly distributed. Colour of the surface, papillae and choanosome is pale pink in alcohol. Cortex dense, leathery, easy detachable. Choanosome friable. Exhalant papillae conical, slightly flattened, 2.8 mm in basal diameter and 0.4 mm in apical diameter. Inhalant papillae are nearly cylindrical. Main choanosomal skeleton is a reticulation of bundles of principal styles. Additional choanosomal skeleton made by free-scattered intermediary styles. The cortex is 2 mm thick, made of two overlapping layers: the external palisade of small styles and the internal confused and dense mass of intermediary styles mixed with small styles.

discussion
The type locality of P. corticata is in the South-West Atlantic, between Pernambuco and Bahia (Figure 25), at 2194 m depth, and yet our specimen very closely matches this species. Similar connections between the northern MAR and the South-West Atlantic have been reported for deep-sea Hexactinellida northern mid-atlantic demosponges (Lopes & Tabachnick, 2013). The specimens described by Topsent (1892Topsent ( , 1904 in the Azores are similar although he reports only one type of papillae. The specimen SMF 9633 (Lesser Antilles, Kahouanne Basin, 1127 m) was identified by D. Janussen (Meyer & Kuever, 2008); we did not examine this specimen but its 28S (D1 -D2) sequence (GenBank#EU005552) groups with the sequences of other identified P. corticata (A. Plotkin, personal communication).

discussion
This specimen's morphology agrees well with the redescription of the species and the type material (Uriz et al., 2011). The only difference we note is that the osculum is not present on the top surface but it is slightly off to the side, as in Stylocordyla longissima Sars, 1872, a species later synonymized with S. borealis (Hansen, 1885;Vosmaer, 1885;Arndt, 1913). This species is widespread across the deep North Atlantic boreo-arctic region, on either side of the Atlantic. However, this is its first record in the Azores region. Many GBIF (www.gbif.org) records (from the Yale Peabody Museum, material not seen) suggest that this species is very common on the eastern US coast shelf down to the Bahamas ( Figure 27). Schmidt (1880) also recorded this species from Grenada (290 m). We found a specimen from Ó lafsfjörður (Iceland, BIOICE campaign, Station 2193) collected at only 18 -22 m depth, which is the shallowest record for this species. The dubious identity of the small S.   Order POECILOSCLERIDA Topsent, 1928 Family COELOSPHAERIDAE Dendy, 1922 Genus Forcepia Carter, 1874 Forcepia (Forcepia) toxafera sp. nov.

discussion
The spicule assemblage of the MAR-Eco specimen is quite similar to the one found in Forcepia (Forcepia) groenlandica and Forcepia (Forcepia) thielei, respectively described from the east coast and from the west coast of Greenland. Both type materials were re-examined for this study and we also made new spicule preparations from the type of F. (F.) groenlandica. Actually, our specimen is even closer to the specimen

1508
paco ca ' rdenas and hans tore rapp described by Topsent (1904) under the name F. (F.) groenlandica, and collected at great depths (2252 m) in the Azores. Topsent's specimen and ours are clearly conspecific (Table 2)  , and again (iii) there are oxeote toxas. All these differences support the creation of a new species, named after the large oxeote toxas. These toxas were overlooked by Topsent (1904) but we found them to be common (albeit in lower numbers than the styles) when we re-examined his specimen (MOM INV-22577). After F. (F.) groenlandica and F. (F.) thielei, F. (F.) toxafera sp. nov. is the third species of Forcepia with a single category of forceps, which are asymmetrical. There are three other North Atlantic Forcepia species with asymmetrical forceps but they also have another category of larger symmetrical forceps: Forcepia (Leptolabis) brunnea (Topsent, 1904), F. (L.) assimilis (Lundbeck, 1910) and F. (L.) forcipula (Topsent, 1904). Forcepia (L.) brunnea and F. (L.) assimilis may actually be the same species according to Topsent (1928).

etymology
This species is named 'paravaridens' to underline its presumed phylogenetic affinities with the Arctic species Iotroata varidens.

discussion
According to their definition, Iotrochotidae species should have birotula microscleres (van Soest, 2002). However, our new species does not have birotulas but two sizes of anchorate chela. We nonetheless decide to allocate this species to the genus Iotroata for the following reasons: (1) the spicule repertoire of this species closely matches that of other Iotroata species (smooth styles, tylotes and unguiferous chelae) and (2) birotulas are probably derived from anchorate chelae (van Soest, 2002). So we consider here that the smaller size of anchorate chelae in our new species is homologous to the small category of birotulas in other Iotroata species. Furthermore, this new species is very similar with respect to spicule morphology and sizes to Iotroata varidens (Lundbeck, 1910), a deep-sea species described from the Denmark Strait at 567 m depth. The main difference between the two species is the presence of birotulas in I. varidens (size of birotulas: 15 -21 mm) vs polydentate anchorate chela in I. paravaridens (size of chela: 14 -25 mm).

discussion
Four species of Spongionella are recorded from the North-East Atlantic/Mediterranean area: S. pulchella (Sowerby, 1806), S. gracilis (Vosmaer, 1883), S. ramodigitata (Topsent, 1901) and S. depressa Topsent, 1929. Based on our measurements of the primary and secondary fibres and on the external morphology, our specimen is closer to S. pulchella. But the distance between the primary fibres (550 -1150 mm) is much bigger than usual (200-350 mm), which is a character found in S. depressa. Moreover, S. pulchella is usually found between 40 and 380 m, the deepest record being off Monaco (Vacelet, 1969 Lendenfeld, 1888;Koltun, 1959). But S. brandtii has a characteristic smooth foliaceous shape (termed 'discoidal' by Koltun) with concentric trenches (Koltun, 1959, plate 41), whereas S. pulchella is often a rather thick lamella with a minute conulose surface (not smooth). We suggest that S. brandtii type material should be re-examined carefully because we suspect this species to be valid.

D I S C U S S I O N
We have identified in the MAR-Eco collection 22 species of Demospongiae, two of which are new for science (Table 1): Forcepia (Forcepia) toxafera sp. nov. and Iotroata paravaridens sp. nov. Seventeen out of the 22 belong to the Tetractinellida order, a group thought to have originated in the deep sea (Cárdenas et al., 2011). In comparison, 16 species of Hexactinellida sponges were identified in the MAR-Eco material with 13 species new for science Tabachnick & Collins, 2008). Only one specimen of Calcarea sponge was found in the MAR-Eco collection: a new calcaronean species (to be described). Station 50 (north of the Azores) was the richest in terms of demosponge 1510 paco ca ' rdenas and hans tore rapp species and biomass: eight species (including four Geodia species) and 24 specimens indicating a sponge ground in this area. Station 70 (north-west of the CGFZ) was also fairly diverse with seven species and 13 specimens. Station 70 actually had the highest diversity of Hexactinellida with nine species identified . Overall, Station 70 (at the southern tip of Reykjanes Ridge) had not only the highest sponge diversity, but also the highest benthic species richness .
The MAR is not a longitudinal barrier for deep-sea demosponges According to our distribution maps (Figures 2, 13 , 20, 25 & 27), 68% of the MAR-Eco species (15 species out of the 22) are amphi-Atlantic. We further note that both boreo-arctic species (with distributions north of the CGFZ) and Lusitanian species (with distributions south of the CGFZ) can be amphi-Atlantic. Although Tetilla longipilis has been essentially collected on the MAR, apart from a record in Hatton Bank (NEA), we have identified specimens collected in the Denmark Strait and on the western side of the MAR so it is likely present further west, and we suspect it to be a true amphi-Atlantic species. Since this is a poorly known species it could have easily been overlooked in the NWA. On other hand, Poecillastra compressa seems genuinely restricted to the NEA and the MAR, since it has numerous records and is a relatively easy species to identify. The rest of the NEA-restricted species are either new species (I. paravaridens sp. nov., F. (F.) toxafera sp. nov.) or poorly known deep-sea species (i.e. with very few records) with distribution ranges south of the CGFZ (C. azorica, T. sandalina, S. aff. pulchella). Therefore these species might very well be amphi-Atlantic but for now they are considered restricted to the NEA and/or the MAR. 68% of amphi-Atlantic sponge species is well in agreement with a faunistic review showing that 61% of the 80 species collected in the Reykjanes Ridge (Madreporaria, Cirripedia, Echinoidea, Asteroidea and Brachiopoda), north of the CGFZ, were amphi-Atlantic (Mironov & Gebruk, 2006). The conspecificity of boreo-arctic amphi-atlantic Geodia species has been further confirmed with genetic data . Seemingly, the status of the amphi-Atlantic MAR-Eco species found south of the CGFZ would need to be tested with genetic data. To conclude, the high proportion of amphi-Atlantic species suggest that the MAR is not a barrier to the dispersal of deep-sea Demospongiae. On the contrary, deep-sea demosponges may use the MAR and neighbouring continental margins to disperse at bathyal depths, to avoid the much deeper abyssal depths (i.e. large abyssal basins) where they are absent (Figures 2, 13 (Mironov & Gebruk, 2006). However, in the case of five out of these eight species, our study showed that the MAR southern populations (i.e. south of the CGFZ) were somehow morphologically different from the northern populations. Southern populations of G. atlantica, G. hentscheli, T. valdiviae, S. fortis and P. compressa had clear external morphology and/or spicule differences. More specimens from this area as well as genetic data are now required to confirm these observations and test if these southern populations represent sister-species. Actually, T. schmidti is probably another example of this since it may represent a southern sister-species of T. muricata, hence our difficulty to discriminate them morphologically. Genetically, the mitochondrial cytochrome oxidase I (COI) Folmer fragment is not enough to differentiate these two populations but the 28S (C1 -D2) marker is more promising (Cárdenas & Rapp, 2012). Likewise, F. (F.) toxafera sp. nov. and I. paravaridens sp. nov. are probably sister species of more northern Greenland species, respectively F. (F.) groenlandica and I. varidens.
Overall, these observations suggest that the CGFZ area represents a major biogeographic barrier for deep-sea demosponges, which limits gene flow between northern and southern populations. This barrier is not limited to the two deep fractures per se (about 4500 m deep) since some of the southern species (C. azorica, S. tuberosa) were actually sampled slightly north of the CGFZ (Station 70 and 72) but it is more likely in combination with environmental changes in the CGFZ area. It is also interesting to notice that two (G. hentscheli, S. rhaphidiophora) out of the five species are strict Arctic species, which manage to extend their southern distribution range to the CGFZ, probably by following the southward flow of cold deep waters (Iceland-Scotland Overflow Water: ISOW) along the Reykjanes Ridge. The CGFZ area has already been considered a major biogeographic transition zone for many planktonic, pelagic or benthic organisms, (Mironov & Gebruk, 2006;Gebruk et al., 2010;Vecchione et al., 2010;Alt et al., 2013). However, one should keep in mind that the CGFZ may not be a boundary for all marine organisms: Kongsrud et al. (2013) do not find a difference in the species composition of benthic polychaetes north and south of the CGFZ and wish for more samples to settle the matter. Watling et al. (2013) delineated global lower bathyal biogeographic provinces (801-3500 m) but were unsure of the boundaries for many of them. In our opinion, the CGFZ could be considered as the MAR biogeographic boundary between the two lower bathyal provinces BY2 (North Atlantic Boreal) and BY4 (North Atlantic) (Watling et al., 2013).

The distribution of deep-sea sponges in the North Atlantic
Numerous studies from shallow water demosponges suggest that sponge larvae are generally short-lived, thus suggesting that short-distance dispersal may be the rule in shallow waters (Uriz & Turon, 2012). There are currently no behavioural or ecological studies on deep-sea demosponge larvae.

northern mid-atlantic demosponges
One study on the Hexactinellida Oopsacas minuta (considered to be a Mediterranean deep-sea species) showed that most larvae settle and metamorphose into the juvenile sponge within 12 -24 h after release from the parent (Leys et al., 2007). So deep-sea sponge larvae may have similar behaviour and short-life expectancies as shallow species and thus have short-dispersal potential. Keeping this in mind, knowing that the CGFZ represents a border between two deep-sea sponge faunas partially hints at the environmental parameters that might influence and limit the distribution of deep-sea sponges. Temperature, salinity and oxygen are very similar at the north and the south of the CGFZ (Table 1) but other environmental parameters related to depth may influence the distributions of deep-sea sponges.
Even though we acknowledge that the distribution maps presented in this study closely mirror overall sampling efforts, one should remember that the North Atlantic is probably the best studied and explored deep-sea area in the world. So we can consider most of these distribution maps (for all the Astrophorina: Thenea spp., Geodia spp., S. tuberosa, P. compressa, S. fortis) as good approximations of the current distribution of these species. Although these maps suggest wide distributions for these deep-sea sponges, depth clearly seems to be a limiting factor. Geodia barretti is the most widely distributed species of our collection, probably because it is the most common large Geodia of the North Atlantic, with one of the widest bathymetric ranges (30 -2000 m) . Geodia phlegraei was commonly observed and collected at 3000 m on Orphan Knoll (north of Flemish Cap)  and we examined a specimen of Geodia megastrella collected at 4152 m depth on the Atlantic continental margin off France (MNHN-DCL2857): both of these species were present in the MAR-Eco material. Thenea valdiviae also has a large track record since it is easily collected in soft sediments, with an even wider bathymetric range (100 -3046 m). So most deep-sea demosponge species have bathymetric records that rarely go beyond 3000 m depth, which thus probably reflects a true lower limit. Of course, there are a few typical deep-sea North Atlantic demosponges that live in the deeper abyssal plains (Barthel & Tendal, 1993), but these species were not found in the MAR so our discussion will focus on bathyal demosponges. The 3000 m depth limit of bathyal demosponges was confirmed by direct observations when exploring the 4500 m deep CGFZ with manned-submersibles, Geodia (called 'round sponges' in Felley et al. (2008)) were not found deeper than 3000 m and most were found shallower than 2500 m depth (Felley et al., 2008;Gebruk & Krylova, 2013). Hexactinellida are also fairly common and diverse around 2500 m depth but they can also occur deeper than 3000 m (Felley et al., 2008;Gebruk & Krylova, 2013). Interestingly, most corals in the CGFZ (Felley et al., 2008) also share this lower limit of 3000 m. So depth is certainly a limiting factor for the dispersal and distribution for most Demospongiae and Hexactinellida. If these species are restricted to the lower bathyal depth layer (800 -3500 m) it means they can disperse without much depth restrictions in the boreo-arctic region, especially in the Shetland-Faroe-Iceland-south Greenland arc where depth does not exceed 3500 m (Figure 1). This would explain why we have so many amphi-Atlantic boreo-arctic sponge species. On the other hand, between 508N and 308N in the North Atlantic, deep basins with plains at abyssal depths (3500 -6500 m) (Figure 1) would greatly limit or even prevent the dispersal of these animals, which usually have short-lived larvae. For instance, at 488N, the MAR is about 1000 km from the closest continental margins (Flemish Cap to the west, Porcupine Bank to the east) (Figure 1). And yet, our results showed that some Lusitanian species were amphi-Atlantic so they manage to disperse across the Atlantic. These deep-sea demosponges would then maybe use the numerous North Atlantic seamounts (Figure 1) and the MAR as 'stepping stones' above the wider ocean basins, although this scenario has never been extensively tested in the North Atlantic and the few genetic studies made tend to give mixed results (Cho & Shank, 2010;Rowden et al., 2010). It is however a fact that Geodia species are commonly found on the slopes and tops of seamounts (e.g. seamounts south of the Azores or off Portugal), or oceanic islands (Bermudas, Azores).
The CGFZ is characterized by two main deep rift valleys 35 km apart; they can be up to 4500 m deep, but are only 10 km wide (Figure 1) . CGFZ sponges significantly show clumped distributions, which could only be partially accounted for by their preference for hard substrates (Felley et al., 2008). So it is possible that either larvae are attracted to other conspecific individuals or that larvae settle close to the parents, a common feature in shallow sponge larvae (Uriz & Turon, 2012). Furthermore, the fact that MAR sponges are mainly restricted to depths ,3000 m and that dispersal may be over short distances, the 10 km wide/ 4500 m deep rifts may represent a true barrier for larvae dispersal. But since Lusitanian species do make it to the northern edge of the CGFZ, larvae must find a way to cross in 1 -2 days the rifts, if we consider them short-lived. Other parameters may be limiting their dispersal further north.
In addition to depth, the distribution of Demospongiae may also be limited by suitable habitats, which may be rocky outcrops (e.g. for Geodia species, P. compressa) or soft bottoms (for Thenea or Stylocordyla). Soft bottoms are the most widely distributed habitats in the MAR so they cannot be a limiting factor. As for rocky outcrops, although they only represent 5% of the MAR, essentially in steep slopes (.308) flanking the MAR (Niedzielski et al., 2013), they are present throughout the MAR and may therefore not significantly limit the dispersal of sponges which prefer these habitats.
Deep-sea currents may also influence deep-sea sponge dispersal and distribution. In the North Atlantic, the Gulf Stream gives the eastward North Atlantic Current (NAC) which crosses the MAR around the CGFZ (Søiland et al., 2008). The northern edge of the NAC (temperate waters) runs alongside the subpolar front (colder waters) which shifts between 488N and 538N (Søiland et al., 2008). This subpolar front is correlated with a clear faunal discontinuity (Vecchione et al., 2010). Meanwhile, there is an opposite deep current of overflowing high salinity bottom cold water coming from the Arctic: the Iceland-Scotland Overflow Water (ISOW). The ISOW runs southward along the flank of the Reykjanes Ridge and explains why deep-sea Arctic sponges (S. rhaphidiophora and G. hentscheli) reach the CGFZ. In the CGFZ area, the ISOW current flows mainly from east to west (in the opposite direction of the NAC)  and may therefore limit the dispersal to the north of species sensitive to cold waters. Furthermore, it has been argued that the ISOW does not act as a dispersal route across the ridge since substantial differences in megafaunal invertebrate 1512 paco ca ' rdenas and hans tore rapp community composition (sponges not included) within the northern part of the CGFZ were observed (Alt et al., 2013), but this needs to be confirmed with a much wider sampling. To conclude, even though the ISOW is the main current along the northern MAR up to the CGFZ, its influence on deep-sea sponge dispersal and distribution is poorly understood. One of the main reasons may be that the ISOW itself is still not fully understood (Kanzow & Zenk, 2014).