Really Onychocellids? Revisions and New Findings Increase the Astonishing Bryozoan Diversity of the Mediterranean Sea

: Investigation of bryozoan faunas collected in two submarine caves in Lesvos Island, Aegean Sea revealed a great number of colonies of three species currently assigned to the cheilostome family Onychocellidae: Onychocella marioni Jullien, 1882, O. vibraculifera Neviani, 1895, and Smittipora disjuncta Canu & Bassler, 1930. All species were ﬁrst described and subsequently recorded on several occasions, from the Mediterranean Sea, particularly from the Aegean Sea. The availability of this material provided the basis for more detailed observations and ﬁrst scanning electron microscopy (SEM) study of some diagnostic characters, including ovicells and ancestrulae, for the well-known species, as well as a few colonies of a species left in open nomenclature (i.e., Onychocellidae sp. 1) in previous works. In this paper we ( i ) update the descriptions of these four species; ( ii ) resurrect the species Floridinella arculifera Canu & Bassler, 1927, which was previously synonymised with Caleschara minuta (Maplestone, 1909), suggesting for it the new combination Tretosina arculifera ; ( iii ) and introduce the new genus Bryobifallax for S. disjuncta .


Introduction
The Mediterranean Sea is one of the best studied marine areas in the world. The first, pioneering investigation started at the end of the 16th century with naturalists, such as Ferrante Imperato, also describing some bryozoan species [1]. The Mediterranean Sea hosts a high proportion of the global biodiversity with approximately 17,000 species [2], including 556 bryozoans which account for about 10% of the world known diversity for the phylum [3]. However, large sectors (mostly in the eastern and southern Mediterranean) and several habitats (e.g., remote and hardly accessible dark habitats) remain understudied, as recently demonstrated [4][5][6][7][8].
In this context, the availability of samples from submarine caves of Lesvos Island, located in the northeastern sector of the Aegean Sea in the NE Mediterranean, was twofold relevant because they yielded colonies of both rare and undescribed taxa (e.g., [6]).
Here, we focus on species of the family Onychocellidae or onychocellid-like taxa. The family Onychocellidae was introduced by Jullien [9] based on Onychocella Jullien, 1882, the type species of the genus, for species having autozooids with extensive cryptocyst, opesiae placed in the autozooidal distal half and lacking spines, inconspicuous ovicells, and large polymorphs (or onychocellaria)

Material and Methods
The material studied here was mainly collected in two submarine caves formed in Triassic carbonate rocks on the southeastern coast of Lesvos Island, N Aegean Sea, Greece. These are: Fara cave (38.969 • N, 26.477 • E), a 32-m-long branch of a cave complex, and Agios Vasilios cave (38.969 • N, 26.541 • E), a 25-m-long funnel-shaped blind cave. Both caves have been widely studied for their morphology and for some taxonomic groups of the sessile benthic component [4,[19][20][21][22]. A summary of the samples collected at Lesvos, with the number of colonies of the four target species yielded in each cave, can be found in Table 1.  (Canu & Bassler, 1930) 4 (1) 2 (1) Tretosina arculifera (Canu & Bassler, 1927) 3 (2) For each species, the number of living and dead colonies (in brackets) is indicated. For each station the following data are reported: sampling location either on the Walls or Ceiling of the cave; distance from the entrance and approximate depth (in metres); biocoenosis and main encrusters indicated by the following abbreviations: SD = Semidark Cave Biocoenosis; Trans = Transitional Zone; Dark = Dark Cave Biocoenosis; Sc = Scleractinian corals; Sp = Sponges; Sr = Serpulid Polychaetes; B = Bryozoans.
Additional Onychocella colonies were collected in three submarine caves located at about 20 m depth in the Plemmirio Marine Protected Area (SE Sicily, western Ionian Sea). Detailed information for these samples can be found in Rosso et al. [23].
Samples were routinely processed at the Palaeoecological Laboratory of the Department of Biological, Geological and Environmental Sciences (DSBGA), University of Catania (Italy). They were washed and sieved, bryozoan colonies and fragments picked, and species initially identified using

Description
Colony encrusting, multiserial, uni-to multilaminar, moderately large, up to 5-6 cm 2 in the examined material, forming knobs of self-overgrowing lobes produced by pseudoancestrulae budded frontally by scattered autozooids (Figures 2A and 3A) of the outer layer; occasionally producing unattached laminar extensions up to one cm in length; light brown with dark brown avicularian mandibles when alive. Autozooids irregularly or quincuncially arranged, communicating through few, large septular pores located at mid-height on the vertical walls ( Figure 3C); large (mean ± Standard Deviation: 538 ± 49 × 442 ± 38 µm) and thick, generally slightly longer than wide (mean L/W: 1.22) but often as long as wide ( Figures 1B and 2E); hexagonal, often arched distally; zooidal boundaries raised, with a median groove in between ( Figure 2B,C and Figure 3D,E). Gymnocyst absent. Cryptocyst (Figures 2 and 3) more extensive proximally, narrowing laterally to the opesia, tapering distally; concave and sloping towards the opesia; evenly and coarsely granular with granules smaller and more densely packed along the margins.
Avicularia common (avicularium-autozooid ratio 1:2.5 and 1:4 in the two examined colonies including about 100 zooids each), vicarious, as long as autozooids but narrower (mean ± SD: 567 ± 69 × 315 ± 40 µm), strongly arched and asymmetrical; gymnocyst usually restricted to the lateral raised walls of the rostrum and occasionally developed proximally if an ovicell is present ( Figure 1B-D, Figure 2C-E and Figure 3E); cryptocyst extensive, similar in appearance to the cryptocyst of autozooids; opesia subcentral, elliptical (mean ± SD: 267 ± 41 × 171 ± 25 µm), the margin beaded as a result of the cryptocystal granules projecting into it; rostrum triangular, about half of the total length of the avicularium with elevated, serrated margins; two delicate knobs with C-shaped fissures at mid-opesia length as a hinge for the mandible. Mandible triangular and falciform, slightly hooked, about 420 µm long, with a median arched sclerite and extensive lateral wings on the convex side. Ancestrula similar to a normal zooid but smaller (mean ± SD: 360 ± 19 × 413 ± 19 µm), budding one distal and two disto-lateral autozooids of about the same size ( Figures 1H and 3B).
In a few colonies, unilaminar expansions extend up to 1 cm over the substrate showing the slightly swollen basal walls of the zooids, their arrangement in somewhat radial rows marked by thin furrows and zooidal lateral connections ( Figure 2F).
Intramural budding common in both autozooids and avicularia, with multiple regenerations visible as piled-up zooidal rims, up to seven in the examined material ( Figure 3C-E).

Remarks
In the colonies examined, peripheral autozooids appear larger and with wider, roundish to elliptical opesiae compared to autozooids placed in other regions of the colony. Differences in the description of these characters and the wide ranges of zooidal and opesial measurements reported in the literature (e.g., [9,11,12,16]) can be explained by the high ontogenetic variability, as already noted by previous authors. The few morphological features available to distinguish species of Onychocella (see also discussion in [29] p. 40), and the high intraspecific and intracolonial variability, possibly contributed to a proliferation of species and confusing synonymies often merging together fossil and living taxa.
In addition to Onychocella vibraculifera Neviani, 1895 (described below), at least three more species have been reported from the Atlanto-Mediterranean area: O. marioni described from recent material from the NW Mediterranean; O. antiqua (Busk, 1858) described from living specimens from Madeira; and O. angulosa Reuss, 1848, described from European Cenozoic material. Taylor et al. [10] summarised the problems related to the synonymy of the fossil species with the modern taxa and the urge for a revision of the type material to assess possible conspecificity. In the Mediterranean, O. antiqua was reported only once from Turkish waters [30] and its genuine occurrence was questioned in Rosso & Di Martino [3]. Based on Reuss' drawings [31] (pl. 11, fig. 10), O. angulosa would differ from O. marioni in having autozooids with centrally placed, polygonal opesiae, and small heterozooids with reduced opesiae. These potential differences, that can be also observed in specimens of O. angulosa illustrated in Zágoršek [32] (pl. 62), prevent the assessment of the long presumed conspecificity between O. angulosa and O. marioni without a careful examination of the type material. Further records from present-day west Africa by Cook [33] (p. 68, fig. 11A,B) seemingly point to a different species showing a distal orificial process projecting into the opesia. Here, we prefer to refer our specimens to O. marioni because of the above-mentioned morphological differences and because the material studied is recent. Onychocella marioni differs from the Mediterranean congener O. vibraculifera in the colour of the living colonies and in the additional skeletal characters discussed below.
We provide the first SEM images of the ancestrula with the first three budded zooids (one distally and two distolaterally), a configuration more commonly observed than the ancestrula alone or with a single distal zooid. Ancestrulae without buds were never described or illustrated for this species, suggesting a possible immediate budding of the first zooid. This is in agreement with observations made by Cook [34] on Onychocella alula Hastings, 1930, in which fully calcified walls of the primary bud were observed after 12 h, while distolateral buds appeared after 108 h from larval settlement. The ancestrula was first described by Canu & Bassler [25] as a "small ordinary zooid". After Gautier [35] erroneously figured an isolated tatiform ancestrula encrusting the lateral proximal wall of the real ancestrula of O. marioni, Gautier [11] himself and other authors, such as Prenant & Bobin [12], described the ancestrula as being either tatiform or similar to a later autozooid, while Zabala [36] reported only a tatiform ancestrula with 11 spines.
Colonies of O. marioni from Lesvos caves form nodular, multilayered, elevated structures [4] similar to those described by Harmelin [37] from the Trémies cave (Marseille area, France). This species also formed small pillar-shaped structures in a shallow-water, dark cave from Lebanon [J.-G. Harmelin, personal communication, October 2020]. Harmelin [37] discussed the perenniality of this species, which is able to maintain the colonised space through the superimposition of subsequent layers, creating new cryptic space for smaller, less competitive species growing around the nodules. In some colonies, we also observed frontal budding with the formation of pseudoancestrulae produced mainly by autozooids.
Unilaminar expansions were never reported in O. marioni to date, and never observed in colonies from submerged caves, although sometimes found in colonies from deep shelf biodetritic bottoms in the Ionian Sea, off eastern Sicily (AR, personal observations). In marine cave habitats, the development of these marginal laminae may allow the colony to overtop neighbouring organisms, an unusual strategy for a species with intrazooidal budding but seemingly advantageous on locally highly colonised hard surfaces such as the walls and ceiling of the caves.

Distribution
Based on the comments above (see Remarks) and adopting the same approach as in Berning [29] (p. 42) for O. angulosa, we refrain from providing a detailed, general distribution of O. marioni. Although recorded from the Atlantic around the Iberian Peninsula [38], it is likely that O. marioni is an endemic Mediterranean species, widely distributed on the whole basin. It has been commonly found in several localities and habitats by one of us (AR) including: coralligenous concretions at 35-55 m depth in the Gulf of Noto, SE Ionian coast of Sicily [39] and pillar-like coralligenous structures at 30 m depth in the same area [26,40]; circalittoral detritic bottoms in the Gulf of Noto (33-78 m living and 33-83 dead) [26], detritic bottoms of the outer shelf in the Ciclopi Island Marine Protected Area (AR, personal observations), and off Ustica Island (60 m living and dead) [41,42]; several submarine caves from the Ionian coast of Sicily [23,28,43]. Additional plausible records are from: Marseille area [9] and other localities along the Mediterranean French coast in: coralligenous habitats, dark and semidark caves, and clastic biogenic bottoms [11,35,44,45]; including the underside of small substrata [45,46] underwater tunnels in Medes Island, Catalan coast [47]; off the coast of Latium, Volcano Isle (S Tyrrhenian Sea), Tunisia [16]; the Aegean Sea in the Karpathos Strait (29-80 m), Kythira Island (66 m) and Santorini (100-128 m) [14]; localities along the southern coast of Chios Island, i.e., Cape Masticho (15-60 m), Venetiko (12-50 m), and Emborios Bay (1-15 m) but reported as O. angulosa [15], and off Milos Island [48]; Cyprus [8,49], and the coasts of Turkey [50] and Lebanon [18] Autozooids irregularly or quincuncially arranged, large (mean ± SD: 427 ± 28 × 326 ± 34 μm) and thick, slightly longer than wide (mean L/W: 1.31); generally ovoidal but often rounded polygonal and arched distally; zooidal boundaries raised and outlined by narrow grooves (Figures 4A and 5A,B-D, 6A,B,E). Gymnocyst forming the lateral walls, visible frontally only on zooidal proximal corners, mostly in irregularly growing areas (Figures 5E,F and 6D). Cryptocyst ( Figures 5 and 6) extensive, occupying the proximal half of the frontal surface, and forming two wings projecting laterally into the opesia at about mid-length, in a few instances tapering gradually; absent distally; depressed, gently sloping from the zooidal rim, flat centrally; coarsely and evenly granular, except for the sloping margins where granules are smaller and more densely packed.
Tubercles at triple zooidal junctions were also described in species of Smittipora, such as Smittipora sawayai Marcus, 1937, redescribed by Winston and Vieira [53], and Smittipora tuberculata (Canu & Bassler, 1928). They are apparently gymnocystal (see [10]), as in the present instance.  (C) Close-up of two zooids, one ovicellate, and a vicarious avicularium contributing to the formation of the gymnocystal ooecium (same as in Figure 4B). Note the slightly dimorphic opesia of the ovicellate zooid. A teratologic zooid is visible on the left top corner. (Scale bar: 200 µm). (D) Close-up of two zooids in B. The ovicellate zooid (on the right side) shows the vestigial ooecium with associated dimorphic opesia, which is terminally placed and larger compared to that of the non-ovicellate zooid.    fig. 46a-c).

Description
Colony encrusting, multiserial, uni-to paucilaminar, usually small sized (<1 cm 2 ) in the examined material, forming small patches or knobs (Figures 5A and 6A) of self-overgrowing lobes produced by pseudoancestrulae budded frontally by scattered autozooids or, more often, avicularia ( Figure 6B-D) of the outer layer; whitish-beige when alive, with darker spots corresponding to sclerites of the avicularian mandibles.
Autozooids irregularly or quincuncially arranged, large (mean ± SD: 427 ± 28 × 326 ± 34 µm) and thick, slightly longer than wide (mean L/W: 1.31); generally ovoidal but often rounded polygonal and arched distally; zooidal boundaries raised and outlined by narrow grooves (Figures 4A and 5A,B-D, Figure 6A,B,E). Gymnocyst forming the lateral walls, visible frontally only on zooidal proximal corners, mostly in irregularly growing areas ( Figure 5E,F and Figure 6D). Cryptocyst (Figures 5 and 6) extensive, occupying the proximal half of the frontal surface, and forming two wings projecting laterally into the opesia at about mid-length, in a few instances tapering gradually; absent distally; depressed, gently sloping from the zooidal rim, flat centrally; coarsely and evenly granular, except for the sloping margins where granules are smaller and more densely packed.
Kenozooids, as rounded tubercles (mean diameter ± SD: 54 ± 11 µm), sporadically occurring at the triple contact between zooids ( Figure 4A,B and Figure 5E-G); often absent in large sectors of the colony.
Ancestrula, similar to later autozooids but smaller; presence uncertain in this material (see putative ancestrula in Figure 5A).

Remarks
Based on the morphological characters, colonies from the submarine caves of Lesvos fit well in the definition of Onychocella vibraculifera, but the size of the polymorphs is remarkably smaller compared to polymorph size reported in Prenant & Bobin [12] and Chimenz Gusso et al. [16] (e.g., autozooids: 520-640 × 420-480 µm).
Onychocella vibraculifera differs from O. marioni in having flatter cryptocyst, a smaller and more bell-shaped, terminal opesia with straight or slightly convex proximal border. Some authors sinonymised O. vibraculifera with O. angulosa Reuss, 1847 (see [12]) but the latter species lacks the interzooidal tubercles and differs also for some other morphological and morphometric characters.

Distribution
Onychocella vibraculifera is an endemic Mediterranean species. It has been recorded from Tunisia [25] and the Sicily Strait [11], Aeolian Islands in SE Tyrrhenian Sea [16], SW Turkey associated with Posidonia rhizomes [16,50], Chios Island in Greece [15], and the coasts of Lebanon [18]. Dead colonies were also reported from submarine caves along the Ionian coast of Sicily [23,28], while both living and dead colonies were reported from the semisubmerged Accademia cave at Ustica Island, S Tyrrhenian Sea [43,54,55]. The specimens described here fall within the known geographical distribution of O. vibraculifera, which is restricted to the eastern and southern sectors of the Mediterranean Sea. This species occurred in the Mediterranean basin (central Italy) at least since the Pleistocene [51]. Its Pliocene record needs to be confirmed following the age updating of the deposits.  . 121B), [60], and especially Coronellina atlantica Souto, Reverter-Gil & Ostrovsky (2014) from Madeira, the latter species also sharing disjointed zooids with tubular connections [61] (figs. 4 and 5). However, because no established genera in the heterogeneous family Microporidae nor in Onychocellidae appear suitable for S. disjuncta, the introduction of a new genus, Bryobifallax gen. nov., was considered necessary. The family placement is also challenging because Bryobifallax gen. nov. shares features with both Onychocellidae and Microporidae. In the latter family, zooids communicate through basal pore chambers or multiporous septula, have extensive pseudoporous cryptocyst pierced by distolateral opesiules, sometimes producing opesiular indentations in the proximal border of the semicircular opesia (e.g., [59]). Furthermore, avicularia, when present, are small and usually interzooidal. However, Microporidae is here preferred to Onychocellidae because of the affinities in ovicell development.  and in colony portions encrusting particularly irregular surfaces; slightly smaller or larger than autozooids and irregularly shaped, with extensive, granular cryptocyst with granules somewhat arranged in radial rows, and a median, subcircular to drop-shaped opesia ( Figure 8C). Ancestrula not observed. Regeneration of autozooid via intramural budding ( Figures 7A,B and 8E,J); kenozooids with median roundish pores budded within autozooids and ovicells ( Figure 8I,J) relatively common.

Diagnosis
Colony encrusting, multiserial, unilaminar, anchored to the substratum through tubular extensions. Autozooids disjointed, connected by short tubes; longer than wide, ovoidal to diamond-shaped, with a raised margin. Gymnocyst extremely reduced. Cryptocyst granular, extensive proximally and surrounding the opesia, depressed. Opesia longer than wide, semielliptical to subtrapezoidal, with a straight or gently arched proximal border; dimorphic and terminal in ovicellate zooids. Muscle scars symmetrical, visible through the opesia on the autozooidal floor. Operculum small, dimorphic. Spines absent. Ovicell subimmersed, hemisphaeric, slightly convex and gently sloping distally; the frontal surface formed by a swelling of the distal autozooid, mainly cryptocystal but with a proximal, narrow band of gymnocystal calcification arched above the opesia. Avicularia rare, vicarious, large, flame-shaped, and symmetrical, with extensive cryptocyst, pear-shaped opesia and elongate triangular rostrum with raised, gymnocystal laminae. Mandible with a straight central sclerite mirroring the shape of the rostrum. Kenozooids rare, irregularly shaped, with an extensive cryptocyst and a median small opesia.

Remarks
The new genus Bryobifallax is here introduced for Smittipora disjuncta (Canu & Bassler, 1930). This species was first placed in Rectonychocella Canu & Bassler, 1917, because of its symmetrical vicarious avicularia, and subsequently still mentioned as Rectonychocella [12,15] or included in Smittipora Jullien, 1882 [3,14,17,56,57]. These two genera, which share the presence of symmetrical vicarious avicularia, were considered synonyms for long time (e.g., [13]). Taylor et al. [10] clarified the differences between the two genera: Rectonychocella has large, ovoidal opesiae, while Smittipora has smaller, semielliptical opesiae with opesiular indentations. However, the type species (and other congeners) of both Smittipora and Rectonychocella, and onychocellids in general (see [10,13]), have immersed ovicells, barely visible, associated with dimorphic autozooids showing larger, cormidial opesia, in contrast with the subimmersed, escharelliform sensu Ostrovsky [58], ovicells of S. disjuncta. The occurrence of a more prominent ovicell compared to typical onychocellids, reported as "hyperstomial", was first noted by Harmelin [14] and later Hayward [15] when fertile colonies from the Aegean Sea became available. The first description of the species was in fact based on a young colony from the Tunisian coast, only consisting of periancestrular zooids. Harmelin [14] and Hayward [15] . 121B), [60], and especially Coronellina atlantica Souto, Reverter-Gil & Ostrovsky (2014) from Madeira, the latter species also sharing disjointed zooids with tubular connections [61] (figs. 4 and 5). However, because no established genera in the heterogeneous family Microporidae nor in Onychocellidae appear suitable for S. disjuncta, the introduction of a new genus, Bryobifallax gen. nov., was considered necessary. The family placement is also challenging because Bryobifallax gen. nov. shares features with both Onychocellidae and Microporidae. In the latter family, zooids communicate through basal pore chambers or multiporous septula, have extensive pseudoporous cryptocyst pierced by distolateral opesiules, sometimes producing opesiular indentations in the proximal border of the semicircular opesia (e.g., [59]). Furthermore, avicularia, when present, are small and usually interzooidal. However, Microporidae is here preferred to Onychocellidae because of the affinities in ovicell development.

Description
Colony encrusting, multiserial, unilaminar ( Figure 8A), up to ca. 6 cm 2 in the observed material; in large colonies, lobes joining but rarely overlapping; anchored to the substratum through tubular extensions about 60-100 µm long and 20 µm in diameter ( Figure 8G,H), visible through the opesia as pits in the autozooidal floor ( Figure 8F); whitish with yellowish to hazel spots, corresponding to the avicularian mandibles, when dried.
Opesia longer than wide (mean ± SD: 223 ± 22 × 208 ± 15 µm), semielliptical to subtrapezoidal, with blunt corners and a straight or gently concave proximal border ( Figure 8B-F,J); dimorphic and becoming subquadrangular in ovicellate autozooids ( Figure 8B,D,E). Frontal membrane covering the whole surface in living colonies. Muscle scars symmetrical, reniform to irregularly shaped, placed on the proximal half of the autozooidal floor visible through the opesia. Operculum small, corresponding to less than half the length and the width of the opesia; shorter but wider in ovicellate autozooids (Figure 7). Spines absent.
Ovicell subimmersed, hemispheric, formed by the enlargement and swelling of the proximal part of the distal autozooid (Figures 7 and 8B-E,I); surface slightly convex and gently sloping distally up to about half length of the frontal cryptocyst of the distal zooid, visible as a zone of more densely spaced granules compared to other frontal regions, including a distal cryptocystal endooecium and a proximal gymnocystal ectooecium consisting of a thin, protruding rim arched above the opesia marked by a band of calcification, narrow in the middle and widening laterally, with no evidence of a median suture.
Kenozooids rare, observed at the colony periphery, along the contact between merging lobes and in colony portions encrusting particularly irregular surfaces; slightly smaller or larger than autozooids and irregularly shaped, with extensive, granular cryptocyst with granules somewhat arranged in radial rows, and a median, subcircular to drop-shaped opesia ( Figure 8C). Ancestrula not observed.

Remarks
The distance between zooids and the length of the connecting tubules, as well as the length of the pillar-like structures for adhering to the substratum, already depicted by Harmelin [14] ( fig. 2.4), vary seemingly in relation to irregularities in the encrusted surface. Otherwise, morphological and morphometric differences with the type material, including the concave proximal border of the opesia and the smaller autozooidal and opesial measurements in the single colony described by Canu & Bassler [25] from Tunisia, that only included the ancestrula and some periancestrular autozooids, could express ontogenetic intraspecific variability.

Distribution
Bryobifallax disjuncta comb. nov. seems to be endemic to the Mediterranean Sea. It was described from Tunisian waters, from calcareous concretions associated with material collected by sponge fishers [25]. Further records are occasional and consistently restricted to the eastern sector of the Mediterranean Sea (i.e., Aegean Sea: [17]). In addition to the material from Lesvos Island, colonies of B. disjuncta comb. nov. were also reported from several localities around Chios Island (Cape Masticho: 40-60 m, Venetiko: 15 m, Dhiaporia: 50 m; Kokkina, Emborios bay: 3-15 m) by Hayward [15]. Harmelin [14] found the species on biogenic concretions collected in the southernmost Aegean localities in the Karpathos Strait (60 m) and near Santorini (110-128 m). The species is also known from Lebanon, reported off Tripoli by Harmelin et al. [18].  9 and 10, Table 3) 648 ± 70 × 418 ± 45 μm), distinct by narrow and deep furrows; irregularly polygonal proximally, arched distally (Figures 10 and 11A), communicating through a row of septular pores in the vertical walls ( Figure 10F). Marginal rim raised, mostly distally, regularly and finely beaded. Gymnocyst extremely narrow, only visible along zooidal boundaries and slightly more at proximal corners, occasionally forming low elevations ( Figure 9B). Cryptocyst extensive, depressed, steeply sloping from the margins, generally flat but sometimes swollen centrally and proximally to the opesia; flanking the opesia, tapering laterally, absent distally ( Figures 10C-E and 11A,B); granular with, variably sized, sparse and randomly distributed granules, which become finer and more closely packed if an ovicell is present. Distal margin straight or slightly convex medially, and then flanked by two, usually shallow, opesiular indentations.

Description
Colony encrusting, unilaminar ( Figures 9A and 10A), with the ancestrular zone located at the periphery; fan-shaped becoming somewhat lobate with evidence of lateral regenerations (Figure 11 A), junctions and possible fusions of lobes in larger colonies (about 2 cm 2 ) and self-overgrowth on senescent-dead zooids.
Autozooids quincuncially arranged ( Figure 9B,C, Figures 10B and 11A), thick and large (mean ± SD: 648 ± 70 × 418 ± 45 µm), distinct by narrow and deep furrows; irregularly polygonal proximally, arched distally (Figures 10 and 11A), communicating through a row of septular pores in the vertical walls ( Figure 10F). Marginal rim raised, mostly distally, regularly and finely beaded. Gymnocyst extremely narrow, only visible along zooidal boundaries and slightly more at proximal corners, occasionally forming low elevations ( Figure 9B). Cryptocyst extensive, depressed, steeply sloping from the margins, generally flat but sometimes swollen centrally and proximally to the opesia; flanking the opesia, tapering laterally, absent distally ( Figure 10C-E and Figure 11A,B); granular with, variably sized, sparse and randomly distributed granules, which become finer and more closely packed if an ovicell is present. Distal margin straight or slightly convex medially, and then flanked by two, usually shallow, opesiular indentations.   from the type material) has much smaller autozooids (400 × 200 μm) with distinctive scales on the cryptocyst, a well-developed median process and narrow elongated opesiular indentations (Voigt 1987, fide [63]).
We interpreted the structures protruding from the lateral walls and having a flat roughly annulated upward-facing surface, as possible bases for the attachment of muscles. Similar structures are present in T. arcifera [63] (fig. 24) and in the autozooids of Parastichopora vanna Cook & Chimonides, 1981, for which the authors hypothesised the same function [71].

Distribution
This is the first record of T. arculifera comb. nov. after its original description by Canu fig. 16), always refer to these same colonies. No obvious morphological differences distinguish our specimens from those figured from Hawaii in addition to the variability of the opesia shape (see Remarks). For this reason, we refrain from introducing a new species. While similarities between deep shelf to upper bathyal habitats and shallow-water caves [28,73,74] may explain the difference in depths between these two records, the great geographical distance is puzzling. More data and possibly phylogenetic analysis are needed either to support the possible transport and introduction of this rare species into the Mediterranean Sea or to reveal a species complex. Opesia large, occupying about half of the frontal surface, semielliptical to subtrapezoidal with blunt corners (Figure 10A-E and Figure 11A), longer than wide (mean ± SD: 282 ± 31 × 263 ± 30 µm). One to three large processes protruding from each side at opesia mid-length (but proximally to the operculum), at the same level as the lateral cryptocyst or more deeply inside the opesia ( Figure 10C,D); usually quadrangular with the flat, frontally-facing surface irregularly to concentrically laminated and etched ( Figure 11C). Occasionally, one or two denticles occur on the distal cryptocystal margin ( Figure 10E). Muscle scars irregularly elliptical, longitudinally elongate or quadrangular, visible very distally through the opesia (located nearly at opercular level) on projections of the lateral walls ( Figure 10D,F). Spines absent. Operculum small, corresponding usually to less than half the length and width of the opesia ( Figure 9B,C).
Ovicell endozooidal, globose but not prominent ( Figure 10D), fully immersed in the thickness of the distal autozooid, formed by a folding of the distal autozooid, lining its frontal surface but leaving a narrow space below its floor, protruding for about the entire cryptocystal length ( Figure 11C); outer surface mostly cryptocystal (and covered by the frontal membrane in living colonies), produced by the distal autozooid and showing a feebly raising, more finely and densely granular cryptocyst than other autozooids; proximal border slightly raised and thickened formed by a markedly developed gymnocystal band. Operculum of the maternal zooid apparently not dimorphic.

Remarks
Our specimens fit well in Floridinella arculifera, as described and figured by Canu & Bassler [63] from Hawaii, although the opesia of the Pacific specimens tends to be more trifoliate, with the proximal border more convex distally and two deeper opesiular indentations than in the Mediterranean colonies. However, Floridinella Canu & Bassler, 1917 (type species F. vicksburgica Canu & Bassler, 1917 from the Oligocene of Alabama, USA) has avicularia, which are missing in this species. Avicularia of F. vicksburgica were observed by Cook & Bock [62] and described as small with triangular rostra and complete crossbar, transversely oriented on ovicells. Furthermore, ovicells are subimmersed in this species, although described [64,65] and sometimes reported (e.g., [66]) as endozooidal.
Based on its endozooidal ovicell and the absence of avicularia, Floridinella arculifera has been placed in Caleschara MacGillivray, 1880 by Cook & Bock [62]. These authors contextually synonymized it with Caleschara minuta (Maplestone, 1909), a species from the Gilbert Islands and with two further Indo-Pacific species, i.e., C. levinseni Harmer, 1926 from the Kei Islands (Moluccas) and C. laxa Canu & Bassler, 1929 from the Philippines. After a careful re-examination of the illustrations and descriptions provided by Cook & Bock [62], as well as Tilbrook [67] (C. minuta), and Gordon [68], who suggested the conspecificity of Caleschara levinseni and C. laxa, we agree to retain the synonymy of Caleschara minuta with C. levinseni and C. laxa but we suggest to reconsider Floridinella arculifera as a separate species. Indeed, the specimen figured by Cook & Bock [62] (fig. 16) lacks the median cryptocystal denticle that is constantly prominent in C. minuta [62] (fig. 15) and, hence, the lateral cryptocystal indentations producing the typical trifoliate opesia. Cryptocystal denticles protruding all along the lateral sides of the opesia are wide in C. minuta but decidedly less developed in F. arculifera from Hawaii and in our specimens from Lesvos. Furthermore, only occasionally denticles have been found at the level with the cryptocyst, whereas they mostly protrude from lateral walls at different heights and all show a flat, frontally-facing surface (see description and morphofunctional comments below). Finally, the cryptocyst is only slightly depressed in relation to the mural rim, and its granules are finer and more densely packed than in C. minuta.
The generic allocation of F. arculifera is challenging. The species shares several characters with both Caleschara and Tretosina Canu & Bassler, 1927 of the family Calescharidae Cook & Book, 2001. Focusing on the external morphology, the absence of a median denticle questions the placement in Caleschara although the oldest representative of Caleschara known to date, from the early Eocene of the Chatham Islands, lacks a median cryptocystal denticle [69]. However, the median cryptocystal denticle typically occurs in species of Caleschara, sometimes forming an extensive shelf leaving only a small, semielliptical opesia with long, denticulate opesiular indentations, as in the genotype C. denticulata (MacGillivray, 1869) (see [62]). Cryptocystal median denticles are missing in two species of Tretosina (i.e., T. moderna Cook, 1985 from present-day West Africa, and T. flemingi (Brown, 1952) from the Pliocene of New Zealand), but in the type species T. arcifera Canu & Bassler, 1927 from the Miocene of Victoria (Australia) it is inconstant and very small. Furthermore, both genera possess a vertical lamina descending from the cryptocyst and separating the internal autozooidal space in two compartments (see [62]) (figs 22,24). This lamina is absent in the Lesvos specimens as it is in T. moderna and T. flemingi. For all the above reasons, we suggest the new combination Tretosina arculifera comb. nov. This is the first record of the genus Tretosina and the family Calescharidae from the Mediterranean and European waters. Only a fossil Danian to Montian species, Caleschara squamosa (Meunier & Pergens, 1886), has been reported from Belgium [70], but this fossil species (only known from the type material) has much smaller autozooids (400 × 200 µm) with distinctive scales on the cryptocyst, a well-developed median process and narrow elongated opesiular indentations (Voigt 1987, fide [62]).
We interpreted the structures protruding from the lateral walls and having a flat roughly annulated upward-facing surface, as possible bases for the attachment of muscles. Similar structures are present in T. arcifera [62] (fig. 24) and in the autozooids of Parastichopora vanna Cook & Chimonides, 1981, for which the authors hypothesised the same function [71].

Distribution
This is the first record of T. arculifera comb. nov. after its original description by Canu & Bassler [63]. The original finding is dated July 1902; a few colonies, some of which alive and fertile, were collected at the depth range of 91-113 m off Hawaii and 142-406 m off Molokai Island, in coral habitats at 20.6 • C. Subsequent citations, including Winston [72] (p. 7), Tilbrook [67] (p. 72, 73) and Cook & Bock [62] (p. 536, fig. 16), always refer to these same colonies. No obvious morphological differences distinguish our specimens from those figured from Hawaii in addition to the variability of the opesia shape (see Remarks). For this reason, we refrain from introducing a new species. While similarities between deep shelf to upper bathyal habitats and shallow-water caves [28,73,74] may explain the difference in depths between these two records, the great geographical distance is puzzling. More data and possibly phylogenetic analysis are needed either to support the possible transport and introduction of this rare species into the Mediterranean Sea or to reveal a species complex.

Discussion
Examination of the bryozoan component from two submarine caves of Lesvos Island, NE Aegean Sea, confirmed the occurrence of two out of the three species of Onychocella known to date from this basin, i.e., O. marioni and O. vibraculifera. The urge for a revision of the type material of both O. marioni and O. angulosa (the third Mediterranean species) to ascertain their conspecificity is remarked once more because necessary to clarify the diversity and distribution of this genus in the Mediterranean, as well as in the near Atlantic from where both species have been reported [13,29]. The proposed new combination, Bryobifallax disjuncta comb. nov., and the placement of this species in Microporidae further decreases the number of species and genera of Mediterranean onychocellids.
It is worth noting that, with the exception of O. marioni, all these species are restricted to the Mediterranean Sea. Onychocella vibraculifera and B. disjuncta comb. nov. are endemic and restricted to the eastern sector of the basin with the exception of the first described colony [25].
Onychocella is a long-living genus with several species known from the Mediterranean area and European regions since the Cretaceous [57], but a reliable stratigraphic distribution of the species here treated, possibly going back to the late Miocene and the Pleistocene for O. marioni and O. vibraculifera respectively, remains to be established (see Remarks for each of these species and [29]).
Tretosina arculifera comb. nov. increases to four the number of species now assigned to this genus, which also includes the Miocene T. arcifera from Australia, the Pliocene T. flemingi [75] from New Zealand, and the Recent T. moderna Cook, 1985 from west Africa [33]. The inclusion of Tretosina arculifera comb. nov. within Tretosina and its finding in the eastern Mediterranean after its historical record from Hawaii [63], widen the present-day geographical distribution of this genus previously only reported from west Africa [34]. The record of this species and genus in the Mediterranean also widens the geographical distribution of Calescharidae previously restricted to the Atlantic (T. moderna) and the Indo-Pacific. However, because information on these taxa is still too fragmentary, any biogeographical hypothesis would be speculative.
A morphological/developmental feature common to all the species/specimens described here is the presence of successive intramural buds affecting both autozooids and vicarious avicularia. Subsequent intramural buds were interpreted either as evidence for high predation pressure [76] or as an effect of the ageing process [6,77]. In submarine cave habitats like those studied here, the recycling of existing modules seems to be linked to changes in nutrient availability, with induced senescence during phases of lower nutrient levels alternating with intramural budding during phases of higher nutrient levels [6,76].