Sexual reproduction of the placental brooder Celleporella hyalina (Bryozoa, Cheilostomata) in the White Sea

Abstract The evolution of parental care is a central field in many ecological and evolutionary studies, but integral approaches encompassing various life‐history traits are not common. Else, the structure, development and functioning of the placental analogues in invertebrates are poorly understood. Here, we describe the life‐history, sexual colony dynamics, oogenesis, fertilization and brooding in the boreal‐Arctic cheilostome bryozoan Celleporella hyalina. This placental brooder incubates its progeny in calcified protective chambers (ovicells) formed by polymorphic sexual zooids. We conducted a detailed ultrastructural study of the ovary and oogenesis, and provide evidence of both auto‐ and heterosynthetic mechanisms of vitellogenesis. We detected sperm inside the early oocyte and within funicular strands, and discuss possible variants of fertilization. We also detail the development and functioning of the placental analogue (embryophore) in the various stages of embryonic incubation as well as embryonic histotrophic nourishment. In contrast to all known cheilostome placentas, the main part of embryophore of C. hyalina is not a single cell layer. Rather, it is a massive “nutritive tissue” whose basal part is associated with funicular strands presumably providing transport function. C. hyalina shows a mixture of reproductive traits with macrolecithal oogenesis and well‐developed placenta. These features give it an intermediate position in the continuum of variation of matrotrophic provisioning between lecithotrophic and placentotrophic cheilostome brooders. The structural and developmental differences revealed in the placental analogue of C. hyalina, together with its position on the bryozoan molecular tree, point to the independent origin of placentation in the family Hippothoidae.

in the boreal-Arctic cheilostome bryozoan Celleporella hyalina. This placental brooder incubates its progeny in calcified protective chambers (ovicells) formed by polymorphic sexual zooids. We conducted a detailed ultrastructural study of the ovary and oogenesis, and provide evidence of both auto-and heterosynthetic mechanisms of vitellogenesis. We detected sperm inside the early oocyte and within funicular strands, and discuss possible variants of fertilization. We also detail the development and functioning of the placental analogue (embryophore) in the various stages of embryonic incubation as well as embryonic histotrophic nourishment. In contrast to all known cheilostome placentas, the main part of embryophore of C. hyalina is not a single cell layer. Rather, it is a massive "nutritive tissue" whose basal part is associated with funicular strands presumably providing transport function. C. hyalina shows a mixture of reproductive traits with macrolecithal oogenesis and well-developed placenta. These features give it an intermediate position in the continuum of variation of matrotrophic provisioning between lecithotrophic and placentotrophic cheilostome brooders. The structural and developmental differences revealed in the placental analogue of C. hyalina, together with its position on the bryozoan molecular tree, point to the independent origin of placentation in the family Hippothoidae.

K E Y W O R D S
colonial aquatic invertebrates, life-history, matrotrophy, oogenesis, placental analogue
The first comprehensive analysis of EEN across the animal kingdom revealed that this phenomenon is established or inferred in at least 21 of 33 animal phyla (Ostrovsky et al., 2016). This number significantly exceeds previous accounts and contradicts the traditional view that matrotrophy is infrequent among invertebrates (see Avise, 2013;Clutton-Brock, 1991;Hogarth, 1976;Trumbo, 2012). Else, the analysis of the distribution and diversity of matrotrophic adaptations (both structural and physiological) in Animalia estimated 140-145 independent origins of this phenomenon (Ostrovsky et al., 2016).
Matrotrophy is associated with all known types of incubation chambers, or performed without them and using five nutritive modes: histotrophy, placentotrophy, oophagy, embryophagy and histophagy, of which the first and the second are the most widespread (Ostrovsky et al., 2016). Nutrient delivery and uptake are performed using secretion, active transport across membranes, facilitated diffusion, endocytosis (pino-and phagocytosis) as well as ingestion of parentally derived nutritive material and sometimes of germ and parental somatic cells. Overall, invertebrate matrotrophic adaptations are less complex structurally than in vertebrates (and chordates, in general), but they are extraordinarily diverse in respect to the sites, modes, mechanisms and structures involved. Despite the current progress in our understanding of this diversity, only few matrotrophic invertebrates have been studied ultrastructurally. This impairs comparative and evolutionary analyses.
The entirely colonial, lophotrochozoan phylum Bryozoa has the widest taxonomic distribution of placental analogues among aquatic invertebrates (Ostrovsky et al., 2016). Among three bryozoan classes, placentation is presumably characteristic to all representatives of Stenolaemata and Phylactolaemata, and is common in the class Gymnolaemata. The distribution patterns as well as the differences in the structure of incubation chambers, in the cell source, position and anatomy of the placental analogues in different clades indicate at least 23 independent origins of matrotrophy within Bryozoa. This makes this phylum an exceptional model to study trends in the evolution of matrotrophy in animals (Ostrovsky, 2013a(Ostrovsky, , 2013bOstrovsky, Gordon, & Lidgard, 2009;Reed, 1991;Ryland, 1976).
The overwhelming majority of independent transitions to EEN occurred within the gymnolaemate order Cheilostomata. This type of nutrition occurs either in internal brood sacs or inside external calcified brood chambers-ovicells (Ostrovsky, 2013a). The opening of the ovicell is normally plugged by the specialized outgrowth of the membraneous wall of the fertile zooid (termed an ooecial vesicle) that in matrotrophic species bears an embryophore, that is, a placental analogue providing nourishment for the embryo. An active embryophore consists of hypertrophied epithelial lining and associated funicular tissue (Hughes, 1987;Moosbrugger, Schwaha, Walzl, Obst, & Ostrovsky, 2012;Woollacott & Zimmer, 1972, 1975. In the internal brooders, the entire wall of the brood sac becomes an embryophore. At present, placental analogues have been recorded in 21 cheilostome species belonging to 10 families (Ostrovsky, 2013a(Ostrovsky, , 2013bOstrovsky et al., 2009), but only three species of two families were studied ultrastructurally (Hughes, 1987;Moosbrugger et al., 2012;Woollacott & Zimmer, 1975). Moreover, sexual reproduction in most placental bryozoans has been studied only fragmentarily (reviewed in Ostrovsky, 2013a).
This study focuses on the reproductive biology of the common boreal-Arctic cheilostome Celleporella hyalina (Linnaeus, 1767). It demonstrates a prominent example of placentation due to its specialized sexually polymorphic zooids. Colonies of this species are simultaneous hermaphrodites comprising feeding autozooids and sexual male and female autozooidal polymorphs that are unable to feed. Embryos are brooded in the ovicells of the female zooids and are supplied by a well-developed placental analogue. As females do not feed, the EEN is provided by the neigbouring autozooids via a transport system of funicular strands/cords connected via interzooidal communication pores (Hughes, 1987;Ostrovsky, 1998).
C. hyalina has been an object of extensive field and experimental studies (predominantly by Hughes with co-authors) focusing on various aspects of fertilization and sex allocation (Bishop, Manríquez, & Hughes, 2000;Hoare, Hughes, & Goldson, 1999;Hughes, Manríquez, & Bishop, 2002;Hughes & Wright, 2014;Hughes, Wright, Carvalho, & Hutchinson, 2009;Hughes, Wright, & Manríquez, 2002;Hunter & Hughes, 1993, 1995Hunter, Hughes, & Goldson, 1996;Manríquez, Hughes, & Bishop, 2001Pemberton, Hughes, Manríquez, & Bishop, 2003). Another focus has been on life-history traits, including growth and fitness, and their plasticity (Atkinson, Morley, & Hughes, 2006;Cancino, 1986;Cancino & Hughes, 1987Eggleston, 1972;Hughes, 1989Hughes, , 1992Hughes & Hughes, 1986;Hughes, Manríquez, Bishop, & Burrows, 2003;Hughes, Manríquez, Morley, Craig, & Bishop, 2004). In contrast, only four morphological studies on the sexual reproduction of this species have been published. Hughes (1987) investigated the formation of the sexual zooids and ovicells, as well as fecundity, gametogenesis and brooding of C. hyalina from the Irish Sea using histological sections and scanning and transmission electron microscopy (SEM and TEM). The development and structure of the ovicells, along with certain aspects of oogenesis and embryonic incubation, were studied on the specimens from the White Sea by Ostrovsky (1998Ostrovsky ( , 2013aOstrovsky ( , 2013b using SEM and histological techniques. However, both oogenesis and placental nourishment, while providing a comparative basis for our study, were described rather superficially. Certain conclusions were only partially supported, calling for more detailed and broader research. The main focus of this study was on the ultrastructure of oogenesis and the development of the placental analogue along with its functioning on various stages of embryonic/larval growth. We also for the first time report the main life-history traits of this bryozoan species in the White Sea, yielding an integral picture of its sexual reproduction.

| MATERIALS AND METHODS
In the White Sea, colonies of C. hyalina (Linnaeus, 1767) range from the intertidal down to 137 m depth, encrusting various substrates, typically, algae (Gostilovskaya, 1978). We collected bryozoans on kelps (Saccharina latissima species-complex) and red algae (Odonthalia dentata, Phycodrys rubens, Coccotylus truncatus) during the ice-free period from 5-10 m depth by boat dredging and SCUBA-diving near the Educational and Research Station 'Belomorskaia', Saint Petersburg State University (Chupa Inlet, Kandalaksha Bay, White Sea).
To study the life-history, the random sampling was performed in 2012 and 2014 (Supporting Information Tables 1 and 2). Vast majority of the colonies were collected between May and September, 2014.
Altogether the state of 1,003 colonies was examined using qualitative parameters, that is, colony shape and relative size, zooidal performance (feeding, budding and polypide degeneration), and presence of sexual polymorphs and embryos in them. Recording of these parameters in different months allowed recognition of main lifehistory traits, number of generations, colony sexual dynamics and lifespan and timing of reproduction.
Alive colonies were photographed with a digital camera Leica DFC295 attached to a Leica M205C stereomicroscope.
For anatomical studies, colonies were collected in 2013 and 2015 (Supporting Information Table 3). They were fixed and decalcified in Bouin's fluid. After dehydration in ethanol series (30-50-70-80-90-96%) they were embedded in resin (Epon 812), sectioned (2.0 μm thick) and stained by Richardson's stain by standard methods (Richardson, Jarrett, & Finke, 1960). Images were made with a Nikon DS-Fi1 photocamera attached to a Leica DM2500 stereomicroscope. Altogether, ovaries from 78 zooids from five C. hyalina colonies were studied. Total preparations of some colonies fixed either in the Bouin's fluid or in 70% ethanol were made after dehydratation and embedding them in epon. They were photographed with a Leica DFC420 photocamera (Leica Microsystems, Wetzlar, Germany) attached to a Leica M205C stereomicroscope to estimate the colony size and the number of female polymorphs.
For ultrastructural studies of oogenesis and placentation 20 colonies were collected in 2013, 2016 and 2017. They were fixed in 2.5% glutaraldehyde (in 0.1 mol L −1 cacodylate buffer with 10% sucrose, pH 7.4) for 3 hr and subsequently rinsed three times in the buffer.
Postfixation was done in a 1% solution of osmium tetroxide (OsO4) in the buffer solution for 1 hr followed by three rinses in the buffer.
Ultrathin sections of 60 nm thickness were placed on the copper grids and contrasted with 2.5% gadolinium triacetate and 3% lead citrate.
Sections were examined with a Zeiss Libra 120 transmission electron microscope (Zeiss, Jena, Germany) and photographed with a digital CCD Olympus Morada G2 (11 MP, in column) camera.
Characteristics of the sexual reproduction in the colonies collected in different years did not differ.

| Life-history and colony sexual dynamics
In the studied population, two age groups were easily distinguished by their appearance: old overwintered colonies formed in the previous year/ice-free period and young colonies formed during the current year ( Figure 1, Supporting Information Tables 1 and 2). The former were characterized by the low transparency of their skeleton, the frequent presence of epibiotic microalgae, infusorians and hydrozoans, and the irregular shape of the colony consisting of the old deteriorating and new budding parts. They were recorded from May to September.
Young colonies were patch-like, with more transparent zooidal walls without microfoulers. They were encountered from June to September being represented by two or, highly likely, three generations. Larval production occurred from June to September and involved both, old and young colonies (Figures 1-2).
Overwintered colonies, collected in May, were inactive without any sign of feeding, budding or reproduction. Only brown bodies (degenerated polypides) were visible through the zooidal walls in some zooids. These overwintered colonies resumed a peripheral growth and began or resumed reproduction in June, giving rise to the small colonies of the daughter (second) generation that appear on algae (Figures 1-2, Supporting Information Table 2). During summer, their old (overwintered) areas were gradually destroyed (mainly in the colony center), but their newly formed parts (often resembling peripheral subcolonies; Figure 2c) continued growth and larval production until late August, and possibly, early September. We did not find evidence of their second hibernation and suppose them to die in winter.
Young colonies appeared in the studied population from June to late September (Figures 1-2a, Supporting Information Tables 1 and 2).
They were actively growing, representing the second and third (and,  Table 2). Throughout September, most colonies of the 'summer (young) generations' grew and reproduced, while others were apparently preparing for dormancy: a few colonies were found that did not grow or feed at the end of that month, and their autozooids possessed either brown bodies or degenerating polypides. Some of them might be dead.
On establishment, the young colonies of C. hyalina are sterile and consist of one layer of autozooids further added by a few additional basal male autozooidal polymorphs. Frontal budding of both, male and female sexual polymorphs, changes male colonies to hermaphrodites ( Figure 2d). Sperm production ends earlier, making colonies female at the end of the reproductive period. In autumn, female gonads are also resorbed, and colonies become sterile again. Overwintered colonies first resume budding of basal autozooids, followed by frontal sexual zooids, thus repeating the same sequence as in young colonies. No repeated establishment of the ovaries in the overwintered female zooids was detected, and their ovicells did not contain embryos.

| Ovary: Development and structure
The earliest germ cells (oogonia) were detected in the young female polymorphic zooids with developing ovicells. They were round or oval, being distinguished from somatic cells due to their markedly larger size (10.0-23.3 × 13.3 μm). The division of the oogonium results in either a pair of oogonia (soon separated from each other) or an early oocyte doublet whose cells are interconnected via a cytoplasmic bridge (see below). Early female gonads contained 2-5 non-paired and dwarf females with ovicells-empty and containing growing embryos of various sizes (outlined by red lines). Ripe oocyte ready for oviposition is shown by arrowhead. Larger autozooidal apertures are interspersed between female polymorphs oogonia and/or one (either oogonial or early oocyte) doublet in our material. Ovaries always had an irregular shape and were suspended in the zooidal coelomic cavity on funicular cords or, sometimes, positioned on the epithelial lining of the zooidal wall (Figures 3 and 4). Germ cells were surrounded by a thin layer of small flattened mesothelial cells (Figures 3a and 4a). One of the funicular cords is connected via a communication pore to the underlying autozooid ( Figure 3d).  and a large nucleolus (average diameter 5.3 μm). Their cytoplasm is electron-dense, containing numerous free ribosomes, some mitochondria and single RER cisternae. Moreover, single cisternae of the smooth ER and Golgi apparatus were detected.
In early vitellogenesis, siblings are still of the same shape, size and have a similar ultrastructure. Small yolk granules (lipid droplets and protein platelets) begin to form in both cells (Figures 4b and 6c). The oolemma is smooth or slightly convoluted (Figures 6b and 7a  Placental analogue (embryophore) is in the early stage of its development (arrow shows fertilization envelope surrounding embryo that partially occupies the brood cavity); ovary is not in the plane of sectioning. (b) Advanced embryo occupying most of the brood cavity. Embryophore is well-developed. Ovary with the early vitelogenic oocyte is in the right part of the female zooid (its nurse cell is out of the section plane). Sclerite is clearly seen in the upper part of the distal zooidal wall in (a) and (b, shown by arrow). (c) Early larva in the ovicell. Embryophore occupies almost half of the female zooid; vitellogenic and previtellogenic oocyte doublets are seen in the ovary in right part of zooid. In all zooids embryophore is developed in association with the distal zooidal wall plugging the entrance to the brood chamber. Abbreviations: a = ascus; ac = coelom of basal autozooid; bc = brood cavity; cu = cuticle of distal zooidal wall; e = embryo; ep = embryophore; fc = coelom of female zooid; fs = funicular strand; o = oocyte in ovary; oc = coelom of ooecium; oe = ooecium (protective outfold of the ovicell); om = opercular muscles; op = operculum; p = rudimentary polypide; pd = previtellogenic doublet 3.4 | Development of the placental analogue and changes in the embryonic epithelium during incubation Each female polymorph containing an ovary is associated with the brood chamber (ovicell) consisting of the spherical protective capsule (ooecium) enveloping the brood cavity, and membranous distal wall of the female zooid plugging the entrance to this cavity. Female polymorphs had 2-4 tentacles and no digestive tract. Polypide retractor and occlusor muscles of the operculum as well as ascus with parietal (dilator) muscles and muscles of the distal wall are well-developed (Figures 3, 5, and 8b,c).
Embryo growth and development takes place in the ovicell. The zygote as well as the young embryo are noticeably smaller than the incubation cavity, being freely suspended in its fluid (Figures 2d and

| Early developmental stage
Brooded embryos and larvae are surrounded by a thick fertilization envelope (Figures 5a, 8e, and 10b). Initially it adjoins the blastomeres of the early embryo, further retreating from them and leaving a substantial space between the envelope and the embryo/larva. The fertilization envelope consists of a thinner, electron-dense external and a thicker, loose internal (lower) layer.
In the early embryo, the peripheral blastomeres have a slightly convoluted plasmalemma and show no signs of endocytosis. Instead, their cytoplasm is filled with large and numerous yolk granules (Figure 7e).
Oviposition and the onset of embryogenesis coincide with the development of the placental analogue in the distal wall of the maternal zooid, whose cells start to grow and divide (Figure 5a). In this process, the initial epithelial lining consisting of cuboidal and prismatic cells   (Figure 8a, b). The cells of the embryophore, some interspersed with the muscles of the zooidal wall, are interconnected by elongated processes constituting a 'loose layer'.
These infoldings are larger and more numerous in the adjacent areas of the neighbour cells. Putative nutrients accumulate on the outer surface of the embryophore cuticle as a thin dark layer of the flocculent material that also spreads into incubation cavity (Figure 8a,d,e).
Some irregularly shaped and sometimes folded cells of the funicular cords (hereafter termed funicular cells) with electron-translucent cytoplasm and lobate nucleus are seen in the proximal part of placental analogue (Figures 8b and 11a). Nutrient-storage cells are visible among the nutritive and funicular cells in the proximal part of the embryophore.

| Mid-developmental stage
When the peripheral embryonic cells begin to develop cilia, the embryophore cells simultaneously greatly increase in size and number; most of them become trapezoid or fusiform and are oriented perpendicular to the distal zooidal wall (Figures 3d and 5b). Cell layers or groups are not recognizable. Instead, the placental analogue is a massive and complex nutritive organ composed of tightly-packed, large cells (Figure 9a-c). Not all of them seem to contact the cuticle of the distal zooidal wall. The muscular bands of the distal wall are embedded in the embryophore.
During growth, the electron density of the cytoplasm of the nutritive cells increases, as does the number of mitochondria and various inclusions, that is, large, round or oval electron-dense granules and smaller vesicles with grayish content. Nutritive cells show a strongly developed synthetic machinery including multiple free ribosomes and numerous cisternae of RER that become longer and more regularly arranged (often stacked). Noteworthy, such cisternae are predominantly situated basally in those nutritive cells that are adjacent to the cuticle of the embryophore (Figures 9a and 10a). In others, apart of the nucleus, the cisternae fill most of available cytoplasm (Figure 9a-

| Embryophore after incubation
After larval release, the placental analogue collapses. Both, nutritive and funicular cells become smaller and fewer, and intercellular spaces appear and expand between them (Figures 3e and 12). Their nuclei remain large ing. Breakage of the substrate is a limiting factor for these colonies, but they can potentially live 1-2 months longer on red algae. Our estimations fit well to the experimental data: colonies of C. hyalina were maintained up to 18 months with repeated reproduction cycles on artificial substrata in the Irish Sea (Cancino & Hughes, 1987). Data on its life-span on, for example, stones or shells, are absent, however.
Thus, although C. hyalina often dominates on ephemeral substrates (Cancino, 1986), at least some colonies living on algae are not ephemeral. Also, our observations, rather than showing a succession, revealed the co-existence of at least three (but more likely four) generations in the White Sea, a situation that presumably exists in the Irish Sea as well. We should add here that a number of genetic studies demonstrated that C. hyalina is a complex of cryptic species   Eggleston (1972) noted that the life-cycle of epibiotic species must be adapted to that of their living substrate. In those C. hyalina colonies that inhabit ephemeral substrates, the ability for early maturation is apparently such an adaptation (Cancino & Hughes, 1987;Hughes, 1989). Indeed, we found the colonies consisting of only 10 autozooids and 1-4 female polymorphs in June-July and, sometimes, in August. Such a very early start of reproduction is also known in other bryozoans, both brooders and broadcasters, living on algae (Bernstein & Jung, 1979;Yoshioka, 1982;Nekliudova, unpubl. data).
Elsewhere, early sexual maturation is characteristic for interstitial bryozoans (Håkansson & Winston, 1985;Winston & Håkansson, 1986), suggesting that life in unpredictable conditions generally promotes early larval production. That strategy can be viewed not only in connection with potential risks (e.g., substrate destruction, Hughes, 1989), but also under favourable conditions when abundant food allows allocating energy to reproduction soon after colony establishment (Nekliudova, unpubl. data). Dyrynda and Ryland (1982) were the first to report that placentation is characteristic for species with ephemeral colony parts. They suggested (although incorrectly argued, see Ostrovsky, 2013a) that placentation could provide faster larval production, enabling more offspring to be released in a shorter time. Ostrovsky supported, but transformed, this idea, arguing that placental brooders combine shorter oogenesis with simultaneous embryonic growth and development during incubation (Ostrovsky, 2013a;Ostrovsky et al., 2009Ostrovsky et al., , 2016. For example, the entire reproductive cycle, from oocyte formation to larval release, takes 4 weeks in Callopora dumerilii (Silén, 1945) and 6 weeks in Chartella papyracea (both non-placental brooding cheilostomes; Dyrynda & King, 1983). In contrast, this period was only 3 weeks in the matrotrophic Bugulina flabellata (Dyrynda & King, 1983;Dyrynda & Ryland, 1982) and B. simplex (Grave, 1930;Ryland, 1974) (both as Bugula). Such faster reproduction could be especially effective in 'seasonal' seas, enable faster occupation of vacant niches after, for example, overwintering. In C. hyalina Hughes (1987) observed that the duration of one reproductive cycle was 14-17 days (comparable with the mentioned placental bugulids). In contrast, Cancino and Hughes (1988) reported 3-4 weeks for embryonic development alone, and this difference could be explained by seasonality in the Irish Sea.
Because the large part of the studied populations of C. hyalina is represented by short-living colonies, this species potentially could use the advantages of placental strategy. Its colonies inhabit a large spectrum of substrates, both stable and ephemeral, organic and not (Gostilovskaya, 1978;Hayward & Ryland, 1999;Kluge, 1975). Considering this, we speculate that rapid larval production supported by placentation is an important factor explaining the success of this species in boreal and Arctic seas.

| Fertilization
In male polymorphic zooid, the polypide consists of a functional lophophore without a digestive tract. There is also a system of organs (polypide retractor muscle, the occlusor muscles of the operculum and parietal muscles that expand the large hydrostatic sac (ascus) (Hughes, 1987; descriptions by Marcus (1938) do not belong to C. hyalina) that serves for the tentacle protrusion followed by the sperm release (Cancino & Hughes, 1988;Hoare et al., 1999;Manríquez et al., 2001). Due to the presence of a similar system of organs for polypide excursion in the female zooids, the female lophophore should be functional, enabling sperm capture and entry (as well as oviposition to the ovicell) (Ostrovsky, 1998; our data; but see Hughes, 1987). As in other cheilostomes, sperm presumably enters the female coelom via the supraneural coelomopore and precociously fuse with the early (previtellogenetic) ovarian oocyte (Bishop et al., 2000;Ostrovsky, 2013a;Ostrovsky & Porter, 2011;Temkin, 1996; our data).
Spermatozoa between two follicle cells in the ovary of C. hyalina were first reported by Hughes (1987). Later, they were found in the ovaries between the follicle cells as well as in the previtellogenic and vitellogenic oocytes (Ostrovsky, 1998(Ostrovsky, , 2013a. In all these cases, the logical route of sperm toward the ovary is through the coelomic cavity. In this respect, our finding of sperm between funicular cells near the ovary is of interest. Could sperm also use funicular cords for this purpose? In Celleporella sp. (as C. hyalina), Marcus (1938) found sperm in all three zooidal types, including autozooids. Elsewhere, it was recorded in the cavity of the ooecium (protective capsule of the ovicell) and in the coelom of an incipient female polymorph that had no vestibule yet (Ostrovsky, 1998). Moreover, experiments showed that alien sperm could be stored by small colonies (three autozooids) for several weeks and used only when the female polymorphs develop (Hughes, Manriquez, & Bishop, 2002). Therefore, in all these cases, the sperm, once caught, somehow travel through the colony. Marcus (1938) suggested that communication pores were the pathway, but this was questioned by Hughes (1987) because of the presence of the pore-cell complexes plugging these pores (see also Ostrovsky, 1998Ostrovsky, , 2008Reed, 1991). In contrast, Hughes et al. (2002a) speculated that the funicular strands could be used for sperm translocation, not considering the fact that these strands are interrupted by the pore-cell complexes. Finally, Ostrovsky (2013a) suggested migration via budding sites prior to the completion of transverse walls (and, thus, communication pores and their cell plugs) between autozooids and female polymorphs.
The current finding, which seemingly supports the suggestion of , is puzzling. Although the central lumen is present in the cheilostome funicular cords (Carle & Ruppert, 1983), no data are available to indicate that the sperm move from this lumen inside the pore-cell complex, further squeezing between its cells, and thus traveling to the neighbour zooid (see Mukai, Terakado, & Reed, 1997 for discussion). If, however, that is possible, then those cords that lead from the pore to the ovary are a potential route for the sperm.
Another possible explanation is that these sperm in trying to reach the ovarian oocytes inadvertently entered the funicular cords adjacent to the ovary.

| Oogenesis and mechanisms of yolk synthesis
Studies on invertebrate oogenesis generally consider major traits such as the origin of the primordial germ cells and the mode of oogenesis, including mechanisms of yolk synthesis (Aisenstadt, 1984;Raven, 1961;Wourms, 1987). Reed (1991) suggested that epigenetic germ cell formation is characteristic for many colonial invertebrates including bryozoans. This reflects the ability of somatic cells to dedifferentiate into totipotential cells transforming to the primordial germ cells and, thus, possibility of germ cell determination throughout ontogeny. Extavour and Akam (2003) also concluded epigenesis to be the basal mode of germ cell specification in Metazoa, including lophophorates, in which primordial germ cells develop in/from either mesenchyme or peritoneal epithelium during late embryogenesis or postembryogenesis. In Bryozoa, female germ cells appear within the mesothelial lining of the forming polypide bud or zooidal wall. Cells of the mesothelial lining surrounding the germ cells form the follicle wall around growing oocytic doublets (reviewed in Ostrovsky, 2013a). This view is consistent with the finding of the presumed oogonial doublet associated with developing female polypide (Ostrovsky, 1998(Ostrovsky, , 2013a, and with our data on the early ovaria in C. hyalina. Although only a few oocytes ultimately develop into the larvae by a single female zooid, the ovarian germ cells can be numerous (up to 25). This points to excessive oogonia/oocyte production in the studied bryozoan. Such a 'surplus', together with the resorption of some germ cells in the ovary, could be an ancestral condition known in broadcasting species (Hageman, 1983), but has never been reported in the placental cheilostomes that normally produce limited number of the germ cells (Ostrovsky, 2013a).
Three modes of metazoan oogenesis can be distinguished regarding the accessory cells: solitary (oocytes develop without such cells), nutrimentary (oocyte is supported by the special nurse cell[s] of either germ or somatic origin), and follicular (each oocyte develops in a follicle, formed by somatic cells, performing either supportive or nutritive function, or both; Aisenstadt, 1984;Wourms, 1987). Most of the brooding cheilostomes combine nutrimentary and follicular modes, although the nutritive role of the follicle cells has been studied ultrastructurally in just three species (Dyrynda & King, 1983;Moosbrugger et al., 2012). Yet, not all bryozoan brooders possess nurse cells (Ostrovsky, 2013a).
The presence of an intercellular bridge connecting the oocyte and its nurse cell suggests an intimate physiological connection between the siblings in C. hyalina. Hughes (1987) proposed that the nurse cell is a nutrient source for the oocyte during early vitellogenesis. In our opinion, such relationships exist during the entire period of yolk accumulation. Several lines of evidence point to the high absorbing and synthetic activities of this sibling that can send yolk precursors, RNA and ribosomes to the oocyte (discussed also in Wourms, 1987). These include the development of microvilli, signs of yolk synthesis, a large active nucleus in the nurse cell (which grows much faster than the cell itself ), and numerous free ribosomes in its cytoplasm. The same relationships between the oocyte and its nurse cell were suggested in the cheilostome brooders Chartella papyracea, Bugulina flabellata and Bicellariella ciliata (Dyrynda & King, 1983;Moosbrugger et al., 2012). It should be stressed that the complex nature of the intercellular bridge, consisting of cytoplasmic and membraneous areas with tight junctions in C. hyalina has been described for the first time in bryozoans.
The follicular cells enveloping the vitellogenetic doublet actively participate in vitellogenesis, using their strongly developed synthetic apparatus, that is, numerous RER cisternae and free ribosomes, mitochondria as well as Golgi complexes. The clathrin-coated pits in the oolemma, suggest that follicular cells synthesize and release nutrients absorbed by the growing oocyte. Note that the electron density of the cytoplasm differs in different follicle cells, suggesting their different functions (e.g., specialization in the synthesis of different products). Consequently, oogenesis in C. hyalina is a combination of nutrimentary and follicular types, as in most incubating cheilostomes (Dyrynda & King, 1983;Moosbrugger et al., 2012). This contrasts to non-brooding and, at least, one brooding species, in which oogenesis is exclusively of the follicular type (Hageman, 1983;Reed, 1991;Shevchenko, unpubl. data).
The mechanism of vitellogenesis depends on the type of yolk precursors obtained by the developing oocyte, and can be autosynthetic, heterosynthetic (Schechtman, 1955) or mixed (Eckelbarger, 1983).
The development of the microvilli and massive synthetic apparatus (large active nucleus with convoluted membrane, RER cisternae as well as multiple mitochondria and Golgi apparatus) in both cells of the vitellogenic doublet indicate active transport of low weight molecular precursors and autosynthetic vitellogenesis in them (Eckelbarger, 1994). Even though the microvilli and ribosomes are still numerous, the Golgi complexes and RER cisternae become less prominent in the final stages of vitellogenesis, indicating a decrease in autosynthetic activity. At the same time, the strongly developed synthetic apparatus in the surrounding follicle cells and the presence of the coated pits in the oolemma of the oocyte point to heterosynthesis. Thus, the vitellogenesis mechanism is mixed in C. hyalina, like in four previously studied cheilostomes (Dyrynda & King, 1983;Hageman, 1983;Moosbrugger et al., 2012).
Interestingly, we detected neither 'direct bathing' of vitellogenic oocyte in the coelomic fluid, nor 'nutrient-storage cells' in the zooidal peritoneal layer opposite to the apical pole of the mature oocyte as described by Hughes (1987, p. 703). In his Plate VII(a) the oocyte is shown to be covered by a thin yet prominent follicular layer. The The cells with numerous large inclusions reported in our study (that we also termed 'nutrient-storage cells') are more similar to the cells described by Dyrynda and King (1983) in zooids of Bugulina flabellata.
In both cases, these cells were associated with the peritoneum of the cystid wall, funicular cords or gonads and contained large spherical yolk-like inclusions.
The increase in oocyte volume during vitellogenesis estimated in the present study (37.5 times) exceeds the calculations made by Ostrovsky (1998) by almost three times, which we explained by the absence of late oocytes in his material. Accordingly, the oocytes were incorrectly described as microlecithal in C. hyalina (Ostrovsky, 1998), although they are in fact macrolecithal (Ostrovsky, 2013a(Ostrovsky, , 2013bour results). Dyrynda and King (1983) described a fibrous 'primary coat' as a precursor of the 'vitelline envelope' surrounding ovarian vitellogenic oocytes in C. papyracea and B. flabellata. According to their description, the oocyte microvilli are embedded in this coat. In contrast, this structure was not recognized around ovarian oocytes in B. ciliata, although the fertilization envelope surrounding the brooded embryo is easily recognizable in this species (Moosbrugger et al., 2012).
Instead, the oocyte microvilli were described as being embedded in a thick matrix of medium electron-density that is actually very similar to the 'coat' described in the two aforementioned species and the fertilization membrane in C. hyalina. Based on this similarity, we suggest that the above mentioned matrix is a fertilization membrane in B. ciliata, permeable for both low and high weight molecular products delivered by the follicle cells.

| Development and functioning of the placental analogue
In matrotrophic cheilostomes, every brooding episode is accompanied by temporal hypertrophy of the embryophore, which collapses after larval release. Ostrovsky (2013a) recently suggested that the embryo produces signal molecules stimulating placental analogue formation and functioning because the embryophore develops soon after oviposition and ceases synthetic activity and degenerates directly after larval release. In C. hyalina, nutritive cells of the placental analogue seem to be its main synthetic part. This is based on the fact that they are much larger and more numerous, and that synthetic organelles develop extensively during embryogenesis. Judging from their position, nutritive (mostly, fusiform) cells could partly originate from the epithelial lining of the distal zooidal body wall and partly from the peritoneum (although it is very loose in gymnolaemate bryozoans, see Mukai et al., 1997). Currently, in C. hyalina we are unable to distinguish the cells of possibly different origin because they are neither organized in prominent layers nor display any distinction in structure.
In the studied species during early embryonic development, the energy costs are apparently covered by the yolk of the egg. Soon thereafter, the developing embryophore starts secreting nutrients into the incubation cavity seen as electron-dense flocculent material. No pores or channels in the cuticle were recorded, which suggests that the nutritive material passes through it in a soluble state (also suggested by Hughes, 1987). As postulated for Bugula neritina and Bicellariella ciliata, diffusion and the osmotic gradient can be the driving forces moving the dissolved nutritive matter across the cuticle (Moosbrugger et al., 2012;Woollacott & Zimmer, 1975). During the period of active nourishment, the fusiform cells make up most of the placental complex and contain a strongly developed synthetic apparatus (see above). The infoldings formed by the apical membranes of nutritive cells at the early and middle stages of embryo incubation are probably a sign of nutrient secretion. This reflects either a surface increase for transmembrane transport or active exocytosis. These infoldings correspond to similar structures formed by nutritive cells of the bugulid placental cheilostomes (Moosbrugger et al., 2012;Woollacott & Zimmer, 1975). Nonetheless, the extensive arrays of foldings developed by embryophore cells in the studied bugulids during the active phase of placentation are not characteristic for C. hyalina. In the latter, the infoldings are developed not so strongly, and present mainly during the early and middle incubation phase. Vesicles filled with flocculent or homogenous material that replace infoldings of the nutritive cells indicate a shift to another mechanism of exocytosis at the advanced stage.
During, active phase of nourishment, the funicular cells that are the part of the embryophore also increase in number and rearrange.
They form a plexus in the basal part of the placental analogue with cytoplasmic processes passing between fusiform cells. Presence of the nutrient-storage cells could also point to accumulation of nutrients in this zone. Contact of the funicular cords with interzooidal pore-cell complexes indicates their main function as pathways for nutrient transport from the neighbouring feeding zooids. At the advanced stages of embryonic development, numerous mitochondria, Golgi complexes and inclusions with different contents, are present in the funicular cells. This may reflect intensified transport activity. An increased contact of the funicular plexus with the basal surface of the hypertrophied placental epithelium has also been detected in Bugula neritina (Woollacott & Zimmer, 1975).
The maximal enlargement (9-fold) of the embryo during the brooding period in our material was less than Hughes (1987) estimated for the Irish Sea (15-fold) but conforms to the data Ostrovsky (2013a) presented (8.8-fold) for the White Sea. Such variation is not surprising and is known within and between populations in the placental cheilostomes (Marshall & Keough, 2003;Ostrovsky, 2013aOstrovsky, , 2013b. As the oocyte size was the same in all these studies (about 80 μm), placentation apparently determines larval size. We predict that such variability should be common in all matrotrophic bryozoans.  (Dyrynda & King, 1983;Ostrovsky, 2013a). The same structure was documented in the con-familiar B. neritina (Mathew, Schwaha, Ostrovsky, & Lopanik, 2018;Ostrovsky, 2013a). In contrast, in the non-placental brooder Chartella papyracea the follicle wall consists of two cell layers, the lower of squamous and the upper (external) of columnar cells. Moreover, the coverage of the vitellogenic doublet by columnar cells is greater, reflecting the greater demand for yolk in oocyte production. The situation is similar in most non-matrotrophic brooders studied that also possess a prominent 'subovarian' or 'intraovarian' space/zone filled by so-called 'basal' cells and numerous intracellular spaces between them (Ostrovsky, 2013a).

| Ovary versus placenta
Noteworthy, the ovarial structure of C. hyalina is reminiscent of both aforementioned variants. The basal part of its follicle on the midvitellogenic stage includes cuboidal cells, thus resembling nonplacental cheilostomes. Conversely, the mature ovary consists of flattened and squamous cells, thus being more similar to the ovaries in most placental species. Also, although we did not find a sub-ovarian space in the female gonad, we did detect intercellular spaces filled with electron-dense material. Importantly, however, the sub-ovarian zone has been recorded in some placental species although an additional ultrastructural study is required to confirm this issue (Moosbrugger et al., 2012;Ostrovsky, 2013a).
Despite the different contribution of the ovary to vitellogenesis in different species, the structure and function of the follicle cells are basically similar in all studied cheilostomes. Judging from their ultrastructure, they potentially obtain the low weight molecular precursors from the funicular cells contacting the follicle wall, although we found no intercellular contact between them. The coelomic fluid is another possible source. These precursors are clearly partially transported and partially modified intracellularly to more complex products. They are subsequently extruded into the intercellular spaces between follicular cells, as well as between them and the vitellogenic oocyte doublet. The presence of microvilli points that the latter absorbs this material via transmembrane transport added by endocytosis. The same processes obviously hold true for the embryophore, whose nutritive cells are also in contact with both the funicular cells and the coelomic fluid. Because intercellular junctions were not found, the nutritive cells presumably obtain low-weight molecular nutrients by facilitated diffusion from the intercellular spaces between them and funicular cells as well as from the coelomic fluid directly. Further, nutritive cells transform and transport them to the embryo. Because of their similar functions, the follicular cells of the ovary and the nutritive cells of the embryophore, although of different origin, are ultrastructurally comparable.
Similar to the ovaries, the hypertrophy of the embryophore cells varies in placental cheilostomes. This is reflected in species with prominent, moderate and weakly developed placentas, which usually correlates with the degree of development and activity of the follicle epithelium in the ovary. Interestingly, the hypertrophy of embryophore cells does not always correlate with a degree of embryonic enlargement, and some species possess moderately developed, but functionally active embryophore (Ostrovsky, 2013a(Ostrovsky, , 2013b. The general structure and functioning of the placental analogues are similar among cheilostomes, involving multiplication and hypertrophy of the epithelial cells of the body wall lining surrounding the brood cavity, that is, either an internal brood sac or ovicell (Ostrovsky, 2013a). These processes are accompanied by the strong development of the synthetic machinery. Nonetheless, in almost all placental species the embryophore consists of one layer of hypertrophied nutritive cells and associated 'sublayer' of funicular cells. C. hyalina is an exception, having a more complex and massive nutritive organ that includes nutritive, funicular (transport) and nutrient-storage cells. When fully developed it occupies a substantial part of the female zooid cavity.
Only the placental analogue of Costaticella solida shows some similarity with this 'nutritive tissue', but its nutritive cells are less numerous and the funicular cells constitute about a half of embryophore (Ostrovsky, 2013a).
Strict polarity in organelle arrangement in hypertrophied cells of embryophore has been documented for the placental analogues of bugulids (Moosbrugger et al., 2012;Woollacott & Zimmer, 1975). This is also true for the nutritive cells adjacent to the cuticle in C. hyalina, whereas the rest (majority) of such cells do not show this polarity.

| Embryonic nutrient uptake
Surface embryonic cells do not show signs of endocytosis until the stage when cilia are formed in the embryos of C. hyalina. Active transmembrane transport of the low weight molecular products is a probable mechanism of nutrient uptake at this early stage. Microvilli developing on the cell surface during late embryogenesis increase its absorption surface, pointing to active transmembrane transport too. Simultaneously, pinocytotic invaginations, channels and vesicles become visible near the microvilli bases. This makes pinocytosis a key mechanism of nutrient intake during larval development (see also Hughes, 1987).
As in B. ciliata (Moosbrugger et al., 2012), microvilli and pinocytotic canals are formed all over the embryo in C. hyalina (except for some cells devoid of cilia and microvilli), that is, not being restricted to the area adjacent to the embryophore. Accordingly, an uptake of nutritive material occurs around the entire embryonic surface. Importantly, in most placental cheilostomes studied so far, growing embryos are suspended in the much larger brood cavity during most of their development; they are not in contact with the embryophore (Moosbrugger et al., 2012;Ostrovsky, 2013aOstrovsky, , 2013b. This makes histotrophy (absorbotrophy) a major nutritive mechanism during this period. The larva occupies the entire brooding cavity and abuts the embryophore only during the final stage of its development (questioned by Hughes, 1987). Thus, the placenta-like system of the apposed embryo-parent tissues providing physiological exchange (Mossman, 1937) 'formally' exists only during the last stage of incubation. Note here that it must provide bidirectional transport of substances, also removing wastes from the developing offspring.
Nutrient uptake in B. neritina reportedly occurs via a specialized region of the embryonic epithelium (presumptive internal sac tissue) directly opposed to the embryophore during most of embryogenesis.
This specific area exhibits apical infoldings, pinocytotic channels and vesicles, pointing to its high absorptive capacity. In contrast, regions of the embryonic epithelia that do not abut the embryophore lack such infoldings (Woollacott & Zimmer, 1975). Published photos of histological sections of incubated zygotes and mid-stage embryos confirm the early establishment of their contact with the embryophore in the related Bugulina flabellata (Ostrovsky, 2013b;Ostrovsky et al., 2009), although it is puzzling why zygotes should stick to the maternal wall. Nonetheless, we assume that absorption could also occur through the rest of the larval surface, although additional study is needed to resolve this question.

| CONCLUSIONS
Matrotrophic nourishment encompasses numerous structural and physiological adaptations reflecting stages and trends in the evolution of parental care. Bryozoa, with their wide distribution of placental analogues that evolved multiple times within various clades, is a unique model showing various manifestations of matrotrophy from both parent and embryo. The placental analogue in C. hyalina strongly differs from all those previously described in matrotrophic cheilostomes. This suggests its independent origin within the family Hippothoidae, in which matrotrophy is known only in the genus Celleporella (Hughes, 1987;Marcus, 1938;Ostrovsky, 1998;Ryland, 1979). This is also confirmed by its position in the bryozoan molecular tree, where it clusters with non-placental taxa (Taylor & Waeschenbach, 2015;Waeschenbach, Taylor, & Littlewood, 2012).
In matrotrophic Cheilostomata, the general similarity in the ultrastructure of the follicle and embryophore cells is not surprising because they all serve to transport and transform nutrients for the growing offspring. At the same time, both the follicle and embryophore cells show various (and correlated) degrees of development (i.e., cell size and number). This illustrates the consecutive stages in the shift from macro-to oligolecithal oogenesesis, accompanied by the placental advancement and transition from the incipient to substantial matrotrophy. Reduced follicle cell size and number, and the production of oligolecithal oocytes with a corresponding strong enlargement of the embryophore and, consequently, embryo, is clearly the most advanced variant of matrotrophic reproduction. It is known in only few species, however. Incipient matrotrophy with weakly developed placenta and macrolecithal oogenesis has been recently described in a few species too. Finally, a number of species, including C. hyalina show a mixture of traits, having macrolecithal, but relatively small, oocytes developing in the ovary formed by the flattened follicle cells, along with a modestly or strongly developed placenta providing a substantial embryo size increase (Moosbrugger et al., 2012;Ostrovsky, 2013aOstrovsky, , 2013bthis study). This  Andrew N. Ostrovsky https://orcid.org/0000-0002-3646-9439