Cleavage modification did not alter early blastomere fates during bryozoan evolution

Stereotypic cleavage patterns play a crucial role in cell fate determination by precisely positioning early embryonic blastomeres. Although misplaced cell divisions can alter blastomere fates and cause embryonic defects, cleavage patterns have changed several times during animal evolution. Here, we analyze the evolutionary transition from spiral cleavage – a stereotypic pattern remarkably conserved in many protostomes – to the biradial cleavage of bryozoans. We characterize the cell lineage, MAPK signaling and expression of several developmental genes in the bryozoan Membranipora membranacea, and found that the fate and the genes expressed in the early bryozoan blastomeres are similar to their putative homologous blastomeres in spiral-cleaving embryos. The data indicate that cleavage geometry evolved independent from other developmental traits during the transition from spiral to biradial cleavage in the bryozoan lineage, revealing that stereotypic cleavage patterns can be evolutionarily modified without major changes to the molecular identity and fate of embryonic blastomeres.

On subsequent stages, the eight animal blastomeres of M. membranacea act as octets, synchronized in the cell divisions ( Figure 3). The first octet (animal pole cells 1q, 4 inner and four outer cells) divides equatorially making a brief 24-cell stage and the octets 1q 1 and 1q 2 (6.5 hpa). This division is shortly followed by an unequal cleavage originating the third quartet (3a-3d) from the four inner vegetal blastomeres at 6.8 hpa ( Cell divisions between the correspondent blastomeres of each quartet are mostly synchronous up to the 64-cell stage (Figure 4). At this point, we observe the fist two significant asynchronous cell divisions in the posterior D quadrant ( Figure S1). The cell 1d i 12 divides approximately 1h before than its partners 1a i 12 , 1b i 12 and 1c i 12 ( Figure S1), while the cell 1d e 11 divides 2h later than its quartet correspondents 1a e 11 , 1b e 11 and 1c e 11 (Figure 4 and Figure S1). Other cells from the D quadrant also exhibit asynchrony, such as the large inner vegetal blastomeres 3Q (3D is delayed) and the 7 each lineage contributing almost equally to the structures on their respective sides: D=posterior, C=right, B=anterior and A=left ( Figure 5A).
Progeny of the first quartet of animal blastomeres (1a-1d) originates apical ectodermal structures such as the apical organ, the aboral epithelia and the corona ( Figure 5A). The apical organ is derived from derivatives of the apical-most cells 1a 1 , 1c 1 and 1d 1 ( Figure 5B; Figure 6A). Cells 1a and 1c form the lateral and anterior most portion of the apical organ while the posterior cell 1d contributes not only to the posterior portion, but also to the tissues at the base of the apical organ ( Figure 6A). Thus, the cell 1b is the only blastomere of the first animal quartet that does not contribute to the apical organ. Epithelial cells between the apical organ and the corona are mostly derived from the octets 1q 11 and 1q 12 . Outer coronal cells originate from 1q 12 and 1q 2 while inner coronal cells (turned inwards after the invagination of the vegetal plate) are derived from 1q 2 ( Figure 6A). For a detailed overview of cell fates see Figure S2.
The vegetal blastomeres 1A-1D form the epithelium of the vestibule, the oral/anal ectoderm as well as the cells internalized during gastrulation, which originate the endoderm and mesoderm of the cyphonautes larva ( Figure 5A). The cellular arrangement at the vegetal plate in a 40-cell embryo consists of 12 outer cells (3a-3d, 2a R/L -2d R/L ) and four large inner blastomeres (3A-3D) ( Figure 6B). Here we define gastrulation as the internalization of these four vegetal cells. It occurs by delamination and epiboly in two rounds of division of the outer vegetal twelve-tets, which divide radially, pushing the 4 larger blastomeres internally and outlining a blastopore (Figure 6B,90 cell). At the 90-cell stage 12 cells are defining the blastopore lip, but this number gets reduced to 8 cells after the next division ( Figure 6B, 120 cell). From the 12 vegetal cells, one 9 M. membranacea.
We investigated the effects of different U0126 concentrations (1, 10, 25 µM) on the development of M. membranacea when applied at the 2-cell stage ( Figure S3A). We found the severity of the phenotype correlates with the concentration of the inhibitor, where the higher concentrations of 10 and 25 µM result in the complete disruption of the normal morphology ( Figure S3A). These embryos show no identifiable larval structures, such as a differentiated apical organ or musculature, have a lower number of cells, and are shorter compared to control samples ( Figure S3B).
The percentage of this severe phenotype decreases when we begin the treatment at later developmental stages (from 4-8 hpa), and embryos treated from 10 hpa onwards show progressively milder phenotypes ( Figure S4). In treatments beginning at 10-16 hpa, the larval structures such as apical organ, ciliated band and gut, are formed but the embryos are shorter and delayed in development in comparison to control embryos, while 18-24 hpa samples have almost normal morphology ( Figure S4). Finally, to identify the developmental defect caused by the MEK inhibitor in the severe phenotype, we recorded M. membranacea embryos treated with 10 µM U0126 under the 4D microscope. We found that embryos with milder phenotypes develop slower when compared to wild type, but do not show any obvious cleavage abnormalities ( Figure S5).
However, embryos exhibiting the severe phenotype have a misguided 4th cleavage (8-16 cell stage), where the animal blastomeres no longer divide in a biradial fashion ( Figure S5). Thus, the developmental defect causing the severe phenotype in U0126-treated embryos occurs before the 28-cell stage, when we first detect MAPK activity in M. membranacea ( Figure 8).

Gene expression
In order to complement the cell lineage studies and elucidate the blastomere identities of the bryozoan embryo, we analyzed the expression of sixteen conserved molecular markers during the development of M. membranacea (Figure 9-11; Figure S6): anterior/neural (six3/6, dlx, otx, pax6 and nk2.1), foregut (foxa and gsc), germline (nanos), hindgut/posterior (bra, cdx, evx and wnt1), endodermal (gata456) and mesodermal (twist, foxc and foxf) markers. 10 mal cell (4B) and in 5 cells of the apical disc ( Figure 9; Figure 11A). Expression of six3/6 clears from 2c R2 and 4B, but persists in the inner cells of the forming apical organ, a central neural region occupied by serotonergic-positive cells in other cyphonautes larvae (Nielsen and Worsaae, 2010). We detected dlx transcripts, a gene involved in neurogenesis and proximodistal patterning (Panganiban and Rubenstein, 2002), in the 8 animal pole cells of the 16 cell stage (1q), broadly in the apical disc during gastrulation and elongation and, finally, localized to the whole apical organ in the late gastrula ( Figure 9).
The gene otx is involved in anterior ectodermal patterning (Arendt et al., 2001;Boncinelli et al., 1993;Lowe et al., 2003;Marlow et al., 2014;Steinmetz et al., 2010;Steinmetz et al., 2011;Umesono et al., 1999) and endomesoderm specification (Harada et al., 2000;Hinman et al., 2003;Mitsunaga-Nakatsubo et al., 2003). In M. membranacea otx is expressed in all blastomeres between 2-8-cell stage and gets restricted to the apical octet of the 16-cell stage (Figure 9). At the 32-cell stage, otx transcripts localize to the 1q 2 octet and during gastrulation there are 3 rows of cells expressing otx with a posterior gap ( Figure 11B). During mid-gastrula two cells in the apical organ express transcripts of otx ( Figure 11B). In the late larva, otx is expressed in the corona and vestibule epithelium ( Figure 11B). Expression of pax6 is first detected during gastrulation, in bilateral patches of the apical ectoderm, and remains as a thin line of expression encircling the embryo above the corona (Figure 9). The gene nk2.1 is involved in the patterning of the neural plate in vertebrates (Shimamura et al., 1995) and is expressed in anterior and ventral territories including the apical/neural plate and anterior endoderm (Lowe et al., 2003;Marlow et al., 2014;Takacs et al., 2004;Venkatesh et al., 1999). Transcripts of nk2.1 are present in the progeny of the vegetal cells 2b and 3b in the early gastrula stage ( Figure 9). These cells occupy an anterior vegetal position abutting the anterior blastopore lip until the edge of the vegetal plate. After the invagination of the vegetal plate, nk2.1-positive cells are lining the anterior portion of the preoral funnel, next to the mouth.
Expression of foxa is related to endoderm specification and commonly associated with the blastopore lip and foregut (Arenas-Mena, 2006;Boyle and Seaver, 2010;Oliveri et al., 2006). At the 16-cell stage, we detected faint expression of foxa in the outer vegetal blastomeres and in 10 11 terior cells 2d L and 2d R ) (Figure 9; Figure 11C). Expression of foxa persisted in the daughter cells of the next division forming 2 rows of cells around the blastopore with a gap at the posterior end ( Figure 11C). With the invagination of the vegetal plate, this region occupies an anterior/lateral position in the vestibule wall, surrounding the mouth region of the late gastrula ( Figure 9; Figure   11D). We only found transcripts of gsc at the early gastrula stage in two anterior and a bilateral pair of cells at the vegetal plate ( Figure 9; Figure 11E). In the late gastrula, gsc is expressed in bilateral domains of the vestibule wall which fuse anteriorly.
The germline marker nanos (Extavour and Akam, 2003;Juliano et al., 2010) is expressed in two posterior cells of the vegetal plate at the 32-cell stage (2d L and 3d) ( Figure 9). In subsequent stages, nanos continues restricted to two cells at the posterior portion of the vegetal plate, localizing to the internal sac region of the cyphonautes larva ( Figure 11F).
All posterior/hindgut and mesodermal markers only initiate expression during gastrulation. The gene bra can have multiple roles, but it is generally related to mesoderm and posterior/hindgut patterning (Technau, 2001). Expression of M. membranacea bra in the early gastrula occurs at the vegetal plate in a posterior band of cells near the blastopore lip ( Figure 10; Figure 11G). It localizes to 6-8 cells at the posterior end of the mid gastrula and a broad portion of the posterior and lateral vestibule ectoderm ( Figure 11G). M. membranacea bra expression domain reaches the posterior portion of the preoral funnel as well as the future hindgut area of the larva ( Figure   10). A single posterior vegetal plate cell (2d R2 ) and its daughter cells (2d R21 and 2d R22 ) express the posterior/hindgut markers cdx and evx at the early gastrula ( Figure 10; Figure 11H; Figure 11J).
During gastrulation, cdx and evx continue to be expressed at the posterior edge of the vegetal plate ( Figure 10; Figure 11J; Figure 11L) and localize to the posterior vestibule ectoderm (hindgut) of the late gastrula ( Figure 10; Figure 11L). At this stage, evx is also found in the posterior region of the gut ( Figure 11L; Figure 11M). We also detected a transient evx expression in the two internalized blastomeres 4a and 4c of the early gastrula. Finally, wnt1 is expressed in a row of 3-5 cells (including 2d L2 , 2d R2 and 3d 2 ) posterior to the blastopore during gastrulation ( Figure 10; Figure 11N). At the late gastrula, wnt1 is detected at the posterior-most vestibule ectoderm, positioned between the corona and hindgut ( Figure 10; Figure 11N). 12 only occurs in a portion of embryos during the development of M. membranacea. We detected colorimetric signal in bilateral internalized cells of the early gastrula -possibly 4a, 4c or derivatives -as well as at the anterior end of the late gastrula ( Figure 10; Figure 11O). Transcripts of foxc, commonly expressed in anterior and posterior mesodermal domains (Häcker et al., 1995;Passamaneck et al., 2015;Shimeld et al., 2010), are present in one unidentified posterior vegetal plate cell of the early gastrula and two similarly positioned cells during mid gastrulation ( Figure   10). In the late gastrula, foxc expression is located in the internal sac area. The gene foxf is a transcription factor involved in mesoderm patterning and expressed mainly in visceral and anterior territories (Mazet et al., 2006;Passamaneck et al., 2015;Pérez Sánchez et al., 2002;Shimeld et al., 2010;Zaffran et al., 2001). In M. membranacea it is expressed in the mesodermal cell 4b in the early and mid gastrula stages ( Figure 10). This cell and its descendants divide subsequently from basal to apical, forming a distinct frontal row of mesodermal cells expressing foxf at the anterior portion of the late gastrula.
We found two copies of the endomesodermal marker gata456 (Patient and McGhee, 2002) in the transcriptome of M. membranacea. While the gene gata456a is not expressed at detectable levels in any of the analyzed stages, gata456b is strongly expressed in endodermal tissues of the bryozoan. The expression of gata456b initiates early, in the vegetal 3D blastomere at the 32-cell stage ( Figure 10; Figure 11P). The expression expands to adjacent lateral blastomeres 4A and 4C in the early gastrula, and in subsequent stages gata456b continues to be expressed in the endodermal tissues forming the gut of the cyphonautes larva ( Figure 10; Figure 11P).

Discussion
The phylogenetic position of bryozoans provides a valuable opportunity to investigate the evolution of developmental traits. Even though the kinship of Bryozoa remains inconclusive -the group is either related to Entoprocta and Cycliophora (e.g. Hejnol et al., 2009)

Specification of the D quadrant
One critical event of animal embryogenesis is the establishment of the dorsoventral polarity. In spiral-cleaving embryos this event is tied to the specification of the D quadrant during development (Freeman and Lundelius, 1992). In species where the first two embryonic cell divisions are unequal, the D quadrant is determined early by the asymmetric distribution of maternal cytoplasmic determinants, while in species that form equal-sized blastomeres at the 4-cell stage, the D quadrant is specified at the 32-cell stage by inductive interactions mediated by cell contacts between micromeres and macromeres (Biggelaar, 1977;Freeman and Lundelius, 1992;Gonzales et al., 2006;Martindale, 1986;Martindale et al., 1985). In the current work, we found evidence that the specification of the D quadrant in the equal, biradial-cleaving bryozoan M. membranacea resembles that of equal, spiral-cleaving molluscs in the timing of specification, pattern of MAPK activation and asynchrony of the D quadrant cell divisions post-specification.
In equal-cleaving molluscs, the specification of the D quadrant correlates with the activation of the MAPK pathway in the 3D macromere only (Koop et al., 2007;Lambert and Nagy, 2003). In M. membranacea, whose equal-sized blastomeres at the 4-cell stage give rise to perfectly symmetrical embryonic quadrants, that are indistinguishable from each other until gastrulation, the earliest molecular asymmetry we could detect is the activation of the MAPK pathway in a single vegetal blastomere that originates the posterior portion of the larval body. As in equal-cleaving molluscs, MAPK is activated in the bryozoan 3D blastomere on the 5th round of cell divisions, suggesting the D quadrant of M. membranacea is specified as early as the 28-cell stage. This might indicate that bryozoans and equal-cleaving molluscs undergo similar developmental mechanisms of D quadrant specification (but see below). Interestingly, most equal-cleaving spiralians studied so far exhibit a single MAPK-activated blastomere during early development, while unequal-cleaving species show diverse deviating patterns (see Table 1), thus suggesting that this pattern of MAPK activity is a common feature of equal-cleaving embryogenesis independent of its cleavage pattern.

14
The developmental role of the MAPK pathway in spiralians, however, is far less clear. Drug treatments to block MAPK activity in molluscs cause radialized larvae that lack muscles, shell and foot, suggesting that the MAPK pathway is involved in the signaling underlying the D quadrant specification (Henry and Perry, 2007;Koop et al., 2007;Lambert and Nagy, 2001;Lambert and Nagy, 2003). While we found severely disturbed M. membranacea embryos in our MEK inhibitor experiments, the abnormal cleavage leading to the severe phenotype occurs before any detectable MAPK activity, as revealed by the 4D recordings. Thus, the developmental defect we observe in the severe phenotype is unlikely to be a direct consequence of the inhibition of MAPK signaling in the 3D blastomere, unless MAPK signaling is active in undetectable levels at the 16cell stage. Instead, most of our experiments indicate that the inhibition of the MAPK pathway does not cause dorsoventral defects in M. membranacea. In fact, when treated with the same MEK inhibitor concentration we used in the bryozoan, equal-cleaving molluscs still show a latent dorsoventral polarity or develop without axial defects (Koop et al., 2007;Lartillot et al., 2002a), which might indicate that MAPK is not the sole signaling agent for the D quadrant specification in spiralians.
Once the D quadrant has been determined, it typically shows asynchronous cell divisions in relation to the other quadrants of spiral-cleaving embryos (Guralnick and Lindberg, 2001). For instance, the 3D macromere in the mollusc Patella vulgata (Biggelaar, 1977) and the 1d derivatives of Ilyanassa obsoleta (Clement, 1952;Goulding, 2009) undergo a late division. Our analyses of M. membranacea cell lineage indicate similar asynchronous cell divisions in the D quadrant, which include the 3D blastomere and 1d derivatives of the bryozoan. Therefore, the specification of the D quadrant seems to be correlated with subsequent changes in the cell cycle timing in both M. membranacea and spiral-cleaving embryos. 15 ertheless, the MAPK pathway is still poorly sampled in spiralians, and other spiral and non-spiral-cleaving groups, such as phoronids, nemerteans, polyclads, rotifers and gastrotrichs, need to be investigated to properly understand the roles and the evolution of MAPK signaling in spiralian development.

Comparative spiralian fate maps
The stereotypic cleavage pattern of spiral cleavage permits to identify putative homologous blastomeres between different spiralians, and to compare their fates in the larval/adult tissues (Conklin, 1897;Mead, 1897;Treadwell, 1901;Wilson, 1892;Wilson, 1898). These studies revealed that homologous early blastomeres share mostly-similar fates in various clades (Hejnol, 2010;Henry and Martindale, 1999;Lambert, 2010;Nielsen, 2004;Nielsen, 2005). The cleavage of M. membranacea clearly differs from the spiral cleavage pattern, which complicates the identification of homologous blastomeres between the bryozoan and a spiral-cleaving embryo. However, we established a common developmental feature to base our comparative cell lineage and gene expression analyses. In both spiral and bryozoan embryogenesis, the vegetal blastomeres sequentially give rise to quartets of daughter cells, while remaining at the vegetal-most portion of the embryo until being internalized during gastrulation. We thus compare the quartets of M. membranacea to the quartets of spiral-cleaving embryos in terms of gene expression and fate in the larval tissues. We find the quartets have a similar molecular identity and contribute to the same set of structures in the larvae of bryozoan and spiral-cleaving groups, and that the subset of blastomeres that gives rise to these structures partially overlap ( Figure 12). This indicates that bryozoans might share a common embryonic patterning of early blastomere fates with other spiralians, and that, in the current phylogenetic scenario, such developmental trait has remained conserved despite the drastic modification in the cleavage pattern from spiral to biradial.

First quartet: apical organ and ciliated band
The first quartet of micromeres in spiral-cleaving embryos contributes to the apical organ, the ciliated band and all the ectoderm in between (Ackermann et al., 2005;Boyer et al., 1998;Child, 1900;Conklin, 1897;Damen and Dictus, 1994;Dictus and Damen, 1997;Hejnol et al., 2007;Henry and Martindale, 1998;Henry et al., 2004;Maslakova et al., 2004;Meyer et al., 2010;Wil- son, 1892). In the M. membranacea, the first quartet of animal blastomeres also gives rise to these ectodermal structures of the cyphonautes larva. This suggests that in both the bryozoan and spiral-cleaving embryos the third cleavage demarcates a split in the embryonic fate map, in which the first quartet of animal blastomeres only gives rise to the ectodermal structures placed towards the animal pole, while the progeny of the vegetal blastomeres (i.e. the 2nd, 3rd and 4th quartets) originates a different set of larval structures (see the next sections). When we compare in more detail the specific fates of the descendants of the first quartet, we find that some blastomeres contributing to the apical organ or ciliated band of the cyphonautes larva indeed contribute to the respective structures of spiral-cleaving larvae -but that this similarity is not complete, and different subsets of blastomeres contribute to the apical organ and ciliated band.
The apical organ, for example, is usually formed by the progeny of the apical-most 1q 1 micromeres in groups with spiral cleavage (Nielsen, 2004;Nielsen, 2005). While the apical organ of M. membranacea larva is also derived from the apical-most subset of 1q i 1 (=1q 11 ), descendants of 1q e 1 (=1q 21 ) contribute to the structure as well. We find a similar situation when comparing the embryonic origin of the corona (i.e. the ciliated band of the cyphonautes larva) with the prototroch -a ciliated band considered to be an ancestral trait for the larval stages of trochozoan spiralians (Rouse, 1999). The prototroch of annelids and molluscs is formed by 1q (accessory and primary trochoblasts) and 2a-c (secondary trochoblasts) descendants (Damen and Dictus, 1994;Hejnol et al., 2007;Henry et al., 2007). Our data reveals that the corona of M. membranacea is formed by blastomeres equivalent to the accessory/primary trochoblasts of the prototroch (Damen and Dictus, 1994). But, unlike spiral-cleaving embryos, the second quartet does not contribute to the ciliated band of the bryozoan larva (see Figure S7). In general, we find that equivalent early blastomeres of the bryozoan and spiral-cleaving embryos contribute to similar larval structures, but that the fate of the progeny of these early blastomeres only partially overlap between the bryozoan and spiral-cleaving embryos. These observations suggest that during the evolution of the bryozoans, shifts in the blastomere fates occurred in late embryogenesis while the early embryonic patterning, presumably inherited from a spiral cleavage ancestor, might have remained conserved.
The bryozoan fate map indicates that the apical organ, outer ectoderm and corona of the cypho-  Table 2). Therefore, the region derived from the first quartet of M. membranacea match the pretroch region of spiralcleaving embryos.
In M. membranacea the blastomeres that form the apical organ express six3/6 and dlx from the 16-cell stage, suggesting that these genes might be involved in the establishment of the embryonic animal/vegetal identities, and possibly in the molecular patterning of the cyphonautes apical organ. The expression of otx in the bryozoan is associated to the corona, the ciliated band of the cyphonautes larva, similar to other spiralians where otx is expressed near or in the larval ciliated band (Arendt et al., 2001;Nederbragt et al., 2002a;Steinmetz et al., 2011). The gene is an interesting example because it provides the opportunity to integrate the cell lineage and gene expression data between the bryozoan and spiral-cleaving embryos. As explained above, the ciliated band of trochophore larvae, the prototroch, is formed by the contribution of first quartet and second quartet blastomeres (Damen and Dictus, 1994;Henry et al., 2007), while the corona of M. membranacea derives solely from first quartet blastomeres, which are the putatively homologous to the primary trochoblasts of the prototroch. In the mollusc Patella vulgata, otx is expressed in all prototroch cells (Nederbragt et al., 2002a). Interestingly, the second quartet blastomeres of M.
membranacea -the set of blastomeres that form the secondary trochoblasts in the prototrochalso express otx, as observed in P. vulgata, even though these cells do not contribute to the Overall, our work reveals that the first quartet of M. membranacea embryo and the first quartet of spiral-cleaving embryos originate a similar set of larval structures, and give rise to a larval body region with similar molecular profile. Thus, the outer ectodermal region of the cyphonautes larva corresponds, in developmental terms, to the head region of other spiralians.

Second and third quartet: mouth
The second and third blastomere quartets of spiral-cleaving embryos contribute to a diverse set of ectodermal structures, such as the foregut, ciliated bands, neurons, the mollusc shell gland and foot, the annelid trunk and nerve cord, as well as ecto-mesodermal muscle cells (Ackermann et al., 2005;Boyer et al., 1998;Dictus and Damen, 1997;Hejnol et al., 2007;Henry and Martindale, 1998;Henry et al., 2004;Lyons et al., 2015;Maslakova et al., 2004;Meyer et al., 2010;Render, 1997). In M. membranacea these blastomeres form the whole vegetal ectoderm that gives rise to the vestibule epithelium, including the preoral funnel and posterior ectoderm of the cyphonautes larva. In most spiralians the second and third quartets are the blastomeres surrounding the blastopore -the orifice formed at the site of endomesoderm internalization (Lankester, 1877), whose developmental fate has been a significant trait for the discussions about metazoan evolution Martindale and Hejnol, 2009;Martín-Durán et al., 2016b). Nevertheless, the fate of the blastopore in bryozoans still remains open to discussion (Gruhl, 2009;Marcus, 1938;Prouho, 1892;Zimmer, 1997).
Even though in most gymnolaemate bryozoans the blastopore closes after gastrulation (Pace, 1906;Prouho, 1892), or in some cases, an orifice is not formed at all (Calvet, 1900), an ultrastructural study in M. membranacea revealed that its blastopore remains open until the larval stage (Gruhl, 2009). Our cell lineage data indicate that cells at the blastopore lip give rise to the preoral funnel of M. membranacea, and that the endodermal cells lining the blastopore form the anterior portion of the larval gut. We also found the foregut marker foxa is expressed around most of the blastopore lip, except for a couple of posterior rows, and that foxa expression persists We found that the vegetal ectoderm -the cells derived from the second and third quartetexhibits an anteroposterior polarity, as revealed by the differential expression of molecular markers. The anterior/foregut markers nk2.1, foxa and gsc are expressed in a region opposed to posterior/hindgut markers bra, cdx, evx and wnt1, which are generally restricted to the D quadrant.  Table 2 for a broader comparison of gene expression within spiralians). This suggests that at least some molecular aspects of the axial patterning remain conserved in the cyphonautes larva.
In some cases, the transcripts of M. membranacea are not only located at a similar position along the anteroposterior axis, but also in the putative homologous blastomeres to spiral-cleaving embryos. An example is the expression of foxa between the bryozoan and the annelid Hydroides elegans (Arenas-Mena, 2006). In both, foxa is expressed in the second quartet blastomeres early in development, and in the cells that surround the blastopore during gastrulation, with a peculiar posterior gap (Arenas-Mena, 2006). Another comparable cellular expression is the gene bra, expressed in the second and third quartet progeny at the posterior lip of the blastopore of the molluscs Patella vulgata (Lartillot et al., 2002a) and Haliotis asinina (Koop et al., 2007). Therefore, the M. membranacea data indicates that the molecular identity of the blastomeres remained conserved to a certain extent, despite the altered cleavage geometry and position of the second and third quartet in the bryozoan, placed more vegetally than spiral-cleaving embryos. Classical works suggest that mesodermal cells derive from endodermal blastomeres, but could not demonstrate the embryonic origin with cellular resolution (Barrois, 1877;Calvet, 1900;Corrêa, 1948;d'Hondt, 1983;Pace, 1906;Prouho, 1892). However, recent ultrastructural data in M. membranacea suggests an ectodermal origin for the bryozoan mesoderm, from the delamination of an ectodermal cell during gastrulation (Gruhl, 2009). Our cell lineage data indicate that the first mesodermal cells of M. membranacea derive from the fourth quartet (4a-4c). The lateral cells 4a A and 4c A form the anterior muscles of the cyphonautes larva while the progeny of 4b 1 gives rise to a stack of mesenchymal cells that express the anterior mesoderm marker foxf. We did not observe the delamination of an anterior ectodermal cell as described by Gruhl (2009) The source of mesodermal tissues in spiral-cleaving embryos is extensively studied and discussed (Ackermann et al., 2005;Boyer et al., 1996;Boyer et al., 1998;Conklin, 1897;Gline et al., 2011;Hejnol, 2010;Hejnol et al., 2007;Henry and Martindale, 1998;Kozin et al., 2016;Lambert, 2008;Lartillot et al., 2002b;Lyons and Henry, 2014;Lyons et al., 2012;Lyons et al., 2015;Meyer et al., 2010;Render, 1997). There are generally two sources, an anterior mesoderm derived from the third quartet blastomeres (known as ectomesoderm) and a posterior mesoderm derived from the 4d blastomere. Even though most spiralians have the 4d as the sole endomesodermal contributor, there are exceptions -in the annelid Capitella teleta 3c and 3d generate mesoderm (Eisig, 1898;Meyer et al., 2010). The blastomeres contributing to the anterior mesoderm (usually 3a and 3b) are more variable (Hejnol et al., 2007;Lyons and Henry, 2014). This indicates that even within spiral-cleaving embryos, particular cell fates can shift to different blastomeres (Hejnol et al., 2007). We find that the source of mesodermal tissues of the bryozoan M. membranacea differs from other spiralians because (1) the third quartet does not contribute to the anterior mesoderm and (2) multiple blastomeres of the fourth quartet give rise to mesodermal tissues. In addition, the blastomeres 4a-4c often give rise to endodermal tissues in spiral-cleaving embryos (Hejnol et al., 2007;Henry and Martindale, 1998). Therefore, our data suggests that the Although only a subset of the genes we analyzed is expressed in the mesoderm of M. membranacea (six3/6, evx, twist, foxf), the patterns indicate that the bryozoan mesoderm is already regionalized at early gastrulation. For instance, we found that 4d expresses evx and cdx, genes commonly associated to posterior mesodermal and hindgut fates in spiralians (Fritsch et al., 2016;Fröbius and Seaver, 2006;Hiebert and Maslakova, 2015b;Hui et al., 2009;Kulakova et al., 2008;Le Gouar et al., 2003;Martín-Durán et al., 2016b;Rosa et al., 2005;Samadi and Steiner, 2010). Even though, we could not resolve the fate of the 4d blastomere in M. membranacea, the expression data indicates that the 4d blastomere might contribute to the posterior mesoderm and hindgut of the bryozoan. The lateral mesodermal cells 4a/4c and derivatives express evx, and possibly twist, in the same manner as the expression of evx orthologs in the annelid Capitella teleta during early development (Seaver et al., 2012). Finally, at the anterior mesoderm we find the expression of foxf, also observed in the brachiopod Terebratalia transversa (Passamaneck et al., 2015). Overall, these molecular data reveal that M. membranacea mesoderm is regionalized and that at least some of the expression patterns are conserved with other spiralians.
The 4d cell and its descendants also form the germline and are known to express nanos in spiralcleaving embryos (Rebscher, 2014). Germ cells have not been identified during the embryogenesis of any bryozoan and were only found in zooids after metamorphosis (Reed, 1991). The expression of nanos in M. membranacea differs from the pattern of spiral-cleaving embryos, since nanos is expressed in two posterior cells of the second and third quartet. These blastomeres divide repeatedly, but nanos expression is always retained in two cells that become part of the larval internal sac -the structure that persists during metamorphosis giving rise to the outer case of the zooid (Stricker, 1988). Thus, we hypothesize the nanos-positive cells might be stem cells contributing to the differentiation of the internal sac, but further analysis in competent larvae and metamorphosed juveniles are needed to clarify the fate and molecular identity these cells. In this context, during the evolution of gymnolaemate bryozoans the ancestral spiral cleavage pattern, characterized by the alternating oblique cell divisions, was modified to biradial cell divisions. While the cleavage pattern changed and the anterior mesoderm was reallocated to the fourth quartet, some aspects of the development have remained conserved, such as the D quadrant specification, MAPK activity and overall fate map of early blastomeres. Spiral cleavage has been modified not only in bryozoans, but in different spiralian branches as well, such as flatworms, molluscs and brachiopods (Hejnol, 2010). In most of these cases, the embryonic development has changed to such extent that no traces of spiral cleavage are found, and the ancestral cleavage geometry can only be inferred by the phylogenetic position of the clade (e.g. the discoidal cleavage of cephalopods (Wadeson and Crawford, 2003)). Remnants of spiral cleavage usually consist of oblique mitotic spindles, as in the flatworm Macrostomum lig-24 beginning to be uncovered using molecular and computational approaches (Brun-Usan et al., 2016;Davison et al., 2016;Kuroda, 2015). Our data suggests that the quartet-divisions, cell fates and other traits commonly associated with a spiral cleavage program were maintained in the bryozoan development despite the evolutionary modification to a biradial cleavage pattern.

Evolution of cleavage patterns
The fate of the early embryonic blastomeres is thought to be causally associated to the cleavage pattern during development (Valentine, 1997;Wray, 1994). In such case, a change in the cleavage geometry would lead to a change in the cell fates. However, the bryozoan cell lineage illustrates a case where the cleavage pattern and blastomere fates are not evolutionary coupled. We find that the fates of the early blastomeres are similar between M. membranacea and spiralcleaving embryos despite the modified biradial cleavage pattern. The relative positioning of the second and third quartets even differs between the bryozoan and a typical spiral-cleaving embryo ( Figure 12), but these blastomeres still contribute to similar tissues, suggesting the early cell fate determination remained relatively conserved during bryozoan evolution. The bryozoan cell lineage illustrates how a widely conserved determinate cleavage pattern -spiral cleavage -can evolve without major changes in other developmental traits, such as the blastomere fates and molecular identity.
We found several developmental genes expressed in a similar spatial arrangement between bryozoans and other spiralians, as revealed by MAPK (3D blastomere), otx (basal blastomeres of the first quartet), and bra and foxa (second and third quartets). This molecular map is similar not only to the typical spiral-cleaving embryos, but also to brachiopod embryos (Martín-Durán et al., 2016b;Passamaneck et al., 2015), whose embryos have a much greater number of cells and no stereotypic cleavage pattern (Long and Stricker, 1991). Thus, a single cell in the bryozoan embryo expressing gata456 might be homologous to a whole region of gata456 expression in the brachiopod embryo (Passamaneck et al., 2015), as suggested by Hejnol (2010). This reinforces the hypothesis that cell fate determination is not tied to a particular cleavage pattern, but depends on the underlying molecular framework established early in development (Henry et al., 1992). This is the case for nematodes, whose cleavage patterns diverged drastically between groups without a corresponding change in the resulting phenotype (Schulze and Schierenberg, 2011). However, a clearer parallel case to the spiral-to-biradial evolution is the transition from the spiral cleavage of polychaete annelids to the derived cleavage of clitellate annelids; despite the differences in the cleavage pattern, the clitellate fate map does not deviate significantly from the annelid ground plan (Kuo, 2017). Overall, our findings support the hypothesis that in evolutionary terms, the causal ontogenetic connection between cleavage pattern and blastomere fates, if any, can be broken (Scholtz, 2005). In the case of the bryozoan M. membranacea, the molecular identity and fate of the early blastomeres might have been maintained, despite the modification in the geometry of cell divisions. Further comparative cell lineage studies with other non-spiral spiralian lineages, such as gastrotrichs and rotifers, will be crucial to establish the ancestral traits of spiralian development and to better comprehend the relation between cleavage patterns and cell fates during evolution.

Conclusions
The embryonic development of M. membranacea provides a unique comparative standpoint to the typical spiral cleavage pattern. It reveals that spiral cleavage is not an all-or-nothing character and has been extensively modified in the diverse spiralian lineages. In particular, we suggest that the cleavage geometry of the bryozoan embryo evolved independent from other spiralian developmental traits. Therefore, modifying spiral cleavage does not require drastic developmental changes such as the ones found in cephalopods or parasitic flatworms. More generally, our data suggests that determinate cleavage patterns can be modified without major changes in the identity of blastomeres and cell fates, which alleviates the idea that the cleavage pattern is evolutionary coupled to the specification of cell fates. In this perspective, the evolutionary conservation of cell fates in spiral-cleaving clades might be a consequence of a conserved underlying molecular patterning, overlaid by a determinate cleavage pattern. Overall, our work highlights the importance of comparative data to better understand the evolution of spiralian development.

Collection, spawning and cultures
We collected M. membranacea in the fjord waters of Hjellestadosen (60°15'23.9" N 5°14'20.1" E) in Bergen, Norway, between May and September. We handpicked kelp blades with ripe bry- water at 10 °C. To induce spawning, we cut a portion of the kelp blade with mature colonies, usually the ones with more opaque whitish/pinkish zooids, and transferred it to glass bowl with sea water sterilized with UV-light and filtered through a 0.2 µm mesh (UVFSW). The bowl was placed under a stereomiscroscope with direct light and a digital thermometer to monitor the water temperature. Ripe colonies begin to spawn in around five minutes or more, or usually when the temperature reached 15 °C. Once the temperature rose to 16 °C, the bowl was cooled down on ice with no direct light. The spawning colony was sequentially transferred to new bowls with UVFSW at 10 °C to distribute the vast amounts of eggs. For each bowl with eggs, we added EDTA to a final volume of 0.1mM (usually ~20 µL of 0.5M EDTA) to induce egg activation (Reed, 1987). The bowl was then placed in a incubator at 15 °C for 30-60 min. Activated eggs were concentrated by swirling, distributed to smaller glass bowls with UVFSW and washed twice to remove the EDTA. We adjusted the amount of eggs per bowl to avoid that the eggs sitting on the bottom of the dish touch each other. We kept the cultures at 15 °C. One colony could be re-used for spawning multiple times. M. membranacea colonies maintained in the flowing tanks remained viable to developmental studies for a week.

4D recordings and cell tracing
We pipetted embryos at the 2-cell stage to a glass slide coated with poly-l-lysine. We mounted the embryos under a cover slip with supporting clay feet, completed the volume with UVFSW and sealed the cover slip with vaseline. The slide was put under an automated 4D microscopy system (Caenotec) (Hejnol and Schnabel, 2006) with a cooling ring around the objective to keep the temperature at 15 °C. We recorded full-embryo stacks (40-60 optical slices) every 40s under differential interference contrast. Development was recorded for approximately 24h, when the embryo became ciliated and swam away from the field of view.
Raw data consists of a sequence focal levels for each time point. We loaded the data into the tracking software SIMI°BIOCELL (SIMI®) and manually traced individual cells. The results of this manuscript are compiled from the cell tracking data of four different embryos ( Figure S8).
Source files with the cell lineages and blastomere synchrony analyses are available at Vellutini et al. (2017).  (Child, 1900;Conklin, 1897;Wilson, 1892). The identical blastomeres at the 4-cell stage were labeled as A, B, C and D according to their correspondent fate, as identified by the video recordings. In such case, the D quadrant was assumed to be the blastomere giving origin to the dorsal/posterior region of the cyphonautes larva. At the 8cell stage, the animal blastomeres were named 1a-1d and the vegetal blastomeres 1A-1D. The first cleavage of these animals blastomeres give rise to the first octet, referred with the subscript o (e.g. 1q o refers to all cells of the first octet). An octet is made from four inner cells, indicated by the subscript i (for internal), and four outer cells indicated by the subscript e (for external). Derivatives of the animal octets received the standard superscripts of spiral cleavage with 1 for the apical progeny and 2 for the basal progeny (e.g. 1a i 1 refers to the inner A quadrant cell of the apical progeny from the first octet). The quartets derived from the four large blastomeres were labeled as in spiral cleavage (2q-4q). And the progeny of the first twelve-tet followed the spiral cleavage 1 and 2 superscripts. Finally, we labeled meridional with the superscript R for the cell to the right and L for the cell to the left, when viewed from the egg axis.

Fixation methods
We fixed representative developmental stages for antibody staining in 4% formaldehyde for 1h at room temperature, washed the embryos in PTw (1x PBS + 0.1% Tween-20) and stored (in PTw) at 4 °C. For in situ hybridization, we fixed the samples in a solution of 4% formaldehyde / 0.2% glutaraldehyde solution to avoid tissue damage during the protocol. After 1h fixation at room temperature, we washed the embryos in PTw, dehydrated through methanol series and kept the samples in 100% methanol at -20 °C.

Gene cloning and in situ hybridization
We assembled the Illumina reads from M. membranacea (SRX1121923) with Trinity (Grabherr et al., 2011) and used known genes to identify putative orthologs in the transcriptome. We performed PCR gene specific primer pairs on cDNA synthesized with the SMARTer RACE cDNA Amplification kit (Clontech). Primers were designed with Primer3 (Untergasser et al., 2012). We

Immunohistochemistry and MAPK antibody
We permeabilized the embryos with several washes in PTx (1x PBS + 0.2% Triton X-100) for 2h and blocked with two washes of 1h in PTx + 0.1% BSA (Bovine Serum Albumin) succeeded by 1h incubation in PTx + 5% NGS (Normal Goat Serum). Samples were incubated with the primary antibody for the MAPK diphosphorylated ERK-1&2 (Sigma M9692-200UL) diluted 1:200, and stored overnight at 4 °C on a nutator. We removed the MAPK antibody with three 5 min and four 30 min washes in PTx + 0.1% BSA, blocked in PTx + 5% NGS for 1h and incubated nutating overnight at 4 °C with the secondary antibody Anti-Mouse-POD conjugate (Jackson) diluted 1:250. We removed the secondary antibody with three 5 min followed by several washes in PTx + 0.1% BSA for 2h. To amplify and develop the signal, we incubated the embryos for 3-5 min with the provided reagent solution and fluorochrome from TSA reagent kit Cy5 (Perkin Elmer).
We stopped the reaction with two washes in a detergent solution (50% formamide, 2x SSC, 1% SDS) at 60 °C, to reduce background, followed by PTw washes at room temperature. We stained nuclei by incubating permeabilized embryos in DAPI 1:500 or Sytox Green 1:1000 for 2h.
Nuclei staining was combined with f-actin staining by the addition of BODIPY FL Phallacidin 5 U/mL previously evaporated to remove methanol.

Microscopy and image processing
We mounted in situ embryos in 70% glycerol in PTw. Embryos from antibody staining were mounted in 97% 2,2'-Thiodiethanol (Asadulina et al., 2012;Staudt et al., 2007), 80% Glycerol in PBS or SlowFade® Gold Antifade (ThermoFisher). We imaged the samples with a Zeiss Axio-Cam HRc mounted on a Zeiss Axioscope A1, using differential interference contrast technique (Nomarski) for in situ hybridizations and a fluorescent lamp for the MAPK antibody staining.
We used a Confocal Leica TCS SP5 to image fluorescent samples. Colorimetric in situ hybridizations were also scanned under the confocal using reflection microscopy (Jékely and Arendt, 2007). We processed all resulting confocal stacks in Fiji (Schindelin et al., 2012). When necessary, we adjusted the distribution of intensity levels to improve contrast with Fiji or GIMP. We created vector graphics and assembled the figure plates using Inkscape.
10µM at 23°C -Disorganized and incomplete larvae. MAPK activation is required on the micromeres themselves, not only in 3D. Alters the cleavage pattern in the third quartet of micromeres and timing of division of the 4d (Lambert and Nagy, 2001).
*Synonyms taken from WoRMS database. 1