A cadherin switch marks germ layer formation in the diploblastic sea anemone Nematostella vectensis

Morphogenesis is a process of tissue and organ formation during organism development (Gilbert, 2013) that is driven by coordinated cell shape changes, cell migration, cell proliferation, cell death and cell adhesion. The key morphogenetic events during early development are gastrulation, germ layer formation, folding of the neural tube and body axis elongation. Cadherins are transmembrane cell adhesion molecules that play an important role in these processes. They not only provide the mechanical connection between cells, but also control cell-cell recognition, cell sorting, tissue boundary formation, signal transduction, formation of cell and tissue polarity, cell migration, cell proliferation and cell death (Gumbiner, 2005; Halbleib and Nelson, 2006). In adult tissues, cadherins preserve stable and ordered tissue integrity (Angst et al., 2001; Halbleib and Nelson, 2006).

Classical cadherins are conserved molecules present in all animals whose genomes have been analyzed (Alberts, 2007). They are major components of the adherens junctions between cells, which are conserved structures of epithelial cells in most animals (Meng and Takeichi, 2009). In adherens junctions, cadherins form homophilic (more rarely heterophilic) calcium-dependent interactions with other cadherin molecules from neighboring cells. The cytoplasmic domain of cadherins is highly conserved among metazoans, distinguishing classical cadherins from other cadherin subfamilies (Hulpiau and van Roy, 2011; Oda and Takeichi, 2011). The cytoplasmic domain contains β-catenin and p120 binding sites. Catenins connect cadherins with the actin cytoskeleton in a dynamic manner (Meng and Takeichi, 2009). In comparison with other cadherin subfamilies, classical cadherins are unique in that they show the most noticeable variation in their extracellular region among different species (Hulpiau and van Roy, 2011). Indeed, the extracellular domain consists of a variable number of cadherin repeats of about 110 amino acids each and, depending on the species, laminin G and epidermal growth factor (EGF)-like domains.

During development, the regulation of specific cadherin expression correlates with the formation of new tissues. For instance, folding of the neural tube in vertebrates occurs in parallel with downregulation of epithelial cadherin (E-cadherin) and upregulation of neuronal cadherin (N-cadherin) (Nandadasa et al., 2009). Such cadherin switches are characteristic of several different morphogenetic processes, such as gastrulation and neural crest development (Basilicata et al., 2016; Dady et al., 2012; Detrick et al., 1990; Giger and David, 2017; Hatta and Takeichi, 1986; Pla et al., 2001; Rogers et al., 2013; Scarpa et al., 2015; Schäfer et al., 2014; Shoval et al., 2007). During mesoderm formation of Drosophila melanogaster, Dme_E-cadherin becomes replaced by Dme_N-cadherin (Oda et al., 1998), similar to the switch from E- to N-cadherin during mesoderm formation in chicken (Hatta and Takeichi, 1986). It has also been shown that N-cadherin expression triggers active endodermal cell migration, which leads to cell segregation and germ layer formation (Ninomiya et al., 2012). Moreover, a cadherin switch allows efficient Wnt, bone morphogenetic protein (BMP) and fibroblast growth factor (FGF) signaling, which are required for proper mesoderm differentiation in both the fruit fly and mouse (Basilicata et al., 2016; Giger and David, 2017; Ninomiya et al., 2012; Schäfer et al., 2014). For example, N-cadherin can interact with the FGF receptor and modulate the signaling pathway (Francavilla et al., 2009; Williams et al., 1994). Therefore, accurate control of the expression of cadherins is important for proper cell movements during gastrulation (e.g. epiboly) and for convergence and extension of the tissue during axis elongation (Babb and Marrs, 2004; Basilicata et al., 2016; Shimizu et al., 2005; Winklbauer, 2012).

Although the role of cadherins has been studied in model bilaterian species, very little is known about their role in diploblastic organisms such as cnidarians. Most of our knowledge on cell adhesion molecules in cnidarians is restricted to genome analyses (Hulpiau and van Roy, 2009, 2011; Tucker and Adams, 2014) or biochemical studies (Clarke et al., 2016). During the last two decades, the sea anemone Nematostella vectensis has become one of the prime model organisms for studying embryonic development (Genikhovich and Technau, 2009a; Layden et al., 2016; Technau and Steele, 2011). Bioinformatic analysis of the available genome sequence of Nematostella vectensis (Putnam et al., 2007) revealed 16 different cadherins from all main groups of the cadherin superfamily (classical, flamingo, FAT, dachsous, FAT-like, protocadherins and cadherin-related proteins) (Hulpiau and van Roy, 2011). It has been shown that adherens junctions in Nematostella ultrastucturally resemble those in bilaterians (Fritzenwanker et al., 2007). However, the molecular composition of these junctions has not yet been described, and a recent report questioned the presence of adherens junctions in the inner layer of Nematostella (Salinas-Saavedra et al., 2018).

Germ layers are formed in Nematostella by invagination of the endoderm at the animal pole (Kraus and Technau, 2006; Magie et al., 2007). However, whether classical cadherins play a role in germ layer formation in non-bilaterians is not known. Here, we show that the classical cadherins of Nematostella, Cadherin 1 (Cdh1) and Cadherin 3 (Cdh3), form the adherens junctions of the epithelium of both germ layers. Germ layer differentiation in Nematostella is marked by a cadherin switch, whereby Cdh3 is downregulated in the inner endodermal layer and Cdh1 is upregulated and remains the only cadherin expressed in the endoderm. Unexpectedly, we found that, in addition to apical adherens junctions, both Cdh1 and Cdh3 are involved in cell junctions between cells on the basal-lateral side. Knockdown of cdh1 and cdh3 indicated important roles of cadherins in cell adhesion and tissue morphogenesis of this diploblastic metazoan.

Three genes encoding classical cadherins have been predicted in the genome of Nematostella vectensis, cadherin1, cadherin2 and cadherin3 (cdh1, cdh2 and cdh3) (Hulpiau and van Roy, 2011). However, our analysis of the cdh2 gene model showed that it is a fusion of two separate gene models for which we have no evidence of its complete transcription; cdh2 was not detectable by in situ hybridization. Furthermore, the hallmarks of the cadherin intracellular domain were absent. Therefore, cdh2 could either be a pseudogene or the result of incorrect assembly. Hence, this gene model was not investigated further in this study.

Hulpiau and van Roy predicted 25-32 extracellular cadherin (EC) repeats for Nematostella cadherins (Hulpiau and van Roy, 2011). However, a more recent publication reported two cadherins with 14 and 30 EC repeats, respectively (Clarke et al., 2016). We cloned both Cdh1 and Cdh2 in overlapping fragments of 2-3 kb length, resulting in full-length cDNA clones of >13 kb, predicting a protein size of about 480 kDa for both cadherins. The protein model suggests a structure similar to classical cadherins, composed of a typical intracellular domain with binding sites for β-catenin and p120, and a large extracellular domain consisting of three EGF-like and two interspaced laminin G (LamG) domains, followed by 30 (Cdh1) or 31 (Cdh3) EC repeats (Fig. 1). This is similar to the original model of Hulpiau and van Roy and we therefore followed their gene terminology. By comparison, fruit fly cadherin has 17 EC repeats, chick cadherin 13 EC repeats and mouse cadherin only 5 EC repeats (Hulpiau and van Roy, 2011).

We also interrogated the genomes and transcriptomes of several other cnidarians and found that all investigated cnidarian cadherins have 30-32 EC domains and the EGF/LamG domains in the extracellular part. Notably, corals and hydrozoans had only a single classical cadherin, whereas the sea anemones underwent a lineage-specific gene duplication (Fig. 1; Fig. S1). Mammals have lost the extracellular EGF and LamG domains and have retained only a few EC domains (Hulpiau and van Roy, 2011). Interestingly, platypus has several paralogs of the short version with no EGF/LamG domains, typical for mammals, but also one classical cadherin with EGF/LamG domains and many EC domains, like other non-mammals. This suggests that an ancestral “long” cadherin gene duplicated in the ancestor of placental mammals and platypus and one of the duplicates underwent a drastic loss of EC and EGF/LamG domains. Platypus has kept both versions, whereas other mammals have retained only duplicates of the short classical cadherin version.

To characterize the pattern of classical cadherin expression during normal development, in situ hybridization was performed on developmental stages from early cleavage through adult polyp. cdh3 was maternally expressed at significant levels, detectable at the earliest cleavage stages. cdh3 was then strongly expressed in all cells from the egg until the gastrula stage (Fig. 2A-D,M). During early gastrulation, cdh3 expression decreased in the presumptive endoderm (Fig. 2B,C) and was completely downregulated in the endoderm by the planula stage (Fig. 2B-E).

By comparison, cdh1 expression could not be detected by in situ hybridization until the early gastrula stage (Fig. 2G-I), although RNAseq data suggest that it is maternally expressed at lower levels (Casper et al., 2018). During gastrulation, cdh1 expression first appeared and intensified in the pre-endodermal plate (Fig. 2H,I). At the late gastrula stage, cdh1 started to be expressed in the aboral ectoderm and then expanded orally during planula development (Fig. 2J,K). Interestingly, at the late planula stage, the strongest cdh1 expression was detected in the endoderm and in a subpopulation of ectodermal cells, which gave rise to a sensory apical organ (Fig. 2K).

In primary polyps, cdh3 expression remained strongly expressed in the tentacles and pharynx but weakly in the body-wall ectoderm (Fig. 2F). Interestingly, cdh3 in juveniles and adults was detectable only in the ectoderm of the pharynx and tentacles, ciliated tract, septal filaments and developing eggs, with hardly any expression in the body-wall ectoderm (Fig. S2A,B). Almost complementary to that, cdh1 was expressed both in the ectoderm and endoderm, but was completely excluded from the ectoderm of the pharynx and tentacles (Fig. 2L). In juveniles, cdh1 was expressed in the endoderm and body-wall ectoderm, but not in the ectoderm of most of the pharynx. Interestingly, the part of the pharynx carrying siphonoglyph and the ciliary tract below the pharynx specifically expressed low levels of cdh1 (Fig. S2C-H). In adults, cdh1 was expressed in the body-wall endoderm and in small oogonia (Fig. 2N; Fig. S2I).

We wished to visualize the subcellular localization of Cdh3 protein during development. We generated specific polyclonal and monoclonal antibodies against two recombinant fragments of Cdh1 and three peptides of Cdh3, respectively (Fig. S3). All antibodies against the different fragments and peptides consistently showed the same pattern for Cdh1 and Cdh3, respectively (Fig. S4) (Madeira et al., 2019). Immunocytochemistry experiments were carried out at all stages of development. Cdh3 protein had already accumulated at the apical cell junctions at the first cell divisions, suggesting a role in establishing early cell polarity. It was also detectable in less confined areas at the lateral contacts between cells (Fig. 3A-C). Interestingly, cells maintained their polarity during cell divisions. In contrast to the Par system proteins (Ragkousi et al., 2017; Salinas-Saavedra et al., 2018), Cdh3 stayed localized at the apical cell junctions at different cell cycle stages (Fig. 4). It is possible that the polarized Cdh3 at the junctions guides the Par system proteins during their transient loss of polarity during cell division. Later, during blastoderm formation, apical cell junctions became more pronounced (Fig. 3D-F). Strikingly, we found that Cdh3 also localized on the basal-lateral side of the cells (Fig. 3D-L), in addition to the apical localization. Ultrastructural analysis by transmission electron microscopy (TEM) revealed that the cell-cell junctions at the basal side resembled the adherens junctions at the apical side (Fig. 3M,N). However, during the blastula stage, asynchronously dividing cells transiently lost the basal-lateral localization of Cdh3 (Fig. 3E, yellow star), similar to the early cleavage stage, when cells divided synchronously (Fig. 3B,C, Fig. 4). Thus, Nematostella has a unique epithelium, where cells form cell-cell contacts on both apical and basal sides. These Cdh3-positive junctions developed before any contact to an endodermal layer or presence of the mesoglea, the extracellular matrix of Cnidaria. This is remarkable and, to our knowledge, has not been described in any other animal. Interestingly, as the pre-endodermal plate began to invaginate and the cells adopted a partial epithelial-to-mesenchymal transition (EMT) phenotype, Cdh3 disappeared from the basal junctions of the invaginating cells (Fig. 3G-I). Meanwhile, ectodermal cells of the blastoderm retained both apical and basal cell contacts. As the pre-endodermal cells lost basal junctions, its epithelium became less rigid and columnar. Pre-endodermal cells formed filopodia and became more motile on the basal side (Fig. 3O). This event is possibly one of the crucial steps of the incomplete EMT, which pre-endodermal cells undergo during gastrulation (Kraus and Technau, 2006; Salinas-Saavedra et al., 2018). Notably, apical cell junctions expressing Cdh3 were preserved in the pre-endodermal cells during the course of gastrulation (Fig. 3J-L).

After the invagination process was complete, Cdh3 fully disappeared from the cell junctions of the endoderm, concordant with the decrease in mRNA expression in the whole endoderm (Fig. 2C-F). Cdh3 remained expressed exclusively in the ectoderm, forming apical and basal cell junctions (Fig. 5A-E). Notably, although the boundary between ectoderm and endoderm is very difficult to discern by morphological criteria, Cdh3 localization at the cell junctions in the pharynx precisely marked the boundary between the last ectodermal cell and the first endodermal cell (Fig. 5B,E). At the polyp stage, Cdh3 remained exclusively expressed in the ectoderm, with especially strong expression in the pharynx and tentacles (Fig. 5F,G).

After completion of gastrulation, Cdh1 protein formed pronounced cellular junctions. In early planula larvae, Cdh1 localized to the apical and basal junctions of the endoderm (Fig. 6A-D). Hence, formation of the endoderm was marked by a cadherin switch from Cdh3 to Cdh1.

It should be noted that the transcriptome data suggested some maternal deposition of cdh1 in the embryo, even though in situ hybridization did not detect cdh1 until the gastrula stage. Indeed, the anti-Cdh1 antibody detected a fuzzy signal beneath the apical cell membrane in all cells at the early gastrula stage, which might be maternal Cdh1 protein that had not yet localized to the cell junctions.

In addition to endodermal expression, Cdh1 was strongly expressed in the apical organ ectoderm and then expanded into a wider domain in the aboral ectoderm, where Cdh1 and Cdh3 were co-expressed (Fig. 6). At the ectodermal surface, expression of Cdh1 decreased along a gradient toward the oral pole (Fig. 6L). Interestingly, the ectodermal cell population that gave rise to the apical tuft was also different from the rest of the ectoderm in terms of cadherin expression. These cells lost Cdh3 basal junctions, but kept the apical junctions (Fig. 5C). This specific arrangement might be connected with the special function of these cells (Fig. S5). Indeed, the loss of Cdh3 expression in the ectodermal apical tuft cells went hand-in-hand with upregulation of Cdh1 in these cells (Fig. 2K, Fig. 6A,C).

We have demonstrated that Cdh3 is the major component of adhesion complexes during cleavage and gastrulation and is present in all cells until the late gastrula stage. Cdh3 formed apical and basal cell junctions in the blastodermal epithelium, which during invagination of the pre-endodermal plate disappeared from basal cell junctions of the future endoderm. Further endoderm differentiation led to complete Cdh3 to Cdh1 replacement. Therefore, there is a distinct boundary between ectoderm and endoderm, which is defined by the localization of Cdh1 and Cdh3.

A recent biochemical study showed that the intracellular domain of classical cadherins can form a ternary complex with α-catenin and β-catenin (Clarke et al., 2016). To explore further the molecular composition of the cadherin cell junctions, we co-stained Nematostella embryos with the antibody against Cdh3 and with the previously described Nematostella antibody against β-catenin (Leclère et al., 2016; Salinas-Saavedra et al., 2018). At blastula stage, Cdh3 was co-localized with β-catenin at the apical junctions, whereas basal junctions did not show such pronounced co-localization (Fig. 7A-C,G-I). Interestingly, at the planula stage, β-catenin was detected only in the body wall ectoderm but not in the ectodermal pharynx nor the endoderm (Fig. 7D-F; Fig. S6). These results could mean that not all the cell contacts of Nematostella epithelium contain β-catenin, in line with other recent findings (Salinas-Saavedra et al., 2018). This is surprising, as no ultrastructural differences in the junctions of endoderm and ectoderm could be detected (Fig. S7).

To examine the function of cadherins, we performed knockdown experiments using morpholinos (MOs) and short hairpin RNA (shRNA). First, we injected independently two non-overlapping translation-blocking cdh3 MOs. However, we could still detect Cdh3 in apical and basal cell junctions in the whole mount MO-injected embryos (Fig. 8). Indeed, on the ultrastructural level, the adherens cell junctions looked similar in morphants and in control embryos (Fig. 7D,H; Fig. 8C,G). These results can be explained by the significant maternal deposition of mRNA and protein (Figs 2M, 3B). However, development of Cdh3 morphants was arrested after the gastrula stage, presumably due to the block of translation of zygotically expressed cdh3. As a result, when Cdh3 protein became limited, post-gastrula embryos were unable to develop further (Fig. 8A,B).

The mild knockdown effect on the presence of Cdh3 in the junctions also suggests that there is relatively little turnover in established junctions. Therefore, to assess the function of Cdh3 in establishing new cell junctions we used an aggregate assay. Nematostella gastrulae can be dissociated into single cells and small clusters and can be re-aggregated by centrifugation into cell aggregates (Kirillova et al., 2018). We followed the establishment of cell contacts and the formation of the epithelium in developing cell aggregates (Fig. 9). Dissociated cells did not show any signs of polarization: Cdh3 was not localized to any side (Fig. 9C). Cdh3 became localized to the apical side of the outer cells of the aggregate only 30 min after re-aggregation, and the first signs of epithelialization became apparent (Fig. 9E,F,M). At 12 h after re-aggregation, the outer epithelial layer was completely formed and Cdh3 was localized at the apical and basal cell junctions (Fig. 9H,I). We have previously reported that the two epithelial layers (ectoderm and endoderm) are formed 24 h after re-aggregation (Kirillova et al., 2018). Both cell layers possessed basal and apical cadherin cell junctions (Fig. 9K,L,S). Cdh1 began to be expressed in both ectoderm and endoderm at 24 h of aggregate development (Fig. 9N,Q,T,W). Similar to the normal embryo, the ectoderm expressed both Cdh1 and Cdh3, whereas the endoderm expressed exclusively Cdh1 (Fig. 9U,X).

To address the question of how Cdh3 downregulation influences the establishment of new cell contacts in the aggregate, we dissociated equal amounts of cdh3 MO-injected gastrulae and standard MO-injected gastrulae (as a control). The first difference we observed was that the size of the aggregates from cdh3 morphant cells was significantly smaller than control aggregates (P<0.0001) (Fig. 10M-O). Moreover, aggregates from cdh3 MO-injected embryos started to fall apart into cells immediately after re-aggregation (Fig. 10; Movies 1 and 2). Cdh3 knockdown in the aggregate at the protein level was shown by immunostaining (Fig. 10I,J). Ultrastructural imaging with TEM confirmed that cells in Cdh3 MO aggregates did not form well-defined subapical adherens junctions, whereas cell contacts were well developed in the control aggregate (Fig. 10K,L). Interestingly, cells in the Cdh3 MO aggregates made lamella-like protrusions extending to the neighboring cell on the apical surface (Fig. 10K). These results show that cdh3 knockdown impairs the de novo formation of cell contacts, although it does not affect the earlier established contacts built from the maternal protein.

To further explore the role of Cdh1 protein, we downregulated cdh1 using an independent approach, shRNA-mediated knockdown (He et al., 2018). As in MO knockdown, shRNA knockdown led to a significant decrease in Cdh1 protein, as assayed by immunohistochemistry (Fig. 11; Fig. S8). Although early development (including gastrulation) appeared largely unaffected, mesenteries did not form upon cdh1 knockdown in the subsequent planula stage. In all MO- and shRNA-injected embryos, mesenteries were absent or impaired, whereas eight mesenteries developed in all control embryos at this stage (Fig. 11, Fig. 12D,H).

In addition to the predominant expression in the endoderm, cdh1 was also expressed in the apical tuft region of the ectoderm (Fig. 2K). Interestingly, cdh1 knockdown abolishes expression of FGFa1, which is responsible for apical organ development (Rentzsch et al., 2008). In most cdh1 MO-injected embryos, the apical organ did not form and there was lack of FGFa1 expression (Fig. 12). These results suggest that Cdh1 is crucial for morphogenesis and differentiation of the endoderm as well as for development of the apical organ.

Although proteins with cadherin domains are present in choanoflagellates, cadherins with intracellular catenin binding domains are an important class of cell adhesion molecules that arose only in metazoans (Nichols et al., 2012). Cadherins mediate not only cell adhesion between epithelial cells, but are strongly involved in the differentiation of specific cell types. Recently, cadherins have also been shown to convey mechanotransduction (i.e. activation of gene expression in the nucleus in response to mechanical stress), which is mediated by β-catenin in Drosophila and Nematostella (Iyer et al., 2019; Pukhlyakova et al., 2018; Röper et al., 2018). However, most studies on the role of cadherins have been carried out in bilaterian model organisms such as mouse or Drosophila. Here, we show the localization and function of both classical cadherins in a representative of the Cnidaria, the sea anemone Nematostella vectensis. Phylogenetic analysis suggests that sea anemones have duplicated an ancestral classical cadherin, whereas corals and hydrozoans have retained a single copy. The two investigated cadherin genes code for large proteins with 31-32 EC domains each, largely confirming previous predictions from the genome (Hulpiau and van Roy, 2011) and gene models based on our transcriptome assembly (Fredman et al., 2013). This significantly extends the structure of the recently published gene model for Cdh3 (termed Cad1 in Clarke et al., 2016). Thus, the classical cadherins of cnidarians and other non-bilaterians are substantially larger than those of most bilaterians and their extracellular domain structure is reminiscent of the FAT-like proteins (Hulpiau and van Roy, 2009, 2011). It will be interesting to determine which extracellular domains are engaged in homophilic or heterophilic interactions.

Interestingly, both cadherins localized to apical junctions as well as to basal cell-cell junctions in the epithelial cells of both ectoderm and endoderm (Fig. 13A). Electron and confocal microscopy analyses showed actin filaments attached to the junction, suggesting that these are adherens junctions (Fig. 3G,N, Fig. 8D,H). This is in contrast to a recent study claiming that the endodermal epithelium does not contain adherens junctions, since neither Par complex components nor β-catenin could be detected (Salinas-Saavedra et al., 2018). Yet, in line with this study (Salinas-Saavedra et al., 2018), we could detect β-catenin in the apical adherens junctions and weakly in the basal junction of the ectoderm, but not in the pharyngeal ectoderm and the endoderm (Fig. 7; Fig. S6). This could indicate that the basal junctions in the ectoderm and all endodermal junctions are qualitatively different. However, apical adherens junctions in the ectoderm and endoderm have a very similar structure at the ultrastuctural level (Fig. S7). As we observed co-localized actin fibers at these junctions, we assume that another protein replaced β-catenin or that β-catenin was not detected at these junctions. Indeed, we note that the antibody also failed to stain nuclear β-catenin after early cleavage stages. Therefore, as a cautionary note, we cannot fully rule out that the failure to stain β-catenin in the pharynx and the endoderm was due to technical reasons.

To our knowledge, the basal-lateral junctions involving specific cadherins have not yet been described in other animals, but fuzzy basal-lateral localization of cadherins has been observed in other systems. For instance, the midgut epithelium of Drosophila shows a basal-lateral localization of a cadherin (Chen et al., 2018), albeit much less defined than described here for Nematostella.

Basal junctions might be an innovation of Cnidaria and play a crucial role in morphogenesis of the epithelium. Formation of the basal junctions might be connected to the special properties and functions of the cnidarian epithelium. For example, Hydra epithelia are composed of multifunctional epithelio-muscular cells. These cells form basal myonemes, connected between neighboring cells by desmosomal-like junctions (Seybold et al., 2016). These basal connections could be associated with the contractile actin bundles and used for the increased synchronized contractile activity within large epithelia sheets. Basal cellular contractions also have a major contribution in the process of bud formation in Hydra (Holz et al., 2017). Such basal contacts are absent from bilaterian embryos, which are mainly connected by apical junction belts.

Establishment of the adherens junctions is crucial for normal development of the embryo. Knockdown of cdh3 in normal embryos does not lead to dissociation of embryonic tissue, suggesting that maternally expressed cadherin protein localized in cell junctions might have a slow turnover and be sufficient for the early stages of development. This is consistent with the results of knockdown of E-cadherin in mouse embryos (Capaldo and Macara, 2007). However, proper formation of epithelial layers is disrupted in embryonic aggregates in response to knockdown of cdh3. Notably, although knockdown of endodermal cdh1 does not disrupt gastrulation, the endoderm does not develop endodermal structures such as mesenteries. Thus, proper development of the inner germ layer is dependent on the expression of Cdh1.

The role of cadherins in the formation of germ layers in a diploblast animal is of particular interest, as we might learn about the evolution and potential homology of germ layers. We found that the formation of the inner layer is accompanied by a stepwise cadherin switch. At the blastula stage, Cdh3 forms apical and basal adherens junctions. The onset of gastrulation is characterized by a change in shape of endodermal cells, which adopt a partial EMT phenotype: apical constriction of cells, loss of Cdh3-positive basal junctions, migration of nuclei basally, development of filopodia and an increase in cellular motility (Fig. 13B). We propose that the changes in the adhesion properties of the endodermal cells are crucial for the morphogenetic behavior and further differentiation. In a second step, after completion of invagination, Cdh3 also disappears from the apical junctions in the endoderm and is replaced by Cdh1, both at apical and basal junctions of the endoderm. Thus, we observed a cadherin switch in Nematostella that is analogous to the cadherin switch in vertebrates and insects. As cdh1 and cdh3, like E- and N-cadherins in mammals and insects, are lineage-specific duplications (Fig. 1), we conclude that the cadherin switch evolved convergently in these animals.

However, although Cdh3 is not expressed in the endoderm after the gastrula stage, Cdh1 shows partially overlapping expression with Cdh3 in the ectoderm. Cdh1 seems to form a decreasing gradient from aboral to oral, but the significance of this gradient is unclear at this point. Notably, the oral region and tentacles are completely devoid of Cdh1 expression. Interestingly, cdh1 knockdown does not disrupt oral patterning. For example, expression of the blastopore marker brachyury was normal. However, expression of the aboral patterning gene FGFa1 was abolished (Fig. 12). Because the aboral part is an area of strong Cdh1 expression, we assume that normal Cdh1 expression is necessary for FGF signaling and apical organ development. Our results show that Nematostella cadherins are important for germ layer morphogenesis and the maintenance of tissue integrity. However, so far we have no evidence that cadherins play a role in initial germ layer differentiation, as shown similarly for the knockdown of α-catenin, another component of the adhesion junction complex (Clarke et al., 2019). We conclude that, as for bilaterians (Basilicata et al., 2016; Giger and David, 2017; Huang et al., 2016; Nakagawa and Takeichi, 1998; Pla et al., 2001; Schäfer et al., 2014; Shoval et al., 2007), different combinations and concentrations of Cdh1 and Cdh3 convey different tissue properties and identities in different regions of the developing embryo. Thus, the combinatorial and differential use of cadherins is a recurring feature of metazoans (Fig. 1, Fig. 13C; Fig. S1), although the paralogous molecules have evolved independently.

Interestingly, our phylogenetic analysis of the classical cadherins showed that hydrozoans and stony corals have only cadherin protein, which groups basally with two classical cadherins of sea anemones, suggesting that cdh1 and cdh3 arose by a lineage-specific gene duplication within the sea anemones (Fig. 1). The expression and function of the single cadherin in other cnidarians is unknown. However, they do have a dachsous gene, which also encodes a cytoplasmic cadherin domain and has a similar structure to classical cadherins, except that they lack the EGF/LamG domains found in most invertebrate cadherins. It remains to be shown whether Dachsous and classical cadherin could interact during early germ layer formation in other cnidarians.

Our study has established that cadherins play an important role in the formation and differentiation of the germ layers in a diploblastic animal. This revives the question of which germ layers in Bilateria these two cell layers are homologous with. Traditionally, they have been homologized with the endoderm and ectoderm, with the mesoderm missing. The identification of a number of mesodermal transcription factors in cnidarians and their expression in the endoderm led to the notion of an inner “mesendoderm” (Fritzenwanker et al., 2004; Kumburegama et al., 2011; Martindale, 2004; Salinas-Saavedra et al., 2018; Scholz and Technau, 2003). However, recent analysis of many endodermal and mesodermal marker genes suggests that segregation has already taken place in the Nematostella polyp. In fact, the inner layer corresponds to mesoderm, whereas all endodermal functions reside in the ectodermally derived extensions of the pharynx, the septal filaments (Hashimshony, 2017; Steinmetz et al., 2017). In the light of those findings, it is interesting to note that Cdh1 is specific to the inner cell layer, which corresponds to the mesoderm of bilaterians. Notably, this cell layer also expresses the zinc finger transcription factor snailA (Fritzenwanker et al., 2004; Martindale, 2004). Snail proteins regulate the downregulation of E-cadherin in vertebrates and insects in the ingressing mesoderm (Nieto, 2002). In line with this, snail genes appear to play a role in regulating invagination and partial EMT in Nematostella (Salinas-Saavedra et al., 2018). It will be of interest to investigate how cadherins are regulated by Snail in Nematostella.

This first analysis of the expression and function of classical cadherins in a diploblast shows that these molecules play a conserved role in cell adhesion, tissue morphogenesis and germ layer specification during embryogenesis. Invaginating cells show partial EMT, accompanied by a cadherin switch. The evolutionarily recurring mechanism of a cadherin switch suggests that the evolution of germ layer formation and tissue morphogenesis is facilitated by the differential expression of cadherins.

Animals were kept in artificial seawater at 18°C the dark. Spawning was induced by temperature shift to 24°C and light exposure over 10 h (Fritzenwanker and Technau, 2002). In vitro fertilized embryos were collected and kept at 21°C as described (Fritzenwanker and Technau, 2002; Genikhovich and Technau, 2009c).

To retrieve the coding sequences of cdh1 and cdh3 genes the 1-3 kb overlapping coding fragments of cdh1 and cdh3 were amplified from cDNA of mixed embryonic stages, cloned using pJet1.2/blunt vector system (Thermo Fisher Scientific) and sequenced. The full-length sequences of Cdh1 and Cdh3 have been deposited in GenBank (accession numbers MK253651 and MK253652).

Assembled cdh1 and cdh3 protein coding sequences were derived in silico using Expasy translation tool (Artimo et al., 2012). Cadherin protein domain annotation was performed using SMART protein domain annotation resource (Letunic and Bork, 2018).

Knockdowns of cdh1 and cdh3 were performed by independent zygote injections of two non-overlapping translation blocking morpholinos (Gene Tools): cdh1MO1 5′-CCGGCCAGCACTCATTTTGTGGCTA-3′, cdh1MO2 5′-ACCCGTGAGTTTAAAAACCCATAGC-3′; cdh3MO1 5′-ACGAGTTGCGGTGAACGAAAATAAC-3′, cdh3MO2 5′-TAGCAGAACCGTCCAGTCCCATATC-3′ at concentrations of 500 μM. Standard morpholino injection at 500 μM was used as a control; SdtMO 5′-CCTCTTACCTCAGTTACAATTTATA-3′.

Non-overlapping morpholinos for cdh1 and cdh3 knockdown had similar phenotypes.

Injection equipment used: FemtoJet (Eppendorf), CellTram Vario (Eppendorf), micromanipulator (Narishige). Needles were pulled from the glass capillaries type GB 100TF-10 (Science Products) with a micropipette puller (Sutter Instrument, Model P-97). We used holding capillaries from Eppendorf for the injection (Renfer et al., 2010).

cdh1 shRNA design and synthesis were performed as described (He et al., 2018). The following primers were used for cdh1 shRNA synthesis: cdh1 shRNA forward, 5′-TAATACGACTCACTATAGAAGCGCGCTCAGGTAAATGTTTCAAGAGA-3′; cdh1 shRNA reverse, 5′-AAGAAGCACGTTCGGGTAAATGTTCTCTTGAAACATTTACCTGAGCGC-3′.

Purified shRNA was injected into zygotes at a concentration of 500 ng/µl. As a negative control, shRNA against mOrange was injected at 500 ng/µl. Uninjected embryos from the same batch were used as a control for injection. After injection, embryos were raised at 21°C.

To generate antibodies against Cdh1, we expressed the protein domains cdh1:domain1 (extracellular) and cdh1:domain3 (intracellular) in E.coli. The fragment sequences were as follows: Cdh1 domain1 (extracellular), NAPKDGSLLIIVNAYDGNFTGGVIGKPYYQDDDFDGDENTYELNSQSPGSYFRVNEGNGDITAAPMIPMGEYNLKIRVTEKKDSPSTVTSSVRVLVRRIDKEAVDNGVAVEFTDMRKVGYFVGDYYKGFEDVLASTLGVPTGDIKIFSVQKAHDNGLAVVVFFTVAAKDSYMPHWDVVSKLVDAKKPLESLGLKVSRLGMD; and Cdh1 domain3 (intracellular), RRPEPVVVYADSTDTGHVHDNVRLYHDDGGGEEDNLGYDITKLMKYTYIETTIAPPSVAPSKASEDKISTSSDQPLLQGRPPDAVFGLTGKEPGPKMPKYMEGDDVGDFITTRVKITDREVFLAVDELHIYRYEGDDTDVD.

The recombinant protein fragments were purified by column-based affinity chromatography and used for immunization. Specifically, the extracellular fragment was used for immunization of two rats (polyclonal Cdh1 antibodies 1 and 2) and the intracellular fragment used for immunization of a rabbit (polyclonal Cdh1 antibody 3). All Cdh1 antibodies resulted in the same staining pattern (see Fig. S4). Cdh1 antibody2 was used for most experiments in this paper.

For visualization of Cdh3, monoclonal antibodies were produced in mice. The following peptides were used for immunization: SSSDRNRPPV (for Cdh3 antibody1) and DEKDPPQFSQ (for Cdh3 antibody2). Both epitopes are located in the extracellular part of Cdh3 in the third and seventh cadherin repeats, respectively. Both antibody clones (Cdh3 antibody1 and Cdh3 antibody2) resulted in the same staining patterns (see Fig. S4). Cdh3 antibody2 was used for antibody staining in this paper.

For Cdh1 antibody staining, embryos were fixed for 1 h at 4°C with Lavdovsky's fixative (3.7% formaldehyde (FA), 50% ethanol, 4% acetic acid). For staining of Cdh3 antibody, β-catenin antibody and phalloidin, embryos were fixed for 1 h with 3.7% FA in PBS at 4°C. Primary polyps were relaxed prior the fixation in 0.1 M MgCl₂ in Nematostella medium for 10 min. After fixation, embryos were incubated on ice in ice-cold acetone (chilled at −20°C) for 7 min followed by five washes with PBSTx 0.2% (PBS with 0.2% of TritonX-100).

Then embryos were incubated in blocking solution [20% sheep serum, 1% bovine serum albumin (BSA) in PBSTx (0.2%)] for 2 h at room temperature (RT). Primary mouse anti-Cdh3 antibody (1:1000), rabbit β-catenin antibody (1:500; Sigma-Aldrich C2206) (Leclère et al., 2016; Salinas-Saavedra et al., 2018) and/or rat/rabbit anti-Cdh1 antibodies (1:500) were diluted in blocking solution and incubated with the embryos overnight at 4°C, followed by washing in PBSTx 0.2% at RT (10×10 min each). After incubation in blocking solution for 2 h at RT, embryos were placed in a secondary antibody solution of goat anti-mouse Alexa Fluor 568 antibodies (1:1000, Thermo Fischer Scientific A11019), goat anti-rat antibody DyLight 488-conjugated (1:1000, Rockland, 612-141-120) and DAPI overnight at 4°C. When fixed with FA, phalloidin Alexa Fluor 488 (1:30, Thermo Fisher Scientific) was added to the secondary antibody solution, because phalloidin staining is not compatible with the Lavdovsky's fixation. Embryos were washed in PBSTx 0.2% at RT (10×10 min each) and infiltrated with Vectashield antifade mounting medium (Vector laboratories) at 4°C overnight. For β-catenin staining of the embryo sections, fixed embryos were embedded in 10% gelatin in PBS. Gelatin blocks were postfixed in 3.7% FA in PBS overnight at 4°C and sectioned on a vibratome Leica VT 1200S. Embryo sections (50 μm) were stained with β-catenin antibody, phalloidin and DAPI as described for the whole-mount embryos. Imaging was performed with a Leica TCS SP5 DM-6000 confocal microscope.

In situ hybridizations of embryos were conducted as previously described (Genikhovich and Technau, 2009b; Kraus et al., 2016). The following regions of the coding sequence of cadherins were used to produce the in situ hybridization probes: 7054-9126 bp for cdh1 and 2728-5091 bp for cdh3. Adult animals and juveniles were relaxed for 20 min in 0.1 M MgCl₂ solution in Nematostella medium followed by fixation and in situ hybridization as described (Steinmetz et al., 2017). After in situ hybridization embryos, adult and juvenile pieces were embedded in 10% gelatin in PBS. Gelatin blocks were postfixed in 3.7% FA in PBS overnight at 4°C and sectioned on a vibratome Leica VT 1200S. Embryos and adult and juvenile 50 μm sections were embedded in 80% glycerol and imaged with a Nikon Eclipse 80i compound microscope equipped with DIC optics and Zeiss AxioCam camera.

Time-lapse imaging was carried out using a Nikon Eclipse 80i compound microscope. Pictures were taken with a Zeiss AxioCam camera. Time-lapse movies were made using FIJI software (Schindelin et al., 2012).

Transmission electron microscopy was performed as previously described (Fritzenwanker et al., 2007).

The protein complements of Mus musculus (GRCm38) (Schneider et al., 2017) and Drosophila melanogaster (FB2018_03) (Thurmond et al., 2019) were downloaded from Ensembl (Zerbino et al., 2018); Strongylocentrotus purpuratus, Capitella teleta, Lottia gigantica and Tribolium castaneum from Ensembl metazoan (Kersey et al., 2018); Hydra vulgaris and Ornithorhynchus anatinus from RefSeq at NCBI (O'Leary et al., 2016); Acropora millepora (PRJNA74409) (Moya et al., 2012), Anemonia viridis (PRJNA260824) (Rachamim et al., 2015), Exaptasia pallida (PRJNA386175) (Baumgarten et al., 2015) and Stylophora pistilata (PRJNA281535) (Voolstra et al., 2017) from the Sequence Read Archive at NCBI (Leinonen et al., 2011); and Acropora digitifera from marine genomics at OIST (Shinzato et al., 2011). Sequences were selected that had a significant domain hit (domE<1×10⁻⁵) to the cadherin cytoplasmic Pfam family (PF01049) according to HMMER 3.2.1 (Finn et al., 2011). When multiple isoforms were present, the longest one was used. The genes were filtered against truncated and misassembled gene models manually. Sequences were aligned using MAFFT v7.307 in E-INS-i mode and a maximum of 1000 iterations of refinement (Katoh and Standley, 2013). The WAG+F+R6 model was determined as optimal by the Bayesian Information Criterion using ModelFinder (Kalyaanamoorthy et al., 2017). This was used to infer a maximum likelihood tree using IQTREE (Nguyen et al., 2015). Support values were determined with 1000 standard bootstrap replicates. Domain architectures were determined using standalone interproscan (Mitchell et al., 2019).

Cdh3 morpholino (MO) embryos and standard MO control embryos (50 of each) were dissociated into cells at the gastrula stage. Cell aggregates were generated by slow centrifugation as described (Kirillova et al., 2018) and photographed immediately after centrifugation. The size of the aggregates was analyzed with FIJI software. (FIJI/Image/Adjust/Threshold tool and FIJI/Analyze/Analyze particles). The threshold was set to 50. The experiment was repeated three times.

Images were processed and adjusted for brightness and contrast using FIJI software (Schindelin et al., 2012). Focus stacking of ISH images was done using Helicon Focus software (Helicon Soft, Kharkov, Ukraine). Images were cropped and assembled into the figures; schemes were made using Adobe Illustrator CS6 software (Adobe, San Jose, USA).

We thank Eduard Renfer and Sarah Streinzer for cloning the cDNAs for the in situ probes of cdh1 and cdh3, and for performing initial in situ hybridizations of these genes. We are very grateful to Prof. Bert Hobmayer for the fruitful discussions and to Dr Alison Cole for critical reading of the manuscript. We thank Rohit Dnyansagar for help with generating the phylogenetic tree and Boris Osadchenko for imaging some of the transmission electron microscopy samples. We thank Vienna Biocenter Facilities (VBCF) for support with recombinant cadherin protein generation and purification, and the Core Facility for Cell Imaging and Ultrastructure Research of the University of Vienna (CIUS) for assistance.