Repetitive accumulation of interstitial cells generates the branched structure of Cladonema medusa tentacles

During development, tissues and organs take various forms according to their particular functions. Branched structures are commonly found in the organs of both animals and plants, a morphology which exposes a relatively larger epithelial surface area available for biological interaction with the surroundings. Recent molecular, cellular and genetic studies on branching morphogenesis in both vertebrates and insects have revealed fundamental mechanisms apparently conserved across organs and species (Goodwin and Nelson, 2020; Varner and Nelson, 2014; Wang et al., 2017). Examples of such mechanisms include the generation of branches in the epithelium through interactions with the adjacent tissues, such as those of mesoderm-derived cells (vertebrates) and of epidermal and/or mesodermal cells (Drosophila), and the use of receptor tyrosine kinase (RTK) signaling, such as that occurring during the action of fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF). Ligands involved in this signaling are produced in the adjacent tissues and act on the epithelial cells to induce a variety of cell behaviors leading to the formation of branched epithelium.

New branches form through budding processes involving mainly two-step spatial and temporal control: determination of branching points, corresponding to branch bud formation, and branch elongation (Goodwin and Nelson, 2020; Varner and Nelson, 2014; Wang et al., 2017). These two steps are regulated by different mechanisms. The first step includes the specification of branch tip cells in the epithelium, which during subsequent branch elongation protrude outward as leader cells. The tip cells are specified in the Drosophila trachea and mouse retinal blood vessels by Branchless (a focally-expressed Drosophila ortholog of FGF) and VEGF, respectively, produced from the surrounding tissues (Gerhardt et al., 2003; Ghabrial and Krasnow, 2006; Sutherland et al., 1996), and in the mammary gland by epithelium-derived FGF receptor 2 (Lu et al., 2008; Parsa et al., 2008). Branch buds are also formed through epithelial folding driven by active forces such as apical constriction (Varner and Nelson, 2014). In the chicken lung, apical constriction to initiate branch formation is regulated by FGF signaling (Kim et al., 2013). In colonial hydroids, stolon branching is initiated by the formation of a plate of columnar ectodermal cells at a branching point (Kosevich, 2005).

In the second step, tip cell protrusion for branch elongation is driven by invasive migratory activities of the tip cells in Drosophila trachea and vertebrate blood vessels (Gerhardt et al., 2003; Samakovlis et al., 1996). Tip cells are also non-invasively pushed outward by cell rearrangement of the branch stalk, as observed in the mouse kidney (Chi et al., 2009; Karner et al., 2009); by oriented cell division of the stalk, in the mouse lung and kidney (Riccio et al., 2016; Schnatwinkel and Niswander, 2013); and by extensive cell proliferation in the tip of the mouse salivary gland, lung, kidney and mammary gland (Bernfield et al., 1972; Riccio et al., 2016; Scheele et al., 2017; Schnatwinkel and Niswander, 2013). Tip cell proliferation leaves behind daughter cells to constitute the branch stalk. In the mammary gland, lineage-restricted mammary stem cells form the majority of the tip cells (Scheele et al., 2017).

Branching morphogenesis in animal organs has been studied extensively in recent years, although the animals studied have been limited mostly to mammals and Drosophila. Therefore, there is currently no information about the extent to which the molecular and cellular mechanisms for branch formation are conserved or diversified across the rest of the animal kingdom. The jellyfish Cladonema pacificum (Phylum Cnidaria) is a diploblastic animal without mesoderm, but it possesses branched tentacles at the medusa stage (Fig. 1A; Fujiki et al., 2019; Schuchert, 2006). Therefore, branch formation must rely on other mechanisms not involving interactions between epithelial and mesodermal-derived cells. The C. pacificum medusa tentacle is a bi-layer tubular structure consisting of ectodermal and endodermal epithelia. Previous studies on medusa tentacle branch formation in this species have revealed the following findings (Fujiki et al., 2019). Firstly, the overall branched architecture of the tentacle is generated by repetitive addition of new branches at the proximal part of the main tentacle as it grows (Fig. 1B), reminiscent of many other branching organs in which complex structures are generated by repetition of a simple mechanism. Secondly, formation of the tentacle branch requires the activity of mitogen-activated protein kinase kinase (MAPKK also known as MEK), as treatment with the MEK inhibitor U0126 inhibits branch formation. Finally, a branch bud is formed consisting of cells with small surface areas in the future branching point of the main tentacle ectoderm layer. The identity of these small cells was unknown.

C. pacificum is a hydrozoan species, which is characterized by possessing pluripotent stem cells called interstitial cells (I-cells) (Gold and Jacobs, 2013; Leclère et al., 2016), which reside primarily in the interconnected spaces of the ectodermal epithelial cells. These I-cells are known to give rise to neurons, nematocytes, gland cells and germ cells but not ectodermal or endodermal epithelial cells in Hydra (Bosch, 2009), whereas I-cells in Hydractinia become germ cells and all somatic cells including ectodermal and endodermal epithelial cells (Gahan et al., 2016; Plickert et al., 2012). In addition to their contribution to maintaining tissue and organismal function by continuously providing various types of differentiated cells throughout life, I-cells are known to play important roles in regeneration (Gahan et al., 2016; Vogg et al., 2019).

The present study was aimed at understanding the branching mechanism in a mesoderm-less diploblastic animal species. Cells contributing to C. pacificum tentacle branch formation were identified by characterizing cells constituting the branch bud and lineage-tracing those cells for their contribution to the branch during its elongation. In addition, signaling pathway regulating accumulation of the branch bud-forming cells at the branching site for the initiation of branch formation was investigated by observing their behaviors under normal conditions and conditions in which branch formation is pharmacologically inhibited.

In the search for genes expressed in a pattern possibly related to medusa tentacle branch formation, C. pacificum orthologs of genes known as stem or germ cell markers (namely Nanos1 and Piwi but not Nanos2; Fig. S1) were found to be expressed in the ectodermal layer of branch buds (arrowheads on days 5 and 7 in Fig. 1C; see also Fig. S2A). All three genes were also expressed in the proximal part of the tentacle, at its expanded attachment to the edge of the medusa: the tentacle bulb (Fig. 1C; Fig. S2A). In Clytia hemisphaerica and Podocoryne carnea, other hydrozoan species with unbranched medusa tentacles, Piwi is expressed in the tentacle bulb but not the counterpart region of the C. pacificum branch bud (Denker et al., 2008; Seipel et al., 2004).

Nanos1 expression was evident in the buds of second (day 5) and third (day 7) branches but was weaker before and between these days (days 3, 4 and 6; Fig. 1B,C), suggesting that Nanos1-expressing cells are repetitively accumulated to form the branch bud. The expression then persisted in the tip of the elongating branch but subsided gradually (arrowheads in bottom images in Fig. 1C for the second branch). At the hydroid stage, Nanos1 expression was observed in the medusa bud emerging laterally from the hydroid body (Fig. S2B), particularly the distal part of the growing bud (top-right image in Fig. S2B). Interestingly, Nanos2 was expressed in domains almost complementary to those of Nanos1 expression (Fig. S2B), but was also not expressed in the hydroid head or the tip of hydroid tentacles (Fig. S2B).

To confirm that the Nanos1-expressing cells accumulated in the branch bud are I-cells, we used anti-β-catenin antibody 7A7 (Plickert et al., 2012; Teo et al., 2006) staining and nuclear Propidium Iodide (PI) counterstaining. I-cells were identified by their cytoplasmic signals from the anti-β-catenin antibody staining as reported also for Hydractinia echinata (Plickert et al., 2012) and by all other morphological characteristics of I-cells such as their spindle-like shape, large nucleus-to-cytoplasm ratio and prominent nucleoli (Fig. 1D). Interestingly, the staining was also able to distinguish nematocytes and epithelial cells (Fig. 1D). The nematocyte contains a large subcellular organelle, the nematocyst, and a crescent-shaped nucleus distorted by the growing nematocyst, while the epithelial cell assumes a polygonal shape.

With this staining method to identify the three cell types in the ectodermal layer, the cell composition at four different locations was firstly examined in day 5 and day 5.5 tentacles. It was found that I-cells had accumulated in the branch bud (Fig. 1D, lower image, and Fig. 1E). Both the absolute number and percentage of I-cells in the 50 µm-diameter circular sampling area of the bud were greater than those in the same-sized sampling area of the tentacle bulb (Fig. 1E). In addition, a higher percentage of I-cells was observed in a smaller circular sampling area of the ectodermal bud (Fig. 1E), further supporting the finding that I-cells are concentrated in the bud region. I-cells were also concentrated at the tip of the elongating branch on day 5.5 (Fig. 1E), consistent with the Nanos1 expression result (Fig. 1C).

The cell composition was further examined in different parts of tentacles on different days. It was found that I-cells were distributed nearly evenly in the main tentacle proximal to the first branch on day 4 (Fig. 2A, top images, Fig. 2B,D), one day before I-cells accumulate for the second branch formation. The percentage of I-cells at four different locations along the length of the proximal region of the main tentacles (Fig. 2A, top diagram) showed no significant differences between one another (Fig. 2D). After day 4, I-cells were found to accumulate in the buds of the second (day 5) and third (day 7) branches but not in the future branch-forming regions (i.e. corresponding areas on days 4 and 6, despite lack of bud formation; Fig. 2A, middle images), judging from both the number and percentage of I-cells in the respective tentacle locations (Fig. 2A, middle images, Fig. 2C,E), consistent with Nanos1-expressing cell distribution in the same period (Fig. 1C). These results support the finding that I-cells are repetitively accumulated in the branch bud regions.

Finally, the number of I-cells in the tip of the elongating branch was counted. It was found that the I-cell number in the 30 µm-diameter circular area of the tip of the second branch stayed nearly unchanged from days 5.5 through 7 (Fig. 2A, bottom images). This is in contrast with the Nanos1 expression result (Fig. 1C), in which the expression level decreased gradually as the branch elongates, suggesting that the molecular nature of the I-cells might change during branch elongation.

To understand the mechanism by which I-cells accumulate in the branch bud region, MEK signaling was inhibited by treatment with the MEK inhibitor U0126, which previously has been shown to cause failure of branch formation (Fujiki et al., 2019). The number and percentage of I-cells dramatically increased from day 4 to day 5 in the future branch-forming region (i.e. the bud on day 5 and a corresponding area on day 4, despite lack of bud formation; Fig. 3A-C). However, the number and percentage of I-cells on day 5 were significantly lower in the presence of the MEK inhibitor (Fig. 3A-C). In addition, reduction of I-cell number and percentage to a similar degree was observed on day 5 in the presence of Hydroxyurea (HU), a cell division inhibitor, and also under the simultaneous presence of both HU and U0126 (Fig. 3A-C). HU was used at a concentration (10 mM) that has been shown to completely block cell proliferation in the medusa of C. pacificum (Fujita et al., 2019). These results suggest that MEK signaling promotes I-cell accumulation through regulation of I-cell proliferation. I-cell accumulation, however, could also be controlled by other mechanisms such as I-cell migration, as the decreased level of I-cell number and percentage after treatment with U0126, HU or both was still a significant increase from that on day 4 (Fig. 3B,C).

An experiment with BrdU labeling was performed to identify the tentacle region in which I-cells proliferate. Cells with BrdU incorporated into the genome during a 1 h period on day 3.5 were found to be almost exclusively I-cells on days 4 and 5 when the labeled cells were evaluated for exclusion of PI staining from nucleoli (Fig. 3D). From day 4 to day 5, the labeled I-cell number in the future branch-forming region increased dramatically (from ∼10 to 35 I-cells; Fig. 3D,E), consistent with the results using anti-β-catenin antibody staining (Figs 2A,C and 3A,B). The number was again decreased by single treatment of HU or U0126 to nearly the same level (∼16 I-cells) but still above that seen on day 4 (Fig. 3E). Therefore, the estimated proliferation coefficient for I-cells in the future branch-forming region was nearly 2.9=[35−(16−10)]/10, suggesting that these cells divided at least once during the one-day period. In contrast, the number of labeled I-cells in the tentacle bulb increased only slightly (from ∼10 to 13 I-cells) and treatment with HU decreased it to almost the same number as that on day 4 (Fig. 3F). Therefore, the estimated proliferation coefficient for I-cells in the tentacle bulb was ∼1.3=(13−10)/10.

These results indicate that I-cell proliferation is more active in the future branch-forming region than elsewhere. Taken together, these results suggest that MEK signaling regulates locally active I-cell proliferation to bring about I-cell accumulation and branch bud formation.

To find out whether other signaling pathways such as those involving RTK play a role in tentacle branch formation, the effect of an FGF receptor inhibitor (SU5402) on branch formation was tested. It should be noted that MEK is a component of canonical FGF signaling pathways. The inhibitor SU5402 has been used in Hydra (Sudhop et al., 2004) and is reported to inhibit PDGF receptor only weakly at a high concentration above 60 µM (Mohammadi et al., 1997), although its specificity in C. pacificum is not known. It was found that the second branch failed to form on day 6 after inhibitor treatment from day 4 to 6 (arrowheads in Fig. 4A). The concentration of the inhibitor used (6 µM) was determined in the same way as that for the MEK inhibitor (Fujiki et al., 2019); that is, it was the lowest to be effective among those examined and did not have any overall effect on medusa growth (Fig. 4A). Further analyses with the use of SU5402 showed that it reduced the I-cell number and percentage in the branch bud on day 5 but not to the extent of when treated with the MEK inhibitor (Figs 3B,C,E and 4B). In fact, co-treatment of HU with SU5402 further reduced the number and percentage to nearly the same level as those when U0126 or HU was each used alone (Fig. 3B,C). It is noteworthy that the effects of U0126 and SU5402 on I-cell accumulation were significantly different despite the fact that the mean number of branches on day 6 after treatment with these inhibitors was comparable: 1.09 (n=162) for SU5402 versus 1.12 (n=189) for U0126 (Fujiki et al., 2019). In addition, treatment of a lower concentration of U0126 (8 µM) reduced the number and percentage of I-cells in the day 5 branch bud compared with those after SU5402 treatment [20.7 out of 35.5 (58.3%) in Fig. S3A versus 23.6 out of 35.2 (67.0%) in Fig. 3B] but allowed the second branch formation in 53.7% (n=108) of the cases (Fig. S3B), indicating that the reduction level in the accumulated I-cell number caused by SU5402 treatment (Fig. 3B) cannot explain the nearly complete absence of the second branch of the treated tentacles. Therefore, these results suggest that FGF signaling regulates branch formation mainly through regulation other than an effect on I-cell accumulation, although it promotes it to some extent (Fig. 3B,C, DMSO versus SU5402).

Considering that a large population of I-cells was still accumulated in the branch bud after SU5402 treatment but that the branch did not form, FGF signaling may regulate a branching step required after the branch bud is formed, possibly a branch elongation step through regulation of self-renewal and differentiation of the accumulated I-cells. In this regard, it is interesting that the accumulated I-cells were found positive for phosphorylated MAPK, a downstream target of FGF signaling, which was nearly eliminated by the SU5402 treatment (Fig. 4C, top and middle images, and Fig. 4D), suggesting that I-cells forming the branch bud may receive FGF signaling. Phosphorylated MAPK was also observed in scattered cells in the tentacle bulb region as well as a population of cells at the proximal end of the tentacle bulb (arrows in Fig. 4C, bottom images).

To further explore whether FGF signaling acts on the accumulated I-cells in the branch bud, two genes encoding FGF ligands were cloned (Fig. S4). One of the genes showing the highest sequence similarity to Hydra vulgaris FGF-1 (Turwankar and Ghaskadbi, 2019) in the BLAST search, namely Fgf1-like, was found to be expressed in the ectodermal layer of branch buds (white arrowheads in Fig. 4E), suggesting that this gene is expressed in the accumulated I-cells and that its protein products may mediate FGF signaling that regulates branch elongation. The H. vulgaris FGF-1 is shown to have structural similarities with vertebrate FGF1 (Turwankar and Ghaskadbi, 2019); however, a phylogenetic analysis has not been carried out to show which FGF subfamily it might group with. In contrast, the expression of the other FGF gene, which showed the highest similarity to H. vulgaris fibroblast growth factor 17-like in the BLAST search, namely Fgf17-like, was evident in a tentacle region at the proximal end of the tentacle bulb (yellow arrowheads in Fig. 4E) and could be responsible for phosphorylated MAPK observed in the same region (arrows in Fig. 4C, bottom images).

To examine whether the I-cells accumulated in the branch bud differentiate into cells that constitute the branch, these cells were traced by DiI labeling (Fig. 5A). Because the branch bud on day 5 was too small to label, the tip region of the second branch on day 5.5 (consisting mostly of I-cells; Fig. 1E) was chosen as a labeling target. Labeling of a single cell was evaluated by the size of DiI fluorescence on the tip of the branch immediately after the labeling (Figs 1D and 5B, left column) and was confirmed by co-staining with Phalloidin (Fig. 5B, middle column). After confirming that the very tip region was labeled in the right size (as shown in Fig. 5B, left column), the labeled cells were traced up to days 7.5 and 15 to see whether or not they produced differentiated cells (Fig. 5C,D). On day 7.5, when the branch had sufficiently elongated, in all instances labeling was found both in ectodermal epithelial cells (Fig. 5C, yellow arrows) and I-cells (Fig. 5C, green arrows) (see Fig. S5A for data). These cells were identified by their cell shape (columnar for epithelial cells and spherical for I-cells) and position within the branch (away from the tip for epithelial cells and interconnected spaces between ectodermal layer cells for I-cells).

In all cases, labeling was also found on day 15 both in nematocytes (Fig. 5D, purple arrows) located in the periphery of the tip region of the branch and in I-cells (Fig. 5D, green arrows) deep within the tip (see Fig. S5B for data). Nematocytes were identified by the presence of a cnidocil within the cell, which stained with Phalloidin (Fig. 5D, white arrows). These results suggest that I-cells in the branch bud produce cell descendants differentiating into epithelial cells and nematocytes as well as those I-cells remaining during branch elongation. As labeled epithelial cells and nematocytes were both found in all the lineage-tracing trials on day 7.5 and 15, respectively (Fig. S5), a single I-cell could produce these various cell types at different stages of branch elongation.

In the present study using a diploblastic animal species, a new mechanism for branching morphogenesis in animals was discovered that does not rely on interactions between the epithelial and mesoderm-derived cells but on pluripotent stem cells, namely the I-cells. Although they reside in the ectodermal epithelial layer, I-cells are not epithelial cells as they are undifferentiated and migrate between the interconnected spaces among cells of the ectodermal layer (Gold and Jacobs, 2013; Leclère et al., 2016; Plickert et al., 2012). It appears that, in the C. pacificum medusa tentacle, branch buds are formed somewhere in the ectodermal epithelial layer, which is composed mainly of epithelial cells and nematocytes (day 4 in Fig. 2B,C), by increasing the I-cell population locally through cell proliferation (day 5 in Figs 2C and 3B). This initial branching step contrasts with the vertebrate and Drosophila models, in which initial changes to form branch buds occur to the epithelial cells themselves, such as leader cell specification and epithelial folding (Varner and Nelson, 2014; Wang et al., 2017), and are regulated by RTK signaling from the adjacent mesodermal and/or epidermal cells (Goodwin and Nelson, 2020; Varner and Nelson, 2014; Wang et al., 2017). Further research is necessary to identify which cells might signal promotion of the initial branching step in the C. pacificum medusa tentacle and which correspond to the ligand-producing mesoderm and undergo regulation by extracellular secreted signals.

Pluripotent stem cell accumulation at a future branching site is observed also in angiosperms (seed-producing plants), where a branch bud composed of axillary meristems with pluripotent stem cells arises in the leaf axils and subsequently outgrows to form a new leaf stem and blade (Wang and Jiao, 2018). Although branch structures in animals and plants were clearly independently acquired during evolution, the use of stem cells as major players in branching morphogenesis in both lineages suggests that this mechanism must be efficient and advantageous; in particular, in organisms where pluripotent stem cells are active with a number of differentiation potentials even in adult stages. Animals such as vertebrates and Drosophila, in contrast, do possess stem cells in the adult but they are lineage-restricted (Scheele et al., 2017; Wuidart et al., 2018).

In the present study of tentacle branch formation using C. pacificum, it was shown that I-cells accumulate in the branch bud but differentiate into at least two different cell types: epithelial cells and nematocytes (Fig. 5). A grown tentacle branch such as the second branch on day 15 has nematocytes not only at the tip of the branch (Fig. 5) but also along the length of the branch stalk, which acquires scattered clusters of nematocytes (Fujiki et al., 2019). However, these nematocytes on the branch stalk were never recognized as labeled cells on day 15 in our DiI labeling experiment, suggesting that the I-cell population in the branch bud may not give rise to the stalk nematocytes. Interestingly, it is known that there is another I-cell population in the tentacle bulb (Fig. 1C; Denker et al., 2008; Seipel et al., 2004), from which it is suggested that nematocytes are differentiated (Condamine et al., 2019; Denker et al., 2008). It is proposed in Clytia that these differentiated nematocytes in the tentacle bulb migrate away to constitute the tentacle nematocytes (Condamine et al., 2019; Denker et al., 2008). Therefore, the nematocytes of the branch stalk in C. pacificum might originate from the tentacle bulb. If so, the C. pacificum tentacle may harbor differently pre-committed subpopulations of I-cells, as suggested for the I-cells of Hydra (David, 2012).

Considering that I-cells play pivotal roles in regeneration and maintenance of organismal function in hydrozoan species (Bosch, 2009; Gahan et al., 2016; Gold and Jacobs, 2013; Leclère et al., 2016; Plickert et al., 2012; Vogg et al., 2019), it is, perhaps, no surprise that these cells are involved in the formation of a new structure; in this instance, branch formation in the medusa tentacle. In contrast, however, it seems more surprising that a number of conserved mechanisms during branching morphogenesis have been determined by this study comparing diploblasts and triploblasts such as mammals and Drosophila. Firstly, it was found that two well-known steps in triploblasts, namely branch bud formation and branch elongation, were also recognizable and regulated by different signaling pathways in the C. pacificum tentacle (Fig. 6). Secondly, RTK signaling is involved in the establishment of branched organs across a wide range of animal species. In the branching of the medusa tentacle, following I-cell accumulation in the branch bud, FGF signaling seems to regulate an event causing bud outgrowth. Essential roles of RTK signaling for branch elongation have been reported also in blood vessels (VEGF) and salivary glands (FGF10) (Gerhardt et al., 2003; Steinberg et al., 2005). Finally, branch elongation depends on an outpushing mechanism of the branch tip as a result of proliferative activities of cells that reside at the tip of the growing branch. In C. pacificum, I-cells remain at the tip of the branch while it elongates (Figs 1C and 2A) and produce descendant cells that differentiate into the branch-forming cells (Fig. 5).

The occurrence of I-cell accumulation is a key step in medusa tentacle branch formation. Not only does it serve as the basis of the forming branch, but also when and where it occurs during tentacle growth could determine the elaborate branching pattern. Branched medusa tentacles are a rare character but are commonly found in the jellyfish of the family Cladonematidae (Schuchert, 2006), and so are considered to be an apomorphic trait within the Medusozoa. It is thus possible that the acquisition of a mechanism by which I-cells are accumulated somewhere in the tentacle could have led to the appearance of tentacle branches as a new trait. One such candidate mechanism could be one regulating budding formation in the hydroid stage, considering that I-cell accumulation was found by way of Nanos1 expression in the medusa bud-forming region of the hydroid stage (Fig. S2B) and that hydroids bud in many cnidarian species belonging to the Class Hydrozoa (which includes Family Cladonematidae). Therefore, it would be interesting in the future to investigate whether or not such I-cell regulation by MEK signaling is operative when budding happens in the hydroids of other hydrozoan species (Sudhop et al., 2004; Lange et al., 2014; Holz et al., 2017; Suryawanshi et al., 2020); as well as when medusa tentacles branch in other cladonematid species such as Staurocladia (Hirano et al., 2006), tentacles of which, unlike those of C. pacificum, branch only once.

The culture and maintenance of C. pacificum Naumov, 1955 (Cnidaria: Hydrozoa) has been described in previous research (Hirai and Kakinuma, 1957). A male strain named UN2 was used in the present study; a strain originally collected near the island of Uratononoshima in Matsushima Bay, Miyagi Prefecture, Honshu, Japan. It originated from a single hydroid by colony expansion and transplantation. Containers with a mass of hydroids that had grown asexually were maintained in filtered seawater (FSW) and stored at 4°C to simulate hibernation conditions. For experiments, they were moved to an incubator at 21°C, when degenerated hydroids revive and begin to produce medusa buds. This ‘medusa bloom’ lasts for 1∼3 weeks. Hydroids and medusae were fed daily until fixed except for day 3, when medusae are released from the polyps (Fig. 1B), with Artemia salina nauplius larvae (A & A Marine Brine Shrimp Eggs, Vietnam) unless otherwise noted, replacing the FSW 3 h after feeding.

Hydroids and medusae were anesthetized in ice-cold 0.2 M MgCl₂ in 50% FSW for 10 min and fixed in 4% paraformaldehyde (PFA) in FSW at 4°C overnight. After fixation, the umbrella was removed from the medusae for ease of further handling.

Presumptive Nanos1, Nanos2, Piwi, Fgf1-like and Fgf17-like sequences were acquired from the annotation of an RNA-seq result obtained from mRNAs extracted from the manubrium of a female strain called ‘6 well’ (Takeda et al., 2018). Using these sequences, PCR primers to clone the genes from UN2 were designed as indicated below. A cDNA library was generated with extracted mRNAs from medusa tentacles isolated from UN2 on days 4.5 and 5 using a SMARTer™ RACE cDNA Amplification Kit (Clontech), and was used as a template for PCR (Nanos1, Nanos2 and Piwi) and RACE (Fgf1-like and Fgf17-like). Amplified PCR products were sub-cloned into the pGEM-T Easy vector (Promega). The resulting plasmids were then used for probe synthesis with either digoxigenin (DIG) or fluorescein labeled for in situ hybridization. Plasmid with the 5′RACE product was used for Fgf17-like, whereas those with the 3′RACE products were used for Fgf1-like. Probes were synthesized in vitro by either T7 or SP6 RNA polymerase (Roche) according to the insert direction.

Primer sequences used for PCR cloning were as follows. For Nanos1: 5′-GCTACAAGGTCAACGTTTTAGAGC-3′ (forward), 5′-TCGTGTGACTTTGGTCACCACTC-3′ (reverse), 5′-AAGAGACACAGTCATTATCAAGCGA-3′ (nested forward) and 5′-AGCACGTAAAATTGGACACGTCG-3′ (nested reverse); for Nanos2: 5′-ACTTCTCCAAAACCTCATGCCGAG-3′ (forward), 5′-GAATGGCGGGCGATTTGACATCC-3′ (reverse), 5′-TGACGAGGAAGCAGATGCATGGT-3′ (nested forward) and 5′-GGCAACGACCCATTTGTGACACG-3′ (nested reverse); for Piwi: 5′-CCAAGCTGCACCACCATCAGAACAGA-3′ (forward), 5′-GTCGAACCACTCTGGTCGTGTCACTG-3′ (reverse), 5′-AAAAGAGCAGCGGCCAGAAAGAAGGC-3′ (nested forward) and 5′-GCGGGTCGCATACTTGTTGGTACTGGC-3′ (nested reverse); for Fgf1-like: 5′-CCCCGACCAAAACGCGTAACGCACA-3′ (forward for 3′RACE), 5′-GGTGGAACTCGAAATAAAGGCGACAG-3′ (nested forward for 3′RACE), 5′-TTTCGAGCACGCAACGTGGTCTCGAT-3′ (reverse for 5′RACE) and 5′-AGCTGAACGACTCCATGAGGAACACA-3′ (nested reverse for 5′RACE); for Fgf17-like: 5′-CGCCAAGACGCGTTTTCTCGGGGAT-3′ (reverse for 5′RACE), 5′-GCAAGAAACCAACCGTTTTCACCGTC-3′ (nested reverse for 5′RACE), 5′-GCAACACCGACAAATGTTCGTCGTGT-3′ (forward for 3′RACE) and 5′-GAGGAGTCGTGAGATTACGAGGTGAA-3′ (nested forward for 3′RACE).

Fixed samples were washed with phosphate buffered saline (PBS) supplemented with 0.1% Tween20 (PBSTw), dehydrated and rehydrated with PBSTw by graded methanol treatment, followed by pre-hybridization in hybridization buffer (5×SSC, 50% formamide, 0.1% Tween20, 50 μg/ml tRNA, 50 μg/ml heparin, 0.1% DMSO) at 55°C overnight. Antisense probes at a concentration of 1 μg/ml in the hybridization buffer were hybridized at 55°C overnight and detected using alkaline phosphatase (AP)-conjugated anti-DIG/fluorescein antibodies (1:500, goat, Roche, 11093274910) by 2-4 h incubation in 0.5% blocking reagent (PerkinElmer) in PBSTw at room temperature. Colorimetric reactions were performed by Nitro Blue Tetrazolium (120 μg/ml)/5-Bromo-4-Chloro-3-Indolyl-Phosphate (120 μg/ml) in alkaline buffer [100 mM NaCl, 100 mM Tris-HCl (pH 9.5), 50 mM MgCl₂, 0.1% Tween20] until the desired signals were observed.

Molecular phylogenetic analyses were carried out with full-length amino acid sequences of Nanos and Piwi-related genes across the animal kingdom. Protein sequence alignments were performed with the software Molecular Evolutionary Genetics Analysis (MEGA) X, using the MUSCLE algorithm. Amino acid residues where gaps occurred in the alignments were excluded manually. Neighbor-joining trees were constructed with 1000 bootstrap replications.

For anti-β-catenin antibody staining, specimens were treated with 3% hydrogen peroxide (H₂O₂) for 30 min and washed three times in PBS with 0.5% Triton X-100 (PBSTr) before incubation in bovine serum albumin (BSA) blocking buffer (2% BSA in PBSTr) at room temperature for 1 h. The anti-β-catenin monoclonal antibody 7A7 (Enzo, ALX-804-259-C100) was used at 1:100 with overnight incubation at 4°C in blocking buffer. Specimens were again washed in PBSTr three times, pretreated in blocking-buffer under the same conditions as described above and incubated at room temperature for 1 h with Mouse MAX PO (Nichirei Corporation, 424131) conjugated with horseradish peroxidase (HRP) diluted 1:1 with blocking buffer as a secondary antibody. Finally, HRP activity was detected using the Tyramide Signal Amplification plus fluorescein amplification kit (PerkinElmer Life Sciences) according to the manufacturer's instructions.

For anti-phosphorylated MAPK antibody staining, specimens with their umbrella removed were washed in PBSTr for 10 min and treated with 2 N HCl for 30 min. After overnight incubation in blocking buffer at 4°C, specimens were incubated with anti-MAPK, activated diphosphorylated extracellular signal-regulated kinase (ERK)-1 and 2 antibody (1:25, Sigma-Aldrich, M9692) at 4°C overnight, followed by rocking slowly on a shaker (TAITEC) at room temperature overnight. After three washes in PBSTr, Alexa fluor 488-conjugated anti-mouse antibody (Invitrogen, A11001) was used at 1:50 in blocking buffer as a secondary antibody.

Day-3 medusae were collected and fed with adequate Artemia nauplii for 1 h. On day 3.5 the medusae were incubated with 5 mM BrdU (Invitrogen) in FSW for 1 h and washed well in FSW. The BrdU-labeled medusae were then cultured up to the desired stage (day 4 or day 5) without feeding, relaxed and fixed as described above. The umbrella was then removed and discarded and the specimens were washed in PBSTr for 10 min and treated with 2 N HCl. After 1 h incubation in blocking buffer, incorporated BrdU was detected by incubating at room temperature overnight with anti-BrdU antibody (Invitrogen, B35128) at 1:100 in blocking buffer. Following three washes in PBSTr, Alexa fluor 488-conjugated anti-mouse antibody (Invitrogen, A11001) was used at 1:100 in blocking buffer as a secondary antibody.

The specimens stained with the antibodies described above (anti-β-catenin, anti-MAPK, activated and anti-BrdU) were treated with 100 μg/ml of RNase A in PBSTr at 37°C for 30 min and incubated with 6 μM PI (Invitrogen) in PBSTr at room temperature for 1 h. The specimens were washed three times in PBSTr for 10 min each and mounted in Vectashield (Vector Laboratories) for observation with a confocal microscope (LSM5 PASCAL; Zeiss).

Images of the adaxial side of the tentacles were captured after antibody and PI staining. The ectodermal epithelial layer within a circle at the place of interest (diameter either 30 or 50 μm, defining the scanning cylinder column) was scanned by confocal microscopy with a z-stack interval set as thin as the slice thickness to enable all slices to be connected seamlessly.

For anti-β-catenin and anit-BrdU antibody staining, cells inside the scanning cylinder column were identified and counted manually. Graphs of cell numbers were created using Rstudio or GraphPad Prism 8.

For anti-MAPK activated-antibody staining, signal intensities for antibody and PI staining were quantified using Fiji ImageJ. In each slice, the integrated density of the phosphorylated MAPK (pMAPK) channel was divided by the mean gray value of the PI channel. The standardized pMAPK intensity of I-cells (identified by the presence of prominent nucleoli) from each slice was then summed to construct the intensity in the scanning cylinder column, and the summed intensity was divided by the total I-cell number in the same column to obtain the standardized pMAPK intensity per I-cell.

Medusae were reared in FSW containing 6 μM FGFR inhibitor SU5402 (Calbiochem) either from day 4 to day 6 for the evaluation of the second branch formation, or from day 4 to day 5 for anti-β-catenin and anti-pMAPK activated-antibody staining, and BrdU labeling. Medusae were also treated with 10 μM MEK inhibitor U0126 (Calbiochem) or 10 mM HU (Sigma-Aldrich) from day 4 to day 5 for the antibody staining and BrdU labeling. Before inhibitor treatment, medusae were fed with Artemia salina nauplii for 1 h and allowed to rest in clean FSW for another hour. Medusae were not fed during inhibitor treatment.

The second branch formation was evaluated on day 6, while medusae were fixed on day 5 for antibody staining and BrdU labeling. To confirm that the inhibitors were effective, some individuals from the same group as those subjected to inhibitor treatment were maintained in culture to day 6 to confirm that a second branch did not form (Fig. 4A; Fujiki et al., 2019).

CellTracker CM-DiI (Molecular Probes) was prepared as described previously (Fujiki et al., 2019). Day 5.5 medusae were relaxed in anesthetic and their movement was further restrained with a thin strip of cover-glass (Fig. 5A). Horizontal injection into the tentacles was performed successfully under a fluorescence microscope (Olympus BH-2), despite the flexibility of the relaxed medusae. The tip region of the newly formed bud was pricked with a sharp needle filled with DiI solution and one drop of DiI was released. Labeling was confirmed through the fluorescence microscope (Olympus BX53) on day 5.5 and day 6 and successfully DiI-labeled medusae were cultured to later stages while being fed every other day after day 6.

On days 5.5 and 7.5 the labeled medusae were fixed in 4% PFA in PBS at room temperature for 1 h for cell membrane staining with Alexa Fluor 488 Phalloidin (Invitrogen); and on day 15 labeled medusae were fixed in the same fixative solution for Phalloidin staining and nuclear staining with SYTOX Green (Invitrogen). After washing in PBS three times for 10 min each, the fixed specimens on days 5.5, 7.5 and 15 were incubated at room temperature for 1 h with a Phalloidin staining solution, which had been prepared at 5 units/ml by dissolving air-dried Phalloidin in PBS+2% BSA. Day-15 specimens were further subjected to nuclear staining, in which they were treated with 100 μg/ml RNase A at 37°C for 30 min, then washed and incubated at room temperature for 1 h with 2 μM SYTOX green solution in PBS. The stained samples were mounted in Vectashield for observation by confocal microscopy.

Normality and comparison tests were performed with GraphPad Prism 8. Normality was assessed using Kolmogorov–Smirnov normality test. Distributed data were analyzed using unpaired two-tailed Student's t-test when two unmatched groups were compared. Alternatively, one-way ordinary ANOVA was used and followed by Tukey's multiple comparisons tests when three or more unmatched groups were compared inside one experiment. P-values were indicated as follows: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. Error bars indicate mean±s.d.

We thank Drs Noriyo Takeda and Ryusaku Deguchi for their help in setting up our Cladonema project and Dr Noriyo Takeda for our use of her RNA-seq data.