Transgenic analysis of a SoxB gene reveals neural progenitor cells in the cnidarian Nematostella vectensis

The complex central nervous systems of bilaterians develop from a relatively small pool of multipotent neural progenitor cells (NPCs). These cells undergo a choreographed program of lineage expansion and diversification to generate an impressive number of specialized cell types in a strict spatio-temporal pattern. A characteristic feature of NPC development is the pattern of cell divisions within a lineage. Symmetric cell divisions can expand a progenitor pool or produce two terminally differentiated neurons, whereas asymmetric divisions produce unalike daughters with varying capacities for self-renewal or differentiation [reviewed by Huttner and Kosodo (2005)]. Well-studied examples, such as Drosophila neuroblasts and mammalian radial glial cells, demonstrate stereotypic lineage progressions, during which they take on distinct temporal identities and generate diverse daughter cells [reviewed by Kohwi and Doe (2013)]. The regulated differentiation of NPCs is thus considered part of the shared cellular framework of bilaterian neurogenesis. However, the evolutionary origin of NPCs is obscure, as a similar NPC-type population has not been identified in the bilaterian sister group, the Cnidaria (Fig. 1A).

Cnidarians (sea anemones, corals, jellyfish and hydroids) possess relatively simple nervous systems, consisting of sensory cells, ganglion cells (the morphological equivalent to interneurons) and nematocytes (stinging cells), which are usually considered to be a third, highly specialized, neural cell type [Fig. 1B; reviewed by Galliot et al. (2009)]. Although their neurons are not uniformly distributed, cnidarian polyps lack brain-like nervous system centralization and their nervous system is often described as a nerve net (Galliot et al., 2009; Watanabe et al., 2009). The cellular origin of neurons has mainly been described in hydrozoans (e.g. Hydra), in which a pluripotent, endodermally derived interstitial stem cell population gives rise to most neurons and nematocytes, as well as to gland cells and gametes (Bosch and David, 1987; Bosch, 2009; David, 2012; Müller et al., 2004). This origin of neurons differs significantly from bilaterian systems, but interstitial stem cells might represent a derived condition, as they have not been described outside hydrozoans.

The sea anemone Nematostella vectensis is an anthozoan cnidarian, and we have previously shown that its neurons derive from both germ layers, i.e. ectoderm and endoderm (Nakanishi et al., 2012). At the molecular level, Nematostella possesses many genes related to conserved bilaterian neural genes (Putnam et al., 2007; Watanabe et al., 2009), and subsets of its nervous system have been examined using antibodies against conserved neuropeptide precursor molecules [e.g. Marlow et al. (2009)]. The generation of neurons in Nematostella has previously been examined through studies of conserved bilaterian transcription factors. However, whereas many of these genes display evocative ‘salt-and-pepper’ expression patterns, and, in the case of the proneural orthologue NvAshA, can regulate the expression of putative neural markers (Layden et al., 2012), thus far none of these expression patterns has been definitively linked to the production of a mature neural cell type [e.g. Marlow et al. (2009); Matus et al. (2007a)]. Here, we address the developmental origins of neural cell types in anthozoans by employing a transgenic approach based on the putative neural regulator NvSoxB(2) (Magie et al., 2005).

NvSoxB(2) [also known as SoxB2, see Magie et al. (2005); NveSoxBa, see Royo et al. (2011)] is one of five Nematostella genes closely related to the bilaterian HMG-box-containing SoxB transcription factor family. The five Nematostella SoxB genes are expressed in overlapping patterns during embryogenesis, with NvSoxB(2) being expressed in a scattered pattern of individual cells in the ectoderm and endoderm (Magie et al., 2005) that is coincident with the timing of neurogenesis in these layers. In bilaterians, SoxB genes are expressed in the neural primordia of diverse organisms [e.g. polychaetes (Kerner et al., 2009), Drosophila (Buescher et al., 2002) and vertebrates (Bylund et al., 2003)]; across these organisms, SoxB genes are commonly expressed transiently in NPCs and are downregulated prior to terminal differentiation. Based on sequence similarity, bilaterian SoxB genes have been classed into two groups, SoxB1 and SoxB2, although phylogenies commonly only resolve a monophyletic SoxB1 group, which resides within an unresolved group of the other SoxB genes [designated SoxB2, see Bowles et al. (2000); Jager et al. (2011); Wilson and Dearden (2008)]. The precise phylogenetic affinities of cnidarian SoxB genes are currently uncertain [e.g. Jager et al. (2011)].

Thus far, functional studies of SoxB genes have been limited to investigations in classic model organisms. In vertebrates, SoxB1 genes have an important role in the maintenance of NPCs (Bylund et al., 2003; Graham et al., 2003; Kishi et al., 2000). The best-described example is the chick neural tube, in which SoxB1 genes (Sox1-3) suppress neurogenesis by maintaining neuroepithelial cells in an undifferentiated state, and SoxB2 genes (Sox21, Sox14) promote neuronal differentiation by counteracting SoxB1 activity (Bylund et al., 2003; Graham et al., 2003; Holmberg et al., 2008; Sandberg et al., 2005). In Drosophila, functional studies have focused on two SoxB genes, soxNeuro (a SoxB1 gene) and dichaete (probably a SoxB2 gene), that are expressed in the neuroectoderm and are involved in the formation of neuroblasts (Buescher et al., 2002; Overton et al., 2007; Zhao and Skeath, 2002). However, their modes of activity do not bear direct comparison to vertebrate SoxB1 and SoxB2 genes, as both genes can positively regulate neural development by promoting neuroblast formation.

In this study, we present an NvSoxB(2)::mOrange transgenic reporter line, which we use to investigate the origin of neural cells and the function of NvSoxB(2) during early neurogenesis in Nematostella. We find that NvSoxB(2)-expressing cells give rise to three major neural cell types: sensory cells, ganglion cells and nematocytes. Using 5-Ethynyl-2′-deoxyuridine (EdU) labelling, we show that NvSoxB(2)⁺ cells are mitotic, and that there are asynchronies in the division of sibling cells in NvSoxB(2)⁺ lineages, which might underpin the generation of neural cell type diversity. With knockdown experiments we demonstrate that NvSoxB(2) regulates development of the nervous system but not ectodermal patterning. Together, these observations reveal the existence of dedicated NPCs in an anthozoan cnidarian and identify SoxB activity as an ancient regulator of neural progenitor populations.

NvSoxB(2) was initially classified as a SoxB2 gene, but has subsequently been shown to be a SoxB gene that cannot be grouped into the B1 or B2 subfamilies [e.g. Jager et al. (2011); Shinzato et al. (2008)]. NvSoxB(2) is expressed in scattered cells from blastula stage on, first in the ectoderm and later also in the endoderm (Magie et al., 2005). To test whether NvSoxB(2)-expressing cells contribute to the nervous system, we generated a stable transgenic line [NvSoxB(2)::mOrange], in which a promoter region of NvSoxB(2) drives the expression of the fluorescent reporter mOrange (Renfer et al., 2010; Shaner et al., 2004). This technique allows the identification of NvSoxB(2)-expressing cells and their progeny via the expression and inheritance of the mOrange reporter protein. Co-labelling of transgenic embryos with probes for both mOrange and NvSoxB(2) mRNA demonstrated strong (ca. 75-85%) correspondence between the expression of the reporter gene and that of endogenous NvSoxB(2) (supplementary material Fig. S1). Although we cannot unequivocally exclude that a specific sub-population of NvSoxB(2)-expressing cells is not represented in the transgenic line, we did not observe anything that would support this suggestion – e.g. no systematic lack of colocalization of NvSoxB(2) and mOrange during a particular time point or in a particular region of the embryo. Subsequent analysis of reporter protein localization enabled us to identify that cells expressing NvSoxB(2) early in their development go on to become sensory, ganglion and nematocyte cells within the Nematostella ectodermal and endodermal nervous systems (Fig. 2; supplementary material Fig. S2).

In gastrula stages, mOrange is localized in scattered ectodermal cells – some are dividing at an apical position within the epithelium (Fig. 2A-F; Meyer et al., 2011). By early planula stage, mOrange localization highlights the development of the basi-epithelial ectodermal nerve net, located predominantly in the aboral two-thirds of the larva, but absent from directly under the aboral pole (where the larval apical organ is located; Fig. 2G-I). The nerve net is composed of neurites extending from the basal side of ectodermal neural cells; the primary orientation of these neurites is lateral, encircling the oral-aboral axis, but this is not strictly uniform (Fig. 2I). In addition to sensory neurons with medially located nuclei and apical cilia (Fig. 2J), NvSoxB(2)::mOrange also labels ganglion neurons, identified by their basal location in the ectoderm (Fig. 2K). In the early planula, activation of the NvSoxB(2) promoter becomes detectable in the endoderm (Fig. 2L). By mid-planula, mOrange is found in a large number of scattered cells throughout the endoderm, in addition to the ectodermal nerve net (Fig. 2M-O).

In later developmental stages, mOrange is localized in the ectodermal nerve net and in the ciliated neurons of the endodermal nervous system (supplementary material Fig. S2A-H). In primary polyps, mOrange localization highlights a close association between endodermal neurons and the longitudinal musculature, with bundles of neurites and cell bodies being aligned along the muscle fibres (supplementary material Fig. S2E-G).

In comparison to the previously characterized line NvElav1::mOrange, in which a promoter sequence of a marker of differentiated neurons drives mOrange expression in a subset of sensory and ganglion cells (Nakanishi et al., 2012), we found a greater number and diversity of cells labelled with mOrange in the NvSoxB(2) line (supplementary material Fig. S2I-L). In double-transgenic animals (NvSoxB(2)::mOrange × NvElav1::cerulean), the NvElav1⁺ cell population is also labelled by the NvSoxB(2)::mOrange transgene (Fig. 2P-S). We speculated that NvSoxB(2)⁺ cells that were NvElav1⁻ might represent nematocytes, the third putative neural cell type of cnidarians. Thus, we performed an immunostaining on the NvSoxB(2)::mOrange line using an antibody against Minicollagen3 (NvNcol3), the major component of the capsule (nematocyst) of nematocytes (Zenkert et al., 2011). This experiment showed that a subset of the NvSoxB(2)::mOrange cells contained nematocysts, thus unambiguously identifying these cells as nematocytes (Fig. 2T,U). These data further support the classification of nematocytes as a bona fide neural cell type of cnidarians.

To better visualize the morphology and neurite projections of individual neurons we used a mosaic analysis based on the injection of an NvSoxB(2)::EGFP plasmid into eggs of the stable NvSoxB(2)::mOrange line (mosaicism of EGFP in the F0 population is due to spatio-temporal differences in construct incorporation/expression within and between embryos). These experiments confirmed the heterogeneity of the NvSoxB(2)⁺ population, with regard to both cell shape and neurite projections arising from individual cells (supplementary material Fig. S2M-V). We found that NvSoxB(2)⁺ cells are multipolar and that they display variation in the number, orientation, length and branching patterns of their basal projections. Variation is also apparent in the number and shape of varicosities found at the base of cells and along neurites. These observations hint at a considerable level of complexity in the neural architecture of this simple nerve net-based nervous system.

Although the NvSoxB(2) transgenic line, through the inheritance of the stable reporter protein mOrange, reveals the contribution of NvSoxB(2)⁺ cells to the three principal neural cell populations in Nematostella, it does not show when NvSoxB(2) is transcribed during the development of a neural cell. We thus extended the previous analysis of NvSoxB(2) transcription across a more detailed time course and tested for co-localization of NvSoxB(2) with neural differentiation markers. We found that the earliest expression of NvSoxB(2) is in the hollow coeloblastula, where it is expressed in discrete groups of cells distributed throughout the epithelium (Fig. 3A,J). The neural marker anthoRFamide [NvRFamide: sensory and ganglion neurons, Marlow et al. (2009)] is expressed in a few scattered single cells at this stage (Fig. 3D). By late blastula, the expression of NvSoxB(2), NvRFamide and the nematocyte marker minicollagen 3 [NvNcol3, Zenkert et al. (2011)], is in scattered single cells throughout the ectoderm (Fig. 3B,E,H), but fluorescent double in situ hybridization shows that there is no co-localization between NvSoxB2 and either of the neural markers (Fig. 3L,N). In the gastrula, all three genes maintain scattered ectodermal expression patterns, with NvSoxB(2) being expressed in a greater number of cells (Fig. 3C,F,I). Again, we found no evidence for co-localization of NvSoxB(2) with NvNcol3; however, we observed some rare instances of co-localization with NvRFamide (Fig. 3M,O). By counter-staining for acetylated tubulin, we identified that some of the NvSoxB(2)-expressing cells possess a mitotic spindle and are thus undergoing cell division (also evidenced by their rounded shape and apical localization within the ectodermal epithelium, Fig. 3K). At later stages, NvSoxB(2) expression continues to be seen in scattered ectodermal cells, some of which are dividing, and in the invaginating pharynx (supplementary material Fig. S3). Expression in endodermal cells is first detected in the early planula stage; a subset of these cells is also undergoing division (supplementary material Fig. S3).

Together with the transgenic analysis, these gene expression data suggest that NvSoxB(2) is transiently expressed in mitotically active cells that give rise predominantly or exclusively to neural cell types. This is in line with data from Bilateria, in which the timing of expression of SoxB genes is commonly restricted to specific phases of neurogenesis [e.g. Bylund et al. (2003); Kerner et al. (2009)].

To further investigate proliferation within the NvSoxB(2)-expressing cell population, we incubated wild-type embryos for 30 min in the presence of the thymidine analogue EdU and then immediately fixed them for in situ hybridization with a probe for NvSoxB(2). Note that, due to the relatively short EdU incubation period and the lack of a chase period prior to fixation, these data present only a snapshot of the total proliferation that will be occurring within the NvSoxB(2)⁺ population. We have used this protocol to ensure that the cells labelled by EdU are those which are in a pre-mitotic state only. For analysis, we selected a 100 μm×100 μm area of ectoderm on the mid-lateral side of each of six embryos/stage and scored the number of NvSoxB(2)-expressing cells with and without EdU incorporation; we also estimated the total number of cells within the area (Fig. 4A-D). These experiments revealed that the number of NvSoxB(2)⁺ cells in the sampled area remains relatively constant over the course of development, from early blastula to mid-planula. However, due to the increase in the total number of cells in the ectoderm during this time, the NvSoxB(2)⁺ population makes up a larger proportion of the sampled ectoderm in earlier stages (e.g. compare 10% in early blastula with 2.5% in mid-planula). Similarly, whereas there is always a subset of NvSoxB(2)⁺ cells incorporating EdU, this proportion is highest in early development and diminishes over time (e.g. compare 30% in early blastula with 10% in mid-planula). Overall, the level of EdU incorporation in the NvSoxB(2)⁺ population is lower than that calculated for the rest of the ectoderm (supplementary material Fig. S4E). However, we cannot consider slow cell-cycling to be a definitive characteristic of the NvSoxB(2)⁺ population, as our measure of cell proliferation in the rest of the ectoderm encompasses a mixture of cell types.

To examine cell division within the progeny of NvSoxB(2)⁺ cells, we treated NvSoxB(2)::mOrange transgenic animals with a 30 min EdU pulse and fixed them immediately afterwards (Fig. 5A). As in the previous experiment, the mid-lateral ectoderm of embryos was examined. We scored for the presence of EdU incorporation in mOrange⁺ cells and counted the number of contiguous mOrange⁺ cells clustered around each EdU⁺ mOrange⁺ cell. As we did not observe expression of NvSoxB(2) mRNA in directly adjacent cells at the stages of analysis, we assume that contiguous transgenic cells predominantly represent sibling daughter cells produced via cell division of a common mother progenitor (supplementary material Fig. S5A). Notably, we recorded a range of differently sized mOrange⁺ cell clusters (2-7 cells), indicating that mitoses of NvSoxB(2)⁺ cells and their progeny can be asynchronous (synchronous cell division would result in even-numbered cell clusters only). As the majority of mOrange⁺ cell clusters did not contain EdU⁺ cells (a likely consequence of the short EdU incubation time, see note above), in the subsequent section we refer exclusively to clusters which contained at least one EdU⁺ cell.

Within the mOrange⁺ EdU⁺ clusters, we commonly observed that only one cell was EdU positive, a further indication of asymmetries in the behaviours of sibling cells in the NvSoxB(2)::mOrange line. We next tested whether the cells in an EdU^– cluster after the short incubation of 30 min would remain EdU⁻ after a longer pulse. As the time from S phase to mitosis at gastrula stage is ca. 2-4 h (supplementary material Fig. S5B), we labelled gastrula stages with EdU for either 30 min or 2 h; a 2 h limit was chosen to ensure that EdU labelling captured pre-mitotic cells only. We further restricted our analysis to clusters containing 2-3 cells, as larger clusters might represent daughters from multiple founder cells, and we wanted to analyse sibling clones only. Our results show that, even when incubated with EdU for longer periods, in the majority of NvSoxB(2)::mOrange⁺ two-cell clusters (97%) across both stages only one cell had incorporated EdU (Fig. 5B). Similarly, the number of cells labelled in the three-cell clusters did not change markedly between the two pulses within a stage; however, in early gastrulae a majority (90%) of three-cell clusters showed one EdU⁺ cell, in contrast to two positive cells (73%) in gastrulae. Taken together, these findings identify differential cell cycling in the daughters of individual NPCs. Specifically, at a given time point, one daughter from an NvSoxB(2)⁺ progenitor can be in S phase whereas the other is not. The cell not in S phase may be quiescent, post-mitotic or residing in G1/G2, with all of these cases representing asymmetry in the developmental progression of daughter cells from NvSoxB(2)⁺ progenitors.

To identify the function of NvSoxB(2), we used a translation-inhibiting morpholino [NvSoxB(2)MO1] that targeted the 5′UTR of NvSoxB(2). NvSoxB(2)MO1 caused a reduction in the expression of the neural markers NvElav1, NvRFamide and NvNcol3 in the ectoderm of early planulae (Fig. 6A-F), but had no effect on the expression of the regional ectodermal markers NvFkh, NvFGFa1 and NvWnt2 (Fritzenwanker et al., 2004; Kusserow et al., 2005; Martindale et al., 2004; Matus et al., 2007b; Rentzsch et al., 2008; see also supplementary material Fig. S6H-M). Quantitative RT-PCR (RT-qPCR) confirmed that the expression of neural marker genes was decreased in NvSoxB(2) morphants, relative to control MO-injected embryos, at both the gastrula and planula stages (Fig. 6G). To determine the effect of this change in gene expression on the morphology of the larval nervous system, we assessed the presence of neurons and nematocytes in morphants by immunostaining against NvRFamide and NvNcol3, and by injecting the NvSoxB(2)MO1 into the NvElav1::mOrange transgenic line. In all cases, neural cells were strongly reduced or absent in the NvSoxB(2) morphants (Fig. 6H-P; for additional control experiments see supplementary material Fig. S6A-G). We conclude that NvSoxB(2) plays an important role in the development of three major neural cell types in Nematostella.

Together with observations in the NvSoxB(2)::mOrange transgenic line, our gene expression analyses suggest the existence of a population of dedicated NPCs in Nematostella. To our knowledge, this is the first description outside the Bilateria of a cell population restricted in fate (in this case neural) but able to generate multiple different cell types (i.e. sensory cells, ganglion cells and nematocytes). Our data are in contrast to studies on a different lineage of cnidarians, the Hydrozoa, in which neurons are mainly derived from pluripotent interstitial stem cells that give rise to both neural and non-neural cells (Bosch and David, 1987; Bosch, 2009; David, 2012; Muller et al., 2004). These observations suggest that Nematostella neural development is more akin to the epithelial progenitor-based neurogenesis of bilaterians than that of hydrozoans, although the levels and extent of homology between these processes remain to be analysed in detail. Indeed, with current data we cannot determine whether the NPCs of Nematostella and bilaterians are homologous, or whether they represent independent events of lineage restriction from an ancestral stem cell population which possessed broader, non-neural potential. Plasticity in the evolution of NPC lineages has previously been observed in the independent acquisition of neural stem cells in the arthropod and vertebrate lineages (Eriksson and Stollewerk, 2010).

By tracing the fate of NvSoxB(2)-expressing cells, we have revealed that sensory and ganglion neurons and nematocytes are linked via a common activation of the NvSoxB(2) promoter during their ontogeny (Fig. 7A). Following on from this discovery, it will be important to determine the potency of the NvSoxB(2)⁺ population – whether it is homogenous [in that all NvSoxB(2)⁺ cells have equal developmental potential] or whether it consists of multiple classes of neural progenitors that will go on to generate different numbers and types of neural cells (Fig. 7B). One possibility is that the NPC pool might be subdivided into nematocyte-producing cells and into cells that produce both sensory and ganglion neurons. Certainly at the molecular and morphological level, sensory and ganglion neurons have more in common with each other than either has with nematocytes [e.g. Marlow et al. (2009); Nakanishi et al. (2012)]. Such putative subsets of NPCs might in turn be defined by specific gene expression signatures, similar to the transcription factor coding of neural fate seen in bilaterian neurogenesis (Guillemot, 2007).

Intriguingly, our analysis of S phase labelling within sibling cells in the NvSoxB(2)::mOrange line demonstrated asymmetric cell behaviours within neural lineages of Nematostella (Fig. 5; supplementary material Fig. S5). In addition to identifying clusters of transgenic cells containing uneven cell numbers [most likely indicating asynchronous cell division from an initial NvSoxB(2)⁺ progenitor], we observed that only one cell in a pair of sibling transgenic cells incorporated EdU, even after a period of prolonged EdU incubation. This probably indicates that either one of the two cells is already post-mitotic (G0), or that it is in an extended G1 phase and thus lagging behind the cell cycle of its sibling. These observations are noteworthy, as, both in mammals and Drosophila, changes in the length of cell cycle phases have been found to mark changes in the future trajectories of NPCs (Bayraktar et al., 2010; Bowman et al., 2008; Calegari et al., 2005; Takahashi et al., 1995). The cell cycle asymmetries we have documented in Nematostella might reflect differential self-renewal and/or the adoption of distinct fates by sibling cells within NvSoxB(2)⁺ lineages. To understand the significance of these asymmetries, and their relationship to the generation of neural diversity in Nematostella, will require the application of live-imaging techniques and/or detailed clonal analysis, in combination with the development of an improved suite of neural cell type-specific markers.

An unexpected finding was that the absolute number of NvSoxB(2)⁺ cells in our sampled area of ectoderm remained roughly the same from blastula to planula stage (Fig. 4). This could indicate that the NvSoxB(2)⁺ population is a stable pool of stem cells residing in the ectoderm, from which differentiated neural cell types [no longer expressing NvSoxB(2)] emerge over time. Inconsistent with this scenario is that NvSoxB(2)⁺ cells labelled with EdU generally display lower levels of NvSoxB(2) mRNA than other NvSoxB(2)-expressing cells. Furthermore, in the transgenic line we did not observe an accumulation of the reporter protein in cells labelled with EdU, as would be expected if this was a stable population of cells in which NvSoxB(2) was being continuously expressed. We therefore favour a scenario in which NvSoxB(2)⁺ progenitors are constantly being produced, either from an unidentified pool of neural stem cells, or directly from ectodermal cells. In the latter case, their number could be kept constant over time by restrictive patterning mechanisms, such as lateral inhibition.

Among bilaterians there is no simple correlation between the orthology and function of specific SoxB family members. Similarly, in cnidarians SoxB genes can be expressed in stem cell populations and/or differentiating cells, in different neural cell types and at different developmental stages (Jager et al., 2011). This suggests that, whereas SoxB function in regulating the developmental progression of neural lineages has been conserved over evolutionary time, considerable variation has arisen in the specific roles that each family member might take during the course of neurogenesis.

The expression and functional data presented here indicate that NvSoxB(2) is required for the proper development of NPCs in Nematostella. NvSoxB(2) may either function as a positive regulator of NPC fate or act in the initiation of a neural differentiation program in an already formed NPC population. In the former scenario, we envisage that the broad expression of NvSoxB(1), NvSox1 and NvSox3 (Magie et al., 2005) provides neural potential to ectodermal cells and that the scattered expression pattern of NvSoxB(2) is indicative of the promotion or stabilization of NPC fate. This scenario might further require that NvSoxB(2) be then downregulated for neural differentiation to proceed, akin to the requirement for SoxB1 downregulation in the differentiation of the chick CNS (Bylund et al., 2003; Sandberg et al., 2005). The latter scenario, in which NvSoxB(2) promotes a neural differentiation program, resembles the function of the SoxB2 group gene Sox21 in chick (Sandberg et al., 2005). Sox21 counteracts the function of the SoxB1 group genes Sox1, Sox2 and Sox3, which maintain NPCs in an undifferentiated state without promoting their proliferative activity (Bylund et al., 2003). Whether a similar antagonism between SoxB genes exists in Nematostella is unknown. The broad expression patterns of NvSoxB(1), NvSox1 and NvSox3 are compatible with a role in maintaining an undifferentiated state that is counteracted by NvSoxB(2), although they do not suggest a role that is specific to neural stem/progenitor cells. We thus favour the first scenario, in which NvSoxB(2) acts in the promotion/stabilization of NPC fate; however, more refined analyses will be required to determine the precise role of NvSoxB(2) in NPC generation and development.

In conclusion, we have identified SoxB genes as ancient genetic components of neurogenesis that are responsible for the regulation of NPC development. Our data further suggest that NPCs feature in the ground pattern of eumetazoan neurogenesis – in that cells destined for a neural fate enter an intermediary phase, whereby they can expand in number and generate an increased diversity of neural cell types via a combination of symmetric and asymmetric division. An improved understanding of the extent of molecular homology underpinning these processes in cnidarians and bilaterians will provide important insights into conserved core aspects of eumetazoan neural development and further illuminate the evolutionary plasticity of neurogenesis.

Animals were maintained in 0.3× filtered seawater (NM) and induced to spawn as described by Fritzenwanker and Technau (2002). Fertilized eggs were removed from jelly packages by incubation for 25 min in 3% cysteine/NM. Embryos were raised at 21°C. Early blastula, 12 h; blastula, 16 h; late blastula, 18 h; early gastrula, 20 h; gastrula, 24 h; late gastrula, 30 h; early planula, 48 h; mid-planula, 72 h; late planula, 96 h; primary polyp, 10 days.

The NvSoxB(2)::mOrange transgenic line was generated by meganuclease-mediated transgenesis as described by Renfer et al. (2010). The NvSoxB(2) promoter sequence (genomic coordinates: minus strand on scaffold 84; 792,491-793,895) was cloned in front of mOrange (Shaner et al., 2004), with the addition of a membrane-tethering CAAX domain to help visualize the boundaries and morphology of cells expressing the reporter protein. Animals used for experiments were progeny from either incrossing of F1 heterozygotes or crossing F2 homozygotes with wild type. In all images, expression of mOrange was visualized via immunostaining with anti-DsRed (rabbit, Clontech, 632496; 1:100).

Microinjections were carried out as described by Rentzsch et al. (2008). Fertilized eggs were injected with 250-500 μM morpholino (Gene Tools) and 40 μg/ml Alexa Fluor-conjugated dextran (Invitrogen) in TAE buffer. Control injections used either a generic control or specific mismatch control [NvSoxB(2)Mm] morpholino. For morpholino sequences see supplementary material Table S2.

Experiments were conducted as described by Nakanishi et al. (2012) and Sinigaglia et al. (2013). For further details see supplementary material. Specimens were imaged on either a Nikon Eclipse E800 compound microscope with a Nikon Digital Sight DSU3 camera or on a Leica SP5 confocal microscope. Figure plates were built using Adobe Design Standard CS5; images were cropped and adjusted for brightness/contrast and colour balance; any adjustments were applied to the whole image, not parts.

Embryos were incubated with 100 μM EdU/DMSO in NM for either 30 min or 2 h at 21°C and then fixed (see methods in the supplementary material) immediately for either ICC or FISH. After ICC or FISH protocols were completed, EdU incorporation was visualized using the Click-iT EdU Alexa Fluor 488 imaging kit (Molecular Probes) following the manufacturer's instructions. For counting NvSoxB(2)⁺ and EdU⁺ cells and cell clusters, a 100 μm×100 μm sampling area was defined in the mid-lateral region of the ectoderm, and serial 1 μm sections through the layer of nuclei within this region were scanned via confocal microscopy. For total cell number estimates, nuclei were counted within multiple 10 μm×10 μm subsamples of the 100 μm×100 μm region; these data were then extrapolated to the whole area.

RNA from embryos from three independent injections was extracted using the RNAqueous kit (Ambion) and DNAase-treated using TurboDNase (Ambion). Quality and quantity of RNA was assessed using a Bioanalyzer (Agilent). cDNA was reverse-transcribed using Superscript III (Invitrogen) primed with random hexamers (Roche). Control reactions without the addition of reverse transcriptase were prepared for all samples. Primer pairs with PCR efficiencies of 90-105% were used for qPCR. Two technical replicates were performed for each of the three biological replicates. Relative expression was calculated using the 2^−ΔΔCt method; control gene stabilities were assessed using RefFinder (http://www.leonxie.com), with NvATPsynthase and NvELF1B being selected as most stable. Mean relative expression and s.e.m. of three biological replicates are plotted. For primer sequences see supplementary material Table S1.

We thank Suat Özbek for generously providing the anti-Ncol3 antibody; Bård Steinar Giezendanner for animal care; Henriette Busengdal and Chiara Sinigaglia for advice on qPCR; and all S8 lab members for constructive discussions and experimental guidance.