FGF signalling controls formation of the apical sensory organ in the cnidarian Nematostella vectensis

All key features of animal body plans are established during embryonic development, and thus an understanding of the early steps in body plan evolution requires an understanding of the embryonic development of phylogenetically basal metazoans. One such metazoan phylum is the Cnidaria,which represents the sister group to the Bilateria. It is estimated that Bilateria and Cnidaria split about 600 million years ago(Fig. 1A). The cnidarian model organism Nematostella vectensis, a sea anemone, is a member of the Anthozoa and is considered to be a representative of the basal group within the Cnidaria (Bridge et al.,1995; Bridge et al.,1992; Collins,2002; Medina et al.,2001) that has retained much of the ancestral genetic complexity of the cnidarian-bilaterian ancestor(Chourrout et al., 2006; Putnam et al., 2007; Ryan et al., 2006; Technau et al., 2005). In contrast to the triploblastic Bilateria, Cnidaria are composed only of ectodermal and endodermal epithelia separated by an acellular extracellular matrix, the mesogloea. Juvenile and adult Nematostella polyps consist of a tube-shaped body column with longitudinal endodermal lamellae, the mesenteries, reaching into the gastric cavity. The pharynx is an ectodermal invagination, the only opening at the oral end is surrounded by a ring of tentacles (Fig. 1E). Nematostella gastrulation (Kraus and Technau, 2006; Magie et al., 2007) is followed by a free-swimming, ciliated planula larva and gradual metamorphosis into a primary polyp(Fig. 1B-E)(Hand and Uhlinger, 1992). During the planula stage, a sensory ciliary tuft develops at the aboral pole(Chia and Koss, 1979). Larvae swim with this apical ciliary organ facing forward and also settle on this pole. Ciliary tuft-bearing apical organs are also located at the aboral pole of marine ciliated larvae of both protostomes (e.g. molluscs and annelids) and deuterostomes (e.g. echinoderms and hemichordates, see Fig. 1A), but their function is not well characterised. Because they are only present in the free-swimming larvae and disappear after metamorphosis, they are thought to be required for the detection of suitable conditions for metamorphosis and/or for directed swimming. However, apical organs are absent in major model organisms, such as Caenorhabditis elegans, Drosophila and vertebrates, and therefore the molecular basis of its development, despite its evolutionary significance, is completely unclear.

In order to identify signalling molecules that are involved in the embryonic patterning of Nematostella, we identified and isolated the complete set of 15 homologous transcripts of Fibroblast growth factors (FGFs)and the only two Fibroblast growth factor receptors (FGFRs) present in the genome. This extends the recent predictive identification and selective cloning of 13 FGF ligands from the Nematostella genome(Matus et al., 2007). FGF signalling is involved in a wide variety of developmental processes in vertebrates and invertebrates (Borland et al., 2001; Bottcher and Niehrs,2005; Huang and Stern,2005; Thisse and Thisse,2005). It regulates the migratory behaviour of cells during gastrulation in vertebrates, sea urchin and Drosophila(Keller, 2005; Leptin, 2005; Rottinger et al., 2008; Wilson and Leptin, 2000),mesoderm formation in vertebrates(Kimelman, 2006), and neural induction in vertebrates and urochordates(Bertrand et al., 2003; Stern, 2005; Wilson and Edlund, 2001). At later developmental stages, it is involved in anteroposterior patterning of the neuroectoderm and the mesoderm in vertebrates(Altmann and Brivanlou, 2001),branching morphogenesis in the Drosophila and mouse respiratory systems (Ghabrial et al.,2003; Metzger and Krasnow,1999; Warburton et al.,2000), limb development in vertebrates(Capdevila and Izpisua Belmonte,2001; Niswander,2002; Tickle,1999), and notochord and heart formation in urochordates(Davidson et al., 2006; Imai et al., 2002; Yasuo and Hudson, 2007). In planarians, a role for FGF signalling in brain development has been proposed(Cebria et al., 2002; Ogawa et al., 2002), and, in the hydrozoan Hydra, a FGF receptor is expressed both in the tip and the foot region of developing buds (Sudhop et al., 2004). However, to date, no functional data about the role of individual FGFs in lower metazoans have been reported.

We show here that two paralogous FGF genes (NvFGFa1 and NvFGFa2) and one FGF receptor gene (NvFGFRa) are expressed at the aboral pole and, later, in the apical organ of the Nematostella embryo. We employ morpholino-mediated knockdown to show that signalling of NvFGFa1 via NvFGFRa is required for specification of the ciliary tuft of the apical organ. By surprising contrast, NvFGFa2 is required to prevent the precocious and ectopic formation of ciliary tuft cells by antagonising NvFGFRa signalling. Furthermore, NvFGFa1 maintains its own expression and that of NvFGFRa and NvFGFa2, indicating that both positive- and negative-feedback loops are involved in development and maintenance of the apical organ. Finally, we show that, whereas absence of the ciliary tuft upon NvFGFa1 knockdown does not affect the swimming behaviour of the planula larvae, it does completely block metamorphosis.

Nematostella culture and the induction of spawning have been described previously (Fritzenwanker and Technau, 2002; Hand and Uhlinger, 1992). Embryos were raised in one-third filtered seawater (Nematostella medium) at 23°C.

All described genes were identified by searching a NematostellaEST database(http://genome.jgi-psf.org/Nemve1/Nemve1.home.html)with corresponding vertebrate sequences and tBLASTN. Incomplete 5′ ends of NvFGFRb and NvFGFRa were obtained by 5′ RACE using GeneRacer (Invitrogen). GenBank Accession numbers are: NvFGFa2,DQ882654; NvFGFa1, DQ882655; NvFGFRb, EF173462; NvFGFRa, EF173463.

Oligo dT-primed cDNAs for developmental RT-PCR were generated by standard procedures. Primer sequences and PCR conditions are available from the authors upon request.

For the rabbit reticulocyte lysate-based assay, the open reading frame and morpholino target site of NvFGFa1, NvFGFa2 and NvFGFRa were cloned into the pCS2+ vector. In each case, 200 ng of plasmid, 0.5 nmol morpholino and 4 μCi ³⁵S-labeled methionine (Amersham, UK) were added to 10 μl of SP6 TnT-Quick coupled transcription/translation reaction mix (Promega, USA) and incubated for 90 minutes at 30°C. Reactions were separated by SDS-PAGE and synthesised proteins were visualised by autoradiography.

Embryos were fixed for 1 hour in cold 4% paraformaldehyde in PBS or 3.7%formaldehyde in Nematostella medium and stored in methanol until use. Hybridisations were carried out as described(Rentzsch et al., 2006). Probes were synthesised from full-length cDNA clones with Megascript Kits(Ambion, USA) and digoxigenin- or FITC-labeled UTP (Roche, Switzerland).

For staining with anti-acetylated tubulin antibody (Sigma T6793), embryos were fixed for 1 hour in cold 4% paraformaldehyde/PBS, followed by 20 minutes in 10% DMSO in PBS and 15 minutes in 2% H₂O₂ in PBS. The antibody was diluted 1:400 in 10% lamb serum, 1% DMSO, 0.1% Triton X-100 in PBS.

Scanning electron microscopy was carried out at the Molecular Imaging Centre (MIC) of the University of Bergen (FUGE, Norwegian Research Council)using a Jeol JSM7400-F microscope. Samples were treated as described previously (Kraus and Technau,2006); the blastopore was used for orientation of the larvae.

For microinjections, fertilised eggs were dejellied with 2.5% cysteine in Nematostella medium(Fritzenwanker and Technau,2002). Injections were done with a Femtojet microinjector(Eppendorf, Germany) on a Nikon TE2000-S inverted microscope.

Morpholinos (MOs) were purchased from Gene Tools, USA. Sequences are: NvFGFa2, CGTTAGCATGGTGATCGTCATGTTG; NvFGFa1,ATAAGGTGGACGCATGACTTTGTAG; NvFGFRa,TCCACCAAGCTCGAAGAGCCGTCAT; and control MO,CATGGAGAAATCGGACTTCATATTT. Nucleotides complementary to the start ATG are underlined. The sequence of the control MO does not yield any hits in the available Nematostella genome assembly or EST database(http://genome.jgi-psf.org/Nemve1/Nemve1.home.html). MOs were diluted in water and injected at 0.25 nmol/μl (NvFGFa2)or 0.5 nmol/μl (NvFGFa1 and NvFGFRa) with 0.5 μg/μl rhodamine-dextran (M_r 10,000, Molecular Probes, USA) as a tracer.

SU5402 (Calbiochem, USA) was applied at a final concentration of 20 μM,UO126 (Promega, USA) at 10 μM, each in 0.1% DMSO. Control animals were incubated in 0.1% DMSO only. Solutions were changed after 8 hours.

Double phosphorylated ERK was detected with monoclonal anti-phospho p42/44 antibody E10 (Cell Signaling Technologies, USA).

Based on genomic and EST resources, we identified 15 FGF-domain containing transcripts (see Fig. S1 in the supplementary material) (J.H.F. and U.T.,unpublished) (Matus et al.,2007), two of which will be described here in detail. They encode putative proteins of 202 amino acids (aa; NvFGFa1)(Matus et al., 2007) and 197 aa (NvFGFa2, this study). The N terminus of each protein contains a predicted signal peptide [SignalP3.0 (Bendtsen et al., 2004)], indicating that they can act as secreted factors. Phylogenetic analyses based on the FGF domain shows that NvFGFa1 and NvFGFa2 belong to a eight-membered paralogous group that cannot be assigned with certainty to a particular subfamily (see Fig. S1 in the supplementary material). Within this paralogous group, NvFGFa1 and NvFGFa2are distantly related. NvFGFa1 has been identified in parallel and named NvFGF1A (Matus et al.,2007), whereas NvFGFa2 has not yet been described. To avoid the impression that NvFGFa1 and NvFGFa2 belong to the FGF1 subfamily, we prefer to use the names NvFGFa1 and NvFGFa2 in the following sections.

We also isolated the complete open reading frame of two paralogous FGF receptors, which we term NvFGFRa and NvFGFRb in accordance with Matus et al. (Matus et al.,2007) (see Fig. S2 in the supplementary material). No other FGF receptors could be identified in the genome. Both contain three extracellular immunoglobulin domains, a single transmembrane domain and an intracellular split tyrosine kinase domain (for domain organisation, see Fig. S3 in the supplementary material). From the hydrozoan Hydra, a single FGFR,termed Kringelchen, has been reported so far (Sudhop et al., 2005), which clusters with the Nematostella FGFRs (see Fig. S2 in the supplementary material). As the duplication leading to FGFRa and FGFRb appears to be lineage specific, we conclude that the common ancestor of Cnidaria and Bilateria had one FGFR, provided that no other FGFR has been lost in the cnidarian lineage.

To determine whether FGF signalling might be involved in embryonic development of Nematostella, we analysed the temporal and spatial expression pattern of NvFGFa1, NvFGFa2, NvFGFRa and NvFGFRbby RT-PCR and in situ hybridisation, respectively. RT-PCR on cDNA from different developmental stages shows that both NvFGF receptors and NvFGFa2 are expressed maternally and zygotically, whereas NvFGFa1 is expressed only zygotically(Fig. 2A). By in situ hybridisation, the earliest localised expression detected for NvFGFa1and NvFGFa2 was in early gastrula stages, as a broad domain encompassing almost the complete aboral half of the embryo(Fig. 2B,F). During gastrulation, this broad domain becomes gradually restricted to the small patch at the aboral pole that marks the site where the apical organ will develop in the early planula larva (Fig. 2C,G). Expression of both genes remains confined to this site throughout planula stages and the first 2 to 3 days after metamorphosis(Fig. 2D,E,H,I), and becomes undetectable afterwards. Double in situ hybridisation did not reveal a difference in the width of the expression domains of NvFGFa1 and NvFGFa2 during midgastrulation, when the expression becomes restricted to the aboral pole (Fig. 2N-P).

The expression of NvFGFRa is very similar to that of NvFGFa1 and NvFGFa2: it commences broadly in the aboral half and then becomes restricted to the aboral pole(Fig. 2J-M). However,fluorescent double in situ hybridisation experiments show that, in contrast to NvFGFa1, the expression domain of NvFGFRa is slightly wider than that of NvFGFa2 during the narrowing of the aboral expression domain (Fig. 2Q-S). In addition, NvFGFRa is expressed in the whole endoderm during planula stages, with pronounced signals detectable in the mesenteries(Fig. 2K,L). The second FGF receptor, NvFGFRb, is expressed uniformly throughout the endoderm during gastrulation and planula stages (data not shown) (see Matus et al., 2007).

Taken together, NvFGFa1, NvFGFa2 and NvFGFRa share a common expression pattern at the aboral pole, compatible with a function in the development of the apical organ. We therefore focused subsequent functional analyses on these genes.

We used morpholino antisense oligonucleotides (MO)(Summerton, 1999) to analyse the function of NvFGFa1, NvFGFa2 and NvFGFRa during Nematostelladevelopment. To test whether the morpholinos can suppress translation of the targeted transcripts, we employed a reticulocyte lysate-based transcription-translation system (see Material and methods). Synthesis of NvFGFa1 protein from a NvFGFa1 encoding plasmid was readily suppressed by addition of the NvFGFa1 MO to the reaction, but not by the NvFGFa2 MO. Similarly, the NvFGFa2 MO, but not the NvFGFa1 MO, suppressed synthesis of NvFGFa2, and translation of NvFGFRa was suppressed by the NvFGFRa MO, but not by the NvFGFRb MO (Fig. 3A).

Embryos injected with a control morpholino or with dextran developed normally into planula larvae, although with a slight developmental delay compared with uninjected embryos. Scanning electron microscopy of mid-planula(96 hpf) embryos revealed that NvFGFa1 MO- or NvFGFRaMO-injected embryos lack the ciliary tuft(Fig. 3B,C; see also Fig. S4 in the supplementary material). By striking contrast, the injection of morpholinos against the co-expressed paralog NvFGFa2 resulted in a pronounced expansion of the ciliary tuft(Fig. 3D). Furthermore,staining of cilia with an antibody against acetylated tubulin showed that, in embryos injected with the NvFGFa2 MO, the long apical cilia develop significantly earlier than in control embryos. At early planula stage (48 hpf), before the apical tuft is visible in control embryos, a vastly oversized tuft of apical cilia is detectable in NvFGFa2 morphants(Fig. 3E-G).

The observation that both FGFs and the FGF receptor are expressed in the apical organ cells throughout planula stages and in the young primary polyp suggested that FGF signalling might still be required after the initial formation of the apical organ. To test this possibility, we used a chemical FGF receptor inhibitor, SU5402, which binds to a region of the tyrosine kinase domain of FGF receptors that is highly conserved in both NematostellaFGF receptors (Mohammadi et al.,1997). Incubation of Nematostella embryos with SU5402 leads to a clear reduction in phosphorylation of the MAP kinase ERK, which is phosphorylated and thereby activated by FGF signalling in various higher metazoans (Fig. 6A). We selected planula larvae after formation of a visible apical tuft (72 hpf), and incubated them in 20 μM SU5402. Within 48 hours, 67% of the SU5402-treated planulae (n=33) completely lost the apical cilia(Fig. 3I) and 24% had a clearly thinner apical tuft; control incubation in 0.1% DMSO had no effect(Fig. 3H; n=38).

We conclude that opposing activities of NvFGFa1 and NvFGFa2 regulate proper development of the apical organ in Nematostella, and that FGFR signalling is required to maintain the apical cilia throughout planula stages.

To obtain a better understanding of the patterning defects that underlie the observed phenotypes, we used a panel of marker genes that demarcate distinct regions along the oral-aboral axis. Expression analysis was carried out at 48 hpf and, thus, about 24 hours before differentiation of the apical organ becomes apparent by the emergence of the apical cilia.

NvCOE is a homolog of the Collier/Olf/EBF family of transcription factors that are implicated in neuronal development in various organisms(Dubois and Vincent, 2001). NvCOE is expressed in the apical organ of the Nematostellaearly planula larvae (Fig. 4A)(Pang et al., 2004). This expression is lost in embryos injected with the NvFGFa1 or NvFGFRa MO (Fig. 4E,I), whereas it is strongly expanded upon NvFGFa2 MO injection (Fig. 4M). NvFoxD1 is a winged helix transcription factor that is expressed in a broad aboral domain of the Nematostella planula larvae and thus includes, but goes beyond the expression domain of NvCOE at this stage (Fig. 4B)(Magie et al., 2005). The expression of NvFoxD1 is unaffected in NvFGFa1 and NvFGFRa MO-injected embryos (Fig. 4F,J), whereas its expression in NvFGFa2 morphants includes the expanded apical organ, but does not exceed the expression domain of NvCOE (Fig. 4N). We also analysed the expression of NvWnt2, which is expressed in a belt-like domain in the central part of the planula(Fig. 4C)(Kusserow et al., 2005), and that of NvFkh, which at this stage is expressed around the blastopore, marking the oral end of the planula(Fig. 4D)(Fritzenwanker et al., 2004; Martindale et al., 2004). Both markers were unaffected by injection of the NvFGFa1, NvFGFa2 or NvFGFRa MOs (Fig. 4G,H,K,L,O,P).

Taken together, knockdown of NvFGFa1 or NvFGFRa leads to a specific loss of the apical organ marker NvCOE, without affecting other markers along the oral-aboral axis, whereas knockdown of NvFGFa2 results in expansion of the apical organ territory into the NvFoxD1-positive aboral region. This suggests that NvFoxD1 specifies a broad aboral identity rather than the apical organ itself, but possibly defines the domain where an apical organ can be formed. As NvFoxD1expression is not affected in knockdown experiments, it appears to act upstream or in parallel to the FGF signalling pathway.

The above results indicate that proper development of the apical organ requires a balance between the opposing activities of NvFGFa1 and NvFGFa2. One possible way to achieve such a balance would be the use of feedback mechanisms. We therefore analysed whether the expression of NvFGFa1,NvFGFa2 and NvFGFRa is regulated by FGF signalling. Indeed, we found that knockdown of NvFGFa1 or NvFGFRa leads to a loss of transcription of NvFGFa1 and NvFGFa2, and to a nearly complete loss of NvFGFRa expression(Fig. 5A-I). By contrast,knockdown of NvFGFa2 resulted in an expansion of the expression domains of NvFGFa1 and NvFGFRa, and of NvFGFa2itself (Fig. 5J-L). The expanded expression domains of NvFGFa1 and NvFGFRa in NvFGFa2 morphants suggest that the expansion of the apical organ in these embryos might be caused by excessive signalling of NvFGFa1 via NvFGFRa. To test this possibility, we co-injected NvFGFa2 and NvFGFRaMOs, and found that co-injection suppressed the expansion of the apical organ territory caused by injection of NvFGFa2 MO alone(Fig. 5M,N). Similarly,incubation of NvFGFa2 MO-injected embryos with the chemical FGF receptor inhibitor SU5402 completely blocked formation of the apical cilia(Fig. 5O).

We conclude that NvFGFa1 signalling via NvFGFRa maintains its own expression, as well as that of NvFGFRa and the antagonistic NvFGFa2 (Fig. 7H), and that the expansion of the apical organ caused by knockdown of NvFGFa2is mediated by NvFGFRa signalling.

Activation of FGF receptors can trigger several intracellular transduction pathways. One of the most prominent among these is a conserved Ras/Raf/MEK/MAP kinase pathway, which mediates FGF signalling in many developmental processes in other animals. Manipulation of this pathway can be achieved by the application of UO126, a chemical compound that specifically blocks the activity of MEK, which is a specific activator of the MAP kinase ERK(Favata et al., 1998). Western blot analysis with an antibody against double-phosphorylated (i.e. activated)ERK shows that UO126 almost completely abolishes ERK activation in Nematostella (Fig. 6A). As ERK is also involved in FGF-independent processes, we applied UO126 only after blastula stage (20 hpf) to minimise the risk of non-FGF related effects that might secondarily affect apical organ formation. Treatment with UO126 blocked apical organ formation and expression of the apical organ markers NvCOE and NvFGFa1, but did not affect the expression of NvFoxD1 (Fig. 6B-D; data not shown). Moreover, UO126 blocked the expansion of apical organ markers caused by the injection of NvFGFa2 MOs(Fig. 6E-G). These data suggest a major role for a Ras-MEK-ERK cascade in FGF-dependent apical organ formation.

The analysis of marker gene expression has shown that knockdown of NvFGFa1 and NvFGFa2 specifically affects formation of the apical organ. This allowed us to examine the possible consequences of loss or expansion of the apical organ for the behavior and later development of the larvae. After gastrulation, Nematostella larvae begin to swim freely,first mainly in circles, later in a more directional manner (see Movies 1, 2 in the supplementary material). In both cases they swim with the apical organ pointing forward. We observed that, despite the absence of the apical organ,larvae injected with the NvFGFa1 MO start to swim freely with the aboral pole forward (see Movie 3 in the supplementary material). Although a minority of them is also capable of directional swimming(Fig. 7A,B; see also Movie 4 in the supplementary material), they predominantly swim in circles. NvFGFa2 morphants with an oversized apical organ are also able to swim with their apical pole forward; however, they swim slowly and only upon agitation (see Movie 5 in the supplementary material). Hence, the apical organ is not essential for swimming, which appears to be solely accomplished by the cilia that cover the whole ectoderm.

As the apical organ of bilaterian larvae has been suggested to be involved in metamorphosis, we next examined whether NvFGFa1 or NvFGFa2 morphants undergo normal metamorphosis into primary polyps. After 12 days, 71% of uninjected (n=156), 67% of control MO-injected(n=133), 64% of NvFGFa2 MO-injected (n=85), but only 2% of NvFGFa1 MO-injected (n=98) embryos had become primary polyps (Fig. 7C-F).

The experiments with the FGF receptor inhibitor SU5402 had shown that continuous FGF receptor activity is required for the maintenance of the apical organ. To support the idea that the apical organ is essential for the induction of metamorphosis, we incubated Nematostella larvae from late planula stages (120 hpf) in SU5402. Whereas 93% (n=159) of the control embryos had become primary polyps at 9 dpf, only 10% (n=165)of the SU5402-treated planula had begun metamorphosis, and none had become primary polyps (Fig. 7G). Although inhibition of FGF receptor signalling by SU5402 is not restricted to the apical organ, these results suggest that loss of the apical organ in NvFGFa1 morphants and upon SU5402 treatment impairs the ability to undergo metamorphosis.

Our results show that development of the apical ciliary organ in Nematostella is controlled by the antagonistic interplay of NvFGFa1 and NvFGFa2. However, our finding that the expression domains of NvFoxD1 and NvWnt2 are essentially unaltered in NvFGFa1 and NvFGFa2 morphants indicates that, despite the early broad expression of NvFGFa1 and NvFGFa2, they are not involved in global patterning of the embryo. Rather, their function is specifically required within the aboral NvFoxD1-positive territory for the proper development of the most aboral cells, which form the apical organ. NvFoxD1 thus probably acts upstream or in parallel to the FGF signalling pathway.

We propose that in the wild-type situation, the apical organ-promoting activity of NvFGFa1 is antagonised by NvFGFa2 until the initially broad expression domain of NvFGFa1 is restricted to a small patch at the aboral pole in the early planula larvae. Because NvFGFa1 activates transcription of both NvFGFa2 and NvFGFa1, it maintains a balance between promoting and suppressing signals, which in turn prevents precocious apical organ formation. Upon suppression of NvFGFa2translation by injection of the NvFGFa2 MO, NvFGFa1 activity is enhanced by the lack of the antagonistic NvFGFa2 protein and by elevated autoregulation of its own transcription(Fig. 7H), resulting in earlier differentiation of the long apical cilia from a broader, i.e. not fully restricted, domain. As in wild-type larvae, expression of NvFGFa1 and NvFGFa2 is then maintained in the differentiated apical organ cells,and this late expression is required for the maintenance of the apical cilia,as even late inhibition of the FGF receptor by the specific inhibitor SU5402 leads to loss of the apical organ. These data are consistent with a reaction-diffusion type of patterning mechanism. It would predict that the inhibitor (NvFGFa2) diffuses faster and thus has a longer diffusion range than the activator (NvFGFa1) in order to delimit the range of signalling and,thereby, the formation of the apical organ to a small area. Although theoretical models have stressed for a long time the power of reaction-diffusion mechanisms to generate spot- or stripe-like patterns(Meinhardt and Gierer, 2000),molecular evidence has remained relatively scarce. The best-studied example is the pair of antagonistically acting TGFβ ligands, Nodal and Lefty, in vertebrate development (for reviews, see Juan and Hamada, 2001; Solnica-Krezel, 2003). Although attracting and repulsive FGF signals have been invoked to act in mesoderm migration during chick gastrulation(Yang et al., 2002), our data are, to our knowledge, the first example of two FGF ligands that are co-expressed, are auto- and crosscatalytic, and have activating and inhibiting effects consistent with a reaction-diffusion type mechanism.

NvFGFa1 and NvFGFa2 are co-expressed from early gastrula stages on, yet only after restriction of their expression domains in the early planula does differentiation of the apical cilia occur. The premature differentiation of the apical cilia upon knockdown of NvFGFa2suggests that it is a quantitative change in NvFGFa1 activity that triggers this differentiation. An excess of NvFGFa1 protein might accumulate over time by a slightly higher rate of synthesis, or a longer half-life of the transcripts or the protein, and finally reach a threshold level for cilia differentiation. Alternatively, a subtle change in expression levels of NvFGFa1 and/or NvFGFa2 at early planula stage might lead to an elevated net NvFGFa1 signalling activity.

The mechanism that drives the restriction of NvFGFa1, NvFGFa2 and NvFGFRa expression into the small aboral domain during gastrulation remains unclear. The activity of NvFGFa2 itself might be required for this process, as an expansion of apical pole markers in NvFGFa2 morphants was already visible at midgastrulation (data not shown). However, expression of apical organ markers is maintained in differentiated apical organ cells, and it is therefore difficult to distinguish whether the early expansion of the expression domains is caused by a direct role of NvFGFa2 in their restriction,or indirectly by the premature onset of differentiation of the apical organ cells. Interestingly, all described apical organ markers in Nematostella either display the same expression dynamics as the FGFs,or their expression commences only at later stages. So far, no gene has been described that is expressed in a spot-like domain at the apical pole during gastrulation, and that would thereby provide a localised source for initiation of the restriction. Dissection experiments have shown that only the oral half of gastrula-stage embryos is able to regenerate a complete planula larva,including a correctly patterned apical tuft(Fritzenwanker et al., 2007; Lee et al., 2007), and it therefore appears possible that the restriction is not triggered by an`attractive' signal from the aboral pole, but rather by a `repulsive' signal from the oral pole. Wnt factors would be a candidate group of molecules for such a function, as several of them are expressed in distinct ring-like domains located exclusively in the oral half of the embryo(Kusserow et al., 2005). A similar mechanism operates in the sea urchin embryo, where the size of the apical organ is restricted by β-catenin signalling from the opposite,vegetal pole (Yaguchi et al.,2006).

The Nematostella genome encodes two FGF receptors, but only NvFGFRa is co-expressed with NvFGFa1 and NvFGFa2 in the apical organ. This co-expression suggests that both FGFs use NvFGFRa as their receptor. As knockdown of NvFGFa1 and NvFGFRa have identical effects on apical organ formation and marker gene expression, one obvious possibility to account for the opposing activities of the two ligands would be that only NvFGFa1 is able to activate the receptor, whereas NvFGFa2 binds to NvFGFRa without activating it and thus acts as a dominant-negative ligand. Alternatively, binding of NvFGFa1 and NvFGFa2 could activate different transduction pathways that antagonise each other intracellularly. To our knowledge, an antagonism of two co-expressed FGFs that signal via the same receptor would represent a novel mechanism for the fine-tuning of FGF activity. Interestingly, a similar mechanism has been described for a member of the TGFβ family of growth factors, whereby BMP3 can bind and block ActRIIB, a receptor that is activated by the binding of BMP4 or Activin(Gamer et al., 2005).

Our results suggest that signalling of NvFGFa1 via NvFGFRa might be mediated by conserved intracellular pathway components. The ability of the MEK inhibitor UO126 to block apical organ formation in wild-type embryos and in NvFGFa2 morphants is compatible with an involvement of a Ras/MEK/ERK signalling cascade in NvFGFa1 signalling. This pathway mediates the response to FGFs in many developmental processes in higher metazoans(Eswarakumar et al., 2005; Thisse and Thisse, 2005). However, the MAP kinase pathway can also be activated by other signals, and indeed blocking of FGF receptor signalling only partially abolishes the phosphorylation of ERK/MAP kinase (Fig. 6A). This indicates that ERK is also activated by another pathway besides the FGF signalling pathway. This other signalling input is apparently not essential for apical organ formation, as expansion and abolishment can all be achieved by manipulation of the corresponding FGF signalling pathway.

Apical organs with a tuft of long cilia have been described in marine ciliated larvae from the two major bilaterian groups, protostomes and deuterostomes, but the evolutionary relationship of cnidarian, protostomian and deuterostomian apical organs is not yet clear.

Strikingly, FGFs or FGF receptors are expressed in the region of apical organ formation in sea urchin and hemichordate embryos [two deuterostomians(Gerhart et al., 2005; Lapraz et al., 2006)], and in the polychaete Platynereis (a protostomian; P. Steinmetz and D. Arendt, personal communication). Thus, although functional data are lacking for these organisms, it is tempting to speculate that regulation of apical organ formation by FGF signalling is common to cnidarian, protostomian and deuterostomian larvae, and might thus represent an ancestral function of FGF signalling that was present in the common ancestor of all eumetazoans. However, broader species sampling and more marker genes are necessary to conclude on the homology of apical organs in cnidarian and bilaterian larvae.

In bilaterian larvae, morphological observations, and pharmacological and cell ablation experiments, have suggested that apical organs are chemo- and/or mechanosensory structures with neuroendocrine functions that might be involved in the induction of metamorphosis(Hadfield et al., 2000; Kempf et al., 1997; Voronezhskaya and Khabarova,2003). Our results strongly indicate that FGF signalling is required for metamorphosis, most likely through the formation of the apical organ. Because metamorphosis of larvae into adults is of pivotal importance for all pelago-benthic directly and indirectly developing organisms, future work will attempt to identify the external cues and their internal processing.

We thank Tom Becker, Daniel Chourrout, Michael Oelgeschlaeger and Patrick Steinmetz for critical reading of the manuscript. We acknowledge Monica Martinussen for excellent care of the Nematostella culture, the Technau group for continuous discussion, Marianne Houssier for help with initial in situ experiments, and Olav Bjørkelund, Anne Nyhaug and Trygve Knag from the Molecular Imaging Center, University of Bergen. Research was funded by the Sars Centre/Unifob budget (U.T.).