Complex functions of Mef2 splice variants in the differentiation of endoderm and of a neuronal cell type in a sea anemone

Among the fundamental questions in biology is the evolutionary origin of the nervous system and the musculature in multicellular animals. The differentiation of these key tissues is governed by a set of conserved transcription factors. Many of these transcription factors have diverse functions in the development of several tissues and developmental processes. Identifying the ancestral role of these transcription factors would be a crucial step in reconstructing the evolution of the tissues involved. One of the key determinants of both muscle and neuronal development in vertebrates and flies is the transcription factor myocyte enhancer factor 2 (Mef2) (Potthoff and Olson, 2007). Although the highest expression levels of Mef2 are in muscle and neuronal tissue (Black et al., 1997), in vertebrates it is also expressed in some other tissues. It has been reported that mouse Mef2 is expressed ubiquitously at low levels (Yu et al., 1992). In combination with Twist, Mef2 is one of the key regulators of muscle differentiation in flies (Sandmann et al., 2007; Sandmann et al., 2006) and vertebrates (Wang et al., 2001), where it also regulates cartilage and bone formation from the mesodermal and neural crest precursor cells (Arnold et al., 2007; Verzi et al., 2007). In addition, in vertebrates, Mef2 plays an important role in neuron and lymphocyte differentiation and survival/apoptosis (Ananieva et al., 2008; Wang et al., 2009), as well as in activity-dependent synapse formation (Flavell et al., 2006; Flavell et al., 2008). Neuronal expression of Mef2 and its role in circadian behaviour have also been shown in insects (Blanchard et al., 2010; Farris et al., 1999; Schulz et al., 1996). Whereas invertebrates possess a single Mef2 gene, vertebrates possess four paralogues, termed Mef2a, Mef2b, Mef2c and Mef2d, which are expressed in overlapping, but distinct, domains (Potthoff and Olson, 2007).

Mef2 binds DNA with its conserved N-terminal MADS domain. The MADS domain is directly followed by a short conserved Mef2 domain. Together, they facilitate DNA binding, dimerisation and co-factor recruitment. The C-terminal transactivation domain shows very little conservation, even between the four vertebrate Mef2 paralogues (Potthoff and Olson, 2007). In both protostomes and deuterostomes, Mef2 mRNA is alternatively spliced to produce multiple isoforms, which, in part, are differentially expressed (Martin et al., 1994; Gunthorpe et al., 1999; Zhu and Gulick, 2004; Zhu et al., 2005). Little is known about the specific functions of these splice variants in Bilateria. A recent report suggests that aberrant splice variants are involved in myogenic disorders (Bachinski et al., 2010). It is unclear what the ancestral function of Mef2 was and whether alternative splicing of Mef2 arose independently in vertebrates and invertebrates.

In this context, investigating the role of Mef2 in cnidarians is of particular importance. These diploblastic organisms are among the earliest diverging animal groups to possess a nervous system and muscles. A cnidarian Mef2 homologue (PcMef2) was first reported in the hydrozoan Podocoryne carnea (Spring et al., 2002). PcMef2 expression is upregulated in the larval endoderm and the endoderm of the polyp at the site of the medusa bud. PcMef2 is also expressed in the plate endoderm, the forming velum, the striated muscle and the tentacle buds of the forming medusa (Spring et al., 2002). Strikingly, in the sea anemone Nematostella vectensis, Mef2 (Nvmef2) was described as an ectodermally expressed gene (Martindale et al., 2004). The transcript was detected in single cells throughout embryonic and larval development (Martindale et al., 2004). These cells are likely to be precursors of neurons or the cnidarian-specific nematocytes. Nematocytes are sensory stinging cells that are considered to be a neuronal cell type (Watanabe et al., 2009).

To gain further insight into the function of cnidarian Mef2, we performed a detailed analysis of this gene in N. vectensis. We identified six alternative splice variants of Nvmef2 that fall into two groups with respect to protein structure and expression pattern. Antisense morpholino knockdown experiments demonstrated that one group of Nvmef2 splice variants is crucial for proper endoderm development. Our data suggest that the dual role of Mef2 proteins in neural and non-neural tissue differentiation predated the cnidarian-bilaterian split.

Animals were cultured and spawning was induced as described (Fritzenwanker and Technau, 2002; Genikhovich and Technau, 2009a).

The Nvmef2 splice variants were cloned by PCR based on the EST and genomic sequences. Accession numbers for splice variants I to VI are: HQ634794, HQ634795, HQ634796, HQ634797, HQ634798 and HQ634799, respectively. In situ hybridisation was performed as described (Genikhovich and Technau, 2009b). The NvNcol3 probe was generated based on GenBank sequence FC288770. RT-PCR analysis of cDNA samples was normalised to β-actin. For β-actin and Nvmef2, 21 and 32 PCR cycles were used, respectively.

The activity and specificity of the Nvmef2 translation-blocking MO (ATG-MO) was confirmed in vitro (TnT Quick Coupled Transcription/Translation System, Promega). The mRNA transcribed from the plasmid carrying six silent mutations (AUGGGGCGAAAGAAGAUCCAGAUAU to AUGGGCAGGAAAAAGAUUCAAAUAU) was not recognised by the ATG-MO, whereas translation from the wild-type mRNA was suppressed in a concentration-dependent manner (supplementary material Fig. S1). The sequences (5′ to 3′) and working concentrations of the microinjected MOs are: ATG-MO, ATATCTGGATCTTCTTTCGCCCCAT, 125 μM; MOex9, GATGTGCCTAGGGTACAACAACAAT, 500 μM; MOex10, GAAGTCTATGGAGATGATACAGGTA, 500 μM; and MOint10, ACCAATAAGAAAGTTCCTTACCCCT, 125 μM.

For rescue experiments, 125 μM MOex9 or MOint10 containing 50 ng/μl Nvmef2-IIwt or Nvmef2-IIΔC2 mRNA was injected into Nematostella zygotes and the embryos were allowed to develop for 6 days. The C2 box deletion in Nvmef2-IIΔC2 was generated by PCR template switching using: Mef2E10_swF, GCAAACAGACACCCTCGATAACTTCCATCTAGACAG; and Mef2E10_swR, CTGTCTAGATGGAAGTTATCGAGGGTGTCTGTTTGC.

The protein coding sequences of Nvmef2-IIwt and Nvmef2-IIΔC2 were directionally cloned into pNvT (Renfer et al., 2010) digested with PacI and SbfI. mRNA synthesis was performed on a template produced by PCR using the mMessage mMachine SP6 Kit (Ambion).

For construction of the MyHCΔMef::mCherry vector, the two Mef2 and one SQUA/Agamous binding sites predicted with CONSITE (Lenhard et al., 2003) (score >9.5) in the MyHC1 promoter sequence used by Renfer et al. (Renfer et al., 2010) were replaced by the following sequences: CTATAAAAAA (position 193) was changed to GGACGGGGGG; CTAAAAATAT (position 298) was changed to GGACGGGGGG; and CCAAATTTAGACAC (position 348) was replaced by GGCGGGGGTTGGCC.I-SceI digestion and microinjection were performed as described (Renfer et al. 2010).

Phalloidin and TO-PRO3 staining were performed as described (Genikhovich and Technau, 2009c). DNA replication was assayed with the Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen). EdU pulses of 25 minutes were used to label the nuclei. The S-phase index was calculated from the percentage of nuclei incorporating EdU among all nuclei per confocal section. A single medial confocal section was counted for each embryo. Nuclei from 8-16 confocal sections were counted for each datum point. The counts from embryos with the highest and lowest S-phase index in the endoderm were removed from the analysis. Imaging was performed with a Leica SP2 confocal microscope. TUNEL staining on 16 μm cryosections was performed with the In Situ Cell Death Detection Kit, Fluorescein (Roche). Transmission electron microscopy and semi-thin sectioning were performed according to standard techniques (Fritzenwanker et al., 2007).

In order to gain insight into the ancestral structure of Mef2, we searched an EST dataset from the sea anemone Nematostella vectensis (Technau et al., 2005) for the presence of a Mef2 homologue. We identified a cDNA homologous to Mef2, which only partly matched the published sequence of Nematostella Mef2 (Nvmef2) (Martindale et al., 2004). This suggested that Nvmef2 is differentially spliced. Based on the alignment of the different Nvmef2 ESTs and the published Nvmef2 mRNA sequence with the genomic sequence on scaffold 140 as well as on PCR analysis, we conclude that Nvmef2 contains at least 11 exons (Fig. 1A). We confirmed six alternative splice variants by PCR (Fig. 1B). All splice variants share the conserved MADS and Mef2 domains as well as a conserved stretch of sequence directly downstream of the Mef2 domain. However, the splice variants differ drastically in the sequence encoding the C-terminal region of the protein and in the 3′ UTRs. Two classes can be distinguished: (1) splice variants I and II containing exons 9-11 (Nvmef2-I and Nvmef2-II); and (2) splice variants III-VI containing exon 7 (Nvmef2-III, -IV, -V and -VI) (Fig. 1B). Exon-intron boundaries are conserved between Nematostella, vertebrates, amphioxus and Drosophila in the two exons encoding the MADS box and the Mef2 domain. The sequence outside of these domains is too divergent to make any confident predictions about the conservation of the exon-intron structure between Nvmef2 and bilaterian Mef2.

The transcripts containing exon 7 differ due to the alternative splice acceptor sites in exon 7 (Fig. 1A,B). Nvmef2-IV represents the published sequence (Martindale et al., 2004). Whereas the exon 9-containing Nvmef2-I and Nvmef2-II encode predicted Mef2 proteins of ‘conventional’ size (>450 amino acids), comparable to bilaterian Mef2 (Fig. 1C), the putative proteins encoded by the exon 7-containing splice variants are significantly shorter, with exon 7 encoding only 33, 5, 9 and 8 amino acids in splice variants III-VI, respectively, followed by long 3′ UTRs of up to 1758 bp.

Alignment of the deduced Nvmef2 protein sequence with Mef2 proteins of other species shows 96-100% amino acid similarity for both the MADS and Mef2 domains within Eumetazoa (Fig. 1C; supplementary material Fig. S2A). The transactivation domain following the Mef2 domain in the eumetazoan proteins can be subdivided into four regions (Fig. 1C). The first region (41 amino acids in Nematostella, which we termed C1) directly following the Mef2 domain is 75% similar to the corresponding C1 domain in bilaterian Mef2 proteins (Fig. 1C; supplementary material Fig. S2A). C1 contains the sequence homologous to the muscle-specific variant of the so-called α-domain located in the alternatively spliced coding exon 3α2 in vertebrates [exon 1b according to Martin et al. (Martin et al., 1994)]. Interestingly, we did not find a splice variant that includes the sequence corresponding to the vertebrate 3α1 variant, nor could we find the homologous sequence in the genomic sequence of Nvmef2. Also, unlike in vertebrates, the Nvmef2 α-domain is split between two exons, neither of which appears to be alternatively spliced. The C1 domain is followed by a large region that shows very little sequence similarity, and which in Nvmef2-I and Nvmef2-II is interrupted by a 44 amino acid serine and threonine-rich motif that we termed C2 (Fig. 1C). C2 shows relatively high evolutionary conservation (76% similarity over 44 amino acids between Nematostella and the corresponding amphioxus protein sequence; supplementary material Fig. S2B). This motif is more derived in the Protostomia Mef2, both in Ecdysozoa and Lophotrochozoa proteins, and is absent from vertebrate Mef2B, which appears to be the most derived vertebrate paralogue. In Nematostella, the conserved C2 box is encoded by the central part of Nvmef2 exon 10. The so-called β- and γ-domains, which are known to be important regulatory modules in the transactivation domain of the mammalian Mef2 proteins and are encoded by alternatively spliced exons (Zhu and Gulick, 2004; Zhu et al., 2005; Hakim et al., 2010), do not align with the Nvmef2 sequence.

In situ hybridisation confirmed that the exon 7-containing splice variants are expressed in a subset of ectodermal cells (Fig. 1D-F), which have been interpreted as putative neuronal or nematocyte precursor cells (Martindale et al., 2004). The number of cells expressing exon 7-containing splice variants increases from the gastrula stage toward the mid-planula and then decreases. During metamorphosis (days 5-7), the expression of the exon 7-containing splice variants diminishes in the body column and localises in the tips of the emerging tentacles (Fig. 1D-F). In line with this, RT-PCR analysis showed very low levels of these splice variants at early and late stages of development, with a peak in expression in the gastrula and planula (supplementary material Fig. S3).

By contrast, RT-PCR showed that Nvmef2-II is a weakly expressed maternal transcript, and its levels are maintained throughout development, whereas Nvmef2-I is even weaker and only detectable from blastula to mid-planula stage (supplementary material Fig. S3). In situ hybridisation revealed that Nvmef2-I and Nvmef2-II expression is upregulated in the endoderm of the gastrulae, planulae and primary polyps, but also shows a slightly weaker expression in the ectoderm (Fig. 1G-I). An upregulation of expression was also observed in the individual ectodermal cells of the gastrula. This short-term increase in expression was, however, over by the planula stage. The area immediately around the blastopore of the gastrula, which gives rise to the future pharynx, was free of Nvmef2-I and Nvmef2-II expression.

Although specific sequences for some individual splice variants were too short to allow preparation of specific probes, we can state that the two classes of splice variants occupy distinct, albeit possibly overlapping, spatiotemporal expression domains.

To gain insight into the function of Nvmef2, we performed antisense MO-mediated knockdown of Nvmef2 translation. Microinjection of the translation-blocking MO (ATG-MO), which affects all splice variants, into Nematostella zygotes strongly affected development. Whereas 88% of the mock-injected embryos (n=105) developed into primary polyps by day 8, most (89%) of the embryos injected with ATG-MO (n=225) did not survive until day 8, and out of those that did (n=28), none underwent metamorphosis (Fig. 2A-D). Transmission electron microscopy (TEM) analysis of 8-day-old morphant embryos showed that they had an ectodermal layer with the normal epithelial and gland cells (supplementary material Fig. S4C,D), but the endoderm was scrambled and no gut lumen was visible (Fig. 2D). The gastric cavity was filled with loose cells packed with electron-dense granules (Fig. 2D,F). The mesogloea appeared to be much thinner than normal. The oral-aboral axis of the morphant embryo, however, appeared intact. The oral side was visible by a residual mouth opening (the former blastopore), and at the aboral side we found typical accumulations of the ectodermal gland cells (Fig. 2D), which are normally observed in the vicinity of the apical organ. Directional swimming of the morphant planulae was not perturbed. These morphants never developed further and died by day 12-14.

One of the hallmarks of endoderm differentiation in Nematostella is the formation of the mesenterial muscles. No muscle fibres could be observed by TEM in the morphants, in contrast to the mock-injected control primary polyps of the same age (Fig. 2E,F; supplementary material Fig. S4I for high magnification). Also, the expression of the retractor muscle-specific Myosin heavy chain 1 (MyHC1) (Renfer et al., 2010) was not detectable in the 5-day-old morphant (supplementary material Fig. S5). Thus, the ATG-MO impairs endoderm differentiation and prevents further development and metamorphosis.

To investigate the endodermal function of Nvmef2 we focused on the endodermally expressed splice variants. We prevented proper splicing of exons 9, 10 and 11 by microinjecting zygotes with MOs designed against the splice acceptor site of exon 9 (MOex9), the splice acceptor site of exon 10 (MOex10), and the splice donor site at the 3′ end of exon 10 (MOint10).

Upon injection of MOex9, exon 9 is spliced out together with intron 8. The resulting mRNA has exon 10 in the correct reading frame, i.e. the Nvmef2 protein is missing the domain encoded by exon 9 but is otherwise intact (Fig. 3A). Although we demonstrated by RT-PCR that MOex9 causes mis-splicing of Nvmef2 mRNA, the microinjected embryos developed normally into primary polyps (Fig. 3A). In contrast to MOex9, microinjection of MOex10 caused a strong phenotype. This MO links exon 9 to exon 11 in the incorrect reading frame, generating a stop codon shortly after the beginning of exon 11 (Fig. 3B). The morphant embryos displayed aberrations of endoderm development but proper oral-aboral axis patterning (Fig. 3B; supplementary material Fig. S4A,B). Hence, the loss of exon 9 (82 amino acids) has no phenotypic effect, whereas truncation of the last two exons (187 amino acids) is sufficient to phenocopy the endodermal phenotype of the ATG-MO. Microinjection of MOint10 caused incorporation of an intron 10 with several stop codons into the Nvmef2 mRNA prior to exon 11. The predicted protein of this mis-spliced mRNA lacks 89 amino acids encoded by exon 11. Morphologically, the MOint10 and MOex10 morphants were indistinguishable (Fig. 3C). Since three independent MOs (ATG-MO, MOex10 and MOint10) all resulted in very similar and strong phenotypes, we conclude that the part of the protein encoded by exon 11 is essential for the role of Nvmef2 in endoderm development. This effect does not seem to rely simply on the shortening of the protein by 89 amino acids, as the removal of exon 9 (which encodes 82 amino acids) has no effect.

The conserved C2 motif is encoded by a sequence located not in exon 11, but in exon 10. To elucidate the function of the C2 motif, we performed rescue experiments by co-injecting different combinations of MOs with capped Nvmef2-II mRNA. Injection of MOex9 together with wild-type Nvmef2-II (Nvmef2-IIwt) mRNA resulted in the development of normal primary polyps by day 6 in 53.5% of cases (Fig. 3F; supplementary material Fig. S6A), showing that co-injection of Nvmef2-IIwt mRNA with a control MO does not perturb normal development. Co-injection of MOint10 together with Nvmef2-IIwt mRNA rescued the MOint10 phenotype (Fig. 3D,F; supplementary material Fig. S6C) and resulted in the development of normal primary polyps in 42.7% of cases by day 6. Strikingly, Nvmef2-IIΔC2 mRNA, which lacks the C2 domain, had no rescue effect when co-injected with MOint10 (Fig. 3E,F) and the morphants were indistinguishable from those injected with MOint10 alone (supplementary material Fig. S6B,D). This allows us to conclude that the effect of the MO is specific and that both the conserved C2 domain encoded by exon 10 and the amino acids encoded by exon 11 are required for the proper function of Nvmef2 in endoderm development.

Since the strongest phenotypic effect is seen in the endoderm, we asked whether it is gastrulation itself or the subsequent differentiation of the endoderm that is affected by Nvmef2 knockdown. The zygotes injected with 500 μM MOex10 underwent normal gastrulation (Fig. 4A,D), but by day 2 the endodermal layer began to lose integrity and the archenteron filled with cells, whereas in controls the endoderm formed a columnar epithelium (compare Fig. 4B,C with 4E,F). The same effect was observed when zygotes were injected with MOint10 at 125 μM (data not shown). We conclude that endodermal Nvmef2 is required for the maintenance of the endodermal layer and its proper differentiation.

We then tested whether the morphological changes in the embryo correlated with changes in cell proliferation. We assayed the morphant and control embryos for the number of cells passing through S phase of the cell cycle. The MOex9 and MOex10 morphants did not show significant differences in the S-phase index (the percentage of nuclei incorporating EdU) of their ectodermal cells as compared with uninjected controls (Mann-Whitney U-test for 2 day planula: P≤0.70, MOex10 versus uninjected; P≤0.34, MOex10 versus MOex9; P≤0.13, MOex9 versus uninjected; MOex10, n=6; MOex9, n=9; uninjected, n=8). However, very few endodermal cells of the MOex10 morphants continued to replicate their DNA after gastrulation was complete (Fig. 4G), and in the 2-day planulae the S-phase index of the endodermal cells remained significantly lower (P≤2.21×10^–7, MOex10 versus uninjected; P≤2.16×10^–6, MOex10 versus MOex9; P≤0.17, MOex9 versus uninjected) than in the controls (Fig. 4H). The reduction in the S-phase index is reflected in the decreased number of endodermal cells in the 2-day MOex10 morphant planulae (supplementary material Fig. S7). Apoptosis does not appear to play a role in this, as no TUNEL-positive nuclei could be identified in 30- to 72-hour-old MOex10 morphants, i.e. during the time when the change in cell proliferation takes place (supplementary material Fig. S8).

To gain further insight into the effect of Nvmef2 knockdown in the endoderm, we analysed the expression of three marker genes (Fig. 5; supplementary material Fig. S9A-C). As MOex10 and MOint10 showed the same phenotype for all genes examined, we will refer to the MOex10 phenotype. The early expression of the pan-endodermal marker NvSnailA (Fritzenwanker et al., 2004; Martindale et al., 2004) was unaffected but faded out gradually in the MOex10 morphants after completion of gastrulation (Fig. 5A-B′). This suggests that the endodermal Nvmef2 splice variants have no effect on the initiation of NvSnailA expression but are necessary for its maintenance. The onset of expression of the pharyngeal marker NvFoxA (Fig. 5C-D′) (Martindale et al., 2004; Fritzenwanker et al., 2004) was delayed in comparison to the control (Fig. 5C,C′). At the 2-day planula stage, NvFoxA expression was detectable around the mouth (Fig. 5D,D′) but did not extend inside, indicating that the pharynx was not forming properly. Expression of the aboral ectodermal marker NvFGFa1 (Fig. 5E-F′) (Rentzsch et al., 2008) was, albeit slightly weaker at later stages of development, largely unaffected in the MOex10 morphants in terms of its expression domain, suggesting that the endodermally expressed Nvmef2 splice variants have no effect on the patterning of the aboral ectoderm. Injection of the ATG-MO had very similar effects on the expression of NvFoxA and NvSnailA as the two endodermal splice MOs (supplementary material Fig. S9D,E and Fig. S10A-D′). However, NvFGFa1 expression showed a significant enlargement of the aboral expression domain as well as ectopic expression at the oral end of the planula in ATG-MO morphant embryos. Moreover, in some cases, NvFGFa1 was expressed throughout the ectoderm of the embryo (supplementary material Fig. S9F and Fig. S10E-F5). Since the ATG-MO affects the translation of all Nvmef2 splice variants, this is likely to be an effect of the ectodermal splice variants.

In triploblasts, one of the most obvious functions of Mef2 is in the regulation of muscle differentiation. In morphant Nematostella lacking functional Nvmef2 protein, the retractor muscles do not form because endoderm development is impaired. However, the effect of the Nvmef2 MOs manifests itself before the muscles start to form. We tested whether one of the characteristic targets of the Mef2 proteins in triploblasts, MyHC1, is under the direct regulation of Nvmef2. The Nematostella MyHC1 gene is specifically expressed in retractor muscles (Renfer et al., 2010). A 1.6 kb regulatory fragment of MyHC1, which reproduces the normal expression pattern of MyHC1 in a transgenesis assay (Renfer et al., 2010), contains three binding sites (predicted with high probability) for MADS box transcription factors: two Mef2 binding sites and one SQUA/Agamous binding site. However, polyps carrying an MyHCΔMef::mCherry transgene in which these three sites were mutated still expressed the mCherry reporter in the retractor muscles (supplementary material Fig. S11), suggesting that Nvmef2 binding to these sites is not essential for the expression of MyHC1. Further research will be required to reveal to what extent muscle development in Nematostella relies on regulation by Nvmef2.

Because it has been proposed that Nvmef2 is expressed in putative neuron or nematocyte precursors in the ectoderm (Martindale et al., 2004), we addressed the ectodermal function of Nvmef2. Unfortunately, producing specific MOs against the individual exon 7-containing splice variants is not technically possible, as this would require simultaneous blocking of four splice acceptor sites. Nevertheless, the effects of the ATG-MO and the endodermal splice MO on neuron and nematocyte differentiation can be compared using neuron and nematocyte markers. Whereas ATG-MO and MOex10 had very similar effects on the differentiation of FMRF amide-positive neurons (data not shown), a very clear difference in their effects on nematocyte differentiation was observed. In situ hybridisation of 4-day-old control and morphant planulae with a probe against the N. vectensis minicollagen gene Nv-Ncol3, which encodes a structural protein in the nematocyst (David et al., 2008), as well as TEM and histological analyses of 8-day-old morphant embryos, revealed that the ATG-MO morphants lack nematocytes in the ectoderm, whereas in MOex10 morphants the nematocytes are still present (Fig. 6A-D; supplementary material Fig. S4F-H and Fig. S9G). This suggests that only the ectodermally expressed, but not the endodermally expressed, Nvmef2 splice variants are involved in the differentiation of nematocytes.

Alternative splicing considerably amplifies the size of the proteome encoded by a genome, and, where alternative splicing is developmentally regulated, the variants are thought to have distinct functions. In humans, up to 90% of all genes are thought to have alternative splice variants (Keren et al., 2010), which is considerably more than in Drosophila, Caenorhabditis and Ciona (Kim et al., 2007). Although the role of alternative splicing is extensively studied in Drosophila and vertebrates, very little is known about alternative splicing in lower Eumetazoa. Recently, the first report that alternative splice variants of the Nematogalectin gene are differentially expressed in different nematocyte types in Hydra was published (Hwang et al., 2010). Here, we demonstrate that alternative splice isoforms of an important developmental transcription factor, Nvmef2, have distinct expression domains and developmental functions in the sea anemone N. vectensis.

Similar to NvMef2, Mef2 pro-mRNAs of flies and mammals are alternatively spliced (Martin et al., 1994; Gunthorpe et al., 1999; Zhu and Gulick, 2004; Zhu et al., 2005). Whereas the splice sites in the MADS and Mef2 domains are conserved between Cnidaria and Bilateria, the sequence coding for the C-terminal region, where the alternative splicing occurs, is too diverged to assess conservation of the splice cites. The two major classes of alternative Nvmef2 splice variants differ in their expression pattern and in the size of the resulting proteins: the endodermally expressed variants are similar in size to the vertebrate Mef2 variants, whereas the ectodermal variants are shorter, lacking a significant part of the transactivation domain. The molecular consequences of these differences in protein structure remain to be investigated.

Our results allow us to reconstruct the domain composition of the ancestral eumetazoan Mef2 protein. Besides the highly conserved N-terminal MADS and Mef2 domains, it contained two conserved motifs in the transactivation domain: the C1 domain and the C2 domain. The muscle-specific form of the C1 domain, encoded in vertebrates by exon 3α2, appears to be the only one present in Nematostella and other invertebrates (i.e. the ecdysozoan Drosophila melanogaster and lophotrochozoan Lottia gigantea) and in the basal chordate Branchiostoma floridae. Thus, the second variant of C1 (exon 3α1) and the alternative splicing of exons 3α1 and 3α2, as well as the alternatively spliced exons encoding the β-domains, appear to be vertebrate-specific innovations. The γ-domain has probably evolved at the base of Bilateria, as a very similar motif can be found in the transactivation domain of Mef2 in both Drosophila melanogaster and Branchiostoma floridae, but in these organisms it is not encoded by a specific alternatively spliced exon. This suggests that alternative splicing of cnidarian and bilaterian Mef2 has either evolved independently or that the ancient exons of the transactivation domain have evolved beyond recognition.

In vertebrates, some of the different Mef2 splice variants are expressed in different tissues. However, the expression of the different splice variants is mostly overlapping, and there might be some redundancy in function. Alternative splicing leads to the presence or absence of the so-called α-, β- and γ-domains, which affect protein function as a transcriptional activator or can even change it into a repressor (Martin et al., 1994; Zhu and Gulick, 2004; Zhu et al., 2005). An extra level of transcriptional, and possibly translational, regulation of Mef2A is conveyed by the use of two alternative promoters autoregulated by Mef2A protein and by the generation of several variants of the 5′ UTR by alternative splicing of the non-coding exons (Ramachandran et al., 2008). The situation in Drosophila is probably different in that it is not the particular splice isoforms but the amount of the protein that defines the identity of the differentiating cells (Gunthorpe et al., 1999). More recently, it has been shown that different levels of Mef2 activity are required for the regulation of the different Mef2 targets in Drosophila (Elgar et al., 2008).

In Nematostella, different Nvmef2 splice variants are expressed in different cell types. Our data support the view (Martindale et al., 2004) that ectodermally expressed Nvmef2 is either directly or indirectly involved in the regulation of the nematocyte and probably also in neuron differentiation. We also found that, in the endoderm, Nvmef2 has another function. The earliest effect of Nvmef2 knockdown is the arrest of cell proliferation in the endoderm after gastrulation. The subsequent endoderm differentiation is also affected by blocking Nvmef2 translation and by improper splicing of the endodermally expressed splice variants. By contrast, splice MOs directed against the endodermally expressed Nvmef2 variants have no effect on the development of ectodermal nematocytes. Surprisingly, despite the presence of the three predicted binding sites in the MyHC1 promoter, so far we have found no evidence for Nvmef2 being a direct regulator of this key structural muscle differentiation gene in Nematostella, although it is a known target for Mef2 regulation in flies and mammals. Further experiments on other structural muscle genes are required to determine whether the role of Mef2 as a key muscle determinant evolved only in Bilateria.

Usually, evolutionary sequence conservation is a good indicator of the functional relevance of a protein domain. Our MO-mediated knockdown experiments revealed that exon 11 is crucial for the function of the endodermal splice variants; however, the amino acid sequence encoded by exon 11 is not conserved. On the contrary, the conserved C2 domain is encoded by exon 10. We hypothesized that both exon 10 and exon 11 are required for the function of endodermally expressed Nvmef2. This view is supported by the deletion study of Janson et al. (Janson et al., 2001), who showed that direct coupling of the region encompassing amino acids 312-367 of human MEF2C (which contains the C-terminal half of the C2 domain) to the MADS and Mef2 domains is sufficient to produce a functional transcriptional activator, whereas proteins lacking amino acids 312-350 are very weak activators. In line with a crucial role for this conserved motif in Nvmef2 function in the endoderm, we found that only the form of Nvmef2 mRNA that contains the sequence encoding the C2 domain could rescue the severe phenotype caused by MOint10, whereas a protein lacking this domain was unable to rescue the phenotype. Thus, our experiments reveal a conserved motif in specific splice variants that is crucial for endoderm differentiation in the sea anemone. In vertebrates, the sequence encoding the C2 domain is split between two constitutive exons. Further biochemical analyses are required to understand the function of this domain.

Our results have demonstrated that Nvmef2 is one of the key players driving post-gastrulation development. Two striking features of the development of Nvmef2 morphants are; (1) their endoderm forms normally during gastrulation but then loses integrity and the proliferation of the endodermal cells becomes suppressed; and (2) the pharynx fails to differentiate during planula development. The latter might be due to the lack of an inducing signal from the endoderm. A genome-wide analysis of Nvmef2 target genes will be necessary to explain the morphant phenotype.

The evolution of the mesoderm is one of the key developmental questions that can be addressed by comparing cnidarians with triploblasts. Surprisingly, diploblastic cnidarians have most of the important mesodermal determinants. The expression of these genes at the blastopore or in the endoderm of Nematostella argues for the common evolutionary origin of the endoderm and the mesoderm in triploblasts (Martindale et al., 2004; Technau and Scholz, 2003). Evidently, the presence of certain conserved transcription factors is not sufficient to induce a mesoderm, suggesting that the function of the individual genes and their gene regulatory network must have evolved. A functional diversification might be conveyed by novel cis-regulatory elements, by the addition of new protein-protein interfaces, or by alternative splicing combined with differential expression. The latter is thought to have played a major role in the diversification of gene function in vertebrates. Although a lower complexity of gene function might be expected in cnidarians on the basis of their simple morphology, our data show that, at least for Mef2, the involvement of alternative splice variants in the specification of neural cell types and the endomesoderm predates the cnidarian-bilaterian split.

We thank Michaela Schwaiger, Johannes Schinko and Patrick Steinmetz for critically reading the manuscript; Irene Lichtscheidl for confocal microscope facility support; Manuel Irimia (University of Barcelona) for help with splice site analysis; Anna Nyhaug and Trygve Knagg (Molecular Imaging Centre, University of Bergen) for help with transmission and scanning electron microscopy; Manfred Walzl and Stefan Jahnel for preparing semi-thin sections; and the staff of the animal facility.