BMPs and Chordin regulate patterning of the directive axis in a sea anemone

Abstract

The TGF-β molecules Dpp/BMP2/4/7 and their antagonist Sog/Chd play a conserved role in establishing the dorso-ventral (DV) axis in bilaterians. Homologues of BMPs and the antagonist, Chordin, have been isolated from Cnidaria and show a striking asymmetric expression pattern with respect to the primary oral-aboral (OA) axis. We used Morpholino knockdowns of Nematostella dpp (bmp2/4), bmp5-8, chordin, and tolloid to investigate their function during early development of the sea anemone Nematostella vectensis. Molecular analysis of the BMP Morpholino phenotypes revealed an upregulated and radialized expression of bmps and chordin in ectoderm and endoderm indicating a negative feedback loop. Our data further suggest that BMP signaling is required for symmetry breaking of bmp and chordin expression during gastrulation. While bmps and chordin marker genes of the ectodermal OA axis extended aborally, other ectodermal markers of the OA axis were not significantly affected. By contrast, expression of other endodermal marker genes marking both the OA and the directive axis were abolished. Our data suggest that the logic of BMP2/4 signaling and the BMP antagonist, Chordin, differs significantly between Cnidaria and Bilateria, yet the double negative feedback loop detected in Nematostella bears systemic similarities with part of the regulatory network of the DV axis patterning system in amphibians.

In the last decade molecular data strongly supported the idea that the mechanisms that establish the dorso-ventral (DV) and anterior-posterior (AP) axes are largely conserved within bilaterians (1–4). In both, vertebrates and insects, opposing gradients of Bone Morphogenetic Protein (BMP) 2/4 and Decapentaplegic (Dpp)/BMP2/4, in conjunction with BMP7/Scw on one side and antagonists, such as Chordin (Chd)/Sog on the other side define the antineural and neural side, respectively (4, 5), and thus, the DV axis of insects is considered an inverted homolog to the DV axis of vertebrates (3). These data strongly suggested that the Urbilateria, the common ancestor of Deuterostomes and Protostomes, had a DV axis established by BMP signaling (3, 4, 6).

The Cnidaria are the undisputed sister group to the Bilateria (7, 8) and they have been traditionally regarded as radially symmetric animals. It is still widely assumed that bilaterality of animals evolved from radial symmetry of more basal metazoans (9). In that respect, Cnidaria may provide important insights into the evolution of bilaterality. However, anthozoan cnidarians have long been recognized to display internal asymmetries reminiscent of bilaterality defining a second axis, termed the directive axis, which is orthogonal to the primary oral-aboral axis. The directive axis is apparent by the slit-like shape of the pharynx, and the asymmetric position of the retractor muscles within the endodermal mesenteries (10). Recent molecular work identified the existence of BMPs in two anthozoans, the coral, Acropora millepora, and the sea anemone, Nematostella vectensis, and showed that the dpp/bmp2/4 homolog is expressed on one side of their directive axis during early embryonic development, indicating a molecularly defined second body axis (11, 12). Surprisingly, subsequent work showed that the BMP antagonist chd is expressed on the same side as bmps (13, 14) suggesting a different logic of BMP signaling in establishing the axes of Nematostella (13). Hence, the evolutionary relationship between the body axes of the Cnidaria and Bilateria is unclear.

To gain insight into the function of the BMP pathway during the development of Nematostella we knocked down dpp, chd, and bmp5–8 using Morpholino oligonucleotides. Our data suggest that BMP signaling is required for the symmetry breaking along the directive axis and indicate a negative control of bmp and chordin expression by BMP signaling. Marker gene expression and analysis of cellular features suggest that BMP signaling has a crucial role in endoderm patterning and differentiation.

Results

The Knockdown of chd, dpp, and bmp5–8 Does Not Inhibit Gastrulation.

Bmps (i.e., homologs of bmp2/4 and of bmp7) and chd start being expressed at the onset of gastrulation in a broad radial pattern circumferential of the blastopore (13, 15). At midgastrula stage, the expression of the genes appears to undergo a symmetry break and they are both expressed on one side of the blastopore (13). We sought to get insight into the function of BMP signaling in Nematostella by Morpholino-mediated gene knockdown (16). Control experiments confirmed the specificity of the Morpholinos used in this article (Fig. S1 and Table S1). Morpholino-injected embryos developed normally through early cleavage and blastula stage. Gastrulation also occurred at the same time as in the controls (17–24 h post fertilization) suggesting that BMP signaling is not required for gastrulation and early germ layer demarcation.

Interestingly, at midgastrula stage, when dpp and chd expression have already shifted toward one side of the blastopore (Fig. 1A), dpp knockdown by a dpp MO (but no other morpholino) (Figs. S1 and S2) led to a radialization and ectopic extension of dpp toward the aboral pole (Fig. 1B), while early gata expression in the ectoderm is unaffected (Fig. S3). Unlike in control embryos, Chd expression appears not shifted toward the aboral end but is radialized upon Dpp MO injection (Fig. 1 C and D). Conversely, treatment of embryos with recombinant human BMP2 protein (hBMP2) in control or Dpp MO injected embryos abolishes chordin expression (Fig. 1 E and F, Fig. S4) suggesting that BMP signaling negatively regulates both dpp and chordin expression in a feedback loop. Since the expression of dpp and chordin remain radial from the earliest time points upon Dpp knockdown, we conclude that a negative feedback of BMP signaling is responsible for the symmetry break of dpp and chordin expression.

Fig. 1.

Dpp signaling regulates symmetry break of dpp and chordin expression during gastrulation. In gastrula stage embryos (24 h) dpp knockdown leads to a radialized expression of dpp (B) and chd (D), both of which are asymmetrically expressed on the same side in controls (A and C), whereas treatment with recombinant human BMP2 protein virtually abolishes chordin expression showing that chordin depends on BMP signaling (E and F). Asterisk marks the oral end. (Scale bar, 100 μm.)

Knockdown of dpp, bmp5–8, and chd Impairs Marker Gene Expression of both OA and Directive Axis.

At the postgastrula stage, the planula larva, the main morphological and molecular body plan features are laid down: the planula shows two adjacent epithelial layers, a well defined pharynx, an apical organ and staggered gene expression along both axes (12–17). On the cellular level, differentiated neurons, nematocytes, and gland cells become apparent. Despite a number of genes that are expressed asymmetrically along the directive axis at this stage, no morphological signs of a second axis are visible (12–15). We therefore wished to know, whether knockdowns of the dpp, bmp5–8, and chd have an effect on the expression of marker genes of both axes at the planula stage (52 hpf). We found that after dpp, bmp5–8, and double knockdown of both, the expression of dpp, bmp5–8, as well as chd remained radialized as in the gastrula stage, yet at a much higher level of expression (Fig. 2 A–D, F–I, and K–N, and Fig. S1C). In addition, dpp and bmp5–8 show an ectodermal expression on the oral side that extends aborally compared with the controls. Furthermore, expression of dpp and bmp5–8 is visible on the aboral side of the planula in the apical organ region. This indicates an alteration of the molecular features of the directive but also of the OA axis (Fig. 2 B–D, G–I, and L–N). By comparison, chd remains expressed radially in the oral ectoderm, but does not spread into the endoderm (Fig. 2 K–N). Conversely, treatment of control or Dpp Mo injected embryos with recombinant hBMP2 strongly diminished or abolished chd expression (Fig. 1 E–F and Fig. S4). In contrast to the ectopic radial expression of dpp, bmp5–8, and chd, expression of gdf-5-like, hoxE and gbx, which are normally expressed on the opposite or lateral side of dpp and chd in the endoderm, is completely abolished upon dpp knockdown (Fig. 2 P–S, U–X, and Z–C′). This suggests that gdf5-like, hoxE, and gbx are regulated by BMP signaling (Fig. 2A′–C′). When chd was knocked down, expression of all of the genes mentioned was abolished, except for chd itself, which remained weakly expressed in a radial ring around the former blastopore (Fig. 2E, J, O, T, Y, and D′). In Bilateria, Chordin can be cleaved by the metalloprotease Tolloid. A homolog of tolloid is expressed in the whole endoderm of Nematostella embryos (15). We knocked down tolloid and found that the asymmetric endodermal markers dpp, hoxE, and gdf5-like are abolished, while the asymmetric expression of chd is diminished (Fig. S5). The effects in the endoderm are reminiscent of the effect of the chordin and Dpp MO, suggesting that Tolloid acts in concert with Chordin and Dpp in the endoderm.

Fig. 2.

Knockdown of dpp, bmp5–8, and chd affects all asymmetrically expressed endodermal genes. Controls show the typical asymmetric expression of dpp (A), bmp5–8 (F), chd (K) along one side of the planula and gdf5-like (P) and hoxE (U) on the opposite side. (Z) Gbx is expressed laterally between dpp and gdf5-like in two broad domains. Note that the specimen in Z is oriented 90° rotated to show the two “lateral” expression domains of gbx. Dpp and bmp5–8 knockdowns, as well as the double knockdown, result in a radialized expression of dpp, bmp5–8 and chd. Furthermore; a shift of the ectodermal dpp and bmp5–8 expression toward the aboral pole is visible in these knockdowns (B–D, G–I, L–N). In contrast, expression of gdf5-like, hoxE, and gbx is abolished (Q–S, V–X, A′–C′). Chd knockdowns show an abolishment of gene expression of all genes tested (E, J, T, Y, D′) except for chd itself (O), where the expression is radialized but weaker than normal. All specimens shown are 52-h post fertilization planula larva. (Scale bar, 100 μM.)

The aborally expanded expression of dpp and bmp5–8 in the dpp and bmp5–8 morphants was a first indication that aside from the directive axis, the OA axis is also affected by the knockdowns. To further investigate this, we analyzed the effect of dpp, bmp5–8, and chd knockdown on several genes marking the OA axis. Among the ectodermally expressed genes, three genes mark the pharynx in various subdomains: fgf8A is expressed at the outer ectodermal margin of the pharynx (Fig. 3A) (18), soxB1 is expressed ectodermally in two domains, in the pharynx/blastopore and in the apical organ region (Fig. 3F) (19), and nk2.1 is expressed in the inner part of the pharynx (Fig. 3K). Interestingly, only the outermost expression by fgf8 is maintained in dpp, bmp5–8, or chd morphants, while both the pharynx expression of soxB1 and nk2.1 is abolished (Fig. 3 A–O). More aborally, neither the circumferential expression of wnt2 (17), (Fig. 3P), nor the apical organ marker fgf1A (18) is affected in dpp or chordin morphants (Fig. 3 Q–T and V–Y). Likewise, the expression of soxB1 in the aboral apical organ is not altered in dpp, bmp5–8, and chd knockdowns (Fig. 3 G–J).

Fig. 3.

BMP pathway inhibition results in differential phenotypes of the OA axis in ectoderm and endoderm. (A–I′) Expression of marker genes of the OA axis in controls, dpp MO, bmp5–8 MO, double dpp/BMP5–8 MO, and chd MO. (A–E) Outer pharynx expression of fgf8A is unaltered in morphants. (F–J) SoxB1 expression in the pharynx but not in the aboral pole is abolished in Morpholino-injected animals. (K–O) Expression of nk2.1 in the inner pharynx is abolished in all Morpholino-injected animals. (P–T) Wnt2 expression and (U–Y) fgf1A expression are unaltered. (Z–D′) Twist and (E′–I′) endodermal otxC expression along the OA axis are abolished in Morpholino-injected embryos. Note that otxC is expressed in two domains, oral ectodermal, and aboral endodermal (E′). Its endodermal expression is abolished after bmp5-8, dpp-bmp5-8-double, and chd knockdown (G′–I′) but remains expressed weakly in the oral region and the expression shifts to the aboral ectoderm after dpp knockdown (F′). All specimens shown are 52 h post fertilization, except in Z–D′ which are 76 hpf. (Scale bar 100 μM.)

In sharp contrast, two endodermal marker genes of the oral-aboral axis are severely affected in all Morpholino injected embryos: both, the oral endoderm marker twist (20), as well as the aboral endodermal expression of otxC (21) are completely abolished upon knockdown of dpp, bmp5–8, or chd (Fig. 3 Z–I′). Together, these results show that of ectodermal OA axis marker genes, only those expressed in the inner pharynx at the oral end are affected. By contrast, in the endoderm, both markers of the directive and the OA axis are strongly affected. In summary, knockdown of dpp, bmp5–8, and chd mainly affects the expression of endodermal genes, while ectodermally expressed genes of the OA axis are only affected at the oral end.

Chd Overexpression Leads to Ectopic dpp and bmp5–8 Expression in the Endoderm.

As shown above, knockdown of bmps and of chd has severe effects on gene expression and differentiation of the endoderm. Since the chd morpholino led to an abolishment of dpp and bmp5–8 expression we were interested to know whether an ectopic overexpression of this gene would lead to an ectopic expression of dpp and bmp5–8. To test this we transiently overexpressed Nematostella chd under the control of a 1-kb fragment of the Nematostella actin promoter. Ectopic overexpression of chd caused an ectopic upregulation of dpp (Fig. 4D) and bmp5-8 (Fig. 4F) in the endoderm. The Mock controls showed no sign of altered gene expression (Fig. 4 A, C, and E). This suggests that Chordin is involved in controlling a negative feedback loop of bmp expression.

Fig. 4.

Ectopic overexpression of chd leads to misexpression of dpp and bmp5–8. Animals injected with a construct carrying the chd gene under the control of a 1 kb actin upstream promoter show transient ectopic expression of chd at 52 hpf (B). Controls expressing mCherry under the same promoter show no altered expression of chd (A). The expression of dpp and bmp5–8 is altered only in animals injected with the chd plasmid (D and F), whereas the control plasmid did not lead to an alteration in expression (C and E). All specimens shown are 52 h post fertilization.

The Role of BMP Signaling in Neuronal Differentiation.

One of the unifying features of BMP signaling in the establishment of the bilaterian DV axis is its antineural effect, defining the position of the central nervous system. To test, whether the conserved role of BMP signaling on the neuronal differentiation extends to Cnidaria, we analyzed two available marker genes, gata and fmrf peptide. We found that in Dpp morphants, the initiation of gata and fmrf peptide in single proneural cells in the ectoderm is not affected, yet expression is lost at planula stage (Fig. S3). Thus, we consider BMP signaling dispensable for the initiation of gata and fmrf peptide expression in the ectoderm; however, it might have a secondary role in its maintenance.

Discussion

The Roles of DPP, BMP, and Chd in Gastrulation and the Formation of the Planula Body Plan.

The knockdowns had no obvious effect on the early development including gastrulation. We conclude that bmp5–8, dpp, and chd do not play a decisive role for the process of gastrulation, or for the early demarcation of germ layers, as was suggested before (15). Our data are in line with observations made in vertebrates, where the mutant phenotypes are often severe and lethal but gastrulation is largely unaffected (22, 23). We share the view, however, that these molecules are necessary for patterning and differentiation of the endoderm and the body axes (12, 14). None of the Morpholino-injected embryos succeeded in metamorphosing into a primary polyp and the morphants eventually died at ≈4–6 days of development. This suggests that proper differentiation of the endoderm during planula stage is required to allow the next developmental step to occur. In line with this, our data also suggest that signals from a properly differentiated endoderm appear also to be required to maintain the expression of neuronal genes in the ectoderm as well as the pharynx. A primary role for BMP signaling in endoderm in cnidarians is supported by data from Hydra where all BMP signaling pathway members are expressed in the endoderm (24–26). Chordin also is affecting the differentiation of the endoderm. A role for Chordin in induction of dorsal endoderm has previously been found in Xenopus (27).

BMP Signaling Is Essential to Establish the Body Axes and Pattern the Endoderm in Nematostella.

dpp/bmp5–8 knockdown led to radial upregulated expression of dpp, bmp5–8, and chordin. In the endoderm, marker genes of both the OA, as well as of the directive axis are strongly affected: dpp and bmp5–8 expand and are expressed throughout the whole endoderm, whereas all other marker genes are abolished. This strongly suggests that BMP signaling in conjunction with Chd establishes the asymmetry defining the directive axis in an autoregulatory way, but also affects the OA axis. Strikingly, such a role of BMP signaling in two body axes is also found in the flour beetle, Tribolium, where loss of the BMP antagonist short gastrulation (sog) not only affects the dorsoventral axis as in Drosophila, but also the formation of the head (28). This is reminiscent to the situation in vertebrates, where the BMP antagonists in the organizer are also crucial for head formation (29–31). Thus, the role of BMP signaling in organizing body axes may actually be ancestral within Eumetazoans. Marker gene expression further revealed that BMP signaling is required for proper differentiation of the directive and OA axis in the endoderm. In line with a primary role for BMP signaling in the endoderm, smad1, smad4, and tolloid expression are restricted to the endoderm (15) and knockdown of tolloid confirms a role for Tolloid in the endoderm (Fig. S5).

Is Chd an Antagonist or an Agonist of BMP Signaling?

Functional tests in zebrafish have shown that Nematostella chd can dorsalize the zebrafish embryo, while Nematostella dpp can ventralize the embryo (13). This shows that the Nematostella proteins are sufficiently conserved on a structural level to mimic the function of the endogenous proteins in vertebrates, suggesting that the proteins might display similar molecular functions in Nematostella. However, while the early symmetry break can be explained by a negative feedback loop of BMP controlled by an antagonistic Chd, later events after gastrulation can be better reconciled with an agonistic role of Chd. Interestingly, in vertebrates both Chd, as well as Crossveinless2 (CV2) have both anti- and pro-BMP functions depending on the concentrations and the cellular context (32, 33). Ternary complexes between Chd, Tsg, CV2, and BMP2/4/7 are modulated by Tolloid metalloproteinases to stabilize and steepen the BMP signaling gradient (33, 34).

In Drosophila, as well as in Tribolium, Short gastrulation (Sog), has been shown to facilitate BMP signaling at a distance by transporting and releasing BMP at more dorsal positions (28, 34–37). Recent work in Xenopus has stressed the idea that shuttling of the BMP ligands by Chd and balancing effects of dorsally expressed ADMP and ventrally expressed antagonists (e.g., CV2) and modulators (ONT1) are key for robust patterning and the ability of scaling (38, 39). Thus, it is conceivable that Chd might also function in shuttling Dpp in Nematostella to exert its effect along the directive and OA axis. Further, ectopic chd expression leads to an ectopic dpp and bmp5–8 expression providing further support to the idea that Chd and Dpp/BMP5–8 may also work synergistically. Like in Bilateria, Dpp/BMP5–8 could be transported in complex with Chd to the endoderm where it is released by the cleavage of Chordin by Tolloid to activate Smad signaling. In line with this, Tolloid acts in the endoderm in conjunction with Chordin and BMP in Nematostella.

Evolutionary Considerations.

The asymmetric expression of dpp, bmp5–8, chordin and other components of the BMP signaling network perpendicular to the oral-aboral axis in a “radially symmetric” animal has prompted suggestions about the homology of the bilaterian DV axis and the cnidarian directive axis (12, 14). Yet, our functional analysis suggests that the wiring of BMP signaling in Nematostella appear to follow a very different logic than in vertebrates: In Xenopus for instance, BMP signaling and Chordin form opposing gradients along the DV axis, due to the combination of a positive autocatalytic loop of BMP signaling and the antagonistic action of Chordin from the dorsal organizer (Fig. 5A). In Nematostella, the data suggest a double negative feedback of BMP signaling: BMP signaling suppresses dpp and bmp5–8 expression, as well as chordin expression. Chordin protein binds to DPP and prevents it from binding to the BMP receptors. However, this seemingly drastically different interaction loop is strikingly similar to the regulatory loop of Chordin and ADMP in the Spemann organizer of amphibians (40; Fig. 5A). ADMP is a ventralizing BMP signaling molecule coexpressed with chordin in the organizer (Fig. 5A).

Fig. 5.

Model for the functional relationships of BMP5–8, Dpp and Chd. (A) Molecular interactions of BMPs and BMP modulators in the Xenopus embryo. Note that on both, the ventral and the dorsal side, pairs of BMPs and BMP antagonists are expressed. The green square highlights the double negative feedback loop of ADMP and Chordin. Scheme modified from ref. 40. (B) Molecular interactions of Chordin and the BMP2/4 homolog DPP (as well as BMP5–8) in the Nematostella embryo. Note that the double negative feedback loop is reminiscent of the molecular interactions in the Spemann organizer (green square).

Notably, another BMP-like molecule, gdf5-like, as well as a BMP antagonist, gremlinA, are expressed on the opposite side (22), just like bmp2/4 and crossveinless2, a BMP binding secreted antagonist, at the ventral organizer in Xenopus (40). Thus, in both Nematostella and Xenopus, at least two BMPs and two BMP antagonists are expressed on either side. Similarly, in zebrafish, the BMP antagonists Follistatin and Noggin are also excluded from the organizer (41). It is therefore conceivable that the same principle of opposing but balanced action of BMPs and shuttling by their antagonists (39) ensures the establishment of a body axis perpendicular to the main body axis.

Another hallmark of the conserved DV axis in Bilateria is the localized CNS defined by Chordin. However, neither does any cnidarian have an indication of an asymmetrically located central nervous system, nor did we observe any drastic effect of BMP loss or gain-of-function on the expression of neuronal markers. We therefore favor the idea that BMP signaling and its modulation through BMP antagonists to create axial asymmetries predates the split of Cnidaria and Bilateria, but the role of opposing gradients of BMPs and their antagonists in establishing a dorso-ventral body axis by defining a central nervous system appears to have evolved only in the common ancestor of the Bilateria (42). Interestingly, in the hemichordate Saccoglossus expression and function of BMPs is also uncoupled of neural development, despite otherwise conserved molecular DV patterning (43) although this is likely derived from a state with a central nervous system (44). Obviously, the molecular system of BMP signaling and control through BMP antagonists is very ancient. Since such a system is simply very well suited to create asymmetries it is therefore plausible that it might have been adapted to various contexts, yet the systemic properties have been conserved.

Materials and Methods

Animal Culture, Embryo Manipulation, and Microinjection.

Animals were cultured and gametes were obtained as described in ref. 45. Microinjections of zygotes were carried out using an Eppendorf Femtojet. Control injections were carried out in each egg-clutch. Injection volumes were approximately 3% of the egg volume. The mixture used for injection was 300 μM Morpholino (in double injections of 300 μM each); 1 mM Tris-HCl, pH 8.0; 0.1 mg/mL Alexa568 Dextran, (Molecular Probes) in autoclaved MiliQ H₂O. After injections, the embryos were transferred in autoclaved Nematostella-medium and let develop at room temperature (22 °C).

Morpholinos.

Antisense morpholino oligonucleotides (GeneTools, LLC) were designed to match the sequence of the Nematostella genes: BMP5–8-MO: gtaacaggtctcgtattctccgcat, CHD-spMO: gatccactcaccatctttgcgagac, DPP-MO: gtaagaaacagcgtaagagaagcat, TLD-spMO: aacagctggcgaggaaaagtgttag. Presence of unspecific binding sites was excluded by searching the Nematostella genome sequence (46) with the Morpholino sequences by BlastN. Binding specificity and efficiency was tested in vitro using a TnT kit (Promega) expressing either the wild-type dpp, or a mismatch dpp with four silent point mutations in the MO binding site (Fig. S1). Chordin was knocked down either with an _ATGMO or a splice MO (spMO) targeting the first intron (Fig. S1) with essentially the same phenotypes. The functionality of the CHD-spMO (splice Morpholino) was tested via PCR with cDNA prepared from RNA from 52 h larvae (Fig. S1).

Treatment with Recombinant Human BMP2.

Embryos were incubated from early cleavage until 52 h of development in recombinant human BMP2 protein (Cell Signaling) 0.5 μg/mL in Nematostella medium/0.1% BSA. The protein solution was replaced after 24 h.

In Situ Hybridization.

In situ hybridization was carried out as described before (16, 20) using 0.8–2 kb DIG-labeled antisense probes. For all experiments embryos were fixed at 24 or 52 hpf except for animals intended to use for in situ hybridization with the twist probe were fixed 76 hpf because twist expression is not detectable before this time. Fixations were performed in ice-cold 4% MEMPFA for 1 h at 4 °C.