Ancient and diverged TGF-β signaling components in Nasonia vitripennis

The transforming growth factor beta (TGF)-β signaling pathway and its modulators are involved in many aspects of cellular growth and differentiation in all metazoa. Although most of the core components of the pathway are highly conserved, many lineage-specific adaptations have been observed including changes regarding paralog number, presence and absence of modulators, and functional relevance for particular processes. In the parasitic jewel wasp Nasonia vitripennis, the bone morphogenetic proteins (BMPs), one of the major subgroups of the TGF-β superfamily, play a more fundamental role in dorsoventral (DV) patterning than in all other insects studied so far. However, Nasonia lacks the BMP antagonist Short gastrulation (Sog)/chordin, which is essential for polarizing the BMP gradient along the DV axis in most bilaterian animals. Here, we present a broad survey of TGF-β signaling in Nasonia with the aim to detect other lineage-specific peculiarities and to identify potential mechanisms, which explain how BMP-dependent DV pattering occurs in the early Nasonia embryo in the absence of Sog.


Introduction
The transforming growth factor beta (TGF-β) signaling pathway controls various processes throughout development, which include dorsoventral axis specification, cell proliferation, appendage formation, as well as the central nervous system patterning (Chang et al. 2002;Parker et al. 2004). It consists of the extracellular ligands and their modulators, which interact with each other to control ligand availability and distribution, of the receptors and of intracellular components, which transduce the signal to the nucleus and regulate target gene expression (Parker et al. 2004;Sieber et al. 2009) (Table 1).
All TGF-β ligands form dimers and bind to a heteromeric receptor complex of two type I and two type II serinethreonine kinase receptors (Sieber et al. 2009). In Drosophila, the BMPs as well as the Activins use Punt as a common type II receptor, while specificity is generated by the type I receptors. BMP ligands bind the type I receptors Orhan Özüak and Thomas Buchta contributed equally.
Upon ligand binding, the type I receptors become phosphorylated, which in turn phosphorylate and thereby activate receptor-regulated SMADs (R-SMADs). Tkv and Sax activate Mothers against Dpp (Mad), and Babo activates SMAD on X (Smox). Once phosphorylated, R-SMADs bind to the Co-SMAD Medea and form a complex that translocates into the nucleus and regulates target gene expression (Parker et al. 2004).
At the extracellular level, a large variety of modulators are involved in controlling ligand availability and distribution. A prominent group of modulators, which usually act as BMP antagonists are characterized by an array of conserved cysteine-rich domains that form cysteine knot structures (Walsh et al. 2010). On the basis of spacing of the cysteine residues within the cysteine ring, they have been grouped into several subgroups, e.g., the Noggins, the Chordin family, the Twisted gastrulation-like proteins (Crossveinless 1), the Crossveinless 2, and the Dan family (Walsh et al. 2010). In Drosophila, the main BMP antagonist in dorsoventral patterning is the Chordin homolog Short gastrulation (Sog), which interacts with Twisted gastrulation and is cleaved by the metalloprotease Tolloid (O'Connor et al. 2006).
Besides diffusible, secreted proteins, several membranebound and extracellular matrix proteins have been shown to influence the efficiency of TGF-β signaling (Ramel and Hill 2012). Thus, in the Drosophila wing, two glypicans, the GPI The NCBI accession number is given with the gene name. The numbers to the right of the Nasonia genes (gray column) indicate sources for phylogenetic analysis, expression studies, and/or functional (RNAi) data. 1, Özüak et al. 2014;2, this study;3, Nunes da Fonseca et al. 2010 anchored heparan-sulfate proteoglycans Dally and Dally-like, are required for efficient BMP signaling activity (Erickson 2011) while type IV collagens control the range of the BMP signaling gradient in the embryo (Ashe 2008). In our previous work, we showed that the parasitic jewel wasp Nasonia vitripennis uses the BMP pathway to pattern the dorsoventral (DV) axis despite the fact that no sog ortholog is present in the Nasonia genome (Özüak et al. 2014). The main goals of this work are to provide an overview of the TGF-β pathway in Nasonia and to identify components which might help to explain how the BMP signaling gradient is established during DV patterning in the wasp embryo.
Similar work was already done for the short germ beetle Tribolium castaneum, which revealed that the beetle retained a more ancestral complement of TGF-β signaling components compared to Drosophila (van der Zee et al. 2006; Van der Zee et al. 2008;Nunes da Fonseca et al. 2010).
Here, we comprehensively identify and describe components of the TGF-β signaling pathway in Nasonia. Interestingly, we found a case of parallel evolution, involving the duplication and divergence of the BMP 5/7 ligands in Nasonia and Drosophila. In addition, we identified the BMP ligand ADMP, which is not present in Drosophila and Tribolium, but plays an important role in vertebrates (Reversade and De Robertis 2005). As we were unable to find a sog homolog in Nasonia, we were especially interested to identify alternative BMP antagonists, which might be expressed at the ventral side of the embryo. However, we failed to identify such inhibitors corroborating our functional studies which indicated that the DV BMP gradient of Nasonia is not shaped by an opposing inhibitory gradient, but rather by diffusion from a dorsal source region (Özüak et al. 2014). Interestingly, the RNA of the type I receptor Tkv is localized to the dorsal midline of the developing Nasonia oocyte. In addition, one of the type II receptors is dorsally expressed in the early embryo. Based on these observations, we discuss a possible scenario of how the embryonic BMP gradient in Nasonia is established in the absence of a ventral inhibitor.

Embryo and ovary collection
All N. vitripennis embryos were collected using the Waspinator and fixed as described by (Buchta et al. 2013).
Nasonia ovaries were dissected and fixed as described by (Lynch et al. 2010a).

ISH
Single-and two-color in situ hybridizations were performed as previously described (Brent et al. 2003;Lynch et al. 2010a).
For list of primer used to produce probes, see Electronic supplementary material (ISH primer).
Identification and phylogenetic analyses of Nasonia orthologs of BMP pathway components Orthologs were identified by reciprocal best BLAST hits of the Drosophila genes of interest to the Nasonia genome or transcriptome assemblies (Werren et al. 2010).

Ligands
So far, seven TGF-β ligands are described in Drosophila and Tribolium. In the wasp Nasonia, we found eight potential TGF-β ligands in the genome. In our previous work, we described the important role of Nv-dpp in patterning the dorsal-ventral axis of the wasp embryo (Özüak et al. 2014). Knocking down Nv-dpp leads to the loss of dorsal fates and an almost complete ventralization of the embryo. In our phylogenetic analysis, Nvdpp groups perfectly with other insect Dpp sequences, as well as their vertebrate homologs Bmp2/4 (Fig. 1a). During oogenesis, Nv-dpp is expressed in the nurse cells and localized at the posterior pole of the oocyte (Fig. 2a). In the early embryo, Nvdpp is ubiquitously expressed and after gastrulation is completed, it is expressed in two stripes that flank the extraembryonic region on the dorsal side of the embryo (Fig. 1b, c).
Another important ligand for embryonic DV patterning in Drosophila is Screw, which represents a diverged paralog of Gbb (Wharton et al. 1991;Arora et al. 1994). Drosophila Gbb plays no role in DV patterning but is required in later developmental processes such as wing formation (O'Connor et al. 2006). We found two paralogs of Nv-Gbb, but unlike in Tribolium, where both Gbb paralogs are closer related to each other than to any other Gbb (Van der Zee et al. 2008), the Nasonia Gbb paralogs are not closely related. Instead, Nv-Gbb1 groups together with other insect Gbbs and the vertebrate homologs BMP5/7, while Nv-Gbb2 shows signs of a very fast evolving gene such as Drosophila screw. Both genes are ubiquitously expressed during oogenesis as well as during early embryogenesis (Fig. 1d, f; Fig. 2b, c). After gastrulation Nv-gbb1 is expressed like Nv-dpp in two stripes at the dorsal side (Fig. 1e), while Nv-gbb2 expression is gone. The knockdown of Nv-Gbb2 leads, like the knockdown of Nv-Dpp to a ventralization of the embryo (Özüak et al. 2014). Interestingly, the knockdown of Nv-Gbb1 has no embryonic phenotype, suggesting that it plays no role in early embryonic DV patterning. It might however play a role in later embryonic or in postembryonic stages, which we could not target by parental RNAi. Thus, the situation might be similar to Drosophila where one Bmp5/7 paralog (screw) was subject to fast evolutionary changes adapting to the special requirements of early embryonic DV patterning while the other one (gbb) retaining ancestral sequence features is required for later (more generic) developmental processes.
Anterior is left (accession numbers are given in Table 1 and Electronic supplementary material) Anti-dorsalizing morphogenetic protein (ADMP) has a BMP-like activity in Xenopus (Reversade and De Robertis 2005). Despite its important role in vertebrates during DV patterning, ADMP is not present in either Drosophila or Tribolium. However, it is present in the honeybee and we found an ortholog of ADMP in the wasp (Fig. 1a). It will be interesting to further investigate ADMP function in Nasonia to see if it also plays a role in self-regulation during embryonic DV patterning.
Orthologs of ligands belonging to the Activin subfamily such as act, daw, and myo are found in Nasonia. All three are ubiquitously expressed in the early embryo and later no expression is detectable (Fig. 1g-i). The last of the eight detected ligands is Maverick, which is not clearly grouped into one of the two large subfamilies. In Drosophila, Mav is required for growth in the wing disc. A knockdown of Mav leads to wing size reduction with normal vein patterns (Hevia and de Celis 2013). We investigated the expression pattern of Nv-mav and found that there is no expression in the early embryo, but later after completion of gastrulation, Nv-mav is expressed in two very narrow stripes flanking the extraembryonic region on the dorsal side (Fig. 1j, k).

Receptors
The extracellular BMP signal is transmitted via tetrameric receptor complexes formed by type I and type II receptors. Altogether, our analyses revealed six Nasonia orthologs of Drosophila serine-threonine kinase receptors: two BMP type I receptors, tkv and sax, and the Activin receptor type I ortholog babo, as well as type II receptors wishful thinking (wit) and two punt orthologs (Fig. 3a). Aside from these receptors, we also found BAMBI (not shown), which in vertebrates is known to be a pseudo-receptor (Onichtchouk et al. 1999) and might play a role in inhibiting BMP signaling in Nasonia as well. Fig. 2 BMP components in Nasonia ovaries. Expression of a dpp, b gbb1, c gbb2, d sax, e punt2, f punt1, and g-i tkv in lateral (g-h) and dorsal (i) views, mad1 (j), mad2 (k), and smox (l) in Nasonia ovaries (accession numbers are given in Table 1 and Electronic supplementary material) Nv-sax is ubiquitously expressed in the nurse cells and ubiquitously distributed in the oocyte (Fig. 2d). Nv-tkv is also expressed in all nurse cells. However, in the oocyte, it shows a striking pattern of localization. Besides an anterior accumulation, the RNA is found in a narrow stripe along the dorsal midline ( Fig. 2g-i). The dorsal side can be identified via the position of the oocyte nucleus. (data not shown, (Lynch et al. 2010b)). Thus, Nv-tkv RNA localization during oogenesis closely resembles that of the Nv-tgf-α, which is involved in establishing the DV axis in Nasonia (Lynch et al. 2010b). In early embryos, Nv-tkv is no longer localized and shows like Nv-sax ubiquitous expression (Fig. 3b, c). Knockdown of both receptors using pRNAi has been performed in previous studies and resulted in the case of Nv-sax, as well as Nv-tkv in a strong ventralization of the embryo (Özüak et al. 2014).
Similar to the type I receptors, Nv-babo shows a uniform expression pattern in early Nasonia embryos (Fig. 3d). The presence of a babo homolog in Nasonia is interesting since previous studies on Apis mellifera revealed a lack of this gene (Van der Zee et al. 2008), indicating that the loss of Activin signaling via Baboon is not a general feature of the hymenoptera.
Furthermore, two homologous genes of the type II receptor Punt are present. punt1 is not expressed during oogenesis (Fig. 2e). However, it shows a distinct expression pattern  Table 1 and Electronic supplementary material) along the dorsal midline in early blastoderm embryos (Fig. 3e), which strongly resembles the early expression patterns of previously described BMP signaling target genes in Nasonia (Buchta et al. 2013).
In contrast to this, punt2 is expressed during oogenesis and is localized at the posterior pole of the oocyte (Fig. 2f). This localization is not present in the embryo where punt2 messenger RNA (mRNA) is distributed ubiquitously (Fig. 3f). Because of its maternal distribution, we knocked down punt2. Enclosed females that were injected with punt2 doublestranded RNA (dsRNA) during yellow pupae stage laid no eggs and had degenerated ovaries (data not shown). This phenotype could be due to defects in Activin or BMP signaling as punt2 might be essential for both pathways.

SMADs
SMADs are characterized by the presence of two Mad homology (MH1 and MH2) domains. In addition, R-SMADs have a C-terminal SXS motif. Upon ligand binding, the receptors pass on the signal by phosphorylating both serines in the SXS motif thereby activating the R-SMADs (Feng and Derynck 2005). We identified four proteins with a MH1 and MH2 domain in the Nasonia genome, three of which contained a SXS motif (Fig. S1). Phylogenetic analysis revealed that two of the SXS motif containing proteins group together with Drosophila Mad and the third one with Drosophila Smox (Fig. 4a). It is worth mentioning that Nv-Mad2 is, similar to Nv-Gbb2 and Nv-Punt1, significantly more diverged in comparison to Nv-Mad1 and the other insect Mad orthologs. The fourth identified protein displayed clear homology to the Co-SMAD Medea (Fig. 4a). We could not identify any orthologs of known inhibitory SMADs like Daughters against dpp (Dad) (Tsuneizumi et al. 1997).
During oogenesis, Nv-mad1 is ubiquitously expressed in the nurse cells and oocytes while Nv-mad2 RNA accumulates at the posterior pole of the oocyte (Fig. 2j, k). Nv-mad1 does not seem to have a conserved role in embryogenesis since it did not show any phenotype after pRNAi mediated knockdown. However, embryos of females injected with Nv-mad2 dsRNA were severely ventralized as seen from the expansion of the ventral twist domain (Fig. 4c). This phenotype closely resembles the previously described knockdown phenotype of Nv-dpp and Nv-gbb2 (Özüak et al. 2014).  Table 1 and Electronic supplementary material) Nv-smox is ubiquitously expressed during oogenesis (Fig. 2l), and pRNAi-mediated knockdown resulted in sterile females (not shown) indicating an important role of Activin signaling during oogenesis. The Nv-smox phenotype suggests that the sterility caused by Nv-punt2 knockdown is due to interference with Activin, rather than BMP signaling.

TGF-β modulators
At the extracellular level, the TGF-β pathway is influenced in many ways to fine tune ligand and receptor activity. While antagonists bind to ligands and inhibit receptor interactions, cell surface proteins modulate the flux of the ligands (Ramel and Hill 2012). In our previous work, we provided evidence that Nasonia is lacking sog, which is the main BMP antagonist in Drosophila and other insects (Buchta et al. 2013). Sog belongs to the Chordin family, one of the subgroups of BMP antagonists. We searched for orthologs of other groups such as Noggin or members of the Dan family. While in Tribolium, two members of the Dan family, Dan and Gremlin, are present, no orthologs of members of the Dan family nor Noggin could be identified in the Nasonia genome using searches based on reciprocal best BLAST hits. However, like in Drosophila and Tribolium, we found an ortholog of Follistatin, which preferentially antagonizes Activin rather than BMP (Pentek et al. 2009). In Nasonia, follistatin is ubiquitously expressed during early embryogenesis and later after gastrulation, it is expressed in two narrow stripes that flank the extraembryonic region on the dorsal side of the embryo (Fig. 5a, e) Finally, we were able to identify four Crossveinless 2 like molecules (Cv2a-Cv2d) (Fig. 5q). Cv2 proteins consist of (1,000 replicates) in percentages. Amino acid substitution model: WAG + i + g. Nv Nasonia vitripennis, Am Apis mellifera, Dm Drosophila melanogaster, Tc Tribolium castaneum, Mm Mus musculus, Hs Homo sapiens, Gg Gallus gallus, Xl Xenopus laevis (accession numbers are given in Table 1 and Electronic supplementary material) similar domains as Sog/chordin proteins, but are thought to tightly interact with the type I receptors and/or membrane anchored proteoglycans and therefore lack diffusibility (Ambrosio et al. 2008). cv2c and cv2d show a uniform expression in early Nasonia blastoderm embryos and weak uniform expression during later embryonic development (Fig. 5c, d, g, h). cv2a is first expressed in small patches on the dorsal side at the cellularized blastoderm stage (Fig. 5b). During development, these patches widen toward the anterior and elongate toward posterior. At the onset of gastrulation, cv2A is expressed in two narrow stripes flanking the presumptive serosa (Fig. 5f). We were not able to amplify cv2b from ovarian or embryonic cDNA, indicating that this gene might be a pseudogene or is expressed only during postembryonic development.
Besides the antagonists, we found several orthologs of cell surface proteins that are candidates for modulating the flux of BMP ligands and facilitating long-range signaling activity. Among these are orthologs of Dally, Glypican4, Collagen IV, and Pentagone (Vuilleumier et al. 2010;Ramel and Hill 2012). In early blastoderm embryos, all are ubiquitously expressed ( Fig. 5i-l). Later, after gastrulation is completed, all of them show a distinct pattern. Nv-dally is expressed in a dotted pattern of two stripes flanking the ventral midline (Fig. 5m). Nv-glypican4 is expressed in segmental stripes on the ventral half of the embryo (Fig. 5n). Nv-collagen IV is expressed in small patches at the anterior and posterior of the embryo (Fig. 5o). The post-gastrulation expression pattern of Nv-pentagone in two ventral stripes resembles strongly that of Nv-ind, a neuroectodermal marker, which we described in our previous work (Fig. 5p) (Buchta et al. 2013).

Discussion
Taken together, this work indicates that Nasonia has retained some components of an ancestral TGF-β signaling network, such as ADMP and BAMBI, that have been lost in other insect lineages (Van der Zee et al. 2008). However, Nasonia also shows high degrees of divergence in some key signaling components with essential functions for embryonic DV patterning like Gbb2, Mad2, and Punt1. The most striking observation is that Nasonia apparently has lost some families of BMP antagonists including Sog which plays a crucial in DV pattering in most bilaterian animals (Mizutani and Bier 2008).
Although the inability to identify a gene within genome and transcriptome data represents only negative evidence, there are several observations supporting the claim that the BMP system in Nasonia operates without Sog: (1) Sog cannot be found in the closely related wasp Trichogramma (JAL personal observation); (2) In Apis, a sog gene has been identified, but it appears to lack expression in early blastoderm embryos when DV patterning is expected to take place (Wilson et al. 2014); and (3) Nv-Tolloid, the metalloprotease universally required to cleave Sog within Sog-BMP complexes in order to release active BMP, has only restricted anterior expression, and pRNAi does not result in DV pattering defects (Özüak et al. 2014).
Despite the absence of sog, three Nasonia orthologs of Twisted gastrulation/Crossveinless1 (Nv-CvA-CvC) were discovered in an earlier study (Nunes da Fonseca et al. 2010). In Drosophila, Tsg is thought to be a part of the Sog-Dpp complex. In this complex, Tsg can have a pro-BMP as well as an anti-BMP function. It was suggested that Tsg acts anti-BMP by enhancing the binding of Sog to Dpp and pro-BMP by enhancing the cleavage of Sog by the metalloprotease Tolloid (O'Connor et al. 2006). In our previous studies with Nasonia using Nv-tsg1 pRNAi, a pro-BMP role for Tsg during embryogenesis was observed since embryos of injected females were severely ventralized (Özüak et al. 2014).
However, our inability to identify sog, as well as the lack of DV patterning function of Tolloid, suggest that the pro-BMP function of Nv-Tsg is Sog-independent. An exclusively pro-BMP role for Tsg was also observed in Tribolium castaneum (Nunes da Fonseca et al. 2010). Interestingly, this pro-BMP function of Tc-Tsg could be shown to be Sog independent, too. Thus, both in Tribolium and Nasonia, Tsg plays a major, albeit Sog-independent role in DV patterning. In Drosophila, Tsg combines Sog-dependent and Sog-independent functions (Wang and Ferguson 2005) while in a more basal insect, we recently observed that Tsg predominantly acts in Sogdependent manner (unpublished observations). Together, these findings illustrate that the interactions of extracellular BMP modulators are subject to considerable evolutionary changes which probably reflect adaptations to the ways by which early BMP gradients from in the embryo.
The most important observation supporting the absence of a ventral BMP inhibitor like Sog in Nasonia is the phenotype caused by loss of Toll signaling (Özüak et al. 2014). In Nasonia, Toll is required to establish ventral-most cell fates, in particular, the mesoderm. However, Toll is not involved in patterning the dorsal half. Loss of function analysis shows that dorsal patterning depends on a BMP signaling gradient which apparently is establish in the absence of Toll, i.e., in the absence of an inhibitor which is ventrally activated by Toll signaling. As in all other insects studied so far, Sog is one of the key target genes activated by Toll signaling at the ventral side, this phenotypic analysis strongly corroborates our claim that the BMP gradient in Nasonia is not established through the formation an opposing inhibitor gradient.
Thus, Nasonia apparently establishes its DV axis in a bipolar fashion using independent signaling sources along the ventral and dorsal midline (Özüak et al. 2014). These signaling sources are likely to be established already during oogenesis. The oocytes of Nasonia like those of Apis, the other hymenopteran species studied so far, show an amazing ability to precisely localize mRNAs not only to the anterior and posterior poles (Fig. 2a, g) but also along the dorsal midline (Lynch et al. , 2010bWilson and Dearden 2013). Thus, in Nasonia and Apis, the RNA of TGF-α is localized in a narrow dorsal stripe and the secreted TGF-α ligand initiates EGF signaling at the dorsal side of the follicular epithelium along the entire egg length (Lynch et al. 2010b;Wilson et al. 2011). In Nasonia, functional studies show that EGF signaling negatively regulates the formation of eggshell cues required to localize a ventral source for embryonic DV patterning (Lynch et al. 2010b;Özüak et al. 2014). However, both hymenopteran species are also able to localize the RNA for BMP signaling components along the dorsal midline of the oocyte.
For Apis, dorsal localization has been observed for the RNA of the ligand Dpp (Wilson et al. 2011) while in Nasonia, as shown in this study, the RNA of the type I receptor Tkv is dorsally localized (Fig. 2g, h). If proteins produced by these localized RNAs are inserted into the membrane and remain there until egg deposition and early embryogenesis, they might become sources for establishing signaling gradients in the embryo. In this context, our finding is interesting that one of the type II receptors Nv-punt1 is expressed along the dorsal midline in blastoderm embryos (Fig. 3e). This pattern might result from a positive feedback mechanism by which BMP signaling initiates local Nv-punt1 expression that in turn potentiates BMP signaling. The broad BMP signaling gradient we detected by pMAD staining in early embryos (Özüak et al. 2014) might result from the maternal localization of signaling components in the oocyte such as Nv-tkv RNA. The later refinement of the gradient, which leads to high levels of BMP signaling along the dorsal midline, might reflect the positive feedback involving Nv-punt1. Further studies are required to investigate whether maternal receptor localization coupled to zygotic positive feedback can account for the longrang effects of BMP signaling in Nasonia which in contrast to Drosophila and Tribolium, are responsible for patterning the entire DV axis of the embryo (Özüak et al. 2014).