Developmental Biology 410 (2016) 108–118

Contents lists available at ScienceDirect

Developmental Biology

http://d0012-16

n Corrmia Sin

E-m

journal homepage: www.elsevier.com/locate/developmentalbiology

Evolution of Developmental Control Mechanisms

Regulatory circuit rewiring and functional divergence of the duplicateadmp genes in dorsoventral axial patterning

Yi-Cheng Chang a,b, Chih-Yu Pai a, Yi-Chih Chen a, Hsiu-Chi Ting a, Pedro Martinez c,d,Maximilian J. Telford e, Jr-Kai Yu a, Yi-Hsien Su a,b,n

a Institute of Cellular and Organismic Biology, Academia Sinica, Taipei, Taiwanb Department of Bioscience and Biotechnology, National Taiwan Ocean University, Keelung, Taiwanc Department de Genètica, Universitat de Barcelona, Barcelona, Spaind Institució de Recerca I Estudis Avançats (ICREA), Barcelona, Spaine Department of Genetics, Evolution and Environment, University College London, London, UK

a r t i c l e i n f o

Article history:Received 18 December 2015Accepted 18 December 2015Available online 21 December 2015

Keywords:BMPAdmpSea urchinAcorn wormAmbulacraria

x.doi.org/10.1016/j.ydbio.2015.12.01506/& 2015 Elsevier Inc. All rights reserved.

esponding author at: Institute of Cellular andica, Taipei, Taiwan.ail address: yhsu@gate.sinica.edu.tw (Y.-H. Su

a b s t r a c t

The spatially opposed expression of Antidorsalizing morphogenetic protein (Admp) and BMP signalscontrols dorsoventral (DV) polarity across Bilateria and hence represents an ancient regulatory circuit.Here, we show that in addition to the conserved admp1 that constitutes the ancient circuit, a secondadmp gene (admp2) is present in Ambulacraria (EchinodermataþHemichordata) and two marine wormsbelonging to Xenoturbellida and Acoelomorpha. The phylogenetic distribution implies that the two admpgenes were duplicated in the Bilaterian common ancestor and admp2 was subsequently lost in chordatesand protostomes. We show that the ambulacrarian admp1 and admp2 are under opposite transcriptionalcontrol by BMP signals and knockdown of Admps in sea urchins impaired their DV polarity. Over-ex-pression of either Admps reinforced BMP signaling but resulted in different phenotypes in the sea urchinembryo. Our study provides an excellent example of signaling circuit rewiring and protein functionalchanges after gene duplications.

& 2015 Elsevier Inc. All rights reserved.

1. Introduction

The DV polarity of bilaterians is established by a BMP gradientgenerated through interactions between BMP ligands, includingBMP and Admp, and their antagonists, such as chordin and noggin(De Robertis, 2008). The BMP/Admp regulatory circuit, in whichadmp expression is repressed by BMP signaling and the Admpprotein generated on the low BMP side further promotes BMPsignaling, has been suggested to be a conserved feature used bybilaterians to establish or regenerate their DV axes (De Robertis,2008; Gavino and Reddien, 2011; Reversade and De Robertis,2005; Srivastava et al., 2014).

The Echinodermata and Hemichordata form a clade calledAmbulacraria (Cannon et al., 2014), and together with Chordataconstitute the three major phyla in Deuterostomes. In echino-derms such as sea urchins, although their adults are pentaradiallysymmetric, they nevertheless develop initially as a bilaterallysymmetric gastrula, in which the DV axis is patterned by BMP

Organismic Biology, Acade-

).

signals as in other bilaterians (Lapraz et al., 2009; McClay, 2011).The sea urchin bmp2/4 gene, one of the major contributors of theBMP gradient, is transcriptionally activated on the ventral side ofthe blastula by Nodal signals (Ben-Tabou de-Leon et al., 2013;Duboc et al., 2004; Saudemont et al., 2010; Su, 2009). Intriguingly,both bmp2/4 and chordin genes are under the control of Nodalsignaling and thus are co-expressed in the ventral ectoderm of seaurchin embryos (Bradham et al., 2009; Lapraz et al., 2009; Su et al.,2009). A BMP activity gradient nevertheless forms that is high onthe dorsal side of the embryo, opposite to the co-expression do-main of bmp2/4 and chordin (Chen et al., 2011; Lapraz et al., 2009).In the hemichordate acorn worm embryos, bmp2/4 is expressed onthe dorsal side, opposite to the expression domain of chordin onthe ventral side, and together they also generate a BMP gradientthat is high on the dorsal side and control the DV polarity of theembryos (Lowe et al., 2006; Rottinger et al., 2015; Rottinger andMartindale, 2011).

The sea urchin genome contains two admp genes, admp1 andadmp2 (Lapraz et al., 2006). The sea urchin admp2 has been shownto be expressed in the dorsal ectoderm with BMP signals requiredfor its expression (Saudemont et al., 2010). Similarly, an admp2 geneidentified from a hemichordate acorn worm is expressed in thedorsal ectoderm, where high BMP activity is observed (Rottinger

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118 109

et al., 2015). The expression of the ambulacrarian admp genes in thedomain with high BMP activity is completely opposite to what havebeen observed in other bilaterians that have been studied, in whichadmp genes are always expressed on the low BMP side due to therepression of admp transcription by BMP signals. These resultsbrought doubts on whether the conserved BMP/Admp regulatorycircuit is altered in Ambulacraria.

In this study, we aim to investigate the evolutionary origin andtranscriptional regulation of the two admp genes, with a furtheranalysis of the functions of their gene products during sea urchin

Fig. 1. Phylogenetic tree of Admp proteins and other TGFβ family members. Protein sequunder a Maximum likelihood analysis. Each TGFβ subfamily forms a distinct group (colorAdmp2 sequences of ambulacrarians and xenacoelomorphs form another distinct groupsister to all bilaterian Admp sequences. No admp genes were found in the ctenophore gennames and accession numbers corresponding to sequence names in the tree are given inper position.

embryogenesis. We survey the presence of admp genes in variousanimal genomes/transcriptomes and reveal their possible evolu-tionary history. We discover that the two admp genes, in both seaurchin and acorn worm embryos, are under opposite transcrip-tional control by BMP signals. Both admp genes are required forestablishing DV polarity in the sea urchin embryo, although thetwo Admp proteins have distinguishable functions when over-expressed. We also discuss the transcriptional rewiring of theadmp genes and the functional divergence of their gene productsafter gene duplications.

ences were aligned and trimmed, and the tree was constructed using the LG modeled boxes). The Admp1 sequences of bilaterians form a group (purple box) while the(pink box). Cnidarian Admp-like proteins (green box) form a group, which is theome, although members of the TGFβ family were identified (red arrow). The speciesTable S1. The level of bootstrap support is indicated. Scale bar shows substitutions

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118110

2. Materials and methods

2.1. Phylogenetic analyses

The sequences were gathered from the National Centre forBiotechnology Information databases, various genomes, EST/tran-scriptome databases, or obtained by PCR cloning. The accessionnumbers or gene IDs of the sequences used in our analyses arelisted in Table S1. The protein sequences were aligned usingClustalW in Mega6 (Tamura et al., 2013), and the ambiguouslyaligned regions were detected and removed with Gblocks (Cas-tresana, 2000) to generate the tree in Fig. 1. The maximum like-lihood analysis was performed using the RAxML Black Box pro-gram with the LG model provided by CIPRES Science Gateway(http://www.phylo.org/) with 1000 bootstrap replicates. Bootstrapvalues higher than 50% are shown.

2.2. Molecular cloning and QPCR analysis

To amplify the sea urchin and acorn worm admp1 and admp2from embryonic cDNA, PCR primers were designed based on thesea urchin genome database (Cameron et al., 2009) and the acornworm transcriptome database (Chen et al., 2014), respectively. Theamphioxus (Branchiostoma floridae) admp cDNA clone was identi-fied from the EST database (Yu et al., 2008). The coding sequencesof admp genes were cloned into the pCS2þ vector and in vitrotranscribed using an mMessage Machine kit (Ambion, Austin, TX,USA). QPCR analysis was performed as previously described (Chenet al., 2011) and the primer sequences are listed in Table S2.

2.3. Embryo preparation and manipulation

Sea urchin (Strongylocentrotus purpuratus), acorn worms (Pty-chodera flava), and zebrafish adults were obtained, spawned, andfertilized as described (Chen et al., 2011; Ikuta et al., 2013; Tsaiet al., 2014; Lin et al., 2016). The sea urchin and zebrafish embryoswere microinjected following published procedures (Ben-Taboude-Leon et al., 2013; Tsai et al., 2014). The sequences of MOs usedto inject sea urchin embryos are listed in Table S3. To perturb BMPsignaling, recombinant BMP4 protein (R&D Systems) or DMH1(Sigma) were added to the embryo cultures after fertilization, andthe same volumes of solvents, 0.1% BSA or DMSO, respectively,were added as controls. The in situ hybridization and im-munohistochemistry were performed, and the embryos were im-aged as described (Chen et al., 2011; Ikuta et al., 2013). All imageswere generated on a Zeiss Axio Imager A1 or a Leica TCS-SP5 APBSinverted confocal system.

3. Results

3.1. Ambulacrarians have two admp genes

Two admp genes, admp1 and admp2, have been annotated inthe sea urchin genome (Lapraz et al., 2006). We searched thegenomes and transcriptomes of a wide range of animals for genesand transcripts encoding Admp-like proteins and performed aphylogenetic analysis of these proteins together with other TGFβfamily members (Fig. 1 and Table S1). Admp orthologs could not beidentified in the ctenophore, sponge, and Trichoplax genomes,suggesting that admp genes have arisen in the common ancestor ofCnidaria and Bilateria: the cnidarian Admp-related proteins areplaced at the base of all bilaterian Admps.

We have been able to find a single admp gene in most bilaterianand cnidarian genomes that we have analyzed; the exceptions arethe sea urchins, hemichordates, the xenoturbellid Xenoturbella

bocki, the acoel Symsagittifera roscoffensis, and the vertebratesXenopus and chick, all of which contain two admp genes. Theambulacrarian Admp1s group with other bilaterian Admps, andthe orthology is corroborated by the conserved linkage with thepinhead gene, as it occurs in various bilaterian genomes in which apinhead gene is found (Fig. S1). This linkage has been shown to beresponsible for controlling the mutually exclusive expression ofadmp and pinhead in an ascidian embryo (Imai et al., 2012). Inaddition to the Admp1 group, the Admp2s of the sea urchins,hemichordates, Xenoturbella, and the acoel form a distinct groupwith a high support value (Fig. 1). The early branching position ofthe single cnidarian Admp-related protein prior to the separationof the bilaterian Admp1 and Admp2 genes suggests that the genesencoding Admp1 and Admp2 arose from a gene duplication eventafter the divergence of Cnidaria and Bilateria. Because the ambu-lacrarian Admp2s group independently from all the Admp1s, wepropose that the duplication occurred in the common ancestor ofBilateria and admp2 genes were subsequently lost in the proto-stome and chordate lineages. Xenopus and chick genomes alsocontain two admp genes (Kumano et al., 2006). The phylogeneticanalysis revealed that the vertebrate Admp2 proteins are notclosely related to the ambulacrarian Admp2s, but are within theAdmp1 group along with their paralogs, supporting their origin ina specific duplication event that took place in the vertebratelineage (Fig. S2).

3.2. Expression of the two admp genes on the opposite sides alongthe DV axis

To investigate the roles of the admp1 and admp2 genes inAmbulacraria, we first analyzed their expression patterns duringembryogenesis of the sea urchin Strongylocentrotus purpuratus andthe acorn worm Ptychodera flava. Quantitative RT-PCR (QPCR)analysis of the sea urchin admp1 revealed that the transcript wasfirst detected at 26 h post fertilization (hpf) and the expressionlevel continued to increase (Fig. S3a). Consistent with the QPCRdata, the admp1 transcript was not observed at the mesenchymeblastula stage (24 hpf) (Fig. 2a); at the early gastrula stage thetranscript was detected in a few cells of the ventral ectoderm(Fig. 2b–b′). As gastrulation proceeds, the admp1 expression do-main expands along the animal-vegetal axis with the expressionbeing stronger in the ventral midline of the ectoderm and weakerin the ventral side of the archenteron (Fig. 2c–e′). Double fluor-escent in situ hybridization (FISH) analysis confirmed that the seaurchin admp1 is co-expressed with brachyury (bra) (Li et al., 2013),the transcript of which marks the stomodeal domain on the ven-tral side of the mid gastrula (Fig. 2f). In the indirectly developinghemichordate acorn worm P. flava, admp1 is expressed on theventral side of the embryo (Fig. 2g–k), similar to the situation seenin the sea urchin. Double FISH confirmed that the Ptychoderaadmp1 expression domain is more towards the animal pole, whencompared to the chordin expression domain in the central ventralectoderm (Rottinger and Martindale, 2011) (Fig. 2l). The expressionof admp1 in the ventral ectoderm resembles that seen in the lategastrula of the direct developing acorn worm, Saccoglossus kowa-levskii (Lowe et al., 2006).

During sea urchin embryogenesis, the high BMP activity on thedorsal side specifies dorsal epidermal cell fate, and the presence ofthe BMP antagonist Chordin maintains BMP activity at a low levelon the ventral side (Angerer et al., 2000; Chen et al., 2011; Laprazet al., 2009). High BMP activity is also observed on the dorsal side ofthe P. flava embryos (Rottinger et al., 2015). The expression ofadmp1 on the side of ambulacrarian embryos with low BMP activityis similar to what is seen in chordates; in the invertebrate chordateamphioxus (Branchiostoma floridae), the admp gene is expressedalong the dorsal midline of the ectoderm and mesendoderm, where

Fig. 2. Expression of sea urchin and acorn worm admp genes during embryogenesis. In situ hybridizations were performed at various developmental stages as indicated ontop (EG, early gastrula; MG, mid gastrula; LG, late gastrula). (a–e”) Sea urchin admp1 transcript is detected initially in the ventral ectodermal cells (solid arrowheads) duringgastrulation and later in the ventral endodermal cells (open arrowheads). (f) Double FISH of sea urchin admp1 and bra was performed at the mid gastrula stage. (g–j)Expression of the acorn worm admp1 gene (solid arrowheads) begins at the early blastula stage and the expression stays on the ventral side throughout gastrulation. (k)Admp1 expression in three patches of the ventral ectoderm is evident at the tornaria larval stage. (l) Double FISH of the acorn worm admp1 and chordin was performed at theearly gastrula stage. (m–q′) The sea urchin admp2 is expressed in the dorsal vegetal ectodermal cells (arrows) at the mesenchyme blastula stage and later in the anal plate ofthe aboral ectoderm at the LG and larval stages. (r) Double FISH of the sea urchin admp2 and irxA were performed at the mesenchyme blastula stage. (s–w) The acorn wormadmp2 transcript (arrows) was detected in the dorsal ectoderm from the blastula to the larval stages. (x) Double FISH of the acorn worm admp2 and dlx was performed at theblastula stage. All embryos/larvae were viewed from the lateral side, except b′-e′ and f from the oral/ventral side, m′–q′ from the vegetal pole, and r from the aboral/dorsalside. The asterisks denote the mouth opening at the larval stages.

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118 111

BMP activity is low (Yu et al., 2007). The expression of ambula-crarian admp1 genes on the ventral side of the embryos, where BMPactivity is low, suggests that the BMP/Admp regulatory circuit isconserved in this lineage. However, we found that the admp2 genes

of both sea urchin and acornworm are expressed not ventrally (as isadmp1), but on the dorsal side, from the blastula to the larval stages,where BMP signaling is highest (Fig. 2m–x; Fig. S3b). Double FISHanalysis confirmed that the sea urchin admp2 is expressed within

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118112

the vegetal part of the irxA expression domain in the dorsal ecto-derm (Chen et al., 2011) (Fig. 2r), while the acorn worm admp2 isexpressed in a broader region of the dorsal ectoderm, similar to thedlx expression domain (Harada et al., 2001) (Fig. 2x). Equivalentexpression patterns of admp2 genes on the dorsal side of the em-bryos have also been described in the European sea urchin Para-centrotus lividus (Saudemont et al., 2010) and in embryos from aHawaiian population of P. flava (Rottinger et al., 2015). The ambu-lacrarian admp1 and admp2 genes are therefore expressed on op-posite sides along the DV axis of the developing embryos.

3.3. Opposite transcriptional control of the two admp genes by BMPand Nodal signals

We next investigated whether these admp genes are regulatedby BMP signals. As expected for the conserved BMP/Admp reg-ulatory circuit, the expression of the admp1 gene in both seaurchin and hemichordate embryos was expanded when the BMP

Fig. 3. Opposite transcriptional control of admp1 and admp2 genes by BMP signals. Thewith the bmp2/4 MO (50 μM), in acorn worm embryos treated with DMH1 (2 μM), and(zBMP4, 100 ng/ml) BMP4 protein. Almost 100% of the embryos showed the phenotype dside to the left and dorsal side to the right), or vegetal (vv) side, as indicated at the bot

activity was down-regulated, either by injecting a bmp2/4 gene-specific antisense morpholino (MO) in the sea urchin embryo or bytreating the acorn worm embryos with DMH1, a BMP signalinginhibitor (Hao et al., 2010) (Fig. 3a–b′). Conversely, elevating BMPactivity by treating the embryos with exogenous recombinant BMPproteins abolished the expression of admp1 genes (Fig. 3c–d′). Instriking contrast, the expression of the novel admp2 genes in bothambulacrarians is regulated by BMP signaling in completely theopposite way compared to the conserved BMP/Admp regulatorycircuit. Ambulacrarian admp2 expression was abolished when theBMP activity was down-regulated and increased if the BMP ac-tivity was elevated (Fig. 3e–h′). These results show that the am-bulacrarian admp1 and admp2 genes are under opposite tran-scriptional control by BMP signals; BMP signaling represses admp1expression but activates that of admp2. It is important to mentionthat we found that short treatments of recombinant BMP proteinsfailed to perturb expression of both admp1 and admp2, but wereable to induce the expression of its direct target gene tbx2/3 (Fig.

in situ hybridization of admp genes were performed in sea urchin embryos injectedin embryos treated with the recombinant mouse (mBMP4, 250 ng/ml) or zebrafishisplayed. The embryos were observed from the oral/ventral (ov), lateral (lv; ventraltom left corner of each panel.

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118 113

S4a–f′), suggesting that both admp1 and admp2 are indirectlyregulated by BMP signals.

Because Nodal signaling is a major player in DV patterning insea urchin embryos, we further investigated whether the twoadmp genes are regulated by Nodal signaling. In embryos injectedwith nodal MO, admp1 expression was abolished while the admp2expression domain was expanded to the whole vegetal ectoderm(Fig. S4g–n). These results suggest that, similar to the oppositeeffects of BMP signals on the expression of admp genes, Nodalsignals induce the expression of the admp1 and repress admp2expression in the ventral ectoderm. Because bmp2/4 is down-stream of Nodal signals and required for admp2 expression(Fig. 3e–e′), the fact that the expression domain of admp2 is ex-panded in the nodal morphants suggests that an additional factoris able to activate admp2 expression in the vegetal ectoderm.

Fig. 4. Admp and BMP molecules act synergistically to generate BMP activity gradient an(100 μM) or admp2 tMO (300 μM), and the phenotypes were observed at the pluteus stawere viewed from the lateral (lv) or vegetal (vv) sides with the ventral side to the left. (MOs against admp1, admp2, and (or) bmp5-8 (100 μM) as indicated. The gastrula embrnumber of cell rows containing the admp1 transcript was quantified. The sample sizespanel. The error bars are the standard deviation, and the p value is o0.001 (T test, ***)

3.4. Admp and BMP molecules act synergistically to generate BMPactivity gradient along the DV axis

To investigate the roles of Admp1 and Admp2 during sea urchinembryogenesis, we conducted knockdown experiments by in-jecting MOs to block RNA splicing or translation. Several splice-blocking (sMO) and translation-blocking (tMO) MOs were de-signed against admp1 and admp2, and the MO efficacy was vali-dated by RT-PCR and the test of GFP fusion constructs, respectively(Fig. S5a–f). In embryos injected with MOs against admp1 oradmp2, the elongation of the embryos along the DV axis wasslightly affected and the two skeletal rods failed to converge at theapex of the embryo, although the DV polarity remained re-cognizable with the mouth opening located on the ventral side(Fig. 4a–c′). Embryos injected with a control MO, with a random

d regulate DV patterning. (a–c′) Sea urchin embryos were injected with admp1 sMOge. More than 90% of the embryos showed the phenotype displayed. The embryosd–i) The in situ hybridizations of admp1 were performed in embryos injected withyos were observed from the ventral side, and the animal pole was on top. (j) Theare indicated inside each column. The injected MOs are listed at the bottom of the.

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118114

sequence at the same concentration, extended normally along theDV axis (Fig. S5g).

Because two BMP ligands, BMP2/4 and BMP5-8, reportedly playa significant role in enhancing dorsal regulatory gene transcription(Ben-Tabou de-Leon et al., 2013), our results support the propo-sition that the two sea urchin Admp proteins play, at least, par-tially redundant roles in patterning the DV axis. We confirmed thesynergistic effects of BMP2/4 and BMP5-8 by analyzing the ex-pansion of the admp1 expression domain as an indicator of theBMP activity level after blocking the translation of bmp2/4 andbmp5-8 by MO injections (Fig. S5h). Similar synergistic effectswere also observed when MOs against admp1 or admp2 were in-jected together with bmp5-8 MO (Fig. 4d–h). The maximum ex-pansion of the admp1 expression domain was observed whenembryos were co-injected with all three MOs (Fig. 4i–j). Takentogether, these results suggest that the dorsal and ventral BMPsignals act synergistically to pattern the DV axis in the sea urchin

Fig. 5. Functional analyses of Admp1 and Admp2 in the sea urchin and zebrafish embryourchin mesenchyme blastula embryos were stained with an anti-pSmad antibody (a–cmarker dlx (g–i), or the pigment cell precursor marker gcm (j–l). The white arrowhead indorsal side of the normal pluteus larva (m), but were more broadly distributed in the admurchin embryos were viewed from the lateral side (lv) with the ventral side to the lefphenotype displayed. (p-z6) Zebrafish embryos were injected with 200 pg of mRNAshybridizations of the dorsal organizer markers gsc and chordin, and the ventral marker evv) or ventral (w–y) sides. The length of the gsc expression domain is indicated by bars. Tphenotypes were observed at 24 hpf (z-z6). The arrowheads and arrows indicate the rcorners indicate the ratios of the displayed phenotypes.

embryo, similarly to what has been observed in the Xenopus em-bryos (Reversade and De Robertis, 2005).

3.5. The two Admps have evolved different functions recognized bythe sea urchin embryo but not by zebrafish embryos

Although the endogenous Admp1 and Admp2 play similar rolesin patterning the DV axis during sea urchin embryogenesis, it isintriguing to examine whether they have evolved different func-tions by introducing them ectopically into the embryos. The over-expression of either admp1 or admp2 in the sea urchin embryoexpanded BMP activity, as evidenced by phospho-Smad1/5/8(pSmad) antibody staining, although the effects differed slightlybetween these two genes. Similar to previously published results(Chen et al., 2011; Lapraz et al., 2009), pSmad signal was detectedon the dorsal half of the blastula embryos (Fig. 5a–a′). In admp1-over-expressing embryos, pSmad was detected in an equatorial

s. Upon injection of mRNAs (300 ng/μl) encoding sea urchin Admp1 or Admp2, sea′) or hybridized with probes against the ventral marker chordin (d–f), the dorsalc indicates the animal pole. Pigment cells (black arrowheads) were restricted to thep1mRNA-injected (n) and absent in the admp2mRNA-injected embryos (o). The seat or from the animal pole (av). Almost 100% of the injected embryos showed theencoding sea urchin Admp1, Admp2, or amphioxus Admp (Bf-Admp). The in situe were performed at 70% epiboly, and the embryos were viewed from the dorsal (p–he white arrow indicates the expression of chordin in the axial mesoderm. Zebrafisheduced tail and absence of tail, respectively. The numbers in the bottom left-hand

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118 115

belt of the ectodermal cells (Fig. 4b–b′). Over-expression of admp2expanded the ectodermal pSmad signal further towards the ani-mal pole and the pSmad signal was also detected in the ingressedmesenchyme cells (Fig. 4c–c′; Fig. S6a). We also noticed that, al-though pSmad signal was expanded when the embryos over-ex-pressing admp1 or admp2, the pSmad signal was generally weakerthan in the controls. It is possible that over-expression of admpsaffects the endogenous BMP signal transduction and that theseAdmp molecules are not as efficient as the BMP ligands in trans-ducing signals via Smad. When a higher concentration of admp1mRNA was injected, pSmad signal was even weaker (Fig. S6a).Admp2 was more effective than Admp1 at expanding pSmad sig-nal to the animal pole of the embryo and its effects on strength-ening BMP signals were dose-dependent. Genes that are normallyexpressed in the ventral ectoderm, such as chordin, gsc, and bmp2/4, were down-regulated in injected embryos, while the expressionof the dorsal ectodermal genes dlx and tbx2/3 was expanded(Fig. 5d–i; Fig. S6b), coinciding with the requirement of BMP sig-naling for their transcriptional activation (Ben-Tabou de-Leonet al., 2013; Saudemont et al., 2010). Clear differences betweenadmp1 and admp2 mRNA-injected embryos were also observed inthe expression of gcm (Fig. 4j–l), a gene that controls the specifi-cation of pigment cell precursors located in the dorsal half of thevegetal plate (Ransick and Davidson, 2012). The expression of gcmexpands to the ventral side in the admp1-over-expressing embryosbut disappears when admp2 is over-expressed. Other distinct rolesof Admp1 and Admp2 include their effects on the expression ofthe dorsal ectodermal marker hox7 and the ciliary band markerhnf6 (Fig. S6b). Embryos injected with admp1 or admp2 mRNAswere observed to lose their DV polarity, as demonstrated by thefailure of the gut to bend towards the ventral side (Fig. 5m–o).Pigment cells were broadly distributed in the admp1-over-ex-pressing embryos but absent when admp2 was over-expressed.These phenotypic changes upon admp mRNA injections showed adose-dependency (Fig. S7). To examine whether the number ofpigment cells was increased in embryos over-expressing admp1,we performed in situ hybridization with a polyketide synthase (pks)probe; pks is a pigment cell-specific marker (Calestani et al.,2003). The results showed that the number of pigment cells re-mained unaltered in the admp1-over-expressed embryos, com-pared to the control embryos injected with the same amount of gfpmRNA (Fig. S8a–b), suggesting that only the distribution of pig-ment cells but not their number was affected.

To extend the comparison, sea urchin embryos injected withmRNA encoding the amphioxus Admp were observed and theydisplayed phenotypes similar to that seen in sea urchin admp1mRNA-injected embryos (Figs. S6b and S7). This functional simi-larity is consistent with their phylogenetic affinities (Fig. 1) andsuggests that the sea urchin Admp1 retains the function seen inother bilaterian Admp proteins and that, by contrast, Admp2 hasacquired a different role. In admp2-over-expressing embryos, thenon-skeletogenic mesoderm-derived pigment and blastocoelarcells, as well as the expression of their respective specificationgenes gcm and ese, completely disappear (Materna et al., 2013)(Fig. 5l and Fig. S8c).

In order to examine whether the different functions of the seaurchin Admp1 and Admp2 can be recognized in a vertebrate sys-tem in which admp2 gene is absent, we over-expressed either ofthese paralogs in zebrafish embryos. Compared to control zebra-fish embryos, all embryos injected with mRNA encoding seaurchin Admp1, Admp2, or amphioxus Admp have smaller organi-zers, as revealed by the expression of the organizer marker gsc,and are shorter due to the partial loss of dorsal structures (Fig. 5p–s). Similarly, the extension of chordin expression in the axial me-sodermwas not observed in the admp1- or admp2-over-expressingembryos (Fig. 5t–v), while the expression of the ventral marker eve

was not affected (Fig. 5w–y). The phenotypes of embryos over-expressing different Admp proteins are indistinguishable (Fig. 5z–z6), and resemble those of zebrafish embryos over-expressing itsown admp mRNA (Lele et al., 2001; Willot et al., 2002). Controlembryos injected with the same amount of gfp mRNA developednormally (Fig. S8d). Therefore, although Admp1 and Admp2 havedifferent effects when over-expressed in the sea urchin embryo,our results show that these Admp family members are functionallyinterchangeable in zebrafish embryos, suggesting that the corre-sponding signaling receiving system for Admp2 has not beenevolved or has been lost in the chordate lineage.

3.6. Both prodomains and TGFβ domains of the Admp proteins areimportant for their functional divergence

Admp proteins, similar to other TGFβ family members, containa conserved C-terminal TGFβ domain and a less well-conservedN-terminal prodomain (Harrison et al., 2011). To investigate if ei-ther domain is likely to be responsible for the difference in thefunction of Admp1 and Admp2 in sea urchin embryos, we createdchimeric constructs by swapping the two functional domains ofboth proteins (Fig. 6).

Embryos injected with the chimeric construct containing theAdmp1 prodomain and the Admp2 TGFβ domain showed a phe-notype similar to that of the admp1-over-expressing embryos,with a disrupted DV axis and a broad distribution of pigment cells(Fig. 6a–b′ and e–f'). Conversely, when injecting, at high con-centration, the chimeric construct consisting of the Admp2 pro-domain and the Admp1 TGFβ domain, half of the injected embryoslost their pigments cells, a phenotype similar to that of the admp2-over-expressing embryos (Fig. 6c–d′ and g–j′). These results sug-gest that the functional divergence of the two sea urchin Admps iscontributed mainly by their prodomains. We next examined theexpression of DV markers in the embryos injected with the chi-meric mRNAs (Fig. S9). The results showed that the distinct effectson marker gene expression observed in admp1 and admp2-over-expressing embryos (Fig. 5j–l; Fig. S6b) were absent in the em-bryos expressing chimeras. Therefore, we concluded that both theprodomains and TGFβ domains are important for the functionaldivergence of Admp1 and Admp2.

4. Discussion

In this study, we have shown that the conserved BMP/Admp1regulatory circuit in which BMP signals repress admp1 expressionand Admp1 contributes to the BMP activity is present in ambula-crarian embryos. In addition, ambulacrarian embryos also use asecond BMP/Admp2 regulatory circuit, in which BMP signals ac-tivate admp2 transcription, and Admp2 further boosts BMP activ-ity. The opposite transcriptional control of the two admp genes byBMP signals shown here confirmed the results seen in anotherspecies of sea urchins (Lapraz et al., 2015). Our results stronglysuggest that both regulatory circuits were present in the commonancestor of the Ambulacraria. Based on our phylogenetic analysis,we propose that a single admp gene was duplicated into admp1and admp2 in the bilaterian ancestor (Fig. 7). Due to the still un-certain phylogenetic positions of Xenoturbellida and Acoelomor-pha (together are called Xenacoelomorpha) as an early branchinggroup of Bilateria or a sister group of Ambulacraria (Dunn et al.,2014), admp2 genes may have been retained independently orpresent in the common ancestor of Ambulacraria and Xenacoelo-morpha, respectively, and then lost in the protostome and chor-date lineages.

Our study further provides a nice example for understandingthe evolution of transcriptional control and the changes of protein

Fig. 6. Phenotypes of sea urchin embryos injected with mRNAs encoding Admp1 (a–b′), Admp2 (c–d′), hybrid of the prodomain from Admp1 and the TGFβ domain fromAdmp2 (Admp1P2T) (e–f′), or hybrid of the prodomain from Admp2 and the TGFβ domain from Admp1 (Admp2P1T) (g–j′). The domain structures of Admp proteinstranslated from the injected mRNAs are illustrated. Pigment cells are indicated with arrowheads. Live pluteus larvae (72 hpf) were observed from the lateral or vegetal side,and the ventral side is to the left. The black box represents the two phenotypes observed in Admp2P1T-over-expressed embryos. The ratios of the displayed phenotypes areindicated at the bottom left corner.

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118116

functions after duplication of genes encoding signaling ligands. InCnidaria, the sister group of Bilateria, BMP signaling is also im-portant for establishing their secondary axis during gastrulation(Saina et al., 2009). Prior to gastrulation, the admp-related gene ofthe cnidarian Nematostella vectensis is expressed in a central do-main located at the animal pole, bmp2/4 is expressed in the centralring that surrounds this central domain, and chordin is expressedin the external ring that lies adjacent to the internal ring (Rot-tinger et al., 2012). It is unclear whether the expression of theNematostella admp-related gene is regulated by BMP signals. It isalso unknown whether the two admp genes of Xenoturbella andacoels are under similar opposite regulation by BMP signals.Nevertheless, our results suggest that, after the ancient admp geneduplicated into admp1 and admp2 in the bilaterian ancestor, thetwo ambulacrarian paralogs were transcriptionally regulated by

the BMP signals in opposite ways (Fig. 7).Changes in transcriptional control after gene duplications are

also found in bmp2 and bmp4 genes, after they duplicated fromtheir ancestral bmp2/4 gene. During zebrafish gastrulation, theexpression of bmp2b and bmp4 is restricted to the future ventralside (Langdon and Mullins, 2011; Ramel and Hill, 2013). Interest-ingly, in the Xenopus embryos, bmp2 is expressed on the dorsalside, opposite to the expression domain of its paralog bmp4 on theventral side (De Robertis and Colozza, 2013; Inomata et al., 2008).Nevertheless, in both animals, the regulatory circuits involvingBMPs and their antagonists robustly generate a ventral to dorsalBMP activity gradient. In most bilaterians, expression of bmp andchordin are on opposite sides of embryos. However, in both seaurchin blastulae (Lapraz et al., 2009; Su, 2009) and the Nematos-tella gastrula (Matus et al., 2006), bmp2/4 and chordin are co-

Fig. 7. Evolutionary scenarios of the admp genes. The common ancestor at each node is indicated with colored circles. The dashed lines indicate two possible phylogeneticpositions of Xenoturbella and acoels. The ovals near the admp loci denote possible enhancers; the purple ovals are negatively and the pink ovals are positively regulated byBMP signals, respectively. The dashed admp2 loci indicate the lost of admp2 genes in the protostomes and chordates.

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118 117

expressed on the same side of the embryos, and their BMP activitygradients are still perfectly formed on the opposite side of thechordin expression domain (Chen et al., 2011; Lapraz et al., 2009;Leclere and Rentzsch, 2014). These examples show that tran-scriptional regulation of bmp genes were modified during evolu-tion without affecting the generation of BMP activity gradients.These data also support the predictions from a mathematicalmodel in which the expression of BMP antagonists (such as chor-din), and not the bmp themselves (such as bmp4), is needed to bespatially restricted to generate a signaling gradient (Genikhovichet al., 2015).

The duplication of the admp genes not only gave rise to twoopposite regulatory circuits, but the functions of their gene pro-ducts have also changed. Our study shows that the two sea urchinAdmps have different effects on the nonskeletogenic mesodermalcells when they are over-expressed in the sea urchin embryos. It isthus clear that functions of the two Admp paralogs diversifiedafter the duplication and we speculate that one of the two para-logs has kept the ancestral function and the other has evolved newfunctions (neofunctionalization) (Carroll, 2005; Force et al., 1999).One clear example of the functional changes in signaling ligandsafter gene duplications is the opposite effect of the two Nema-tostella paralogous FGFs on apical organ formation (Rentzsch et al.,2008). Neofunctionalization is probably an important and com-mon feature for duplicated genes to be retained in genomes (Assisand Bachtrog, 2013).

Taken together, our study on admp genes provides a good ex-ample of both the rewiring of signaling circuits and the changes ofprotein functions resulting from the duplication of genes encodingsignaling molecules.

Author Contributions

Y.C. Chang designed and performed most of the experimentsconducted in sea urchin and zebrafish embryos in consultationwith Y.H.S. C.Y.P. generated chimeric constructs, and Y.C. Chen andH.C.T. performed experiments in hemichordate embryos in con-sultation with Y.H.S. and J.K.Y. P.M, M.J.T, and J.K.Y provided thesequences obtained from database mining. Y.C. Chang and Y.H.S.wrote the manuscript, and P.M., M.J.T, and J.K.Y commented on it.

Acknowledgments

We thank Yu-Hsiu Liu in Dr. Sheng-Ping L. Hwang's laboratoryfor her technical assistance in microinjection of zebrafish embryosand Drs. Jen-Chieh Lin and Bon-chu Chung for providing zebrafishcDNA clones for in situ hybridization. Y.H.S and J.K.Y. are supportedby the intramural fund from Institute of Cellular and OrganismicBiology, Academia Sinica and grants from Ministry of Science andTechnology, Taiwan (103-2311-B-001-030-MY3 and 103-2627-B-001-001 to Y.H.S. and 102-2311-B-001-011-MY3 and 104-2923-B-001-002-MY3 to J.K.Y.). M.J.T is supported by a Royal SocietyWolfson Research Merit Award. P.M. is supported by a Grant of theSpanish “Ministerio de Economia y Competitividad (MINECO)(BFU2012-32806)”.

Appendix A. Supplementary material

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.ydbio.2015.12.015.

References

Angerer, L.M., Oleksyn, D.W., Logan, C.Y., McClay, D.R., Dale, L., Angerer, R.C., 2000.A BMP pathway regulates cell fate allocation along the sea urchin animal-ve-getal embryonic axis. Development 127, 1105–1114.

Assis, R., Bachtrog, D., 2013. Neofunctionalization of young duplicate genes inDrosophila. Proc. Natl. Acad. Sci. USA 110, 17409–17414.

Ben-Tabou de-Leon, S., Su, Y.H., Lin, K.T., Li, E., Davidson, E.H., 2013. Gene regulatorycontrol in the sea urchin aboral ectoderm: spatial initiation, signaling inputs,and cell fate lockdown. Dev. Biol. 374, 245–254.

Bradham, C.A., Oikonomou, C., Kühn, A., Core, A.B., Modell, J.W., McClay, D.R.,Poustka, A.J., 2009. Chordin is required for neural but not axial development insea urchin embryos. Dev. Biol. 328, 221–233.

Calestani, C., Rast, J.P., Davidson, E.H., 2003. Isolation of pigment cell specific genesin the sea urchin embryo by differential macroarray screening. Development130, 4587–4596.

Cameron, R.A., Samanta, M., Yuan, A., He, D., Davidson, E., 2009. SpBase: the seaurchin genome database and web site. Nucleic Acids Res. 37, D750–D754.

Cannon, J.T., Kocot, K.M., Waits, D.S., Weese, D.A., Swalla, B.J., Santos, S.R., Halanych,K.M., 2014. Phylogenomic resolution of the hemichordate and echinodermclade. Curr. Biol. 24, 2827–2832.

Carroll, S.B., 2005. Evolution at two levels: on genes and form. Plos Biol. 3, e245.Castresana, J., 2000. Selection of conserved blocks from multiple alignments for

their use in phylogenetic analysis. Mol. Biol. Evol. 17, 540–552.Chen, J.H., Luo, Y.J., Su, Y.H., 2011. The dynamic gene expression patterns of tran-

scription factors constituting the sea urchin aboral ectoderm gene regulatorynetwork. Dev. Dyn. 240, 250–260.

Chen, S.H., Li, K.L., Lu, I.H., Wang, Y.B., Tung, C.H., Ting, H.C., Lin, C.Y., Lin, C.Y., Su, Y.H., Yu, J.K., 2014. Sequencing and analysis of the transcriptome of the acorn

Y.-C. Chang et al. / Developmental Biology 410 (2016) 108–118118

worm Ptychodera flava, an indirect developing hemichordate. Mar. Genom. 15,35–43.

De Robertis, E.M., 2008. Evo-devo: variations on ancestral themes. Cell 132,185–195.

De Robertis, E.M., Colozza, G., 2013. Development: scaling to size by protease in-hibition. Curr. Biol. 23, R652–R654.

Duboc, V., Rottinger, E., Besnardeau, L., Lepage, T., 2004. Nodal and BMP2/4 sig-naling organizes the oral-aboral axis of the sea urchin embryo. Dev. Cell. 6,397–410.

Dunn, C.W., Giribet, G., Edgecombe, G.D., Hejnol, A., 2014. Animal phylogeny and itsevolutionary implications. Annu. Rev. Ecol., Evol., Syst. 45, 371–395.

Force, A., Lynch, M., Pickett, F.B., Amores, A., Yan, Y.L., Postlethwait, J., 1999. Pre-servation of duplicate genes by complementary, degenerative mutations. Ge-netics 151, 1531–1545.

Gavino, M.A., Reddien, P.W., 2011. A Bmp/Admp regulatory circuit controls main-tenance and regeneration of dorsal-ventral polarity in planarians. Curr. Biol. 21,294–299.

Genikhovich, G., Fried, P., Prunster, M.M., Schinko, J.B., Gilles, A.F., Fredman, D.,Meier, K., Iber, D., Technau, U., 2015. Axis Patterning by BMPs: Cnidarian Net-work Reveals Evolutionary Constraints. Cell Rep. 10, 1646–1654.

Hao, J., Ho, J.N., Lewis, J.A., Karim, K.A., Daniels, R.N., Gentry, P.R., Hopkins, C.R.,Lindsley, C.W., Hong, C.C., 2010. In vivo structure-activity relationship study ofdorsomorphin analogues identifies selective VEGF and BMP inhibitors. ACSChem. Biol. 5, 245–253.

Harada, Y., Okai, N., Taguchi, S., Shoguchi, E., Tagawa, K., Humphreys, T., Satoh, N.,2001. Embryonic expression of a hemichordate distal-less gene. Zool. Sci. 18,57–61.

Harrison, C.A., Al-Musawi, S.L., Walton, K.L., 2011. Prodomains regulate the synth-esis, extracellular localisation and activity of TGF-beta superfamily ligands.Growth Factors 29, 174–186.

Ikuta, T., Chen, Y.C., Annunziata, R., Ting, H.C., Tung, C.H., Koyanagi, R., Tagawa, K.,Humphreys, T., Fujiyama, A., Saiga, H., Satoh, N., Yu, J.K., Arnone, M.I., Su, Y.H.,2013. Identification of an intact ParaHox cluster with temporal colinearity butaltered spatial colinearity in the hemichordate Ptychodera flava. BMC Evol. Biol.13, 129.

Imai, K.S., Daido, Y., Kusakabe, T.G., Satou, Y., 2012. Cis-acting transcriptional re-pression establishes a sharp boundary in chordate embryos. Science 337,964–967.

Inomata, H., Haraguchi, T., Sasai, Y., 2008. Robust stability of the embryonic axialpattern requires a secreted scaffold for chordin degradation. Cell 134, 854–865.

Kumano, G., Ezal, C., Smith, W.C., 2006. ADMP2 is essential for primitive blood andheart development in Xenopus. Dev. Biol. 299, 411–423.

Langdon, Y.G., Mullins, M.C., 2011. Maternal and zygotic control of zebrafish dor-soventral axial patterning. Annu. Rev. Genet. 45, 357–377.

Lapraz, F., Besnardeau, L., Lepage, T., 2009. Patterning of the dorsal-ventral axis inechinoderms: insights into the evolution of the BMP-chordin signaling net-work. Plos Biol. 7, e1000248.

Lapraz, F., Haillot, E., Lepage, T., 2015. A deuterostome origin of the Spemann or-ganiser suggested by Nodal and ADMPs functions in Echinoderms. Nat. Com-mun. 6, 8434.

Lapraz, F., Rottinger, E., Duboc, V., Range, R., Duloquin, L., Walton, K., Wu, S.Y.,Bradham, C., Loza, M.A., Hibino, T., Wilson, K., Poustka, A., McClay, D., Angerer,L., Gache, C., Lepage, T., 2006. RTK and TGF-beta signaling pathways genes inthe sea urchin genome. Dev. Biol. 300, 132–152.

Leclere, L., Rentzsch, F., 2014. RGM regulates BMP-mediated secondary axis for-mation in the sea anemone Nematostella vectensis. Cell. Rep. 9, 1921–1930.

Lele, Z., Nowak, M., Hammerschmidt, M., 2001. Zebrafish admp is required to re-strict the size of the organizer and to promote posterior and ventral develop-ment. Dev. Dyn. 222, 681–687.

Li, E., Materna, S.C., Davidson, E.H., 2013. New regulatory circuit controlling spatialand temporal gene expression in the sea urchin embryo oral ectoderm GRN.Dev. Biol. 382, 268–279.

Lin, C.Y., Tung, C.H., Yu, J.K., Su, Y.H., 2016. Reproductive periodicity, spawning in-duction, and larval metamorphosis of the hemichordate acorn worm Ptycho-dera flava. J Exp Zool B 326, 47–60.

Lowe, C.J., Terasaki, M., Wu, M., Freeman Jr., R.M., Runft, L., Kwan, K., Haigo, S.,Aronowicz, J., Lander, E., Gruber, C., Smith, M., Kirschner, M., Gerhart, J., 2006.Dorsoventral patterning in hemichordates: insights into early chordate evolu-tion. Plos Biol. 4, e291.

Materna, S.C., Ransick, A., Li, E., Davidson, E.H., 2013. Diversification of oral andaboral mesodermal regulatory states in pregastrular sea urchin embryos. Dev.Biol. 375, 92–104.

Matus, D.Q., Thomsen, G.H., Martindale, M.Q., 2006. Dorso/ventral genes areasymmetrically expressed and involved in germ-layer demarcation duringcnidarian gastrulation. Curr. Biol. 16, 499–505.

McClay, D.R., 2011. Evolutionary crossroads in developmental biology: sea urchins.Development 138, 2639–2648.

Ramel, M.C., Hill, C.S., 2013. The ventral to dorsal BMP activity gradient in the earlyzebrafish embryo is determined by graded expression of BMP ligands. Dev. Biol.378, 170–182.

Ransick, A., Davidson, E.H., 2012. Cis-regulatory logic driving glial cells missing:self-sustaining circuitry in later embryogenesis. Dev. Biol. 364, 259–267.

Rentzsch, F., Fritzenwanker, J.H., Scholz, C.B., Technau, U., 2008. FGF signallingcontrols formation of the apical sensory organ in the cnidarian Nematostellavectensis. Development 135, 1761–1769.

Reversade, B., De Robertis, E.M., 2005. Regulation of ADMP and BMP2/4/7 at op-posite embryonic poles generates a self-regulating morphogenetic field. Cell123, 1147–1160.

Rottinger, E., Dahlin, P., Martindale, M.Q., 2012. A framework for the establishmentof a cnidarian gene regulatory network for “endomesoderm” specification: theinputs of ss-catenin/TCF signaling. Plos Genet. 8, e1003164.

Rottinger, E., DuBuc, T.Q., Amiel, A.R., Martindale, M.Q., 2015. Nodal signaling isrequired for mesodermal and ventral but not for dorsal fates in the indirectdeveloping hemichordate, Ptychodera flava. Biol. Open. 4, 830–842.

Rottinger, E., Martindale, M.Q., 2011. Ventralization of an indirect developinghemichordate by NiCl suggests a conserved mechanism of dorso-ventral (D/V)patterning in Ambulacraria (hemichordates and echinoderms). Dev. Biol. 354,173–190.

Saina, M., Genikhovich, G., Renfer, E., Technau, U., 2009. BMPs and chordin regulatepatterning of the directive axis in a sea anemone. Proc. Natl. Acad. Sci. USA 106,18592–18597.

Saudemont, A., Haillot, E., Mekpoh, F., Bessodes, N., Quirin, M., Lapraz, F., Duboc, V.,Rottinger, E., Range, R., Oisel, A., Besnardeau, L., Wincker, P., Lepage, T., 2010.Ancestral regulatory circuits governing ectoderm patterning downstream ofNodal and BMP2/4 revealed by gene regulatory network analysis in an echi-noderm. Plos Genet. 6, e1001259.

Srivastava, M., Mazza-Curll, K.L., van Wolfswinkel, J.C., Reddien, P.W., 2014. Whole-body acoel regeneration is controlled by Wnt and Bmp-Admp signaling. Curr.Biol. 24, 1107–1113.

Su, Y.H., 2009. Gene regulatory networks for ectoderm specification in sea urchinembryos. Biochim. Biophys. Acta 1789, 261–267.

Su, Y.H., Li, E., Geiss, G.K., Longabaugh, W.J., Kramer, A., Davidson, E.H., 2009. Aperturbation model of the gene regulatory network for oral and aboral ecto-derm specification in the sea urchin embryo. Dev. Biol. 329, 410–421.

Tamura, K., Stecher, G., Peterson, D., Filipski, A., Kumar, S., 2013. MEGA6: MolecularEvolutionary Genetics Analysis version 6.0. Mol. Biol. Evol. 30, 2725–2729.

Tsai, M.Y., Lu, Y.F., Liu, Y.H., Lien, H.W., Huang, C.J., Wu, J.L., Hwang, S.L., 2014.Modulation of p53 and met expression by Kruppel-like factor 8 regulateszebrafish cerebellar development. Dev. Neurobiol. 75, 908–926.

Willot, V., Mathieu, J., Lu, Y., Schmid, B., Sidi, S., Yan, Y.L., Postlethwait, J.H., Mullins,M., Rosa, F., Peyrieras, N., 2002. Cooperative action of ADMP- and BMP-medi-ated pathways in regulating cell fates in the zebrafish gastrula. Dev. Biol. 241,59–78.

Yu, J.K., Satou, Y., Holland, N.D., Shin, I.T., Kohara, Y., Satoh, N., Bronner-Fraser, M.,Holland, L.Z., 2007. Axial patterning in cephalochordates and the evolution ofthe organizer. Nature 445, 613–617.

Yu, J.K., Wang, M.C., Shin, I.T., Kohara, Y., Holland, L.Z., Satoh, N., Satou, Y., 2008. AcDNA resource for the cephalochordate amphioxus Branchiostoma floridae. Dev.Genes. Evol. 218, 723–727.