RIPPLY3 is a retinoic acid-inducible repressor required for setting the borders of the pre-placodal ectoderm

Retinoid signaling acting through the nuclear retinoic acid receptors (RARs) is crucial to the establishment of the anteroposterior (AP) axis during embryonic development. RARs function as ligand-modulated transcription factors that activate expression of their target genes (Chambon, 1996; Mangelsdorf et al., 1995; Mark et al., 2009). Studies in a variety of systems have shown that retinoic acid (RA) is an important component of the neural posteriorizing signal (Blumberg et al., 1997; Durston et al., 1989; Lloret-Vilaspasa et al., 2010; Papalopulu and Kintner, 1996; Shiotsugu et al., 2004) as well as of the AP patterning of the neural crest, somites and limbs (Moreno et al., 2008; Moreno and Kintner, 2004; Niederreither et al., 2002; Stratford et al., 1999; Villanueva et al., 2002). Retinoid signaling also establishes AP identity in the endoderm and mesoderm (Bayha et al., 2009; Deimling and Drysdale, 2009; Pan et al., 2007).

RAR signaling has been studied in some detail in the neural crest (Dupe and Pellerin, 2009; Li et al., 2010; Villanueva et al., 2002); RA is required for eye morphogenesis, otocyst development and specification of the olfactory epithelium (Bok et al., 2011; Lupo et al., 2011; Matt et al., 2005; Radosevic et al., 2011; Rawson and LaMantia, 2006; Romand et al., 2006). Relatively little is known about RAR function in the early patterning of sensory organs of the head. The pre-placodal ectoderm (PPE) is an anterior lateral crescent-shaped region that forms at the boundary of the neural plate and neural crest (Brugmann and Moody, 2005; Moody, 2007; Schlosser and Ahrens, 2004). Sensory placodes are epidermal thickenings derived from the PPE that give rise to characteristic sensory organs (e.g. olfactory, lens and otic placodes) and specialized ganglia (e.g. trigeminal, profundal, epibranchial and otic) of the vertebrate head (Ahrens and Schlosser, 2005; David et al., 2001; Schlosser and Ahrens, 2004). Much is known about the molecular pathways that subdivide the PPE into individual sensory placodes and how these give rise to adult structures. Less is known about the mechanisms through which the PPE is established and segregated from the adjacent neural crest. Whether the PPE is derived from a common precursor shared with neural crest (Brugmann and Moody, 2005; Streit, 2007) or is instead a distinct entity with a different origin (Ahrens and Schlosser, 2005; Pieper and Schlosser, 2009; Schlosser, 2008) is currently controversial.

Our previous work suggested that RAR is involved in the AP patterning of the PPE. Knockdown of RARα2 led to posterolateral expansion of Fgf8 and Fgfr4 expression in the PPE (Shiotsugu et al., 2004). Fgf8, together with BMP inhibitors, induces expression of key factors required for placode development (Ahrens and Schlosser, 2005; Litsiou et al., 2005). Therefore, our result suggested that RARα2 is necessary to restrict the posterolateral boundary of the PPE, probably by inducing repressors of Fgf8/Fgfr4 expression. Our published microarray analysis (Arima et al., 2005) identified two interesting RA-induced genes expressed in the PPE: Tbx1, a T-box transcription factor; and Dscr6, a novel member of the Ripply/Bowline family, designated as Ripply3. Members of the Ripply/Bowline family are Groucho-associated co-repressors that regulate regional boundaries of gene expression, via interaction with T-box proteins (Kawamura et al., 2008; Moreno et al., 2008). We hypothesized that RAR signaling controls the borders of Fgf8 expression in the PPE by regulating the expression of Ripply3 and Tbx1.

Here, we show that RAR signaling is required for the AP patterning of the PPE. RA upregulates Ripply3 throughout early development, whereas Tbx1 is upregulated by RA before (in contrast to published findings), but inhibited after, neurogenesis. TBX1 has a dual function in the PPE downstream of RAR. TBX1 induces PPE gene expression in regions where RIPPLY3 is absent. However, TBX1 restricts the posterolateral boundaries of PPE gene expression in areas where its expression overlaps with Ripply3 along the lateral edge of the anterior crescent demarcating the PPE. RIPPLY3 represses the ability of TBX1 to activate reporter gene constructs in vivo and this inhibition depends on the association of RIPPLY3 with its co-repressor GROUCHO and with TBX1. In agreement with our predictions, RIPPLY3 knockdown perturbs the borders of PPE marker expression. These results demonstrate a novel role for RAR in the precise positioning of the PPE boundaries and establish RIPPLY3 as a key factor that demarcates the boundaries of the PPE.

Ripply sequences were obtained from Genbank and Uniprot databases (Benson et al., 2008; Uniprot Consortium, 2009), aligned with MAFFT (L-INS-i algorithm) (Katoh et al., 2009; Katoh et al., 2005) and a phylogenetic tree constructed with PROml, version 3.69 (Protein Maximum Likelihood) (Felsenstein, 2005). Default settings were used, global rearrangements (–G) were performed, and the outgroup (–O) was set to amphioxus. The resultant tree was drawn with FigTree (Rambaut, 2007). Conserved domains of the Ripply gene family were visualized with WebLogo (Crooks et al., 2004; Schneider and Stephens, 1990).

Xenopus eggs were fertilized in vitro as described previously (Blumberg et al., 1997; Koide et al., 2001) and embryos staged according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1967). Embryos were maintained in 0.1× MBS until appropriate stages or treated with 1 μM agonist (TTNPB) and 1 μM antagonist (AGN193109) as described (Arima et al., 2005).

Embryos were injected bilaterally or unilaterally at the two-cell stage with combinations of gene specific morpholinos (MO), mRNAs and 100 pg/embryo β-galactosidase mRNA lineage tracer. MOs used for this study are found in supplementary material Table S1. Control embryos were injected with 20 ng standard control MO: CCT CTT ACC TCA GTT ACA ATT TAT A (GeneTools). The following plasmids were constructed by PCR amplification of the protein-coding regions of the indicated genes and cloning into the expression vector pCDG1: xRARa2.2 (Sharpe, 1992), xTbx1 (Ataliotis et al., 2005) and xRipply3 (Hitachi et al., 2009). xRipply3^WRPW→AAAA, xRipply3^FPVQ→AAAA and Ripply3^{WRPW→AAAA, FPVQ→AAAA} were constructed by two-fragment PCR and cloned into pCDG1. pCS2-mCherry was provided by Dr Thomas Schilling (University of California, Irvine, CA, USA). All pCDG1 plasmids were linearized with NotI and mRNA was transcribed using mMessage mMachine T7 (Ambion). pCS2-mCherry was linearized with NotI and transcribed using mMessage mMachine SP6 (Ambion).

Whole-mount in situ hybridization was performed as described previously (Blumberg et al., 1997; Koide et al., 2001). Six1 (courtesy of Sally Moody, George Washington University, Washington DC, USA) was cloned into pBluescript II SK and linearized with BamHI. Fgf8 (courtesy of Nancy Papalopulu, University of Manchester, UK) cloned into pCS2 was linearized with BamHI. Ripply3 (clone Xl018m04) was linearized with NotI and transcribed with T7 RNA polymerase. Tbx1, Eya1 and Sox3 probes were prepared via PCR amplification of coding regions from cDNA: xTbx1 (Ataliotis et al., 2005), xEya1 (David et al., 2001) and xSox3 (Koyano et al., 1997). A T7 promoter was added to the 3′ end and probes were transcribed with MEGAscript T7 (Ambion) in the presence of digoxigenin-11-UTP or dinitrophenol-11-UTP as previously described (Arima et al., 2005). Forward primers and reverse primers containing a T7 promoter are listed in supplementary material Table S2. For double whole-mount in situ hybridization, genes were visualized with BM Purple (Roche) and either Fast Red (Roche) in 0.1 M Tris (pH 8.2) or BCIP (0.1875 mg/ml) and Tetrazolium Blue (0.5 mg/ml) in 100 mM Tris (pH 9.5), 50 mM MgCl₂, 100 mM NaCl and 2 mM levamisol. The first alkaline-phosphatase conjugated antibody was deactivated by washing embryos with 0.1 M glycine (pH 2) and 0.1% Tween for 40 minutes. The embryos were washed (four times, 20 minutes each) with MAB before staining for the second gene proceeded.

Total embryo RNA was extracted using Trizol reagent (InVitrogen Life Technologies), DNAse treated, LiCl precipitated and then reverse transcribed using Superscript III reverse transcriptase according to the manufacturer-supplied protocol (InVitrogen Life Technologies). The resultant first-strand cDNA was diluted 10-fold, then amplified and quantitated using a DNA Engine Opticon Continuous Fluorescence Detection System (Bio-Rad) with SYBR green detection (Roche Applied Science). Primer sets used are listed in supplementary material Table S3. Each primer set amplified a single band by gel electrophoresis and melting curve analysis. QPCR data was analyzed by employing the ΔΔCt method (Livak and Schmittgen, 2001) normalizing to Histone H4, which is insensitive to retinoic acid (Arima et al., 2005). Mann-Whitney statistical analysis was performed using MATLAB.

Embryos from whole-mount in situ hybridization were cleared with Histoclear and embedded in Paraplast+ using Sakura Finetek disposable base molds (15 mm × 15 mm × 5 mm) and yellow embedding rings. Serial sections were mounted onto Superfrost+ microscope slides, dried overnight, then dewaxed with xylene and photographed.

Two copies of the Brachyury T-box sites were cloned into tk-luc (Glass et al., 1989) using primers shown in supplementary material Table S4 via ExoIII digestion (Li and Evans, 1997) into the HindIII and BamHI sites of pTK-luc. Embryos derived from the same frog were injected at the two-cell (2/2 blastomeres) or four-cell stage (4/4 blastomeres), with 50 pg TBRE-TK-Luc (WT or MUT) construct and specific amounts and combinations of Tbx1 mRNA, Ripply3 mRNA or Ripply3^{WRPW→AAAA, FPVQ→AAAA} mRNA (indicated in Fig. 9). mCherry mRNA was added to each injection solution to keep the total amount of mRNA constant. Embryos were allowed to develop until stage 11, and then homogenized in 10 μl/embryo lysis buffer [0.1 M phosphate (pH 7.2), 1 mM DTT, 0.4 mM AEBSF], by two cycles of freeze thawing interspersed with vigorous mixing. A Bradford assay (BioRad) determined total protein concentration. Sodium luciferin solution (100 μl; 1× luciferase base buffer, 0.5 mM ATP, 5 mM DTT, 0.15 mg/ml coenzyme A, 0.5 mM sodium luciferin) and 20 μl of cleared lysate were added to a 96-well plate in triplicate. Relative light units were measured using an ML3000 luminometer, then normalized to total protein (Milnes et al., 2008). Error bars represent biological replicates (multiple pools of five embryos derived from the same female frog).

Previous results showed that RARα2 knockdown led to a posterolateral expansion of Fgf8 expression in the Xenopus neurula that was rescued by co-injection of Rarα2 mRNA (Shiotsugu et al., 2004) (supplementary material Fig. S1). Our microarray study identified two RAR target genes, Tbx1 and DSCR6/Ripply3, expressed in the PPE that we hypothesized might restrict Fgf8 expression (Arima et al., 2005). Whole-mount in situ hybridization revealed that Tbx1 and Ripply3/DSCR6 are expressed in the early PPE (Fig. 1A,B). Expression is ubiquitous until stage 16, when specific staining begins to occur at the horns of the anterior crescent. Ripply3 separates into a PPE domain circumscribed by intermediate and lateral plate mesoderm (Fig. 1B, stage 20, inset). Both genes are expressed in the epibranchial placodes (Fig. 1A,B; supplementary material Fig. S2) of tailbud embryos. Tbx1 is also expressed in the otic placode, whereas Ripply3 is expressed in the pronephros. Similar results were reported for Tbx1 (Ataliotis et al., 2005; Showell et al., 2006) and for Ripply3 (Hitachi et al., 2009). Quantitative real-time RT-PCR (QPCR) analysis showed that Ripply3 is expressed maternally. Zygotic transcription of both Ripply3 and Tbx1 first becomes prominent in the early neurula and continues to be expressed until tailbud stages (Fig. 1C,D).

RIPPLY3 has two highly conserved regions found in all Ripply family genes: a WRPW motif and a C-terminal Ripply homology domain (also called the bowline-DSCR-Ledgerline conserved ‘BDLC’ region) (Kondow et al., 2006). The WRPW motif facilitates interaction with GROUCHO (Fisher et al., 1996; Paroush et al., 1994). The ‘BDLC’ domain (supplementary material Fig. S3A) is predicted to mediate contacts with T-box proteins (Kawamura et al., 2008).

We aligned the available full-length sequences from Ripply genes and then generated a phylogenetic tree using the PROml maximum likelihood-based method (Felsenstein, 2005) (Fig. 2). The resulting tree properly segregates the three Ripply families. X. laevis Ledgerline and Bowline belong to the Ripply2 family, and there are two corresponding X. tropicalis genes (Ripply2.1=Ledgerline; Ripply2.2=Bowline), which are adjacent on scaffold 8, suggesting a local duplication event. As X. tropicalis is a true diploid organism, the similarities between Ripply2.1 and Ledgerline, and Ripply2.2 and Bowline suggest that these X. laevis genes are not pseudo-alleles resulting from pseudo-tetraploidy of X. laevis, but rather are distinct genes. In support of this, we identified the tetraploid copies of Ledgerline and Bowline (supplementary material Fig. S3B) and these are labeled as 2.1A and 2.1B, and 2.2A and 2.2B in Fig. 2. We were unable to identify a Xenopus Ripply1 ortholog and propose that the duplicated Ripply2 genes may subsume its function. To date, Ripply sequences are only found in Deuterostomes, with the exception of the sea anemone Nematostella vectensis (Nv genome v1.0, scaffold_76:151284-156895).

Research in several laboratories has linked RA with Tbx1 expression, albeit not in the PPE. Studies using mouse and zebrafish embryos demonstrated that application of high concentrations of RA leads to downregulation of Tbx1. Hence, the prevailing view is that endogenous RA inhibits the expression of Tbx1 in vitro and in vivo (Ataliotis et al., 2005; Guris et al., 2006; Roberts et al., 2005; Zhang et al., 2006). Our microarray results revealed that Tbx1 was upregulated by RA treatment in stage 15-18 (neurula) Xenopus laevis embryos (Arima et al., 2005). Time course experiments resolved this seemingly contradictory effect; the RAR-selective agonist TTNPB induced expression of Tbx1 prior to the late neurula (stage 18), but Tbx1 levels were downregulated at subsequent time points (Fig. 3A). Hence, the current view that Tbx1 is repressed by RA depends on when the expression is analyzed. Ripply3 showed consistent induction by TTNPB, and repression by the RAR-selective antagonist AGN193109 throughout early development (Fig. 3B).

These results establish that Ripply3 expression and early Tbx1 expression are stimulated by RA in the PPE. We tested for the necessity of RA signaling by unilaterally microinjecting embryos with a morpholino oligonucleotide (MO) directed against Rarα2 (Rarα2 MO). Whole-mount in situ hybridization at neurula stages revealed that expression of Tbx1 and Ripply3 was strongly downregulated in the injected side (Fig. 4B,E). MO specificity was demonstrated by rescuing the phenotype with Rarα2 mRNA (Fig. 4C,F). Thus, RARα2 is essential for correct expression of Tbx1 and Ripply3 in the PPE.

Next, we asked whether RARα2 knockdown would affect Eya1 and Six1, known markers of the PPE (Litsiou et al., 2005; Schlosser and Ahrens, 2004). Eya1 and Six1 expression was inhibited by Rarα2 MO (Fig. 4H,K) and rescued by Rarα2 mRNA (Fig. 4I,L). In partial knockdown embryos, where Eya1 and Six1 were mildly visible on the injected side, there was a posterolateral shift in their expression (not shown). We would expect the boundaries of Eya1 and Six1 to be shifted by partial RARα2 loss of function because this leads to a posterolateral shift of the Fgf8 expression border (supplementary material Fig. S1) (Shiotsugu et al., 2004). More complete RARα2 loss of function should lead to loss of Eya1 and Six1 expression as discussed below. These results demonstrate that RARα2 is crucial to restrict the border of Fgf8 and is a positive regulator of Eya1 and Six1 expression.

We performed double whole-mount in situ hybridization to ascertain the spatial relationships between Ripply3 and other PPE genes. The Ripply3 expression domain skirts the lateral edge of the anterior crescent occupied by Tbx1 (Fig. 5A, Fig. 6A) and Fgf8 (Fig. 5D, Fig. 6D). Tbx1 significantly overlaps with Ripply3 (Fig. 5A, Fig. 6A) and Fgf8 (Fig. 5E, Fig. 6E); however, expression of Ripply3 and Fgf8 do not overlap (Fig. 5D, Fig. 6D). Ripply3 expression is not contiguous with Six1, and would probably not directly confine its posterolateral border (Fig. 5B, Fig. 6B). Tbx1 overlaps Six1 in the horns of the crescent but not the arc (Fig. 5F, Fig. 6F). Eya1 has two PPE expression domains: the posterior placodal (pp) area and the profundal placodal (pPrV) area (David et al., 2001). Ripply3 and Tbx1 expression overlap with the former, but not with the latter (Fig. 5C,G, Fig. 6C,G). Fgf8 expression parallels the expression of Six1, except in the bottom of the crescent (where Fgf8 expression is lateral to Six1). Fgf8 expression is largely localized between the pp and pPrV domains of Eya1, although there is some overlap of expression (supplementary material Fig. S4). We infer from these expression patterns that RIPPLY3 and TBX1 influence Six1 and Eya1 indirectly, probably via Fgf8. Fig. 6H summarizes the expression domains of the PPE genes tested in Figs 5 and 6.

Fgf8 is essential for establishing the PPE (reviewed in Moody, 2007) where it also mediates Eya1 and Six1 expression (Ahrens and Schlosser, 2005). Hence, it was not surprising that MO-mediated knockdown of FGF8 (leading to 70% reduction in Fgf8 mRNA levels; supplementary material Fig. S5) decreased Tbx1 and Ripply3 expression (Fig. 7B,D). TBX1 knockdown resulted in a reduction of Fgf8, Ripply3, Six1 and Eya1 (Fig. 7E,G,I,K) that was rescued with Tbx1 mRNA (Fig. 7F,H,J,L). QPCR analysis of embryos uniformly injected with the Tbx1-MO also revealed reduced expression of PPE genes (supplementary material Fig. S6). Hence, although Fgf8 induces Tbx1 and other PPE genes, TBX1 is required to maintain Fgf8 expression, thus promoting the continued expression of genes within the PPE.

The results presented above show that Tbx1 and Ripply3 are RAR target genes expressed in overlapping patterns in the PPE, supporting the idea that they function together to restrict PPE boundaries. Recently, it was shown in Cos7 cells that TBX1 could repress reporter gene activity in the presence of RIPPLY3 (and that repression was dependent on the WRPW domain), whereas TBX1 would activate reporter gene activity in the absence of RIPPLY3 (Okubo et al., 2011). We tested the effects of Ripply3 on the ability of Tbx1 to activate reporter gene constructs in vivo, and whether the effects required the interaction of RIPPLY3 with GROUCHO and TBX1.

No high-affinity TBX1 consensus target DNA sequence has been identified, but most Tbox proteins bind the same core motif, AGGTGTGA (Wilson and Conlon, 2002), which is derived from the consensus Brachyury-binding site (an inverted palindromic repeat of the half-site sequence: AGGTGTGAAATT) (Kispert and Herrmann, 1993). The GTG triplet is a key point of protein-DNA contact in the TBX3 crystal structure (Coll et al., 2002). Most T-box proteins bind as monomers, but BRACHYURY and TBX1 dimerize and possess high binding affinity for the palindromic repeat (Muller and Herrmann, 1997; Papapetrou et al., 1997; Sinha et al., 2000).

We created two constructs with Brachyury palindromic repeats (one wild type, one mutated) preceding a thymidine kinase minimal promoter, driving firefly luciferase (Fig. 8A). The wild-type construct is expected to bind Tbx1 and associated factors, whereas TBX1 should not bind well to the mutant construct (Fig. 8A). The mutant construct was unresponsive to Tbx1, except slightly at the highest dose in injected embryos (supplementary material Fig. S7). The wild-type construct was injected into embryos alone and in combination with Tbx1, Ripply3 or Ripply3 mutant mRNAs. Microinjected Tbx1 activated the reporter (Fig. 8B, columns 2 and 6) and co-injection of Ripply3 repressed luciferase activity (Fig. 8B, columns 3 and 7). This repression was lost when the WRPW and FPVQ domains of RIPPLY3 were mutated (Fig. 8B, columns 4 and 8). Mutation of only the WRPW motif did not consistently relieve repression (data not shown), which suggests that WRPW is not the entire interaction domain. We infer that RIPPLY3 inhibits TBX1-mediated induction of the reporter gene, that RIPPLY3 converts TBX1 from an activator into a repressor and that the ability of RIPPLY3 to inhibit reporter gene activity was dependent upon both its TBX1 and GROUCHO interaction domains.

Our hypothesis that RIPPLY3 sets the posterolateral border of the PPE predicts that RIPPLY3 knockdown would disrupt this boundary. To test this possibility, we generated two different translation inhibiting MOs (one targeting the 5′ UTR and the other targeting the coding region). These produced similar phenotypes with respect to Fgf8 expression – a posterolateral shift and a broadening of the expression domain (supplementary material Fig. S8). Owing to the three-dimensional nature of the embryo, a lateral shift in PPE gene expression necessarily also causes a posterior shift. As the phenotypes resulting from combining the Ripply3 MOs was stronger than individual MOs, all subsequent experiments used the combination.

Ripply3 MO-injected embryos showed a posterolateral shift and expansion in Fgf8, Tbx1, Eya1 and Six1 expression (Fig. 9I-P), compared with the uninjected contralateral side and the control MO-injected embryos (Fig. 9A-H). We quantitated axis length and the extent to which marker gene expression was shifted in Ripply3 MO-injected embryos (supplementary material Fig. S9). The posterolateral shifts in Fgf8, Tbx1, Eya1 and Six1 expression were significant, whereas the axis length did not differ significantly between the injected and uninjected sides. Therefore, RIPPLY3 is required to establish, or to allow, the formation of a sharp posterolateral boundary of the PPE. We expected to see perceptible changes in the expression of placode markers in Ripply3 MO-injected tailbud stage embryos. These embryos demonstrated alterations in placode morphology, notably changes in the spacing of the epibranchial placodes, blurring of the lateral line placodes, and altered staining of Eya1 and Six1 in and around the otic vesicle (supplementary material Fig. S10). Although the otocyst seems devoid of an otic pit, transverse sections of Ripply3 MO-injected embryos showed that the otic vesicle is present (supplementary material Fig. S11).

Ripply3 is orthologous to the Down Syndrome Critical Region 6 (DSCR6) transcript in humans (Kawamura et al., 2005; Shibuya et al., 2000). Little is known about the function of DSCR6 in humans or the effects of having three copies of this gene, in the case of Trisomy 21 (Down Syndrome). In human DiGeorge Syndrome, 22q11.2, a region carrying TBX1, is deleted. We mimicked the increased dose of DSCR6 in Down Syndrome by microinjecting Ripply3 mRNA into Xenopus embryos. We found that Tbx1 expression was lost (Fig. 10A,B); however, when the WRPW or FPVQ domain was mutated (Ripply3^WRPW→AAAA or Ripply3^FPVQ→AAAA), Tbx1 expression was essentially normal (Fig. 10C-F). This confirms that the function of RIPPLY3 is dependent on both the WRPW domain (GROUCHO interaction) and the FPVQ domain (TBX1 interaction) (Fig. 8).

We identified two genes expressed in the PPE that are induced by RA and require RARα2 in the Xenopus neurula: Tbx1 and Ripply3. Tbx1 is a T-box gene that functions primarily as a transcriptional activator to regulate the expression of target genes such as Fgf8, Fgf10, Pitx2, Foxa2 and Gbx2, which are important in craniofacial patterning, and development of the aortic arches, thymus and parathyroid glands (Ataliotis et al., 2005; Hu et al., 2004; Ivins et al., 2005; Nowotschin et al., 2006; Packham and Brook, 2003). Disruption of Tbx1 results in placode and pharyngeal defects. For example, the zebrafish mutation, van gogh (Tbx1) has small otic vesicles owing to an underdeveloped placode (Whitfield et al., 1996). Xenopus Ripply3 is orthologous to human DSCR6 and belongs to the Ripply/Bowline gene family. Ripply3 is primarily associated with the development of the thymus, parathyroid and thyroid gland in mice (Okubo et al., 2011) but it is notable that Down Syndrome is associated with defects in sensory organs. We demonstrate a novel role for RIPPLY3 in regulating the boundaries of the PPE, with TBX1 being an integral player in this process, downstream of RA signaling.

Tbx1 and Ripply3 are expressed in the anterior lateral crescent that marks the PPE (Fig. 1). Ripply3 is lateral to Tbx1, but they significantly overlap at stage 18 (Figs 5, 6), which led us to investigate interactions between these genes in PPE development. At later stages, Tbx1 and Ripply3 are expressed in the epibranchial placodes (supplementary material Fig. S2), which are derived from the PPE, and give rise to cranial sensory neurons. Tbx1 is also expressed in the otic placode (Ataliotis et al., 2005; Vitelli et al., 2003). Ripply3 is also expressed in the lateral plate and intermediate mesoderm (Fig. 1), and, subsequently, the pronephros (supplementary material Fig. S2). DSCR6 is expressed in the human fetal kidney and fetal brain (Shibuya et al., 2000).

We were initially surprised when our microarray results revealed that Tbx1 was upregulated by RA at neurula stages, contrary to publications showing that Tbx1 expression is inhibited by RA (Zhang et al., 2006). We found that suppression of Tbx1 by RA occurs in later stages of embryonic development, whereas early (post-gastrulation) expression of Tbx1 is upregulated by RA (Fig. 3A). This is supported by evidence demonstrating that early Tbx1 expression is perturbed in the absence of RA. For example, in vitamin A-deficient quail embryos, Tbx1 expression was disturbed early and eventually lost (Roberts et al., 2005). RA bead implants led to strong downregulation of Tbx1, but only after 8-12 hours, and de novo protein synthesis was required to achieve full suppression of Tbx1 expression by RA (Roberts et al., 2005). This suggests that an additional, unidentified, factor(s) is induced by RA and acts as a negative regulator of Tbx1 expression after neurogenesis.

Ripply3 is also upregulated by RA (Fig. 3B); however, unlike Tbx1, the increase in expression is greater and is maintained through later stages of development. In the zebrafish presomitic mesoderm, Ripply1 and Ripply2 expression sets somite boundaries and was expanded in the presence of RA (Moreno et al., 2008). The RA-inducibility of Ripply3 provides further evidence that Ripply family genes are sensitive to RA signaling.

Phylogenetic analysis of Ripply sequences (Fig. 2) showed that Xenopus Ripply3 segregated appropriately with the Ripply3 family. Although it was believed that Xenopus Bowline and Ledgerline were divergent duplicated genes in the pseudo-tetraploid genome of X. laevis, we found that these genes each corresponded to two distinct genes in the diploid X. tropicalis. Bowline/Ripply2.2 and Ledgerline/Ripply2.1 each have their own pseudo-allele, designated Ripply2.1A, Ripply2.1B and Ripply2.2A, Ripply2.2B (supplementary material Fig. S3B). We remain unable to find a Xenopus Ripply1 ortholog, suggesting that a duplication and loss may have occurred during Xenopus evolution. We hypothesize that the Ripply2.1 and Ripply2.2 genes have functionally compensated for the lost Ripply1 gene.

Two intriguing anomalies were noted in the analysis of Ripply family genes. First, Ripply genes appear only in Deuterostome sequences; however, a likely Ripply homolog appears in the genome sequence of the sea anemone Nematostella vectensis. No other related sequences appear in any of the other sequenced invertebrate genomes, including Hydra, C. elegans and Drosophila. It is possible that Ripply family genes originated before the divergence of Protostomes and Deuterostomes but have subsequently been lost in Protostomes and most Cnidarians. Second, although a Ripply gene appears in the Cephalochordate Branchiostoma floridae, no Ripply genes were identified in the sequenced echinoderm or Urochordate genomes. This would be consistent with an early origin, subsequent loss model.

Several lines of evidence show that FGF signaling acts upstream and downstream of T-box proteins. Brachyury was shown to be regulated by FGF signaling, and vice versa, in Xenopus mesoderm formation (Casey et al., 1998; Isaacs et al., 1994; Schulte-Merker and Smith, 1995). Zebrafish mutant for the T-box protein spadetail showed a reduction in FGF signaling, but inhibition of the FGF receptor exacerbated the spadetail phenotype (Griffin and Kimelman, 2003). Tbx1-FGF interdependency was also discovered in tooth development (Mitsiadis et al., 2008). Our data support the possibility that Tbx1 and Fgf8 are engaged in a regulatory loop in the PPE as Tbx1 and Fgf8 expression overlap in the PPE (Figs 5, 6) and knockdown of one gene affects the other (Fig. 7). It is unlikely that Tbx1 initiates Fgf8 expression as Fgf8 expression precedes that of Tbx1 during development. Therefore, Tbx1 probably is required for maintenance, rather than induction, of Fgf8. As Fgf8 is required for the expression of Eya1 and Six1 (Ahrens and Schlosser, 2005), it follows that Tbx1 loss of function would indirectly lead to loss of Eya1 and Six1, as we observed (Fig. 7).

Although much is known about Tbx1, Ripply3 has primarily been studied in the pharyngeal apparatus, and in heart development (Okubo et al., 2011). Zebrafish Ripply1 (Kawamura et al., 2005), mouse Ripply2 (Biris et al., 2007; Chan et al., 2007) and Xenopus Bowline/Ledgerline (Ripply2) (Chan et al., 2006; Kondow et al., 2006) participate in setting somite boundaries during somitogenesis. In Ripply1 MO-injected zebrafish embryos, somites were not partitioned into distinct divisions (Kawamura et al., 2005; Kawamura et al., 2008). In Ripply2^–/– mice, the spinal column and ribs were fused, and the caudal myotome had indefinite segmental borders, indicative of defects in early somite segmentation (Chan et al., 2007; Morimoto et al., 2007). As other Ripply genes appear to function in boundary formation, we hypothesized that Ripply3 might set or refine the posterolateral border of the PPE.

Ripply3 is spatially positioned to regulate the posterolateral boundary of the PPE because it is expressed lateral to known PPE markers, Fgf8, Eya1 and Six1, as well as Tbx1 (Figs 5, 6). Ripply3 knockdown led to a posterolateral expansion of Tbx1, Fgf8, Eya1 and Six1 expression boundaries at stage 18 (Fig. 9). RIPPLY2 knockdown caused a similar effect in the presomitic mesoderm, where it caused a shift in Delta2 and Thylacine1 expression (Chan et al., 2006). The blurring of somite boundaries led to the fusion of ribs and vertebral components (Chan et al., 2007; Morimoto et al., 2007). RIPPLY3 knockdown led to noticeable changes in placode marker expression in tailbud stage Xenopus embryos (supplementary material Fig. S10). The intricate expression patterns of the lateral line, epibranchial and trigeminal placodes became distorted, and the otocyst seemed devoid of an otic pit. However, transverse sections of these embryos showed that the otic vesicle is present in Ripply3 MO-injected embryos (supplementary material Fig. S11).

Ripply3 does not overlap with Six1 and Eya1; however, Fgf8 does (Figs 5, 6) and is known to regulate their expression in the PPE (Ahrens and Schlosser, 2005). Thus, the shifts observed in Six1 and Eya1 expression in Ripply3 MO embryos (Fig. 9) probably result from perturbation of the Fgf8 expression boundary.

Mutually inhibitory interactions between RAR and FGF signaling are a common theme in developmental biology (Diez del Corral et al., 2003; Maden, 2006; Niederreither and Dolle, 2008). RA signaling plays crucial roles in the developing lens and otic placodes, in part by regulating Fgf8 (Bhasin et al., 2003; Mic et al., 2004; Romand et al., 2006; Song et al., 2004) and Fgf8 regulates components of RA biosynthesis (Mercader et al., 2000; Schneider et al., 2001; Shiotsugu et al., 2004). We have previously explored the relationship of FGF and RARα in the central nervous system in the Xenopus laevis embryo (Shiotsugu et al., 2004) and showed that loss of RARα shifted the border of Fgf8 expression. Knocking down Ripply3 phenocopied this effect (Fig. 9I,J), suggesting that RIPPLY3 normally restricts the borders of Fgf8 expression. Because RARα2 is required for Ripply3 expression (Fig. 4), we hypothesized that Ripply3 expression is lost in the absence of RAR, and is unable to restrict Fgf8 expression.

Double whole-mount in situ hybridization revealed that Ripply3 does not directly overlap with Fgf8. As Tbx1 regulates Fgf8 (discussed above), we believe that RIPPLY3 regulates Fgf8 expression via TBX1. Previous studies showed that Ripply proteins convert T-box proteins from transcriptional activators into repressors. Tbx6 activated the Thylacine1 promoter and this activation was blunted by co-transfection with Bowline/Ripply2 and Groucho (Kondow et al., 2007). Similarly, Tbx24-mediated repression of the Mesp-b promoter was conferred by co-expression of Ripply1/2/3 in luciferase assays (Kawamura et al., 2008). This effect was dependent on the WRPW (required for the Groucho interaction) and FPVQ (required for the T-Box interaction) motifs in RIPPLY, as well as HDAC activity (Kawamura et al., 2008).

We asked whether Tbx1 transcriptional activity could be reduced by RIPPLY3, and which domains of RIPPLY3 mediated this process. Whole-embryo luciferase assays showed that RIPPLY3 converted Tbx1 to a transcriptional repressor, and that this was dependent on the WRPW and FPVQ motifs of RIPPLY3 (Fig. 8). Interestingly, Tbx1 also promoted Ripply3 expression (Fig. 7). TBX1 knockdown reduced Ripply3 expression, which is consistent with the observation that Tbx6 is required for Bowline/Ledgerline/Ripply2 expression (Hitachi et al., 2008b). TBX6 binds the Bowline promoter at a T-Box Response Element (TBRE), in conjunction with THYLACINE1 and E47 to induce expression of Bowline (Hitachi et al., 2008b). Thus, in areas where TBX1 is co-expressed with Ripply3, their interaction converts TBX1 into a transcriptional repressor that restricts Fgf8 expression and defines the posterolateral boundary of the PPE. TBX1 further enhances this process by positively regulating Ripply3.

Although much is known about the transcriptional cascades that regulate placode identity and the subsequent development of the sensory organs, less is known about the molecular details underlying the initial establishment and patterning of the PPE. The PPE domain is defined specifically in time and space by a combinatorial pattern of signals (BMP, WNT, FGF) (Moody, 2007; Schlosser, 2006; Schlosser, 2008), and we propose that RA signaling through RARα is another important signal in PPE patterning. Accordingly, we further explored the transcriptional mechanisms downstream of RARα in the PPE by asking how RARα regulated expression of two PPE genes: Tbx1 and Ripply3. Fig. 11 summarizes the molecular interactions in PPE formation identified in this paper, together with those previously published by other laboratories.

Precise positioning of the PPE later translates to proper morphological identity of the placodes, thus early border formation is vital. RAR signaling was linked with the regulation of borders, such as in the delineation of somite and rhombomere boundaries, and Ripply/Bowline family members set borders in somitogenesis (Biris et al., 2007; Chan et al., 2007; Hitachi et al., 2008a; Kawamura et al., 2005; Kawamura et al., 2008; Moreno et al., 2008; Morimoto et al., 2007). A border-setting gene within the PPE was proposed to exist (Moody, 2007) and we provide in vivo evidence that this is likely to be Ripply3. Ripply3, which is regulated by RAR, sets the posterolateral border of the PPE, much as other Ripply genes do in somite development. It is likely that RAR induces both Ripply3 and Tbx1, which function together to regulate the spatial expression of genes, such as Fgf8, along the border of the PPE.

Ripply3 and Tbx1 are associated with Down Syndrome (DS) and DiGeorge Syndrome, respectively. Intriguingly, DiGeorge Syndrome (caused by a deletion in 22q11.2 containing Tbx1) and DS (caused by a trisomy in a large region of chromosome 21 that includes Ripply3) share common craniofacial and cardiovascular abnormalities (e.g. small malpositioned ears, flat facial profile, and atrial and septal defects), growth delays (owing to pituitary dysfunction), hearing loss, ocular abnormalities, immunodeficiency (owing to thymic aplasia), and learning disabilities (Korenberg et al., 1994; Ryan et al., 1997). Inner ear defects are common in DS, and most children with DS develop hearing loss (Balkany et al., 1979; Sacks and Wood, 2003; Shott et al., 2001). DS children also have a reduced ability to react to environmental cues and sensory stimuli. Although this is often attributed to cognitive problems related to sensory processing (Bruni et al., 2010; Wuang and Su, 2011), some evidence points to the fact that DS children have impaired tactile, vestibular, nociception and proprioceptive senses (Brandt, 1996; Brandt and Rosen, 1995; Chen and Fang, 2005; Shumway-Cook and Woollacott, 1985). This suggests that early placode defects occur in DS children. Considering the interactions, other researchers and we have demonstrated between RIPPLY3 and TBX1 and the similar phenotypes elicited by Tbx1 loss of function and Ripply3 gain of function (Fig. 10), a plausible hypothesis is that the signaling pathways underlying these two diseases converge at the level of TBX1 and RIPPLY3 interaction. Future work will explore the molecular basis of this interaction in vivo and place RIPPLY3, TBX1 and RA signaling in the context of human placode development.

We thank Dr Ira Blitz for critical comments on the manuscript. We also thank UCI undergraduate students Sophia Liu and Rachelle Abbey for outstanding technical help in the execution of some experiments.