FGF signaling establishes the anterior border of the Ciona neural tube

The generation of diverse cell types from a fertilized egg remains a fundamental problem for developmental biologists. The simple chordate Ciona intestinalis is well suited to address this question owing to its small cell number, detailed cell lineage characterization, experimental tractability and provisional gene regulatory networks (Passamaneck and Di Gregorio, 2005; Satou et al., 2009). Ciona, an ascidian, belongs to the urochordates, which represent the closest living relatives of the vertebrates (Delsuc et al., 2006). Ascidians are sessile, filter-feeding animals with incurrent and excurrent siphons that develop following the settlement and metamorphosis of a swimming tadpole larva. Ascidians diverged from vertebrates before the two rounds of genome duplication that created the large multigene families that complicate genetic analysis in vertebrates (Dehal and Boore, 2005).

The Ciona central nervous system (CNS) develops via neurulation of cells within the neural plate, forming a simple brain, which is also called the sensory vesicle, and a caudal nerve cord (Nicol and Meinertzhagen, 1988; Lemaire et al., 2002). Ciona also has neurogenic tissues that derive from the anterior neural plate but do not undergo neurulation and instead contribute to the peripheral nervous system (PNS) (Meinertzhagen et al., 2004). The corresponding region in vertebrate embryos forms the anterior placodes, namely the lens, olfactory and adenohypophyseal placodes (Toro and Varga, 2007; Streit, 2008; Schlosser, 2010). The anteriormost neural plate derivatives in Ciona are the adhesive papillae, or palps. The palps are three bumps on the rostral end of the tadpole that contain sensory neurons that project to the brain (Imai and Meinertzhagen, 2007b). The palps secrete adhesive substances that secure attachment of the tadpole to a solid substrate in preparation for the metamorphosis that creates the sessile adult body plan. Morphological, embryological and molecular evidence suggest that ascidian palps represent a rudimentary placode (Manni et al., 2004; Mazet et al., 2005). Another placode-like structure, the stomodeum, arises from the neural plate territory sandwiched between the palp placode and the anterior neural tube (Manni et al., 2005; Christiaen et al., 2007). The stomodeum gives rise to the oral siphon (mouth) of the adult and derives from the anterior neuropore (Veeman et al., 2010). In vertebrates, the stomodeum invaginates to form Rathke’s pouch and, ultimately, the anterior pituitary.

Fibroblast growth factor (FGF) signaling plays an important role in neural induction and subsequent development throughout the vertebrates (Ribisi et al., 2000; Streit et al., 2000; Dono, 2003; Delaune et al., 2005; Paek et al., 2009). Likewise, the FGF signaling pathway also functions in neural development in ascidians. In Ciona, neural induction begins with Otx expression at the 32-cell stage in the anterior neurectoderm via GATA and Ets transcription factors in response to FGF signaling (Bertrand et al., 2003). The requirement for FGF signaling persists in later development, when it is required for patterning posterior components of the nervous system (Hudson et al., 2003; Imai et al., 2009; Stolfi and Levine, 2011). Thus, FGF signaling acts at both early and late stages of Ciona nervous system development.

Much progress has been made in recent years toward uncovering the gene regulatory networks that control the development of the posterior nervous system. Less is known about the early patterning mechanisms in the anteriormost nervous system. Here, we examine the role of FGF signaling in a cell fate choice between the palp placode and the anterior CNS, which express FoxC and ZicL, respectively. We find that FGF-MAPK signaling is required to establish the posterior boundary of the FoxC expression domain, as abrogation of this pathway results in ectopic FoxC expression and loss of CNS markers. We identified an Ets binding site in a minimal FoxC enhancer that is required to repress FoxC in the CNS progenitors. Strikingly, we observe a truncated anterior neural tube in tailbud stage embryos that experience FGF-MAPK perturbation, as well as a posterior shift in the position of the neuropore. Thus, FGF signaling delineates the boundary between the palp placode and the anterior neural tube.

Ciona intestinalis adult animals were collected at Pillar Point Harbor in Half Moon Bay, California, or obtained from M-Rep (San Diego, CA, USA). Dechorionation, fertilization and electroporation of eggs were performed as described (Christiaen et al., 2009b; Christiaen et al., 2009c). Each experiment was performed at least twice. Electroporation was with 40-100 μg of each plasmid. Fluorescent images were obtained with a Zeiss Axio Imager.A2 upright fluorescence microscope or a Zeiss 700 laser-scanning confocal microscope.

The DMRT enhancer was amplified from Ciona intestinalis genomic DNA with DMRT –1k PstF (5′-GAAGTACTGCAGTAGTAGGGTGGAGGAAGATGGGAC) and DMRT NR (5′-AAAGCGGCCGCCATGCCAGTTAAACGAACTGTTTG); underlining indicates restriction sites used for cloning. Enhancers for FoxC were isolated with FoxC –2.1 BF (5′-GAAGTAGGATCCGACTTTCAGGCCGCCTACGTAGGGTAAATA) and FoxC NR (5′-GAAGTAGCGGCCGCCATTATAGAGAATCAAACCAAGGATTCGGT). The ‘minimal’ FoxC enhancer was amplified with FoxC-375XhF (5′-GAAACTCGAGCAGGTGTTTCGTAAGGCGGC) and FoxC –50 BR (5′-TAATGGATCCAACTTACACTAACACTGCCCG). Enhancer fragments were cloned upstream of lacZ in the CesA vector (Harafuji et al., 2002) or into modified vectors in which lacZ was replaced with GFP, mCherry or derivatives thereof. The ZicL –1.5 kb enhancer and the FoxB enhancer have been described previously (Shi and Levine, 2008; Imai et al., 2009).

RNA was isolated from staged embryos using Trizol (Invitrogen) and reverse transcribed with oligo(dT) and Superscript II reverse transcriptase (Invitrogen). The FoxC coding sequence was amplified with FoxC cds NF (5′-GAAGTAGCGGCCGCAATGACAATGCAAATCCGGTACCCGTCGTCC) and FoxC cds ER (5′-GAAGTAGAATTCTCAGTACTTAGTGTAATCGTAAGCAG) and cloned downstream of the DMRT enhancer. The RorA coding sequence was amplified with RorA NF (5′-GAAGAAGCGGCCGCAATGTTTCGTCATACCGGTTACGTC) and RorA ER (5′-ATAGAATTCCGTGGTCTGGTTTGCATCGCGAC) and cloned into the pCRII (Invitrogen) dual promoter vector to prepare the in situ probe. ELK coding sequence oligos were ELK cds PspF (5′-ATAAGGGCCCATGATGACATTGAAAATCGACACAG) and ELK cds MfeR (5′-GAATACAATTGTTACGATGAAGTTAGTGTCGTGACGG). The ZicL coding sequence was gift from L. Christiaen (New York University, USA) and subcloned downstream of the DMRT enhancer using NotI and BlpI restriction sites. The DN-FGFR and CA-FGFR transgenes (Davidson et al., 2006; Shi and Levine, 2008) were subcloned downstream of the DMRT enhancer by standard methods.

Ets sites were deleted from the FoxC minimal enhancer using the QuikChange Lightning Mutagenesis Kit (Stratagene/Agilent) with delEts-295 F (5′-CATTAGCGCGCCATTGCGCGAAAGTTGATTGG), delEts-295 R (5′-CCAATCAACTTTCGCGCAATGGCGCGCTAATG), delEts-255 F (5′-GATTATGACGCTCCTGCTATTGTTTAAGGGGAGATG) and delEts-255 R (5′-CATCTCCCCTTAAACAATAGCAGGAGCGTCATAATC).

The following in situ hybridization probes were obtained from plasmids contained in the Ciona intestinalis gene collection release I: FoxC (GC44e14), ZicL (GC08g08), ELK (GC27g08), Ephrin A-d (GC01j20), Epi1 (GC25g21) and Six3/6 (GC11m13). PCR amplification of transcription templates was with the following oligos: BSK-MT (5′-GCAAGGCGATTAAGTTGGGTAACG) and M13R (5′-CAGGAAACAGCTATGAC). DIG-labeled probes were transcribed with T7 RNA polymerase (Roche) using DIG labeling mix (Roche) and purified with the RNeasy Mini Kit (Qiagen).

Double fluorescent in situ hybridization and β-galactosidase (β-gal) antibody staining was performed as follows. Mid-gastrula stage embryos were fixed in MEM-PFA (4% paraformaldehyde, 0.1 M MOPS pH 7.4, 0.5 M NaCl, 1 mM EGTA, 2 mM MgSO₄, 0.05% Tween 20) overnight at 4°C. Embryos were rinsed in cold PBS and dehydrated as described (Christiaen et al., 2009d) and kept at –20°C. Prior to rehydration, embryos were treated with 2% H₂O₂ in methanol for 30 minutes at room temperature, and then rehydrated and manipulated as described through the post-hybridization washes (Christiaen et al., 2009d). Embryos were then equilibrated in TNT (100 mM Tris pH 7.5, 150 mM NaCl, 0.1% Tween 20) and blocked in TNB (0.5% Roche blocking reagent in 100 mM Tris pH 7.5, 150 mM NaCl) for 1-2 hours. Anti-β-gal (Promega) and anti-DIG-POD (Roche) antibodies were diluted 1:1000 in TNB and incubated with embryos overnight at 4°C. Embryos were washed extensively in TNT. For tyramide signal amplification, Cy3-coupled or fluorescein-coupled TSA Plus reagent (PerkinElmer) was diluted 1:100 in amplification diluent and added to embryos for 5-10 minutes at room temperature. Embryos were washed in TNT and incubated with secondary antibody (anti-mouse Alexa Fluor 488 or Alexa Fluor 555, Molecular Probes) at 1:1000 dilution overnight at 4°C. Embryos were washed in TNT and equilibrated and mounted in 50% glycerol in PBS containing 2% DABCO (Sigma).

The MEK inhibitor U0126 (Promega) was resuspended in DMSO to prepare a 10 mM stock, and diluted in filtered artificial sea water (FASW) to 10 μM working concentration. Embryos at the 76-cell stage were transferred to FASW containing U0126 or DMSO as control.

For antibody staining, embryos were fixed in MEM-FA (3.7% methanol-free formaldehyde, 0.1 M MOPS pH 7.4, 0.5 M NaCl, 2 mM MgSO₄, 1 mM EGTA) for 30 minutes, rinsed several times in PBT (PBS with 0.1% Tween 20), and incubated with anti-dpERK antibody (Sigma M9692) at 1:1000 with 2% normal donkey serum (NDS) in PBT at 4°C overnight. Embryos were washed in PBT and then incubated with donkey anti-mouse secondary antibody (1:1000) coupled to Alexa Fluor 555 (Molecular Probes) in PBT with 2% NDS for 2 hours at room temperature, then washed in PBT. For phalloidin staining, phalloidin (Molecular Probes) was diluted 1:500 in PBT, and embryos were incubated for 2-4 hours at room temperature, followed by PBT washes. For Hoechst staining, Hoechst (Molecular Probes) was diluted 1:1000 in PBT, and embryos were incubated for 0.5-2 hours, then washed in PBT. All embryos were equilibrated and mounted in 50% glycerol in PBS containing 2% DABCO.

Neurons of the swimming tadpole can be traced to individual blastomeres in the neural plate, which reflect the ascidian lineage-based cell nomenclature of Conklin (Conklin, 1905). The neural plate in Ciona emerges at the mid-gastrula stage on the dorsal surface of the embryo in a characteristic grid of cells with eight columns and six rows (our schematic in Fig. 1B is simplified to exclude the lateral-most columns that derive from the posterior ectoderm). The posterior neural plate, comprising rows I and II, derives from the anterior vegetal lineage (denoted as A-line) and contribute to the caudal nerve cord, motor ganglion and posterior sensory vesicle (we use the term posterior brain for simplicity) (Imai et al., 2009). Cells in rows III-VI derive from the anterior ectodermal lineage known as the a-line. Progeny of row III contribute to the anterior sensory vesicle (hereafter called anterior brain), whereas cells of row IV contribute to the stomodeum as well as the anterior brain. Rows V and VI comprise the anteriormost neural plate, and the progeny of these cells contribute to the palps as well as to epidermal sensory neurons in the trunk (see Fig. 1B,D). It is prudent to note that the derivatives of rows V/VI generate neurons of the PNS, and that the CNS derives from cells in rows I-IV, which undergo primary neurulation. In vertebrates, the neural plate is defined on the basis of neurulation to form the CNS; the neurogenic tissues anterior to the plate are called placodes and their neurons contribute to the PNS. Thus, the simple Ciona neural plate, as classically defined, includes a neural plate proper as well as a placodal territory (rows V/VI).

The adhesive papillae of the Ciona tadpole derive from the a8.18 and a8.20 blastomeres of 112-cell stage embryos, which express the forkhead transcriptional repressor FoxC (Imai et al., 2006; Lamy et al., 2006). FoxC expression is maintained in the progeny of these cells, namely a9.35, a9.36, a9.39 and a9.40, which comprise the medial four columns of the anteriormost neural plate, rows V and VI (Fig. 1A,B). We have isolated a ∼2 kb FoxC enhancer upstream of the FoxC coding sequence and found it to recapitulate the native expression pattern in gastrula stage embryos (Fig. 1C). Posterior to the a8.18/a8.20 palp precursor cells are the a8.17/a8.19 cells, which contribute to anterior brain and stomodeum. The a8.17/a8.19 blastomeres express ZicL, but not FoxC. An enhancer for ZicL has been described (Shi and Levine, 2008). It drives expression in the a8.19/a8.17 cells, as well as in their lateral neighbor a8.25, and also in the vegetal neural progenitors (rows I/II) and mesodermal lineages (Fig. 1A-C). Thus, FoxC and ZicL and their enhancers are active in complementary patterns in the anterior neural plate and allow visualization of the boundary between the CNS and PNS.

The FoxC-expressing and ZicL-expressing cells derive from a common progenitor that expresses DMRT (Fig. 1B). Both FoxC and ZicL require DMRT for their activation (Imai et al., 2006; Tresser et al., 2010). DMRT is homologous to the Drosophila doublesex gene, which, in addition to a role in sex determination, also functions in the development of neurons controlling sex-specific behavior (Rideout et al., 2007; Rideout et al., 2010). In vertebrate embryos, DMRT genes mark anterior neural regions, including the nasal placode and presumptive forebrain (Winkler et al., 2004; Huang et al., 2005; Veith et al., 2006; Wen et al., 2009; Yoshizawa et al., 2011). In Ciona, DMRT is first expressed at the 64-cell stage in the a7.10, a7.9 and a7.13 blastomeres, which are progenitors of the anterior neural plate (a-line, rows III-VI) (Imai et al., 2006; Tresser et al., 2010). The medial two cells, a7.10 and a7.9, undergo a cell fate choice to produce daughters that express either ZicL posteriorly (a8.17/a8.19) or FoxC anteriorly (a8.18/a8.20) (see Fig. 1A,B). We have isolated a ∼1 kb enhancer upstream of the DMRT gene and used this cis-regulatory DNA to perturb gene expression and fluorescently label cells of the anterior neural plate (rows III-VI).

What determines the outcome of the FoxC-ZicL cell fate choice? The FGF signaling pathway functions in multiple cell fate choice events during nervous system development in Ciona embryos (Hudson et al., 2007; Picco et al., 2007; Stolfi and Levine, 2011; Stolfi et al., 2011). The Map kinase (MAPK) signaling cascade can mediate signal transduction downstream of the FGF receptor, and active MAPK signaling can be detected with an antibody that recognizes a dual-phosphorylated (dp) form of the extracellular regulated kinase 1/2 (ERK1/2) (Yao et al., 2000; Hudson et al., 2003; Hudson et al., 2007). To determine the timing and levels of MAPK activation in developing neural tissue, we stained staged embryos with dpERK antibody. These embryos were electroporated with a FoxB reporter plasmid (driving H2B:CFP or GFP) to label the vegetal neural precursors that express FGF9/16/20 (A8.7, 8.8, 8.15, 8.16) (Imai et al., 2002; Tassy et al., 2010); this also assists in correct identification of the cells of interest.

Shortly after the cell fate choice, at ∼4.75 hpf, we detected dpERK staining in both the FoxC-expressing and ZicL-expressing cells, as well as in the mother cells on the right-hand side of the embryo, which have yet to divide (Fig. 2Aa,b). The level of staining appeared uniform in all nuclei. By contrast, when we looked a little later in the cell cycle, at 5.17 hpf, we noticed that the dpERK signal appeared graded (Fig. 2Ba,c,d). The strongest signal was seen in the vegetal A-line neural cells, in which it appeared to be both nuclear and cytoplasmic. In the a8.17/a8.19 cells, however, the dpERK signal was concentrated in the nuclei, whereas the anteriormost FoxC-expressing cells, a8.18 and a8.20, showed no detectable staining. The non-neural ectoderm adjacent to the FoxC-expressing cells showed elevated staining, comparable to that seen in the A-line neural cells. As a control, we stained embryos treated with the MAPK pathway inhibitor U0126, and found that the dpERK antibody signal was abolished (supplementary material Fig. S1). These results suggest that the level of MAPK activation is differentially regulated between the FoxC-expressing and ZicL-expressing cells, and might be an important factor in determining the downstream gene expression program.

The localized dpERK we observed in the ZicL-expressing cells is consistent with reports that FGF-MAPK signaling is important for its activation (Imai et al., 2006). We perturbed FGF-MAPK signaling by misexpressing a dominant-negative FGF receptor (DN-FGFR, which lacks the intracellular tyrosine kinase domain) using the DMRT 5′ enhancer and assayed the activity of the ZicL and FoxC enhancers. Although the cell fate choice is complete at the 112-cell stage, we assayed expression at mid-gastrula stage because the shape of the embryo makes it easier to visualize the cells of interest. DN-FGFR expression led to loss of ZicL in rows III/IV of the neural plate, although sometimes cells in the lateral-most columns escaped repression (compare Fig. 3Aa with 3Ab). Additionally, we observed ectopic expression of FoxC in rows III/IV of the neural plate, as compared with the control, in which FoxC is expressed only in rows V and VI (compare Fig. 3Ac with 3Ad). To confirm this result, we used the irreversible MAPK pathway inhibitor U0126 to block the MAPK signaling cascade downstream of the FGF receptor, and examined the expression of FoxC and ZicL transcripts. As with DN-FGFR, U0126 treatment led to loss of ZicL in rows III and IV (compare Fig. 3Ba with 3Bb), concomitant with ectopic FoxC expression (compare Fig. 3Bc with 3Bd). Expression of a constitutively active (CA) form of FGFR had no effect on the FoxC expression pattern. However, co-expression of CA-FGFR and DN-FGFR was able to rescue the ectopic FoxC pattern observed with DN-FGFR alone (supplementary material Fig. S2). We conclude that FGF-MAPK signaling is required for ZicL expression and for setting the posterior boundary of the FoxC expression domain.

We next asked whether abrogation of MAPK signaling had a broader impact on neural plate patterning, beyond altered expression of FoxC and ZicL. Ephrin A-d, which encodes a membrane-localized signaling molecule, is co-expressed with FoxC in rows V and VI of the neural plate. Upon U0126 treatment, however, the Ephrin A-d expression pattern expands to include rows III and IV, as does FoxC (Fig. 4A,B). RorA, which encodes an orphan receptor tyrosine kinase (RTK) that modulates non-canonical Wnt signaling, shows a graded pattern with strong expression in rows V/VI and weaker signal in rows III/IV (Fig. 4C) (Auger et al., 2009). U0126 treatment disturbs the graded nature of the RorA pattern, which becomes uniform across rows III-VI (Fig. 4C,D). Six3/6, which encodes a homeodomain transcription factor, is normally expressed in row IV of the neural plate (Moret et al., 2005; Christiaen et al., 2007), but its expression is lost upon U0126 treatment (Fig. 4E,F). Likewise, the Ets family transcription factor gene ELK, a marker for rows III/IV, is lost with U0126 treatment (Fig. 4G,H). These results indicate that the MAPK signaling cascade is broadly required for proper neural plate patterning, in particular to establish the proper identity of cells in rows III and IV. These cells give rise to the anteriormost derivatives of the CNS, including the anterior brain and stomodeum.

Because inhibition of FGF-MAPK signaling led to ectopic FoxC and loss of ZicL expression, and because the lateral-most cells a9.49/a9.50 do not experience ectopic FoxC and sometimes escape repression of ZicL, we asked whether FoxC itself might repress ZicL. FoxC is known to act as a transcriptional repressor in various metazoans by virtue of a conserved Engrailed homology repressor domain (Yaklichkin et al., 2007). Interestingly, this conserved repressor motif is absent from the Ciona FoxC ortholog. Nonetheless, when full-length FoxC protein was misexpressed using the DMRT enhancer, ZicL was lost from rows III and IV (Fig. 5A). A truncated form of FoxC that lacks the C-terminal 293 amino acids downstream of the forkhead DNA-binding domain was also sufficient to repress ZicL (data not shown). These results suggest that, despite lacking an obvious transcriptional repression motif, FoxC can inhibit ZicL expression. Whether the repressive effect of FoxC on ZicL is direct or indirect will require further study.

We then asked whether ZicL could repress FoxC expression, reasoning that a mechanism of mutual repression could account for the establishment of the boundary between rows IV and V of the neural plate. However, misexpression of ZicL had no effect on the endogenous FoxC expression pattern (Fig. 5B). We verified this result by misexpressing ZicL with the Bmp2b enhancer, which is active much earlier, at the 16-cell stage, and should allow sufficient time to accumulate transgene expression (Christiaen et al., 2009a). Again, ectopic ZicL expression had no effect on FoxC (data not shown).

How does MAPK signaling keep FoxC repressed in rows III/IV of the neural plate? Ets family transcription factors are known effectors of FGF-MAPK signaling and can act as transcriptional repressors in various contexts (Oikawa and Yamada, 2003; Hollenhorst et al., 2011). We identified a ‘minimal’ FoxC enhancer by making 5′ and 3′ deletions of the original 2 kb fragment, to facilitate analysis of potential transcription factor binding sites. We isolated a 325 bp fragment that maps from –375 to –50 bp relative to the start codon, and cloned it upstream of a heterologous promoter (Rothbacher et al., 2007) (from the FOG gene) to drive lacZ expression. This minimal FoxC enhancer is conserved in the related species C. savignyi and was sufficient to drive a wild-type FoxC expression pattern (Fig. 6B).

We then searched for Ets sites in the minimal FoxC enhancer. Ets family transcription factors recognize a well-defined core consensus sequence: GGAA/T (Hollenhorst et al., 2011). We found two candidate sites that map to –255 bp and –295 bp relative to the native start codon and deleted them (Fig. 6A). Deletion of the Ets site at –295 caused an otherwise normal FoxC-lacZ fusion gene to exhibit derepressed expression in rows III and IV of the neural plate (Fig. 6B,C). Deletion of the Ets site at –255 had no effect on reporter activity (Fig. 6D). This result suggests that an Ets family transcription factor might repress FoxC in rows III/IV of the neural plate.

The Ciona genome encodes ∼15 Ets domain transcription factors. ELK was an intriguing candidate repressor of FoxC because it is expressed in rows III/IV of the neural plate in a MAPK-dependent manner (Fig. 4D). Furthermore, it has been documented that ELK can repress transcription in response to MAPK signaling (Usenko et al., 2002). However, when we misexpressed ELK with the DMRT enhancer, FoxC expression was unaffected (data not shown). Thus, although our data collectively suggest that an Ets family factor might function to repress FoxC expression in rows III/IV, the identity of such a factor remains unknown.

We next looked at tailbud stage embryos that developed under conditions of perturbed FGF signaling to examine the morphology of the sensory vesicle and palp placode. We predicted that the ectopic FoxC might lead to the specification of ectopic palps at the expense of anterior brain. In control embryos, cells of the anterior brain are marked by co-expression of DMRT and ZicL (pink cells in Fig. 7A), whereas cells expressing DMRT but not ZicL (row V/VI derivatives) label the anterior trunk ectoderm (palp placode and progeny of a9.51/a9.52; blue cells in Fig. 7A; see also Fig. 1B). In the presence of DN-FGFR, we observe a dramatic expansion of the anterior epidermis at the expense of anterior brain cells, as evidenced by a larger proportion of DMRT⁺ cells lacking ZicL expression (Fig. 7B). This was accompanied by an expansion of epidermis marker gene expression (supplementary material Fig. S3). The few remaining cells that co-express DMRT and ZicL derive from the lateral columns of the neural plate, a9.49/a9.50, which do not experience ectopic FoxC when FGF signaling is perturbed (Fig. 3Ad,Bd).

Phalloidin staining of DN-FGFR-expressing tailbud embryos reveals a smaller sensory vesicle, which is rounded rather than elliptical in shape (Fig. 7C,D). This truncated brain is entirely derived from vegetal neural plate cells, as evidenced by the lack of DMRT expression (Fig. 7D). DMRT⁺ cells remain on the surface of the embryo, rather than being internalized to form the brain vesicle (compare DMRT pattern in Fig. 7C and 7D). Interestingly, we observed a neuropore positioned on top of the truncated brain; note that this neuropore is shifted posteriorly compared with that of the control (asterisks, Fig. 7C,D). Taken together, these results suggest that, normally, the boundary between the palp placode and the CNS corresponds to the boundary between the FoxC and ZicL expression domains. Shifting the domain of FoxC expression posteriorly induces a similar posterior shift of the palp placode.

We next looked at tailbud embryos misexpressing FoxC to determine whether this manipulation recapitulates the DN-FGFR phenotype. Fig. 7E shows that FoxC misexpression indeed leads to truncation of the anterior brain (compare Fig. 7D and 7E), suggesting that FoxC expression is sufficient to promote a placodal rather than CNS cell fate. The FoxC-misexpressing cells, labeled with the DMRT reporter, remain in a columnar epithelium on the dorso-anterior surface of the embryo. Notably, we found no evidence of anterior neuropore formation. The DMRT enhancer drives FoxC expression across all six columns of the anterior neural plate, whereas the DN-FGFR perturbation leads to ectopic FoxC only in the medial four columns. This suggests that proper specification of the lateral-most columns of the neural plate, in part through exclusion of FoxC, is necessary for neuropore formation (Nicol and Meinertzhagen, 1988).

Ectopic expression of FoxC via inhibition of FGF signaling does not cause transformation of the anterior brain into supernumerary palps. On the contrary, palp development is impaired (Fig. 7F). This is in agreement with a previous study showing that MAPK signaling is required at least through neurula stages for proper palp development (Hudson et al., 2003). Thus, although FoxC is a marker of the palp lineage, its expression alone does not suffice to direct morphogenesis of the palps. Our results show that abrogation of MAPK signaling causes a posterior shift in the boundary between the palp placode and the presumptive stomodeum, and suggest that FoxC expression might function in part to position the anterior neuropore (see below).

We have presented evidence that FGF-MAPK signaling is required to specify anterior CNS fates (ZicL) in rows III/IV of the neural plate, and that disruption of this pathway causes derepression of FoxC, which is normally restricted to rows V/VI (Fig. 3). Inhibition of FGF-MAPK signaling also causes the loss of other determinants of the anterior neural tube, including Six3/6, which marks the presumptive stomodeum (Fig. 4) (Mazet and Shimeld, 2005; Christiaen et al., 2007). A similar situation is seen in the posterior neural plate comprising rows I/II, which derive from the vegetal A-line and contribute to the posterior CNS, including the motor ganglion and caudal nerve cord. Hudson et al. showed that MAPK signaling is required for the expression of row I markers such as Mnx, Cdx and Ephrin A-b; the loss of row I markers in U0126-treated embryos was accompanied by the posterior expansion of row II markers into row I (Hudson et al., 2007). Thus, FGF-MAPK signaling broadly promotes posterior cell fates over anterior cell fates across the developing nervous system in Ciona. This trend is also seen in vertebrates, where neural induction generally promotes nervous tissue of anterior character and FGF signaling acts on this specified tissue to induce posterior fates (Gamse and Sive, 2000; Ribisi et al., 2000; Stern, 2001; Dono, 2003; Fletcher et al., 2006).

FoxC maintains differential expression of ZicL and FoxC by inhibiting ZicL expression in rows V and VI. Ectopic expression of FoxC in rows III and IV results in the loss of ZicL. Thus, regulatory interactions between FoxC and ZicL appear to be hierarchical, rather than mutual. That is, FoxC inhibits ZicL, but misexpression of ZicL in rows V and VI does not alter the expression of FoxC (Fig. 5). Hierarchical repression is observed in the dorsal-ventral patterning of the Drosophila melanogaster CNS, with ventral repressors such as Snail and Vnd inhibiting the expression of more dorsal determinants such as Ind and Msh (Drop), but not vice versa (Cowden and Levine, 2003). By contrast, anterior-posterior patterning depends on mutual cross-repression, particularly among pairs of gap genes (hunchback/knirps and giant/Kruppel) (Kraut and Levine, 1991).

The palps derive from an embryonic territory sandwiched between the non-neural ectoderm and progenitors of the CNS (stomodeum and anterior brain). The corresponding region in vertebrate embryos forms the anterior placodes, which give rise to the lens, olfactory epithelium and adenohypophysis (anterior pituitary) (Streit, 2008; Schlosser, 2010). In chick embryos, FoxC is expressed in the lens placode, and in mammals it functions in the development of the anterior eye (Saleem et al., 2003; Bailey et al., 2006; Berry et al., 2006). FoxC is also expressed in the developing Xenopus eye, although it is not clear whether this expression is in the lens placode and/or the surrounding tissue (Bowes et al., 2010). It is noteworthy that FGF signaling functions in the chick to repress lens fate and promote olfactory placode development (Bailey et al., 2006). A parallel situation is seen in the Ciona palp placode, which expresses FoxC and β/γ-crystallin, another lens marker (Shimeld et al., 2005), and is repressed by FGF signaling.

We do not suggest homology of the Ciona palp placode with the vertebrate lens placode. Ciona has a light-sensing apparatus that consists of a pigmented ocellus and numerous photoreceptor cells, and these are located in the anterior brain (Imai and Meinertzhagen, 2007a; Horie et al., 2008). The palp placode also expresses olfactory placode markers such as DMRT, Coe and Dlx genes (Caracciolo et al., 2000; Winkler et al., 2004; Huang et al., 2005; Mazet et al., 2005; Veith et al., 2006; Wen et al., 2009). Thus, we refer to the Ciona palp primordium simply as an anterior placode, which might be derived from an ancestral structure that evolved into separate, specialized anterior placodes in the vertebrate lineage.

In summary, the boundary between rows IV and V of the Ciona neural plate is important because it delineates the neural tube as distinct from the anterior placode region. We have shown that FGF-MAPK signaling is required to establish this boundary and that it does so by repressing FoxC expression in the presumptive brain. Disruption of FGF signaling leads to a truncated anterior brain, reminiscent of the phenotype of triple FGFR knockout mice, which lack the telencephalon (Paek et al., 2009). Ciona embryos develop from defined cell lineages, and the gene expression patterns that underlie important embryological boundaries form relatively early in development. In vertebrates, neural territories are composed of mixed fate cell populations, which segregate and form boundaries according to fate at comparatively later developmental stages in response to diverse signals, including FGFs (Puelles et al., 2005; Toro and Varga, 2007; Cajal et al., 2012; Sanchez-Arrones et al., 2012). Despite these differences, a common requirement for FGF signaling highlights the relevance of Ciona embryos to the study of patterning and the morphogenetic mechanisms underlying vertebrate development.

We thank Alberto Stolfi for critical reading of the manuscript and sharing many reagents and Lionel Christiaen for helpful comments and the Bmp2b>ZicL plasmid.