Activation of Pax3 target genes is necessary but not sufficient for neurogenesis in the ophthalmic trigeminal placode

Summary

Vertebrate cranial neurogenic placodes are relatively simple model systems for investigating the control of sensory neurogenesis. The ophthalmic trigeminal (opV) placode, for which the earliest specific marker is the paired domain homeodomain transcription factor Pax3, forms cutaneous sensory neurons in the ophthalmic lobe of the trigeminal ganglion. We previously showed that Pax3 expression in avian opV placode cells correlates with specification and commitment to a Pax3+, cutaneous sensory neuron fate. Pax3 can act as a transcriptional activator or repressor, depending on the cellular context. We show using mouse Splotch^2H mutants that Pax3 is necessary for the normal neuronal differentiation of opV placode cells. Using an electroporation construct encoding a Pax3-Engrailed fusion protein, which represses Pax3 target genes, we show that activation of Pax3 target genes is required cell-autonomously within chick opV placode cells for expression of the opV placode markers FGFR4 and Ngn2, maintenance of the preplacodal marker Eya2, expression of Pax3 itself (suggesting that Pax3 autoregulates), neuronal differentiation and delamination. Mis-expression of Pax3 in head ectoderm is sufficient to induce FGFR4 and Ngn2 expression, but neurons do not differentiate, suggesting that additional signals are necessary to enable Pax3+ cells to differentiate as neurons. Mis-expression of Pax3 in the Pax2+ otic and epibranchial placodes also down-regulates Pax2 and disrupts otic vesicle closure, suggesting that Pax3 is sufficient to alter the identity of these cells. Overall, our results suggest that activation of Pax3 target genes is necessary but not sufficient for neurogenesis in the opV placode.

Introduction

Vertebrate cranial placodes are bilateral patches of thickened head ectoderm that form the paired peripheral sense organs (olfactory epithelia, inner ears, lateral line system), the eye lenses, and a wide variety of sensory neurons in cranial sensory ganglia, as well as the endocrine adenohypophysis (reviewed in Baker and Bronner-Fraser, 2001; Schlosser, 2006). They provide a relatively simple, accessible set of model systems in which to investigate the control of sensory neurogenesis and, ultimately, the generation of different neuronal subtypes.

Cranial placodes originate from a “pre-placodal region” at the border between the prospective anterior neural plate and epidermis, characterised by expression of specific members of the Six and Eya familes of transcription factors/co-factors (reviewed in Baker and Bronner-Fraser, 2001; Rebay et al., 2005; Schlosser, 2006; Streit, 2007). Paired domain transcription factors of the Pax family, which can act in complex regulatory networks with Six, Eya and Dach (reviewed in Pappu and Mardon, 2004; Schlosser, 2006; Streit, 2007), are also expressed in different subsets of cranial placodes: Pax6 in the olfactory and lens placodes; Pax3 in the ophthalmic trigeminal placode; Pax2 and Pax8 in the otic and epibranchial placodes (reviewed in Baker and Bronner-Fraser, 2001; Schlosser, 2006), leading to the suggestion that Pax genes may be important for aspects of individual placode identity (Baker and Bronner-Fraser, 2000; Baker and Bronner-Fraser, 2001; Streit, 2004).

Here, we investigate the role of Pax3 in the development of the ophthalmic trigeminal (opV) placode, which forms cutaneous sensory neurons in the ophthalmic lobe of the trigeminal ganglion. In the chick, Pax3 is the earliest specific marker for the chick opV placode, being broadly expressed from the 7–8 somite-stage in caudal midbrain/rostral hindbrain-level ectoderm that is fated to form opV placode-derived neurons (Stark et al., 1997; Baker et al., 1999; Xu et al., 2008).

Pax3 is expressed in the opV placode in all vertebrates analysed (Stark et al., 1997; Schlosser and Ahrens, 2004; O'Neill et al., 2007). Using explant culture and heterotopic grafting experiments, we previously showed that Pax3 expression in avian opV placode ectoderm correlates with commitment to a Pax3+ cutaneous sensory neuron fate (Baker et al., 1999; Baker and Bronner-Fraser, 2000; Baker et al., 2002), but the precise role of Pax3 in opV placode development was unclear. Pax3 has a paired domain, a paired-type homeodomain, and a C-terminal transactivation domain; it generally acts as a transcriptional activator during development (Relaix et al., 2003), though it can repress some target genes (e.g. (Kioussi et al., 1995; Kwang et al., 2002). Homozygous and heterozygous Pax3 mutant phenotypes in Splotch mice show that Pax3 is required in different cells for functions as diverse as migration, differentiation, and survival (reviewed in Machado et al., 2001; Chi and Epstein, 2002). Heterozygous Splotch mutants are viable, with patchy pigmentation due to defective development of neural crest-derived melanocytes, while homozygous Splotch mutations are mid-gestation lethal with defects in neural tube closure, somite derivatives, limb muscles and trunk neural crest cell migration (Auerbach, 1954; Epstein et al., 1991; Epstein et al., 1993; Schubert et al., 2001). Pax3 has a wide variety of developmental roles, being required for the delamination and migration of hypaxial and limb muscle precursor cells (reviewed in Buckingham and Relaix, 2007), for the migration of trunk neural crest cells (Serbedzija and McMahon, 1997; Mansouri et al., 2001), for the differentiation of muscle cells (Tajbakhsh et al., 1997; Relaix et al., 2003; Bajard et al., 2006) and neural crest-derived melanocytes (Watanabe et al., 1998; Lang and Epstein, 2003; Lang et al., 2005), and for cell survival in the somites (Borycki et al., 1999).

The ophthalmic nerve in Splotch embryos is missing or reduced at embryonic day 12.5 (Tremblay et al., 1995), suggesting a role for Pax3 in development of the ophthalmic lobe of the trigeminal ganglion, from which the ophthalmic nerve projects. However, the ophthalmic lobe contains not only Pax3+ opV placode-derived neurons, but also neural crest-derived neurons which originate from Pax3+ neural crest precursor cells (Baker et al., 2002), while the opV placode is itself induced by a signal(s) from the Pax3+ dorsal neural tube (Stark et al., 1997; Baker et al., 1999). Hence, it is unclear which of these Pax3+ cell populations requires Pax3 function to enable proper formation of the ophthalmic nerve. Here, by analysing trigeminal ganglion development in Splotch^2H embryos, and by using in ovo electroporation to manipulate Pax3 function specifically in surface ectoderm, we show that Pax3 is necessary but not sufficient for neurogenesis in the opV placode.

Materials and Methods

Mouse embryos

Sp^2H mice were maintained as a random-bred colony, and genotyped by PCR using primers to Pax3 (Epstein et al., 1991). Mice were killed by cervical dislocation and embryos dissected from the uterus in Dulbecco’s Modified Eagle’s Medium (Invitrogen), rinsed in diethylpyrocarbonate (DEPC)-treated phosphate-buffered saline (PBS) and fixed in cold 4% paraformaldehyde in DEPC-PBS. After extensive washing in cold DEPC-treated PBS, they were dehydrated into 100% methanol and stored at −20ºC. Embryos were genotyped by PCR using primers to Pax3 (Epstein et al., 1991) on DNA extracted from tail samples.

In situ hybridisation

Wholemount in situ hybridisation on mouse embryos was performed as in O'Neill et al. (2007) with minor modifications; a detailed protocol is available on request. In situ hybridisation on paraffin wax sections was performed as described in Lassiter et al. (2007).

Mouse Ngn1, Ngn2 and SCG10, and chicken Ngn2, were kind gifts of David Anderson (Caltech, Pasadena, USA). Chicken Pax2 was a kind gift of Dr. Andrea Streit (King’s College London, UK). A 400 bp fragment of chicken (Gallus gallus) FGFR4 cDNA, corresponding to base-pairs 54–454 of chicken FGFR4 (GenBank accession number AF083063), was PCR-amplified from a cDNA library of homogenised 3–10 somite-stage chick embryos using degenerate primers (5′-GGAGATGGAGCCAGACTCG-3′ and 5′-ACCTCTCCAGCACRTCCA-3′) and cloned into the pGEM-T easy vector (Promega).

Immunohistochemistry and confocal microscopy

The following primary antibodies were used: GFP (rabbit; Molecular Probes/Invitrogen); Islet1 (mouse IgG2b; Developmental Studies Hybridoma Bank: DSHB), NeuN (mouse IgG1; Chemicon), Pax2 (rabbit; Zymed/Invitrogen), Pax3 (mouse IgG2a; kind gift of Marianne Bronner-Fraser, Caltech, Pasadena; also available from the DSHB), TuJ1 (mouse IgG2a; Covance/Invitrogen). The Developmental Studies Hybridoma Bank was developed under the auspices of the NICHD and is maintained by the University of Iowa, Department of Biological Sciences, Iowa City. Appropriately matched Alexa488- or Alexa594-conjugated goat anti-mouse or goat anti-rabbit secondary antibodies, and Alexa350-NeutrAvidin, were obtained from Molecular Probes/Invitrogen; biotinylated goat anti-mouse secondary antibodies were obtained from Southern Biotech.

Immunohistochemistry on sections was performed as described in Lassiter et al. (2007). The whole-mount immunostaining protocol for mouse embryos was modified from Guris et al. (2001). Serial sagittal images of TuJ1-immunostained mouse embryo heads were collected to a depth of about 200μm (2.8μm spacing) using an inverted laser scanning LEICA DM IRBE confocal microscope. Composite images showing the maximum intensity projection of the stack were prepared using ImageJ software (National Institutes of Health).

Electroporation constructs

All constructs were prepared using an Endotoxin-free Maxi preparation kit (Qiagen) according to the manufacturer’s instructions and used at 5.0 μg/μl. The mPax3En-nGFP electroporation construct was made in the pCIG vector (Megason and McMahon, 2002). It encodes the first 374 amino acids of mouse Pax3 fused to the Drosophila Engrailed repressor domain (Bajard et al., 2006; Relaix et al., 2006), with IRES-nuclear localized green fluorescent protein (GFP). The empty pCIG vector, encoding nuclear GFP only (nGFP), was used for control electroporations.

The cPax3/GFP construct was made by inserting full-length chicken Pax3 cDNA (Matsunaga et al., 2001) into a modified version of the pCL-GFP electroporation vector (Scaal et al., 2004), in which the CMV promoter and associated intron sequence were replaced with the RSV promoter sequence from pRc/RSV (Invitrogen). GFP expression from this construct is under the separate control of the SV40 promoter. The empty vector (GFP) was used for control electroporations.

In ovo electroporation

Fertilized chicken (Gallus gallus) eggs from local farms were incubated in a humidified atmosphere at 38ºC until the desired stage. Black ink (Fount India, Pelikan) was diluted to 10–20% in Howard Ringer's solution and injected beneath the blastoderm to visualise the embryo. Electroporation constructs (5.0 μg/μl) were introduced into surface ectoderm by in ovo electroporation as described in Lassiter et al. (2007).

Results

Pax3 is necessary for normal neuronal differentiation in the mouse ophthalmic trigeminal (opV) placode

In both chick and mouse, trigeminal placode-derived neurons differentiate long before neural crest-derived neurons (Verwoerd and van Oostrom, 1979; D'Amico-Martel and Noden, 1980; Nichols, 1986; Covell and Noden, 1989; Moody et al., 1989; Stainier and Gilbert, 1990). Hence, one way to distinguish between Pax3 function in opV placode-derived neurons versus trigeminal neural crest-derived neurons is to examine neuronal differentiation in Splotch (Pax3-deficient) embryos at a stage when only placode-derived neurons have differentiated. Cells begin delaminating from the mouse trigeminal placodes at E9 (10–12 somite-stage [ss]); almost all delamination is completed by E10 (30ss) (Nichols, 1986). The first placode-derived neurons differentiate at E9 (14ss), while a second wave of neurogenesis, presumed to be neural crest-derived, is seen from E10 (30ss), lasting for 2 days (Stainier and Gilbert, 1990).

We performed whole-mount TuJ1 immunostaining to detect neuronal beta-III tubulin in wildtype and homozygous Splotch^2H (Sp^2H) embryos at E9 (17–18ss), when only placode-derived neurons have differentiated (Fig. 1A-D). Neurons in the ophthalmic lobe of the trigeminal ganglion are missing in Sp^2H/Sp^2H embryos at this stage (Fig. 1C,D). Hence, Pax3 is necessary for the differentiation of opV placode-derived neurons. Expression of the proneural basic-loop-helix transcription factor Neurogenin1 (Ngn1) (Ma et al., 1998) in Sp^2H/Sp^2H embryos, however, was essentially indistinguishable from wildtype, both at E9 (Fig. 1E,F) and also at E8.5 and E9.5 (data not shown). Thus, in Sp^2H/Sp^2H embryos at E9, opV placode-derived Ngn1+ neuronal precursors are present (Fig. 1F) but fail to differentiate as neurons (Fig. 1C,D). By late E9.5 (28ss), expression of the neuronal marker SCG10 (Anderson and Axel, 1985) revealed some recovery of neuronal differentiation in the ophthalmic lobe of Sp^2H/Sp^2H embryos, though this was still reduced relative to wildtype (Fig. 1G,H).

Fig. 1. Pax3 is necessary for normal neuronal differentiation of mouse ophthalmic trigeminal (opV) placode cells. (A-D).

At embryonic day 9 (E9), before neural crest-derived neurons have differentiated, neuronal beta-III tubulin expression (detected by the TuJ1 antibody) shows neurons in the ophthalmic lobe of the trigeminal ganglion (arrowheads) projecting towards the eye in a 17-somite wildtype embryo (green channel only in A; green channel superimposed on brightfield in B; higher-power images of the same embryo in A1,B1), but absent in a 17-somite homozygous Splotch^2H (Sp^2H) embryo (green channel only in C; green channel superimposed on brightfield in D; higher-power images of the same embryo in C1,D1). The same difference in phenotype was also seen at the 18-somite-stage (not shown). (E,F) At E9, Ngn1 is expressed in the ophthalmic lobe of the trigeminal ganglion (arrowheads) in both wildtype (E) and homozygous Sp^2H embryos (F). (G,H) At late E9.5, expression of the neuronal marker SCG10 shows neurons in the ophthalmic lobe of the trigeminal ganglion (white arrowheads) and along the ophthalmic nerve projecting towards the eye (black arrowheads), both in wildtype embryos (G), and also, though in reduced numbers, in homozygous Sp^2H embryos (H). The lefthand trigeminal ganglion can be seen, out of focus, in both embryos.

E, embryonic day; ma, mandibular arch; ss, somite-stage.

Activation of Pax3 target genes is necessary for chick opV placode development

To investigate the requirement for Pax3 specifically during opV placode development, and to test the hypothesis that Pax3 acts as a transcriptional activator in the opV placode, we used an electroporation construct, mPax3En-nGFP, encoding a fusion protein between mouse Pax3 and the Drosophila Engrailed repressor (mPax3En) and nuclear GFP. mPax3En retains both Pax3 DNA-binding domains and has been shown to repress Pax3 target genes in mouse (Bajard et al., 2006; Relaix et al., 2006). The mouse Pax3 paired domain is 96.9% identical to the chick paired domain at the amino acid level, while the homeodomain is 100% identical, so we predicted that this fusion protein should compete for endogenous Pax3 binding sites and repress Pax3 target genes in the chick embryo. Electroporating this construct into cranial ectoderm enabled us to test the cell-autonomous requirement for activation of Pax3 target genes during the development of opV placode cells. nGFP was used as a negative control.

We electroporated either mPax3En-nGFP or nGFP into the cranial ectoderm of 4–7ss chick embryos, fixed the embryos 24 hours later at the 19–26ss, and performed in situ hybridisation on sections for the chick opV placode markers fibroblast growth factor receptor 4 (FGFR4; Marcelle et al., 1994; Stark et al., 1997) or Neurogenin2 (Ngn2; Begbie et al., 2002). Within the chick neurogenic placodes, FGFR4 is opV placode-specific (Stark et al., 1997); Ngn2 is initially opV placode-specific (Begbie et al., 2002), although by the 27ss it is also expressed in other placodes (Xu et al., 2008). We then immunostained the sections for GFP, Pax3, and the Lim-domain transcription factor Islet1, which is expressed in all trigeminal neurons and has been extensively used as a definitive marker for differentiating trigeminal neurons (Begbie et al., 2002; Fedtsova et al., 2003; Lassiter et al., 2007; Shiau et al., 2008; Xu et al., 2008). Islet1 is expressed before Brn3a in the opV placode (Begbie et al., 2002; Fedtsova et al., 2003; Lassiter et al., 2007; Shiau et al., 2008; Xu et al., 2008), and delaminating Islet1+ opV placode-derived neurons are apparently post-mitotic (Begbie et al., 2002). Using a nuclear neuronal antigen, rather than axonal or cytoplasmic neuronal antigens, also enabled us to count cells precisely in order to determine the proportion of mPax3En-nGFP or nGFP-targeted cells that differentiated as neurons.

We counted all GFP+ cells in the opV placode region of electroporated embryos to determine the proportion of targeted cells expressing a particular marker. Since electroporation is mosaic, the opV placode region could be identified by the presence of untargeted cells expressing Pax3, Ngn2 or FGFR4: GFP+ cells were only counted in sections containing such cells. For statistical purposes, we counted each opV placode with GFP+ cells as a separate targeting event; cells were counted across a minimum of 5 placodes from 5 embryos (see Table 1). [The same control cell counts are used here as for the experiments in Lassiter et al. (2007), in which the effects of blocking canonical Wnt signalling were investigated by electroporating dominant negative human TCF4 into the opV placode, followed by sectioning, in situ hybridisation for FGFR4 or Ngn2, and immunostaining for GFP, Pax3 and Islet1. All these experiments were performed at the same time.]

Table 1.

Cell counting data across all embryos electroporated in the opV placode with either mPax3En-nGFP (which represses Pax3 target genes) or nGFP as a control, and assayed a day later for FGFR4, Ngn2 or Islet1 expression, cell delamination, and expression of Pax3 itself. The mean and standard error about the mean (s.e.m.) are presented graphically in Fig. 3H.

Marker assayed	Construct	Age when electroporated/analysed (ss)	No. opV placodes counted (no. embryos)	No. GFP+ cells counted overall	Mean % GFP+ cells per placode +ve for marker	s.d.	s.e.m.
FGFR4	mPax3-En	6–7/21–25	7 (5)	554	1.6	2.0	0.8
	nGFP	5–7/20–26	6 (5)	305	26.9	12.3	5.0
Ngn2	mPax3-En	4–7/19–23	6 (5)	801	3.7	2.1	0.9
	nGFP	4–7/20–26	5 (5)	657	48.5	20.8	9.3
Islet1	mPax3-En	4–7/21–25	11 (8)	1110	1.1	1.7	0.5
	nGFP	4–7/20–26	10 (9)	894	14.3	12.5	4.0
Cell delam.	mPax3-En	4–7/20–26	13 (10)	1215	3.1	4.0	1.1
	nGFP	5–7/19–25	11 (10)	962	20.8	12.6	3.8
Pax3	mPax3-En	4–7/20–26	13 (10)	1215	4.3	4.4	1.2
	nGFP	4–7/19–25	11 (10)	962	40.8	12.8	3.9

Abbreviations: delam., delamination; mPax3-En, mPax3En-nGFP; No., number of; s.d., standard deviation; s.e.m., standard error of the mean; ss, somite stage; +ve, positive.

(NB The cell counts are lower for Islet1 than for delamination because the Islet1 immunostaining did not work on all slides.)

Very few mPax3En-nGFP-targeted opV placode cells expressed either FGFR4 (Fig. 2A,B; compare with Fig. 2C,D) or Ngn2 (Fig. 2E,F; compare with Fig. 2G,H). The mean percentages/placode of mPax3En-nGFP-targeted cells that expressed FGFR4 or Ngn2 were very significantly lower than for nGFP-targeted cells (p<0.0005, Student’s t-test; Table 1; Fig. 3H). Hence, activation of Pax3 target genes is necessary for the expression of both FGFR4 and Ngn2 in the opV placode, or these genes are directly downstream of Pax3 in these cells. Very few mPax3En-nGFP-targeted opV placode cells expressed Islet1 (Fig. 3A,B; compare with Fig. 3C). The mean percentage/placode of mPax3En-nGFP-targeted cells that expressed Islet1 was very significantly lower than for nGFP-targeted cells (p<0.005, Student’s t-test; Table 1; Fig. 3H). Similarly, very few mPax3En-nGFP-targeted opV placde cells delaminated into the mesenchyme (Fig. 2A,B,E,F; Fig. 3A,B; compare with Fig. 2C,D,G,H; Fig. 3C). The mean percentage/placode of mPax3En-nGFP-targeted cells that delaminated into the mesenchyme was very significantly lower than for nGFP-targeted cells (p<0.0005, Student’s t-test; Table 1; Fig. 3H). Hence, activation of Pax3 target genes is necessary for neuronal differentiation and delamination.

Fig. 2. Repressing Pax3 target genes in the chick ophthalmic trigeminal (opV) placode blocks expression of the opV placode markers FGFR4 and Ngn2.

All panels show transverse sections through the opV placode region of chick embryos a day after electroporating either mPax3En-nGFP or nGFP at the 4–7 somite-stage. (A,B) Most mPax3En-nGFP-targeted cells (nuclear GFP, green) do not express FGFR4 (arrowheads show examples). (C,D) Control nGFP-targeted cells (nuclear GFP, green), including cells that have delaminated from the ectoderm, express FGFR4 (arrowheads show examples). (E,F) Most mPax3En-nGFP-targeted cells (nuclear GFP, green) do not express Ngn2 (arrowheads show examples). (G,H) Control nGFP-targeted cells (nuclear GFP, green) express Ngn2. Much more extensive Ngn2 expression is seen in controls (compare panels E and G).

hb, hindbrain.

Fig. 3. Repressing Pax3 target genes in the ophthalmic trigeminal (opV) placode blocks neuronal differentiation, delamination, expression of Pax3 itself and of the preplacodal marker Eya2.

Panels A-E show transverse sections through the opV placode region of chick embryos a day after electroporating either mPax3En-nGFP or nGFP at the 4–7 somite-stage. The opV placode region can be identified by the expression in untargeted cells of Pax3 and the opV placode-specific markers FGFR4 (A) or Ngn2 (B,C). (A,B) Most mPax3En-nGFP-targeted cells (nuclear GFP, green) in the opV placode region do not express Pax3 itself (red nuclei) or the neuronal differentiation marker Islet1 (white nuclei in A3,B3 for easier visualisation; blue nuclei in merged images in A4,A5 and B4,B5), and do not delaminate. (C) Control nGFP-targeted cells (nuclear GFP, green) in the opV placode region (identified by expression of Ngn2) express Pax3 (red nuclei), differentiate as neurons (white nuclei in C3 for easier visualisation; blue nuclei in merged images in C4,C5), and delaminate into the mesenchyme (arrows indicate examples of delaminated nGFP-targeted Pax3+Islet1+ cells). (D,E) Most mPax3En-nGFP-targeted cells (nuclear GFP, green) in the opV placode region do not express the preplacodal marker gene Eya2. Arrows indicate untargeted Eya2+ cells near Eya2-negative mPax3En-nGFP-targeted ectoderm. (F,G) mPax3En-nGFP (nuclear GFP, green) has no obvious effect on Pax2 expression in the otic vesicle and epibranchial placodes (arrowheads indicate examples of Pax2+ mPax3En-nGFP-targeted cells). (H) Summary graph showing the mean and s.e.m. of the percentage per opV placode of mPax3En-nGFP-targeted versus control nGFP-targeted cells expressing FGFR4, Ngn2, Islet1, and Pax3 itself, and delaminating (see Table 1 for data). The number above each column indicates the number of opV placodes counted for that case.

hb, hindbrain; opV, ophthalmic trigeminal; ov, otic vesicle.

Interestingly, we found that expression of Pax3 itself was also downregulated in mPax3En-nGFP-targeted cells (Fig. 3A,B; compare with Fig. 3C), suggesting that Pax3 regulates its own expression in the opV placode. The mPax3En fusion protein does not seem to be detected by the anti-Pax3 antibody, which was raised against the C-terminal region (amino acids 298–481) of quail Pax3 (Venters et al., 2004): most of the corresponding region in mPax3En has been replaced by the Engrailed repressor domain. Therefore, we could detect endogenous Pax3 expression in targeted cells. The mean percentage/placode of mPax3En-nGFP-targeted cells that expressed Pax3 was very significantly lower than for nGFP-targeted cells (p<0.0005, Student’s t-test; Table 1; Fig. 3H).

We previously showed that blocking canonical Wnt signalling in the opV placode not only led to the loss of Pax3 expression and subsequent opV placode differentiation, but also downregulated expression of the preplacodal marker Eya2 (Lassiter et al., 2007). Although we did not quantify this, we found that electroporating mPax3En-nGFP also led to the downregulation of Eya2 in the opV placode (Fig. 3D,E), suggesting that activation of Pax3 target genes is required to maintain Eya2 expression in the opV placode. The slight thickening of the ectoderm usually seen in the opV placode (slight as compared to the thickening seen in the epibranchial and otic placodes) also seemed to be absent after blocking Pax3 target genes (compare mPax3En-nGFP-targeted versus untargeted sides of the embryo shown in Fig. 2A; compare Fig. 2E with Fig. 2G; compare mPax3En-nGFP-targeted versus untargeted ectoderm in Fig. 3A).

Repressing Pax3 target genes did not seem to affect cell survival, since mPax3En-nGFP-targeted cells expressed nuclear GFP without any obvious changes in nuclear morphology (Fig. 2B,F; Fig. 3A,B; compare with Fig. 2D,H; Fig. 3C). Repressing Pax3 target genes in the Pax2+ otic and epibranchial placodes (which do not themselves express Pax3) had a small but statistically insignificant effect on Pax2 expression, showing that the construct does not indiscriminately quench gene expression: a day after electroporation at the 7–9 somite-stage, the mean percentage of targeted cells counted within Pax2+ otic or epibranchial placode ectoderm that co-expressed Pax2 was 68.8% ± 10.7 for mPax3En-nGFP (965 cells counted across 2 embryos; Fig. 3F,G), versus 85.2% ± 6.2 for GFP controls (686 cells counted across 2 embryos; Fig. 5E,F). This difference is not statistically significant (Student’s t-test).

Fig. 5. Pax3 mis-expression in epibranchial and otic placode ectoderm downregulates Pax2 and disrupts otic vesicle closure.

All panels show transverse sections through the epibranchial placode and/or otic vesicle region of chick embryos either 1 day (A-F) or 2 days (G-J) after electroporating cPax3/GFP or GFP at the 6–8 somite-stage. (A,B) Most cPax3/GFP-targeted cells (Pax3, red nuclei) in the epibranchial placode region do not co-express Pax2 (green nuclei). Compare with unelectroporated side in A. White arrowheads in B show Pax3+Pax2-negative cells; yellow arrowheads show Pax3+Pax2+ cells. (C,D) Most cPax3/GFP-targeted cells (Pax3, red nuclei) in the otic vesicle, as well as expressing Ngn2, fail to co-express Pax2 (green nuclei), and the otic vesicle fails to close properly (compare with Fig. 4G,H). White arrowheads in D show Pax3+Pax2-negative cells; yellow arrowheads show Pax3+Pax2+ cells. (E,F) Electroporating GFP (cytoplasmic GFP, green) has no effect on Pax2 expression in the epibranchial placodes: white arrowheads show examples of Pax2+GFP+ cells. (NB All these embryos were fixed for in situ hybridisation on wax sections. Since only the rabbit anti-GFP antibody works well in our hands after this procedure, and the anti-Pax2 antibody was also raised in rabbit, for the GFP controls we performed in situ hybridisaton for Pax2, followed by immunostaining for GFP.)

hb, hindbrain; ov, otic vesicle; ph, pharynx.

We also attempted to knock down Pax3 expression by electroporating fluorescein-tagged anti-Pax3 morpholinos into opV placode ectoderm (J. R. Bradshaw, PhD thesis; http://contentdm.lib.byu.edu/u?/ETD,590). This consistently reduced or eliminated expression of opV placode markers, in general agreement with the results shown here, but it was difficult to identify targeted cells on sections, and control morpholinos did not always yield consistent results (data not shown).

Pax3 is sufficient to induce FGFR4 and Ngn2 in head ectoderm, but neurons do not differentiate

We investigated whether mis-expression of Pax3 was sufficient to direct opV placode development in non-opV placodal head ectoderm, including ectoderm fated to form other placodes, such as the Pax2+ otic and epibranchial placodes (Baker and Bronner-Fraser, 2000). Note that at early somite stages (at least up to the 4 somite-stage), all cranial ectoderm rostral to the first somite expresses pre-placodal marker genes such as Six4, Eya2 and Dach1 (Litsiou et al., 2005). A day after electroporating cPax3/GFP (encoding full-length chick Pax3 and GFP under separate promoters) unilaterally into head ectoderm from the level of the midbrain to the third somite in 3–10ss embryos, 82% of targeted embryos (41/50; Fig. 4A-C) showed widespread ectopic expression of FGFR4 after whole-mount in situ hybridisation, including in the otic vesicle and pharyngeal arch ectoderm, where the epibranchial placodes form (Fig. 4A-C). No ectopic FGFR4 expression was seen after electroporating GFP (0/8 embryos; Fig. 4D).

Fig. 4. Pax3 mis-expression is sufficient to induce FGFR4 and Ngn2 expression in non-opV placode head ectoderm.

(A-C) Ectopic FGFR4 expression is seen in head ectoderm a day after electroporating cPax3/GFP on the righthand side (black arrowheads), including in the otic vesicle (ov) and nearby pharyngeal ectoderm, where the epibranchial placodes form. Normal FGFR4 expression is seen in the opV placode on the unelectroporated side (white arrowhead) and in the lens (Marcelle et al., 1994). Panels A and B show two different embryos; panel C shows a transverse section through another embryo after wholemount in situ hybridisation for FGFR4. (D) No ectopic FGFR4 expression is seen after electroporating GFP; normal FGFR4 expression is seen in the lens and opV placode (white arrowhead). (E,F) In situ hybridisation directly on sections shows strong FGFR4 expression in cPax3/GFP-targeted cells (Pax3, red nuclei) in the epibranchial placode region, a day after electroporation at the 6–7 somite-stage. (This is a different embryo from those shown in A-C.) (G) FGFR4 expression is not seen in GFP-targeted cells (cytoplasmic GFP, green) in the otic and epibranchial placode region, a day after electroporation at the 7–8 somite-stage. (H) Normal Ngn2 expression is seen in GFP-targeted cells (cytoplasmic GFP, green) in the otic and epibranchial placode region, a day after electroporation at the 7–8 somite-stage. (I,J) Ngn2 expression is seen in cPax3/GFP-targeted cells (Pax3, red nuclei) in surface ectoderm in the vicinity of the otic vesicle, a day after electroporation at the 6–7 somite-stage. (K,L) The neuronal nuclear protein NeuN (red nuclei) is expressed in the vestibuloacoustic and geniculate ganglia, but not in a cluster of cPax3/GFP-targeted cells (Pax3, blue nuclei; GFP, green cytoplasm) at the edge of the geniculate ganglion, 36 hours after electroporation at the 16ss.

gen, geniculate ganglion; h, heart; hb, hindbrain; ov, otic vesicle; va, vestibuloacoustic ganglion.

We quantified the upregulation of FGFR4 in the Pax2+ otic/epibranchial placode region by performing in situ hybridisation and immunostaining directly on sections of five additional embryos (Fig. 4E,F; Table 2). Only Pax3+ cells located within a larger Pax2+ domain were included in the dataset. A day after electroporating cPax3/GFP, the mean percentage/embryo of cPax3/GFP-targeted cells that clearly expressed FGFR4 above background in the Pax2+ otic/epibranchial placode region was 27.6% (Table 2). In contrast, in five control embryos (electroporated at the 7–8ss and fixed a day later at the 23–29ss), GFP-targeted cells in otic/epibranchial placode region ectoderm did not obviously upregulate FGFR4 expression above background (Fig. 4G).

Table 2.

Cell counting data across all embryos electroporated in otic/epibranchial placode ectoderm with cPax3/GFP and assayed a day later for FGFR4, Ngn2, Islet1, NeuN or Pax2 expression.

Marker assayed	Age when electroporated with cPax3/GFP analysed (ss)	No. embryos	No. cPax3/GFP- targeted cells counted overall	Mean %/embryo of cPax3/GFP- targeted cells +ve for marker	s.d.	s.e.m.
FGFR4	6–7/23–28	5	1296	27.6	22.9	10.2
Ngn2	7–8/21–26	4	2148	32.4	6.2	3.1
Islet1	6–8/21–28	8	3167	1.7	2.9	1.0
NeuN	12–19/26–36	3	430	3.1	3.4	2.0
Pax2	6–8/21–28	9	3444	6.2	5.5	1.8

Abbreviations: No., number of; s.d., standard deviation; s.e.m., standard error of the mean; ss, somite stage; +ve, positive.

(NB The cell counts and embryo numbers are lower for Islet1 than for Pax2 because the Islet1 immunostaining did not work on slides from one embryo.)

At the stages analysed, Ngn2 is normally expressed in the neurogenic region of the otic vesicle and in the epibranchial placodes (Xu et al., 2008); this normal expression was seen after in situ hybridisation and immunostaining on sections of five control embryos (electroporated with GFP at the 7–8ss and fixed a day later at the 23–29ss) (Fig. 4H). A day after electroporating cPax3/GFP, the mean percentage/embryo of cPax3/GFP-targeted cells that expressed Ngn2 in the Pax2+ otic/epibranchial placode region was 32.4% (Fig. 4I,J; Table 2). Hence, Pax3 induces Ngn2 as well as FGFR4, and in a similar proportion of targeted cells (Table 2).

We co-immunostained the sections analysed for FGFR4 and Ngn2 expression with Islet1, to detect differentiating neurons. Rather surprisingly, only 0.85% of the 3167 cPax3/GFP-targeted cells counted clearly expressed Islet1. The mean percentage/embryo of cPax3/GFP-targeted cells that expressed Islet1 in the Pax2+ otic/epibranchial placode region was only 1.7% (Table 2). We also analysed expression of the neuron-specific nuclear protein NeuN (Mullen et al., 1992) in three embryos fixed a day after electroporation at the 12–19ss. In these older embryos, the epibranchial placodes were identified by morphology and position, rather than by Pax2 expression. The mean percentage/embryo of cPax3/GFP-targeted cells that expressed NeuN in the epibranchial placode region was only 3.1% (Table 2). In another embryo, electroporated at the 16ss and incubated for 36 hours (to stage 20; Hamburger and Hamilton, 1951) to allow a longer time-period for differentiation, we found a cluster of NeuN-negative cPax3-targeted cells that had delaminated from the epithelium and were associated with, though excluded from, the NeuN+ geniculate ganglion (Fig. 4K,L). Overall, these results suggest that although a subset of cPax3-targeted cells in the otic/epibranchial placode region expresses Ngn2 and FGFR4, very few differentiate as neurons.

Pax3 is sufficient to alter the identity of otic and epibranchial placode cells

Strikingly, the mean percentage/embryo of Pax3+ cells in the otic/epibranchial placodes that co-expressed Pax2 was only 6.2% (Fig. 5A-D; Table 2), versus 85.2% ± 6.2 for GFP controls (686 cells counted across 2 embryos; Fig. 5E,F). Furthermore, when otic-level ectoderm was extensively targeted with cPax3/GFP, otic vesicle closure was often disrupted (Fig. 5C,D), when compared with the normal otic vesicle development seen after electroporating GFP into otic-level ectoderm (Fig. 4G,H).

Overall, therefore, Pax3 mis-expression in otic and epibranchial placode ectoderm not only upregulates the opV placode markers FGFR4 and Ngn2, but also downregulates and/or prevents Pax2 expression and disrupts otic vesicle closure, suggesting that Pax3 expression is sufficient to alter the identity of otic and epibranchial placode cells.

Discussion

Pax3 is necessary for normal neuronal differentiation in the mouse opV placode

The Splotch^2H (Sp^2H) mutation is a 32-basepair deletion in the Pax3 homeobox, yielding a truncated protein lacking the whole of the C-terminal transactivation domain and part of the homeodomain (Epstein et al., 1991). The Sp^2H phenotype is indistinguishable from the original Splotch allele (in which genomic mutations in intron 3 prevent normal splicing; Epstein et al., 1993), and both are considered functionally null alleles (Epstein et al., 1991; Epstein et al., 1993). We found that at E9, before mouse neural crest-derived trigeminal neurons have differentiated (Verwoerd and van Oostrom, 1979; Nichols, 1986; Stainier and Gilbert, 1990), opV placode-derived neurons in the developing ophthalmic lobe of the trigeminal ganglion are missing in homozygous Sp^2H embryos, although the proneural transcription factor Ngn1 is expressed normally. Hence, in the absence of wildtype Pax3 function, Ngn1+ opV placode cells fail to differentiate as neurons. At late E9.5, we saw some recovery of neuronal differentiation in the ophthalmic lobe and along the ophthalmic nerve projection towards the eye. It is possible that these neurons are derived from early-differentiating trigeminal neural crest cells, or even that they represent a rostral expansion of maxillomandibular trigeminal placode-derived neurons.

Activation of Pax3 target genes is necessary for chick opV placode development

We tested the cell-autonomous requirement for activation of Pax3 target genes during the development of the chick opV placode by electroporating a construct encoding a fusion between mouse Pax3 and the Drosophila Engrailed repressor domain (previously shown to repress Pax3 target genes in the mouse; Bajard et al., 2006; Relaix et al., 2006). We found that activation of Pax3 target genes is necessary for expression of the opV placode markers FGFR4 and Ngn2, maintenance of Pax3 itself and of the preplacodal marker gene Eya2, expression of the neuronal differentiation marker Islet1, and delamination. We also found that the normal slight thickening seen in the opV placode (slight as compared to the much thicker ectoderm of the otic and epibranchial placodes) seemed to be lost after blocking Pax3 target gene expression. Hence, activation of Pax3 target genes is necessary for all aspects of chick opV placode development. These results also confirm that Pax3 acts as a transcriptional activator in the opV placode, as it does in the neural tube, neural crest and somites (Relaix et al., 2003; Bajard et al., 2006).

We previously found that blocking canonical Wnt signalling in opV placode ectoderm led to the loss of Pax3 expression and the failure of opV placode development, including expression of FGFR4, maintenance of Eya2, neuronal differentiation (as assayed by Islet1, NeuN and neurofilament expression) and delamination (Lassiter et al., 2007). Here, we showed that activation of Pax3 target genes is necessary for all of these events, suggesting that the effects of blocking canonical Wnt signalling are at least partly due to the loss of Pax3 expression.

Pax3 autoregulates in the opV placode

Since activation of Pax3 target genes is needed to maintain Pax3 in the opV placode, this suggests that Pax3 regulates its own expression. In P19 embryonal carcinoma cells, in which Pax3 is necessary and sufficient to drive skeletal (but not cardiac) myogenesis, in vitro experiments using a similar Pax3-Engrailed fusion protein (in this case between the first 286 amino acids of Pax3 and the mouse Engrailed2 repressor domain) have also shown that activation of Pax3 target genes is required for Pax3 expression (Ridgeway and Skerjanc, 2001). There is precedent for direct autoregulation of Pax genes: Pax6 positively autoregulates in the lens and cornea by directly binding a head ectoderm-specific enhancer in the Pax6 gene (Ashery-Padan et al., 2000; Aota et al., 2003), and in the diencephalon by binding to a conserved enhancer in intron 7 of the Pax6 gene (Kleinjan et al., 2004). Interestingly, Pax6 also negatively autoregulates at high levels: over-expression of Pax6 protein leads to the downregulation of Pax6 expression (Manuel et al., 2007).

Pax3 upregulates FGFR4 and Ngn2 expression in head ectoderm

Both FGFR4 and Ngn2 are expressed in the chick opV placode from the 10ss (Stark et al., 1997; Begbie et al., 2002), a few hours after Pax3 protein first becomes detectable by immunostaining, at the 8–9ss (Baker et al., 1999). We found that Pax3 mis-expression in non-opV placode head ectoderm, including the otic and epibranchial placodes, upregulates FGFR4 and Ngn2 in a subset of targeted cells. In the mouse, FGFR4 and Ngn2 were recently shown to be direct transcriptional targets of Pax3 (Lagha et al., 2008; Nakazaki et al., 2008), so it is possible that Pax3 directly activates both genes during chick opV placode development. In the mouse, we found that Pax3 is not necessary for Ngn1 expression in the opV placode. However, the expression of Ngn1 and Ngn2 is different in mouse and chick trigeminal placodes: in the mouse, Ngn1 is expressed in both opV and mmV placodes (Gradwohl et al., 1996; Sommer et al., 1996; Fode et al., 1998; Ma et al., 1998), with Ngn2 only transiently expressed later in what appears to be (though was not reported as such) specifically the opV placode (Fode et al., 1998). In the chick trigeminal placodes, Ngn2 is specifically expressed in the opV placode, at least initially (Xu et al., 2008), while Ngn1 is only expressed in the mmV placode (Begbie et al., 2002). It is not clear what changes in gene regulation have led to this difference between mouse and chick (Furlong and Graham, 2005).

The absolute levels of Pax3 are important for its function

It is striking that only a minority of targeted ectoderm cells (around 30%) expressed FGFR4 or Ngn2 following Pax3 mis-expression. One possibility that could be considered is the importance of absolute Pax3 expression levels for the activation of gene expression. Heterozygous mutations in Pax3 have a phenotype primarily due to defects in melanocyte development (pigmentation defects in mice, pigmentary-auditory deficits in human Waardenburg syndrome I or III), while it was recently shown that Pax3 hypomorphs have deficits in limb and hypoglossal (but not other hypaxial or epaxial) musculature (Zhou et al., 2008). Thus, the absolute levels of Pax3 are critical for the appropriate development of specific subpopulations of Pax3-expressing cells. Furthermore, in vitro experiments have shown that low levels of Pax3 activate expression of a reporter construct, while higher levels inhibit its expression (Chalepakis et al., 1994). The change from activation to inhibition occurred over a relatively narrow concentration range: a 2-fold increase in concentration was enough for the switch (Chalepakis et al., 1994). When Pax3 is mis-expressed outside opV placode ectoderm, each electroporated cell is likely to receive a different amount of plasmid, and thus be exposed to different levels of Pax3 expression. One could speculate that only cells where Pax3 levels are within a particular range are likely to activate (rather than repress) expression of Ngn2, FGFR4, and other direct or indirect Pax3 gene targets. However, although we did not quantify Pax3 levels after mis-expression, a qualitative analysis shows that both cells with very bright and cells with relatively faint levels of Pax3 immunofluorescence upregulated FGFR4 or Ngn2 (e.g. Fig. 4F,J).

Pax3 mis-expression does not promote neuronal differentiation

Although Pax3 target gene activation is necessary for neuronal differentiation in the opV placode, and Ngn2 is upregulated in a subset of Pax3-targeted cells in non-opV placode head ectoderm, the great majority of these cells do not differentiate as neurons. Retroviral-mediated mis-expression of Ngn2 in the chick embryo leads to ectopic sensory neurogenesis in neural crest cell derivatives and the dermomyotome (Perez et al., 1999). However, we found in previous experiments that blocking canonical Wnt signalling in either trunk or head ectoderm broadly upregulates Ngn2 expression without any accompanying neuronal differentiation (Lassiter et al., 2007). Indeed, Ngn2 may be more important for specifying neuronal subtypes than for directly promoting neuronal differentiation: genetic lineage-tracing has shown that Ngn2 expression biases (but does not commit) mouse trunk neural crest cells to a sensory fate (either neuronal or glial) (Zirlinger et al., 2002). Similarly, genetic experiments in which the coding sequences for Mash1 and Ngn2 were swapped suggested that Ngn2 acts permissively, rather than instructively, to specify neuronal phenotype in combination with other transcription factors (Parras et al., 2002).

The failure of Pax3-targeted cells to differentiate as neurons suggests that additional signals are required for neuronal differentiation, and/or that the maintenance of high levels of Pax3 expression blocks subsequent neuronal differentiation. In myogenesis, FGFR4, which is expressed after Pax3, is required for chick muscle differentiation and mouse muscle regeneration (Marics et al., 2002; Zhao et al., 2006). Hence, signalling through FGFR4 seems to be necessary for a Pax3+ muscle cell to differentiate. It is possible that FGFR4 plays a similar role in the opV placode, i.e., that only Pax3+ cells in which signalling through FGFR4 has been activated are able to differentiate as neurons. Various different FGFs are known to activate FGFR4 (Ornitz et al., 1996; Zhang et al., 2006). If the appropriate ligand(s) is not present in the targeted region, then even FGFR4+ Pax3-targeted cells may not differentiate as neurons.

There is increasing evidence that Pax3 can both activate expression of genes required for a specific differentiation pathway, and block differentiation until appropriate signals are received that downregulate Pax3 expression levels or activity. Pax3 activates myelination genes, and its expression in Schwann cell precursors is downregulated immediately prior to terminal differentiation (Kioussi et al., 1995). Enforced maintenance of Pax3 expression in neural crest cells blocks neural crest-derived osteoblast differentiation (Wu et al., 2008). In melanocytes, Pax3 initiates differentiation by activating MITF transcription, but blocks differentiation until beta-catenin displaces Pax3-recruited Grg4 from the dopachrome tautomerase promoter (Lang et al., 2005). Transfecting antisense Pax3 RNA into the undifferentiated neuronal cell line ND7 results in differentiation, suggesting that Pax3 blocks neuronal differentiation in this cell line (Reeves et al., 1999). Antisense-oligonucleotide down-regulation of Pax3 in neural crest cultures inhibits sensory neuron differentiation without affecting survival of sensory neurons or precursor populations (Koblar et al., 1999), and neural crest cells undergo premature neurogenesis in Splotch mice (Nakazaki et al., 2008). Although Pax3 is not expressed in neural crest-derived neurons in either the trigeminal or dorsal root ganglia, it is expressed in non-neuronal neural crest cells in these ganglia, albeit at much lower levels than in opV placode-derived cells (Baker et al., 2002). This suggests that Pax3 must be downregulated as soon as (or before) neural crest cells differentiate as neurons. Similarly, in spinal cord neurogenesis, Pax6 upregulates Ngn2, which promotes cell cycle exit and neuronal commitment, but the cells do not differentiate as neurons until Pax6 is itself repressed by increasing levels of Ngn2 (Bel-Vialar et al., 2007).

How might this model apply to Pax3+ opV placode-derived neurons, which maintain high levels of Pax3 expression long after differentiation in the chick (Baker et al., 2002)? Interestingly, in the neuronal cell line ND7, the DNA-binding activity of Pax3 is downregulated within an hour of the induction of neuronal differentiation, before any drop in Pax3 expression levels can be detected (Reeves et al., 1998). Phosphorylation of the linker region between the paired domain and homedomain of Pax3 inhibits DNA binding without affecting nuclear localisation or stability (Amstutz et al., 2008). Thus, it is possible that in normal opV placode development, neuronal differentiation only occurs when the DNA-binding activity of Pax3 is inhibited, perhaps by phosphorylation of the linker region in response to an extracellular or intracellular signal. If this signal were only present in the opV placode, this model might explain why Pax3+ ectodermal cells outside the opV placode do not differentiate as neurons, even when FGFR4 and Ngn2 are expressed. Alternatively, high levels of Pax3 after mis-expression might swamp the protein-modification machinery, short-circuiting the neuronal differentiation pathway.

Pax3 is sufficient to alter the identity of otic and epibranchial placode cells

After Pax3 mis-expression in the Pax2+ otic and epibranchial placodes, very few targeted cells co-express Pax2. Since Pax2 should have been expressed at the time of electroporation (6–8ss; Groves and Bronner-Fraser, 2000; Streit, 2002), this suggests that Pax3 mis-expression down-regulates Pax2. This is reminiscent of the mutual repression seen between Pax2 and Pax6 in the mammalian eye that establishes the optic cup/optic stalk boundary (Schwarz et al., 2000). In vitro work using reporter constructs has demonstrated that Pax2 can repress Pax6 expression, while Pax6 can repress Pax2 transcription, as well as activate its own transcription (Schwarz et al., 2000). Furthermore, Pax3 mis-expression in the otic and epibranchial placodes induces expression of the opV placode-specific marker FGFR4 and upregulates Ngn2, and disrupts closure of the otic vesicle. Overall, these experiments show that Pax3 can change the fate of other placodal cells, suggesting that it is important for placodal cell identity.

Model for opV placode development in the chick embryo

At the 3-somite stage, all cranial ectoderm rostral to the first somite is competent to respond to a Pax3-inducing signal from the dorsal neural tube (though competence is rapidly lost by specified otic placode ectoderm) (Stark et al., 1997; Baker et al., 1999). This region of competence correlates with expression of the pre-placodal marker genes Six4, Eya2 and Dach1 (Litsiou et al., 2005). Both canonical Wnt signalling (Lassiter et al., 2007) and PDGF signalling from the neural folds (McCabe and Bronner-Fraser, 2008) are necessary for Pax3 expression in the opV placode and subsequent opV placode development, but neither is sufficient to induce Pax3 in competent ectoderm (Lassiter et al., 2007; McCabe and Bronner-Fraser, 2008). FGF signalling from the neural folds is a likely additional candidate; this is under investigation. Canonical Wnt signalling may serve to stabilise Pax3 expression, since it is required not only for Pax3 expression but also for its maintenance (Lassiter et al., 2007). Thus, we propose a model in which Pax3 expression is stabilised in those Pax3+ cells that receive canonical Wnt signals (just as, in the otic placode, canonical Wnt signals are required for Pax2+ cells to adopt an otic rather than epidermal fate; Ohyama et al., 2006). Activation of Pax3 target genes, directly or indirectly, is necessary for the expression of FGFR4 and Ngn2, to maintain expression of Pax3 itself (and the pre-placodal marker gene Eya2), and to enable delamination and neuronal differentiation. However, additional signals (which may abrogate Pax3 function) are necessary to enable Pax3+ opV placode cells to differentiate as neurons. This model is summarised in Fig. 6.

Fig. 6.

Model for opV placode development in the chick embryo, based on work presented here and in Lassiter et al. (2007), as well as McCabe and Bronner-Fraser (2008). See text for details.

fb, forebrain; ov, otic vesicle.

Summary

We have shown that Pax3 is necessary for normal neuronal differentiation of Ngn1+ opV placode cells in the mouse. In the chick, we have shown that activation of Pax3 target genes is necessary for all aspects of opV placode development: expression of the opV placode markers FGFR4 and Ngn2, maintenance of Pax3 itself (suggesting Pax3 autoregulates) and the preplacodal marker gene Eya2, placodal thickening, neuronal differentiation and delamination. Mis-expression of Pax3 in non-opV placode head ectoderm is sufficient to induce FGFR4 and Ngn2, downregulate Pax2 in the otic/epibranchial placodes and disrupt otic vesicle closure, showing that Pax3 is sufficient to alter the fate of other placode cells. However, most Pax3-targeted cells in the otic or epibranchial placodes do not differentiate as neurons. Taken together, these results support a model in which activation of Pax3 target genes is necessary for opV placode development, including neuronal differentiation, but additional signals are required to enable Pax3+ opV placode cells to differentate as neurons. Overall, therefore, Pax3 is necessary but not sufficient for sensory neurogenesis within the opV placode.