Mesendodermal signals required for otic induction: Bmp-antagonists cooperate with Fgf and can facilitate formation of ectopic otic tissue

Abstract

Induction of otic placodes requires Fgf from surrounding tissues. We tested the hypothesis that mesendodermally derived Bmp-antagonists Chordin, Follistatin-a and Crossveinless-2 cooperate in this process. Injecting morpholinos for all three genes, or treatment with the Nodal inhibitor SB431542 to block mesoderm-formation, reduces otic induction and strongly enhances the effects of disrupting fgf3 or fgf8. In contrast, using a lower dose of SB431542, combined with partial loss of Fgf, causes a dramatic medial expansion of otic tissue and formation of a single, large otic vesicle spanning the width of the hindbrain. Under these conditions, paraxial cephalic mesoderm forms ectopically at the midline, migrates into the head and later transfates to form otic tissue beneath the hindbrain. Blocking expression of Bmp-antagonists blocks formation of medial otic tissue. These data show the importance of mesendodermal Bmp-antagonists for otic induction and that paraxial cephalic mesendoderm can facilitate its own otic differentiation under certain circumstances.

INTRODUCTION

The vertebrate inner ear develops from a simple ectodermal thickening, the otic placode. Induction of the otic placode is critically dependent upon Fgfs from adjacent hindbrain tissue and subjacent mesendoderm. In zebrafish, fgf3 and fgf8 are expressed prominently in the hindbrain and at a lower level in paraxial cephalic mesendoderm. Disrupting either gene alone permits some otic induction to occur whereas loss of both fgf genes blocks otic induction entirely (Phillips et al., 2001; Léger and Brand, 2002; Maroon et al., 2002; Liu et al., 2003). Similarly, in chick and mouse embryos, Fgf3 in the hindbrain and Fgf8 and Fgf10 in subjacent mesenchyme are critical for otic induction, and mesodermal Fgf19 also contributes to otic induction in chick. In these species, too, disrupting any single Fgf gene has little or no effect on otic development, but disrupting Fgf3 together with any one of the mesendodermal Fgf genes severely impairs or blocks otic induction (Ladher et al., 2000; Alvarez et al., 2003; Wright and Mansour, 2003; Ladher et al., 2005; Zelarayan et al., 2007). Thus, Fgfs from the hindbrain and mesendoderm act at least partially redundantly in initiating otic development.

In addition to the above redundant functions, it is likely that functions unique to either the mesendoderm or hindbrain also contribute to otic induction. For example, Wnt8 is expressed at relevant stages in the developing hindbrain in chick, mouse and zebrafish, wherein it has been implicated in stabilizing otic fate following induction (Ohyama et al., 2006; Freter et al., 2008) or in regulating timely expression of fgf3 and fgf8 in the hindbrain (Phillips et al., 2004). Less is known about specific mesendodermal signals. However, in zebrafish paraxial cephalic mesendoderm expresses a number of Bmp-antagonists, including Chordin (Chd) (Miller-Bertoglio et al., 1997) and Follistatin-a (Fsta) (Bauer et al., 1998), which we hypothesize contribute to preotic development. Another Bmp-antagonist, Crossveinless-2 (Cv2), has been variously reported to be expressed in paraxial mesendoderm (Rentzsch et al., 2006) or in preplacodal ectoderm (Esterberg and Fritz, 2008), a zone of general preplacodal competence from which all placodes arise. Although the discrepancy in spatial expression requires resolution, Fritz and Esterberg (2008) have shown that Cv2 activity does indeed contribute to otic development by potentiating Fgf signaling. Furthermore, misexpression of various Bmp-antagonists in Xenopus and chick embryos expands the domain of preplacodal ectoderm (Brugmann et al., 2004; Glavic et al., 2004; Ahrens and Schlosser, 2005; Litsiou et al., 2005). However, given the likelihood that there is functional redundancy between diverse Bmp-antagonists, the full significance of such factors for otic development remains an open question, and each needs to be tested individually and in combination by loss-of-function approaches.

Other techniques for addressing the relative importance of mesendoderm for otic induction include genetic or surgical ablations. In chick, surgical removal of paraxial cephalic mesoderm underlying the presumptive otic ectoderm prevents induction of the otic placode, despite continued interactions with the hindbrain (Kil et al., 2005). In zebrafish, ablation of head mesendoderm by disrupting Nodal signaling delays otic induction and results in a small malformed otic vesicle (Mendonsa and Riley, 1999; Phillips et al., 2001). The milder phenotype in mesoderm-depleted zebrafish relative to chick possibly reflects the additional Fgf-redundancy in the zebrafish hindbrain. To test this notion, we previously examined otic induction in fgf8 mutants that were depleted of mesendoderm by morpholino-mediated knockdown of One-eyed pinhead (Oep), an essential co-receptor for Nodal (Riley and Phillips, 2003). Surprisingly, rather than further reducing otic induction, a variable fraction of such embryos developed with a dramatic medial expansion of otic tissue, resulting in a single large otic vesicle spanning the width of the hindbrain. This paradoxical finding remains unexplained.

Here we readdress the role of mesendoderm by using a pharmacological inhibitor of Nodal signaling, SB431542 (Ho et al., 2006; Sun et al., 2006; Hagos and Dougan, 2007). This drug offers the ability to precisely titrate the level of Nodal signaling, giving a more reproducible phenotype than oep-morpholino oligomer (Nasevicius and Ekker, 2000). We find that loss of one hindbrain signal, Fgf3 or Fgf8, combined with SB431542-treatment can ablate both paraxial cephalic mesendoderm and otic tissue. In contrast, a lower dose of SB431542 combined with heterozygous loss of fgf3 or fgf8 resulted in a variable fraction of embryos forming medially expanded otic tissue. Lineage tracing showed that in such embryos paraxial mesendoderm forms at the midline and subsequently trans-differentiates into otic tissue near the end of gastrulation. Bmp-antagonists expressed in paraxial cephalic mesendoderm are required for its otic-inducing properties. Knockdown of chd, fsta and cv2 strongly impaired induction of endogenous otic placodes and fully blocked formation of ectopic (medial) otic tissue in inhibitor-treated embryos.

RESULTS

Cooperation of mesendoderm and hindbrain signals in otic placode induction

To further address the role of mesendodermal signals in zebrafish otic induction, we examined embryos lacking head mesendoderm. Disruption of Nodal signaling in zebrafish results in loss of head and trunk mesendoderm (Gritsman et al., 1999). Previously, we reported that in embryos depleted for Oep, an essential co-receptor required for Nodal signaling, otic induction is delayed by 1–2 hours and the otic vesicle is small and malformed (Mendonsa and Riley, 1999; Phillips et al., 2001). Here we have extended those studies and, to avoid the phenotypic variability associated with oep-morpholino oligomer (oep-MO), we used a pharmacological inhibitor of Nodal signaling. SB431542 specifically blocks Nodal signaling and phenocopies disruption of maternal and zygotic oep functions in a dose-dependent manner (Ho et al., 2006; Sun et al., 2006; Hagos and Dougan, 2007). In our wild-type background, exposure to 30 μM SB431542 beginning at the 1-cell stage (0.5 hpf) phenocopied loss of zygotic oep function (Zoep mutants), whereas exposure to 100 μM SB431542 phenocopied loss of maternal and zygotic oep function (MZoep mutants). Consistent with earlier studies, 100 μM SB431542 delayed otic induction and led to formation of small malformed otic vesicles (Fig. 1B). To examine the effects of disrupting mesendodermal signals and one hindbrain signal, we treated fgf8^−/− mutants or fgf3 morphants with 100 μM SB431542. This resulted in a severe deficiency (Fig. 1D, n=14/21) or complete loss of otic tissue (Fig. 1E, n=7/21; Fig. 1F, n=7/27). These data support the hypothesis that mesendodermal signals cooperate with hindbrain signals to induce otic development.

Figure 1. Loss of mesendoderm and fgf8 or fgf3 disrupts otic induction.

Lateral views (anterior to left) of otic vesicles in live embryos at 35 hpf. A, wild-type control. B, wild-type embryo treated with 100 μM SB431542. C, fgf8^−/− mutant. D, E, fgf8^−/− mutants treated with 100 μM SB431542 showing either a micro-vesicle (D, arrow) or no morphologically detectable vesicle (E, asterisk). F, fgf3-morphant embryo treated with 100 μM SB431542 showing no otic morphologically detectable vesicle (asterisk). Scale bar, 25 μm.

Bmp-antagonists expressed in subjacent mesendoderm potentiate otic induction

Potential otic inducing signals expressed by paraxial cephalic mesendoderm include Fgfs (Reifers et al., 2000; Phillips et al., 2001; Nechiporuk et al, 2007; Nikaido et al., 2007), as well as various Bmp-antagonists (Miller-Bertoglio et al., 1997; Bauer et al., 1998; Rentzsch et al., 2006). The role of Fgf has been extensively studied. In contrast, while attenuation of Bmp has been implicated in specification of preplacodal ectoderm (Brugmann et al., 2004; Glavic et al., 2004; Ahrens and Schlosser, 2005; Litsiou et al., 2005), relatively little is known about which Bmp-antagonists are involved or whether they regulate later stages of placodal development. In zebrafish, at least three Bmp-antagonists are expressed in mesendodermal domains likely to influence otic development: chordin (chd), follistatin-a (fsta), and crossveinless 2 (cv2) (Miller-Bertoglio et al., 1997; Bauer et al., 1998; Rentzsch et al., 2006). chd is broadly expressed in dorsal tissue, including axial and paraxial mesendoderm, through much of gastrulation (Fig. 2A). By 9 hpf, when otic induction begins, chd-expressing cells lie just medial to the preotic domain of foxi1 (Fig. 2D). fsta is expressed throughout gastrulation in paraxial mesendoderm, including cells lying directly beneath the preotic domain of pax8 (Figs. 2B, E). Near the end of gastrulation, cv2 begins to be expressed in paraxial tissue near the prospective ear (Fig. 2C). There has been disagreement as to whether this domain is mesendodermal (Rentzsch et al., 2006) or ectodermal (Esterberg and Fritz, 2008). However, we find in sections of 10.5 hpf embryos show that the bulk of cv2 expression is mesendodermal (Fig. 2F), though there is also a minor ectodermal domain, possibly corresponding to the domain reported by Esterberg and Fritz (2008). Thus, we hypothesized that Chd, Fsta and Cv2 secreted primarily from mesendoderm surrounding the prospective ear provide redundant functions required for normal otic induction.

Figure 2. Loss of Bmp-antagonists impairs otic induction.

A, D expression of chd at 7.5 hpf, dorsal view (A), and chd and foxi1 at 9 hpf, lateral view (D). The preotic domain of foxi1 is indicated (ot). B, E, expression of fsta at 10 hpf, dorsal view (B) and fsta and pax8, lateral view (E). The preotic domains of pax8 is indicated (ot). C, F, expression of cv2 at 10.5 hpf as shown in a wholemount specimen, dorsal view (C) and in cross section (F). The plane of section (f) and preplacodal ectoderm (ppe) are indicated. G-I, Otic vesicle at 26 hpf in a wild-type control (D), chd/fsta double-morphant (E) and chd/fsta/cv2 triple-morphant (F). J-L, dorsal views of cldna expression in otic cells at 27 hpf in an fgf8^−/− mutant (J), fgf8^−/− mutant injected with chd/fsta-MO (K) and fgf8^−/− mutant injected with chd/fsta/cv2-MO (L). M-R, dorsal views of pax8 expression in pre-otic cells at 11 hpf (boxed area) in a wild-type control (M), chd/fsta double-morphant (N), chd/fsta/cv2 triple-morphant (O), fgf8^−/− mutant (P), fgf8^−/− mutant injected with chd/fsta-MO (Q), and fgf8^−/− mutant injected with chd/fsta/cv2-MO (R). Anterior is to the top in all images except (G-I) in which anterior is to the left. Scale bar, 25 μm. S, schematic diagram of the fsta exon/intron structure, the fsta-MO2 target site (bar), primer-binding sites used for RT-PCR (P1, P2 arrows), and the results of RT-PCR confirming that fsta-MO2 injection severely reduces fsta transcript levels.

We tested the roles of chd, fsta and cv2 by morpholino-mediated knockdown. MOs for chd and cv2 have been previously characterized (Nasevicius and Ekker, 2000; Rentzsch et al., 2006). To knockdown fsta, we designed both translation-blocking and splice-blocking MOs, which gave identical results (see Experimental Procedures). RT-PCR was used to demonstrate the efficacy of the fsta splice-blocker (Experimental procedures, and Fig. 2S). Knockdown of any single Bmp-antagonist had little or no effect on ear development. However, coinjection of chd-MO and fsta-MO (chd/fsta double-morphants) reduced the size of the otic vesicle by roughly half (Fig. 2H). In chd/fsta/cv2 triple-morphants, the size of the otic vesicle was further reduced and no otoliths were produced (Fig. 2I). While several markers of vesicle patterning were detectably expressed in chd/fsta/cv2 triple-morphants, expression was usually not spatially restricted (Table 1). About one-third of chd/fsta/cv2 triple-morphants failed to express the ventrolateral marker otx1 (Table 1), and in most of the remainder otx1 expression was not localized. The anterior (utricular) marker pax5 was not expressed in any chd/fsta/cv2 triple-morphants and most specimens did not produce hair cells (Table 1). Thus, patterning in these micro-vesicles was highly aberrant. At an earlier stage during otic induction, the number of pax8-expressing cells in the prospective ear region was substantially reduced in chd/fsta double-morphants, and was further reduced in chd/fsta/cv2 triple-morphants (Fig. 2N, O). Loss of these Bmp antagonists did not disrupt expression of fgf3 or fgf8 (not shown), but we speculated that under these conditions otic development might be more sensitive to loss of fgf function. In support, knockdown of chd and fsta, or chd, fsta and cv2, in fgf8^−/− null mutants or fgf3-morphants severely reduced the preotic domain of pax8 or ablated it altogether (Fig. 2P–R, and data not shown). Severely affected specimens produced no morphological otic vesicles, though a few disorganized cldna-expressing otic cells were still evident at 24 hpf (Fig. 2L). These data support the hypothesis that otic induction requires Chd, Fsta and Cv2, which appear to potentiate the otic inducing activity of Fgfs.

Table 1.

Fraction of chd/fsta/cv2 triple morphants expressing otic markers at 24 hpf.

Marker	dlx3b	nkx5.1	otx1	pax2a	pax5	brn3c:gfp
Detectable expression	10/10	12/12	10/14	10/10	0/14	2/11
Proper localization	5/10	0/12	2/14	8/10	0/14	2/11

Previous studies have shown that Bmp-antagonists, working in parallel with Fgfs, stimulate formation of preplacodal ectoderm from which all placodes are derived (Ahrens and Schlosser, 2005; Litsiou et al., 2005). We therefore tested whether knockdown of Bmp-antagonists impair formation of preplacodal ectoderm. chd/fsta/cv2 triple-morphants expressed preplacodal markers dlx3b, eya1, and six4.1 on time, although the level of expression was variably reduced (Fig. 3A–F). To test whether other placodal derivatives still formed, we examined expression of pitx3 and cxcr4b, which mark various anterior cranial placodes (Dutta et al., 2005; Knaut et al., 2005; Zilinski et al., 2005; Miyasaka et al., 2007). pitx3 was expressed appropriately in the pituitary and lens placodes at 19 hpf, albeit in smaller domains, whereas the level of expression in the trigeminal placode was blocked or strongly reduced (Fig. 7H). Expression of cxcr4b was relatively normal in the nasal placode at 19 hpf, while the trigeminal domain was much smaller than normal (Fig. 7J). Thus, loss of chd, fsta and cv2 does not block formation of preplacodal ectoderm, possibly reflecting redundancy provided by other dorsally expressed Bmp-antagonists (Fürthauer et al., 1999; Nicoli et al., 2005). It is possible that the reduced level of preplacodal gene expression in chd/fsta/cv2 triple-morphants contributes to the small size of most cranial placodes. In addition, the especially severe perturbation of trigeminal and otic development (Figs. 2O and 3H) indicates that these placodes have more stringent requirements for these Bmp-antagonists.

Figure 3. Formation of preplacodal ectoderm and various cranial placodes in chd/fsta/cv2 triple-morphants.

A, B, six4.1 expression at 10.5 hpf in a wild-type control (A) and chd/fsta/cv2 triple-morphant (B). C, D, eya1 expression at 10.5 hpf in a wild-type control (C) and chd/fsta/cv2 triple-morphant (D). E, F, dlx3b expression at 10.5 hpf in a wild-type control (E) and chd/fsta/cv2 triple-morphant (F). G, H, pitx3 expression at 19 hpf in a wild-type control (G) and chd/fsta/cv2 triple-morphant (H). The anterior pituitary (ap), lens placode (lp) and trigeminal placode (tp) are indicated. I, J, cxcr4b expression at 19 hpf in a wild-type control (I) and chd/fsta/cv2 triple-morphant (J). The nasal placode (np) and trigeminal placode (tp) are indicated.

Figure 7. Patterning within medial otic vesicles.

Expression of various markers in control embryos, or wild-type embryos treated with 70 μM SB431542 + 1 μM SU5402. A-F, Expression of AP markers pax5 (A, B), nkx5.1 (C, D) and fsta (E, F). G-J, Expression of DV markers dlx3b (G, H), otx1 (I, J) and ngn1 (K, L). M-P, fluorescence/DIC overlays showing cranial ganglia in live embryos at 28 hpf, as visualized by islet2b:GFP. Neurons of the statoacoustic ganglia (sag) are indicated. Q, R, Bright field and fluorescence images of a live inhibitor-treated specimen at 30 hpf reveals the presence of sensory epithelia, marked by otoliths (Q, arrows) and brn3c:GFP expression in hair cells (R, arrows). Images show lateral views with anterior to the left (A-L, O and P) or dorsal views with anterior to top (M, N, Q, R). Scale bar, 25 μm.

Hypomorphic reduction of Nodal and Fgf signaling results in medial expansion of otic tissue

In sharp contrast to the loss of ear tissue seen in fgf8^−/− null mutants treated with 100 μM SB431542, we noticed that ~ 6% of fgf8^+/− intercross progeny exposed to 70 μM SB431542 produced a single large otic vesicle spanning the width of the hindbrain at the level of rhombomere 4 (Fig. 4B, Table 2). This is similar to the phenotype previously reported using oep-MO to impair Nodal signaling (Riley and Phillips, 2003). PCR-based genotyping revealed that embryos forming medial otic vesicles were either fgf8^+/− heterozygotes or fgf8^−/−homozygotes, though heterozygotes were disproportionately represented over homozygotes by a margin of nearly 10:1 (n=57). The size of the otic vesicle was also distinctively larger in fgf8^+/−heterozygotes compared to fgf8^−/− homozygotes (not shown). Similar results were also obtained by treating fgf3^+/− intercross progeny with 70 μM SB431542 (Fig. 4M, Table 2), or by treating wild-type embryos with 70 μM SB431542 plus a very low dose (1–2 μM) of SU5402, an inhibitor of Fgf signaling (Fig. 4E and Table 2). The latter experiment used only one-twentieth of the dose of SU5402 required to fully block Fgf signaling during gastrulation (Léger and Brand, 2002; Maroon et al., 2002). Higher doses of SU5402 (≥5 μM) yielded no medial ear phenotype (Table 2). Analysis of early otic markers revealed that preotic domains of pax8 and pax2a already showed medial expression at earliest stages of otic induction (Fig. 4M, P). Moreover, otic-competence factors dlx3b and foxi1 (Solomon et al., 2004; Hans et al, 2004) were also expressed at the midline (Fig. 4G, J). Perturbing Nodal signaling alone, without partial impairment of Fgf signaling, was not sufficient to produce a medial ear phenotype in the vast majority (99.9%) of embryos (Fig. 4C, etc., and Table 2). Thus, partial loss of Nodal signaling combined with a modest reduction in Fgf signaling can result in ectopic induction of otic tissue at the midline. Nodal-Fgf-deficient embryos failing to produce medial ear tissue resembled embryos treated with 70 μM SB431542 alone. The reason for this variable response is not known.

Figure 4. Partial loss of Nodal and Fgf signaling results in medial expansion of otic tissue.

In the indicated columns, embryos were treated with 70 μM SB431542 beginning at 0.5 hpf (nodal ↓). A-C, otic vesicles at 28 hpf in wild-type embryos (A, C), and an fgf8^+/− heterozygote (B). D, E, Expression of ear markers dlx3b and pax5 (blue) and the hindbrain marker krox20 (red) at 28 hpf in a wild-type control (D) and a wild-type embryo treated 1 μM SU5402 beginning at 5.5 hpf (E). F-H, expression of foxi1 at 11 hpf in wild-type embryos (F, H) and an fgf8^+/− heterozygote (G, 2/42 affected). I-K, expression of dlx3b at 11.5 hpf in wild-type embryos (I, K) and an fgf8^+/− heterozygote (J, 4/56 affected). L-N, expression of pax2a at 12 hpf in wild-type embryos (L, N) and an fgf3^+/− heterozygote (M, 7/87 affected). O-Q, expression of pax8 and shh in wild-type embryos (O, Q) and an fgf8^+/− heterozygote (P, 4/51 affected). R-T, expression of fsta (blue) and krox20 (red) at 11 hpf in wild-type embryos (R, T) and an fgf8^+/−heterozygote (S, 40/52 affected). U-W, Expression of islet1 in the trigeminal placode at 11 hpf in wild-type embryos (U, W) and an fgf8^+/− heterozygote (V, 3/46 affected). All images show dorsal views (anterior to top). Scale bar, 50 μm.

Table 2.

Effects of varying Nodal- and Fgf-inhibition on production of medial otic vesicles.

			medial otic vesiclesc
Genotype	SB431542a	SU5402b	number	mean ± SEM	# expt
fgf8×15/+ intercross	30 μM	0	0/60	0% ± 0%	2
	45 μM	0	0/84	0% ± 0%	2
	70 μM	0	82/1260	6.5% ± 1.5%	15
	85 μM	0	0/31	0%	1
	95 μM	0	0/150	0% ± 0.0%	3
	70μM at 5 hpf	0	0/61	0%	1
fgf8ti282a/+ intercross	70 μM	0	55/1482	3.8% ± 0.5%	23
	70 μM at 2 hpf	0	0/31	0%	1
fgf3t24152/+ intercross	70 μM	0	7/240	5.4% ± 0.3%	3
oeptz257/+ intercross	70 μM	1 μM at 5.5 hpf	169/912	20.3% ± 2.8%	18
+/+	70 μM	0	1/915	0.1% ± 0%	12
	70 μM	1 μM at 5.5 hpf	24/309	8.7% ± 1.5%	8
	70 μM	2 μM at 5.5 hpf	7/211	4.1% ± 1.3%	7
	70 μM	5 μM at 5.5 hpf	0/90	0% ± 0%	2
+/+	70 μM	1 μM at 1 hpf	0/30	0%	1
	70 μM	10 μM at 1 hpf	0/60	0%	1
	70 μM	1 μM at 3.5 hpf	0/90	0%	1
	70 μM	2 μM at 6.5 hpf	1/121	0.8%	1
	70 μM	2 μM at 8 hpf	0/90	0%	1
	70 μM	1 μM 5–6 hpf	0/30	0%	1
	70 μM	2 μM 5–6 hpf	4/57	7.4%	1
+/+	70 μM, 90 μM at 2.8 hpf	1 μM at 5.5 hpf	1/16	6.3%	1
	70 μM, 180 μM at 2.8 hpf	1 μM at 5.5 hpf	0/29	0%	1
	70 μM, 240 μM at 2.8 hpf	1 μM at 5.5 hpf	0/27	0%	1
	70 μM, 90 μM at 4.0 hpf	1 μM at 5.5 hpf	2/28	7.1%	1
	70 μM, 180 μM at 4.7 hpf	1 μM at 5.5 hpf	2/27	7.4%	1
	70 μM, 240 μM at 4.7 hpf	1 μM at 5.5 hpf	2/31	6/5%	1
	70 μM, 400 μM at 4.7 hpf	1 μM at 5.5 hpf	1/53	1.9%	1

^aSB431542 added at 0.5 hpf and maintained thereafter, or added at a later time or followed by an increase in concentration as indicated.

^bSU5402 added at the indicated times and concentrations. Where indicated, embryos were treated for only one hour between 5–6 hpf.

^cNumber of embryos showing a medial otic vesicle at 24 hpf/total number of embryos, the percentage of affected embryos ± SEM, and the number of independent experiments (expt). SEM is not given for data based on one experiment.

Adjusting the time of addition and withdrawal of Nodal- and Fgf-signaling inhibitors implicated changes in axial development and early mesendoderm formation. To generate the medial ear phenotype, SB431542 had to be added by 0.5 hpf and maintained through at least 5.5 hpf. In contrast, SU5402 was required only during a window from 5–6 hpf, after which it could be removed (Table 2). Addition of SU5402 before 5 hpf potentiated the effects of SB431542 and blocked medial ear formation, and addition of SU5402 after 6 hpf also did not result in medial ear formation (Table 2). Because the period of sensitivity to SU5402 corresponds to the time when the organizer and mesendoderm are first forming, we examined mesendoderm formation to delineate its possible involvement in formation of medial otic tissue. As expected from the effects of inhibiting Nodal signaling, embryos treated with 70 μM SB431542 (including those induced to form medial otic tissue) lacked axial mesoderm as shown by the absence of shh expression (Fig. 4P, Q, data not shown). In contrast, paraxial cephalic mesendoderm, marked by fsta expression, formed as a single longitudinal band beneath the neural plate (Fig. 4S, T). These data are consistent with the possibility that redistribution of paraxial cephalic mesendoderm along the midline facilitates formation of medial otic tissue.

Paraxial cephalic mesoderm transfates to form medial otic tissue in Nodal-Fgf-deficient embryos

To better characterize the spatial relationships between preotic cells and surrounding tissues, we sectioned inhibitor-treated and control embryos after staining for pax8 expression. Control embryos showed the expected bilateral ectodermal domains of preotic pax8 expression (Fig. 5A, B). In embryos with medial otic tissue, however, only the lateral edges of the preotic domain of pax8 marked ectodermal cells, whereas all medial staining was located in mesendoderm, the layer between the neurectoderm and yolk (Fig. 5C, D). Double staining for pax8 and fsta showed that all embryos with medial otic cells coexpress both markers at 11 hpf (Fig. 5F). Between 11 and 12 hpf, however, fsta expression was lost from the head in most embryos while otic markers increased in intensity (Fig. 5G, H). In contrast, expression of fsta was maintained in developing somitic tissue (Fig. 5H), which proceeded to express myoD and form morphological somites (Fig. 5L and data not shown). Treated embryos also showed cranial expression of fibronectin1, a marker of lateral plate mesoderm, showing that there is not a general deficiency of head mesoderm (Fig. 5J). Based on changes in pax8 and the anterior domain of fsta, we hypothesized that cells derived from paraxial cephalic mesoderm, which normally do not participate directly in cranial placode formation, transfate to form otic tissue in Nodal-Fgf-deficient embryos. To test this possibility, we conducted lineage-tracing experiments. Embryos were injected at the one-cell stage with caged fluorescein-dextran, and nascent mesendodermal clones were marked by laser-irradiating small groups of marginal cells at 5 hpf (Fig. 5M). Subsequent fluorescein-staining confirmed that Nodal-Fgf-deficient embryos produced mesendodermal cells that involuted and migrated towards the animal pole during gastrulation in a manner similar to control embryos (Fig. 5N–O, Q–R). After gastrulation, some embryos that had formed medial otic tissue contained lineage-labeled cells within the preotic domain of pax2a at 13 hpf (Fig. 5P), or within the wall of the otic vesicle at 1 dpf (Fig. 5T). These data support the hypothesis that, in Nodal-Fgf-deficient embryos, paraxial cephalic mesendoderm migrates into the head and then switches fate to form otic tissue. In contrast, in control embryos we observed no descendants of marginal cells within the otic vesicle (n = 33, data not shown). Likewise, marginal cells in embryos exposed to 85 μM SB431542 failed to leave the margin and migrate towards the developing ear (Fig. 5S), similar to mutants with severe loss of Nodal signaling (Feldman et al., 2000; Carmany-Rampey and Schier, 2001; Dougan et al., 2003).

Figure 5. Paraxial cephalic mesoderm transfates to otic tissue in Nodal-Fgf-deficient embryos.

A, B, Transverse section of a wild-type control stained for pax8 and shh. (B) shows an enlargement of the boxed region in (A), with the mesodermal layer indicated (bracket). C, D, Transverse section through the otic domain of pax8 in a fgf8^+/− heterozygote treated with 70 μM SB431542. (D) shows an enlargement of the boxed area in (C). E-H, Double staining for pax8 and fsta expression (dorsal views) in a wild-type control at 11.5 h (E), and wild-type embryos treated with 70 μM SB431542 + 1 μM SU5402 and fixed at 11 hpf (F) or 11.5 hpf (G, H). The normal otic domain of pax8 (ot) and somitic expression of fsta (som) are indicated. I, J, Two-color staining for fn1 (blue) and krox20 (red) at 10 hpf in a wild-type control (I) and a wild-type embryo treated with 70 μM SB431542 + 1 μM SU5402 (J). K, L, double staining for pax2a and myoD in a wild-type control (K) and a wild-type embryo treated with 70 μM SB431542 + 1 μM SU5402 (L). The otic domain of pax2a (ot) and myoD-expressing somites (som) are indicated. M-T, lineage-tracing of prospective mesendoderm. Laser irradiation was used to photo-uncage fluorescein-dextran in small clusters of marginal cells wild-type embryos treated with 70 μM SB431542 + 1 μM SU5402 (M), and embryos were fixed and stained with anti-fluorescein antibody (blue) at 9.5 hpf (O), 13 hpf (P), or 24 hpf (T). Older specimens (P, T) were also stained by wholemount in situ hybridization for pax2a expression (red). The specimen in (T) shows a sagittal section (anterior to left) of a medial otic vesicle containing scattered lineage-marked cells (arrows). For comparison, lineage-labeled cells are shown in wild-type controls at 9.5 hpf (N) and 13 hpf (Q), and at 13 hpf in fgf8^+/− heterozygotes treated with 70 μM SB431542 (R) or fgf8^+/− heterozygotes treated with 85 μM SB431542 (S). The specimen in (R) also shows krox20 expression (red). Wholemount specimens show dorsal views with anterior to the top. Scale bar, 25 μm.

In addition to forming ectopic otic tissue, Nodal-Fgf-deficient embryos also produced an enlarged trigeminal placode that spanned the midline at the level of the anterior hindbrain, as shown by a transverse stripe of islet1-expressing cells at 11 hpf (Fig. 4V). This corresponds to the anterior limit of fsta expression (Fig. 4N), raising the possibility that mesendoderm could also play a role in forming ectopic trigeminal tissue. No other cranial placodes showed similar patterns of development (data not shown).

Inactivation of Bmp-antagonists blocks formation of medial otic tissue

We hypothesized that redistribution of mesodermally derived Bmp-antagonists in Nodal-Fgf-deficient embryos facilitates formation of medial otic tissue. To test this, we treated chd/fsta double- or chd/fsta/cv2 triple-morphants with 70 μM SB431542 plus 1 μM SU5402. To increase the incidence of the medial ear phenotype, we used oep^+/− intercross progeny for this experiment. Under these conditions, about 20% of non-morphant embryos produced medial otic vesicles (Fig. 6A, Table 2). The reason for this phenotypic enhancement is not known but we could not mimic the effect by simply increasing the concentration of SB431542 at later stages (Table 2). Despite the enhanced oep+/− background, the incidence of morphologically detectable medial otic vesicles declined to only 1% of chd/fsta double-morphants (n = 126), and none were detected in chd/fsta/cv2 triple-morphants (n = 65). We also examined expression of the otic-specific marker cldna (Kollmar et al., 2001), which provides a more sensitive assay for the presence of otic tissue. In chd/fsta double-morphants, 6 % (n = 165) of embryos showed a disorganized band of cldna-expressing cells spanning the midline (Fig. 6B, E). Strikingly, however, medial cldna expression was never detected in chd/fsta/cv2 triple-morphants (Fig. 6B, F; n = 44). A possible trivial explanation for the loss of medial otic tissue is that knockdown of Bmp-antagonists in inhibitor-treated embryos further ventralizes the embryo, thereby ablating paraxial cephalic mesendoderm. However, the presence of fsta expression at the midline during gastrulation indicates that paraxial cephalic mesendoderm is still produced under these conditions (Fig. 6G, H). Thus, expression of Bmp-antagonists by paraxial cephalic mesoderm is essential for these cells to trans-differentiate into otic tissue.

Figure 6. Development of medial otic tissue requires mesendodermal Bmp-antagonists.

A, B, Frequency of inhibitor-treated embryos producing medial otic vesicles as determined by morphological criteria (A) or expression of cldna (B). Embryos obtained from outcrosses of oep^+/− heterozygotes were treated with 70 μM SB431542 + 1 μM SU5402. Where indicated, embryos were also injected with 2.5 ng chd-MO, 5 ng fsta-MO and 5 ng of cv2-MO2 (1X), or with half-doses of morpholino (0.5X). Data are presented as the mean and SEM of 2–8 independent experiments, with total sample sizes indicated. C-F, Representative specimens showing cldna expression at 24 h (C-F). G, H, expression of fsta at 11 hpf shows that paraxial cephalic mesendoderm is not ablated by the combination of SB432542, SU5402 and 1X morpholinos. All images are dorsal views (anterior to top). Scale bar, 50 μm.

Patterning within medial otic vesicles

The unusual origin of medial otic tissue, and its ectopic positioning beneath the hindbrain, led us to examine patterning within medial otic vesicles. We previously showed that expression of fgf3 in rhombomere 4 is required for proper expression of anterior otic markers pax5 and nkx5.1 (Kwak et al, 2002). Both anterior markers show expression throughout otic vesicles that span the midline, whereas the posterior otic marker fsta is lost (Fig. 7A–F). These changes in AP patterning are consistent with close apposition to rhombomere 4 (Fig. 4D). In contrast, dlx3b and otx1 were localized normally to the dorsal and ventral walls, respectively, indicating that DV patterning is relatively normal in such otic vesicles (Fig. 7G–J).

Patches of hair cells and associated otoliths also formed within medial otic vesicles, though the number and positioning were highly variable (Figs. 4B, 7Q). Expression of various markers of hair cell differentiation, including pax2a and brn3c:GFP, indicate that hair cell differentiation occurs relatively normally (Fig. 7R, and data not shown). Neurons of the statoacoustic ganglion (SAG), which are derived solely from cells in the ventral wall of the otic vesicle, also form in embryos with medial otic vesicles (Fig. 7L). However, rather than forming bilateral patches of neurons (Fig. 7M, O), medial otic vesicles are associated with transverse stripes of neurons that coalesce along the anterior edge of the vesicle (Fig. 7N, P). The identity of these neurons is not certain, but we infer they are SAG neurons based on their proximity to the otic vesicle and because ngn1- and neuoD-expressing neuroblasts can be seen delaminating from the otic vesicle (Fig. 7L, and data not shown).

DISCUSSION

Many aspects of otic induction are conserved in all vertebrates, especially the requirement for Fgfs secreted by the developing hindbrain and subjacent mesendodermal (Ladher et al., 2000; Phillips et al., 2001; Maroon et al., 2002; Alvarez et al., 2003; Léger and Brand, 2003; Liu et al., 2003; Wright and Mansour, 2003; Ladher et al., 2005; Zelarayan et al., 2007). However, there is still much to learn about the underlying molecular mechanisms and signaling interactions governing this process. Our findings have clarified the role of mesendoderm in otic induction in zebrafish, and we have identified additional inductive signals likely to play a conserved role in all vertebrates.

A role for Bmp-antagonists

While Fgfs constitute the predominant otic-inducing signals from mesendoderm, we have shown that Bmp-antagonists expressed in head mesendoderm are also vital for this function. In chick and Xenopus embryos, it has been shown that misexpression of various Bmp-antagonists can widen the domain of preplacodal ectoderm from which all placodes are derived (Brugmann et al., 2004; Glavic et al., 2004; Ahrens and Schlosser, 2005; Litsiou et al., 2005). In the present study, we used loss-of-function approaches to test the roles of three Bmp-antagonists in vertebrate ear development. We focused on zebrafish chd, fsta and cv2 because they are all expressed at the appropriate time and place to affect otic induction (Fig. 2; Miller-Bertoglio et al., 1997; Bauer et al., 1998; Rentzsch et al., 2006). Knocking down all three genes severely restricts otic induction. This does not reflect a failure to specify preplacodal ectoderm since chd/fsta/cv2-triple morphants show timely expression of preplacodal markers dlx3b, eya1 and six4.1 (Fig. 3). Moreover, these embryos produce a variety of placodal derivatives, albeit in smaller domains. The most severe deficiencies are seen in the trigeminal and otic placodes, as shown by near loss of some early markers, as well as the very small size and gross patterning defects in the otic vesicle. The fact that these defects ensue despite initial formation of preplacodal ectoderm suggests that subsequent stages of placodal development require increasingly stringent inhibition of Bmp. Indeed, Esterberg and Fritz (2008) have shown that expression of cv2 along the edges of the neural plate requires prior activity of dlx3b and dlx4b in the preplacodal ectoderm. Thus, newly specified preplacodal ectoderm actively enhances local antagonism of Bmp, presumably facilitating subsequent placodal development. Combining chd/fsta/cv2-knockdown with loss of either fgf3 or fgf8 ablates preotic expression of pax8 and blocks formation of a morphological vesicle, though a few disorganized cldna-positive cells persist (Fig. 2L). These findings support a model in which mesendoderm-derived Bmp-antagonists, acting in parallel with Fgf, are essential for normal otic placode induction.

The requirement for Bmp-antagonism reflects dynamic changes in signaling requirements for ectodermal patterning. Bmp is initially required to establish non-neural ectoderm during late blastulation/early gastrulation (Kishimoto et al., 1997; Neave et al., 1997; Wilson et al., 1997; Nguyen et al., 1998). By mid-gastrulation, both Fgf and Bmp-antagonists are required to establish preplacodal ectoderm (Glavic et al., 2004; Ahrens and Schlosser, 2005; Litsiou et al., 2005) and these signaling conditions must be maintained past the end of gastrulation in order to induce and maintain the otic placode (Léger and Brand, 2002; Maroon et al., 2002; Esterberg and Fritz, 2008). It is interesting that Nodal-Fgf-deficient embryos do not appear to produce ectopic epibranchial placodes, which normally abut the lateral edge of the otic placode and require similar signaling interactions (Nechiporuk et al., 2007; Nikaido et al., 2007; Sun et al., 2007). However, it is possible that epibranchial placodes require relatively lower Fgf and higher Bmp levels, conditions that are not met beneath the hindbrain in part because of the influence of mesendodermal Bmp-antagonists.

There has been some controversy as to whether Cv2 antagonizes or promotes Bmp signaling (Moser et al., 2003; Coles et al., 2004; Kamimura et al., 2004; Ikeya et al., 2006). However, recent studies reveal that the function of Cv2 is highly context-dependent and, in addition, separate domains within Cv2 mediate its opposing functions (Rentzsch et al., 2006; Ambrosio et al., 2008; Serpe et al., 2008; Zhang et al., 2008). At high concentrations, Cv2 can bind directly to Bmp and/or type-I receptor to block receptor-ligand interactions, and this antagonistic function is further stimulated by local interactions with extracellular matrix. Such conditions are likely to be met nearest sites of active cv2 expression. We find that knockdown of cv2 enhances the developmental defects caused by chd-MO and fsta-MO, supporting the notion that Cv2 normally works together with Chd and Fsta to inhibit Bmp signaling in preplacodal ectoderm.

It is as yet unclear how broadly conserved the above functions are amongst vertebrates. Chordin expression in chick and mouse embryos is limited to axial mesoderm during gastrulation (Chapman et al., 2002; Bachiller et al., 2003) while cranial expression of Cv2 is limited to the dorsal hindbrain after gastrulation (Coffinier et al., 2002; Moser et al., 2003). Although both proteins can diffuse over considerable distances, it is not known to what extent they affect preplacodal or otic cells, respectively. In contrast, Follistatin is expressed in paraxial cephalic mesendoderm during gastrulation in chick and mouse (Feijen et al., 1994; van den Berg et al., 2007), indicating it could play a conserved role in preplacodal and/or placodal development. A similar role has also been proposed for Cerberus and Dan in chick (Litsiou et al., 2005) based on early expression in head mesendoderm (Ogita et al., 2001; Chapman et al., 2002). Loss-of-function studies are needed to test the functions of these genes in other vertebrates.

Relative roles of hindbrain and mesendoderm

We have resolved an apparent contradiction regarding the requirement for mesendoderm- and hindbrain-derived signals in otic induction in zebrafish. We previously used oep-morpholino to disrupt Nodal-dependent mesoderm formation. This diminishes and delays otic induction. Unexpectedly, however, when oep-MO is injected into progeny of ace (fgf8) heterozygotes, up to 5% of embryos show a dramatic medial expansion of otic tissue (Riley and Phillips, 2003). We now attribute this paradoxical phenotype to variable efficacy of oep-MO (Nasevicius and Ekker, 2000). In the current study we have used the pharmacological inhibitor SB431542 to more precisely and reproducibly control the level of Nodal signaling. Treating embryos with a high dose that abolishes Nodal signaling, combined with disruption of either fgf3 or fgf8, results in complete abrogation of otic induction. This shows that the remaining Fgf signal from the hindbrain is not sufficient to induce otic tissue. Similar results are seen in chick, in which extirpation of paraxial cephalic mesoderm blocks otic induction despite continuing expression of Fgf3 in the hindbrain (Kil et al., 2005). Thus, there appears to be a threshold of Fgf signaling below which otic induction cannot occur. Mesendodermal Fgfs clearly play a vital role in achieving this threshold, a function that could also be enhanced by Bmp-antagonists.

Transfating of paraxial mesendoderm to placodal tissue

In sharp contrast to the effects of fully blocking Nodal and Fgf8 (or Nodal and Fgf3), partial inhibition of Nodal and Fgf leads to production of a single large otic vesicle spanning the width of the hindbrain beneath r4. Under these conditions, paraxial mesendoderm forms at the midline at the expense of axial mesendoderm. Lineage-tracing and analysis of gene expression patterns show that paraxial mesendoderm initially forms and migrates beneath the hindbrain and then, surprisingly, transdifferentiates into placodal tissue and contributes directly to formation of the otic vesicle. Analysis of marker genes reveals a dynamic shift in expression between 11 to 12 hpf, when fsta begins to fade and pax8 increases in intensity. Bmp-antagonists expressed by these cells as they migrate are essential for stimulating their subsequent transdifferentiation, as knock-down of chd, fsta and cv2 blocks formation of medial otic tissue. Hindbrain-expression of fgf3 and fgf8 in r4 also appears critical, since ectopic otic tissue is restricted to cells beneath r4. Additionally, full loss of fgf3 or fgf8 reduces the frequency and amount of medial otic tissue induced. Therefore the signaling requirements for forming medial otic tissue are virtually the same as for induction of endogenous otic placodes, further highlighting the importance of signals from both hindbrain and mesendodermal tissue.

We still do not understand the mechanistic basis for transfating of paraxial cephalic mesendoderm into ectoderm, but studies on mesoderm induction provide some clues. In Zoep^−/−mutants, a number of regional mesodermal markers are initially induced but are not properly maintained (Warga and Kane, 2003). In MZoep^−/− and cyc-sqt mutants, early mesodermal markers are expressed only briefly, after which most cells adopt ectodermal fates (Carmany-Rampey and Schier, 2001). Similarly, loss of Fgf signaling results in failure to maintain axial mesodermal fates (Fletcher and Harland, 2008). We speculate that partial loss of Nodal and Fgf modifies early mesoderm induction such that paraxial cephalic cells undergo involution and migration but are not stably committed to mesendodermal fates. The period of sensitivity to Fgf-inhibition is between 5–6 hpf when the first mesendodermal cells are being specified. Fgf and Nodal signaling normally engage in mutual feedback amplification through Fgf-mediated upregulation of Oep (Mathieu et al., 2004). Hence, partial inhibition of both Nodal and Fgf would likely weaken this feedback loop. Subtle modulation of the Fgf-Oep feedback loop could also explain why the incidence of the medial ear phenotype is significantly elevated in oep^+/−heterozygotes (Table 2). Interestingly, during normal development there is one population of cells that undergoes a mesendoderm-to-ectoderm fate switch in zebrafish embryos: A subset of cells in nascent chordamesoderm are induced by Delta-Notch signaling to rejoin the epiblast and contribute to the floorplate of the neural tube (Appel et al., 1999). Therefore the ability of head mesendoderm to transfate in Nodal-Fgf-deficient embryos may reflect modification of a normal developmental program.

The ability of mesendodermal cells to switch to placodal fates has a number of implications. Normally, otic induction is a prolonged process involving multiple signaling interactions (Gallagher et al., 1996; Groves and Bronner-Fraser, 2000; Riley and Phillips, 2003). Mesendodermal cells clearly follow a very different developmental history than do normal placodal cells, indicating that early regulatory events are much less critical than later events. In some ways mesendodermal transfating is reminiscent of grafting experiments showing that ectoderm from foreign sites can be induced to form otic tissue when transplanted to the ear-forming region (Yntema, 1933, 1950; Gallagher et al., 1996; Groves and Bronner-Fraser, 2000). The zone of otic-competence is initially widespread throughout the ectoderm but becomes increasingly restricted as cells begin to commit to other fates. In the case of Nodal/Fgf-deficient embryos, commitment of paraxial cephalic mesendoderm is impaired, possibly allowing these cells to remain plastic in their developmental potential until they encounter the otic-inducing environment near r4.

EXPERIMENTAL PROCEDURES

Fish Strains, Staging and Genotyping

Wild-type zebrafish strains were derived from the AB line (Eugene, OR). Embryos were developed in fish water containing methylene blue at 28.5°C and staged according to standard protocols (Kimmel et al., 1995). The mutant alleles used in this study include fgf8^ti282a (or ace), fgf3^t24152 (or lia) and oep^tz257, which have been described previously (Brand et al., 1996; Hammerschmidt et al., 1996; Herzog et al., 2004). A new allele of fgf8, fgf8^×15, was induced with ENU and identified in a non-complementation screen (B. Riley, unpublished observations). PCR amplification and sequencing of the fgf8^×15 sequence revealed a C-to-T transition in the coding region that introduces a premature stop after amino acid residue 95. This results in loss of the C-terminal 115 amino acids, suggesting that fgf8^×15 is a null allele. PCR genotyping was carried out as described previously (Mathieu et al., 2004) with minor modifications. The x15 allele was identified using forward primer, 5′ CTTCGGATTTCACATATTTATGCCCGTATGTATGCATATC 3′ and reverse primer, 5′ CAGTTTTAGTAAGTCACAAAAGTGATGACTTTTTCACATA-3′. PCR products were digested with pleI, yielding wild-type fragments of 190 bp and 110 bp, while the mutant PCR product remains intact at 300 bp. The Tg(brn3c:Gap43-GFP)^s356t transgenic line was provided by H. Baier (Xiao et al., 2005). The Tg(islet2b:GFP)^zc7 transgenic line was provided by A. Pittman and C.-B. Chien (Pittman et al., 2008).

Inhibitor Treatment

SB431542 (Sigma) and SU5402 (Calbiochem) were dissolved in DMSO to prepare a 10 mM stock solution. The stock was diluted to a working concentration in fish water. Embryos were treated without removing their chorions. Treatment was carried out in 24-well plates, with 40 embryos in 0.5 ml of solution per well. Controls were incubated in embryo water containing an equal concentration of DMSO to that of treated embryos. No effects were observed by exposure to DMSO vehicle alone.

Morpholino Injection

Morpholino oligomers (MOs) were obtained from Gene Tools, Inc. MOs for chd, cv2 and oep were described previously (Nasevicius and Ekker, 2000; Phillips et al., 2001; Rentzsch et al., 2006). Unless stated otherwise, 5 ng of each morpholino was injected into one-cell embryos. chd-MO was injected at 2.5 ng per embryo. For most experiments involving cv2-knockdown, we injected 5 ng of cv2-MO2 (Rentzsch et al., 2006), which in our hands reproducibly caused mild ventralization indicative of an anti-Bmp role for Cv2. The effects of cv2-MO1 were more variable, but at the standard concentration this MO caused nonspecific embryonic lethality by mid-somitogenesis. At sub-lethal concentrations (≤2.5 ng/embryo) cv2-MO1 behaved similarly to cv2-MO2 (5 ng/embryo) and enhanced the effects of chd-MO and fsta-MO. To knockdown fsta, either of two MOs was used: Translation blocker fsta-MO1, 5′ CGCTTTAGCATCCTTAGCATGTTTA 3′ ; and splice blocker fsta-MO2, which targets the exon3-intron3 junction (E3I3), 5′ TGTGTTACCTACTTTTGCATTTGCC 3′. By themselves, fsta-MOs had little effect on embryonic patterning, but both showed identical additive interactions with chd-MO and cv2-MO (described in Results). Efficacy of fsta-MO2 was evaluated using RT-PCR (Fig. 2S) with primers as follows: P1 (forward primer), 5′ CTAAAGCGTCAGCAGCTCC 3′ and P2 (reverse primer), 5′ CACAGGACATCACGACACG 3′. To knockdown fgf3, a splice blocker was used to target the exon2-intron2 (E2I2) splice junction: 5′ AAACGGTTTACTCACTTAAAGGGTT 3′, which produced a phenotype identical to the null allele of fgf3, lia^t24152 (Herzog et al., 2004).

Whole-Mount In Situ Hybridization and Sectioning

Whole-mount in situ hybridization was carried out as described previously (Phillips et al. 2006). For sectioning, embryos were embedded in Immunobed resin (Polysciences No. 17324) and cut into 8 μm sections.

Lineage Tracing

Lineage tracing using laser-uncaging of caged fluorescein dextran was performed as described (Carmany-Rampey and Schier, 2001) with minor modifications. Uncaged embryos were allowed to develop to the appropriate developmental stage and fixed and processed by in situ hybridization (Phillips et al., 2004) and/or immunohistochemistry. Fast Red (Roche) was used in the alkaline phosphatase reaction for in situ hybridization. To detect labeled cells in the same specimen, alkaline phosphatase from the first reaction was inactivated in 0.1 M glycine (pH 2.2). Embryos were washed 5 times with PTW (Phosphate-Buffered Saline containing 0.1% Tween-20) and incubated overnight at 4°C with a 1:3000 dilution of anti-fluorescein alkaline phosphatase Fab fragment (Roche). NBT and BCIP were used as substrates in successive alkaline phosphatase reactions.