FoxN3 is required for craniofacial and eye development of Xenopus laevis

Authors


Abstract

A functional knockdown of FoxN3, a member of subclass N of fork head/winged helix transcription factors in Xenopus laevis, leads to an abnormal formation of the jaw cartilage, absence or malformation of distinct cranial nerves, and reduced size of the eye. While the eye phenotype is due to an increased rate of apoptosis, the cellular basis of the jaw phenotype is more complex. The upper and lower jaw cartilages are derivatives of a subset of cranial neural crest cells, which migrate into the first pharyngeal arch. Histological analysis of FoxN3-depleted embryos reveals severe deformation and false positioning of infrarostral, Meckel's, and palatoquadrate cartilages, structural elements derived from the first pharyngeal arch, and of the ceratohyale, which derives from the second pharyngeal arch. The derivatives of the third and fourth pharyngeal arches are less affected. FoxN3 is not required for early neural crest migration. Defects in jaw formation rather arise by failure of differentiation than by positional effects of crest migration. By GST-pulldown analysis, we have identified two different members of histone deacetylase complexes (HDAC), xSin3 and xRPD3, as putative interaction partners of FoxN3, suggesting that FoxN3 regulates craniofacial and eye development by recruiting HDAC. Developmental Dynamics 236:226–239, 2007. © 2006 Wiley-Liss, Inc.

INTRODUCTION

Neural crest cells originally delaminate from the neural tube, migrate laterally, and differentiate to multiple cell types. Many organs and tissues in vertebrates are derivatives of the neural crest, such as neurons and glia cells of the peripheral nervous system, eye tissues like the sclera, the cornea, the choroidea, endocrine cells, and several bones of the craniofacial cartilage system. Graft experiments in chicken embryos with quail (Le Douarin, 1980) and later in amphibians using the Xenopus laevis/Xenopus borealis chimeric system (Thiebaud, 1983) allowed cell fate mapping and observation of cell migration. New techniques of microscopy, new vital dyes and fluorescence proteins facilitated and amended these experiments (Sadaghiani and Thiebaud, 1987; Borchers et al., 2000). The discovery and application of inducible photoconvertible proteins like Eos-FP (Wiedenmann et al., 2004; Wacker et al., 2006) in combination with fluorescence microscopy eliminates the need of grafting embryos, thereby offering the advantage of less mortality and less artificial results. Despite this progress, the complex molecular mechanisms underlying neural crest determination, migration, and differentiation are still not fully understood. Growth factors and transcription regulators, which play a role in development, migration, and differentiation of neural crest, include the FGFs (Mayor et al., 1997), BMPs (Nguyen et al., 1998), Wnts (Chang et al., 1998), Pax3 (Bang et al., 1999), Zinc finger proteins Zic3 and Slug (Nakata et al., 1997; LaBonne and Bronner-Fraser, 1998), Cadherin 11 (Borchers et al., 2001), HoxA2, HoxA3, and HoxD3 (Manley et al., 1997; Pasqualetti et al., 2000), Sox9 and Sox10 (Spokony et al., 2002; Aoki et al., 2003), as well as the fork head factors XBF2 and FoxD3 (Gomez-Skarmeta et al., 1999; Dottori et al., 2001; Pohl and Knöchel, 2001).

In Xenopus embryogenesis, the cranial neural crest cells migrate ventrally in four mainstreams to the pharyngeal arches (Baltzinger et al., 2005). The most anterior segment is the mandibular arch. From stage 21 onward, mandibular crest cells move ventrally and some from the dorsal part rostrally over the eye vesicle. Thereby, they constitute elements of the upper and lower jaw including Meckel's cartilage and palatoquadrate. Additionally, they take part in forming non-cartilage structures, like cornea and sclera of the eye, parts of the ear and the gasserian and geniculate ganglia (Mayor et al., 1999; Cerny et al., 2004; Baltzinger et al., 2005). The second (hyoid) arch arises from the anterior part of the rhombencephalon, migrates ventrally to the bottom of the pharynx, and forms the ceratohyale. Moreover, it forms cartilage structures involved in the respiration tract and takes part in the preotic region to form the acusticofacial ganglion complex (Sadaghiani and Thiebaud, 1987; Mayor et al., 1999; Baltzinger, 2005). The third and fourth pharyngeal arches arise from the posterior part of the rhombencephalon, form the anterior as well as the posterior branchial arches, and take part in the development of the thymus, the epithelium of the pharynx, and the heart (Mayor et al., 1999).

The chondrocranium of tadpoles has been analysed in detail (Dreyer, 1915; Trueb and Hanken, 1992; Haas, 1995; Svensson and Haas, 2005). It represents a complex structure that shows various alterations in different types of anurans. In addition, the cranial cartilage of anuran larvae differs from other vertebrates by the presence of an infra- and a suprarostral cartilage. While the morphology of cartilage is well described, the molecular mechanisms responsible for jaw formation in vertebrates are less understood. However, recent reports have shown that several genes and gene groups, like the Dlx genes, the Hox genes, and the Nkx3 genes, are involved in this process (Depew et al., 2005; Baltzinger et al., 2005; Tucker et al., 2004). In the present study, we demonstrate that FoxN3, a member of subclass N of fork head/winged helix factors, is an additional regulator of cranial cartilage development. The fork head box (Fox) family represents a large family of transcription factors that is subdivided into several subclasses (Kaestner et al., 2000; Katoh and Katoh, 2004). The FOXN3 gene was initially described in human by screening for genes that are able to suppress DNA damage-activated checkpoint mutations in Saccharomyces cervisiae by restoring G(2)/M cell cycle arrest (Pati et al., 1997). FOXN3 has further been shown to be downregulated in oral squamous cell carcinoma (OSCC) (Chang et al., 2005). Functional analyses revealed that FOXN3 could bind to yeast Sin3 (Scott and Plon, 2003). Sin3 is a highly conserved factor in all eukaryotes from yeast to human. It is known as a co-repressor in the histone deacetylase complexes I and II.

We have recently analysed the temporal and spatial expression of FoxN genes in the early development of Xenopus laevis (Schuff et al., 2006). FoxN3 transcripts are found within the animal half of gastrulating embryos. In post-gastrula embryos, neural crest cells and the early eye field show strong expression of FoxN3. At late tadpole stages, the pharyngeal arches and the eyes are stained. We here show by loss-of-function experiments that FoxN3 is required for the correct development of the jaw and the eye. Histological analysis of FoxN3-depleted embryos demonstrates not only a reduced eye size and oedema, but also a deformed structure and false positioning of craniofacial cartilage elements derived from the mandibular and hyoid arches. We demonstrate that the reduced size of eyes results from an increased rate of apoptosis. By GST-pulldown analysis, we have identified the histone deacetylase complex (HDAC) components Sin3 and RPD3 as putative interaction partners indicating that FoxN3 might act as a repressor by recruiting HDAC to repress target genes.

RESULTS

FoxN3-MO Inhibits Translation of FoxN3/GFP Fusion Transcripts In Vivo

A loss of function study was used to investigate the role of FoxN3 during Xenopus development. Because different splice variants of FoxN3 are composed of different leader exons and by the alternative splicing of a coding exon (Schuff et al., 2006), we used an antisense morpholino beginning with the translation start codon present in all types of transcripts. We first tested the ability of FoxN3-MO to specifically inhibit FoxN3 protein synthesis in vivo. Therefore, we fused a 300-bp 5′-coding sequence of FoxN3 containing the putative morpholino binding site to GFP (FoxN3/GFP). For control, we generated a FoxN3/GFP construct containing six silent mutations in the first 24 bases also beginning with the translation start (FoxN3Res/GFP) (Fig. 1A). In sum, 500 pg RNA of these GFP fusion constructs were injected either alone or together with varying amounts of FoxN3-MO or Co-MO, respectively, in each blastomere of 2-cell stage embryos.

Figure 1.

Inhibition of in vivo translation of a FoxN3/GFP fusion construct by FoxN3-MO. The 5′-coding sequence (300 bp) of FoxN3 including the translation start codon were fused to GFP (FoxN3/GFP). Alternatively, six silent mutations in the putative morpholino binding site were introduced (FoxN3Res/GFP). RNA of these constructs was injected into both blastomeres at the 2-cell stage and embryos were analysed at stage 31. A: 5′-coding sequence (25 bp) of FoxN3 is reverse complementary to the FoxN3 morpholino (FoxN3-MO) sequence. Silent mutations in the FoxN3Res/GFP construct are indicated. B,D,F: Lateral views on embryos injected with 0.5 ng FoxN3/GFP RNA (B), co-injected with 13 ng Co-MO (D) or with 13 ng FoxN3-MO (F) into each blastomere. C,E,G: Lateral views on embryos injected with 0.5 ng FoxN3Res/GFP RNA containing the silent mutations (C), co-injected with 13 ng Co-MO (E) or with 13 ng FoxN3-MO (G) into each blastomere. Embryos are photographed under normal light (left) and under UV light (right). Fluorescence is visible in embryos with FoxN3Res/GFP RNA injection and FoxN3-MO (G), but not with FoxN3/GFP RNA and FoxN3-MO (F).

Injection of FoxN3/GFP or FoxN3Res/GFP and co-injections of these RNAs with Co-MO resulted in bright fluorescence (Fig. 1B–E). In contrast, co-injection of FoxN3/GFP with 13 ng FoxN3-MO completely abolishes fluorescence (Fig. 1F). However, the same amount of FoxN3-MO does not affect the translation of injected FoxN3Res/GFP RNA (Fig. 1G). These results indicate that the FoxN3-MO can specifically repress the translation of FoxN3 transcripts within the embryo.

Knockdown of FoxN3 in Xenopus Results in Small Eyes and Craniofacial Malformations

The injection of 13 ng FoxN3-MO into each blastomere of 2-cell stage embryos does not lead to any visible phenotype until stage 39. The first morphological changes can be observed in the eye (shown for stage 41 in Fig. 2A). From stage 43 onwards, FoxN3-MO-injected embryos show malformation in the most anterior head region (Fig. 2B). Interestingly, when FoxN3-MO-injected embryos were grown until stage 45–48, we observed additional developmental abnormalities, like smaller brains and oedema in the head and trunk region (Fig. 2C,D). At stage 48, the vast majority of FoxN3-MO-injected embryos show a malformation of the craniofacial cartilage. Moreover, many of these embryos show false positioning of the eyes, which might be the consequence of malformed cartilage structures, smaller brain and oedema. Injection of 13 ng FoxN3-MO together with GFP RNA in only one blastomere of a 2-cell stage embryo results in half-side deformed cartilage and reduced eye, while the oedemas are also present in the uninjected side (Fig. 2D). All affected embryos are not able to take up food and die before stage 50.

Figure 2.

Phenotype of FoxN3-MO injected embryos during embryogenesis. Embryos were injected with morpholino (13 ng/blastomere) either in two (A–C) or in one blastomere (D) at 2-cell stage. A: Comparison of FoxN3-MO-injected embryos (right side) and Co-MO-injected embryos (left side) at stage 41, lateral view. The FoxN3-MO-injected embryos reveal smaller eyes, while the trunk appears to be quite normal. B: Same analysis at stage 43. The control embryos (left side) grow normal, while FoxN3-MO-injected show severe defects. The endophthalmus of the eyes is clearly visible. The head becomes malformed and the whole embryos become oedemic. C: Magnification of the head of a stage-45 embryo (dorsal view). Cranial structure of Co-MO-injected embryo (left side) is normal. The FoxN3-MO-injected embryo (right side) is smaller. The nose and mouth region is invaginated. D: Dorsal view of an unilaterally FoxN3-MO-injected embryo at stage 45 (bottom side is injected). The left eye is clearly reduced, as well as the anterior cartilage. E: Diagram of eye and cartilage phenotypes at stage 48 presented as dose dependency of FoxN3-MO. Indicated amounts of FoxN3-MO were injected into each blastomere at the 2-cell stage. Embryos were scored for eye phenotype (red) and for cartilage phenotype (blue). The numbers at the bottom indicate injected embryos and the percentage of surviving embryos at stage 48.

To investigate the dose dependency of these effects, increasing amounts of FoxN3-MO were injected and embryos were cultivated until stage 48. Resulting malformations were scored as percentages of injected embryos (Fig. 2E). With less than 1 ng per blastomere, no effect was observed. Injection of 5 ng FoxN3-MO results in 14% abnormalities of the injected embryos, while the Co-MO had no effect. With higher amounts of FoxN3-MO, the rate of malformations increased, but also the mortality rate, especially during gastrulation. The eye size is reduced in 62% of embryos injected bilaterally each with 13 ng FoxN3-MO, but this effect cannot be enhanced by using higher amounts of the morpholino. When 20 ng FoxN3-MO was injected into each blastomere of 2-cell stage embryos, all embryos died during gastrulation (data not shown).

The Small Eye Phenotype Does Not Result From a Reduction of the Early Eye Field

To investigate the molecular basis of the small eye phenotype, we have investigated FoxN3-depleted embryos by whole mount in situ hybridization using xRx as a marker for the early eye field (Mathers et al., 1997). FoxN3-MO was co-injected with GFP into one blastomere of 2-cell stage embryos. At stage 15, embryos were sorted for left/right injections, grown to the desired stage, and fixed for the hybridization procedure. However, no alteration of the size of the eye field could be observed in comparing the injected side to the uninjected side at stage 16 or stage 30 (Fig. 3A,B). Even at stage 35, the embryos did not exhibit any difference in the size of the forming retina (Fig. 3C,C'). An alteration becomes obvious at stage 39 (Fig. 3D,D') and is clearly documented by sections through the eye at stage 48 (Fig. 3E,F). The eye at the FoxN3-depleted side is smaller, but it contains all the elements of the normal eye including the lens and all retinal layers. It seems that the small eye phenotype does not reflect any failure of differentiation of distinct cells forming the retina layers, but the number of cells (which can be counted by red-stained nuclei) is dramatically reduced. Therefore, we suppose that FoxN3 is primarily not required for the differentiation of eye structures, as it is known for mouse foxn4, another member of the FoxN subclass (Li et al., 2004). However, depletion of FoxN3 leads to a significant reduction in the number of retina cells, even though the size of the early eye field is not affected.

Figure 3.

Analysis of the FoxN3 small eye phenotype. Thirteen nanograms of FoxN3-MO were co-injected with 200 pg GFP RNA in one blastomere at 2-cell stage. Asterisks denote the injected side. A,B: Whole mount in situ hybridization of unilaterally FoxN3-MO-injected embryos showing xRx staining in the eye field. A: Frontal view of a stage-16 embryo. B: Dorsal view of a stage-30 embryo. The insert shows a magnification of the right and left eye. No difference in xRx expression between injected and uninjected side is noted. C,C': Comparison of the injected (FoxN3-MO) and uninjected side of an embryo at stage 35. C' shows the injected side. No difference in the size of the forming eyes is observed. D,D': Right and left view of an unilaterally injected embryo, stage 39. D' shows the injected side. The eye at the FoxN3-MO injected side is clearly reduced in comparison to the uninjected side. E, F: Sections of Alcian blue–stained eyes of the uninjected control side (E) and the FoxN3-MO-injected side (F) of an embryo at stage 48. F: The eye is smaller but shows a normal structure. The different layers of the retina are clearly visible indicating that the differentiation of the retina is not affected by FoxN3-MO injection. The black arrows denote the optic nerve, indicating identical planes of section in E and F.

Histological Analysis of FoxN3-Depleted Embryos Reveals Defects of Mandibular and Hyoid Structures

To evaluate which cartilage structures are affected by FoxN3 knockdown, histological staining was performed (Fig. 4). Alcian blue–stained skeletal preparations from injected tadpoles at stage 48 reveal severe cartilage defects in craniofacial elements (Fig. 4C,D) compared to Co-MO-injected embryos (Fig. 4A,B). Increasing severity and frequency of phenotypes was observed in a dose-dependent manner. Deformations of affected craniofacial cartilage structures appear in a graded way in malformed embryos. The most anterior jaw elements derived from the mandibular arch, the Meckel's, and the infrarostral cartilage are strongly reduced and deformed (Fig. 4C). The ceratohyale, which is derived from the hyoid arch, is deformed at the anterior side. The right and the left part are widely separated from each other (Fig. 4C,D). In contrast to the controls (Fig. 4E), the eyes are separated from the cranial cartilage by a distinct cleft (Fig. 4G). The palatoquadrate cartilage is strongly reduced. The branchial arches of FoxN3-MO-injected embryos appear to be normally positioned.

Figure 4.

Craniofacial cartilage defects of FoxN3-depleted embryos. Embryos were injected with 13 ng FoxN3-MO in each blastomere at 2-cell stage. A: Dorsal view on an Alcian blue–stained wild-type embryo at stage 48. B: The corresponding schematic drawing of the craniofacial cartilage. C: Dorsal view on a FoxN3-MO injected and Alcian blue–stained embryo at stage 48. D: The corresponding schematic drawing of the craniofacial cartilage. Frontal view (E, G) and sections (F, H) of Alcian blue–stained Co-MO (E, F) and FoxN3-MO-injected embryos (G, H). FoxN3-MO-injected embryos reveal an invagination of the olfactory region, characterised by a large cleft. Many of the cranial cartilage elements are malformed or absent. The white lines in E and G denote plane of sections shown in F and H. ac, auditory cap; apc, anterior parachordal cartilage; ba, branchial arches; bb, basobranchiale; br, brain; ch, ceratohyale; fo, foramen olfactorium; ic, infrarostal cartilage; it, intestine; mc, Meckel's cartilage; mpp, muscular process of palatoquadrate; nc, notochord; pcqm, processus cornu quadratus medialis; pq, palatoquadrate; sbp, subocular bars of the palatoquadrate; sfen, subocular fenestra; sp, suprarostral plate; ta, tectum anterius; tr, trabecula; tenc, tentacular cartilage; vpp, ventrolateral process of palatoquadrate.

To analyse the observed defects in more detail, tissue sections of embryos at stage 48 were prepared (Fig. 4F,H). After staining with Alcian blue, almost all examined embryos exhibited an invagination at the anterior head localised in the olfactory region. The embryos suffer from a complete loss of nose structures (Fig. 4H). The branchial arches are less affected and false positioning of these structures was never observed. This result indicates that FoxN3 is predominantly necessary for cartilage structures derived from the first and second pharyngeal arches.

Analysis of the Development of Cranial Nerves in FoxN3-Depleted Embryos

The head structures, which show malformations in FoxN3-depleted embryos, are innervated by cranial nerves originating from the brain. Therefore, we have analysed the arrangement of nerves in Xenopus laevis embryos by staining with the monoclonal neurofilament antibody 3A10. Embryos injected with Co-MO exhibited a normal arrangement of cranial nerves (Fig. 5B,D). Injection of FoxN3-MO embryos leads to a loss of the hypoglossal nerve (cranial nerve XII), which innervates in mammals the intrinsic and extrinsic muscles of the tongue (Fig. 5A). In addition, branches of the trigeminal nerve (cranial nerve V), notably the frontal branch of the ophthalmic nerve, are completely absent. Also, the maxillary, the mandibulary, and the nasociliary nerves innervating the lower jaw, the maxillary antrum, the nasal cavity, and the eye, are partially reduced or missing (Fig. 5C). Other cranial nerves are less or not affected (e.g., nervus facialis, Fig. 5C,D).

Figure 5.

Analysis of cranial nerve structures in FoxN3-depleted embryos. Thirteen nanograms FoxN3-MO was injected into each blastomere of 2-cell stage embryos. Developing nerves were visualised by 3A10 monoclonal antibody staining in whole embryos at stage 48. A: Ventral view on a bilaterally injected embryo shows a complete absence of hypoglossal nerves at both sides of the body. B: Ventral view on wild-type embryo. C: Dorsal view on a FoxN3-depleted embryo demonstrates a diffuse arrangement of cranial nerves. The nervus nasociliaris, a branch of the trigeminal nerve, and the maxillary nerve are strongly reduced, while the facial nerve shows normal appearance. The nervus frontalis is completely absent. D: Dorsal view on wild-type embryo. fa, nervus facialis; fo, nervus frontalis; hg, nervus hypoglossus; max, nervus maxillare; md, nervus mandibulare; nc, nervus nasociliaris. Roman numbers indicate the cranial nerves II (optic), III (oculomotorius), and V (trigeminus).

FoxN3 Depletion Does Not Affect Early Neural Crest Markers But Interferes With Late Mandibular Arch Markers

To determine the molecular basis of the defects observed in cranial structures, we investigated whether loss of FoxN3 function would affect the expression of neural crest–specific genes and other genes involved in branchial arch development. To investigate early neural crest development, in situ hybridization with the transcription factors FoxD3 (XFD6), AP2α, snail, and slug were performed. FoxD3 is known to play a role in neural crest cell migration (Pohl and Knöchel, 2001) and the knockdown of FoxD3 in zebrafish revealed a loss of cartilage elements (Lister et al., 2006). Snail and slug are also involved in the specification and migration of neural crest cells. AP2α is required for neural crest induction (Luo et al., 2003) and mice lacking the AP2α gene also show craniofacial defects (Schorle et al., 1996).

FoxN3-MO was injected together with GFP RNA into one blastomere at the 2-cell stage. Embryos were fixed between stages 23 and 32, when the neural crest markers are expressed. Comparisons of the injected with the uninjected side show no alternation in expression of the analysed neural crest markers (Fig. 6A–E), indicating that the observed loss of function phenotype is not due to early determination events of cranial crest cells (Aybar et al., 2003). To examine the fate of crest cells at later stages, we used Sox3 and several Hox genes as markers of the pharyngeal arches. Since HoxB1 and HoxD3 expression is restricted to posterior pharyngeal arches, it is not impaired by FoxN3-MO injection (data not shown). Surprisingly, HoxA2, which is known to induce a homeotic transformation of the second branchial arch cartilage from mandibular to hyoid fate in Xenopus and mouse (Rijli et al., 1993; Pasqualetti et al., 2000; Baltzinger et al., 2005), is also not affected (Fig. 6F). No alteration of Sox3 expression was observed (Fig. 6G). Therefore, we decided to analyse early jaw formation by using jaw markers. Xbap, an Nkx transcription factor homologue of the Drosophila gene bagpipe, is expressed in the precursor of the palatoquadrate and the lateral part of Meckel's cartilage, whereas the paralogous gene zampogna (zax) is expressed in the region corresponding to the developing infrarostral cartilage (Newman and Krieg, 1999). Goosecoid expression is similar to that of Xbap, but is more restricted towards the midline (Baltzinger et al., 2005). In contrast to the Hox gene expression, we found that Xbap expression is significantly altered in the anterior head region. Similarly, zax and goosecoid expression domains are reduced (data not shown). Beginning with stage 38, FoxN3-MO-injected embryos show a strongly reduced expression of Xbap compared to Co-MO-injected embryos (Fig. 6H,I). The ventral region of the Xbap expression domain is completely absent (see Fig. 6H). At stage 43, Xbap expression is visible in the forming Meckel′s cartilage. However, in FoxN3-depleted embryos, this expression is strongly reduced and restricted to more lateral positions (Fig. 6J,K).

Figure 6.

Effects of FoxN3 depletion on neural crest formation. Embryos were unilaterally injected with 13 ng FoxN3-MO and 500 pg GFP RNA. After sorting for right/left injections, embryos were submitted at different developmental stages to in situ whole mount hybridization. The asterisks indicate the side of injection. A:FoxD3, stage 23. B:AP2α, stage 23. C:snail, stage 23. D:AP2α, stage 29. E:snail, stage 29. No difference in the expression domains between injected and uninjected sides can be observed. F,G: Whole mount in situ hybridization of unilaterally injected embryos stained for HoxA2 (F) and Sox3 (G). Embryos were injected at their right sides. H–K: Whole mount in situ hybridization of bilaterally FoxN3-MO or Co-MO-injected embryos (13 ng pro blastomere). H:Xbap, Co-MO-injected, stage 38. I:Xbap, FoxN3-MO injected, stage 38. Xbap expression is visible close to the cement gland in the controls (see arrow). The FoxN3-MO-injected embryos show reduced expression of Xbap in the prospective jaw (see arrow). The ventral part of this expression domain is completely absent. J:Xbap, Co-MO-injected, stage 43, ventral view. K:Xbap, FoxN3-MO-injected, stage 43, ventral view. The Xbap expression in the controls is restricted at stage 43 within the head to Meckel's cartilage. FoxN3-MO-injected embryos show severe reduction of staining (see arrows).

Effects of FoxN3-MO on Neural Crest Cell Migration and Differentiation

Results from in situ hybridization experiments have shown that depletion of FoxN3 does not influence the expression of neural crest and pharyngeal arch markers. We next elucidated if the abnormalities observed in jaw formation result from a migration failure of precursor cells of Meckel's cartilage. Therefore, lineage-labeling experiments using the recently described green to red photoconvertible fluorescent protein EosFP (Wiedenmann et al., 2004) were performed. EosFP mRNA was injected in combination with FoxN3-MO or Co-MO. Cells from the mandibular arch were photoconverted at a stage when marker analysis did not indicate any abnormal development (stage 26, Fig. 7A–C). Formation of Meckel's cartilage was then analysed at stage 46. In the injected embryos, the labelled cells were found in a position corresponding to Meckel's cartilage and adjacent epidermis (Fig. 7D–F). In contrast to Co-MO-injected embryos, which formed Meckel's cartilage from the labelled cells (Fig. 7E), FoxN3-MO-injected embryos did not form this element of the lower jaw. In summary, these experiments indicate that FoxN3 is not required for early mandibular crest migration. Rather, the effect results from a failure of cartilage formation.

Figure 7.

Analysis of cranial crest cell migration during embryogenesis. Embryos were injected with 200 pg EosFP RNA and 13 ng FoxN3-MO into each blastomere at 2-cell stage. At stage 26, cells of the mandibular arch were photoconverted. A: Lateral view of FoxN3-MO-injected embryo, stage 26. B: Lateral view of Co-MO-injected embryo, stage 26. C: Schematic representation of cranial crest migration at stage 26 (lateral view, converted area is shown as dashed circle). D: Embryo shown in A at stage 46, ventral view. The FoxN3-MO injected embryos show an unshaped structure of Meckel's cartilage. E: Embryo shown in B at stage 46, ventral view. The Co-MO-injected embryos revealed a normal structure of Meckel's cartilage. The dotted red cells in D and the red cells at the anterior front of the embryo in E represent converted epidermal cells. F: Cartilage structures (ventral view) at stage 48 (modified from: Sadaghiani and Thiebaud, 1987; Baltzinger et al., 2005). The rectangle in F denotes the area seen in D and E. 1, mandibular arch (blue); 2, hyoid arch (yellow); 3, anterior branchial arch (green); 4, posterior branchial arch (brown). ba, branchial arch; ch, ceratohyale; l, lens; mc, Meckel's cartilage; pq, palatoquadrate.

FoxN3 Depletion Affects Apoptosis But Not Cell Proliferation

The human FOXN3 gene is known to be a checkpoint suppressor in the cell cycle (Pati et al., 1997). Therefore, we have investigated whether the observed defects in FoxN3-depleted embryos are due to cell cycle defects. Unilaterally injected embryos between stages 24 to 45 were analysed for proliferation by BrdU and for apoptosis by TUNEL assays. From stage 36/37 onwards, the deeper cell layers of whole embryos were found to be less accessible for these stainings. Therefore, we analysed cell proliferation and apoptosis on tissue sections.

At stage 24, almost no apoptotic cells are found in the FoxN3-MO-injected and uninjected sides of embryos (Fig. 8A,A') as well as in Co-MO-injected embryos (data not shown). At stage 32, the difference in the intensity of TUNEL staining between the uninjected and injected side is clearly visible (Fig. 8B,B'). While only a few cells are stained in the eye and the otic vesicle on the control side (Fig. 8B), the number of apoptotic cells significantly increased in these tissues on the FoxN3-MO-injected side (Fig. 8B').

Figure 8.

Analysis of apoptosis and cell proliferation in FoxN3-depleted embryos. Embryos were co-injected with 13 ng FoxN3-MO and 200 pg GFP RNA into one blastomere at 2-cell stage. An asterisk denotes the injected side of embryos. A, A': Left and right view on an unilaterally FoxN3-MO-injected and TUNEL-stained embryo at stage 24. A' shows the injected side. Almost no apoptotic cells can be observed on either side. B, B': Left and right view on the head of an unilaterally FoxN3-MO-injected and TUNEL-stained embryo stage 32. The injected side (B') shows a higher rate of apoptosis than the uninjected side (B). C: Section of an unilaterally FoxN3-MO-injected embryo at stage 41. An increased number of apoptotic cells is found on the injected side, especially within the eye. D, E: Tissue sections of the eye of a unilaterally injected embryo at stage 41. The eye of the injected side (E) shows an increase of apoptosis in the retina as compared to the uninjected side (D). F–H: Tissue sections of unilaterally injected embryos analysed by BrdU immunohistochemistry. F: Section of the head of an embryo at stage 39/40. No difference in the proliferation rate of the injected side in comparison to the uninjected side can be observed. G, H: Sections of BrdU stained eyes of an unilaterally injected embryo at stage 45. The labeled cells of the retina indicate the ciliary marginal zone (arrowheads), but are equally stained in the uninjected (G) and in the FoxN3-MO-injected side (H).

Cryo-sections of embryos at stage 41 show an increased level of apoptosis in the developing retina at the FoxN3-MO-injected side as compared to the uninjected side (Fig. 8C). Eye sections of the uninjected side (Fig. 8D) and FoxN3-MO-injected side (Fig. 8E) reveal an increase of apoptosis in the retina at the FoxN3-MO-injected side. Histological analysis of unilaterally FoxN3-MO-injected embryos at stage 41 (14 tissue sections of 4 embryos) and at stage 43 (10 tissue sections of 4 embryos) revealed a 64 and 65% increase of TUNEL-stained cells at the injected side. Unilaterally, Co-MO-injected embryos at stage 41 and 43 do not show an increase of apoptosis at the injected side (10 tissue sections of 4 embryos for each stage; data not shown). These experiments clearly demonstrate that FoxN3 depletion increases the number of apoptotic cells in the eye.

To investigate the effect of loss of FoxN3 function on cell proliferation, we performed BrdU assays on tissue sections of unilaterally FoxN3-MO-injected embryos between stages 36 to 45. Interestingly, no increase or decrease of cell proliferation can be observed at all analysed stages as shown for stage 39/40 (Fig. 8F). The late phenotypic eye effect could also arise from degeneration of retinal stem cells. We, therefore, analysed whether FoxN3 depletion affects the proliferation in the ciliary marginal zone (CMZ), a known retinal stem cell pool in amphibians. Tissue sections of stage-45 embryos revealed identical rates of proliferation in both the CMZ of the uninjected and the injected side (Fig. 8G,H). Thus, FoxN3 depletion did not have any visible effect on cell proliferation in the eye.

FoxN3 Interacts With Components of HDAC

We were further interested in the mechanisms by which FoxN3 controls gene regulation. It has already been reported that human FOXN3 (Ches1) interacts with yeast Sin3, a component of the histone deacetylase complex (Scott and Plon, 2003). To test whether the Xenopus Sin3 also interacts with the two Xenopus FoxN3 isoforms, FoxN3a and FoxN3b, we performed pull-down analyses with GST-tagged FoxN3a/b and, for comparison, with GST-tagged FoxN2. Sin3 binds FoxN2 and both isoforms of FoxN3 (Fig. 9B). Moreover, we prepared deletion constructs of FoxN3a to delineate the binding motif of a putative Sin3 interaction domain (Fig. 9A). While the C-terminal truncated deletion constructs show no binding, the binding domain could be localised within the C-terminal 43 amino acids of FoxN3 (Fig. 9C). The deletion mutant FoxN3ΔEAA lacks an EAA motif that is conserved between FoxN3 and FoxN2 paralogues. Furthermore, these amino acids are part of the Sin3 binding motif (Fig. 9E), which has previously been proposed for proteins of the Mad family (van Ingen et al., 2003).

Figure 9.

FoxN3 interacts with components of the histone deacetylase complex (HDAC). A: Scheme of FoxN3 deletion constructs used for GST-pull down assays. The fork head/winged helix domain is indicated by a black box. B: [35S] methionine-labeled Sin3 protein was incubated with GST-fusions of the two isoforms of FoxN3 or with GST-FoxN2, respectively. The pull-down reactions were analysed using a 10% SDS-PAGE. C: Pull-down analysis of GST-FoxN3 deletion mutants with [35S] methionine-labeled Sin3 protein. Sin3 requires for binding the C-terminus of FoxN3 but not the N-terminus. D: [35S] methionine-labeled RPD3 protein was incubated with GST-FoxN3 mutants. RPD3 binds to N-terminal, but not to C-terminal region of FoxN3 deletion constructs. E: Alignment of putative Sin3 binding sites of FoxN2, FoxN3, and human Mad1 (see: van Ingen et al., 2004). Conserved amino acids are underlined and shown in grey.

Histone deacetylase complexes (HDAC) are involved in repressing genes by chromatin remodelling. They contain several components, which cannot bind directly to DNA, but bind via specific transcription factors. In human, there is a multitude of HDACs that are subdivided into three subclasses by their deacetylase component. The HDAC I subclass contains the highly conserved histone deacetylase RPD3, which is present from yeast to human, where it is found in HDAC 1, 2, 3, 8, and 11 (Grozinger et al., 1999). We have investigated whether FoxN3 can bind to Xenopus RPD3 and might thereby target the HDAC to DNA. We found that both isoforms of FoxN3 can bind to RPD3. Moreover, this interaction requires the first 135 amino acids within the N-terminal part of FoxN3. The C-terminal part of the protein does not bind to RPD3 (Fig. 9D). These results suggest that FoxN3 may probably act as a repressor by recruiting HDAC I and the co-repressor Sin3. Thus it is concluded that a functional HDAC requires both the C-terminal and the N-terminal region of FoxN3.

DISCUSSION

FoxN3 Is Required for the Development of Craniofacial Structures

Although the function of human FOXN3 as checkpoint suppressor in cell cycle regulation and its downregulation in oral carcinoma is reported (Pati et al., 1997; Scott and Plon, 2003; Chang et al., 2005), little is known about the function of FoxN3 during development. We have, therefore, investigated the role of FoxN3 during embryogenesis of Xenopus laevis by loss of function experiments using an antisense morpholino oligonucleotide derived from the initial 25 nucleotides of the FoxN3 coding sequence. We demonstrate that this FoxN3-MO inhibits the translation of a FoxN3/GFP construct in vivo.

Embryos injected with FoxN3-MO do not exhibit any visible abnormalities until stage 38. From stage 39 onwards, a reduction of the size of the eye and, notably, from stage 41 onwards, malformations of the developing jaw, can be observed. These effects are dose dependent. In addition, we frequently observed oedema in the head and trunk region. The small eye phenotype is not due to a reduced size of the early eye field but arises rather late between stages 35 and 39. While all normal morphological structures of the retina layers are present, the number of cells is dramatically reduced. For a more detailed characterisation of the FoxN3-MO cartilage phenotype, we have performed histological analyses. Alcian blue staining of whole embryos and tissue sections reveals an invagination in the mouth region and the complete absence of nose structure. The craniofacial cartilage showed an abnormal structure with increasing malformations from posterior to anterior. While Meckel's cartilage, palatoquadrate, and ceratohyale derived from the most anterior mandibular and hyoid arches are severely reduced and deformed, the branchial arches derived from the posterior pharyngeal arches are less affected. Moreover, depletion of FoxN3 also affects cranial nerve development. The hypoglossal nerve is lacking. A reduction or malformation of several branches of the trigeminal nerve can be observed. We do not know if this is a direct effect of FoxN3 knockdown on cells forming these nerves. It is more likely that these cranial nerve defects are indirect effects, mainly due to the malformations of innervated structures. This is supported by the fact that the affected branches are normally innervating those structures, which are malformed in FoxN3-MO-injected embryos, namely the nervus hypoglossus (lower jaw), the nervus mandibularis (lower jaw), the nervus maxilliaris (upper jaw), and branches of the nervus ophtalmicus (innervating the lens and eye). Other cranial nerves, innervating structures that are less or not affected, form in an almost normal way (e.g., nervus facialis innervating the musculus interhyoideus posterior in anurans) (Gradwell, 1972).

FoxN3 Is Not Involved in Cranial Crest Cell Migration

Most parts of the vertebrate jaw derive from cranial crest cells, migrating lateral from the neural tube into the first two pharyngeal arches. This migration requires several processes, like cell determination, delamination, apoptosis, and cell rearrangements that are governed by a multitude of genes. Although FoxN3 is expressed in the neural crest during early neurula stages (Schuff et al., 2006), we demonstrate here by whole mount in situ hybridization that the spatial expression patterns of the neural crest markers AP2α, FoxD3, slug, and snail are not affected by a functional FoxN3 knockdown. This result correlates with the observation that FoxN3-MO-injected embryos develop normally up to stage 39. Therefore, we suggest that FoxN3 does not influence early neural crest cell migration. We have further analysed crest cell migration into the mandibular arch by a photoconvertible fluorescent protein. In FoxN3-MO-injected embryos, we observed that the precursor cells of Meckel's cartilage migrated to correct position, but the cartilage does not differentiate at the correct extent. We conclude that this deficiency is not due to a positional migration defect of cranial crest cells, but rather to a failure in differentiation probably at the level of chondroblasts. Moreover, the expressions of the Hox genes A2, A3, B2, and D3, which pattern the pharyngeal arches, are not changed by FoxN3-MO injections. This observation gains special interest in the light of previous reports that Hox genes are involved in skull morphogenesis. An alteration of the most anterior Hox genes causes severe defects in the facial skeleton, even the transformation in jaw fate (Alexandre et al., 1996; Pasqualetti et al., 2000; Creuzet et al., 2002; Baltzinger et al., 2005; Kuratani, 2004, 2005). However, the expression of Nkx3 genes in the presumptive jaw is significantly reduced.

These results indicate that FoxN3 knockdown does not initialise homeotic cell transformation as it is reported for HoxA2 in mouse and Xenopus (Alexandre et al., 1996; Pasqualetti et al., 2000; Baltzinger et al., 2005). The malformation of anterior jaw elements in FoxN3-depleted embryos might be due to disordered differentiation events.

FoxN3 Is Involved in Cell Survival

It has been shown that many fork head transcription factors are involved in cell cycle regulation and cell survival (Wijchers et al., 2006). We, therefore, have investigated the role of FoxN3 in the cell cycle by BrdU assays and TUNEL staining during different stages of embryonic development. The results did not reveal any difference in the rate of cell proliferation for FoxN3-MO-injected embryos. However, we observed an increase of apoptotic cells in the eye, when FoxN3 is depleted. This effect becomes stronger at later stages of development. While at stage 24, no increase of cell death by FoxN3-MO injection can be observed, the number of apoptotic cells is significantly increased from stage 32 onwards. TUNEL analysis of tissue sections reveals an increase of apoptosis in the eye. This renders an explanation for the reduced eye size. In addition, since FoxN3 is expressed in the lens (Schuff et al., 2006), it is reasonable to assume that FoxN3 depletion in the lens leads to a lack of signals being required for prevention of apoptosis in the retina. Thus, the reduced size of the retina might also result from the reduced size of the lens (Yamamoto and Jeffery, 2000). The discrepancy between the appearance of apoptosis at stage 32 and the first phenotypic defects observed at stage 39 could be explained by an accumulation of the apoptotic effect.

Whether an increased rate of apoptosis or a reduction of proliferation is involved in the abnormal jaw development in FoxN3-depleted embryos remains to be elucidated. We neither succeeded in demonstrating a change in the rate of apoptosis nor in the rate of proliferation for the craniofacial cartilage by functional knockdown of FoxN3. However, it should be mentioned that the jaw malformation might originate much earlier in development. We cannot exclude that cell death or reduced proliferation of distinct progenitor cells, which are not detectable by marker gene analysis, may be responsible for this phenotype.

FoxN3 May Act as a Repressor by Recruiting Histone Deacetylase Complexes

Downregulation of FOXN3 in oral squamous cell carcinoma (OSCC) (Chang et al., 2005) and its function as a checkpoint suppressor (Pati et al., 1997) suggests that this factor might play a role in gene regulation and chromatin remodelling. Indeed, FOXN3 interactions with yeast Sin3 and human Ski-interacting protein (SKIP) have been demonstrated (Scott and Plon, 2003, 2005). Both proteins are known to repress transcriptional activity (Jones et al., 1998; Vermaak et al., 1999; Xu et al., 2000; Prathapam et al., 2001). Interestingly, Sin3 as well as SKIP have been proposed as co-repressors of HDAC (Laduron et al., 2004). Additionally, SKIP has been shown to interact with signal transducers, like Smads or Notch (Xu et al., 2000). It is known that HDAC itself cannot directly bind to target genes but only by recruitment of transcription factors. We, therefore, investigated whether Xenopus FoxN3 might also interact with components of HDAC, like Sin3 or RPD3, the histone deacetylase in HDAC I and II. We could demonstrate by GST-pulldown that FoxN3 can bind to Sin3 and to RPD3, respectively. Additionally, by use of deletion mutants we have delineated the sequences of FoxN3, which are required for these bindings. We found that the interaction between xSin3 and FoxN3 is restricted to the C-terminal region of FoxN3 protein. This region maps to a previously localized Sin3 interaction domain of the Mad family and other transcription factors containing a minimum consensus motif EAAxxL (van Ingen et al., 2003). The interaction with xRPD3 could be localized to the N-terminal region of FoxN3 protein. Interestingly, it has been shown that mutation of the RPD3 gene as well as knockdown studies in zebrafish revealed specific defects in eye, craniofacial cartilage, and pectoral fin development (Golling et al., 2002; Pillai et al., 2004; Yamaguchi et al., 2005). We, therefore, suggest that histone deacetylases might regulate development via a subset of transcription factors mediating transcriptional repression of target genes. Interestingly, it has been shown that HDAC inhibitors increase the rate of apoptosis (Sonnemann et al., 2006; Peart et al., 2003). This renders an explanation for how the knockout of FoxN3 leads to programmed cell death. Assuming that FoxN3 recruits HDAC to bind to DNA, the loss of FoxN3 function might prevent DNA binding and, therefore, has an inhibitory effect on HDAC. Analysis of interactions between FoxN3 and other members of HDAC and the identification and characterisation of FoxN3 target genes will be the next step to understand the role of FoxN3 in eye and jaw development.

EXPERIMENTAL PROCEDURES

Antisense Morpholino Oligonucleotide

A FoxN3 antisense morpholino oligonucleotide (FoxN3-MO) derived from 25 nucleotides beginning at the translation start site had the sequence 5′-TACTAGGAGGCATGACTGGACCC- AT-3′ (Gene Tools). It matches to all splice variants of FoxN3 (Schuff et al. 2006). Doses of 1–20 ng antisense oligonucleotide were injected into single blastomeres of 2- or 4-cell stage embryos. A control morpholino oligonucleotide (Co-MO) 5′-CCTCTTACCTCAGTTACAATTTATA-3′ directed against the sequence of the human β-globin gene (Gene Tools) was injected under identical conditions.

Whole Mount In Situ Hybridization and Immunohistochemistry

Whole mount in situ hybridizations were done according to standard procedures (Harland, 1991). Digoxigenin-labeled antisense probes were synthesized using AP2α, FoxD3, snail, HoxA2, HoxA3, HoxB1, HoxD3, Sox3 and Xbap cDNAs. For staining of cranial nerves, we used the monoclonal antibody 3A10 (Developmental Studies Hybridoma Bank, University of Iowa), which recognizes a neurofilament-associated protein.

Embryo Manipulation

In vitro fertilization and embryo culture were done according to standard protocols. Embryonic stages were determined according to Nieuwkoop and Faber (1967).

In Vivo Assay for the Morpholino Function

The N-terminus coding region of FoxN3 that contained the putative FoxN3-MO binding site was fused to GFP (FoxN3/GFP) (Carl et al., 2002). For control, the N-terminus of FoxN3 containing six silent mutations was fused to GFP (FoxN3Res/GFP). RNA was transcribed and injected either alone or with different amounts FoxN3-MO or Co-MO, respectively. Embryos were analysed by fluorescence microscopy.

Cartilage and Tissue Staining of Embryos and Tissue Sections

Embryos at stage 48 were fixed in MEMFA (0.1 M MOPS, pH 7.4, 2 mM EGTA, 1 mM MgSO4, 4% formaldehyde), transferred into 1% Alcian blue (Sigma) solution containing 0.5% acetic acid, and incubated for 95 min, followed by incubation in ethanol/glacial acetic acid (80/20) overnight. The embryos were washed for 3 hr in a 1% KOH, 3% H2O2 solution and incubated in saturated sodium tetraborate solution containing 0.05% trypsine for 2 hr. Finally, embryos were washed in PBT (PBS (137 mM NaCl, 2.7 mM KCl, 8.5 mM Na2HPO4 × 2 H2O, 1.5 mM KH2PO4) + 0.1% Tween), incubated with 6 μl/ml Proteinase K (20 mg/ml) in PBT for 1–3 hr, washed in PBT again, and refixed in MEMFA. Alternatively, embryos (stage 48) were fixed in MEMFA, embedded in OCT resin (Sakura Finetek), and cryo-sectioned (10 μm) according to standard protocols. After incubation in 1% acetic acid for 2 min, tissue slides were transferred into Alcian blue solution (1% Alcian blue, 0.15% acetic acid in H2O) for 10 min. The sections were washed in H2O and incubated for 14 min in Nuclear Fast Red (Merck). After washing in H2O, they were transferred to 96% ethanol (1 min), 100% ethanol (2 min), isopropanol, (2 min), and xylene (6 min). Slides were covered with Histo Kitt II (Roth).

BrdU and TUNEL Assay

Embryos (stages 36 to 40) were injected with 10 nl of BrdU labeling reagent (Roche) posterior to the eye and incubated for 2 hr according to standard protocols (Hardcastle and Papalopulu, 2000). Embryos at stage 45 were incubated in 250 μl 0.1 × MBSH containing 20 μl BrdU labeling reagent for 4 hr. Embryos were fixed in MEMFA, embedded in OCT resin, cryo-sectioned (14 μm) as described before, and analysed by using the 5-Bromo-2′-deoxy-uridine Labeling and Detection Kit II (Roche). BrdU assay on tissue slides was done according to the manual.

TUNEL assay on whole embryos was done according to standard procedures (Hensey and Gautier, 1998). For TUNEL assays on tissue sections, embryos at stage 36 to 45 were fixed in MEMFA, embedded in OCT resin, and cryo-sectioned. Tissue slides were washed in PBS containing 0.1% Tween (2 × 10 min), followed by an incubation in PBS containing 10 μg/ml Proteinase K for 4 min, washed in PBS containing 0.2% glycine for 7 min, followed by washing in PBS/Tween (2 × 5 min). Tissue sections were refixed in 4% paraformaldehyde/PBS for 1 hr and incubated in PBS/Tween (4 × 5 min). After washing in PBS/Tween, the sections were incubated in TdT-buffer (Invitrogen) for 2 hr, incubated overnight in TdT-buffer containing 0.1 μl digoxigenine dUTP (Roche) and 10 μl Terminal deoxynucleotide Transferase (TdT 150U; Invitrogen) per ml TdT-buffer, followed by washing with PBS containing 1 mM EDTA. Tissue slides were incubated 1 hr in PBS/EDTA at 65°C. Incubation with AP-antibody (1:2,000) (Roche), blocking, and staining with BM Purple (Roche) were done according to standard protocols. Tissue sections were refixed in MEMFA.

EosFP Analysis of Neural Crest Cell Migration and Differentiation

mRNA (200 pg) encoding the green to red photoconvertible fluorescent protein EosFP was injected together with Co-MO or FoxN3-MO into one or two blastomeres at the 2-cell stage. Embryos were kept in the dark to prevent unwanted photoconversion. At stage 26, cells from the mandibular arch were converted. Embryos were finally analysed at stage 46. All converting and documentation procedures are described (Wacker et al., 2006).

GST-Pulldown Assay

Fusion proteins were expressed in Escherichia coli BL2 (DE3) Plus (Stratagene, La Jolla, CA) and purified in batch under native conditions. GST-tagged FoxN2, FoxN3, and FoxN3 deletion mutants were purified with glutathione-Sepharose (Amersham Biosciences) according to the manufacturer's protocol. The purified proteins were dialysed overnight at 4°C (50 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 20% glycerol). 35S-labeled proteins were prepared using the TNT-coupled transcription-translation system (Promega, Madison, WI).

Acknowledgements

We thank P. L. Jones, M. Ori, H. Jansen, and M. Jamrich for gifts of plasmids.