Distinct requirements for wnt9a and irf6 in extension and integration mechanisms during zebrafish palate morphogenesis

Palate development provides an excellent example of conservation of genetic programs and morphogenesis events across vertebrate ontogeny. Five distinct facial prominences converge to shape the facial form. The median frontonasal prominence (FNP) contributes the columella and the central region of the primary palate, and coalesces with the paired maxillary prominences in a Y-shaped seam, where Bmp and a Pbx-dependent Wnt-p63-Irf6 regulatory module acts (Tamarin and Boyde, 1977; Jiang et al., 2006; Depew and Compagnucci, 2008; Ferretti et al., 2011). The secondary palate forms from midline fusion of bilateral palatal shelves, which are derived from the maxillary prominences (MXP), where midline fusion is characterized by distinct palatal shelf movements, apoptosis and epithelial-mesenchymal transition (Cuervo and Covarrubias, 2004; Helms et al., 2005; Bush and Jiang, 2012). Dysregulation of the highly choreographed molecular and morphogenetic events results in orofacial clefts in humans, which represent the most common structural birth defects (Mulliken, 2004; Rahimov et al., 2011).

Wnt signaling is crucial to craniofacial morphogenesis, where a delicate balance of co-receptor LRP6 and the transcriptional co-activator β-catenin mediate head formation at large, and refines specific facial features such as the lip (Brugmann et al., 2007; Fossat et al., 2011). In the mouse, Wnt9b is involved in fusion of the medial and lateral nasal processes (Juriloff et al., 2006; Lan et al., 2006; Ferretti et al., 2011). Variation of Wnt genes is associated with non-syndromic cleft lip and/or palate (CLP) in human and mouse studies (Juriloff et al., 2006; Chiquet et al., 2008). Knockdown of zebrafish wnt9a results in an abridged palate, where the bilateral trabeculae are joined in the midline but the ethmoid plate does not extend (Curtin et al., 2010). Conservation of genetic regulation and analogy of facial prominences between amniotes and zebrafish has been carefully described (Swartz et al., 2011).

The IRF6 transcription factor is a key regulator of palatogenesis, implicated in both syndromic and non-syndromic CLP (Kondo et al., 2002; Beaty et al., 2010). Irf6 regulates the transition of keratinocyte between proliferation and differentiation, and the craniofacial abnormalities observed in the Irf6 mouse mutants may be secondary to the abnormal differentiation of epithelial tissues, including epidermis and oral epithelium (Ingraham et al., 2006; Richardson et al., 2006; Richardson et al., 2009). Expression of irf6 in zebrafish has been reported in the pharyngeal arches and oral epithelium, and its embryonic requirement has been examined using dominant-negative mRNA overexpression that resulted in epiboly arrest, precluding analysis of its potential role in palatogenesis (Ben et al., 2005; Sabel et al., 2009). The function of Irf6 has been described for epithelial development in murine and zebrafish models, but whether and how this gene may directly regulate palatal fusion remains unclear (Sabel et al., 2009; Restivo et al., 2011; de la Garza et al., 2012).

This work presents detailed analysis of zebrafish palatogenesis, demonstrating that wnt9a and irf6 are required for distinct morphogenetic processes. wnt9a is required for coordinated extension, cell proliferation, morphological change and intercalation of the palatal chondrocytes. Meanwhile, irf6 is expressed in the bilateral MXP chondrocytes and required for integration with the median FNP. The morphogenetic context in which to understand a clinically relevant zebrafish orofacial cleft model is proposed, which could be applied towards future studies of palatogenesis in this highly tractable vertebrate system.

Zebrafish husbandry was performed as described previously (Westerfield, 1993). The Tg(sox10:kaede) line was generated using the 7.2 kb genomic region upstream of zebrafish sox10 (a gift from R. Kelsh, University of Bath, UK) and the kaede protein sequence using Gateway cloning (Invitrogen) (Kwan et al., 2007; Dutton et al., 2008).

Site-directed mutagenesis of irf6 was performed using the QuikChange II XL Kit (Agilent Technologies), which substitutes the critical arginine 84 residue: R84C, R84H and R84K. Each mutant transgenic, Tg(sox10:irf6^R84C,H,K), was generated with irf6 driven by the sox10 promoter, with germline integration mediate by Tol2 transposase.

Whole-mount RNA in situ hybridization was performed as described previously (Thisse and Thisse, 2008), with DIG- and DNP-labeled irf6 probe using Cyanine5 on sox10:GFP embryos, followed by incubation with anti-GFP Alexa Fluor 488 conjugate (Invitrogen).

Antisera against human IRF6 (anti-IRF6 peptide, 358-HQKGQIEKQPPFEIY-372 Genway Biotech) and against zebrafish Irf6 (i.e. anti-Irf6 peptides 137-TYETDGDEDDIPDTPE-152 and 213-ANVWPKKEPEDVEM-226, affinity purified; B. Schutte and R.C., unpublished) were used as previously reported (Fakhouri et al., 2012).

Proliferation was detected using the Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen). Apoptosis was detected using the Click-iT TUNEL Alexa Fluor 488 Imaging Assay. Detection was performed in the background of sox10:mcherry in analysis of wnt9a morphants. The Tg(sox10:irf6^R84C,H,K) and Tg(sox10:kaede) lines were crossed to label neural crest cells in the irf6 mutants.

Photoconversion was performed using the UV laser (404.3 nm) for 5-30 seconds, performed unilaterally so that the contralateral side serves as internal control. For time-lapse imaging, embryos were mounted in low-melt agarose, continuously imaged (2-4 minute intervals) over 24-36 hours.

Lineage tracing of migrating cranial neural crest cells (CNCCs) was carried out using the sox10:kaede transgenic (Dougherty et al., 2012). We confirmed that at 16 hours post-fertilization (hpf), the median and lateral ethmoid plate could be targeted by photoconversion of the CNCCs anterior and medial to the developing eye, respectively (Fig. 1A,B) (Wada et al., 2005; Eberhart et al., 2006). More specific CNCC labeling could be achieved by photoconverting at 19 hpf, where a small group of cells has already extended ventrally at the most anterior boundary of the embryo; these are the cells that will stream to form the median ethmoid plate (Fig. 1C,F,G; supplementary material Movies 1-3). This anterior group of CNCCs has been suggested to be analogous to the frontonasal process (FNP) (Swartz et al., 2011).

Time-lapse confocal analysis revealed that migration of the anterior-most CNCCs to the median ethmoid plate can be divided into three stages. First, CNCCs in discrete paired lateral anterior populations merge at the midline, while transitioning from a rostral to caudal direction of migration (supplementary material Movie 1). Second, the cells converge towards the midline as they extend into the space over the oral cavity and in between the developing paired maxillary process (MXP) positioned laterally (supplementary material Movie 2). Third, the frontonasal CNCCs make a sharp 180° turn at the anterior boundary of the embryo, while condensing to form a single median stream that then migrates caudally to meet the paired MXP that extend cephalad (supplementary material Movie 3).

When the entire length of the extending MXP was photoconverted, one can observe that the palate extension occurs via two distinct mechanisms: morphological change and proliferation. Extension of the lateral palate element results from morphological change of the chondrocytes into a columnar shape and intercalation in a cobblestone stacked pattern, leading to a gain in the length of the photoconverted area (bracket, Fig. 1E). Meanwhile, additional unlabeled green cells are observed to proliferate beyond the photoconverted red region, extending anteriorly to form the distal part of the ethmoid plate (asterisk bracket, Fig. 1E; supplementary material Movie 4). In fact, supplementary material Movie 4 shows both accumulation of cells in the proliferative front (anterior of the red photoconverted region) and stacking of columnar cells proximally. Palatal extension occurs by a similar mechanism to that previously described for the lower jaw, where mandible extension is mediated in part by morphological change and arrangement of chondrocytes into an intercalated pattern (Kimmel et al., 1998; Kimmel et al., 2001).

As the median ethmoid plate merges with the paired lateral MXP, a morphologically discrete transition can be delineated, which represents a morphologically defined border between juxtaposed columnar-shaped cells of the lateral ethmoid plate element and the cuboidal-shaped cells of the median ethmoid plate (broken line, Fig. 1H). Along this morphological transition, the red/green fluorescence appears to respect the delineation of median versus lateral ethmoid plate.

The amniote upper lip and primary palate forms from fusion of the MXP with the FNP derivatives of medial and lateral nasal processes. In the zebrafish, the ethmoid plate is also formed from convergence of three parts: the paired MXP (trabeculae) extending from lateral to medial and the FNP (median ethmoid plate) that streams along the midline in a rostral to caudal direction (supplementary material Movies 2, 3). We propose the V-shaped fusion seam between the median FNP and the bilateral MXP of the zebrafish palate is analogous to the Y-shaped junction found in the palate of amniotes, supported from molecular and morphogenetic evidence presented here and previously (Ferretti et al., 2011; Swartz et al., 2011). In this light, zebrafish palatogenesis is analogous to formation of the amniote lip/primary palate unit. With this model of zebrafish palatogenesis in mind, we explored whether mechanisms resulting in cleft malformation are also conserved.

In mouse models, cleft lip and palate can form as a result of defects in cell proliferation where the palatal shelves fail to extend, thereby precluding midline union, exemplified by targeted disruption of Tgfbr2 (Ito et al., 2003). The analogous defect in zebrafish palatogenesis would be failure in extension mechanisms of the MXP, which one may predict will result in a short ethmoid plate. In fact, when genes required for murine palatogenesis are targeted via morpholino knockdown in zebrafish, such as fgf10, tgfb2 and pax9, shortened ethmoid plate phenotypes were observed (Swartz et al., 2011). Similarly, when zebrafish wnt9a (ortholog of human WNT9B) was disrupted by morpholino knockdown, a short ethmoid plate phenotype was reported (Curtin et al., 2010). Knockdown of wnt9a in col2a:GFP transgenic embryos confirmed that the short ethmoid plate phenotype is a result of absent chondrocytes, not a consequence of immature chondrocytes that failed to deposit glycosaminoglycans and therefore were undetected by Alcian Blue staining (Fig. 2A,B). Likewise, Wnt regulation of the convergence extension process during zebrafish embryogenesis has been described, where disruption of Wnt/planar cell polarity in the zebrafish gpc4 mutant also resulted in foreshortened ethmoid plate (Verkade and Heath, 2008; LeClair et al., 2009).

Extension of the palate requires morphological change of the chondrocytes and cell proliferation, both of which were disrupted in the wnt9a morphant. In the wnt9a morphant, chondrocytes failed to adopt a columnar shape and do not stack in an intercalated pattern (Fig. 2). Lineage analysis of the wnt9a morphant confirmed that the short ethmoid plate phenotype did not result from failed migration of the frontonasal CNCC population, which in fact still migrated to the median ethmoid plate (supplementary material Fig. S1). However, proliferation of cells in the ethmoid plate was decreased in the wnt9a morphant (Fig. 2C, arrow). Lack of extension is not due to increased cell death, confirmed by TUNEL assay in the morphant (Fig. 2E,F). The proliferation defect can also be demonstrated by time-lapse confocal imaging in the Tg(sox10:kaede) embryo, where new green chondrocytes accrued distal to the red photoconverted region in the wild type (supplementary material Movie 4), but were absent in the wnt9a morphant (supplementary material Movie 5), as the photoconverted region failed to lengthen (Fig. 2G,H). Together, these results show that wnt9a is required for ethmoid plate extension, regulating mechanisms of morphological change and cell proliferation.

Human recessive or dominant-negative mutations in IRF6 result in cleft lip and palate, confirmed by similar mutations in the mouse (Kondo et al., 2002; Ingraham et al., 2006; Richardson et al., 2006). An analogous cleft defect in zebrafish may be predicted to manifest as a cleft in the seam where the facial prominences converge, as median ethmoid plate (FNP) and the lateral ethmoid plate (MXP) converge and integrate along an analogous V-shaped junction.

Embryos injected with dominant-negative truncation mutant irf6 mRNA resulted in epiboly arrest, precluding analysis of irf6 function in palatogenesis (Sabel et al., 2009; de la Garza et al., 2012). It has been reported that irf6 is expressed in pharyngeal arch epithelium and that sox10 is expressed in pharyngeal arch mesenchyme (Ben et al., 2005; Dutton et al., 2008; Sabel et al., 2009). We confirmed that irf6 is expressed in the pharyngeal arch epithelium at 55 hpf but also detected an overlap between irf6 and GFP in the pharyngeal arch mesenchyme of Tg(sox10:gfp) transgenic embryos at this stage (Fig. 3A-C). Moreover, at 72 hpf, anti-irf6 immunoreactivity was clearly demonstrated in ethmoid plate chondrocytes (Fig. 3D,E). Therefore, to circumvent the gastrulation lethality of dominant-negative irf6 mRNA injection, we expressed irf6 dominant-negative mutant transgenes under transcriptional control of the sox10 promoter, to ascertain their function during palatogenesis.

Dominant-negative irf6 mutants were generated based on known human mutations that abrogate DNA binding, where arginine 84 plays a crucial role (Escalante et al., 1998; Kondo et al., 2002; Item et al., 2005). Three mutant irf6 transgenes were generated: two transgenes harbor disruptive amino acid substitutions (R84C, R84H), and one transgene contained a conserved amino acid change (R84K) to serve as negative control. Stable transgenic lines of each arginine mutant were generated, where Tg(sox10:irf6^R84C,H,K) embryos assayed using irf6 riboprobe at 10-somites shows sox10 pattern of expression, a time when endogenous irf6 is normally absent, thereby confirming high level transgene expression (supplementary material Fig. S2). Expression of irf6 mutants under transcriptional control of sox10 did not perturb early neural crest patterning, with wild-type expression of dlx2a and sox9a (supplementary material Fig. S2).

The irf6 mutant embryos appeared grossly normal during early embryogenesis; however, the overall head shape of the R84C and R84H mutants become noticeably smaller than wild type past the pharyngeal stages (Fig. 3F-I). When stained with Alcian Blue, R84C and R84H mutants exhibited clefts between the median FNP and paired MXP elements, along the V-shaped seam by which they are joined (Fig. 3J-Q). In particular, clefts in the R84C mutant were more severe than the R84H mutant, whereas R84K appeared wild type, consistent with degree of non-conserved amino acid substitutions. Time-lapse visualization of the ethmoid plate demonstrated that although new chondrocytes appeared in the distal front of the paired MXP as the median FNP chondrocytes emerged, the median FNP and paired MXP failed to integrate and form a continuous ethmoid plate (supplementary material Movie 6).

The lower jaw was also dysmorphic in R84C and R84H mutants with shortened Meckel’s cartilage and smaller palatoquadrate structures, again with R84C>R84H in severity of phenotype (Fig. 3R-U). It is important to note that the dominant-negative irf6 mutants affected integration of the median FNP with the paired MXP specifically, whereas extension of the formed ethmoid plate was not affected; the chondrocytes were able to differentiate and adopt columnar morphology, and organized in the intercalated pattern.

The CNCC lineage analysis presented here illustrates conservation of the fundamental morphogenetic events between amniote primary palate and zebrafish ethmoid plate. Moreover, we describe distinct mechanisms of palatal extension at the cellular level: morphological change, intercalation, proliferation and integration (Fig. 4). Distinct requirements of wnt9a and irf6 in palatal extension and integration, respectively, reveal the complex molecular control of palatogenesis. These findings raise the issue of whether orofacial clefts in individuals with Van der Woude syndrome, where IRF6 is mutated, result from disruption of IRF6-mediated events within the palatal chondrocytes or within oral epithelium, or within both.

We are grateful to Brian Schutte for generating and sharing the anti-zebrafish-irf6 sera, to Greg Bonde for technical assistance, to Robert Kelsh for kindly sharing the zebrafish sox10 promoter genomic DNA, and to Renee Ethier for excellent management of our aquatics facility.