The Polycomb Group Protein Ring1b/Rnf2 Is Specifically Required for Craniofacial Development

Polycomb group (PcG) genes are chromatin modifiers that mediate epigenetic silencing of target genes. PcG-mediated epigenetic silencing is implicated in embryonic development, stem cell plasticity, cell fate maintenance, cellular differentiation and cancer. However, analysis of the roles of PcG proteins in maintaining differentiation programs during vertebrate embryogenesis has been hampered due to the early embryonic lethality of several PcG knock-outs in the mouse. Here, we show that zebrafish Ring1b/Rnf2, the single E3 ubiquitin ligase in the Polycomb Repressive Complex 1, critically regulates the developmental program of craniofacial cell lineages. Zebrafish ring1b mutants display a severe craniofacial phenotype, which includes an almost complete absence of all cranial cartilage, bone and musculature. We show that Cranial Neural Crest (CNC)-derived cartilage precursors migrate correctly into the pharyngeal arches, but fail to differentiate into chondrocytes. This phenotype is specific for cartilage precursors, since other neural crest-derived cell lineages, including glia, neurons and chromatophores, are formed normally in ring1b mutants. Our results therefore reveal a critical and specific role for Ring1b in promoting the differentiation of cranial neural crest cells into chondrocytes. The molecular mechanisms underlying the pathogenesis of craniofacial abnormalities, which are among the most common genetic birth defects in humans, remain poorly understood. The zebrafish ring1b mutant provides a molecular model for investigating these mechanisms and may lead to the discovery of new treatments or preventions of craniofacial abnormalities.


Introduction
Craniofacial abnormalities are amongst the most common genetic birth defects and result from defects in cranial neural crest patterning and morphogenesis. However, the molecular mechanisms regulating cranial neural crest differentiation are not completely understood. Most of the cartilage and bones of the skull in vertebrate embryos are formed by cranial neural crest cells (CNCs), a subpopulation of the neural crest (NC), a vertebratespecific, ectoderm-derived, multipotent cell population that is induced at the beginning of neurulation at the border between epidermal and neural ectoderm [1,2]. Fate-mapping studies in fish, amphibians, birds and mammals have shown that streams of NC cells from the hindbrain form most of the pharyngeal skeleton, while the more anterior NC cells form the brain-case or neurocranium [2,3]. The NC gives rise not only to cartilage and bone, but also to neurons, glia, smooth muscle cells and chromatophores. Distinct cell types are formed according to their position along the anteroposterior axis. For instance, in teleosts, NC cells that give rise to chromatophores originate from trunk NC, whereas NC cells that will form craniofacial cartilage are located at the level of the mid/hindbrain [4,5].
During cartilage development, the most anteriorly-located CNC-derived cartilage precursors will mainly form skeletal elements that encase the brain (neurocranium), whereas more posteriorly-located NC cells give rise to the ventral viscerocranium. Viscerocranial precursors migrate ventrally as three streams to populate the pharyngeal arches, embryonic structures that will form several tissues of the head-neck region in higher vertebrates. The mandibular and hyoid arches form the jaw and its supportive elements, respectively, whereas the five posterior (branchial) arches give rise to the cartilaginous elements of the gills.
Cranial bones are also derived from CNC cells and ossification occurs through two distinct processes: endochondral and dermal ossification. In endochondral ossification, a skeletal template comprising of replacement cartilage is formed first, which is then replaced by bone. In dermal ossification, bones form directly from osteogenic condensations. Both processes are essential for cranial bone development.
Skeletogenesis is a complex process guided by specific genetic and epigenetic programs. The Polycomb group (PcG) proteins are transcriptional repressors that mediate epigenetic silencing of target genes and thereby regulate differentiation programs. PcG proteins are key regulators of cell proliferation, cell-fate maintenance during embryonic development; and in adults, they are implicated in tissue homeostasis, cellular differentiation and cancer [6,7]. Mechanisms underlying PcG-mediated gene silencing include the organization of higher-order chromatin structure, post-translational modifications on nucleosomes and interference with the transcription machinery [7,8,9]. The ubiquitin E3-ligase activity of RING-domain-containing proteins in Polycomb Repressive Complex 1 (PRC1) leads to the mono-ubiquitilation of histone H2A at lysine 119 (H2AK119), a histone modification that is associated with gene repression [10,11].
The study of PcG proteins in maintaining differentiation programs during vertebrate embryonic development is hampered by the early lethality of mice defective for the core proteins of PRC2 and PRC1 (Ring1b). We have recently generated Ring1bdeficient zebrafish that survive early development (in contrast to mouse ring1b knock-outs [12]) and show surprisingly specific defects in tissue differentiation [13].
In this study we show that Ring1b-deficient zebrafish display a severe craniofacial phenotype, including an almost complete absence of all cranial cartilage, bone and musculature. We show that specification and initial migration of CNC-derived cartilage precursors into the pharyngeal arches is largely unaffected, however, cartilage differentiation is blocked. Moreover, the loss of Ring1b specifically affects CNC-derived cartilage precursors since other NC-derived lineages including neurons, glia and chromatophores are present in ring1b mutants. Finally, both endochondral and dermal ossification are disrupted due to loss of Ring1b function.
Our study highlights the importance of epigenetic control in cartilage differentiation and unravels a critical and specific role for PcG-mediated gene regulation in craniofacial development.

Ring1b Mutants Lack The Jaw Elements
The ring1b/rnf2 mutants were generated by zinc finger nuclease (ZFN)-mediated targeted gene inactivation [14] and were previously described [13]. Briefly, two indel alleles were generated, which resulted in an open reading frame-shift that led to a premature stop codon. No Ring1b protein was produced and no ubiquitination of H2A was detected in the ring1b mutants [13], indicating that they are functional nulls.
Visible craniofacial defects in homozygous ring1b mutants were first observed at around 40 hours post fertilization (hpf), and consisted of a reduction in the amount of tissue at the level of the anterior pharyngeal arches. In contrast to WT embryos at 72 hpf ( Fig. 1A), ring1b mutants showed a complete absence of tissue under the eye at the location of the jaw (Fig. 1B). Indeed, Alcian-Blue staining to visualize cartilage, revealed that almost all craniofacial and jaw cartilage elements were absent in ring1b mutants at 72 hpf. This dramatic phenotype led us to examine whether cartilage elements were formed at an earlier developmental stage and then deteriorated or whether any cartilage was formed at all. To address these issues, we stained WT and ring1b siblings at different developmental stages with Alcian blue.
In WT embryos at 56 hpf, the trabeculae cranii, a cartilage element of the prospective neurocranium, had fused posteriorly to the parachordals and the ethmoid plate starts to emerge anteriorly ( Fig. 1E). At 65 hpf, the ethmoid plate had elongated in WT embryos and the first viscerocranial cartilage structures were also formed ( Fig. 1G) as well as Meckel's cartilage and the palatoquadrate (derived from cartilage precursors from the mandibular arch [15]. The ceratohyal and hyosymplectic cartilage elements, derived from the hyoid arch [15] had also differentiated in WT embryos (Fig. 1G). At 72 hpf, craniofacial cartilage structures had differentiated further and the elaborate cartilaginous skeleton of the head had been established (Fig. 1I). In contrast to WT embryos, craniofacial cartilages were absent in ring1b mutants at both 56 and 65 hpf (Fig. 1F, H). At 72 hpf, the entire head cartilage skeleton was missing with the exception of two cartilage deposits located at both sides of the anterior notochord (Fig. 1J). The cartilaginous pectoral fin girdle was also missing. Together, these results show that loss of Ring1b severely disrupts head skeleton formation.

Cranial Muscle Development is Compromised in ring1b Mutants
Because both the viscerocranium and neurocranium are essential for the patterning of associated craniofacial musculature [16], we examined whether cranial muscle development was affected in ring1b mutants. To detect the developing muscles we used MF20, an antibody that recognizes the myosin heavy chain of vertebrate striated muscles [17].
The vertebrate craniofacial musculature is of paraxial mesoderm origin [4]. Muscle precursors residing in the first two pharyngeal arches give rise to muscles that are associated with the jaw and its support elements whereas precursors residing in arches III-VII form muscles associated with gill cartilage.
The anterior mandibularis that originates from the mandibular arch had formed by 56 hpf in WT embryos ( Fig. 2A). At 65 hpf, cranial musculature associated with the jaw started to emerge (Fig. 2C). The hyohyoideus and interhyoideus, derived from the hyoid arch, matured further at 72 hpf and various other muscles, including the mandibular arch-derived intermandibularis anterioris, intermandibularis posterioris, levator arcus palatine and dilator operculi as well as the hyoid arch-derived adductor hyomandibulae and adductor opercule, were readily visible in WT embryos (Fig. 2E). In contrast, most craniofacial muscles were missing in ring1b mutants. Although ring1b mutants lack pectoral fins, a few MF20-positive cells, presumably pectoral fin muscle progenitors, were detected in-between the sternohyoideus and posterior hypaxial muscle (Fig. 2D). Only some irregular MF20positive patches of craniofacial musculature were detected in 72 hpf ring1b mutants (Fig. 2F). We also analyzed development of somite-derived muscles and found that whereas initial formation and migration of those was unaffected, later elongation and differentiation was impaired in ring1b mutants (Fig. S1).
Our findings indicate that loss of Ring1b severely impairs the formation of craniofacial musculature, whereas initial formation and migration of somite-derived muscles is unaffected. This dramatic absence of cranial neural crest derived-tissues and head muscles in ring1b mutants prompted us to investigate whether all neural crest differentiation was affected.

Development of Trunk Neural Crest Derivatives is Largely Unaffected in ring1b Mutants
NC cells give rise to a wide array of cell lineages including cartilage, bone, neurons, glia, smooth muscle cells and pigment cells. Since the differentiation of CNC cells into cartilage was abolished in ring1b mutants, we addressed whether overall NC differentiation, is perturbed by loss of Ring1B.
To examine whether migration of trunk NC cells was affected, we stained embryos for sox10, a marker of NC lineages that is involved in the specification of non-ectomesenchymal NC derivatives, such as chromatophores, neurons and glia [18]. In WT embryos at 24 hpf, Sox10-positive NC cells were observed migrating ventrally from their dorsal pre-migratory position in a rostrocaudal fashion (Fig. 3A). At 24 hpf, we observed that ring1b trunk NC cells had migrated as far as in WT siblings and the localization of sox10-expressing NC cells remained indistinguishable from WT siblings at 32 hpf ( Fig. 3A-D). Thus, the migration of trunk NC cells and their specification is normal in ring1b mutants.
To address whether glial differentiation was hindered, temporal expression of the forkhead transcription factor foxD3 [19] was evaluated. In 32 hpf ring1b mutants, foxD3 was expressed at nearnormal levels in the cranial ganglia-associated glia of the developing trigeminal ganglion as well as in the preotic and postotic ganglia (Fig. 3E, F). At 56 hpf, foxD3 was present in all domains of expression in ring1b mutants, albeit slightly weaker when compared to WT siblings (Fig. 3H). These data show that migration and initial differentiation of cranial glial precursors was largely normal. Next, we analyzed the expression of neuroD, a marker for all neurogenic placodes [20]. At 32 hpf, neuroD was expressed in the trigeminal ganglia, octaval/statoacustic ganglia, medial lateral line ganglia and the posteriolateral line ganglia in WT embryos (Fig. 3I). In ring1b mutants, neuroD expression was present but weaker than in WT (Fig. 3J). Ring1b mutants at 56 hpf were expressing neuroD in the octaval/statoacustic ganglia (gVIII) as well as in a single domain in which the vagal nerve (gX) and the posteriolateral line ganglia (gP) are located (Fig. 3K, L), which together showed that cranial ganglia development is largely unaffected.
Trunk neural crest gives rise also to pigment cells and therefore we morphologically inspected chromatophore development. Three NC-derived chromatophores are produced in zebrafish: melanophores (black), iridophores (iridescent) and xanthophores (yellow). While melanophores are found in a wide variety of vertebrates, iridophores and xanthophores are not [21]. At 32 hpf, melanophore pigmentation in ring1b mutants was indistinguishable from WT siblings. (Fig. 3M, N) During later stages of development, melanophores in WT embryos matured further and changed their morphology from spindly to a more compacted shape (Fig. 3S). In contrast, ring1b melanophores remained spindly (Fig. 3T). It has been reported however, that melanophores in 72 hpf WT embryos from a different genetic background displayed a very similar morphology to those in 60 hpf ring1b mutants [22]. Iridophores terminally differentiate at around 42 hpf and initially populate the dorsal and ventral stripe [23]. At 60 hpf, iridophores were present in both WT and ring1b mutants, albeit the number of iridophores was slightly reduced in the mutants (Fig. 3Q, R, U, V). Yellow xanthophores also emerged at around 42 hpf [24]. Since individual xanthophores are difficult to distinguish, qualitative observations can be made by staining for methylene blue, which is specifically taken up by xanthophores [25]. Methylene blue staining showed that xanthophores were present in the tail of both WT and ring1b mutants (Fig. 3Q, R). In addition, while melanocytes in WT embryos formed compacted islands, they appeared more 'diffuse' in the ring1b mutants (Fig. 3T) although that could be due to genetic background differences. Xanthophores visualized with methylene blue were present but displayed a different shape in the ring1b mutants (Fig. 3V).
Together, these results allow us to conclude that migration and differentiation of non-ectomesenchymal NC derivatives is only modestly affected by the loss of Ring1b. Glia, neurons and the three chromatophore lineages were formed in ring1b mutants. Thus, the arrest in cartilage formation appears to be a specific developmental defect in ring1b CNC cells.

Ring1b is not Required for Early CNC Specification and Migration
Because craniofacial cartilage development was impaired in ring1b mutants, we investigated whether CNC cells that will form the prospective viscerocranium migrated correctly into the pharyngeal arches.
We analyzed the expression of two transcription factors that are required for pharyngeal arch development, dlx2a that is expressed in CNC cells and hand2, whose expression is confined to a ventral subset of dlx2a-positive cells in post-migratory CNC cells [26,27]. We observed that at 32 hpf, dlx2a-expressing CNC cells had migrated into the pharyngeal arches in both WT and ring1b embryos (Fig. 4A, B). In WT embryos, CNC cells had begun to separate into five cell groups that populate the branchial arches: three distinct cell groups were readily visible. The three branchial dlx2a-expressing cell groups were also detected in ring1b mutants but their separation appeared slightly affected. Hand2 expression in the pharyngeal, hyoid and the two anterior branchial arches of ring1b mutants was indistinguishable from WT embryos (Fig. 4C,  D). We also analyzed the expression of hoxa3a, a reported direct target of PcG proteins in mammals [28] that is expressed in the branchial arches (Fig. 4E, F) [29]. Hoxa3a expression was slightly reduced in ring1b mutants, but the overall size of the expression domain was unaffected.
Branchial pouches are outpockets of the foregut endoderm that play a role in directing CNC cell migration [15] and express cyp26a1. Expression of cyp26a1 was reduced in ring1b mutants (Fig. 4G, H), indicating that pharyngeal pouch development was impaired.
Together, these data show that Ring1b is dispensable for the specification and migration of CNC cells into the pharyngeal arches. In addition, pharyngeal pouch development is affected, as assessed by reduced cyp26a1 expression, which might contribute to the defect in cartilage development.

Ring1b is Required for Chondrocyte Differentiation and Ossification
We next investigated whether differentiation of post-migratory CNC cells into chondrocytes was impaired. We stained WT and ring1b embryos with markers revealing differentiating cartilage. The transcription factors sox9a/sox9b have a critical role in cartilage morphogenesis [30,31]. In 50 hpf WT embryos, sox9a  (2) and the posterior branchial arches (3-7) (Fig. 5A, A', B, B'). In addition, both genes were expressed in the developing trabeculae cranii. In ring1b mutants, sox9a was expressed in the pharyngeal arches (Fig. 5B, B'), albeit the domain of expression was smaller. Interestingly, ring1b mutants lacked trabecular staining. Sox9b was expressed similarly to sox9a in the pharyngeal arches of WT embryos, whereas the trabecular expression domain was more extended (Fig. 5C, C'). Sox9b was expressed at reduced levels in arches 1,2 in ring1b mutants, whereas expression in arches 3-7 and at the presumptive trabeculae was missing (Fig. 5D, D'). Notably, sox9a and sox9b expression in the hindbrain domains was unaffected in the mutants, suggesting a specific defect in cartilage differentiation (and not a general downregulation of sox9a/b gene expression).
Similar results were obtained when examining the pan-neural crest markers sox10 and dlx2a that remained expressed in postmigratory CNC cells. Both genes were expressed in the two anterior pharyngeal arches of ring1b mutants, whereas trabecular expression was completely absent (Fig. 5E-H, E'-H').
To further examine chondrogenesis in ring1b mutants, we analyzed the expression of alpha I chain of type II collagen (col2a1), the major collagen in cartilage and marker of differentiating chondrocytes in zebrafish [32]. Col2a1 expression is controlled by sox9a/b [31]. In 50 hpf WT embryos, col2a1 was strongly expressed in the trabeculae cranii, otic vesicle and to a lesser extent in the mandibular and hyoid arch (Fig. 5I, I'). Col2a1 was not expressed at the location of the presumptive trabeculae cranii in ring1b mutants (Fig. 5J, J'). Interestingly, col2a1 was not expressed in ring1b mandibular and hyoid arches, despite the observed sox9a/b expression in these structures.
Overall, these results demonstrate that Ring1b is required for proper execution of the chondrocyte differentiation program.
To address to which extent ossification is affected in ring1b mutants, we assayed for expression of the two runx2 genes that are present in zebrafish, runx2a and runx2b. The transcription factor runx2 is a key regulator of osteoblast differentiation and is also critically involved in the maturation step from immature chondrocytes to hypertrophic chondrocytes during the process of endochondral ossification [33,34,35]. In WT embryos, runx2a and runx2b were expressed in hypertrophic chondrocytes and dermal ossification centers. Runx2a was expressed in the cleithrum, dentary, maxilla, operculum, pharyngeal arches and parasphenoid whereas runx2b expression was detected in differentiating osteo- blasts of the branchiostegial ray, cleithrum, operculum, palatoquadrate, parasphenoid and pharyngeal arches (Fig. 6A, C).
Runx2a and runx2b expression was greatly reduced or absent in presumptive cartilaginous elements of the viscerocranium in 68 hpf ring1b mutants (Fig. 6B, D) indicating that endochondral ossification is abolished.
In addition, weak runx2a/b expression was detected only at the presumptive parasphenoid, suggesting that dermal ossification may also be disrupted. To confirm this observation, we stained for type X collagen (col10a1), an extracellular matrix component known to be directly regulated by runx2 [36]. Col10a1 expression is specific to developing dermal bones at 72 hpf. Indeed, strong staining was observed in the cleithrum, operculum and parasphenoid of WT embryos (Fig. 6E). Strikingly, col10a1 expression was completely absent in ring1b mutants (Fig. 6F). These results show that dermal ossification is also severely impaired in ring1b mutants.
The marked reduction in the expression of genes involved in chondrocyte differentiation prompted us to investigate whether the cartilage progenitors could have been eliminated by apoptosis. To this end, we stained WT and ring1b embryos for apoptotic cells using TUNEL and Acridine Orange staining. We stained embryos at 36 hpf, since at this developmental stage the ring1b CNC cells express markers at similar levels as the WTs. As shown in Fig. 7A, WT embryos displayed two small clusters of TUNEL-positive apoptotic cells in the pharyngeal arch area and although in ring1b embryos the same clusters appeared to contain more apoptotic cells (Fig. 7B), the induction of apoptosis observed in this particular region was not considered significant. Nonetheless, ring1b embryos exhibited a robust overall increase in apoptosis when compared to WT siblings, particularly in the trunk and tail (Fig. 7C, D). Acridine orange staining was conducted in WT and ring1b embryos at 48 hpf when deregulations of gene expression were apparent in the mutants. Consistently with the enhanced cell death detected in ring1b embryos, we observed that a small number of apoptotic cells persisted at 48 hpf in the prospective jaw region, whereas no apoptotic cells could be detected in the WTs at this stage (Fig. S2A, B). Overall, these results indicate that the slight upregulation of apoptosis occurring in ring1b mutants may contribute to the elimination of few chondrocyte progenitors, however, it may not be sufficient to account for the almost complete lack of cartilage observed in ring1b mutants.
We show here that zebrafish deficient in the Polycomb gene ring1b/rnf2 display severe craniofacial defects and lack cartilage elements of the neurocranium, pharyngeal arches, and pectoral girdle.
Craniofacial birth abnormalities result from defects in cranial neural crest patterning and morphogenesis. A better understanding of the molecular mechanisms that regulate chondrogenesis will facilitate development of therapies and/or prevention. The rapid development of the jaw cartilages in the zebrafish, its transparency and amenability to genetic manipulations makes it an excellent model for studying the genetic basis of craniofacial development. In addition, zebrafish mutants isolated in forward and reverse genetic screens have enabled the identification of critical genes involved in cartilage development [37,38,39]. Interestingly, several of those serve as models for human cartilage diseases, such as campomelic dysplasia (sox9) [39].
The dramatic ring1b cartilage phenotype is quite rare as only few mutants showing a total lack of pharyngeal elements and neurocranium have been identified. Remarkably, the ring1b mutant is the only one among those with cartilage defects that is missing the cleithrum, the pectoral girdle element that is not of Neural Crest origin. Not surprisingly, the sox9a/sox9b double mutant displayed severe craniofacial abnormalities. Zebrafish mutants in Trap230/Med12, a transcriptional co-activator for Sox9, were characterized by disrupted cartilage development [40,41]. Furthermore, knockdown experiments showed that runx2b and runx3 are critically involved in chondrogenesis: runx2b morphants lack the entire pharyngeal skeleton whereas only rudimentary trabeculae are formed in runx3 morphants [42].
Finally, in the lim-absent (lia)/fgf3 mutant, posterior arch formation is abrogated [43]. Potent inhibition of Fgf signaling, either chemically or by morholino-mediated knockdown of both Fgf3 and Fgf8, led to an almost complete loss of cartilage [44].
Our data indicate that Ring1b acts genetically upstream of Sox9 and Runx2, as their expression in (pre)chondrocytes is markedly decreased in ring1b mutants. We demonstrated previously that the Fgf-target genes dusp6, pea3 and spry4 were expressed at near normal levels in the pharyngeal arches of 32 hpf ring1b mutants [13], indicating that impaired Fgf signaling is unlikely to be underlying the cartilage defect.  Chondrogenesis involves a cascade of events that requires the concerted action of an intricate gene regulatory network. Although recent studies have provided insight into the molecular mechanisms underlying chondrogenesis, much remains unknown. For instance, the role of (epigenetic) chromatin modifications such as chromatin remodeling, DNA (de) methylation and histone modifications has largely been overlooked. Our results illustrate the critical role of epigenetic control in cartilage differentiation and highlight for the first time an essential and specific role for Polycomb genes in craniofacial development.
Despite the numerous molecular targets of Polycomb-mediated repression, the ring1b cartilage phenotype affects specific groups of cells and at a specific developmental point, namely the late stages of differentiation of the cranial neural crest cells into cartilage and muscle of the head. Our results indicate that epigenetic regulation of transcription in specific cell populations is very important for late differentiation events.
The cartilage phenotype in ring1b mutants consists of a block in terminal differentiation of chondrocytes. We propose that this is the result of deregulated expression of developmental programs rather than individual genes. This is in line with the proposed role of Ring1b to control differentiation programs in ES cells and of the Polycomb complex to regulate programs rather than individual genes.
Since no cartilage is formed in ring1b mutants, they provide an attractive platform to interrogate the function of (human) candidate genes for skeletogenesis, dissect possible genetic interactions, as well as screening of compounds to identify regulators of cartilage development.

Zebrafish Strains and Genotyping Methods
Zebrafish (Danio rerio) were maintained as previously described [45]. They were cared for in accordance with the NKI institutional guidelines and as approved by the Animal Experimentation Committee (DEC) of the Royal Netherlands Academy of Arts and Sciences (KNAW). Ring1b founder fish were out-crossed to AB and TL genetic backgrounds and genotyped as described [13] The mutation in ring1b/rnf2 that led to the insertion of 4 base pairs (bp) within the ZFN target site was named rnf2 ibl30/ibl30 and the mutation that caused deletion of 14 bp at the ZFN target site was named rnf2 ibl31/ibl31 (www.zfin.org).