Ciliopathic micrognathia is caused by aberrant skeletal differentiation and remodeling

Primary cilia are ubiquitous microtubule-based organelles that serve as signaling hubs for multiple molecular signaling pathways (Goetz and Anderson, 2010). Disruptions in the structure or function of primary cilia result in a class of disorders known as ciliopathies. Clinically, ciliopathies commonly present with a range of phenotypes including polydactyly, situs inversus, retinitis pigmentosa and renal cystic disease (Baker and Beales, 2009). Approximately 30% of all ciliopathies can be classified as craniofacial ciliopathies owing to the craniofacial complex being the primary organ system affected (Zaghloul and Brugmann, 2011). Craniofacial ciliopathies are characterized by the common presentation of several phenotypes including cleft palate, craniosynostosis and micrognathia (Schock and Brugmann, 2017). Although some progress has been made towards understanding the pathologies associated with craniofacial ciliopathies (Cela et al., 2018; Chang et al., 2016a; Chang et al., 2014; Kawasaki et al., 2017; Millington et al., 2017; Schock et al., 2015; Tian et al., 2017; Watanabe et al., 2019), the cellular and molecular etiologies of ciliopathic craniofacial skeletal anomalies remain unclear.

Development of the mandible begins with the generation of cranial neural crest cells (NCCs) from the rostral neural tube. A subset of Hox-negative NCCs migrate from the dorsal neural tube into the first branchial arch where they proliferate and populate the mandibular prominence (MNP) (Couly et al., 1996; Kontges and Lumsden, 1996). After initial formation and patterning of the MNP, NCCs begin to differentiate into skeletal derivatives. A subset of NCCs condense and differentiate into chondrocytes to form a bilateral cartilaginous structure called Meckel's cartilage. Meckel's cartilage serves as a template for proper growth of the mandible, but it is not necessary for mandibular bone development (Mori-Akiyama et al., 2003). Other NCCs in the MNP undergo intramembranous ossification, in which NCCs condense and directly differentiate into osteoblasts. During intramembranous ossification, osteoblast differentiation is initiated when distal-less homeobox 5 (DLX5) induces the expression of runt-related transcription factor 2 (RUNX2), the master transcriptional regulator of bone development (Holleville et al., 2007; Lee et al., 2003). RUNX2+ NCCs differentiate into osteoblasts and secrete a specialized extracellular matrix called osteoid tissue. Osteoblasts embedded within the osteoid mature into osteocytes (reviewed by Franz-Odendaal, 2011).

After bone deposition, the developing mandible begins to take its characteristic shape via bone remodeling. Within the craniofacial skeleton, both mesoderm-derived osteoclasts and neural crest-derived osteocytes contribute to bone resorption via the secretion of tartrate-resistant acid phosphatase (TRAP) (Minkin, 1982; Qing et al., 2012; Tang et al., 2012), osteoclast-specific matrix metalloproteinase 9 (MMP9) (Engsig et al., 2000; Reponen et al., 1994), or osteocyte-specific matrix metalloproteinase 13 (MMP13) (Behonick et al., 2007; Johansson et al., 1997). TRAP is secreted into the bony matrix where it dephosphorylates the structural phosphoproteins osteopontin and bone sialoprotein (Ek-Rylander et al., 1994), whereas matrix metalloproteinases (MMPs) enzymatically degrade extracellular matrix components such as collagen and elastin to remodel tissues (reviewed by Cui et al., 2017). Levels of bone resorption are inversely proportional to jaw size and play an important role in determining overall mandibular size between and among species (Ealba et al., 2015). Despite numerous studies demonstrating that ciliary dysfunction leads to micrognathia (Adel Al-Lami et al., 2016; Brugmann et al., 2010; Cela et al., 2018; Gray et al., 2009; Kitamura et al., 2020; Kolpakova-Hart et al., 2007; Zhang et al., 2011), whether or how loss of ciliary function affects bone resorption has yet to be described.

Owing to ample in ovo accessibility of NCCs and conserved organization of facial prominences, avian embryos have long been used to study craniofacial development (reviewed by Schock et al., 2016). Improved genome sequencing coupled with the existence of naturally occurring avian mutants have recently allowed researchers to make significant advances in understanding the etiology of developmental disorders using an avian model system. Perhaps the most used mutant avian lines have been those of the talpid family (Abbott et al., 1959, 1960; Ede and Kelly, 1964a,b). talpid² (ta²) is a naturally occurring avian mutant that is characterized by its striking presentation of polydactyly and craniofacial phenotypes (Abbott et al., 1959, 1960; Brugmann et al., 2010; Chang et al., 2014; Dvorak and Fallon, 1992; Muñoz-Sanjuán et al., 2001; Schneider et al., 1999). Our previous studies revealed that the causative mutation in the ta² was a 19 bp deletion in C2 calcium-dependent domain containing 3 (C2CD3) (Chang et al., 2014), a distal centriolar protein-coding gene required for ciliogenesis (Hoover et al., 2008; Ye et al., 2014). Impaired C2CD3-dependent ciliogenesis in the ta² results in facial clefting, ectopic archosaurian-like first generation teeth, hypo- or aglossia and micrognathia (Chang et al., 2014; Chang et al., 2016b; Harris et al., 2006; Muñoz-Sanjuán et al., 2001; Schock et al., 2015). Genetic, molecular and bioinformatic data from these studies determined that ta² was a bona fide model for the human craniofacial ciliopathy, oral-facial-digital syndrome 14 (OFD14) (Schock et al., 2015). Although micrognathia is a common and severe craniofacial phenotype associated with OFD14 (Boczek et al., 2018; Cortés et al., 2016; Thauvin-Robinet et al., 2014), the molecular and cellular etiology of micrognathia in OFD14 patients and in ta² embryos has yet to be explored.

In this study, we present the first in-depth analysis of the onset of ciliopathic micrognathia using the ta² model. Our data demonstrate that, although there is an expansion of the SOX9+ osteochondroprogenitor population in the ta², osteoprogenitors undergo precocious and incomplete differentiation, resulting in a reduced number of mature osteoblasts. Furthermore, analysis of bone remodeling markers demonstrated that there is increased bone resorption in the ta² mandible. Thus, our data suggest that ciliopathic micrognathia is due to the combinatorial effect of aberrant skeletal differentiation and remodeling.

To understand the etiology of ciliopathic micrognathia, we first characterized the onset of micrognathia in ta² embryos. The mandibular skeleton was first detectable via Alizarin Red staining at Hamburger-Hamilton stage (HH) 36. At this stage, skeletal elements that form the mandible, including the angular, dentary, splenial and surangular bones, were visible (Fig. 1A,A′). Relative to controls, Alizarin Red staining of HH36 ta² mandibles revealed a reduction in calcified tissue (Fig. 1B,B′). Volumetric measurements of the ta² angular, dentary, splenial and surangular bones confirmed a significant reduction in size relative to analogous control skeletal elements (Fig. 1C). ta² embryos possessed a medial ectopic skeletal element adjacent to the splenial bone, as revealed via Alizarin Red (Fig. 1A′,B′) and micro-computed tomography (μCT) analysis (Fig. 1D-E′). Although the volume of the endogenous splenial bone alone was significantly reduced in ta² embryos relative to controls, including the ectopic skeletal element in the analyses (the combined volume of the endogenous and ectopic ta² splenial bone) eliminated a significant difference in volume (Fig. S1A). Whereas total mandibular volume and surface area were not significantly different at this stage (Fig. S1B,C), total length measurements revealed that the ta² mandibles were significantly shorter than control mandibles at HH36 (Fig. 1F).

We further characterized mandibular development in control and ta² embryos at HH39, the latest stage at which ta² embryos can be consistently harvested before embryonic lethality (Abbott et al., 1959, 1960). At this stage, the developing ta² mandible was reduced both in volume and surface area when compared with stage-matched control mandibles (Fig. 1G-J). μCT analysis confirmed the presence of an ectopic skeletal element medially adjacent to a severely reduced splenial bone in HH39 ta² mandibles (Fig. 1H, yellow). These findings were further supported by whole-mount skeletal staining with Alizarin Red and Alcian Blue (Fig. S1D,E). Compared with HH39 control embryos, stage-matched ta² mandibles were dysmorphic and possessed reduced Alizarin Red staining, indicative of decreased bone mineralization (Fig. S1E). Length measurements between controls and stage-matched ta² mandibles confirmed that the micrognathia first observed at HH36 persisted in HH39 embryos (Fig. 1K). Although some elements in the mandible were dysplastic and thicker (a phenotype commonly seen in ciliopathic mutants; Brugmann et al., 2010; Cela et al., 2018; Kitamura et al., 2020; Kolpakova-Hart et al., 2007), the overall length of the mandible was reduced.

Avians have a distinct glossal apparatus that contains osteogenic and cartilaginous elements supported by the lingual process of the hyoid bone and rudimentary lingual muscles. Unlike the skeletal elements of the developing mandible that form through intramembranous ossification, the avian glossal and hyoid apparatus have skeletal elements that form through endochondral ossification, including the ceratobranchial bone (Lillie and Hamilton, 1952). To determine whether endochondral ossification was also affected in ta² embryos, we examined the development of the ceratobranchial bone via Alizarin Red and Alcian Blue staining. Total length measurements revealed that the ta² ceratobranchial was significantly reduced in size relative to controls (Fig. S1F-H). Although the hypoplastic ceratobranchial bones correlated with hypoglossia in ta² embryos, it is interesting that hypo/aglossia was a defining phenotype of OFD14 patients despite humans lacking skeletal elements in the glossal apparatus (Boczek et al., 2018; Chang et al., 2014; Schock et al., 2015; Thauvin-Robinet et al., 2014). Thus, taken together, these analyses suggested that both endochondral and intramembranous ossification were impaired in ta² mutants. Considering these findings, we next sought to examine the etiology of ciliopathic micrognathia by examining both cellular processes and molecular pathways required for mandibular development.

Although the mechanistic link between a number of signaling pathways and the cilium is not well understood, it is well-established that Gli-mediated Hedgehog (Hh) signaling requires primary cilia (Huangfu and Anderson, 2005; Huangfu et al., 2003). We previously reported that Hh signaling and Gli activity were impaired in ta² embryos (Chang et al., 2014). Thus, to better understand the impact that aberrant Gli processing had on cellular processes essential for mandibular development, we examined genes associated with Gli-bound loci in the MNP, as determined by ChIP-seq (GSE146961) (Elliott et al., 2020; Lorberbaum et al., 2016) and cross-referenced this dataset against our previously published bulk RNA-seq dataset from control and ta² MNPs (GSE52757) (Chang et al., 2014). These analyses revealed a total of 1609 differentially expressed genes that were in close proximity to a Gli-bound locus, with an approximately equal number of target genes being either down- or upregulated (Fig. S2, Table S1). Gene ontology (GO) analysis for phenotypes associated with the 1609 differentially expressed presumptive Gli target genes in the ta² MNP tightly correlated with the observed phenotype and included ‘abnormality of the mandible’, ‘abnormal jaw morphology’ and ‘abnormal facial shape’. Further GO analyses revealed the top biological processes impacted in the ta² MNP included ‘positive regulation of cell cycle’, ‘ossification’, ‘regulation of microtubule cytoskeleton organization’ and ‘regulation of osteoblast differentiation’ (Fig. 2). Based on these results we examined three cellular processes (cell-cycle progression, ossification and osteoblast differentiation) during mandibular development in ta² embryos in an attempt to better understand the molecular and cellular basis for ciliopathic micrognathia.

Development of the mandibular skeleton initiates with an expansion of NCCs within the developing MNP. Ciliopathic mutants are particularly vulnerable to cell cycle disruptions as the centrioles required for ciliogenesis are the same organelles required for formation of the mitotic spindle. In vitro experiments have established that ciliogenesis is initiated at G1/G0. A mature/fully extended cilium is present at G0 and, upon cell cycle re-entry, microtubule stability is reduced, resulting in cilia disassembly (Fig. 3A) (Doxsey et al., 2005; Izawa et al., 2015; Kim et al., 2015; Kim and Tsiokas, 2011; Pugacheva et al., 2007; Tucker et al., 1979; Yeh et al., 2013). After disassembly, the basal body is released from its role in ciliogenesis, thereby freeing up centrioles to function in forming the mitotic spindle (Fig. 3A) (Kobayashi and Dynlacht, 2011; Nigg and Stearns, 2011). Our unbiased GO-analyses suggested that gene expression changes associated with ‘positive regulation of the cell cycle’ were enriched in ta² mutants. Given the impaired ciliogenesis in ta² mutants, we examined expression of genes associated with the G1/S checkpoint, a point at which ciliary retraction is necessary for cell passage into S phase and DNA replication (Tucker et al., 1979). We hypothesized that the lack of ciliogenesis in ta² mutants would correlate with unchecked progression from G1 to S. We first examined p21 expression: p21 arrests the passage from G1 to S phase if DNA damage or microtubular aberration is detected (Barr et al., 2017; el-Deiry et al., 1993). Thus, we hypothesized that reduced ciliogenesis in ta² mutants would correlate with reduced p21 expression and unchecked passage through the cell cycle. qRT-PCR analysis revealed that p21 expression was significantly decreased in the ta² MNPs relative to controls (Fig. 3B). We next examined expression of two positive G1/S regulators, cyclin D3 (CCND3) and ribosomal protein L23 (RPL23) (Bartkova et al., 1998; Qi et al., 2017). Expression of CCND3 and RPL23 were upregulated in ta² mutants by 28% and 43%, respectively (Fig. 3B). Together, these results suggest that impaired ciliogenesis in ta² mutants allows for unchecked progression through the G1/S cell cycle checkpoint.

Given these results, we next sought to determine whether and how unchecked cell cycle progression impacted cell proliferation rates in the osteochondroprogenitor population of the ta² MNP. Immunostaining for SOX9 (Ng et al., 1997) and phosphorylated histone H3 (PHH3) in HH26 control and ta² MNPs revealed that ta² MNPs had an increased SOX9+ population relative to control MNPs (Fig. 3C-G), despite there being no significant difference in total MNP area between control and ta² embryos (representing 21% of the total area in controls and 26.9% in ta² MNPs) (Fig. 3H). Double immunostaining for PHH3 and SOX9 further revealed that proliferation was significantly increased specifically within the SOX9+ population (Fig. 3I), whereas the proliferation was significantly decreased within the SOX9− mesenchyme (Fig. 3J). Cell death analysis, as assessed by cleaved caspase 3 (CC3) expression, revealed apoptosis was significantly decreased in ta² MNPs across both SOX9+ and SOX9− populations (Fig. S3A-F). Together, these data suggest that unchecked cell cycle progression leads to hyperproliferation specifically within SOX9+ osteochondroprogenitor cells in the ta² MNP. Previous studies have suggested that positively altering cell cycle progression in NCCs would result in premature formation of a mesenchymal condensation and cause the subsequent premature initiation of osteogenic differentiation (Hall et al., 2014). To further examine this possibility, we next analyzed the onset of osteogenic differentiation in ta² mutants.

After sufficient proliferation, multipotent NCCs form a mesenchymal condensation and differentiate into several lineages, including the chondrogenic and osteogenic derivatives (Le Douarin, 1982). As expression for genes associated with the GO term ‘regulation of osteoblast differentiation’ were altered in ta² embryos, we sought to examine this process in a stepwise fashion. RUNX2 functions as the master transcriptional regulator for osteoblast differentiation, as increased RUNX2 expression is required for the initial skeletal differentiation of NCCs into preosteoblasts (Komori et al., 1997; Otto et al., 1997). Once NCCs express RUNX2, they proceed down the osteoblast lineage (Fig. 4A) (Kobayashi et al., 2000). Previous studies have reported that RUNX2 overexpression in earlier stages of development resulted in micrognathia in chick embryos (Hall et al., 2014). In the MNP, RUNX2 expression is regulated both positively and negatively by additional transcription factors expressed in the developing MNP. DLX5 induces the expression of RUNX2 (Holleville et al., 2007; Lee et al., 2003), whereas heart and neural crest derivatives expressed 2 (HAND2) inhibits RUNX2 expression (Funato et al., 2009). To determine whether osteoblast differentiation was disrupted in ta² MNPs, we spatially and quantitatively examined expression of these genes. Both whole-mount and single molecule fluorescent in situ hybridization (Wang et al., 2012) revealed that DLX5 expression was shifted distally around Meckel's cartilage in the HH29 ta² MNPs relative to controls (Fig. 4B-E′). Despite this spatial shift in expression, qRT-PCR analysis did not detect a significant change in the level of DLX5 expression in the ta² MNP (Fig. 4F). Conversely, HAND2 expression was lost distally and significantly decreased within the developing ta² MNP (Fig. 4G-K). Lastly, examination of RUNX2 expression demonstrated a significant and robust proximal expansion in the developing ta² MNP (Fig. 4L-O′). This observation was confirmed quantitatively via qRT-PCR, which revealed that RUNX2 expression was significantly increased by 38% in the ta² MNP (Fig. 4P). Sagittal z-stack projections of the MNP revealed that transcripts for all three genes were detected ventrally adjacent to the SOX9+ population (Fig. S4A-F; Movie 1). Together, these data suggest that excessive expression of RUNX2, correlated with the decreased expression of the negative regulator HAND2, resulted in ectopic production of pre-osteoblasts in the ta² MNP.

Increased RUNX2 expression could be indicative of increased bone deposition; however, in the ta², increased RUNX2 expression accompanied a micrognathic phenotype. We hypothesized that, despite an expansion of pre-osteoblasts, transition towards a mature osteoblast was impaired in ta² mandibles. Osteoblast maturation occurs in three stages. In the first stage, pre-osteoblasts proliferate and express collagen type 1 (COL1A1) protein. In the second stage, pre-osteoblasts exit the cell cycle and differentiate, while expressing alkaline phosphatase (ALPL) and COL1A1 (Rodrigues et al., 2012). In the last phase, mature osteoblasts express osteocalcin (OCN; also known as BGLAP), a protein that is essential for matrix mineralization (Fig. 5A) (reviewed by Rutkovskiy et al., 2016). First, to confirm the specificity of our reagents to delineate these stages of osteoblast maturation, we performed RNAscope in situ hybridization for ALPL and immunohistochemistry for COL1A1 and OCN on sagittal sections of an HH35 control femur (Fig. S5). We readily identified ALPL+ cells in the bony collar of the developing femur, COL1A1+ pre-osteoblasts within the extracellular matrix and OCN+ cells in the bony matrix (Fig. S5C-G). In frontal sections through the ta² surangular bone (Fig. 5B), COL1A1 expression was increased relative to controls (Fig. 5C,D). qRT-PCR analysis supported this finding and revealed a significant, twofold upregulation in the expression of COL1A1 (Fig. 5E). ALPL expression, which was localized to cortical surfaces of the developing skeletal elements in HH35 control mandibles, was reduced in ta² mandibles (Fig. 5F,G). Quantification of ALPL transcripts in HH35 mandibles revealed a 35% downregulation in ta² mandibles in comparison with control mandibles (Fig. 5H). Lastly, we analyzed the expression of OCN. OCN expression was reduced in the developing ta² mandible when compared with controls (Fig. 5I,J). qRT-PCR analysis supported this finding and revealed that OCN expression was significantly reduced by ∼72% in ta² mandibles relative to controls (Fig. 5K). Taken together, these results suggest that, despite an expansion of the pre-osteoblast population, osteoblast maturation was severely impaired. Thus, our data suggest that impaired osteoblast maturation and subsequently reduced bone deposition contributes to ciliopathic micrognathia.

Several studies have reported that proper determination of jaw length requires regulated bone remodeling. Bone remodeling is a dynamic process that consists of resorption of bony matrix by osteoclasts and osteocytes and deposition of new bony matrix by osteoblasts. We next sought to determine whether aberrant bone remodeling also contributes to ciliopathic micrognathia, as it was previously reported that the amount of bone resorption was inversely proportional to jaw length in avian embryos (Ealba et al., 2015). To test the hypothesis that increased bone resorption contributes to ciliopathic micrognathia, we examined the expression of several markers for this process in sections through the HH39 surangular bone. We first assayed TRAP activity. When compared with controls, TRAP staining was increased in HH39 ta² mandibles, indicative of increased bone resorption (Fig. 6A,B). We next examined expression of MMPs, a family of proteases necessary for osteoclast recruitment and solubilization of the osteoid during bone remodeling (reviewed by Cui et al., 2017). During mandibular remodeling, MMP13 is expressed in NCC-derived osteoblasts and osteocytes (Behonick et al., 2007; Johansson et al., 1997). In situ hybridization and qRT-PCR revealed increased MMP13 expression in the HH39 ta² mandible when compared with controls (Fig. 6C,D). To confirm this result, we analyzed a second marker of bone resorption, secreted phosphoprotein 1 (SPP1). SPP1 is responsible for osteoclast adhesion and bone resorption across various species (Choi et al., 2008; Pinero et al., 1995). SPP1 was more prominently expressed in the medullar region of the developing ta² mandible relative to controls (Fig. 6E,F). These data were verified via qRT-PCR, revealing a significant twofold increase in both MMP13 and SPP1 expression in the ta² mandibles relative to controls (Fig. 6G,H). Finally, we examined the ratio of receptor activator of NF-κB ligand (RANKL; TNFSF11) to osteoprotegerin (OPG; TNFRSF11B) expression. OPG protects bone from excessive resorption by binding to RANKL and preventing it from binding to its receptor, RANK (TNFRSF11A) (Fig. 6I). qRT-PCR analyses revealed that both RANKL and OPG expression were significantly increased in the developing ta² mandible (Fig. 6J). Aberrant expression of both markers resulted in an increased ratio of RANKL/OPG expression (Fig. 6J), further suggesting that excessive bone resorption contributes to the onset of ciliopathic micrognathia in ta² embryos.

Herein, we explored the etiology of ciliopathic micrognathia using the ta² model as our guide. We found that, despite increased proliferation and expansion of osteochondroprogenitors and preosteoblast populations, osteoblast maturation and subsequent bone deposition was impaired in the ta² mandible. Furthermore, excessive amounts of bone resorption were also detected following impaired bone deposition (Fig. 7). These studies are significant as they increase the understanding of the cellular processes and molecular pathways that can be targeted for future therapeutic approaches for ciliopathies.

Micrognathia is a significant biomedical burden present in almost 20% of ciliopathies (Schock and Brugmann, 2017). The predominant treatment for micrognathia is distraction osteogenesis, an invasive and traumatic procedure in which the mandible is cut and gradually separated to allow new bone growth until the mandible reaches a desired size (Ilizarov, 1988). To develop less invasive therapeutic options for ciliopathic micrognathia, it is important to gain a more extensive understanding of its etiology. Although studies have shown that loss of primary cilia leads to micrognathia, a thorough analysis of how the cellular and molecular mechanisms essential for mandibular development are impaired in ciliopathic mutants has yet to be described. To this end, we used the avian ciliopathic mutant ta², a bona fide model of the human ciliopathy OFD14, to investigate how ciliopathic micrognathia arose. Whereas the majority of ciliopathic studies use conditional knockout murine models, the naturally occurring ta² model, in which the ciliopathic insult is ubiquitous throughout the organism, allowed for analyses more relevant to human ciliopathic conditions. Our previous data, together with that reported herein, revealed that loss of C2CD3-mediated ciliogenesis resulted in impaired Gli-mediated Hh signal transduction. Impaired Hh signaling (via direct or indirect mechanisms) resulted in unchecked cell cycle progression, subsequent increased proliferation of SOX9+ osteochondroprogenitors and impaired expression of gene regulatory networks necessary for the maturation of NCC-derived osteoblasts. Furthermore, excessive bone resorption was detected in ta² mandibles later in development (Fig. 7A,B). To our knowledge, this work represents the first in-depth description of the etiology of ciliopathic micrognathia.

The cell cycle and ciliogenesis both require centriolar function and are thus inextricably linked. Previous studies solidified this relationship, reporting that knockdown of the ciliary protein IFT88 promoted cell cycle progression and proliferation (Robert et al., 2007). Furthermore, previous studies have also shown that cell cycle progression and the onset of osteogenesis are tightly correlated and autonomously controlled within NCCs (Hall et al., 2014). Our current study revealed that NCC-derived skeletal progenitors in the MNP were specifically affected, as seen by increased cell proliferation in the SOX9+ population of ta² MNPs. Integrating these data with those previously reported (Ealba et al., 2015; Hall et al., 2014), we hypothesize that loss of cilia allows for the unchecked passage of NCCs in the MNP through the cell cycle, which subsequently results in premature formation of a mesenchymal condensation and early execution of the skeletal differentiation program. Although, our previous studies examining cell proliferation in early, undifferentiated NCC populations of the frontonasal prominence did not reveal any alteration in proliferation rates between control and ta² embryos (Schock et al., 2015), these data still beg the question: are NCC-derived skeletal progenitors more sensitive to a ciliopathic insult than other cell types?

There is some precedent for the idea of cell-specific sensitivities. Treacher-Collins syndrome (TCS) is a congenital craniofacial disorder characterized by hypoplasia of facial bones, high arched palate and ear defects. Seminal studies addressing the etiology of TCS determined that NCCs, owing to their high energetic demands as highly proliferative and migratory multipotent cells, were especially sensitive to deficient ribosome biogenesis and subsequent cellular stress (Dixon et al., 2006; Jones et al., 2008). Studies on various cell types are beginning to uncover a connection between the cilium and energy homeostasis (Arsov et al., 2006; Collin et al., 2005; Davenport et al., 2007; Marion et al., 2009; Pampliega et al., 2013; Tang et al., 2013). In addition, the process of osteogenic differentiation, and maturation of osteoblasts in particular, has a heavy requirement of oxidative phosphorylation and glycolysis (Guntur et al., 2014; Regan et al., 2014). Thus, the lack of primary cilia on skeletal progenitors could result in a similar type of cellular stress, preventing ta² NCCs from executing proliferation and differentiation programs that allow for proper skeletogenesis in the developing MNP.

Gli transcription factors are post-translationally processed into functional activator (GliA) or repressor (GliR) isoforms at the primary cilium (Goetz and Anderson, 2010; Haycraft et al., 2005; Liu et al., 2005). Loss of functional Gli isoforms results in altered transcription of Hh pathway target genes (Chang et al., 2014). We have previously shown that all facial prominences of ta² mutants have perturbed Gli processing and subsequent disruptions in Hh signal transduction (Chang et al., 2014). Here, we advanced our understanding of the molecular etiology of ciliopathic phenotypes by comparing Gli bound loci with genes that were differentially expressed in ta² MNPs. Our results suggested that loss of proper Gli-mediated transcription impacted several cellular processes necessary for mandibular development, including positive regulation of the cell cycle, ossification and osteoblast maturation. These data not only explain the phenotypic similarities frequently observed between Hh and ciliopathic mutants (Abzhanov et al., 2007; Billmyre and Klingensmith, 2015; Elliott et al., 2020; Jeong et al., 2004; Lenton et al., 2011; Xu et al., 2019), but also provide new insight into pathways that may be available for therapeutic intervention.

Although these data strongly suggest that impaired Gli-mediated Hh signaling is molecularly causal for the micrognathic phenotype, the exact molecular mechanisms by which this occurs remains unknown. For example, it is possible that the onset of micrognathia is due to a failure of Gli isoforms to directly regulate transcription of targets necessary for skeletal differentiation. Our GLI2 and GLI3 ChIP-seq data reveal Gli peaks near the transcriptional start site of HAND2 and within the RUNX2 locus (Elliott et al., 2020). It is also possible, however, that impaired ciliary-dependent Gli processing impairs the ability of Gli proteins to interact with skeletal transcription factors. GLI2 was reported to physically interact with RUNX2 to direct osteoblast differentiation (Shimoyama et al., 2007) and GLI3 was reported to utilize HAND2 as a co-factor to drive the mandibular patterning and osteogenic transcriptional programs (Elliott et al., 2020). Understanding the exact molecular mechanisms associated with ciliopathic phenotypes is the focus of our ongoing work.

Despite our focus on the Hh pathway, other signaling pathways, such as the Wnt, fibroblast growth factor (FGF), and bone morphogenetic protein (BMP) pathways, are also known to be involved in skeletal development (Jiang et al., 2014; Merrill et al., 2008; Mina et al., 2007); however, the mechanisms by which these pathways require cilia-dependent signaling transduction is less clear. The role of the cilium in canonical Wnt signaling has been heavily disputed. Initial studies demonstrated that numerous Wnt pathway components, such as inversin, Vangl2 and Apc, localized to the ciliary axoneme or basal body, suggesting a role for primary cilia in Wnt signal transduction (Morgan et al., 2002; Ross et al., 2005; Simons et al., 2005). Subsequent studies demonstrated that the cilium functions to repress Wnt signaling, as loss of cilia correlated with increased Wnt activity and Wnt target gene expression (Corbit et al., 2008; Lancaster et al., 2011). Concurrent studies in mice and zebrafish that lack various ciliary proteins retained normal levels of canonical and non-canonical Wnt signaling transduction (Huang and Schier, 2009; Ocbina et al., 2009). Further, in the NCC-derived facial mesenchyme, conditional loss of the ciliary protein Kif3a in NCCs did not appear to alter Wnt activity (Brugmann et al., 2010). Recently, it was found that deletion of Kif3a plays a role in Wnt signaling in a ciliary-independent fashion by activating the pathway in an autocrine manner (Kim et al., 2016). If and how C2CD3 regulates Wnt activity in the ta² during craniofacial skeletal development has not yet been described. Our future research will examine the impact loss of C2CD3-mediated ciliogenesis has on Wnt signaling.

BMP and FGF signaling pathways are also required for mandibular skeletogenesis (Ashe et al., 2012; Merrill et al., 2008; Mina et al., 2007), but there is a lack of understanding of how these signals are affected in ciliopathic backgrounds. The ciliopathic Fuz^−/− mouse mutant exhibits craniofacial features that closely resemble FGF hyperactivation syndromes, such as craniosynostosis and high arched palate (Tabler et al., 2013). In addition, our previous studies with the ta² mutant have demonstrated that the developing MNP and maxillary prominence have increased FGF signaling (Schock et al., 2015). However, the mechanistic link between cilia and the FGF signaling pathway has not yet been established. In addition, it is unclear whether, and how, a BMP signal is transduced through the primary cilium (reviewed by Kaku and Komatsu, 2017). Our future goals include gaining a broader understanding of how other signaling pathways essential for skeletal development are affected in ciliary mutants.

Bone is constantly remodeled through bone deposition and bone resorption. Disturbances in the skeletal environment caused by excessive bone remodeling lead to decreased jaw length and density (Ealba et al., 2015; Feng and McDonald, 2011). As bone remodeling requires both bone deposition and bone resorption, defects in either program can lead to bone remodeling diseases such as osteoporosis and osteopetrosis (Feng and McDonald, 2011). It has also been long established that osteoblasts and osteoclasts communicate to control bone remodeling through physical interaction and release of various growth factors and chemokines (Matsuo and Irie, 2008); however, how primary cilia mechanically or molecularly contribute to this communication has not yet been discovered. Previous studies have reported that primary cilia may play an antiresorptive role in osteocytes through increasing the RANKL/OPG ratio (Malone et al., 2007). Although our studies confirm this finding, as loss of cilia resulted in increased bone resorption in the ta² mandible, future work on the molecular and mechanical control of bone resorption by primary cilia is necessary.

The discovery that bone remodeling contributes to ciliopathic micrognathia is a significant finding as it opens a potential avenue of therapeutic intervention. Studies have shown that bone resorption can be pharmacologically induced or depleted with recombinant forms of OPG and RANKL in avian models (Ealba et al., 2015). In addition, bisphosphonates have previously been used to decrease bone resorption by preventing osteoclast formation and inducing osteoclast apoptosis (Boonekamp et al., 1986; Löwik et al., 1988; Sato and Grasser, 1990). Lastly, specific αNAC polypeptides have the potential to induce osteoblastic maturation in vivo (Meury et al., 2010). Treating ciliopathic micrognathia models, including the ta², with these compounds during discrete temporal windows of remodeling is a focus of our ongoing research.

It is well-established that increased mechanical loading stimulates bone formation (reviewed by Robling and Turner, 2009), and numerous studies have demonstrated a role for primary cilia as mechanosensory organelles during skeletal development (Leucht et al., 2013; Malone et al., 2007; Temiyasathit and Jacobs, 2010; Temiyasathit et al., 2012). Conditional loss of cilia in osteoblasts and osteocytes causes reduced bone formation due to reduced loading (Temiyasathit et al., 2012), decreased osteogenic response during fracture healing (Leucht et al., 2013) and inhibition of fluid flow-induced expression of osteogenic markers (Malone et al., 2007). When cilia on murine osteoblasts were removed with chloral hydrate, there was a lack of mineral deposition as a response to oscillatory fluid flow, demonstrating the importance of cilia for bone development (Delaine-Smith et al., 2014).

Several studies have also suggested that a key role of the primary cilium is to function as a chemical sensor for extracellular Ca²⁺ (Delling et al., 2013; Nauli et al., 2016; Zayzafoon, 2006). Primary cilia possess membrane-based ion channels including PKD2L1 (TRPP3), which allow for intracellular flux of Ca²⁺ (DeCaen et al., 2013; Nauli et al., 2016). Although this theory is controversial (Delling et al., 2016), this is an interesting direction for future experimental work given that bone mineralization requires Ca²⁺. Ca²⁺ is essential for bone mineralization as it precipitates with inorganic phosphate to form hydroxyapatite crystals in collagen-rich extracellular matrix (reviewed by Murshed, 2018). Low Ca²⁺ intake and incorporation lead to severe skeletogenic disorders, such as osteoporosis and osteopenia, which are presented with low osteoblast maturation and increased bone remodeling (Anderson, 1996; Dymling, 1964). Could C2CD3-mediated loss of cilia prevent Ca²⁺ processing or integration into skeletal elements? This question is particularly intriguing for the fact that bone deposition, differentiation and remodeling are connected to Ca²⁺ signaling (Delling et al., 2013; Nauli et al., 2016; Zayzafoon, 2006) and that C2CD3 contains several C2 Ca²⁺-dependent domains that are very poorly understood. Future studies will help to address whether and how Ca²⁺ integration is affected in ciliopathic mutants, and will be crucial for the understanding of how micrognathia and other ciliopathic skeletal disorders can be treated.

Fertilized control and ta² eggs were supplied from University of California, Davis, CA, USA. Embryos were incubated at 38.8°C for 5-13 days and then harvested for analysis. All embryos collected were staged according to the Hamburger-Hamilton staging system and genotyped as previously described (Chang et al., 2014; Hamburger and Hamilton, 1951). For all experiments, three control^+/+ and three ta² embryos at each stage were used, unless noted otherwise in the text or figure descriptions. Embryos were fixed in 4% paraformaldehyde (PFA) overnight at 4°C, unless noted otherwise.

Whole-mount skeletal staining was performed as previously described (Rigueur and Lyons, 2014) with several modifications. Briefly, embryos were collected in PBS and fixed in 95% ethanol (EtOH) overnight. To remove excess adipose tissue, embryos were incubated in acetone overnight. To stain cartilage, embryos were placed in 0.03% Alcian Blue solution (Sigma-Aldrich, A5268) for 2 h. Following cartilage staining, samples were washed with seven parts EtOH and three parts acetic acid until the solution was clear and incubated in 95% EtOH overnight. To stain bone, samples were incubated in 0.005% Alizarin Red S (Sigma-Aldrich, A5533) in 1% KOH for 3 h at room temperature and cleared in 1% KOH. Once cleared, samples were incubated in 50% glycerol/50% KOH solution. For imaging and long-term storage, samples were kept in 100% glycerol. Stained specimens were imaged using a Leica M165 FC stereo microscope system.

Mandibular and ceratobranchial length measurements were performed using Leica LAS X software. Alizarin Red-stained mandibles were measured from the distal tip of the dentary bone to the proximal edge of the surangular bone for each side of the mandible and then averaged. For the mandibular length and area measurements, Alizarin Red-stained mandibles were measured in mm or bones were traced in Fiji (Schindelin et al., 2012) and areas of each bone were measured and calculated in mm². Volumetric analysis of individual skeletal elements (voxel measurements) were taken from the μCT-analyzed embryos. Unpaired one-tailed Student's t-test was used for statistical analysis. P<0.05 was determined to be significant.

HH36 and HH39 mandibles were harvested, fixed in 4% PFA and stained with 0.005% Alizarin Red S in 1% KOH solution for 3 h at room temperature. The samples were further processed at the Preclinical Imaging Core in the University of Cincinnati Vontz Center for Molecular Studies, OH, USA. Mandibular surface and volume area were extracted from DICOM files and measured in voxels. These files were analyzed and processed in Imaris 9.3 (Bitplane, Oxford Instruments).

Hematoxylin and eosin (H&E) staining was performed using standard protocols. Pentachrome staining was performed as previously described with slight modifications, as methyl cellosolve acetate washes were substituted with xylene washes and slides were transferred to alkaline ethanol solution for 90 min (Olah et al., 1977). TRAP staining was performed on 8 μm thick frontal sections of undecalcified HH39 mandibles using the Acid Phosphatase Leukocyte (TRAP) kit (Sigma-Aldrich, 387A) following the manufacturer's protocol, with Fast Red Violet LB Solution (Sigma-Aldrich, 912) used in place of provided Fast Garnet GBC Solution.

Whole mandible immunofluorescence was carried out as previously described with minor modifications (Williams et al., 2018). Femur and mandibles were sagittally and frontally sectioned at 8 µm. Triton-X concentration was increased to 1%, and the sections and embryos were incubated in blocking solution supplemented with 5% normal goat serum. Cleaved Caspase-3 (1:100, Cell Signaling Technology, 9661), Collagen Type I Alexa-Fluor 488 Conjugate (1:200, SouthernBiotech, 1310-30), Osteocalcin (1:20, Abcam, 17.0021) and Phospho-Histone H3 (1:200, Cell Signaling Technology, 9706) primary antibodies were used. The secondary antibodies used were goat anti-mouse-568 (1:250, Thermo Fisher Scientific, A-11004), goat anti-rabbit-647 (1:250, Thermo Fisher Scientific, A21244) and Sox9-Conjugate AlexaFluor-488 (1:100, Cell Signaling Technology, 94794). Nuclei counterstaining was carried out by incubation in 1 μg/ml DAPI in PBS+Triton 1% for 2 h and processed for clearing with Ce3D solution overnight (Li et al., 2017).

HH29 heads, HH35 heads and HH39 mandibles were fixed in 4% PFA at 4°C for 16-24 h. HH35 heads and HH39 mandibles were decalcified in 19% EDTA solution at 4°C for 2 and 5 days, respectively. Samples were dehydrated in an ethanol series, washed in xylene, embedded in paraffin and sectioned at 8 μm thickness using a Sakura Accu-Cut SRM 200 microtome. Transcripts of RUNX2 (ACD 571591), HAND2 (ACD 571571-C2) and DLX5 (ACD 571561) were detected using the RNAscope Multiplex Fluorescent V2 kit per the manufacturer's instructions. Transcripts of SPP1 (ACD 571601), ALPL (ACD 837811) and MMP13 (ACD 571581) were detected using RNAscope 2.5 HD Duplex Assay (ACD) per the manufacturer's instructions. Briefly, slides were baked in a hybridization oven for 1 h at 60°C, deparaffinized using xylene, dehydrated using 100% EtOH and allowed to dry completely at room temperature. Endogenous peroxidase activity was quenched using RNAscope Hydrogen Peroxide and washed using distilled water. Target retrieval was performed using 1× RNAscope Target Retrieval Buffer in an Oster steamer for 15 min, washed in distilled water, dehydrated in 100% EtOH and dried at 60°C. Slides were treated with ACD Protease Plus in an HybEZ II oven for 30 min at 40°C. Probes were hybridized for 2 h at 40°C in an HybEZ II oven. Amplification steps were performed as described by the manufacturer. Signal development for RUNX2, HAND2 and DLX5 were carried out using Cyanine 3 (PerkinElmer, NEL752001KT) diluted 1:500 in RNAscope Multiplex TSA Buffer. Signal amplification for SPP1, ALPL and MMP13 was carried out using provided HRP-based Green chromogen. Slides were counterstained and mounted per the manufacturer's instructions and imaged using a Leica DM5000B upright microscope system.

Fluorescent whole mount in situ hybridization (FISH) was performed using digoxigenin-labeled riboprobes as previously described (Denkers et al., 2004). Antisense riboprobes against DLX5 (NM_204159.1; primers F 5′-ATCAGGTCCTCCGACTTCCA-3′ and R 5′-ATACGACTCACTATAGGGGATTTTCACCTGCG TCTGCG-3′), HAND2 (NM_204966.2; primers F 5′-CGAGGAGAACCCCTACTTCC-3′ and R 5′-TAATACGACTCACTATAGGGCCTGTCCGCCCTTTGGTTT-3′) and RUNX2 (NM_204128.1; primers F 5′-CGCATTCCTCATCCCAGTAT-3′ and R 5′-TAATACGACTCACTATAGGGTATGGAGTGCTGCTGGTCTG-3′) were designed to range from 600 to 800 bp. TSA Plus HRP-Fluorescence kits were used for both Cyanine 3 and Cyanine 5 fluorescence channels (PerkinElmer, NEL752001KT). For simultaneous SOX9 immunofluorescence, embryos were washed with Tris-NaCl-Tween (TNT) and incubated overnight in 5% normal goat serum in TNT. Nuclei were counterstained with DAPI. For the vector rendering, confocal images were loaded in Bitplane Imaris and the surface algorithm was run. The selection of the surface was made by subtracting the background and adjusting the brightness until no signal was detected outside of the surface. The spots rendering for PHH3 and CC3 were adjusted for the detection of 2 µm signals, and background subtraction was applied.

RNA was extracted using TRIzol reagent (Invitrogen) and cDNA was synthesized using SuperScript III (Invitrogen). HH39 mandibles were first frozen with liquid nitrogen and ground using a mortar and pestle to ensure homogenous extraction. SYBR Green Supermix (Bio-Rad) and a Quant6 Applied Biosytems qPCR machine were used to perform qRT-PCR. All the genes were normalized to GAPDH expression. Negative controls were performed by omitting the cDNA in the mixture. The level of expression for each gene was calculated using the 2^−ΔΔCq method (Livak and Schmittgen, 2001). Unpaired one-tailed Student's t-test was used for statistical analysis. P<0.05 was determined to be significant.

We thank Mary Delany and the University of California, Davis, Avian Facility, Jackie Pisenti and Kevin Bellido for maintenance and husbandry of the talpid² colony. Technical assistance was given by Dr Matt Kofron and Evan Meyer for image acquisition and analysis (Confocal Imaging Core at Cincinnati Children's Hospital Medical Center) and Dr Lisa Lemen for μCT acquisition and analysis (Preclinical Imaging Core at the University of Cincinnati). We also thank members of the Brugmann lab for helpful comments and feedback, especially Dr Kelsey Elliott for her valuable help analyzing the RNA-seq data.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.194175.reviewer-comments.pdf