Nucleolin loss of function leads to aberrant Fibroblast Growth Factor signaling and craniofacial anomalies

The craniofacial complex consists of the primary sense organs, central and peripheral nervous systems, and musculoskeletal components of the head and neck. Craniofacial development is therefore an intricate process, which involves the coordinated interaction of all three germ layers, but it is also sensitive to environmental and genetic insults, which often result in craniofacial disorders. Despite advances in genomic sequencing, many affected individuals have an unknown genetic diagnosis. Therefore, it is necessary to identify novel genetic factors and better understand the cellular and molecular mechanisms that regulate normal craniofacial development, which may also aid in identifying potential therapeutic targets to prevent or ameliorate craniofacial diseases.

Ribosome biogenesis is essential for cell growth and survival because ribosome quantity and quality dictate the translation of mRNA into proteins. Transcription of ribosomal DNA (rDNA) by RNA Polymerase (Pol) I in the nucleolus generates a 47S pre-ribosomal RNA (pre-rRNA). The 47S pre-rRNA is then cleaved and processed into 18S, 5.8S and 28S rRNAs. These rRNAs, together with Pol III-transcribed 5S rRNA, associate with ribosomal proteins and accessory proteins to form ribosomes (Laferté et al., 2006). The transcription and processing of pre-rRNA requires RNA Pol I together with associated proteins such as UBTF and SL-1 (Friedrich et al., 2005; Panov et al., 2006), as well as rRNA-processing proteins including Treacle (encoded by Tcof1) (Jones et al., 2008; Valdez et al., 2004), Nol11 (Freed et al., 2012), Wrd43 (Zhao et al., 2014) and Fibrillarin (Tollervey et al., 1991). Interestingly, when Pol I subunits or associated factors are disrupted or mutated in zebrafish, frogs or mice, it results in developmental defects that mostly affect craniofacial cartilage and bone differentiation (Achilleos and Trainor, 2015; Bouffard et al., 2018; Dixon et al., 2006; Griffin et al., 2015; Watt and Trainor, 2014; Watt et al., 2016, 2018; Yelick and Trainor, 2015; Zhao et al., 2014). This raises the question of why disruptions in these ubiquitously expressed genes, which are required in a global process, result in tissue-specific craniofacial anomalies. One hypothesis is that neural crest cells (NCCs), which are the progenitors of most of the craniofacial bone and cartilage, are more proliferative or metabolically active than non-NCCs (Falcon et al., 2022). In addition, NCCs form via an epithelial-to-mesenchymal transition, which involves major cytoskeletal changes, thus requiring high levels of new protein synthesis and, accordingly, more rRNA transcription. Another hypothesis is that the RNA Pol I subunit, associated proteins and rRNA-processing proteins have other non-ribosomal functions, which, together with the regulation of rRNA synthesis, make the craniofacial skeleton more susceptible to disruption (Sakai and Trainor, 2016; Sakai et al., 2016; Warner and McIntosh, 2009).

Nucleolin is a major nucleolar protein, rRNA-processing protein and nucleotide-binding protein (Ghisolfi-Nieto et al., 1996; Ginisty et al., 1998). Among the top ten highly enriched genes in NCCs (Pijuan-Sala et al., 2019), Ncl is the only one involved in rRNA transcription. Therefore, in this study we explored the hypothesis that Nucleolin regulates rRNA transcription, NCC development and craniofacial-specific gene expression and function during embryogenesis. We discovered that Nucleolin is essential for zebrafish embryo survival and is required for NCC-derived craniofacial bone and cartilage development. Furthermore, Nucleolin regulates rRNA transcriptionally, fgf8a mRNA post-transcriptionally and the p53 protein post-translationally. Consistent with this model, we demonstrate that exogenous human recombinant FGF8 can ameliorate the cranioskeletal defects as well as recover rRNA transcription in ncl^−/− mutant embryos. Our work, therefore, has uncovered previously unreported tissue-specific functions for Nucleolin in regulating rRNA transcription and FGF signaling in osteochondroprogenitor differentiation during vertebrate embryogenesis, each of which is crucial for craniofacial development.

To understand the function of Nucleolin in vertebrate development, we characterized its protein expression during zebrafish embryogenesis (Fig. 1). Nucleolin was maternally expressed at 1.5 h post fertilization (hpf) and remained ubiquitously expressed through blastulation (3 hpf), early neurulation (12 hpf) and axial segmentation (18 hpf) (Fig. 1A-D). While at early stages, Nucleolin was present in the cytoplasm, and by 18 hpf, its expression became nuclear (Fig. 1D″). At 24 hpf, the expression of Nucleolin was still ubiquitous; however, it was enriched in the eye and the midbrain-hindbrain boundary (MHB) (Fig. 1E). In 36 and 72 hpf zebrafish embryos, elevated expression of Nucleolin was observed within the eye, pharyngeal arches and the brain (Fig. 1F,G). These expression analyses demonstrated that Nucleolin is dynamically expressed during embryogenesis with enriched expression in craniofacial tissues. Further, this suggests that Nucleolin may be required for proper craniofacial morphogenesis.

To test our hypothesis that Nucleolin is functionally required for proper craniofacial development, we characterized the phenotype of the ncl zebrafish line ncl^hi2078Tg that was generated by insertional mutagenesis of exon 1, which disrupts ncl transcription (Amsterdam et al., 2004). ncl^{hi2078T/hi2078Tg} zebrafish exhibited dramatically reduced levels of the Nucleolin protein between 18 hpf and 24 hpf (Fig. 1C′-G′), thereby establishing them as ncl^−/− mutants.

ncl^−/− embryos were phenotypically distinguishable from their wild-type (WT) siblings at 24 hpf by their smaller heads and the presence of necrosis in the craniofacial region (Fig. 2A,B). At 36 hpf, ncl^−/− embryos had a misshapen frontonasal prominence and MHB (Fig. 2C-D′). By 3 days post fertilization (dpf), ncl^−/− mutants displayed craniofacial anomalies, such as mandibular hypoplasia, in addition to a misshapen MHB (Fig. 2E-F′). At 5 dpf, ncl^−/− embryos continued to exhibit craniofacial anomalies and failed to inflate their swim bladders (Fig. S1A,A′), which collectively led to their lethality between 6 and 10 dpf (Fig. S1D,D′).

To further characterize the craniofacial defects in ncl^−/− embryos, we performed skeletal staining with Alcian Blue and Alizarin Red for cartilage and bone, respectively. At 3 dpf, ncl^−/− embryos displayed hypoplastic Meckel's and ceratohyal cartilages compared with their ncl^+/+ siblings (Fig. S1B,B′). In 5 dpf ncl^−/− embryos, the craniofacial cartilages were severely hypoplastic (Fig. 2G-M). In the neurocranium, the trabecula and the parasphenoid, which forms the base of the skull (Cubbage and Mabee, 1996), were smaller in ncl^−/− mutants compared with their ncl^+/+ siblings (Fig. 2I,J). In ncl^+/+ larvae, both edge and medial chondrocyte populations were present and were demarcated by differential Alcian Blue staining. However, in ncl^−/− larvae, cells in the ethmoid plate exhibited uniform Alcian Blue staining, suggesting that the differentiation of medial chondrocytes in the ethmoid plate may be perturbed (Fig. 2K,L). In addition, the trabecula and the parasphenoid were smaller in ncl^−/− larvae compared with ncl^+/+ larvae (Fig. 2M), whereas the length of the ethmoid plate was comparable between ncl^+/+ and ncl^−/− larvae. In the viscerocranium, the Meckel's cartilage was misshapen and the ceratohyal exhibited reverse polarity, whereas the basihyal was missing in ncl^−/− larvae. Furthermore, the posterior pharyngeal arch-derived ceratobranchials were hypoplastic (Fig. 2N,O) and osteogenesis of the teeth was incomplete in ncl^−/− larvae. At 8 dpf and 10 dpf, the polarity of the ceratohyal was similar between ncl^+/+ and ncl^−/− larvae, suggesting that ncl^−/− larvae may have delayed cranioskeletal development. However, the Meckel's cartilage remained misshaped and the ceratobranchials remained hypoplastic in 8 dpf ncl^−/− mutant zebrafish (Fig. S1B-C′ and E-F′). This establishes Nucleolin as being essential for proper cranioskeletal development, and ncl^−/− mutant zebrafish as a new model for understanding the etiology and pathogenesis of craniofacial anomalies.

Molecularly, Nucleolin binds to and modifies histones on the promoter of rDNA, and thereby regulates rRNA transcription, which is a rate-limiting step of ribosome biogenesis (Cong et al., 2012). In addition, Nucleolin interacts with the U3 small nucleolar RNA (snoRNA), and thus binds to pre-rRNA to process and splice it (Ginisty et al., 1998; Roger et al., 2003; Shao et al., 2020). Therefore, we hypothesized that ncl loss of function would lead to diminished rRNA transcription.

To validate our hypothesis, we quantified rRNA transcription by assaying for its 5′ external transcribed spacer (ETS) by quantitative reverse transcription PCR (qRT-PCR) at four different stages between 18 and 36 hpf, and observed that 5′ETS expression was significantly reduced beginning at 18 hpf, and gradually diminished at later developmental stages (Fig. S2A). We further assayed for the intergenic regions of the 47S pre-rRNA – 5′ETS and the internal transcribed spacer (ITS) 1 and ITS2 – as well as 18S rRNA at 36 hpf, and observed a significant reduction of all four amplicons in ncl^−/− embryos compared with their ncl^+/+ siblings, suggesting that Nucleolin is necessary for rRNA transcription in zebrafish (Fig. 3A). We then performed RNA immunoprecipitation using a Nucleolin-specific antibody, and observed that, at 28 hpf, Nucleolin bound to the 5′ETS and ITS1, but not to ITS2 and the 18S regions of the 47S pre-RNA (Fig. 3B), consistent with the known function of Nucleolin in vitro (Cong et al., 2012).

Decreased rRNA transcription in vitro has previously been shown to result in increased free ribosomal proteins in cells, which then interact with Mdm2, changing its conformation such that Mdm2 can no longer bind to p53 and ubiquitylate it for degradation (Donati et al., 2011). Consequently, p53 accumulates in the cell and results in p53-dependent cell death. To test whether this mechanism holds true in vivo in ncl^−/− embryos, we first performed quantitative PCR (qPCR) and western blotting to assess for p53 expression. We observed that p53 (also known as tp53) mRNA transcript levels were not significantly changed in ncl^−/− embryos compared with their ncl^+/+ siblings between 18 and 36 hpf (Fig. 3C; Fig. S2B). However, in contrast, p53 protein levels were significantly upregulated in ncl^−/− mutants compared with ncl^+/+ controls at 24 hpf, but the difference subsided by 36 hpf (Fig. 3E,I). The downstream target of p53, p21 (or cdkn1a), was initially upregulated between 24 and 30 hpf, but its expression also subsided by 36 hpf (Fig. 3C; Fig. S2C). Nucleolin regulates p53 activity by various mechanisms, including increasing p53 mRNA and/or p53 protein stability (Saxena et al., 2006; Takagi et al., 2005). We observed that, in wild-type tissues, Nucleolin did not bind to zebrafish p53 mRNA at 28 hpf (Fig. 3D). Instead, Nucleolin bound to the p53 protein (Fig. 3F).

To test our hypothesis that the Mdm2-p53 protein interaction might be reduced in ncl^−/− embryos as a result of ribosomal stress, we performed immunoprecipitation for Mdm2 in control and ncl^−/− embryos and immunoblotted for both Mdm2 and p53. Prior to examining the binding efficiency of Mdm2 and p53 in control and ncl^−/− embryos, we confirmed that Nucleolin was indeed absent in ncl^−/− embryos at 28 hpf (Fig. 3G). In our immunoprecipitation experiments, we observed decreased pulldown of Mdm2 and p53 in ncl^−/− embryos compared with control embryos, indicating that the initial accumulation of p53 was a result of reduced interaction between Mdm2 and p53 (Fig. 3H-J). The subsidence of p53 activity at later stages might therefore be due to a lack of Nucleolin-dependent stabilization of the p53 protein, which would normally increase its half-life.

To examine whether increased p53 activation and accumulation resulted in cell death and therefore necrotic craniofacial tissue in ncl^−/− mutants (Fig. 2A), we performed terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining at 24 hpf, and observed a general increase in apoptosis in ncl^−/− embryos compared with ncl^+/+ controls (Fig. 3K-L′). However, by 36 hpf, TUNEL+ cells were restricted to the MHB in ncl^−/− embryos (Fig. 3M-N′).

p53 activation can lead to increased apoptosis as well as decreased proliferation that collectively result in tissue hypoplasia. Therefore, we performed immunostaining with the G2/M phase marker phospho-histone H3 (pHH3), but observed no significant change in pHH3+ cells in 24 hpf ncl^−/− mutants compared with controls (Fig. S3A,B). Consistent with the subsidence of increased p53 and p21 expression at 36 hpf, the number of pHH3+ cells was significantly increased in ncl^−/− embryos, especially at the MHB, at 48 hpf (Fig. S3C,D,G). We also performed 5-ethynyl-2′-deoxyuridine (EdU) staining on ncl^+/+ and ncl^−/− zebrafish at 48 hpf to label cells in the S phase of the cell cycle, and observed that the mutants had a higher number of EdU+ cells (Fig. S3E,F,H). The increase in proliferation may explain how ncl^−/− embryos survived beyond 24 hpf, even though they exhibited elevated apoptosis.

We hypothesized that the increased apoptosis in ncl^−/− embryos might be p53 dependent and that genetically inhibiting p53 would suppress cell death and rescue the craniofacial anomalies in ncl^−/− embryos. Therefore, we crossed the tp53^M214K/M214K allele (hereafter referred to as p53^−/−) into the background of ncl^−/− mutant zebrafish to generate ncl^−/−;p53^−/− double mutants. Consistent with our hypothesis, TUNEL staining revealed a reduction in the number of apoptotic cells in 24 hpf ncl^−/−;p53^−/− embryos, compared with ncl^−/− embryos (Fig. 3O-P′). However, EdU labeling indicated that ncl^−/−;p53^−/− embryos had increased proliferation at 48 hpf, similar to ncl^−/− embryos (Fig. S3I,J), suggesting that the proliferation effects were not p53 dependent.

Removal of both the copies of p53 altered jaw morphology in ncl^−/− embryos (n=82) (Fig. S4). Although the ceratohyal remained smaller compared with control siblings, its polarity was restored to normal. The size of the head and the shape of the Meckel's cartilage improved in ncl^−/−;p53^−/− embryos compared with ncl^−/−;p53^+/+ larvae. However, chondrogenesis of the ceratobranchials was not improved, and in addition, chondrogenesis was disrupted in the neurocranium. ncl^−/−;p53^−/− mutant zebrafish failed to inflate their swim bladders, similar to ncl^−/− mutant zebrafish, and died between 10 and 12 dpf. This suggests that, although p53 accumulation resulted in apoptosis during early development, the skeletal differentiation defects in ncl^−/− mutant zebrafish were not p53 dependent.

In vertebrates, NCCs differentiate into most of the bones and cartilages of the craniofacial skeleton. To determine whether defects in NCC induction and migration underlay the cranioskeletal malformations in ncl^−/− mutants, we labeled NCCs by crossing sox10:egfp transgenic zebrafish into ncl^−/− mutant zebrafish. In addition, we immunostained ncl^+/+ and ncl^−/− embryos with the antibody Zn-8, which labels the endodermal pouches to demarcate individual pharyngeal arches (Trevarrow et al., 1990; Warga and Nüsslein-Volhard, 1999). sox10:egfp labeling of pre- and post-migratory NCCs and volumetric rendering of the pharyngeal arches revealed no significant change in the size of the pharyngeal arches in ncl^−/− mutant embryos compared with ncl^+/+ embryos (Fig. S5A,B). Comparisons of the sox10:egfp fluorescence intensities in the pharyngeal arches of ncl^+/+ embryos with those of the ncl^−/− embryos suggested that NCC induction and migration into the pharyngeal arches were unaffected in ncl^−/− mutant embryos (Fig. S5C). However, the pineal gland as well as the nasopharyngeal regions of ncl^−/− mutants had reduced sox10:egfp expression, suggesting that, although NCC migration into the arches was unaffected, migration to other cranial regions may have been altered. We further validated sox10;gfp transgene expression data by assessing endogenous Sox10 protein expression in ncl^+/+ and ncl^−/− embryos. The number of Sox10+ cells was similar between 36 hpf ncl^+/+ and ncl^−/− mutant embryos (Fig. S5D), confirming that NCC induction and migration into the pharyngeal arches occurred normally in ncl^−/− mutants. We observed no alteration in the formation or segregation of the endodermal pouches in 36 hpf ncl^−/− mutants, as evidenced by normal patterns of Zn-8 expression (Fig. S5A). This implied that the osteochondrogenic anomalies in ncl^−/− mutant embryos were not the result of abnormal pharyngeal pouch development.

Given the chondrogenic defects observed in ncl^−/− mutant embryos, we next investigated the basis for altered chondrogenesis in ncl^+/+ and ncl^−/− embryos. Sox9a is a transcription factor that promotes NCC differentiation into chondrocytes (Yan et al., 2002), and the mRNA levels of sox9a and its downstream target col2a1a were reduced in ncl^−/− mutant embryos at 36 hpf (Fig. 4A). The overall expression of the Sox9a protein was also reduced in ncl^−/− mutants, but most significantly in pharyngeal arches 2-5, which differentiate into the ceratohyal and ceratobranchial cartilages (Fig. 4B,B′). This provides a molecular explanation for the cartilage hypoplasia in ncl^−/− mutant embryos.

To determine the basis for the osteogenic defects in ncl^−/− mutant embryos, we evaluated the progression of osteogenesis in ncl^+/+ and ncl^−/− embryos. Runx2a and Runx2b are transcription factors that govern NCC differentiation to bone (Flores et al., 2006). runx2a mRNA expression was upregulated in ncl^−/− mutant embryos (Fig. 4D,D′), as was runx2b, but runx2b upregulation was not statistically significant at 36 hpf (Fig. 4C). We examined the protein expression of Runx2a and Runx2b in ncl^−/− and ncl^+/+ embryos by using a pan-Runx2 antibody that was designed to recognize human RUNX2 in the 300-400 amino acid region, and has 85% similarity with the zebrafish Runx2a and Runx2b proteins. Similar to the results seen for runx2a mRNA expression, Runx2 protein levels at 3 dpf were upregulated in the craniofacial region, especially in the palatoquardate and the parasphenoid (Fig. 4E,E′). Runx2 overexpression has previously been observed to result in premature osteoblast differentiation (Liu et al., 2001), which leads to a reduced pool of osteoblasts and, consequently, abnormal skeletal development. In agreement with these findings, the gene expression of the Runx2a and Runx2b downstream targets col1a2 and col10a1a was reduced in ncl^−/− embryos compared with ncl^+/+ embryos. Furthermore, we observed an increase in the expression of the early osteoblast markers spp1 (osteopontin) and sp7, and a concomitant decrease in the expression of the late osteoblast marker bglap (osteocalcin) in 36 hpf ncl^−/− mutant embryos (Fig. 4F). To confirm the presence of prematurely differentiated osteoblasts, we stained ncl^+/+ and ncl^−/− mutant embryos for alkaline phosphatase activity, which is endogenously high in primary osteoblasts and in arteries (Ohlebusch et al., 2020). At 3 dpf and 5 dpf (Fig. 4G-H′), alkaline phosphatase staining was significantly diminished in the ceratobranchials of ncl^−/− embryos compared with ncl^+/+ embryos, and overlaid the pattern of Alcian Blue staining in ncl^−/− embryos (Fig. 2O) as well as fli1a:gfp expression, which labels NCCs and endothelial cells (Askary et al., 2017). Overall, our data reveal that Nucleolin loss of function results in a reduced pool of osteochondroprogenitors that fail to mature or differentiate properly.

In mice, Fgf8 signaling upregulates chondrogenic genes such as Sox9 and Col2a1 while inhibiting osteogenic genes such as Runx2 (Xu et al., 2018; Zehentner et al., 1999). Similarly, Fgf signaling regulates sox9a in zebrafish (Esain et al., 2010), and interestingly, fgf8a mutant zebrafish have craniofacial defects comparable with those seen in ncl^−/− embryos, especially in the cartilages of the viscerocranium (Crump et al., 2004). This suggested a potential link between Nucleolin and Fgf signaling during chondrogenesis, and we therefore examined the expression of Fgf8a in ncl^+/+ and ncl^−/− embryos. We observed a significant general downregulation of Fgf8a in ncl^−/− mutant embryos (Fig. 5A), and, in examining the expression of fgf8a mRNA by in situ hybridization at 36 hpf (Fig. S6) and qPCR at four stages between 18 and 36 hpf in ncl^−/− mutant embryos, we discovered that as the expression of ncl decreased in the mutants, the expression of fgf8a also declined (Fig. 5B; Fig. S2D,E). This was suggestive of co-regulation of ncl and fgf8a expression. In silico analysis of fgf8a mRNA revealed that the fgf8a 5′ untranslated region (UTR) contains a Nucleolin consensus binding site – UCCCGA (Ghisolfi-Nieto et al., 1996). We tested for Nucleolin binding to fgf8a mRNA through RNA immunoprecipitation and observed that fgf8a mRNA was pulled down with Nucleolin, whereas that of the housekeeping gene actb was not (Fig. 5C). actb is an abundant RNA molecule and is readily available in zebrafish embryos. This suggests that Nucleolin specifically binds to fgf8a mRNA and post-transcriptionally regulates fgf8a expression.

To determine whether the phenotype of ncl^−/− mutant embryos was a direct consequence of fgf8a downregulation, we treated ncl^+/+ and ncl^−/− embryos at 18 hpf with human recombinant FGF8, which has 76% identity to the zebrafish Fgf8a protein. Compared with untreated ncl^−/− mutant embryos, we observed a considerable rescue of the craniofacial cartilage phenotype at 60 hpf (n=15/15) (Fig. S7A), the degree of which correlated with the concentration of FGF8. The same was true for 5 dpf mutants, with restoration of ceratohyal polarity at 0.25 μg/μl FGF8 (n=45/45) (Fig. 5D,E), and rescue of basihyal (n=44/45) and ceratobranchial (n=45/45) chondrogenesis at 1 μg/μl (Fig. 5F). In addition, the lengths of the trabecula and the parasphenoid were restored in FGF8-treated ncl^−/− larvae (n=45/45). However, although osteogenesis of the tooth improved, the mutant embryos were still missing their 4V¹ teeth (n=45/45), indicating that FGF8 partially rescued the ncl^−/− tooth phenotype. At 8 dpf, the phenotypes seen for the craniofacial cartilages (Fig. S7C) as well as the putative posterior swim bladder (Fig. S7B) were rescued in FGF8-treated ncl^−/− larvae. The FGF8-treated ncl^−/− fry survived at least until 15 dpf, which was 5 days longer than untreated ncl^−/− fry, and all the craniofacial skeleton elements were rescued (Fig. S7D,E). However, these fry were smaller than ncl^+/+ fry, and their anterior swim bladders failed to inflate. Nonetheless, our data indicate that exogenous FGF8 was sufficient to rescue the craniofacial skeleton defects and increase the lifespan of ncl^−/− mutants.

Fgf8 is expressed in the endoderm-derived epithelium of the pharyngeal arches and oral ectoderm, and Fgf receptors are expressed in the pharyngeal endoderm as well as in osteochondroprogenitors (Eberhart et al., 2006; Hall et al., 2006; Larbuisson et al., 2013; Stock et al., 2006; Walshe and Mason, 2003). In addition, Fgf8 has been shown to have mitogenic potential in mice (Cruz-Martinez et al., 2014; Mahmood et al., 1995; Storm et al., 2006). Therefore, FGF8 rescue of cranioskeletal defects in ncl^−/− embryos could be a result of: (1) Fgf signaling through FGF8 interactions with Fgf receptors on osteochondroprogenitors; (2) FGF8-mediated rescue of rRNA transcription; (3) increased proliferation in ncl^−/− mutants as a result of FGF8 treatment; or (4) FGF8-mediated regulation of other mesenchymal signaling pathways important for osteochondroprogenitor differentiation.

To test whether the addition of FGF8 rescued rRNA synthesis, we examined rRNA transcription in FGF8-treated ncl^−/− embryos, and observed an upregulation in 47S rRNA in these embryos (Fig. 6A). This effect could be the result of FGF8 interacting with rDNA either directly, or through Fgfr2, which has been shown to positively regulate rRNA transcription (Neben et al., 2017). Consistent with the rescued levels of rRNAs, the FGF8-treated ncl^−/− embryos also had fewer TUNEL+ cells compared with ncl^+/+ embryos (Fig. 6B), indicating that FGF-mediated rRNA transcription is important for cell survival.

To test whether this initial rescue of apoptosis in FGF8-treated ncl^−/− embryos was enough to rescue the cartilage and bone hypoplasia, we treated ncl^+/+ and ncl^−/− embryos with FGF8 at 30 hpf, which is when ncl^−/− embryos no longer exhibit high apoptotic and necrotic tissue (Fig. S8). We observed a similar rescue of the ncl^−/− embryo phenotype at 5 dpf when treatment with FGF8 began at 18 hpf and 30 hpf (Fig. S9), suggesting that exogenous FGF8 improved chondrogenesis and osteogenesis in ncl^−/− embryos.

To analyze whether FGF8 treatment rescued the cranioskeletal phenotype of ncl^−/− mutant embryos via restoration of rRNA synthesis, we treated the ncl^+/+ and ncl^−/− embryos at 18 hpf with BMH21 in concert with FGF8. BMH21 is a specific inhibitor of Pol I and dramatically reduces rRNA transcription (Peltonen et al., 2014). With BMH21 treatment alone, the ncl^+/+ embryos exhibited a similar loss of cartilage development as seen in ncl^−/− embryos (n=45/45), whereas the phenotype of BMH21-treated ncl^−/− embryos remained unchanged (n=45/45). The chondrogenic hypoplasia phenotype in BMH21-treated ncl^+/+ embryos may have been due to the loss of rRNA synthesis and p53-dependent apoptosis. However, the ncl^−/− embryos already had reduced rRNA synthesis, whereas p53 was increased transiently such that BMH21 treatment did not affect their phenotype. FGF8 plus BMH21 treatment led to severe craniofacial cartilage and bone anomalies in ncl^+/+ larvae at 5 dpf, in association with diminished rRNA transcription (n=45/45) (Fig. 6D; Fig. S8) and possibly p53-dependent apoptosis and necrosis in the presence of Nucleolin (Fig. S8C,D). However, the cranioskeletal defects in the ncl^−/− larvae were rescued in a manner similar to that seen in the larvae treated with FGF8 alone (n=45/45) (Fig. 6D), suggesting that this mechanistically occurs independent of rRNA transcription.

It is well known that the Fgf and Bmp signaling pathways interact synergistically during craniofacial development (Lovely et al., 2016). More specifically, Fgf8 and Bmp2 regulate osteochondroprogenitor differentiation by positively regulating Sox9 expression (Chen et al., 2019; Dash and Trainor, 2020; Shao et al., 2015). Therefore, we examined the expression of bmp2 mRNA in untreated and FGF8-treated ncl^−/− embryos at 30 hpf. We observed that, whereas bmp2 mRNA expression was downregulated in untreated ncl^−/− embryos, its expression was increased in FGF8-treated ncl^+/+ and ncl^−/− embryos, suggesting that exogenous FGF8 stimulates Bmp signaling in osteochondroprogenitors (Fig. 6C).

Because bmp2 expression was downregulated in ncl^−/− embryos, but was restored upon FGF8 treatment alone, we hypothesized that the FGF8-based rescue occurs via BMP2 signaling. To test this hypothesis, we treated the ncl^+/+ and ncl^−/− embryos with BMP2 alone, and observed a rescue of craniofacial cartilage and bone in ncl^−/− larvae, similar to that seen in embryos with FGF8 treatment alone (n=45/45) (Fig. 6D). Furthermore, we quantified rRNA transcription in BMP2-treated embryos and observed downregulation of the 5′ETS and 18S, similar to what was seen in DMSO-treated ncl^−/− embryos at 28 hpf (Fig. S10).

Taken together, our data suggest that Nucleolin regulates craniofacial development through two different pathways, one by controlling rRNA transcription, which appears to be important for cell survival, and the other by post-transcriptionally regulating FGF8 signaling during NCC-derived osteochondroprogenitor differentiation.

Nucleolin is the most abundant phosphoprotein in the nucleolus (Tajrishi et al., 2011); however, its activity and localization are not limited to the nucleolus. Nucleolin has been observed in the nucleus, cytoplasm and plasma membrane, where, depending on the cell type and the environmental condition, Nucleolin performs various functions including DNA replication and repair (Kobayashi et al., 2012), chromatin remodeling (Cong et al., 2012), rRNA transcription and processing (Ginisty et al., 1998; Roger et al., 2003), mRNA turnover and translation (Abdelmohsen et al., 2011; Chen et al., 2000; Fähling et al., 2005; Jiang et al., 2006; Takagi et al., 2005; Zaidi and Malter, 1995; Zhang et al., 2006, 2010) and viral entry and replication (Bose et al., 2004; Izumi et al., 2001; Nisole et al., 2002). Although the cellular and molecular functions of Nucleolin have been previously studied in vitro, its role in vertebrate development remains to be explored. Here, we show that Nucleolin is crucial for zebrafish craniofacial development, and specifically for proper differentiation of NCCs into cartilage and bone. Consistent with our findings, Ncl knockdown in Xenopus has also been shown to result in craniofacial cartilage hypoplasia (Delhermite et al., 2022).

Similar to other rRNA modifying proteins such as Nol11, Wrd43 and Fibrillarin, the absence of which results in chondrogenesis defects in zebrafish and frog (Bouffard et al., 2018; Griffin et al., 2015; Zhao et al., 2014), it is interesting that perturbation of a global process such as ribosome biogenesis results in craniofacial defects. One possible reason for this tissue specificity could be that the genes required for rRNA transcription, such as those encoding subunits of RNA Pol I, and fbl and ncl, are highly expressed in NCCs and craniofacial tissues, which makes them more susceptible to disruptions in rRNA transcription (Bouffard et al., 2018; Watt et al., 2016, 2018; Weaver et al., 2015). rRNA transcription is essential not only because rRNA forms the catalytic core of ribosomes, but also because, in its absence, cells undergo proteotoxic stress as a result of ribosomal protein accumulation (Albert et al., 2019; Tye et al., 2019). Under such circumstances, it has been proposed that free ribosomal proteins bind to Mdm2, resulting in the accumulation of p53 and, consequently, cell cycle arrest and cell death (Donati et al., 2011; Lohrum et al., 2003; Scala et al., 2016; Zhang et al., 2003) (Fig. 7A,B).

Contradicting previous in vitro studies in human cell lines (Chen et al., 2012), we observed that Nucleolin does not bind to p53 mRNA in vivo in zebrafish, most likely owing to the differences in the 5′ and 3′ UTR sequences of human and zebrafish p53. However, the Nucleolin and p53 proteins bind to each other, suggesting that Nucleolin is required to stabilize the p53 protein (Saxena et al., 2006). In the absence of Nucleolin, p53 accumulates temporarily and apoptosis is initiated, however, the effect is not sustained, and it does not result in lethality. This is consistent with an initial upregulation of p21 mRNA expression from 18 to 24 hpf, followed by a steep decrease by 36 hpf, and is indicative of a momentary surge in the transcriptional activity of p53. p21 is a cell cycle inhibitor, which, when significantly reduced, releases cell cycle arrest (Gartel and Radhakrishnan, 2005; Hemmati et al., 2005), resulting in higher proliferation in ncl^−/− embryos compared with ncl^+/+ embryos. The increase in proliferation in the absence of Nucleolin is contrary to published literature, which suggests that Nucleolin promotes proliferation in various cell types (Jiang et al., 2015; Ugrinova et al., 2007). This indicates that a different mechanism regulates proliferation in the absence of Nucleolin in mammalian cells and developing zebrafish. Over time, apoptosis decreases, and proliferation increases in ncl^−/− embryos, which results in embryonic survival until 5 dpf.

Given the ubiquitous expression of Nucleolin in early zebrafish embryos and its requirement in rRNA synthesis, which is essential for ribosome biogenesis and protein translation in all cell and tissue types, we hypothesize that Nucleolin is crucially required during the cleavage, gastrulation and segmentation stages. However, the maternal loading of Nucleolin in the ncl^−/− embryos until 18 hpf is sufficient for survival of the mutant embryos and for the NCCs to form and migrate to the pharyngeal arches. Further studies are required to understand the function of Nucleolin in early NCC development by either generating maternal zygotic ncl mutants or establishing an NCC-specific ncl conditional deletion mutant. From our current analysis, however, we can conclude that Nucleolin is required for NCC differentiation into chondrocytes and osteocytes in the zebrafish craniofacial region.

Defects in cartilage and bone formation, such as the cartilage hypoplasia, inverted ceratohyal and underdeveloped ceratobranchials, which are observed in ncl^−/− mutants, are common phenotypes typically associated with defects in NCC formation and migration. However, in ncl^−/− mutants, NCC formation and migration to pharyngeal arches appears to be normal, based on lineage-labeling with sox10:egfp as well as endogenous Sox10 expression. Instead, the chondrogenic hypoplasia phenotype in ncl^−/− mutants can be linked to sox9a downregulation, whereas the osteogenic defects arise due to the upregulation of Runx2. This leads to premature osteoblast differentiation, which reduces the pool of cells that can subsequently differentiate into osteoblasts. Thus, early osteoblast markers such as spp1 and sp7 are more highly expressed in ncl^−/− embryos compared with their ncl^+/+ siblings, while late osteoblast markers such as bglap are downregulated. However, both sox9a and runx2 are indirect targets of Nucleolin based on in silico analysis of their 5′ and 3′ UTRs, which do not contain a Nucleolin consensus binding site. fgf8a, which is also downregulated in ncl^−/− mutants and is required for osteochondroprogenitor differentiation, does contain a Nucleolin-binding site in its 5′ UTR. The 5′ UTR of an mRNA is responsible for both stability and translation of mRNA (Dash et al., 2016). An RNA-binding protein binding to the 5′ UTR of an mRNA has two probable functions: (1) it can stabilize/degrade mRNA; or (2) it can promote/hinder protein translation. Combined with our data that fgf8a mRNA decreases over time in the ncl^−/− mutants, this suggests that fgf8a is post-transcriptionally regulated by Nucleolin, which results in reduced translation, indicating that Nucleolin must stabilize fgf8a mRNA. In addition, the maternal expression of Nucleolin in zebrafish until 18 hpf possibly prevents early embryonic phenotypes such as agenesis of the cerebellum and MHB organizer, which have previously been associated with fgf8a downregulation (Reifers et al., 1998). This is further corroborated by the restoration of osteogenesis and chondrogenesis in ncl^−/− mutants with exogenous FGF8, suggesting that the crucial time for Fgf8a function in ncl^−/− zebrafish is around 18 hpf. FGF8 treatment increased the survival of ncl^−/− larvae by 5 days, implying continued treatment may further augment lifespan, and this will be tested in the future.

In mice, Fgf8 has been shown to preferentially bind to Fgfr1 (Mott et al., 2010). However, in zebrafish, Fgf receptors are functionally redundant with respect to Fgf8a ligand binding (Leerberg et al., 2019). Therefore, it is likely that Fgf8a could activate Fgfr2, which is known to bind to the promoter of rDNA, resulting in histone modification and activation of rRNA transcription, similar to Nucleolin (Neben et al., 2017). This suggests that, upon FGF8 treatment, either FGF8 or Fgfr2 translocate to the nucleolus and change the state of the rDNA chromatin from closed to open. However, given that the upstream core elements of rDNA, to which RNA Pol I and UBTF bind, are as yet unannotated in zebrafish, and further work is required to test our hypothesis. This will involve identifying the upstream core elements using a combination of assays for transposase-accessible chromatin using sequencing (ATAC-seq) and chromatin immunoprecipitation followed by sequencing (ChIP-seq), followed by elucidating the mechanism of Fgf8 and Fgfr2 regulation of rRNA transcription in the absence of Nucleolin.

Exogenous FGF8 restores rRNA transcription in ncl^−/− mutants, but the rescue of cranioskeletal defects in ncl^−/− mutants by FGF8 can still occur in the absence of rDNA transcription, as evidenced by BMH21 treatment, possibly because the ribosomal stress-induced p53 response is not sustained in ncl^−/− embryos. This suggests that the effect of FGF8 on osteochondroprogenitor differentiation is mechanistically independent of rRNA synthesis and occurs via its downstream regulation of Bmp2 signaling, which was corroborated by the BMP2-mediated rescue of the cranioskeletal anomalies in ncl^−/− mutants.

Interestingly, Fgf8 has been previously shown to regulate variance in facial shape in a dose-dependent manner (Green et al., 2017), and rDNA transcription is known to be essential for craniofacial development (Terrazas et al., 2017; Trainor and Merrill, 2014) and variation (Claes et al., 2018). Therefore, as an upstream regulator of fgf8, and with a role in rDNA transcription, Nucleolin, by extrapolation, may also play an important role in determining facial shape. Overall, our work has uncovered that Nucleolin regulates Fgf8 signaling as well as rRNA transcription, making craniofacial development especially susceptible to Nucleolin loss of function.

Adult zebrafish (Danio rerio) were housed and maintained in the Stowers Institute Zebrafish Facility according to an Institutional Animal Care and Use Committee (IACUC)-approved protocol (#2021-124). Zebrafish embryos were raised at 28.5°C and staged using standard procedures (Kimmel et al., 1995). To prevent pigment development for immunostaining experiments, 0.002% 1-phenyl-2-thiourea was added to the embryo media. ncl^hi2078Tg zebrafish were obtained through the Zebrafish International Resource Center (ZIRC) and maintained as heterozygotes on the AB/TU background and incrossed to generate homozygous mutant embryos. PCR was used to detect for the presence or absence of the insertional mutation. The ncl wild-type allele was detected using the following primers: forward, 5′-TTACATGTGGTGAGAAGGCCC-3′, and reverse, 5′-AACACCTCCCCTGGGTTTAT-3′. The ncl mutant allele was detected using the following primers: forward, 5′-TTACATGTGGTGAGAAGGCCC-3′, and reverse, 5′-GCTAGCTTGCCAAACCTACAGGT-3′. The ncl heterozygous mutant lines were crossed with the reporter line Tg(7.2kb-sox10:gfp), referred to as sox10:egfp, as well as with the tp53^M214K line.

Embryos were anesthetized with MS-222 (Sigma-Aldrich, A5040) and mounted in 2% methyl cellulose while submerged in E2 media. Embryos were imaged using a Leica MZ16 microscope equipped with a Nikon DS-Ri1 camera and NIS Elements BR 3.2 imaging software. When appropriate, manual z-stacks were taken and the images were assembled using Helicon Focus software.

Alcian Blue staining and Alizarin Red staining to label the cartilage and bone, respectively, were performed according to Walker and Kimmel (2007). Embryos were cleared in glycerol and potassium hydroxide and dissected for neurocranium and viscerocranium images. Imaging was performed as described above. Length measurements were performed using ImageJ.

Whole-mount immunostaining was performed according to standard protocols (Westerfield, 2000) using primary antibodies against Nucleolin (1:500, Abcam, ab22758), GFP (1:500, Life Technologies, A6455), Zn-8 (1:250, Developmental Studies Hybridoma Bank), Sox10 (1:500, GeneTex, GTX128374), Sox9a (1:500, GeneTex, GTX128370), Runx2 (1:500, Abcam, ab23981), Fgf8a (1:500, GeneTex, GTX128126), pHH3 (1:2000, Millipore, 06-570). Fluorescent secondary antibodies, either Alexa Fluor 488 or Alexa Fluor 546 (1:500, Invitrogen) were used for detection. The TUNEL and EdU Click-IT assays were performed according to the manufacturer's instructions with slight modifications. Embryos were incubated for 1 h on ice and 1 h at 37°C in the reaction buffer for TUNEL assay. Embryos were imaged using a Zeiss upright 700 confocal microscope, and images were captured and processed using Zen software. ImageJ software was used to quantify the fluorescence intensity and area. The pHH3 and EdU+ cells quantified were in the craniofacial region between the frontonasal prominence and the otic vesicle.

For section staining, embryos were collected at 18 hpf and fixed with 4% paraformaldehyde overnight, followed by equilibration in 30% sucrose overnight. The embryos were then embedded in tissue-freezing media and sectioned at 10 μm thickness. The sections were then immunostained for Nucleolin (1:500, Abcam, ab22758) and imaged using a Zeiss upright 700 confocal microscope.

In situ hybridization was performed using probes against runx2a and fgf8a according to standard protocols. Briefly, the embryos were treated with 10 µg/µl proteinase K and hybridized in the probes diluted to 1 ng/μl overnight at 68°C. Following hybridization, the embryos were washed, blocked and incubated with anti-digoxigenin-AP Fab fragments (1:5000, Roche, 11093274910). Signals were detected using nitro blue tetrazolium (NBT)/5-bromo-4-chloro-3-indolyl phosphate (BCIP) and the embryos were imaged using a Nikon DS-Ri1 camera.

Fixed embryos were washed in TBS and NTMT buffer (100 mM NaCl, 100 mM Tris-HCl pH 9.5, 50 mM MgCl₂, 1% Tween 20), followed by incubation with NBT (3.5 µl) and BCIP (5 µl) in NTMT for the desired color revelation time. The images were collected using a Nikon DS-Ri1 camera.

Protein samples consisting of five fish/sample were collected at appropriate stages. Embryos were homogenized and suspended in sample buffer containing Tris-HCl pH 8.0, NaCl, SDS, sodium deoxycholate, NP-40 and protease inhibitor and used for western blotting according to standard protocols (Dash et al., 2021). Protein quantity was estimated via a bicinchoninic acid (BCA) assay. The primary antibodies used were: γ-tubulin (1:1000, Sigma-Aldrich, T6557), α-tubulin (1:10,000, Sigma-Aldrich, T5168), p53 (1:500, Cell Signaling Technology, 2524S) and Nucleolin (1:500, Abcam, ab22758). Western blots were imaged and quantified using a CLx-Scanner (Li-COR) and Odyssey Software. For quantification, band intensities for p53 were compared with the housekeeping control protein γ-tubulin. Two-tailed, paired Student's t-test was performed for statistical analysis.

Protein lysates from 28 hpf control and ncl^−/− zebrafish were used (n=25 per biological replicate; total three biological replicates) for pre-conjugation with the p53 antibody (Cell Signaling Technology, 2524S, 1:100) and IgG with magnetic beads. The same number of embryos were used for each group. Equal volumes of the immunoprecipitates were then loaded on two SDS-PAGE gels and transferred to polyvinylidene difluoride (PVDF) membranes to immunoblot for Mdm2 (1:1000, Cell Signaling Technology, 86934) and p53 (1:1000, Cell Signaling Technology, 2524) separately. Because the embryos were grouped and lysed immediately after collection, genotyping by conventional methods could not be performed. Instead, the lysate was used in a western blot for Nucleolin to substitute as genotyping. Following immunoprecipitation, the samples were used in western blotting and immunoblotted for both Mdm2 and p53.

For sample collection, embryos were genotyped individually using the tip of the tail tissue at the required developmental stage. After the identification of mutants and wild types from a clutch, ten embryos of the same genotype were pooled together for RNA isolation (n=1). RNA was collected from zebrafish embryos using the Qiagen miRNeasy Micro Kit and tested for quality on an Agilent 2100 Bioanalyzer. The Superscript III kit (Invitrogen) was used to synthesize cDNA for qPCR using random hexamer primers. The following primers were used: ncl forward, 5′-ATATCGAGGGCAGGAGTATT-3′, and reverse, 5′-GTTTTCGTAGGTCCAGAGTT-3′; tp53 forward, 5′-CGAGCCACTGCCATCTATAAG-3′, and reverse, 5′-TGCCCTCCACTCTTATCAAATG-3′; p21 forward, 5′-GACCAACATCACAGATTTCTAC-3′, and reverse, 5′-TGTCAATAACGCTGCTACG-3′; sox9a forward, 5′-GGAGCTCAGCAAAACTCTGG-3′, and reverse, 5′-AGTCGGGGTGATCTTTCTTG-3′; col2a1 forward, 5′-GCGACTTTCACCCCTTAGGA-3′, and reverse, 5′-TGCATACTGCTGGCCATCTT-3′; runx2a forward, 5′-AACTTTCTGTGCTCGGTGCT-3′, and reverse, 5′-AACTTTCTGTGCTCGGTGCT-3′; runx2b forward 5′-CAAACACCCAGACCCTCACT-3′, and reverse, 5′-GTATGACCATGGTGGGGAAG-3′; col1a2 forward, 5′-CTGGCATGAAGGGACACAG-3′, and reverse, 5′-GGGGTTCCATTTGATCCAG-3′; col10a1a forward, 5′-CCTGTCTGGCTCATACCACA-3′, and reverse, 5′-AAGGCCACCAGGAGAAGAAG-3′; fgf8a forward, 5′-GCCGTAGACTAATCCGGACC-3′, and reverse, 5′-TTGTTGGCCAGAACTTGCAC-3′; bglap forward, 5′-TGAGTGCTGCAGAATCTCCTAA-3′, and reverse, 5′-GTCAGGTCTCCAGGTGCAGT-3′; spp1 forward, 5′-TGAAACAGATGAGAAGGAAGAGG-3′, and reverse, 5′-GGGTAGCCCAAACTGTCTCC-3′; sp7 forward, 5′-GGATACGCCGCTGGGTCTA-3′, and reverse, 5′-TCCTGACAATTCGGGCAATC-3′; bmp2 forward, 5′-AGCTTCCACCATGATGAATCTACA-3′, and reverse, 5′-TCAGGTTGAAGAGGAACCGC-3′; actb forward, 5′-TTCCTTCCTGGGTATGGAATC-3′, and reverse, 5′-GCACTGTGTTGGCATACAGG-3′; and canx forward, 5′-ACGATACCGCAGAGAATGGAGACA-3′, and reverse, 5′-TCCTGTTTCTGGGAGACCTCCTCA-3′. Previously published rRNA primer sequences were used for qPCR (Azuma et al., 2006). Power Sybr (Life Technologies) reaction mix and the ABI 7900HT real time PCR cycler was used to measure cDNA amplification. Three biological replicates were run in technical triplicate for each experiment. Samples without the template and without reverse transcriptase were run as negative controls.

Lysates from 28 hpf wild-type zebrafish were used (n=100 per biological replicate; total three biological replicates) for pre-conjugation with the Nucleolin antibody (Abcam, ab22758, 1:100) and IgG with magnetic beads. Immunoprecipitation was performed using the manufacturer's instructions for the EZ-Nuclear RIP Kit (EMD Millipore, 17-10523), followed by cDNA synthesis and qRT-PCR for 5′ETS, ITS1, ITS2, 18S rRNA, fgf8a, p53 and actb.

ncl^+/+ and ncl^−/− embryos were treated with 0.0625, 0.25 and 1 ng/μl human recombinant FGF8 (Thermo Fisher Scientific, PHG0184) as well as 1 ng/μl FGF8 along with 2 ng/μl BMH21 (Sigma-Aldrich, SML1183) and 1 μg/μl BMP2 (Stemcell Technologies, 78004) diluted in E2 media by immersion starting at 18 hpf. In addition, ncl^+/+ and ncl^−/− embryos were treated with 1 ng/μl FGF8 by immersion beginning at 30 hpf. The drugs in the media were replaced every 24 h until 5 dpf. The FGF8-treated larvae were added to the tank system and allowed to develop until 15 dpf.

For brightfield imaging and skeletal stain, 15 embryos of each genotype for a particular stage were assayed from one clutch of embryos, and this was repeated three times to ensure adequate power for statistical analysis. For immunostaining, western blotting and drug treatments, five embryos per genotype from a clutch were assayed and analyzed. For qPCR, immunoprecipitation and RNA immunoprecipitation, 10, 25 and 100 embryos per clutch were used, respectively. All experiments were repeated three times to ensure reproducibility. The embryos were chosen at random from a clutch size of at least 200 embryos. None of the data collected were excluded from analysis. For phenotypic analysis, two-tailed, unpaired t-test with Welch's modification was performed. For molecular quantifications (western blot, RNA immunoprecipitation and immunoprecipitation), two-tailed, paired t-test was used. Two-way ANOVA was performed for statistical analysis of qPCR data. P-values for all experiments are provided in Table S1.

We are thankful to all members of the Trainor lab for their input and suggestions during the course of this work. We thank Dr Tom Schilling for the gift of the sox10:egfp zebrafish, Dr Tatjana Piotrowski for the gift of the tp53 zebrafish as well as in situ hybridization probes and the Stowers Institute for Medical Research (SIMR) Aquatics staff, especially Carrie Carmichael for zebrafish care and maintenance. We also thank Mark Miller for illustrating Fig. 7 and Dr Cathy McKinney for imaging support.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200349.