FGFR2 directs inhibition of WNT signaling to regulate anterior fontanelle closure during skull development

The calvarial bones of the infant skull are linked by transient fibrous joints called sutures and fontanelles. Sutures, which join two adjacent calvarial bones, and fontanelles, which occupy large, unossified regions between multiple bones where sutures will eventually form, are crucial for calvarial reshaping during birth and calvarial expansion during postnatal brain growth. Numerous genetic disorders present with craniofacial deformities when sutures fuse prematurely (craniosynostosis), or fontanelles remain patent. For example, syndromes associated with genetic variants in fibroblast growth factor receptor 2 (FGFR2), including Apert, Pfeiffer, Crouzon, and bent bone dysplasia, often present with multi-suture craniosynostosis, as well as a persistently open anterior fontanelle (AF) (Azoury et al., 2017; Merrill et al., 2012; Reardon et al., 1994; Wilkie et al., 1995). The co-occurrence of these seemingly opposite calvarial joint phenotypes suggests important differences in the way FGF signaling regulates development of sutures versus fontanelles.

The AF, or a baby's ‘soft spot’, is a broad region of fibrous connective tissue at the apex of the developing calvaria where the paired frontal bones will ultimately meet the parietal bones. During postnatal development, the frontal bones undergo rapid apical expansion though intramembranous ossification, replacing the AF with the posterior frontal suture (PFS), which connects the frontal bones from the bregma to the jugum limitans. As the PFS forms, it differentiates into two distinct bony layers, with the endocranial layer undergoing fusion through endochondral-like ossification that requires Sox9 and local inhibition of WNT signaling (Bradley et al., 1996; Behr et al., 2013; Manzanares et al., 1988; Sahar et al., 2005). The fate of the AF connective tissue and its role in frontal suture development, as well as its potential contribution to the etiology of calvarial defects, has remained unclear.

While calvarial phenotypes in FGFR2-related syndromes indicate a key role for FGF signaling in the developing AF, studies to date have largely focused on the role of the receptor in bone formation during embryogenesis when it is expressed within the mid-suture and advancing osteogenic bone fronts. Genetic studies in mice have shown that FGFR2, likely activated by FGF18, is crucial for maintaining the balance of proliferation and osteoblast differentiation within sutures (Ohbayashi et al., 2002). In mouse models that harbor gain-of-function Fgfr2 variants associated with human craniosynostosis syndromes, mechanisms of premature fusion of the coronal suture include altered proliferation and induction of ectopic osteoblast differentiation (Holmes et al., 2009; Wang et al., 2005; Yin et al., 2008; Pfaff et al., 2016; Hajihosseini et al., 2001). These mouse models also exhibit a midline gap between the frontal bones in a region anatomically homologous to the human AF. Studies using the Apert syndrome mouse model have shown that this midline gap is linked to decreased osteoblast differentiation, as well as transcriptomic changes in the cells within the osteogenic fronts of the frontal bones (Wang et al., 2005; Holmes et al., 2020). A midline ossification defect was also reported in mice with conditional loss of Fgfr2 in skeletogenic mesenchyme, but the mechanism underlying the phenotype has yet to be investigated (Yu et al., 2003).

Here, we show that Fgfr2 is necessary for AF closure and, subsequently, frontal suture joint formation. We find that cells of the AF express the tendon/ligament marker Scx and represent at least two transcriptionally distinct populations spatially restricted to either the ecto- or endocranial domains, which give rise to bone and cartilage, respectively, in an FGF-dependent manner. We provide evidence that FGF signaling in the osteogenic fronts of the frontal bone promotes expression of the secreted WNT inhibitor Wif1, which non-autonomously regulates WNT signaling in AF cells. In the absence of FGF signaling, Wif1 expression is lost, and cells of the AF fail to differentiate into bone and cartilage, instead maintaining a connective tissue-like identity. Correspondingly, elevation of WNT/β-catenin signaling induces a persistent AF phenotype, and downregulation of WNT/β-catenin rescues AF closure in Fgfr2 mutant mice. Together, these results suggest that FGF-mediated regulation of WNT signaling regionally influences cell fate and suture formation within the AF.

Genetic lineage analysis in mice using Wnt1-Cre previously showed that neural crest-derived mesenchyme gives rise to the frontal bones, as well as the tissue that occupies the AF (Fig. 1A) (Jiang et al., 2002; Yoshida et al., 2008). Therefore, to investigate a role of FGFR2 within the AF development, we performed Wnt1-Cre-mediated deletion of Fgfr2 using a floxed allele that produces a functional null copy of the receptor (Yu et al., 2003). We examined calvarial bone and cartilage development in Wnt1-Cre;Fgfr2^−/− and littermate controls using wholemount Alizarin Red and Alcian Blue staining. The AF of Wnt1-Cre;Fgfr2^−/− mice was wide and patent beginning at embryonic day (E) 18.5 compared to control (Fig. 1B,E). In control mice at postnatal day (P) 3, the AF began to progressively close, whereas that of Wnt1-Cre;Fgfr2^−/− mice remained wide open (Fig. 1C,F). By P5, the frontal bones of control mice had closely approximated to form the PFS, whereas the AF remained persistent in the mutant (Fig. 1D,G). Quantitative measurement showed that the normalized average area between the frontal bones (total area of AF/length of PFS from jugum limitans to bregma) was significantly larger in Wnt1-Cre;Fgfr2^−/− at P3 (P=0.0167, n=7) and P5 (P=0.0003, n=9) compared to littermate controls (Fig. 1H). By P10, the PFS of control mice began to form cartilaginous fusions that follow a joint-gap pattern (Fig. 1I, arrowheads), whereas the AF failed to resolve into the PFS in Wnt1-Cre;Fgfr2^−/− mice (Fig. 1L). Micro-computed tomography (μCT) analysis showed that the PFS of control mice fused by P30, with distinct endocranial and ectocranial layers visible in orthoslice views (Fig. 1J,K). In Wnt1-Cre;Fgfr2^−/− mice, the AF persisted, and regions of the frontal and sagittal sutures that it prefigures failed to form (Fig. 1M,N). Together, these results show that Fgfr2 is necessary for AF closure.

The posterior frontal suture (PFS), which forms within the AF, undergoes normal fusion in the endocranial domain through a cartilage intermediate. To identify histogenic changes in PFS development, Hall-Brunt Quadruple stain (HBQ) was used to distinguish bone and cartilage (Hall, 1986). In control mice at P5, the paired frontal bones approximated in the ectocranial domain and secondary osteogenic fronts formed within the endocranial domain (Fig. 2A). The frontal bones of Wnt1-Cre;Fgfr2^−/− mice, however, remained far apart and the secondary osteogenic fronts within the endocranial domain were missing (Fig. 2B). In control mice at P7, the cartilage template that prefigures the fused endocranial layer of the PFS began to condense, becoming more differentiated at P11 and P13 (Fig. 2C,E,G). Conversely, the frontal bones of Wnt1-Cre;Fgfr2^−/− mice remained separated and showed no evidence of cartilage formation in the endocranial domain (Fig. 2D,F,H). Together, these results show that Fgfr2 is necessary for normal formation and fusion of the PFS within the AF.

To better understand Fgfr2⁺ cell populations in the developing AF, we analyzed a previously published single-cell RNA sequencing (scRNA-seq) dataset of the developing frontal suture at E18.5, which included the paired osteogenic fronts of the frontal bones and intervening AF connective tissue (FaceBase dataset 1-4TT6; Holmes et al., 2020). Re-analysis of this dataset using Seurat identified nine mesenchymal clusters with similar gene enrichment to those previously reported (Fig. 3A,C) (Holmes et al., 2020). Violin plots showed enriched Fgfr2 expression in cell clusters we designated AF1 and AF2 based on their expression of genes associated with ligament-like connective tissue, including Scx, Tnmd, Mkx, and Matn4 (Fig. 3B). Interestingly, AF2 was distinct from AF1 based on its enriched expression of the chondrogenic gene Sox9 and tenogenic gene Egr1 (Fig. 3B). This is consistent with a previous study that showed Sox9 to be indispensable for closure of the PFS through endochondral-like ossification (Sahar et al., 2005). Fgfr2 expression was most enriched in the osteogenic populations, including the osteoblasts (OB) cluster, marked by Ibsp expression, and the osteogenic front (OF) 1 and 2 clusters, which expressed pro-osteogenic genes Npnt and Sp7 (Fig. 3B). Interestingly, Sp7 has been implicated in regulating cellular recruitment at advancing bone fronts (Kague et al., 2016; Wang et al., 2022). Compared to OF1, the OF2 cluster was enriched for AF2-related genes including Sox9 and Egr1. Relatively lower levels of Fgfr2 were expressed in the dura mater (DM) cluster, which was marked by expression of Cxcl12 (Fig. 3B) (Holmes et al., 2020). While Fgfr2 expression was not enriched in the ectocranial mesenchyme (EM) cluster, marked by C1qtnf3 and Igfbp3, or the hypodermis (HD) cluster, marked by Clec3b, these clusters expressed the FGFR2 ligand Fgf18 (Fig. 3B) (Farmer et al., 2021). Fgf9 and Fgf2, the only other FGF ligands detected in the dataset, were also enriched in the EM cluster. RNAScope fluorescent in situ hybridization at E18.5 validated that the spatial distribution of Fgfr2-IIIc (the mesenchymal-specific variant of Fgfr2) expression overlapped with expression domains for the AF1 and AF2 marker Tnmd and the OF1 and OF2 marker Sp7 (Fig. 3D,E). Pseudotime analysis performed with Monocle 3 predicted that the AF1 and AF2 clusters follow distinct developmental trajectories to give rise to OF1 and OF2, respectively, and converge on an osteogenic identity (Fig. 3F) (Cao et al., 2019). This likely represents the distinct contribution of Sox9⁺ AF2 cells to endochondral-like bone that forms within the endocranial domain of the PFS. Cellular communication modeling using CellChat indicated that FGFs expressed in the EM signal to FGFR-expressing cells in the AF and OF populations (Fig. 3G,H) (Jin et al., 2021).

To determine how loss of Fgfr2 impacts the cell populations identified by scRNA-seq, we examined changes in cluster-enriched gene expression in the developing AF at E18.5 using RNAScope in situ hybridization. In the control, Scx, which marks the AF1 and AF2 clusters, was expressed in the AF connective tissue between the bone fronts with highest expression in the ectocranial region (Fig. 4A). The expression domain of Scx expression was expanded in Wnt1-Cre;Fgfr2^−/− littermates, likely indicative of the increased size of the AF in mutants (Fig. 4B, asterisks). An expanded domain of Scx-GFP expression was also seen at this stage in Wnt1-Cre;Fgfr2^−/− compared to controls (Fig. 4C,D, asterisks). Sp7, a marker for the OF1 and OF2 clusters, was expressed in osteogenic fronts of controls and reduced in Wnt1-Cre;Fgfr2^−/− littermates (Fig. 4E,F,G,H). In control, Sox9, a marker for the AF2 and OF2 clusters, was expressed in the endocranial region of the AF and within the osteogenic fronts (Fig. 4E,E′). Expression of Sox9 was reduced in the AF and osteogenic fronts of Wnt1-Cre;Fgfr2^−/− littermates (Fig. 4F,F′, arrowhead). By contrast, no difference in the expression of Fgf18, a marker for the EM cluster, was detected (Fig. 4G,H). Together, these results provide evidence that transcriptionally distinct populations of AF cells occupy spatially separate domains, and their markers are differentially impacted by loss of Fgfr2.

Very little is known about the AF connective tissue cells and their role in frontal suture formation. To observe AF connective tissue cells during calvarial development, we employed a Scx-GFP transgenic reporter mouse that has been previously used to study tendon and ligament development (Pryce et al., 2007). At E18.5, Scx-GFP was expressed in all calvarial sutures and fontanelles, as well as in the immediately adjacent bone (Fig. S1A,C). At P3, when the frontal suture had largely replaced the AF, Scx-GFP expression was confined to the suture connective tissue (Fig. S1B,D). Sections through the AF at these stages revealed a bi-layered expressional pattern of Scx-GFP representing the ecto- and endocranial layers of the AF at E18.5 and P3, as well as some GFP⁺ cells observed within the frontal bones at P3 (Fig. S1E,F). At P3, Scx-GFP expression in the ectocranial layer likely marked the developing suprasutural ligament.

We next examined osteogenic differentiation within the AF of Wnt1-Cre;Fgfr2^−/− mice using immunofluorescent detection of RUNX2, an early marker of osteoblasts, along with Scx-GFP to discriminate the AF connective tissue. At P0, RUNX2⁺ cells were enriched within the osteogenic fronts of the paired frontal bones, and Scx-GFP expression was localized to cells in the ecto- and endocranial layers of the AF in both control and Wnt1-Cre;Fgfr2^−/− mice (Fig. 5A,B). In control mice at P3, RUNX2 expression was activated in SCX⁺ cells located in the endocranial layer of the AF connective tissue, and SCX⁺/RUNX2⁺ double-positive cells were found in the advancing frontal bones (Fig. 5C, arrowheads). SCX⁺/RUNX2⁺ cells failed to form in the endocranial layer of Wnt1-Cre;Fgfr2^−/− mice, remaining exclusively SCX⁺ within the AF (Fig. 5D, asterisks). The ectocranial layer of SCX⁺ cells spanning and overlaying the frontal bones at this stage remained similar between control and mutant mice (Fig. 5C,D, arrows). In controls at P5, RUNX2⁺ cells coalesced at the approximating endocranial osteogenic fronts to form the PFS, and SCX⁺/RUNX2⁺ cells showed contribution to the approximating bone fronts (Fig. 5E, arrowheads). Meanwhile, a portion of the ectocranial layer in control mice remained occupied by SCX⁺ connective tissue that will become the suprasutural ligament (Fig. 5E, arrow). Conversely, in Wnt1-Cre;Fgfr2^−/− mice the AF remained patent and very few SCX⁺/RUNX2⁺ were detected in the osteogenic fronts or bones (Fig. 5F, asterisks).

To detect chondrogenic differentiation, we performed immunofluorescent detection of SOX9, an early marker for chondrocytes. SOX9 was not detected in the PFS at P5 in control or Wnt1-Cre;Fgfr2^−/− mice (Fig. 5G,H), but a population of SOX9⁺ cells appeared in the endocranial region of the suture midline in controls by P7 (Fig. 5I). At this stage, the SCX⁺ suprasutural ligament was clearly distinguishable (Fig. 5I, arrow). In Wnt1-Cre;Fgfr2^−/− mice, however, SOX9 expression remained undetectable, and the persistent AF was occupied by SCX⁺ cells that structurally resembled suprasutural ligament in both the ecto- and endocranial layers (Fig. 5J). We then investigated whether the loss of SOX9⁺ cells could be caused by cell death in the Wnt1-Cre;Fgfr2^−/− mice. Immunofluorescence for cleaved-caspase showed no cell death in controls or Wnt1-Cre;Fgfr2^−/− mutants at P0, P3, or P5 (Fig. S2). Together, these results strongly indicate that the SCX⁺ cells of the AF connective tissue normally undergo regionally selective differentiation to contribute to intramembranous bone, endochondral bone, and ligament in the PFS, and that the persistent AF in Wnt1-Cre;Fgfr2^−/− mice is largely caused by failed differentiation of SCX⁺ cells into RUNX2⁺ and SOX9⁺ skeletogenic cells.

Fgfr2 has a well-established role in regulating proliferation and differentiation in the developing sutures. We next investigated how loss of Fgfr2 impacts proliferation during AF closure using 5-ethynyl-2-deoxyuridine (EdU) pulse-chase experiments, since persistence of the AF in Wnt1-Cre;Fgfr2^−/− mice could be due to increased proliferation in the AF and/or decreased proliferation in the osteogenic fronts of the frontal bone. At E18.5, no statistically significant differences were identified in the percentage of proliferating cells within the AF connective tissue or osteogenic fronts of the paired frontal bones (P>0.02) (Fig. S3A-C). When these regions were combined, a statistically significant difference in proliferation of <1% was seen between control and Wnt1-Cre;Fgfr2^−/− mice (6% in total control versus 5.1% in total mutant; P=0.02, n=7) (Fig. S3C). At P3, EdU pulse-chase showed a significant difference in proliferation of 2.3% between control and mutant mice (9.6% in total controls versus 7.3% in total mutant; P=0.004, n=6) (Fig. S3D-F). This difference was primarily driven by a decrease in the proliferative rate of the Wnt1-Cre;Fgfr2^−/− AF connective tissue cells (6.8% in control versus 4.9% in mutants; P=0.001, n=6). Proliferation rates in the bone fronts were not statistically significantly different between mutant and control (Fig. S3F). Therefore, it is unlikely that a changes in proliferation in Wnt1-Cre;Fgfr2^−/− mice lead to the persistent AF phenotype.

To determine the fate of AF cells during PFS formation, we traced Scx-lineage (Scx^LIN) and Sox9-lineage (Sox9^LIN) cells at P4 and P7. In Scx-Cre;TdTomato:Ai9 mice at P4, Scx^LIN cells occupied the osteogenic fronts of the frontal bones, as well as the ecto- and endocranial layers of the AF (Fig. S4A). At P7, the contribution of Scx^LIN cells to the ecto- and endocranial layers of the AF had increased, occupying the mid-suture, osteogenic fronts of the frontal bones, and the suprasutural ligament (Fig. S4B). In Sox9-CreERT2;TdTomato:Ai9 mice induced with tamoxifen at P2 and P3, Sox9^LIN cells were enriched in the endocranial domain of the AF and also contributed to the frontal bones at P4 (Fig. S4C). At P7, the Sox9^LIN cells contributed to the condensing cartilage in the endocranial domain of the PFS (Fig. S4D). These results show that different cell populations of the AF make direct and region-specific contributions to the PFS.

We next tested the extent to which Fgfr2 is required in Scx- and Sox9-expressing cells during AF development. Wholemount skeletal staining of Scx-Cre;Fgfr2^−/− mice at P5 and P7 showed that the AF closure is indistinguishable from littermate controls (Fig. 6A-D). Since feature plots of the scRNA-seq dataset identified cells within the AF1 and AF2 clusters that co-express Fgfr2 and Scx (Fig. S4E), we concluded that loss of Fgfr2 in Scx⁺ cells was not sufficient to recapitulate the AF and PFS phenotypes seen in Wnt1-Cre;Fgfr2^−/− mice. Wholemount skeletal staining of Sox9-CreERT;Fgfr2^−/− mice at P5 following tamoxifen induction at E17.5 and 18.5 revealed that AF closure was indistinguishable from littermate controls (Fig. 6E,F). The same result was found when Sox9-CreERT;Fgfr2^−/− mice were induced at P0 and collected at P7 (Fig. 6G,H). However, histological examination of the PFS in Sox9-CreERT;Fgfr2^−/− mice at P9 following tamoxifen inductions at P2 and P3 showed loss of the endocranial cartilage (Fig. 6I,J). Feature plots of the scRNA-seq dataset identified cells within the AF2 and OF2 clusters that co-express Fgfr2 and Sox9 (Fig. S4F). Thus, although there is some autonomous requirement for Fgfr2 during endochondral-like fusion of the PFS, it is not required in Sox9⁺ cells for AF closure.

To determine the molecular mechanisms through which Fgfr2 regulates differentiation within the AF, we performed bulk RNA-seq of the control and Wnt1-Cre;Fgfr2^−/− AF at E18.5, when a phenotypic difference was first detected. Heat map analysis of the differentially expressed genes showed that RNA expression profiles of control and Wnt1-Cre;Fgfr2^−/− littermates have significant, genotype-specific differences (Fig. 7A). Gene ontology (GO) analysis of these genes indicated differential expression of members of the WNT signaling pathway, as well as genes related to focal adhesion, extracellular matrix–receptor interaction, and axon guidance (Fig. 7B). Previous studies have identified a role for WNT signaling in the developing suture, where this pathway works cooperatively with FGF signaling to determine skeletogenic cell fate (Liao et al., 2022; Quarto et al., 2010; Maruyama et al., 2010). Dot plots of WNT pathway members showed that secreted inhibitors Sfrp4, Notum, and Wif1, along with the WNT potentiator Lgr6 were downregulated in the Wnt1-Cre;Fgfr2^−/− AF (Fig. 7C). By contrast, the WNT/β-catenin target gene and WNT signaling potentiator Lgr5, Wnt11, and the context-dependent WNT modulator Sfrp2 were upregulated in the Wnt1-Cre;Fgfr2^−/− AF (Fig. 7C). String analysis of the WNT pathway genes found in our sequencing predicted likely interactions between the FGF and WNT pathways (Fig. 7D).

By comparing differentially expressed genes identified in the RNA-seq analysis with the cluster-enriched genes from the scRNA-seq dataset, we found that 40 of 42 AF1-enriched genes, 19 of 25 AF2-enriched genes, and 46 of 48 EM-enriched genes identified were upregulated in the Wnt1-Cre;Fgfr2^−/− AF (Fig. S5). We found that 12 of 13 OF1-enriched genes, 16 of 16 OF2-enriched genes, and 57 of 57 OB-enriched genes identified were downregulated in the Wnt1-Cre;Fgfr2^−/− AF. This suggests that loss of Fgfr2 leads to increased expression of genes associated with AF connective tissue cell identity and decreased expression of gene associated with osteoblast identity.

Violin plots of WNT pathways members from the scRNA-seq dataset showed that the AF1, AF2, and EM clusters had enriched expression of the WNT receptor Fzd1, as well as Wnt9a, Wnt11, and Lgr5 (Fig. 8A,A′). Cellular communication modeling using CellChat predicted that WNT ligands from the EM regulate WNT signaling in AF1 and AF2 (Fig. 8B). The OF1, OF2, and OB clusters, by contrast, were enriched for Wif1 in cells that also expressed Fgfr2 (Fig. 8A,A′,C). WNT-related gene expression changes identified through RNA-seq were validated using RNAScope in situ hybridization at E18.5. Lgr5 expression, which was enriched in the endocranial domain of the control AF, was expanded in the Wnt1-Cre;Fgfr2^−/− AF (Fig. 8D,E, asterisks). Lgr6 expression, which was observed in both the endo- and ectocranial domains of the control AF, was reduced within the endocranial domain of the Wnt1-Cre;Fgfr2^−/− AF (Fig. 8F,G, arrowhead). Wif1 expression was localized to cells within the osteogenic fronts in the control and greatly reduced in Wnt1-Cre;Fgfr2^−/− embryos (Fig. 8H,I, arrowheads). Together, these results indicate that Fgfr2-dependent activation of Wif1 expression in cells of the osteogenic front non-autonomously modulates WNT signaling in Lgr5- and Lgr6-expressing cells of the AF. LGR5 is a known stem cell marker in multiple epithelial tissues, and in this context Lgr5-expressing cells may represent connective tissue progenitors (Leung et al., 2020; Ng et al., 2014; Shi et al., 2012).

To test the impact of increased WNT/β-catenin signaling in Scx⁺ cells of the AF, we crossed mice harboring the Scx-Cre driver to mice with a stabilizing mutation in exon 3 of β-catenin [Ctnnb1^lox(Ex3)] (Harada et al., 1999). Stabilization of the β-catenin protein blocks its degradation through the destruction complex and mimics conditions of constitutively active WNT signaling. Wholemount skeletal staining of Scx-Cre;Ctnnb1^lox(Ex3)/+ (Scx-Cre;Ctnnb1^GOF) mice at E18.5 showed that the AF is larger compared to that of littermate controls (Fig. 9A,B). At P5, when the frontal bones of control mice approximate to form the PFS, the AF remained patent and the PFS failed to form in Scx-Cre;Ctnnb1^GOF mice (Fig. 9C,D). The observation that increased WNT/β-catenin signaling in Scx⁺ cells recapitulate the phenotype of Wnt1-Cre;Fgfr2^−/− mice supports a mechanistic connection between FGF and WNT signaling during AF closure and PFS formation.

To test the idea that elevated WNT signaling in Wnt1-Cre;Fgfr2^−/− mice underlies the AF closure defect, we attempted a rescue through WNT signaling inhibition. To inhibit WNT signaling, we administrated a cocktail of mouse WIF1 recombinant protein and vantictumab, a humanized monoclonal antibody that inhibits WNT pathway signaling through competitive binding of FZD receptors. At P3, Wnt1-Cre;Fgfr2^−/− pups and littermate controls were treated with a single injection of the WNT inhibitor cocktail just under the scalp at the location of the AF. Samples were collected at P7 and processed for wholemount skeletal staining. Following treatment, AF closure was largely restored in 9 out of 12 treated Wnt1-Cre;Fgfr2^−/− mutants (Fig. 9E-G). Quantitative measurement showed that the median area of the AF was reduced in treated Wnt1-Cre;Fgfr2^−/− (487.75; s.d.=398.49, n=12) versus untreated Wnt1-Cre;Fgfr2^−/− mice (837; s.d.=275.43, n=11) (Fig. 9G). These results indicate that elevated WNT signaling is a crucial driver of the persistent AF phenotype in Wnt1-Cre;Fgfr2^−/− mice.

This study characterizes a previously unappreciated role for FGF signaling in calvarial development. Using a combination of mouse genetics and single-cell transcriptomics, we demonstrate that Fgfr2 is necessary for closure of the AF and development of the PFS it prefigures. We find that cells of the AF, which are marked by Scx and differentially express Sox9, are regionally organized into ecto- and endocranial domains that give rise to distinct lineages within the PFS. We provide evidence that FGFR2 signaling in the osteogenic fronts of the frontal bones, through induction of the expression of Wif1, induces local inhibition of WNT signaling in the AF (Fig. 10). Within the zone of WNT inhibition, SCX⁺ cells in the ectocranial domain are recruited to form bone, and SCX⁺/SOX9⁺ cells in the endocranial domain are recruited to form cartilage. A subset of ectocranial SCX⁺ cells that remain outside the domain of WNT inhibition give rise to suprasutural ligament. Upon loss of FGFR2 signaling, Wif1 expression is lost, WNT signaling remains, and cells of the AF retain a connective tissue-like fate instead of forming the PFS.

Our results reveal that FGF and WNT pathways interact to coordinate frontal bone advancement with suture joint formation. We show that differential expression of FGF and WNT pathway members in cells of the ectocranial mesenchyme, osteogenic fronts, and AF establish a signaling center that promotes a unique lineage potential along the ectocranial-endocranial axis. Ectocranial mesenchymal cells express FGF and WNT ligands, suggesting that these cells are an essential source of secreted signals that act on Fgfr2⁺ cells in the osteogenic fronts and Fgfr2⁺ and Fzd1+ cells in the AF. FGF18 is likely the ligand responsible for promoting osteoblast cell fate in Fgfr2⁺ cells, as prior studies show that loss of Fgf18 leads to wide-open fontanelles and ectopic FGF18 accelerates approximation of the bone fronts in the coronal suture (Ohbayashi et al., 2002; Sahar et al., 2005; Nagayama et al., 2013). WNT, by contrast, likely acts to maintain Fzd1⁺ cells in the AF in an undifferentiated state. This is supported by multiple studies showing that, in the PFS, canonical WNT signaling is dynamic and sharply downregulated immediately preceding cartilage formation in the endocranial domain while continuous activation of WNT leads to a wide-open frontal suture and failure of cartilage formation (Behr et al., 2010, 2011, 2013). We propose that Fgfr2⁺ cells at the osteogenic fronts downregulate WNT in the AF to promote chondrogenic fate through their expression of the secreted WNT inhibitor WIF1. That WNT promotes connective tissue over cartilage fate is also supported by studies in the developing limb (ten Berge et al., 2008).

The calvaria develops from a layer of head mesenchyme that is derived from mesoderm or neural crest, and regionally patterned into osteogenic and non-osteogenic domains. Mesenchyme that initially resides in the supraorbital ridge just above the eyes, called the supraorbital mesenchyme (SOM), generates the ossification centers for the frontal and parietal bones (Ishii et al., 2015; Deckelbaum et al., 2012; Jiang et al., 2002; Yoshida et al., 2008). In contrast, mesenchyme that lies apical to the SOM, called the early migrating mesenchyme (EMM), remains non-osteogenic through embryonic stages of calvarial development (Cesario et al., 2018; Roybal et al., 2010). Growth of the frontal and parietal bone rudiments towards the vertex of the head in embryogenesis is fueled by the SOM through intrinsic growth rather than recruitment of adjacent EMM mesenchyme (Yoshida et al., 2008). Instead, the EMM contributes to non-osteogenic connective tissue that occupies the sutures and fontanelles (Cesario et al., 2018). Our findings elaborate on this model by showing that, in the postnatal period, non-osteogenic SCX⁺ cells within the AF directly contribute to bone, cartilage, and suprasutural ligament within the PFS. This suggests that formation of the PFS within the AF is developmentally distinct. Our results indicate that, as the frontal bones approximate, they recruit SCX⁺ cells of the AF to initiate frontal suture joint formation, which is characterized by the formation of a secondary osteogenic front within the endocranial layer that subsequently undergoes endochondral fusion. This may be a key factor that distinguishes the bilayered, fusing murine PFS from the overlapping coronal suture or abutting sagittal suture, which do not fuse during normal mouse development. Importantly, SCX⁺ cells in the AF appear to be distinct from those we previously identified in the non-fusing murine coronal suture. In the coronal suture, SCX⁺ cells are highly restricted to the overlapping edges of the frontal and parietal bones in the ectocranial domain and appear to remain non-osteogenic (Farmer et al., 2021).

Overall, the results presented in this study offer further insight into regional differences in the development of sutures that are prefigured by fontanelles. How might the mechanism of AF closure identified here explain why both an increase and decrease in FGFR2 signaling lead to failed formation of sutures that develop within fontanelles and why an increase in FGFR2 signaling causes AF patency along with premature fusion of other cranial sutures? Although it is not yet clear how FGFR2 gain-of function variants impact the mechanism for AF closure we described here, our study indicates that suture-specific requirements for FGF and/or WNT signaling levels could be the key to answering these questions. One possibility is that the AF phenotype observed in a model of FGFR2 loss of function leads to excess unbound ligand that can diffuse and activate signaling in regions where FGF signaling is normally low, recapitulating a gain of function via other receptors. Specifically, whereas FGFR2 is expressed in the osteogenic fronts and AF near a source of FGF ligands, FGFR1 and FGFR3 are expressed in osteoblasts located in an adjacent domain (Johnson et al., 2000; Iseki et al., 1999; Rice et al., 2000). Another possibility is that an increase or decrease in FGFR2 signaling leads to WNT activation through distinct mechanisms. Interestingly, in a mouse model for Apert syndrome, which harbors the gain-of-function variant Fgfr2^S252W that causes both craniosynostosis and a persistent AF, the premature fusion of the coronal suture has been linked to activation of WNT signaling through increased expression of the Frizzled co-receptors Lrp5 and Lrp6 (Min Swe et al., 2020). Increased WNT signaling may also underlie persistent AF patency in other craniosynostosis syndromes. A previous study found that Lgr5 is upregulated in the developing frontal suture of Twist^+/− mice, a model for Saethre–Chotzen syndrome, which presents with persistent AF and premature coronal suture fusion (Holmes et al., 2020).

To conditionally knockout Fgfr2 in neural crest cell (NCC)-derived tissue, Fgfr2^flx/flx (The Jackson Laboratory, stock #007569) mice were crossed with the Wnt1-Cre driver (The Jackson Laboratory, stock #003829), the Sox9-CreERT2 driver (The Jackson Laboratory, stock #035092) (Akiyama et al., 2005), or the Scx-Cre driver, which was generously provided by Dr Ronen Schweitzer (Shriners Hospital for Children, Oregon Health and Science University; Portland, OR, USA). The Fgfr2^flx/flx and Wnt1-Cre lines were previously described (Lewis et al., 2013; Yu et al., 2003). Scx-GFP was used to mark dense connective tissue fibroblasts (Pryce et al., 2007). Scx-Cre and Sox9-Cre-ERT2 lines were described previously (Xu et al., 2015; Sugimoto et al., 2013). The Ai9 allele (The Jackson Laboratory, stock #007909) was used as a lineage marker for those tissues targeted by Scx-Cre and Sox9-Cre-ERT2 (Madisen et al., 2010). For Sox9-Cre-ERT2 mice, tamoxifen (Sigma-Aldrich; dissolved in peanut oil overnight at 40°C) was delivered via oral gavage to either timed pregnant females (for embryonic inductions; 1 mg/induction/dam) or neonatal mice (for postnatal inductions; 0.1 mg/induction/pup). Finally, constitutive activation of β-catenin in fibroblasts was driven by crossing Ctnnb^lox(Ex3) mice (Harada et al., 1999) to mice carrying the Scx-Cre driver. Embryonic samples were collected from timed pregnant females. Postnatal samples were staged according to the date of birth. All experimental protocols were approved by the USC's Institutional Animal Care and Use Committee.

Mice processed for wholemount imaging were collected as above and following a PBS rinse were fixed in 95% ethanol for a minimum of 4 days. Fixed samples were stained for cartilage using 0.15 mg/ml Alcian Blue (Sigma-Aldrich, A5268) in 80% ethanol and 20% glacial acetic acid. Samples were then de-stained in 95% ethanol for a minimum of 2 days. Tissue clearing was performed using KOH between 0.5 and 1% w/v for 2-8 days depending on the size of the sample. Calcified tissue was then discriminated using 0.02 mg/ml Alizarin Red S (Sigma-Aldrich, A5533) in 0.5-1% KOH. Destaining and further clearing was performed as needed in 0.5-1% KOH. Samples were equilibrated, stored, and imaged in 75% glycerol. Measurements of the AF area using ImageJ and each individual area measurement were normalized to the length of the PFS from the jugum limitans to the coronal suture. These were then averaged within each group and the average AF area of control and Wnt1-Cre;Fgfr2^−/− mice was compared by age using an unpaired, two-tailed t-test assuming unequal variance.

All μCT scans were performed by the USC Molecular Imaging Center using a μCT50 (Scanco Medical). Samples were rotated 360° and X-ray settings were standardized to 90 kV and 155 µA with an exposure time of 0.5 s per frame to yield a nominal resolution of 20 μm. A 0.5-mm-thick aluminum filter was employed to minimize beam-hardening artifacts. Morphometric analysis was performed using Amira 6.2 and VG Studio MAX 3.0 software packages. Isosurface renderings with equal threshold were measured using the 3D measuring tool. All jaw measurements and landmarks were measured as previously described (Guerreiro et al., 2013). These experiments were performed on at least three biological replicates, which we defined as three same-sex littermate pairs (control and mutant) derived from three different litters. Statistical significance was determined using unpaired, two-tailed Student's t-tests.

Samples used for histology were processed as follows. Animals were euthanized humanely and decapitated. Heads were then fully or partially skinned, rinsed in PBS, and fixed in 10% neutral buffered formalin at room temperature for 24-36 h depending on stage. Following fixation, samples were decalcified in 10% EDTA (pH 7.4) at 4°C for 3-7 days, until calvaria began to concave. Samples were then dehydrated in serial ethanol washes of increasing concentration, equilibrated in Citrus Clearing Solvent (VWR, 72060-044), and embedded following three washes in pure paraffin wax at 65°C with vacuum pressure. Embedded samples were then cut into 8-µm-thick sections. Sections were stained using the Hall-Brunt quadruple stain (Hall, 1986).

Samples for immunofluorescent staining were harvested and fixed in 4% paraformaldehyde at 4°C for 30-90 min depending on stage. Decalcification of bone was performed by incubation in 10% EDTA (pH 7.4) at 4°C for 3-7 days until calvaria began to concave. Subsequently, samples were equilibrated sequentially in 15% and 30% sucrose/PBS at 4°C until they sank, then embedded in optimal cutting temperature (O.C.T.) compound (Electron Microscopy Sciences). Cryosectioning was performed at a section thickness of 8 μm onto SuperFrost Plus slides. Sections were washed three times with PBS, permeabilized with 0.1% Triton X-100 in PBS, then blocked for 1 h at room temperature in 10% serum (either donkey or goat from Sigma-Aldrich). Slides were incubated in primary antibody overnight at 4°C (Table S1), then washed in PBS and incubated in Alexa Fluor 568 secondary antibody (Thermo Fisher Scientific A-11036) at a concentration of 1:200 for 1 h at room temperature. Finally, slides were washed in PBS and mounted in VECTASHIELD with DAPI (Vector Laboratories). Slides used for lineage tracing were washed three times in PBS and counterstained with DAPI as described above. Slides were imaged using either a Leica TCS SP5/8 or Stellaris 5 confocal system. These experiments were performed on three biological replicates, which we defined as three littermate pairs (control and mutant) derived from two different litters.

Data for scRNA-seq analysis was obtained from FaceBase (accession 1-4TT6) (Holmes et al., 2020). Seurat v.3 R-Package was used to explore QC metrics and filter for high-quality cells (Stuart et al., 2019). The filtered dataset representing 3366 cells (median of 3750 genes per cell) was normalized (log normalized), scaled (linear transformation), and clustered into 16 clusters (dims 1:13, resolution=0.7) using unsupervised graph-based clustering based on principal component analysis scores. The identities of the clusters were resolved using previously reported markers. Pseudotime analysis was conducted using Monocle 3 (Cao et al., 2019). Cell–cell communication predictions were made using CellChat (Jin et al., 2021).

Samples for in situ hybridization were collected and processed in the same manner as histological samples, described above. Transcripts were detected using RNAScope Fluorescent Multiplex Assay v2 (ACD) as per manufacturer's instructions. Briefly, slides with paraffin sections were baked in an embedding oven for 60 min at 65°C. Slides were then deparaffinized using xylenes, dehydrated in 100% ethanol, and allowed to completely air dry. Endogenous peroxidase activity was quenched using hydrogen peroxide provided within the kit and slides were washed in deionized water. Antigen retrieval was performed using the provided Target Retrieval Reagent for 15 min in an Oster steamer heated to 99°C. Slides were then dehydrated in 100% ethanol and allowed to completely air dry. The slides were treated with ACD Protease plus in a humidified slide chamber for 15 min at 40°C. Probe hybridization was also performed in a 40°C humidified slide chamber for 2 h (Table S1). Slides were then stored in 5× SSC solution overnight. Amplification steps were performed as prescribed by the manufacturer and signal development was carried out using Opal 570 (Akoya Biosciences, FP1488001KT) Opal 620 (Akoya Biosciences, FP1495001KT) or Opal 690 (Akoya Biosciences, FP1497001KT) diluted 1:1000 in the ACD provided TSA buffer. Slides were then counterstained using VECTASHIELD mounting medium with DAPI and imaged using the Leica TCS SP5/8 or Stellaris 5 confocal system. These experiments were performed on three biological replicates, which we defined as three littermate pairs (control and mutant) derived from two different litters.

Cell proliferation was assayed using an in vivo EdU Click kit according to the manufacturer's instructions (Sigma-Aldrich, BCK647-IV-IM-M). Briefly, E18.5 samples were obtained by intraperitoneally injecting timed pregnant females with 1.5 mg EdU dissolved in PBS, and P3 mice received subcutaneous injections of 0.1 mg EdU dissolved in PBS. In both cases, a chase period of 4 h was observed before sample collection. Samples were cryo-embedded and sectioned at 8 μm as described above. Detection of EdU took place in the dark, at room temperature for 30 min. Slides were then counterstained and mounted using VECTASHIELD mounting media with DAPI. For analysis, littermate control and Wnt1-Cre;Fgfr2^−/− pairs were selected, and a minimum of six different sections from each sample were imaged and uploaded to ImageJ. Each image was quantified for total AF cells by counting individual nuclei, and the number of proliferating cells was counted. Statistical analysis was performed on total proliferation in mutants compared with controls, proliferation of AF fibrous connective tissue, and proliferation in bone front cells. Statistical significance was determined using unpaired, two-tailed t-tests assuming unequal variance.

The region of the AF was excised from E18.5 embryos (including both suture connective tissue and bone but excluding skin and brain) and added to 300 µl of DNA/RNA Shield (Zymo Research). RNA was isolated using the Quick-RNA Miniprep Plus Kit (Zymo Research) according to the manufacturer's instructions with the addition of an optional DNase digestion and a 1 min dounce homogenization step following proteinase-K digestion. Samples were eluted from the column using 80 µl of DNase/RNase-free water and then analyzed using a NanoDrop spectrophotometer and to check concentration and purity. Library preparation and sequencing was performed at the UCLA TCBG facility using the Kapa Stranded Kit (Roche). Sequencing was performed using HiSeq 3000 at 1×50 read length and 30 million reads per sample. Differential gene expression analysis was performed using Partek Flow.

Stock vantictumab (Fisher Scientific, MA5-42006) and recombinant WIF1 protein (R&D Systems, 135-WF-050) were diluted in sterile water to a final concentration of 1.1 mg/ml and 0.5 mg/ml, respectively. At P3, pups were treated with a single dose of the inhibitor cocktail (a total of 2 μg vantictumab and 2 μg Wif1) using a pre-filled 29 G insulin syringe. For the treatment, pups were placed on a nitrile glove on ice for 30-45 s to slow movement, and the WNT inhibitor cocktail was injected subcutaneously just under the scalp above the area of the AF. Volume of liquid per injection did not exceed 20 μl. Following treatment, pups were placed on a heating pad and observed for several minutes to ensure recovery from the procedure before returning to their dams.

Some of the text and figures in this paper formed part of Lauren Bobzin's PhD thesis in the Development, Stem Cells, and Regenerative Medicine Doctoral Program at the University of Southern California in 2024 (Bobzin, 2023). The UCLA Technology Center for Genomics & Bioinformatics provided support for cDNA library creation and bulk RNA-sequencing.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.204264.reviewer-comments.pdf