Temporally regulated FGFR2-retinoic acid signaling mediates dura mater-suture mesenchyme interactions to prevent craniosynostosis in mice

Abstract

Congenital anomalies often arise during critical developmental time windows, yet the underlying mechanisms remain unclear. Here, we identify a time-specific FGFR2-retinoic acid signaling axis regulating postnatal coronal suture development through dura mater-suture mesenchyme interactions. FGFR2 signaling gradually declines after birth, leading to reduced retinoic acid signaling by moderating the expression of Aldh1a3 in the dura mater and suture mesenchyme alongside that of Rbp1 in the dura mater, thereby preserving GLI1+ progenitors and restraining osteogenesis to maintain suture patency. Importantly, FGFR2 overactivation in Fgfr2IIIc mutant mice within an early time window, disrupts this physiological FGFR2-retinoic acid signaling decline by upregulating Rbp1 and Aldh1a3 through enhanced P38 signaling. This leads to increased retinoic acid synthesis, premature osteogenic differentiation of GLI1+ progenitors, and coronal suture craniosynostosis. Significantly, genetic restoration of retinoic acid signaling rescues craniosynostosis in Fgfr2IIIc mutant mice, confirming retinoic acid signaling as a key downstream effector of FGFR2 signaling. Notably, restoration of suture patency alone rescues neurocognitive dysfunctions in Fgfr2IIIc mutant mice despite Fgfr2 mutation in the brain, demonstrating that the neurocognitive impairments primarily arise from cranial structural constraints and elevated intracranial pressure rather than intrinsic neural defects. This critical postnatal time window in mice parallels the timing of FGFR2-related craniosynostosis onset in humans, underscoring the importance of this study in advancing our understanding of the molecular and cellular mechanisms in craniosynostosis. Our findings define a temporally regulated FGFR2-P38-retinoic acid signaling axis and highlight retinoic acid signaling as a promising therapeutic target in FGFR2 overactivation-related craniosynostosis.

Introduction

The impacts of congenital anomalies are often shaped by time-specific developmental windows during which morphogenetic processes occur. A broad spectrum of congenital malformations, from neural tube defects to craniofacial anomalies, arise only when perturbations occur within precise embryonic or early postnatal time windows.^1–3 Cranial sutures provide a powerful model for studying the temporal dynamics of developmental pathogenesis due to their prolonged postnatal patency, well-defined molecular and cellular boundaries, accessibility in both animal and human studies, and clinical significance in humans.^3–5 Craniosynostosis is the premature fusion of cranial sutures and can result in skull shape abnormalities, elevated intracranial pressure (ICP), and neurocognitive impairments.^3,4,6–8 Among the syndromic forms of craniosynostosis, gain-of-function mutations in fibroblast growth factor receptor 2 (FGFR2) represent one of the most common and clinically significant etiologies, with different variants resulting in Apert and Crouzon syndromes.^9–13 Although previous work has explored the role of FGFR2 signaling in osteogenesis and suture biology during embryonic development, the precise timing and downstream signaling mechanisms that drive craniosynostosis at postnatal stages remain incompletely understood.

FGFR2 signaling plays a key role in regulating progenitor cell dynamics in multiple organs, including the lung, salivary gland, and skin, by supporting epithelial-mesenchymal interactions and maintaining stem cell niches.^14–18 However, its role in cranial suture progenitor regulation remains unclear. Specifically, whether and how FGFR2 signaling affects the differentiation trajectory of progenitors in the coronal suture has not been elucidated. The coronal suture contains a distinct population of GLI family zinc finger 1 (GLI1)+ progenitors that are essential for preserving postnatal suture patency.¹⁹ This GLI1+ progenitor pool interfaces with the osteogenic fronts and other surrounding tissues, presenting a unique model for investigating compelling questions about spatially coordinated cell-fate regulation through tissue-tissue interactions. Our study addresses these critical knowledge gaps and explores how aberrant FGFR2 signaling alters the fate of GLI1+ progenitors to drive the development of craniosynostosis.

Tissue-tissue interactions are fundamental to developmental patterning, especially in spatially distinct structures like the cranial vault.^5,20,21 The dura mater, a meningeal layer classically known for protecting the central nervous system, has been suggested to influence cranial suture development.^20,22–24 The dura mater’s unique embryonic origin and anatomical proximity to the suture mesenchyme suggest it may contribute to craniofacial morphogenesis, supported by the observation that surgical ablation or transposition of dura mater can impact cranial suture patency in animal models.^21,25,26 However, the detailed molecular communication between the dura mater and suture mesenchyme, particularly in the regulation of progenitor cell fate, remains unclear.

In this study, we define a temporally specific role of FGFR2 signaling in coronal suture development. In mice, embryonic Fgfr2 mutation induces craniosynostosis, yet the postnatal requirement for FGFR2 activity declines markedly between postnatal day 3.5 (P3.5) and P7.5. Overactivation of FGFR2 signaling in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice during the P3.5 time window disrupts coronal suture development through aberrant activation of retinoic acid (RA) signaling, perturbing the synergistic functions of the dura mater and coronal suture mesenchyme. Mechanistically, overactivation of FGFR2-P38 mitogen-activated protein kinase (P38 MAPK) signaling axis upregulates the RA synthesis genes retinol binding protein 1 (Rbp1) in the dura mater and aldehyde dehydrogenase 1 family member A3 (Aldh1a3) in both the coronal suture mesenchyme and dura mater, thereby disrupting the normal postnatal decline of local RA levels and resulting in craniosynostosis. Furthermore, we demonstrate that restoring RA signaling alleviates craniosynostosis by restoring GLI1+ progenitors and reducing abnormal osteogenic differentiation in Fgfr2IIIc mutant mice. Our study offers new insights into cranial suture biology and suggests RA signaling as a promising therapeutic target in FGFR2 mutation-related craniosynostosis.

Results

Fgfr2 regulates coronal suture development via GLI1+ progenitors during a specific postnatal time window

FGFR2-mediated signaling is critical for embryonic cranial suture formation and is implicated in a number of syndromic forms of craniosynostosis, especially those involving the coronal suture.^10,27,28 To investigate whether FGFR2 signaling is activated in GLI1+ progenitors at early postnatal stages, we analyzed the Fgfr2 expression in GLI1+ progenitors at P1.5, P3.5, and P7.5 using Gli1-LacZ reporter mice, with β-Gal+ cells representing GLI1+ cells. We focused on the coronal suture because it is the primary site affected in FGFR2 mutation-related syndromic craniosynostosis.^16,29 At these stages, Fgfr2 expression robustly overlapped with β-Gal+ cells in the dura mater and coronal suture mesenchyme, especially in the coronal suture mesenchyme near the dura mater (Fig. 1a). Notably, Fgfr2 expression levels within β-Gal+ cells were comparable at P1.5 and P3.5, but significantly declined by P7.5 (Fig. 1a–e). These findings indicate that FGFR2 signaling is activated within GLI1+ progenitors during early postnatal coronal suture development and is physiologically sustained through P3.5 but declines by P7.5, defining a temporally restricted window of FGFR2 signaling activity.

Fig. 1.

Fgfr2 overactivation in GLI1+ progenitors leads to coronal suture craniosynostosis. a Hematoxylin and eosin (H&E) staining (a1–a3) and colocalization analyses of β-Gal+ cells and Fgfr2 expression by immunostaining of β-Gal and RNAscope of Fgfr2 (a1’–a3’, a1”–a3”) at P1.5, P3.5 and P7.5 in Gli1-LacZ mice. Dotted/dashed lines outline the frontal bone, the parietal bone, and the dura mater. Triangles highlight the colocalization of β-Gal+ cells and Fgfr2 expression. b, d Quantification of β-Gal+Fgfr2+ cell percentage in coronal suture mesenchyme and dura mater. c, e Quantification of the Fgfr2 RNAscope-positive area within β-Gal+ cells, expressed as the percentage of β-Gal+ area occupied by Fgfr2 signal, representing relative Fgfr2 expression within the GLI1+ cells. f Colocalization analyses of β-Gal+ cells and Fgfr2 expression by immunostaining of β-Gal and RNAscope of Fgfr2 in Fgfr2IIIc^fl/+;Gli1-LacZ (f1, f1’) and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Gli1-LacZ (f2, f2’) mice at P5.5 after tamoxifen induction at P3.5. Triangles highlight the colocalization of β-Gal+ cells and Fgfr2 expression. g–j Quantification of the Fgfr2 RNAscope-positive area within β-Gal+ (g, i) and β-Gal- (h, j) cells, expressed as the percentage of β-Gal+ and β-Gal- area occupied by Fgfr2 signal, representing relative Fgfr2 expression within the GLI1+ and GLI1- cells. k MicroCT images of control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice showing the skull (k1, k2), coronal suture ortho-slices (k1’, k2’), H&E staining of the coronal sutures (k3, k4), and head size in lateral view (k5, k6). Triangles highlight the normal coronal suture, while the asterisks highlight the fused coronal suture. Arrows highlight skull length of the mice, which is shortened in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. Scale bars: 20 μm for (a, f), 50 μm for (k3, k4), 1 mm for (k1, k1’, k2, k2’), 1 cm for (k5, k6). Data are mean ± SEM, n = 20 in (k) and n = 3 in the other figures. For Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice in (k), phenotype penetrance = 100%. *P < 0.05, **P < 0.01, ns: not significant, calculated by one-way ANOVA (analysis of variance) with Tukey’s multiple comparisons test in (b–e), and by two-tailed unpaired t test (g–j)

To assess the biological response of the coronal suture to this decline in FGFR2 signaling, we selected P3.5 as a stage of high FGFR2 activity and P7.5 as a stage of reduced FGFR2 activity. Accordingly, we generated Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, as FGFR2-related syndromic craniosynostosis in humans is predominantly caused by FGFR2 heterozygous mutations,^3,29 and induced them with tamoxifen at these time points to conditionally overactivate FGFR2 signaling within the GLI1+ cell lineage. We first analyzed mouse samples collected at P5.5 following Cre^ERT2 activation by tamoxifen injection at P3.5, representing an early activation time point. We confirmed that Fgfr2 was selectively overexpressed within GLI1+ cells in both coronal suture mesenchyme and dura mater, whereas its expression in GLI1- cells was not significantly increased (Fig. 1f–j). To evaluate the effects of early activation of FGFR2 signaling, we collected mouse samples at postnatal 1 month (P1m). Notably, when induced at P3.5, the Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice developed bilateral coronal suture craniosynostosis with 100% phenotype penetrance (n = 20). Meanwhile, these mice showed a reduction in skull size and abnormal cranial morphology, as confirmed by micro-computed tomography (microCT) and principal component analysis (PCA) (Fig. 1k; Fig. S1a, b). Furthermore, the ICP was significantly elevated in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice at postnatal 3 months (P3m) (Fig. S1c). These findings suggest that early postnatal overactivation of FGFR2 signaling impairs coronal suture development. Interestingly, tamoxifen-induced Cre^ERT2 activation at P7.5 failed to affect suture patency, with both skull shape and ICP in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (n = 6) remaining comparable to controls (Fig. S1d–g). These results indicate that FGFR2 signaling overactivation influences postnatal coronal suture development within a critical window encompassing P3.5 but ending before P7.5. Based on this temporal specificity, all subsequent mechanistic studies were performed with tamoxifen induction at P3.5.

To determine whether other cranial sutures were affected in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, we examined sagittal and lambdoid sutures at P1m after tamoxifen induction at P3.5. Unlike the coronal sutures, both sagittal and lambdoid sutures remained patent and comparable to controls in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (n = 20) (Fig. S2a, b), indicating that FGFR2 overactivation in GLI1+ progenitors selectively disrupts coronal suture development rather than inducing more general cranial suture defects. Furthermore, to determine whether homozygous mutation of Fgfr2IIIc produces more severe phenotypes in coronal suture and other cranial sutures, we compared Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ and Gli1-Cre^ERT2;Fgfr2IIIc^fl/fl mice at P1m following tamoxifen induction at P3.5. Both heterozygous and homozygous mutants developed coronal suture craniosynostosis with 100% phenotype penetrance (n = 20 for heterozygous group, n = 6 for homozygous group), with more pronounced fusion in homozygous mutants (Fig. S2a–c). The sagittal and lambdoid sutures remained patent in both mutant groups (Fig. S2a–c). Survival rates were comparable between the two mutant groups, with approximately 80% surviving to P1m (Fig. S2d), indicating that homozygous Fgfr2IIIc mutation in GLI1+ progenitors does not increase lethality. Together, these findings indicate that while homozygous mutation of Fgfr2IIIc produces more severe coronal suture craniosynostosis, heterozygous mutation is sufficient to induce coronal suture craniosynostosis with complete phenotype penetrance. We therefore used the Fgfr2IIIc heterozygous mutation model for subsequent mechanistic studies to more closely mimic the heterozygous mutation typically seen in human FGFR2-related craniosynostosis.^3,29

Fgfr2 gain-of-function mutation in GLI1+ progenitors triggers coronal suture fusion from the dura mater side

To investigate the etiology and mechanism of craniosynostosis in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, we performed histological analysis following FGFR2 overactivation at P3.5 and found that fusion was detectable as early as P10.5 (Fig. S3a, b). Interestingly, craniosynostosis consistently initiated on the dura mater side, rather than the periosteal side, of the skull bones in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. 1k), suggesting the potential contribution of the dura mater to early postnatal coronal suture development.

To investigate the potential function of the dura mater during early postnatal coronal suture development, we conducted a co-culture assay using conditioned medium (CM) from control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ dura mater cells (Fig. S3c). Compared to control coronal suture mesenchymal cells treated with CM from control dura mater cells, those treated with CM from Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ dura mater cells showed enhanced osteogenic differentiation indicated by increased expression of osteogenic differentiation markers runt-related transcription factor 2 (RUNX2) and SP7 (also known as osterix). Notably, the strongest osteogenic differentiation was observed when both the dura mater CM and coronal suture mesenchymal cells were derived from Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. S3d–f). These results indicate that the osteogenic differentiation of coronal suture mesenchymal cells increases after FGFR2 overactivation, with the dura mater promoting this aberrant osteogenic differentiation.

To investigate cellular changes underlying craniosynostosis, we first analyzed GLI1+ progenitors¹⁹ using Fgfr2IIIc^fl/+;Gli1-LacZ and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Gli1-LacZ mice. Notably, we found a significant decrease in β-Gal+ cell percentage in the coronal suture mesenchyme of Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Gli1-LacZ mice as early as P6.5 (Fig. 2a, c), which intensified by P8.5 (Fig. 2b, d). To further determine whether the reduced GLI1+ cell population in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice resulted from altered Hedgehog signaling, we analyzed the expression of key Hedgehog ligands sonic hedgehog (Shh), Indian hedgehog (Ihh), and desert hedgehog (Dhh), and core pathway components patched 1 (Ptch1), GLI family zinc finger 2 (Gli2), and GLI family zinc finger 3 (Gli3) at P6.5. None of these genes showed significant changes in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice compared with controls (Fig. S4), indicating that the decreased proportion of GLI1+ cells is not due to altered Hedgehog signaling activity.

Fig. 2.

Fgfr2 overactivation increases osteogenic differentiation of GLI1+ progenitors in the coronal suture. a, b Immunostaining analyses of β-Gal+ cells in coronal suture at P6.5 (a) and P8.5 (b) in Fgfr2IIIc^fl/+;Gli1-LacZ and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Gli1-LacZ mice. Dotted/dashed lines outline the frontal bone, parietal bone, and dura mater. Triangles highlight the β-Gal+ cells. c, d Quantification of the β-Gal+ cell percentage in coronal suture mesenchyme at P6.5 (c) and P8.5 (d). e Colocalization analyses of tdTomato and immunostaining of SP7 in coronal sutures of Gli1-Cre^ERT2;tdTomato and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;tdTomato mice. Triangles highlight the tdTomato+SP7+ cells in the coronal suture mesenchyme. f Colocalization of EdU staining and immunostaining of SP7 in coronal sutures of control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. Triangles highlight the EdU+SP7+ cells in the coronal suture mesenchyme. g Quantification of the SP7+ cell percentage in coronal suture. h, i Quantification of the percentage of tdTomato+SP7+ cells out of the total tdTomato+ cells in the coronal suture mesenchyme osteogenic front near dura mater (h) and near periosteum (i). j, k Quantification of the percentage of EdU+SP7+ cells out of the total EdU+ cells in the coronal suture mesenchyme osteogenic front near dura mater (j) and periosteum (k). Scale bars: 20 μm. Data are mean ± SEM, n = 3. *P < 0.05, **P < 0.01, ns: not significant, calculated by two-tailed unpaired t test c, d, g–k

We further analyzed osteogenic differentiation potentials of these GLI1+ progenitors using Gli1-Cre^ERT2;tdTomato and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;tdTomato mice, in which GLI1+ progenitors and their progeny were labeled with tdTomato. Significantly, we found an increase in the SP7+ osteoblast percentage, as well as in the percentage of tdTomato+SP7+ cells out of the total tdTomato+ cells in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;tdTomato mice as early as P6.5, especially in the suture mesenchyme osteogenic front near the dura mater (Fig. 2e, g–i). These results indicated increased osteogenic differentiation of GLI1+ progenitors in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. To further verify these results, we used 5-ethynyl-2′-deoxyuridine (EdU) to label proliferating cells and assessed their osteogenic differentiation potential through the colocalization of SP7 and EdU 3 days after EdU labeling. Notably, we found an increase in the percentage of EdU+SP7+ cells among the EdU+ cells in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. 2f, j, k), indicating increased osteogenic differentiation of proliferating cells. Together, these results suggest that FGFR2 overactivation in the coronal suture mesenchyme diminishes the GLI1+ progenitor pool, potentially through increased osteogenic differentiation of GLI1+ progenitors.

Transcriptomic profiling identifies RA signaling as a key downstream pathway of FGFR2 signaling in coronal suture development

To investigate the molecular mechanisms underlying these phenotypes in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, including craniosynostosis initiating from the dura mater side, we performed two complementary bulk RNA-seq analyses. First, we compared the dura mater and coronal suture mesenchyme between control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice at P5.5 (2 days after tamoxifen induction at P3.5) to identify genotype-driven transcriptional changes in both tissues. Second, we compared the dura mater and coronal suture mesenchyme of control mice at P3.5 and P7.5 to capture stage-specific transcriptional dynamics during development. Hierarchical clustering heatmaps and volcano plots revealed distinct gene expression profiles separating Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice from controls, and P3.5 from P7.5 stages (Fig. 3a, b, d, e; Fig. S5). In the genotype-wise comparison, we identified 386 upregulated and 199 downregulated genes in the Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ coronal suture mesenchyme, along with 228 upregulated and 255 downregulated genes in the Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ dura mater. Stage-wise analysis revealed 384 upregulated and 578 downregulated genes in the P3.5 control coronal suture mesenchyme when compared to P7.5, along with 367 upregulated and 538 downregulated genes in the P3.5 control dura mater when compared to P7.5.

Fig. 3.

Transcriptomic analyses identify upregulated RA signaling in dura mater and coronal suture mesenchyme upon Fgfr2 overactivation. a, b, d, e Heatmaps of the differentially expressed genes in bulk RNA-seq analysis of coronal suture mesenchyme and dura mater tissue comparing Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ and control mice at P5.5 after tamoxifen induction at P3.5 (a, b) and comparing P3.5 and P7.5 control mice (d, e). The differentially expressed genes were selected using cutoffs of P < 0.05 and fold change < −1.2 or >1.2. c, f Ingenuity Pathway Analysis (IPA) of enriched signaling pathways when comparing differentially expressed genes of coronal suture mesenchyme tissue and dura mater tissue between Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ and control mice as well as between P3.5 and P7.5 control mice. g–i The Least Squares mean (LS mean) values of the retinoic acid (RA) signaling member genes in the dura mater and coronal suture mesenchyme comparing Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ and control mice. Data are LSMean, n = 3. *P < 0.05, ***P < 0.001, calculated by Wald test followed by Benjamini-Hochberg correction to control the false discovery rate (g–i)

To further identify signaling pathways altered by FGFR2 overactivation, we performed Ingenuity Pathway Analysis (IPA) on differentially expressed genes from control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. This analysis revealed 98 enriched pathways in the coronal suture mesenchyme and 120 in the dura mater (Fig. 3c). More importantly, 31 pathways were shared between the two tissues, including BMP signaling, ID1 signaling, the RAF/MAPK cascade, FGFR2 signaling, and RA signaling, indicating coordinated pathway dysregulation in response to FGFR2 overactivation (Fig. 3c). We next examined pathway changes across developmental stages by comparing control samples at P3.5 and P7.5. This analysis identified 133 enriched pathways in the coronal suture mesenchyme and 138 in the dura mater, with 39 pathways overlapping between the two tissues (Fig. 3f). These shared pathways included extracellular matrix organization, VEGF signaling, TR/RXR activation, TP53 signaling, and RA signaling, suggesting common developmental regulation of these two tissues during the P3.5 to P7.5 time window (Fig. 3f). Across both analyses, RA signaling was the only pathway consistently ranked among the top five (Fig. 3c, f), highlighting its potential role in the pathogenesis of temporally restricted, FGFR2-driven craniosynostosis during this temporal window.

To further investigate the molecular basis of RA signaling changes, we analyzed transcriptomic data and identified two key RA synthesis genes, Rbp1 and Aldh1a3, as significantly upregulated in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. 3g, h). In addition, retinoic acid receptor gamma (Rarg), a canonical RA-responsive transcriptional target, was also significantly upregulated in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. 3i). To verify these observations, we investigated their in vivo expression and found that RBP1, which is an essential retinol carrier,^30–33 showed strong enrichment and robust upregulation specifically in the dura mater, both beneath bone and coronal suture mesenchyme, in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. 4a–c). Aldh1a3, which encodes a rate-limiting enzyme that converts retinaldehyde to RA,^34–36 was significantly upregulated in both the dura mater and coronal suture mesenchyme, particularly near the osteogenic surface, in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. 4a, d, e). Other RBP and ALDH1A family members including Rbp4, Rbp7, and Aldh1a2 showed minimal expression (Fig. S6), supporting the selective involvement of Rbp1 and Aldh1a3 in this context.

Fig. 4.

Natural decline of RA signaling is disrupted upon Fgfr2 overactivation in early postnatal coronal suture development. a Immunostaining of RBP1 expression (a1–a1”, a2–a2”) and RNAscope analyses of Aldh1a3 expression (a3–a3”, a4–a4”) in coronal suture of control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. Dotted/dashed lines outline the frontal bone, parietal bone, and dura mater. Triangles highlight the protein location or gene expression, while the asterisk highlights rare expression. b, c Quantification of the RBP1 immunostaining-positive area in dura mater under coronal suture mesenchyme (b) and under frontal bone (c), expressed as the percentage of area occupied by RBP1 signal, representing relative RBP1 expression. d, e Quantification of the Aldh1a3 RNAscope-positive area in the dura mater (d) and the coronal suture mesenchyme (e), expressed as the percentage of area occupied by Aldh1a3 signal, representing relative Aldh1a3 expression. f, g RNAscope analyses of Rarg expression in the dura mater and coronal suture mesenchyme at P3.5 (f) and P7.5 (g) in control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. Triangles highlight gene expression in the coronal suture mesenchyme and dura mater. h, i Quantification of Rarg expression by fluorescence area fraction in the dura mater and coronal suture mesenchyme showing a decrease in expression from P3.5 to P7.5 in control mice and elevated expression at P7.5 in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. j, k Quantification of RA concentration by ELISA in the dura mater and the coronal suture mesenchyme. Scale bars: 20 μm. Data are mean ± SEM, n = 3. *P < 0.05, **P < 0.01, ***P < 0.001, ns: not significant, calculated by two-tailed unpaired t test in (b–e), and by one-way ANOVA with Tukey’s multiple comparisons test in (h–k)

Overactivated FGFR2 signaling disrupts the normal decline of RA signaling specifically in coronal suture

To investigate whether altered RA signaling activity contributes to craniosynostosis in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, we assessed RA signaling activity at P3.5 and P7.5 using two complementary readouts: Rarg expression and RA concentration. Retinoic acid receptors (RARs) function as ligand-inducible transcriptional activators and are widely accepted as reliable indicators of RA signaling activity.^37–39 Among the RAR isoforms, Rarg was identified as the predominant receptor in the coronal suture (Fig. 4f, g), justifying it as a tissue-relevant transcriptional readout. Furthermore, RA functions as a morphogen whose local concentration gradient governs downstream pathway activation.³⁸ Therefore, quantifying tissue RA level provides a biologically valid assessment of RA signaling activity.

In control mice, both Rarg expression and RA concentration were significantly reduced at P7.5 compared to P3.5 in the coronal suture mesenchyme and dura mater (Fig. 4f–k), indicating a physiological decline of RA signaling during normal suture development. However, such a decline was not seen in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, which exhibited sustained Rarg expression and RA levels in the dura mater and coronal suture mesenchyme at P7.5 (Fig. 4f–k). These findings demonstrate that FGFR2 overactivation leads to abnormal activation of RA signaling activity during a period when it should physiologically decline. Together, these findings indicate that FGFR2 overactivation enhances RA synthesis through upregulation of Rbp1 in the dura mater and Aldh1a3 in both the suture mesenchyme and dura mater, disrupting the normal postnatal decline of RA signaling activity in the coronal suture.

Given that the craniosynostosis phenotype in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice was restricted to the coronal suture, we next investigated if differential downstream responses contributed to this regional heterogeneity. Following tamoxifen induction at P3.5, Fgfr2 expression was significantly increased in sagittal and lambdoid sutures of Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. S7a, b, e, f), indicating that FGFR2 overactivation also occurred in these sutures. However, unlike in the coronal suture, RA signaling activity was not significantly elevated, as neither Rarg expression nor RA concentration showed significant increases in the sagittal or lambdoid sutures in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. S7a, b, g–j). Furthermore, at P8.5, GLI1+ progenitors were not significantly affected in the sagittal or lambdoid sutures of Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Gli1-LacZ mice compared to Fgfr2IIIc^fl/+;Gli1-LacZ mice (Fig. S7c, d, k, l). These findings indicate that although FGFR2 overactivation occurred in all the three cranial sutures, downstream RA signaling was selectively overactivated in the coronal suture. This differential molecular response is consistent with the suture-specific craniosynostosis phenotype in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, supporting the hypothesis that coronal suture-specific RA signaling overactivation contributes to FGFR2-driven craniosynostosis.

FGFR2-P38 signaling axis regulates RA synthesis

To determine how FGFR2 signaling controls RA synthesis, we assessed readouts of its canonical downstream pathways, including PLCγ (phospholipase C gamma), AKT (protein kinase B), JNK (c-Jun NH₂-terminal kinase), ERK (extracellular signal-related kinase), and P38. The results showed that only P38 signaling activity was significantly altered, with elevation in both the coronal suture mesenchyme and dura mater of Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. 5a, b), suggesting FGFR2 signaling may regulate RA signaling through P38 signaling. To test whether P38 signaling directly regulates RA signaling, we inhibited P38 signaling activity using the specific inhibitor adezmapimod.^40,41 Upon inhibition, we found reduced expression of Rbp1 and Aldh1a3 in dura mater cells, and decreased Aldh1a3 expression in coronal suture mesenchymal cells (Fig. 5c–e, g, h). Furthermore, RA secretion by dura mater cells and RA synthesis by coronal suture mesenchymal cells also decreased after inhibiting P38 signaling (Fig. 5f, i). These results demonstrate that P38 signaling acts as a functional downstream effector of FGFR2 signaling to upregulate RA synthesis genes Rbp1 and Aldh1a3, leading to increased RA synthesis, thereby translating FGFR2 overactivation into RA signaling elevation within coronal suture.

Fig. 5.

FGFR2 overactivation upregulates RA signaling through P38 MAPK signaling pathway. a Western blot analysis of indicated genes in the dura mater and the coronal suture mesenchyme in control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. b Immunostaining of p-P38 expression in coronal sutures of control and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice. Dotted/dashed lines outline the frontal bone, the parietal bone, and the dura mater. Triangles highlight the p-P38+ cells in the dura mater and the coronal suture mesenchyme. c, g Western blots of p-P38 in dura mater cells and coronal suture mesenchymal cells after DMSO or adezmapimod treatment. d, e, h Quantification of the real-time qPCR of Rbp1 (d) and Aldh1a3 (e, h) expression. f Quantification of RA concentration by ELISA in the medium after culturing dura mater cells. i Quantification of RA concentration by ELISA in the coronal suture mesenchymal cells. Scale bars: 20 μm. Data are mean ± SEM, n = 3. **P < 0.01, calculated by two-tailed unpaired t test (d–f, h, i)

Dura mater synergizes with suture mesenchyme to regulate osteogenic differentiation of coronal suture mesenchymal cells

To evaluate the function of Rbp1 and Aldh1a3 in regulating the osteogenic differentiation of coronal suture mesenchymal cells, we developed co-culture systems for investigation. Since Rbp1 is derived specifically from the dura mater while Aldh1a3 is present in both coronal suture mesenchyme and dura mater, two co-culture systems were used: one to investigate dura mater-derived Rbp1 and Aldh1a3, and a second to investigate coronal suture mesenchyme-derived Aldh1a3.

First, for dura mater-derived Rbp1 and Aldh1a3, the coronal suture mesenchymal cells from Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice were cultured with dura mater CM from control, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+, Rbp1-knockdown Gli1-Cre^ERT2;Fgfr2IIIc^fl/+, or Aldh1a3-knockdown Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice (Fig. S8a). We found increased expression of Rbp1 and Aldh1a3 in dura mater cells of Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice and higher RA secretion in CM from their cultures than in control CM, consistent with our in vivo results (Fig. S8b–d). Meanwhile, the coronal suture mesenchymal cells treated with dura mater CM from Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice showed increased Rarg expression, decreased Gli1 expression, and increased RUNX2+ and SP7+ cell percentages (Fig. S8e–i). Conversely, after treatment with dura mater CM from Rbp1-knockdown or Aldh1a3-knockdown Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, the coronal suture mesenchymal cells showed partially restored expression of the above markers (Fig. S8b–i). These findings indicate that dura mater-derived Rbp1 and Aldh1a3 regulate osteogenic differentiation of coronal suture mesenchymal cells via mediating RA signaling activity.

Next, for coronal suture mesenchyme-derived Aldh1a3, we cultured coronal suture mesenchymal cells from control or Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice and treated them with control siRNA or Aldh1a3 siRNA (Fig. S9a). We found elevated Aldh1a3 expression in coronal suture mesenchymal cells of Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, which was effectively knocked down after Aldh1a3 siRNA treatment (Fig. S9b). Significantly, the Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ coronal suture mesenchymal cells with control siRNA treatment had a stronger RA synthesis ability, and showed increased Rarg expression, decreased Gli1 expression, and increased RUNX2+ and SP7+ cell percentages (Fig. S9c–h). Conversely, with Aldh1a3 siRNA treatment, the coronal suture mesenchymal cells showed restored expression of the above markers (Fig. S9c–h). These findings indicate that coronal suture mesenchyme-derived Aldh1a3 also contributes to regulating osteogenic differentiation of coronal suture mesenchymal cells. Together, results from the two co-culture systems indicate that the dura mater synergizes with the suture mesenchyme to regulate osteogenic differentiation of coronal suture mesenchymal cells by upregulating RA synthesis.

Restoring RA signaling rescues craniosynostosis in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice

To assess whether elevation of RA signaling is causative of craniosynostosis in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, we generated Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ (compound mutant) mice to genetically reduce RA synthesis under FGFR2 overactivation. Heterozygous deletion of Aldh1a3 was utilized to restore RA signaling activity while avoiding potential additional craniofacial abnormalities associated with homozygous mutation of Aldh1a3.^34,36,42,43 Following tamoxifen induction at P3.5, Rarg expression and RA levels in the dura mater and coronal suture mesenchyme were restored in the compound mutant mice, indicating normalization of RA signaling activity (Fig. 6a, c, d). Importantly, this signaling restoration was accompanied by the recovery of GLI1+ progenitors and reduction of SP7+ osteoblast percentage at P6.5, particularly in the suture mesenchyme near the dura mater (Fig. 6a, b, e, f). Notably, the compound mutant mice showed normal coronal sutures, skull morphology, and ICP (n = 20 out of 24, phenotype penetrance = 83.3%) (Fig. 6g–i). These results demonstrate that RA signaling is a key effector downstream of FGFR2 signaling, with its restoration functionally alleviating the consequences of FGFR2 overactivation, preserving GLI1+ progenitors, preventing enhanced osteogenic differentiation, and protecting coronal suture patency.

Fig. 6.

Restoring RA signaling normalizes GLI1+ progenitors and osteogenic differentiation upon FGFR2 overactivation. a RNAscope analyses of Rarg expression (a1–a3) and immunostaining of SP7 expression (a4–a6) in coronal sutures of control, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+, and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ mice. Dotted/dashed lines outline the frontal bone, the parietal bone, and the dura mater. Triangles highlight the gene expression or protein location. b Immunostaining of β-Gal in coronal sutures of Fgfr2IIIc^fl/+;Gli1-LacZ, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Gli1-LacZ, and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+;Gli1-LacZ mice. Triangles highlight the β-Gal+ cells. (c, d) Quantification of RA concentration by ELISA in the dura mater (c) and the coronal suture mesenchyme (d). e Quantification of the SP7+ cell percentage in the coronal suture mesenchyme. f Quantification of the β-Gal+ cell percentage in the coronal suture mesenchyme. g MicroCT analysis of control, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+, and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ mice showing the skull in axial (g1–g3) and sagittal (g1’–g3’) views, coronal suture ortho-slices (g1”–g3”), and H&E staining of the coronal sutures (g1”’–g3”’). Triangles highlight the patent coronal sutures, while the asterisks highlight the fused coronal sutures. h PCA dot plot of skull shape. i ICP measurement over a 60-s period. Scale bars: 20 μm for (a, b), 50 μm for (g1”’–g3”’), 1 mm for (g1–g3, g1’–g3’, g1”–g3”). Data are mean ± SEM, n = 20 in (g) and n = 3 in the other figures. For Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ mice in (g), phenotype penetrance = 83.3% (n = 20 out of 24). *P < 0.05, **P < 0.01, ns: not significant, calculated by one-way ANOVA with Tukey’s multiple comparisons test (c–f)

Collectively, these findings reveal a mechanistic pathway in which FGFR2 overactivation enhances RA synthesis through P38 signaling. This process is achieved by interactions between the coronal suture mesenchyme and dura mater. Ultimately, the elevated RA levels pathologically promote osteogenic differentiation of GLI1+ progenitors, resulting in craniosynostosis.

Neurocognitive functional deficits stem from cranial structural malformation rather than brain-intrinsic abnormalities in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice

Brain development is closely influenced by skull shape and ICP, making it susceptible to deleterious effects secondary to craniosynostosis.^6,44,45 To investigate whether intrinsic Fgfr2 mutation in the brain can affect neurocognitive functions, we first examined the expression of Gli1 and Fgfr2 in the brain. We found that Gli1 and Fgfr2 were co-expressed in cortex, hippocampus, and cerebellum at both early postnatal (P3.5) and adult (P1m) stages (Fig. S10), suggesting that intrinsic Fgfr2 expression in the brain may play a role in neurocognitive functions.

To differentiate the potential neurocognitive impact of craniosynostosis and elevated ICP from that of brain-intrinsic Fgfr2 mutation, we implanted GLI1+ suture mesenchymal stem cells (MSCs) in biodegradable gelatin methacryloyl (GelMA) to regenerate patent coronal sutures in P21.5 Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice⁴⁶ (Fig. 7a, b, c3–c3”’). The transplanted MSCs were GLI1+ suture mesenchymal stem cells isolated from coronal suture mesenchyme of P1m wild-type mice. The GLI1+ cell population in the suture mesenchyme is known to exhibit clonogenicity and multilineage differentiation capacity (osteogenic, chondrogenic, and adipogenic) and functions as a resident progenitor pool supporting cranial suture patency and regeneration.⁴⁶ We performed sham surgery on another group of P21.5 Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice and compared them to controls for any influence of the surgical operation itself (n = 20) (Fig. 7c4–c4”’).⁴⁶ Notably, we found that the Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice with coronal suture regeneration showed restored skull morphology and normalized ICP, while the sham group showed no restoration of either skull morphology or ICP (n = 20) (Fig. 7c, d). Next, we evaluated the neurocognitive performance of control mice, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice without surgery, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice with suture regeneration, and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice with sham surgery. We found the Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice without surgery exhibited notable behavioral deficits in cognitive memory, sociability and social novelty recognition, anxiety-related behaviors, spatial working memory, and motor learning abilities (Fig. 7e–j). Notably, the Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice with coronal suture regeneration showed rescued neurocognitive performance, while the sham group mice showed no restoration (Fig. 7e–j). These results indicate that the neurocognitive impairments we observed predominantly stem from craniosynostosis and the resulting elevated ICP in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice, rather than the direct effects of intrinsic Fgfr2 mutation in the brain.

Fig. 7.

Impaired neurocognitive functions in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice are rescued by MSC implantation and restoration of RA signaling. a Diagram of the MSC implantation surgery. b Surgical images showing exposed coronal suture (b1) and MSC implantation (b2) in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ with suture regeneration group. c MicroCT images of P3.5 m control, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ without surgery, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ with suture regeneration, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ with sham surgery, and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ mice showing the skull in axial (c1–c5) and sagittal (c1’–c5’) views, coronal suture ortho-slices (c1”–c5”), and H&E staining of the coronal sutures (c1”’–c5”’). Triangles highlight the patent coronal sutures, while the asterisks highlight the fused coronal sutures. d ICP measurement over a 60-s period. (e-j) Behavioral assays of the five groups in object recognition memory (e), sociability (f), social novelty recognition (g), anxiety-like behavior (h), spatial working memory (i), and motor learning (j) capabilities. Scale bars: 50 μm for (c1”’–c5”’), 500 μm for (b), and 1 mm for (c1–c5, c1’–c5’, c1”–c5”). Data are mean ± SEM, n = 20, female: male = 10:10. **P < 0.01, ***P < 0.001, ns: not significant, calculated by one-way ANOVA with Tukey’s multiple comparisons test (e–j). Illustration created with BioRender.com for (a)

Furthermore, to test whether restoring RA signaling could alleviate neurocognitive dysfunctions under FGFR2 overactivation, we performed behavioral tests on Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ mice. Notably, these mice with restored skull shape and ICP showed marked improvement in neurocognitive abilities, comparable to the control mice (Fig. 7c5–c5”’, d–j). Our findings indicate that restoring RA signaling in Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ mice functionally rescues neurocognitive performance under FGFR2 overactivation.

Discussion

Temporal sensitivity to growth factor signaling is a fundamental principle of development, with defined windows during which tissues are particularly vulnerable to perturbation.^47–51 However, how such timing governs postnatal cranial suture biology has remained unclear. Here, by investigating Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ and other mouse models, we identify a discrete early postnatal window during which FGFR2 signaling regulates interactions between the suture mesenchyme and underlying dura mater, defining a period of heightened susceptibility to pathological disruption. We further found that the coronal suture is more vulnerable to such disruption than the sagittal or lambdoid sutures are, which aligns with clinical observations that FGFR2 mutations preferentially affect coronal sutures.^3,52,53 This intrinsic heterogeneity among cranial sutures likely reflects differences in their embryonic origin,^22,54,55 progenitor composition, dura mater-suture mesenchyme interactions, and baseline signaling states.

Tissue-tissue interactions are essential for establishing local microenvironments that regulate cell fate. In limb, lung, and craniofacial development, signaling pathways such as RA, BMP, and FGF coordinate morphogenesis.^47,56–58 Importantly, our study identifies the dura mater as an active signaling contributor that promotes osteogenic differentiation of GLI1+ progenitors through an FGFR2-RA signaling axis. Craniosynostosis has been associated with abnormal RA signaling, including teratogenic RA exposure^59–64; dysregulation of the RA-related genes retinaldehyde dehydrogenase 1 (Raldh1), Raldh2, Raldh3, Rbp4,^65,66 RARα,⁶⁷ and dehydrogenase/reductase 3 (DHRS3)^68,69; as well as defects in the RA catabolism gene cytochrome P450 family 26 subfamily b polypeptide 1 (Cyp26b1).^60,70–72 Furthermore, MAPK pathway downstream of FGFR2 signaling contributes to craniosynostosis, as P38 inhibition has been shown to ameliorate skull abnormalities in mice with global FGFR2 overactivation.^11,73 Notably, our study identifies a direct mechanistic link between FGFR2 activation and RA synthesis via P38 signaling, specifically through regulation of Rbp1 in the dura mater and Aldh1a3 in both the suture mesenchyme and dura mater, thereby coupling receptor-mediated signaling to metabolic control of morphogen production. This FGFR2-P38-RA axis highlights a mechanism by which dysregulated dura mater-suture mesenchyme interactions drive craniosynostosis.

Mechanistically, sustained RA signaling promotes premature osteogenic differentiation of GLI1+ progenitors, leading to depletion of the progenitor pool required for suture maintenance. Although the direct transcriptional regulation of Rbp1 and Aldh1a3 remains to be defined, P38 signaling is known to regulate transcription factors including members of the myocyte enhancer factor 2 (MEF2) family and the activating transcription factor/cAMP response element-binding protein (ATF/CREB) family^73–78 through P38-activated kinases like mitogen- and stress-activated kinase 1/2 (MSK1/2),^73–75 regulating downstream genes related to cell differentiation.^76–78 Future studies using chromatin immunoprecipitation or promoter-reporter assays will determine whether these transcription factors directly control RA synthesis genes in the coronal suture.

Importantly, our findings clarify the origin of neurocognitive dysfunction in Fgfr2 mutation-related craniosynostosis. We show that neurocognitive impairments can be rescued by restoring suture patency alone, without directly targeting the brain, indicating that these Fgfr2 mutation-mediated neurocognitive deficits are primarily driven by restricted cranial growth and elevated ICP rather than intrinsic neural abnormalities. This is further supported by evidence from twist family basic helix-loop-helix transcription factor 1 (Twist1) mutant mice (Twist1^+/-), in which rescue of craniosynostosis restores neurocognitive functions,^46,79 as well as clinical evidence showing that timely cranial vault expansion alleviates elevated ICP and improves neurocognitive outcomes.^8,44,45,79 Importantly, although the timing of intervention in craniosynostosis has been subject to debate,^11,80–82 our findings establish a general principle that cranial sutures exhibit temporally defined sensitivity to signaling perturbations, underscoring the importance of timely intervention within defined developmental windows to improve both skull growth and neurocognitive outcomes.

This study defines an FGFR2-P38-RA signaling axis that governs dura mater-suture mesenchyme interactions to regulate GLI1+ progenitor fate and cranial suture integrity within an early postnatal developmental time window. Disruption of this axis drives coronal suture craniosynostosis and in turn leads to impaired neurocognitive functions (Fig. 8). These findings refine the mechanistic understanding of craniosynostosis and highlight early intervention as a key strategy to restore the skull growth trajectory while preserving neurocognitive functions.

Fig. 8.

FGFR2-RA signaling governs dura mater-suture mesenchyme interaction during early postnatal development to control coronal suture development and craniosynostosis. During normal early postnatal development, FGFR2 signaling expression gradually decreases from P3.5 to P7.5, accompanied by a corresponding decline in RA signaling activity within the coronal suture region. Specifically, by controlling P38 signaling activity, physiological FGFR2 signaling induces moderate expression of Rbp1 and Aldh1a3 in the dura mater, as well as Aldh1a3 in the adjacent suture mesenchyme, leading to controlled RA synthesis and RA signaling activity within the local environment. This coordinated signaling reduction as postnatal development progresses ensures balanced RA levels, supporting appropriate osteogenic differentiation of GLI1+ progenitors and preserving their population, thereby maintaining suture patency. However, when FGFR2 signaling is overactivated at P3.5, it disrupts the normal postnatal decline of RA signaling. Instead, upon the regulatory function of increased P38 signaling, RA signaling remains abnormally elevated due to upregulation of Rbp1 and Aldh1a3 in the dura mater and Aldh1a3 in the suture mesenchyme. The resulting excessive RA drives premature and elevated osteogenic differentiation of GLI1+ progenitors, decreasing their population and ultimately leading to craniosynostosis. Tightly regulated FGFR2-RA signaling dynamics during a critical postnatal time window extending through P3.5, but closing before P7.5, governs the proper interactions between dura mater and coronal suture mesenchyme, disruption of which leads to craniosynostosis. Illustration created with BioRender.com

Materials and methods

Study approval

All animal experiments were performed following a protocol approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Southern California. Mice were housed in pathogen-free conditions under a 12-h light/dark cycle. All procedures followed institutional and federal guidelines for the care and use of laboratory animals.

Animals

Fgfr2^tm2.2Dsn/SbelJ (JAX no. 033934), Gli1^{tm3(cre/ERT2)Alj}/J (JAX no. 007913), tdTomato conditional reporter (JAX no. 007905), and Gli1-LacZ reporter (JAX no. 008211) mice were purchased from the Jackson Laboratory (Bar Harbor, Maine, USA). The Aldh1a3^fl/+ mouse strain was a kind gift from Dr. Jinsook Son and Dr. Domenico Accili at Columbia University (New York, USA).³⁵ To generate Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ (Fgfr2IIIc mutant) mice, we crossed Gli1^{tm3(cre/ERT2)Alj}/J mice with Fgfr2^tm2.2Dsn/SbelJ mice. To generate Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ mice, we crossed Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice with Aldh1a3^fl/+ mice. In this study, the genotypes of control mice were Fgfr2IIIc^fl/+ or Fgfr2IIIc^fl/fl. The tdTomato conditional reporter mice and Gli1-LacZ reporter mice were crossed with control, Gli1-Cre^ERT2;Fgfr2IIIc^fl/+, and Gli1-Cre^ERT2;Fgfr2IIIc^fl/+;Aldh1a3^fl/+ mice as needed. Prior to tamoxifen induction, the Gli1-Cre^ERT2;Fgfr2IIIc^fl/+ mice at P3.5 were indistinguishable from control littermates.

Tamoxifen (TMX) (Sigma-Aldrich, St. Louis, Missouri, T5648) was used for Cre^ERT2 activation. It was dissolved in corn oil (Sigma-Aldrich, C8267) at 20 mg/mL and injected intraperitoneally at 1.5 mg per 10 g body weight on P3.5 or P7.5. The mice were euthanized by carbon dioxide overdose followed by decapitation (older than P6.5) or directly decapitated (P6.5 or younger).

The mice were identified by ear tags (P2w or older) or paw tattoos (younger than P2w) using the NEO-9 Neonate Rodent Tattooing System (AIMS Lab Animal Tattoo Identification, Elk Grove Village, Illinois, USA). Genotyping was conducted as described.⁴⁶ Briefly, ear or tail biopsies were lysed overnight at 55 °C in DirectPCR solution (Viagen 102T, Viagen Biosciences, Los Angeles, California, USA) and heat-inactivated at 85 °C for 1 h. PCR-based genotyping was performed using GoTaq Green Master Mix (Promega, Madison, Wisconsin, USA) on a C1000 Touch Cycler (Bio-Rad, Hercules, California, USA).

Our study examined both male and female mice across all experimental groups. No sex-specific differences were observed in phenotype or molecular outcomes. Therefore, data from male and female animals were pooled for analysis.

Author contributions

L.G. and Y.C. conceptualized the study and designed the experiments. L.G., P.C., J.F., T.G., and M.Z. performed experiments and acquired data. L.G., J.F., T.G., and M.Z. carried out data analysis. L.G. drafted the manuscript. J.F., T.G., M.Z., J.C., and Y.C. contributed to manuscript editing and revision. Y.C. supervised the project. Y.C. and J.C. provided funding.

Data availability

All supporting data are available in the article and the supplementary information file. The RNA-seq datasets have been deposited in the FaceBase Consortium Data Hub (10.25550/8B-9CNW). Additional raw data are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41368-026-00449-0.