Dysregulated PDGFRα signaling alters coronal suture morphogenesis and leads to craniosynostosis through endochondral ossification

Bone formation is a complex process in which mesenchymal cells become osteoblasts and deposit bone matrix. There are generally two pathways to form bone. The mesenchymal cells may first form a cartilaginous template that is later replaced by osteoblasts that mineralize into bone, as hypertrophic chondrocytes in the growth plate are thought to be an end state of differentiation. This pathway of endochondral ossification occurs in the axial and appendicular skeleton (Kronenberg, 2003; Ornitz, 2005). Recent studies have however challenged this traditional concept of lineage segregation, as hypertrophic chondrocytes in long bones have been shown by lineage tracing to be able to transdifferentiate into osteoblasts (Yang et al., 2014a,b; Zhou et al., 2014). Alternatively, mesenchymal cells can aggregate and directly differentiate into osteoblasts. This pathway of intramembranous ossification is observed in the formation of the flat bones in the skull, collarbones and jaw bones (Hall and Miyake, 2000).

The skull is composed of the neurocranium and the viscerocranium (facial skeleton). The neurocranium plays an important role in encasing and protecting the brain and other sense organs. Both endochondral ossification and intramembranous ossification occur during neurocranium formation. The base of the neurocranium forms through endochondral ossification (McBratney-Owen et al., 2008), whereas the vault (calvarium) is thought to develop exclusively through intramembranous ossification (Hall and Miyake, 2000). The calvarium is composed of paired frontal bones, paired parietal bones, the squamous region of the occipital bone, and the sutures, which are the fibrous tissues that separate and joint the calvarial bones. Lineage-tracing studies show that the frontal bones are derived from the neural crest, the parietal bones originate from the paraxial mesoderm, while the coronal sutures that separate these bones are heterogeneous but derive mostly from mesoderm cells (Deckelbaum et al., 2012; Jiang et al., 2002; Lenton et al., 2005; Yoshida et al., 2008). During mouse development, the progenitors of sutures and calvarial bones arise at around embryonic day (E) 11 in the supraorbital regions. These cells undergo active proliferation and differentiation while migrating towards their dorsal destination. They develop coordinately in close proximity to the adjacent dura, a thick membrane that surrounds the brain, and surrounding mesenchymal cells through tissue-tissue interactions (Deckelbaum et al., 2012; Opperman, 2000). In addition, the sutures also serve as a growth site and stem cell niche to support continuous growth of calvarial bones after birth (Zhao et al., 2015).

Deformities of the calvarial bones or sutures usually lead to premature fusion of calvarial bones, or craniosynostosis, a common human birth defect with an occurrence of 1/2500 (Senarath-Yapa et al., 2012; Wilkie and Morriss-Kay, 2001). Craniosynostosis may occur as part of a syndrome or more frequently as isolated (non-syndromic) anomalies (Cohen, 2000). Among the genetic alterations identified in humans, mutations are prevalent in genes encoding receptor tyrosine kinases (RTKs), their ligands or downstream signaling components (Johnson and Wilkie, 2011). These include fibroblast growth factor receptor 2 and 3 (FGFR2 and FGFR3), ephrin B1 (EFNB1) and KRAS. Of these, mutations in FGFR2 are most frequent and account for 32% of all cases (Johnson and Wilkie, 2011). Such mutations lead to activation of FGFR2 and frequently cause Apert syndrome, which is characterized by bi-coronal synostosis and other defects (Ibrahimi et al., 2005). These studies suggest that RTKs and their common signaling targets may be involved in the normal development of calvaria and in pathogenesis.

Platelet-derived growth factor receptors (PDGFRs) constitute another family of RTKs. In mouse and human, there are two PDGFRs (α and β) and four PDGF ligands (A-D) (Hoch and Soriano, 2003). While both receptors are required for normal development and homeostasis (Andrae et al., 2008), PDGFRα is essential in neural crest cell (NCC) development through its downstream PI3K/AKT and Rac1 signaling pathways (Fantauzzo and Soriano, 2014, 2016; He and Soriano, 2013; Klinghoffer et al., 2002; Soriano, 1997; Tallquist and Soriano, 2003). Of the four ligands, PDGFA and PDGFC are specific endogenous activators of PDGFRα, and inactivating both ligands causes embryonic lethality at mid-gestation and severe craniofacial malformations, phenocopying Pdgfra null mutants (Ding et al., 2004; Soriano, 1997). It has been reported that both the expression level and activity of PDGFRα are significantly increased in Apert osteoblasts (Miraoui et al., 2010), and generalized activation of PDGFRα signaling has been shown to lead to craniosynostosis (Moenning et al., 2009).

In the present study, we show that PDGFRα and its ligand PDGFA are expressed in the developing calvarial tissues of mice. Expression of an activatable form of PDGFRα knocked into its own locus leads to abnormal ossification and to craniosynostosis of the coronal sutures in neonates. Interestingly, we find that enhanced PDGFRα signaling activity causes abnormal expansion and premature differentiation of cartilage anlagen underlying the coronal sutures, and that endochondral ossification contributes directly to calvarial morphogenesis. Lineage tracing reveals that these chondrocytes derive from both neural crest and mesoderm cells, and activating PDGFRα signaling in chondrocytes, neural crest or mesoderm causes distinct calvarial phenotypes. Further analysis reveals that both PI3K and PLCγ signaling activity are elevated in NCCs expressing constitutively active PDGFRα, and that PDGFRα/PI3K signaling is required for the normal development of coronal sutures and frontal bones.

To examine the consequence of dysregulated PDGFRα activity in calvarial development, we drove expression of a conditionally activatable D842V Pdgfra allele utilizing its endogenous promoter by crossing Pdgfra^+/K knock-in mice (Olson and Soriano, 2009) with the general epiblast deleter line Meox2Cre (Tallquist and Soriano, 2000). Because Pdgfra^+/K; Meox2Cre mice exhibit perinatal lethality (Olson and Soriano, 2009), we analyzed skeletal preparations at E18.5.

In wild-type embryos, the frontal bones and the parietal bones remained separated by coronal sutures (Fig. 1A,C). In Pdgfra^+/K; Meox2Cre embryos, however, these bones were fused and the coronal sutures were absent (Fig. 1B,D). Alkaline phosphatase (AP) staining of sagittal sections through these embryos highlighted differentiating osteoblasts and revealed that, in wild-type embryos, the frontal and parietal bones are separated by AP-negative coronal sutures (Fig. 1E). In Pdgfra^+/K; Meox2Cre embryos, the frontal and parietal bones were fused, as indicated by continuous AP staining across both bones, which also appeared thicker than in the wild-type controls (Fig. 1E,F).

These results indicate that PDGFRα activity needs to be tightly regulated for normal calvarial osteogenesis and coronal suture development, and that increased PDGFRα activity leads to craniosynostosis of the coronal suture.

Using in situ hybridization and immunohistochemistry, we examined the expression pattern of Pdgfra and its endogenous ligands Pdgfa and Pdgfc. The rudiments of the frontal bones and the parietal bones are composed of differentiating osteoblasts, which we identified using AP and RUNX2 as markers (Fig. 2A,D,H,K). Pdgfa and Pdgfc exhibited differential expression patterns in developing calvarial tissues. Pdgfa mRNA was expressed in a similar domain to that of AP in the rudiments of the frontal and parietal bones at E13.5 and E15.5 (Fig. 2B,E). Pdgfc mRNA was expressed in the ectoderm, but was barely detected in the bone rudiments (Fig. 2C,F). Pdgfra expression, as detected by EGFP expression in Pdgfra^GFP/+ knock-in embryos, was identified in the mesenchymal cells surrounding RUNX2⁺ osteoblasts at E13.5 (Fig. 2G-I). At E15.5, Pdgfra expression overlapped with RUNX2 in some osteoblasts as well as in coronal suture cells (Fig. 2J-L).

Together with the craniosynostosis phenotype observed in Pdgfra^+/K; Meox2Cre embryos (Fig. 1), these results are consistent with a role for PDGFA-PDGFRα signaling during calvarial development and coronal suture morphogenesis.

To trace the onset of the Pdgfra^+/K; Meox2Cre craniosynostosis phenotype, we analyzed sections across the coronal sutures at different stages. At E13.5, AP activity became detectable in the differentiating osteoblasts within the frontal and parietal bone rudiments, which were separated by the AP-negative coronal sutures through E15.5 (Fig. 3A,C,E). By contrast, the coronal sutures of Pdgfra^+/K; Meox2Cre embryos showed AP activity by E15.5 (Fig. 3B,D,F), suggesting that these sutural cells are transitioning to osteoblasts. In addition, the calvarial rudiments in Pdgfra^+/K; Meox2Cre embryos became thicker than in wild-type counterparts at E15.5. The AP-positive domain exhibited sharp edges flanking AP-negative tissues underneath the coronal sutures (Fig. 3F). Alcian Blue (AB) staining of alternate sections indicated that the AP-negative and AB-positive cells were cartilage cells, which were also observed in the control but significantly more so in Pdgfra^+/K; Meox2Cre embryos starting at E13.5 (Fig. 3G-L). BrdU labeling results showed proliferation rates for mutant chondrocytes (63%) that trended higher than those for controls (53.6%; Fig. S1, n=9, P<0.01).

To confirm the identity of these cartilage cells, we analyzed the expression of a series of molecular markers. In situ hybridization results showed that Sox9 and Col2a1 were expressed in these cells, indicating that they were chondrocytes and that many more chondrocytes were formed in Pdgfra^+/K; Meox2Cre embryos (Fig. 4A-D). Expression of the hypertrophic chondrocyte marker Col10a1 was also detected, suggesting that these chondrocytes underwent differentiation in both wild-type and Pdgfra^+/K; Meox2Cre embryos (Fig. 4E,F). Expression of the osteoblast marker Ocn (Bglap) in the Pdgfra^+/K; Meox2Cre cartilage indicates that these cells begin to differentiate into osteoblasts (Fig. 4G,H). In line with this result, immunohistochemistry showed that the osteoblast markers RUNX2 and osterix (SP7) were expressed in the chondrocytes of Pdgfra^+/K; Meox2Cre, but not wild-type, embryos (Fig. 4I-L), indicating that activation of PDGFRα signaling promotes endochondral ossification. By E18.5, the chondrocytes underlying the parietal bony rudiments were replaced by AP-positive osteoblasts, although some chondrocytes underlying the frontal bone remained (Fig. 4M,N).

These results indicate that PDGFRα regulates the timing and extent of endochondral ossification during calvarial development, and that constitutive PDGFRα activation promotes chondrocyte formation and subsequent differentiation into osteoblasts.

To confirm our observation that endochondral ossification contributes to calvarial morphogenesis in Pdgfra^+/K; Meox2Cre embryos, we used Col2a1Cre mice, which express Cre recombinase specifically in chondrocytes (Sakai et al., 2001), to fate map these cells. Lineage tracing with the R26R^lacZ reporter line (Soriano, 1999) showed that at E15.5 the lacZ reporter was expressed in both the frontal cartilage and the perichondrium underneath the coronal sutures (Fig. 5A,C). At postnatal day (P) 20, chondrocyte-derivatives were detected in the nasal bones and tissues at the junctions between calvarial bones, including the coronal sutures (Fig. 5E). Compared with littermate controls, Pdgfra^+/K; Col2a1Cre embryos exhibited many more chondrocyte-derived cells (Fig. 5B,D) and ectopic chondrocyte-derived tissues adjacent to the coronal sutures at E15.5. At E18.5, skeletal preparations showed enlargement of Pdgfra^+/K; Col2a1Cre skulls (Fig. S2). The total length of Pdgfra^+/K; Col2a1Cre skulls was 6.9±1.2% (P<0.01) larger than those of the wild type, the mutant frontal bones were 11.7±2.5% (P<0.01) longer, and the mutant parietal bones were 3.1±3% (P>0.05) longer than in the control (n=5). At P20, Pdgfra^+/K; Col2a1Cre mice accumulated more chondrocyte derivatives in the nasal bones and calvarial junctions (Fig. 5E,F) and exhibited a domed skull, indicating abnormal calvarial development (Fig. 5G,H). Skeletal preparations revealed abnormal morphology of coronal sutures in Pdgfra^+/K; Col2a1Cre mice compared with control littermates (Fig. 5I,J). AP staining of sagittal sections of these skulls revealed that Pdgfra^+/K; Col2a1Cre frontal and parietal bones were thinner than in controls (Fig. 5K,L). Further analysis using the R26R^tdT reporter (Madisen et al., 2010) for lineage tracing revealed that many cells formerly expressing Col2a1Cre in Pdgfra^+/K; Col2a1Cre mice were RUNX2 positive and aligned with calvarial osteoblasts (Fig. 5M,N).

These results indicate that ectopic PDGFRα activity in the chondrocyte lineage disrupts normal calvarium development, but does not cause craniosynostosis by P20.

The calvarial bones are derived from two different origins: the neural crest and the mesoderm (Chai et al., 2000; Jiang et al., 2000; Yoshida et al., 2008). To examine in which lineage overactive PDGFRα signaling causes craniosynostosis, Pdgfra^+/K mice were crossed to the Wnt1Cre-2 (Lewis et al., 2013) or Mesp1Cre (Saga et al., 1999) drivers. Lineage tracing showed that the cartilage underlying the coronal suture was derived from both neural crest and mesodermal cells (Fig. S3). Constitutive activation of PDGFRα in NCCs with Wnt1Cre-2 led to perinatal lethality. Skeletal preparations of E18.5 Pdgfra^+/K; Wnt1Cre-2 embryos showed ectopic cartilage formation at the interfrontal, sagittal and coronal sutures (Fig. 6A-D). Lineage tracing revealed that in Pdgfra^+/K; Wnt1Cre-2; R26R mice, the neural crest domain expanded posteriorly compared with control littermates (Fig. 6E,F). Constitutive activation of PDGFRα in mesoderm cells with Mesp1Cre, on the other hand, did not cause embryonic lethality. Pdgfra^+/K; Mesp1Cre mice survived beyond P20 (n=5) and showed a domed skull (data not shown). Skeletal preparations of Pdgfra^+/K; Mesp1Cre mice revealed craniosynostosis of the coronal and lambdoid sutures (Fig. 6G,H). At E16.5, ectopic cartilage was observed posterior to the parietal bone (Fig. 6I,J), and lineage tracing did not detect a displacement of the mesoderm boundary into the anterior skull, although structures resembling vascular plexus appeared increased in the anterior skull (Fig. 6K,L, arrows).

These data demonstrate that activating PDGFRα signaling in two distinct lineages leads to different outcomes during calvarial morphogenesis, and that the phenotype observed in Pdgfra^+/K; Meox2Cre embryos could be caused by the synergistic effect of overactive PDGFRα activity in both lineages. In addition, the ectopic cartilage at the suture observed using Wnt1Cre-2 indicates that PDGFRα signaling in NCCs is an important contributor in coronal suture morphogenesis.

PDGFRα engages the phosphorylation and activation of multiple downstream signaling pathways, including the SRC, ERK1/2 (MAPK3/1), STAT, PI3K/AKT and PLCγ pathways. To identify the mechanisms of PDGFRα regulation in NCCs, we assayed the activity of these signaling pathways in primary cultures prepared from E11.5 frontonasal prominence cells, which are predominantly derived from neural crest.

Western blot analysis revealed that AKT and PLCγ phosphorylation levels are significantly increased in the NCCs of Pdgfra^+/K; Wnt1Cre-2 embryos relative to control littermates (Fig. 7A; 1.45±0.14-fold increase in phospho-AKT and 2.98±0.13-fold increase in phospho-PLCγ), as observed previously in E13.5 Pdgfra^+/K; Meox2Cre whole-embryo protein lysates (Olson and Soriano, 2009), whereas the activity of SRC, ERK1/2 and STAT3 did not change (Fig. S4). Immunostaining showed increased phospho-AKT and phospho-PLCγ staining in Pdgfra^+/K; Meox2Cre embryos compared with control littermates at E13.5 (Fig. 7B-E).

PI3K signaling has been established as a crucial mediator of PDGFRα signaling during craniofacial and skeletal development (Fantauzzo and Soriano, 2014, 2016; He and Soriano, 2013; Klinghoffer et al., 2002; Pickett et al., 2008). We therefore asked whether activation of PI3K/AKT signaling in NCCs causes abnormal calvarial development. Our results showed that disruption of the PI3K/AKT inhibitor Pten in NCCs caused enlargement of the head and calvarial bones (Fig. 7F,G) and perinatal lethality (data not shown; n=3). AP and AB staining of Pten^fl/fl; Wnt1Cre-2 embryos at E14.5 revealed enhanced cartilage formation underlying the coronal sutures, and increased AP signal at the coronal sutures and tissues surrounding the cartilage (Fig. 7H,I). Inactivating Pten in the whole embryo led to embryonic lethality at E14.5 accompanied by edema, as also observed in Pdgfra^+/K; Meox2Cre embryos (Olson and Soriano, 2009) (Fig. 7J,K). AP staining of the coronal sections showed that the coronal sutures were partially closed and the underlying cartilage was expanded (Fig. 7L,M). By contrast, Pdgfra^PI3K/PI3K mutant embryos, in which the ability of PDGFRα to activate PI3K signaling is abrogated (Klinghoffer et al., 2002), displayed a wider separation of the frontal and parietal bones at the coronal sutures, and a foramen between the two frontal bones (Fig. 7N,O). AP and AB staining showed hypoplastic cartilage formation in the Pdgfra^PI3K/PI3K embryos at E15.5 (Fig. 7P,Q).

Taken together, these results indicate that PI3K/AKT signaling plays an essential role in calvarial development and coronal suture morphogenesis.

Craniosynostosis is a prevalent human birth defect that arises from genetic or environmental disruptions of normal calvarium development (Wilkie and Morriss-Kay, 2001). Our studies establish PDGFRα and its downstream signaling pathways as crucial regulators of calvarial growth. The calvarial bones are derived from two separate lineages: the neural crest and the mesoderm. Our data showed that increased activity of PDGFRα in both lineages causes the abnormal formation of sutures, as well as expansion and premature ossification of the underlying cartilage. We have identified the origin of these cartilage anlagen, and further demonstrate by marker analysis and lineage tracing that, upon activation of PDGFRα signaling, cartilage contributes to calvaria morphogenesis through endochondral ossification. We found that this process was under the regulation of PDGFRα-PI3K/AKT signaling in the neural crest lineage. Our study also showed that activating PDGFRα signaling in either the neural crest or the mesoderm caused distinct phenotypes, indicating a synergistic role of signaling in both lineages during calvaria development and craniosynostosis.

A role for PDGFRα in craniosynostosis was first suggested by reports showing its increased expression and activity in the osteoblasts from patients with Apert syndrome (Miraoui et al., 2010), as well as premature fusion of the interfrontal and coronal sutures by constitutive activation of PDGFRA in the neural crest (Moenning et al., 2009). Although this particular phenotype is different from that we observe at the coronal suture only, it is likely that this is due a difference in the domain or level of expression of the activated PDGFRα, as the earlier study used constitutive expression from the ROSA26 locus instead of restricted expression from the endogenous Pdgfra locus. Consistent with this interpretation, conditional activation of PDGFRα from these two alleles throughout the epiblast or in NCCs led to different phenotypic outcomes (Kurth et al., 2009; Moenning et al., 2009; Olson and Soriano, 2009). Additional differences in the previous study include the use of a human PDGFRA cDNA, which might signal differently than a mouse cDNA, and the use of the original Wnt1Cre mice (Danielian et al., 1998) that misexpress Wnt1, instead of the Wnt1Cre-2 driver (Lewis et al., 2013). This might confound some results, as ectopic Wnt signaling activity has been shown to affect normal calvarium development and suture morphogenesis (Liu et al., 2007; Maruyama et al., 2010; Mirando et al., 2010). These differences notwithstanding, both our study and the previous report (Moenning et al., 2009) support an important role for PDGFRα signaling in the regulation of cranial sutures and craniosynostosis. We note, however, that PDGFRα activation has not been implicated in human patients with craniosynostosis, possibly because PDGFRα activation leads to very pleiotropic phenotypes (Olson and Soriano, 2009).

RTKs regulate cell fates during development and disease by engaging a number of shared intracellular signaling pathways. Among these, the involvement of the ERK1/2 pathway has been well documented in animal models of craniosynostosis. For example, mice carrying active mutations of Fgfr2 exhibit craniosynostosis correlated with enhanced ERK1/2 signaling activity in the suture cells and osteoprogenitors (Wang et al., 2005; Yin et al., 2008), and inhibiting ERK1/2 phosphorylation or engagement can rescue the craniosynostosis phenotype (Shukla et al., 2007; Eswarakumar et al., 2006). Increased ERK1/2 signaling activity has also been identified in the calvarial bone marrow of Twist1 mutant mice (Connerney et al., 2008). Interestingly, decreased ERK1/2 signaling activity is also observed in Epha4^−/− and Twist1^+/− craniosynostosis models (Ting et al., 2009). These studies indicate that ERK1/2 signaling must be tightly regulated at multiple stages and in different tissues during calvarium development. Another RTK downstream signaling pathway, PLCγ, is activated with increased PDGFRα activity and has been implicated in osteogenesis of neural crest-derived osteoprogenitors (Moenning et al., 2009; Olson and Soriano, 2009). Our work highlights the role of PI3K/AKT signaling in normal calvarium development and its pathogenesis. In a previous study, we have reported that PDGFAA induces AP activity in mouse embryonic palatal mesenchymal cells via PI3K signaling (Vasudevan et al., 2015). In the present work, we show that PDGFRα-engaged PI3K/AKT signaling regulates development of the coronal sutures and underlying cartilage, and dysregulated PI3K signaling activity causes ossification of suture cells and expansion of chondrocytes. Together, these data reveal an important role for PI3K/AKT signaling in calvarial development and disease.

Bone can be formed by either endochondral ossification or intramembranous ossification, and it is generally believed that calvaria form only through intramembranous ossification. However, cartilage has also been implicated in calvarial ossification. The parietal cartilage has been shown to be remodeled into membranous bone through a non-endochondral ossification pathway regulated by the membrane type 1 matrix metalloproteinase (MT1-MMP; also known as MMP14). In 5-day-old mice, these parietal chondrocytes express the chondrocyte marker Col2a1 but not the hypertrophic chondrocyte marker Col10a1, and finally undergo apoptosis and are replaced by bone cells by day 10 (Holmbeck et al., 2003). In the present study, we show the existence of cartilage anlagen underlying the coronal sutures derived from both the neural crest and mesoderm (Fig. S3). These cells express Col2a1, Sox9 and Col10a1, and exhibit AP activity in E18.5 Pdgfra^+/K; Wnt1Cre-2 embryos, indicating that they contribute to the calvaria via the hypertrophic chondrocyte pathway. Interestingly, this pathway does not necessarily follow the traditional endochondral ossification process, since we show by lineage tracing using a chondrocyte-specific Col2a1 Cre driver that descendants of these cartilage cells are incorporated into calvarial osteoblasts and suture cells at P20 (Fig. 5N). This observation is in line with an increasing number of recent studies showing that chondrocytes are not necessarily terminally differentiated, but can instead be transformed into osteoblasts during long-bone development (Enishi et al., 2014; Park et al., 2015; Yang et al., 2014a,b; Zhou et al., 2014). We suggest that this alternate differentiation pathway has also been redeployed in calvarial osteogenesis, at least in part under the regulation of PDGFRα. In addition, identification of Col2a1Cre lineage-traced cells in the coronal suture suggests that chondrocyte progenitors might be multipotent, since the sutures not only function as connective tissues but also provide growth sites and signaling centers of the growing calvaria (Lana-Elola et al., 2007; Morriss-Kay and Wilkie, 2005; Opperman, 2000). A recent study has identified the suture as a mesenchymal stem cell niche for calvarial bone homeostasis and repair (Zhao et al., 2015). Another recent report supports this notion by showing that hypertrophic chondrocytes express pluripotency marker genes and are involved in fracture healing (Hu et al., 2017). In addition, since the perichondrium is the major source of osteoblast precursors, expansion of this tissue primordium might contribute to the enhanced growth of calvarial bones in Pdgfra^+/K; Meox2Cre mice.

Compared with the endochondral ossification that occurs in normal posterior frontal suture closure (Moss, 1958; Sahar et al., 2005), dysregulated growth factor signaling pathways have been shown to cause ectopic cartilage formation in the anterior frontal, sagittal and lambdoid sutures. For example, ectopic cartilage formation at the sagittal sutures has been reported in mouse models including Fgfr2^S252W (Holmes and Basilico, 2012; Wang et al., 2005) and Axin2^−/−; Fgfr1^+/− mutants (Maruyama et al., 2010). Inactivating Gli3 in mice causes ectopic cartilage in the lambdoid suture (Rice et al., 2010). Unlike these models, Pdgfra^+/K; Wnt1Cre-2 embryos exhibit ectopic cartilage in the coronal sutures (Fig. 6A-D). The prominent cartilage phenotype in Pdgfra^+/K; Wnt1Cre-2 could be caused by the persistence and expansion of cartilage that forms transiently within the sutures; alternatively, increased PDGFRα activity might direct mesenchymal stem cell differentiation within the sutures. The fact that we could not phenocopy the ectopic sutural cartilage observed in Pdgfra^+/K; Wnt1Cre-2 mice using a chondrocyte Cre driver (Fig. S2) supports this latter possibility. This notion is further supported by transcriptome analyses showing that PDGF signaling may be involved in directing mesenchymal stem cell differentiation towards a chondrogenic lineage (Ng et al., 2008). These observations, together with other reports that examine the effects of PDGFR activation in fibro/adipogenic progenitors or multipotent perivascular cells, underscore the need for tight regulation of PDGFR activity in regulating the homeostatic state of various mesenchymal progenitor cells (Iwayama et al., 2015; Olson and Soriano, 2009, 2011). Together, our findings establish a crucial role for PDGFRα in craniosynostosis, and identify endochondral ossification as a mechanism that can operate during calvarial morphogenesis.

All animal experiments were approved by the Institutional Animal Care and Use Committee at Icahn School of Medicine at Mount Sinai. The following strains were maintained on a C57BL/6J background: Pdgfra^tm12Sor (Olson and Soriano, 2009), referred to as Pdgfra^+/K in the text; Pdgfra^{tm11(EGFP)Sor} (Hamilton et al., 2003), referred to as Pdgfra^+/GFP; Gt(ROSA)26Sor^tm1Sor (Soriano, 1999), referred to as R26R^lacZ; and Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze} (Madisen et al., 2010), referred to as R26R^tdT. All other strains were maintained on C57BL/6J; 129SvJaeSor (MGI:3044540) mixed genetic backgrounds: Pdgfra^tm8Sor (Tallquist and Soriano, 2003), referred to as Pdgfra^+/fl; Pdgfra^tm5Sor (Klinghoffer et al., 2002), referred to as Pdgfra^+/PI3K; Meox2^tm1(cre)Sor (Tallquist and Soriano, 2000), referred to as Meox2Cre; Tg(Wnt1-cre)2Sor (Lewis et al., 2013), referred to as Wnt1Cre-2; Mesp1^tm2(cre)Ysa (Saga et al., 1999), referred to as Mesp1Cre; and Pten^tm1Hwu (Groszer et al., 2001), referred to as Pten^+/fl.

Staged embryos were sacrificed, dissected in PBS and their skin and internal organs removed. Following incubation in 95% ethanol overnight at room temperature, skeletons were stained with 0.015% Alcian Blue (Sigma, A5268), 0.005% Alizarin Red (Sigma, A5533), 5% glacial acetic acid in 70% ethanol for 72 h at 37°C. The skeletons were then washed briefly in 95% ethanol, cleared with 1% KOH, and subsequently in increasing concentrations of glycerol in KOH until reaching 80% glycerol. For adult mice, the skin was removed using a scalpel, and the cranial base and brain were dissected before staining. Staining and clearing were performed with longer incubation times using the same protocol as for embryos.

For whole-mount staining, the ectoderm and superficial dermis were carefully peeled off, and embryos were fixed and stained with X-gal (Soriano, 1999). X-gal staining on cryosections was performed using standard procedures. For Alcian Blue staining, sections were quickly dipped in 1% Alcian Blue 8GX (A5268, Sigma) in 0.1 N HCl. For alkaline phosphatase (AP) staining, paraffin sections were deparaffinized in Histoclear (Fisher Scientific, 50-899-90147), rehydrated in decreasing ethanol solutions, and incubated in a solution containing 0.03% nitro-blue tetrazolium chloride (NBT; Roche, 11383213001) and 0.02% 5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt (BCIP; Roche, 11383221001) to detect AP activity.

In situ hybridization was performed as previously described (He and Soriano, 2013). Pdgfa and Pdgfc probes were provided by Andras Nagy (Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital, Toronto), Sox9, Col2a1 and Col10a1 probes originate from YiPing Chen (Tulane University, New Orleans), and the Ocn probe was from Fanxin Long (Washington University, St Louis). Immunostaining was performed according to a standard protocol and antibodies to anti-phospho-AKT (1:25, Cell Signaling Technology 9271), anti-phospho-PLCγ1 (1:400, Novus Biologicals, MAB74542), RUNX2 (1:1000, Cell Signaling Technology, 12556) and osterix (SP7) (1:5000, Abcam, ab94744).

BrdU/PBS solution was administered by intraperitoneal injection into pregnant females at 3 mg/100 g body weight. Embryos were dissected in ice-cold PBS 1 h after injection. Embryonic heads were fixed in Carnoy's fixative for 1 h and dehydrated and embedded in paraffin. Immunostaining was performed on 10 μm transverse sections using the BrdU Labeling and Detection Kit II (Roche, 11299964001) following the manufacturer's instructions. Sections were counterstained with nuclear Fast Red (Sigma, N8002) to facilitate quantification. Results from three embryos of each phenotype were collected. Statistical data are presented as mean±s.e.m., and subjected to two-tailed Student's t-tests.

Protein lysates were prepared from primary cultures of frontonasal prominences dissected from E11.5 embryos in RIPA buffer. Western blot was carried out using the following primary antibodies at 1:1000: anti-p44/42 MAPK (9102), anti-phospho-p44/42 MAPK (9101), anti-AKT (9272), anti-phospho-AKT (Ser473; 9271), anti-PLCγ (2822) and anti-phospho-PLCγ (2821) from Cell signaling Technology; anti-SRC (44-655G) and anti-phospho-SRC (44-660G) from Life Technologies; and anti-STAT3 (sc-8019) and anti-phospho-STAT3 (sc-8059) from Santa Cruz Biotechnology. Primary antibodies were detected by horseradish peroxidase-conjugated secondary antibodies (1:10,000; Jackson ImmunoResearch Laboratories). Western blot results were quantified using ImageJ from NIH.

We thank Tianfang Yang, Tony Chen and Matthew Hung for excellent genotyping work; Francesco Ramirez for Col2a1Cre mice and Greg Holmes for the anti-osterix antibody; and our laboratory colleagues and Greg Holmes for constructive comments on the manuscript.