Dlx5-augmentation in neural crest cells reveals early development and differentiation potential of mouse apical head mesenchyme

Neural crest cells (NCCs) give rise to various tissues including neurons, pigment cells, bone and cartilage in the head. Distal-less homeobox 5 (Dlx5) is involved in both jaw patterning and differentiation of NCC-derivatives. In this study, we investigated the differentiation potential of head mesenchyme by forcing Dlx5 to be expressed in mouse NCC (NCCDlx5). In NCCDlx5 mice, differentiation of dermis and pigment cells were enhanced with ectopic cartilage (ec) and heterotopic bone (hb) in different layers at the cranial vertex. The ec and hb were derived from the early migrating mesenchyme (EMM), the non-skeletogenic cell population located above skeletogenic supraorbital mesenchyme (SOM). The ec developed within Foxc1+-dura mater with increased PDGFRα signalling, and the hb formed with upregulation of BMP and WNT/β-catenin signallings in Dermo1+-dermal layer from E11.5. Since dermal cells express Runx2 and Msx2 in the control, osteogenic potential in dermal cells seemed to be inhibited by an anti-osteogenic function of Msx2 in normal context. We propose that, after the non-skeletogenic commitment, the EMM is divided into dermis and meninges by E11.5 in normal development. Two distinct responses of the EMM, chondrogenesis and osteogenesis, to Dlx5-augmentation in the NCCDlx5 strongly support this idea.

. In mice, the calvarium and chondrocranium malformations have been shown to associate with Dlx5-downregulation 7,22,23 . Meanwhile, cranial base cartilages derived from NCC are enlarged by Dlx5-overexpression, but calvarial bones have not been examined 11 . In calvarial development, formation of the frontal and parietal bones start with the aggregation of mesenchymal cells in the area of the supraorbital ridge at embryonic day (E) 10.5 3 , referred to as the supraorbital ridge mesenchyme 5,[24][25][26] or supraorbital mesenchyme (SOM) 27 . The SOM proliferates and differentiates into osteoblasts from E11.5, then intrinsically expands to the apex of the head to form the bone from E13.5 4,5 . Importantly, due to the intimate association and mutual support of cranial bones and the dura mater, the defects in the dura mater affect calvarial bone formation and maintenance [28][29][30][31] . Before the SOM begins apical growth, a population of head mesenchyme, termed as early migrating NCC 32 or early migrating mesenchyme (EMM) 27 , is established above the SOM to contribute to the sutures or soft tissue layers such as the dermis and the meninges 4,25,32 . Transcriptome analysis revealed that the SOM and the EMM exhibit different gene expression profiles by E12. 5 33 , and the development of the skull vault is achieved by interactions between the apical (EMM) and basal (SOM) cell populations 28 . Although the EMM is normally non-osteogenic, previous reports demonstrated that the EMM can generate bone in genetic disorders 27,32 .
NCC-specific Dlx5-augmentation results in a switch of the jaw identity 11 , but the effect on NCC differentiation potential has not been examined. In this study, we further investigated the effect of Dlx5-overexpression in NCCs with special reference to early development and differentiation potential of the EMM.

Results
Dlx5 expression and NCC distribution in the NCC Dlx5 . NCC-specific forced expression of Dlx5 was confirmed in Wnt1-Cre;R26R CAG-flox-Dlx5/+ (hereafter NCC Dlx5 ) mice at E9.5 with ectopic Dlx5 expression in the pharyngeal arch 11 . We further examined the expression of Dlx5 in later stages by comparison with X-gal staining of Wnt-Cre;R26R lacZ/+ (NCC LacZ ) (n = 3) to demonstrate the NCC distribution. NCCs of wild-type located at the maxillary process and supraorbital ridge at E10.5, but hardly detected in the surrounding of the brain at the vertex ( Fig. 1a). At E11.5, NCCs made up the mandibular and maxillary processes, also the head mesenchyme surrounding the brain (Fig. 1b). Endogenous Dlx5 expression in head mesenchyme at E10.5 was found only in the mandibular process, whilst in the NCC Dlx5 , Dlx5 was additionally expressed in the maxillary process and the SOM (n = 3) (Fig. 1c,d). At E11.5, endogenous Dlx5 expression was seen in the frontal bone primordium of the SOM (Fig. 1e). In the NCC Dlx5 , Dlx5 was expressed in the EMM besides the SOM (n = 3) at E11.5 (Fig. 1f, arrowhead). Therefore, Dlx5 expression was successfully induced in NCCs, including EMM, in the NCC Dlx5 .
We examined the NCC distribution in the NCC Dlx5 by whole-mount in situ hybridization (WISH) for Snail family transcriptional repressor 1 (Snai1), a NCC specifier 1 , at E9.0 (n = 4). Snai1 expression was shown in a comparable pattern between the control and the NCC Dlx5 (Fig. 1g,h). We next examined the Snai1 expression by section ISH (n = 3). We confirmed that post-migratory NCC-derived mesenchyme at the surrounding of the forebrain and the mandibular process similarly expressed Snai1 in both of the control and the NCC Dlx5 (Fig. 1i,j). In the dorsal region of the rhombencephalon, Snai1 expression was detected in migrating NCCs in the control and the NCC Dlx5 (Fig. 1k,l), the data revealed no difference between the two in this area. These results indicated that Dlx5-overexpression do not affect the migration and distribution of cranial NCCs.
Predisposition of NCC differentiation in the NCC Dlx5 mouse. Dlx5 is normally expressed in the trigeminal ganglion 14 , and the size is reduced in Dlx5 knock-out mice 7 . Acetylated tubulin staining at E11.5 demonstrated that neuron localization did not show obvious difference between the control and the NCC Dlx5 (n = 3) (Fig. 2a-d). Reconstructed trigeminal ganglion from the serial histological sections at E17.5 illustrated the similar shape and size of the control and the NCC Dlx5 (Fig. 2e,f) and no significant difference in volume (n = 3) (Fig. 2g). We also examined the pigment cell, another NCC-derivative 1 , by expression of dopachrome tautomerase (Dct) 34 on frontal sections at E15.5 (n = 3) (Fig. 2h,i). The number of Dct-positive cells in the head dermis was significantly higher in the NCC Dlx5 than that of the control (p < 0.05) (Fig. 2j), suggesting that the NCC potential for pigment cell differentiation was enhanced by Dlx5-augmentation.
We next examined bone and cartilage formation in the calvarium. In the control, chondrocranium cartilages were observed at the skull base and lateral walls, while no cartilage and bone was observed at the apical part of the head at E13.5 and E14.5 (Fig. 2k,m). Interestingly, cartilage was formed at the vertex of the NCC Dlx5 (Fig. 2l,n). This calvarial cartilage was newly introduced to the region that usually has no cartilage, hence it is an ectopic cartilage 35 (hereafter ec). The ec did not connect with cartilages at the skull base or the lateral wall and appeared like a bridge connecting bilateral hemispheres at E13.5 (n = 4) and E14.5 (n = 4) (Fig. 2l,n, arrowhead). As the frontal and parietal bones further developed at E14.5, the coronal suture was identified between the frontal bone and parietal bone in the control (Fig. 2m). The ec was seen anterior to the prospective coronal suture and seemed to outline the posterior border of NCC-derived frontal bone (Fig. 2n, arrowhead). The ec remained unossified (n = 5) (Fig. 2p, arrowhead), whereas no cartilage was detected at the same region in the control even at postnatal day 0 (P0) (Fig. 2o).
At P0, calvarial bone formation was distinct between the control and the NCC Dlx5 . The frontal bone developed toward the midline, forming the interfrontal suture in the control (Fig. 2o). In the NCC Dlx5 , not only ec but bony islands were also found in the interfrontal suture (Fig. 2p, asterisk) and the posterior part of frontal bone forming area (Fig. 2p, double asterisk). In some NCC Dlx5 mice, "patchy" bones with holes ( Supplementary Fig. S1a,b) were formed randomly in the frontal bone and interfrontal area. These irregular bones seemed to fuse to the frontal bone, therefore we called them heterotopic bones 35 (hereafter hb). In summary, Dlx5 ectopic expression in head mesenchyme induced ec and hb formations. The development of the endogenous frontal bone was comparable between the control and the NCC Dlx5 at P0 (Fig. 2o,p) www.nature.com/scientificreports/ but the relative position of bilateral parietal bones in the cranium was abnormal at P0, and the anterior edge of parietal bone, which comprises the coronal suture was more inclined to the posterior (Fig. 2p). We reason that the inclination was caused by the appearance of ec and hb that interfered the normal parietal bone development. Given the significant effects of Dlx5 on NCC, we attempted to analyze MES-specific Dlx5-augmented mice by crossing Mesp1-Cre mice 36 and R26R CAG-flox-Dlx5/+ . Four litters at E11.5-13.5 were examined, but all Mesp1-Cre;R26R CAG-flox-Dlx5/+ fetuses were lethal, making following analyses impossible.
NCC potential for chondrogenesis and osteogenesis was increased in the calvaria. Skeletal staining analysis of the NCC Dlx5 revealed that both chondrogenesis and osteogenesis were promoted simultaneously at the same region of the calvarium, which has not been reported in any other calvarial bone mutants. We thus further analysed the phenotype. Because the ec and hb were present at the NCC-MES junction, we first confirmed the cell origin of the misregulated structures. The NCC domain was visualized by enhanced yellow fluorescent protein (EYFP) in the Wnt1-Cre;R26R CAG-flox-Dlx5/EYFP (NCC Dlx5/EYFP ) and in the littermate control Wnt1-Cre;R26R EYFP/+ (NCC EYFP ) at E17.5.
In bright-field images, the baso-lateral part of the coronal suture was comparable between the NCC EYFP and the NCC Dlx5/EYFP (n = 5) (Fig. 3a,b, brackets). However, the coronal suture at the vertex seemed to be shifted more posterior in the NCC Dlx5/EYFP compared to the NCC EYFP (Fig. 3a,b, dashed line). Under the fluorescent microscope, we found that the frontal bone of the NCC EYFP was highlighted while the parietal bone was not, and the "NCC tongue" 3 protruded to the sagittal suture (n = 5) (Fig. 3c, arrowheads). The ec and hb were formed within the fluorescent NCC-derived domain in the NCC Dlx5/EYFP (n = 5) (Fig. 3d). In sagittal sections, the ec and hb were clearly detectable by fluorescence (n = 3) (Fig. 3f). The hb was formed in line with the frontal bone and parietal bone, and the ec was always seen underneath the bone-forming layer (Fig. 3f). In the NCC EYFP , osteogenic fronts  , h), showing the forebrain, the mandibular process (i, j), and the dorsal area (k, l). fr, frontal bone; md, mandibular process; mx, maxillary process; so, supraorbital ridge; EMM, early migrating mesenchyme; SOM, supraorbital mesenchyme; op, optic vesicle; ot, otic vesicle, rho, rhombencephalon. Scale bars; 500 μm (a, b, c, e), 200 μm (g), and 100 μm (i, k).   www.nature.com/scientificreports/ of the frontal and parietal bones normally overlap at the coronal suture ( Fig. 3e), however, in the NCC Dlx5/EYFP , the suture was established in a widely opened end-to-end type (Fig. 3f, arrowheads). Therefore, the extended NCC-derived area hosted the NCC-derived ec and hb, and the position of the coronal suture was shifted backwards. Computed X-ray microtomography (μCT) data of P0 (n = 3) revealed that the NCC-derived frontal bone length at the midline of calvaria in the NCC Dlx5 increased significantly by 18.5% compared to control (p < 0.001) ( Fig. 3g-k). Furthermore, the NCC-derived calvarial bone volume rose by 10.2% (p < 0.05) (Fig. 3l). The volume of MES-derived parietal bone, meanwhile, did not significantly decrease (n = 3) ( Supplementary Fig. S1c). Besides, the ec was found at the cranial vertex, the chondrogenic potential of the NCC Dlx5 obviously increased in this area. Thus, the NCC-derived apical head mesenchyme increased chondrogenic and osteogenic potentials in response to Dlx5-overexpression.
The ec developed inside of the dura mater. Histological analysis at E15.5 showed that cranial bones had not reached to the midline, and cartilage was absent in the vertex in the control (Fig. 4a-c). In the NCC Dlx5 , the thickness of the ec was comparable to that of cranial base cartilages (n = 3) (Fig. 4d-f). The hb was ossified on top of the ec in the calvaria (Fig. 4d-f). As expected from the skeletal staining data (Fig. 2), the endogenous frontal bone and parietal bone of the NCC Dlx5 observed on HE sections illustrated similar bone quality in terms   Fig. S2). We conducted a more detailed investigation by using transmission electron microscope (TEM) at E15.5 (n = 3). Using toluidine blue stained semi-thin sections, we chose the relevant area of the control and the NCC Dlx5 for TEM analysis (Fig. 4g-j). In the control, the dura mater located just underneath the bone layer, characterized by longitudinally arranged fibroblast-like cells (Fig. 4i, arrow), and collagen bundles filling intercellular spaces (Fig. 4i, arrowhead) 37 . The arachnoid mater was clearly seen next to the dura mater, which contains more loosely attached cells, and numerous free ribosomes 28,37,38 (Fig. 4i). In the NCC Dlx5 , the ec occupied a large space between the bone and the brain (Fig. 4j). On its outer and inner surfaces, similar structures that had the characteristics of the dura mater were found (Fig. 4j). Besides, the arachnoid mater was recognized under the dura mater structure (Fig. 4j). Therefore, our histological analyses demonstrated that the ec developed within the dura mater.
The ec and hb were derived from the EMM. It was reported that apical mesenchyme has both osteogenic and chondrogenic potential in vitro 39 . Double conditional knock-out of Msh homeobox 1/2 (Msx1/2) in the mouse NCC (Msx1/2 cko/cko ) generated heterotopic bones from the EMM at abnormal positions including the suture area 32 . More recently, in vivo loss and gain of function experiments of LIM homeobox transcription factor 1 beta (Lmx1b), which is expressed in the EMM but not in the SOM, demonstrated the inhibitory function of Lmx1b on osteogenic specification in the EMM 27 . Lmx1b loss-of-function in head mesenchyme (Lmx1b LOF HM ) induced osteogenic marker expression in the vertex mesenchyme, future interfrontal suture and expanded boneforming area resulting in synostosis 27 . These previous studies suggest that the EMM has osteogenic potential, which is inhibited in normal context. . Arrows and arrowheads in (i, j) point to longitudinal-arranged fibroblasts and collagen fibrils, respectively. Ectopic cartilage appears in the meninges and is flanked by dura mater (j). ar, arachnoid mater; br, brain; du, dura mater; ec, ectopic cartilage; fr, frontal bone; hb, heterotopic bone; pa, parietal bone; pi, pia mater. Scale bars; 500 μm (a, d), 100 μm (b, c, e, f), 2 μm (i, j). www.nature.com/scientificreports/ We explored the gene expression change that led to ec and hb formations. At E10.5, few mesenchymal cells were detected at the apical head, and histological difference between the control and the NCC Dlx5 was not noticed (Fig. 1a,b). We found that there were differences in gene expression as well as histology from E11.5. SRY-Box transcription factor 9 (Sox9) and Runt-related transcription factor 2 (Runx2) were used for evaluating mesenchymal condensation of cartilage and bone, respectively. At E11.5, Sox9 and Runx2 were substantially upregulated in the EMM region of the NCC Dlx5 compared to the control ( Fig. 5a-d, arrowheads). Mesenchymal condensation for the ec and hb was found at E11.5, which was around the same time with the beginning of original cranial base and calvarial development. The ectopic Sox9 expression domain was not connected to any part of the future skull base domain (n = 5) (Fig. 5b). This result confirmed that the ectopic Sox9 expression was not due to the extension of skull base primordium. In contrast, Runx2 expression in the EMM seemed to be continuous with the SOM by a thin expression line in the NCC Dlx5 (n = 4) (Fig. 5c,d). To test whether the developing hb was independent of the SOM, we examined expression of Sp7, an early osteoblast marker and downstream of Runx2, at E14.5 by WISH (n = 5). The development of frontal and parietal bones was visualized by Sp7 expressing domain at this stage (Fig. 5e,f, dotted line). In the EMM area of the NCC Dlx5 , several Sp7 expression islands were independent of the SOM (Fig. 5f, arrowheads). These results strongly suggested that the hb in the NCC Dlx5 is formed in the EMM and independent of the endogenous frontal bone.
The EMM layer was thickened in the NCC Dlx5 at E11.5, which contained expanded Sox9 and Runx2 expression domains (Fig. 5g,h). Our BrdU incorporation assay showed a significantly increased BrdU + cells in the EMM of E11.5 NCC Dlx5 (n = 3, p < 0.001) (Fig. 5i). Immunohistochemical (IHC) staining for cell death showed no signals in the EMM of both the control and the NCC Dlx5 at E11.5 (n = 3) (Supplementary Fig. S3). Therefore, the thickened EMM in the NCC Dlx5 was caused by increased cell proliferation.
Early development of the EMM in the control and the NCC Dlx5 . We examined gene expression in early development of the ec and hb (n = 4) at E11.5. In the control, expression of Forkhead Box C1 (Foxc1), transcribed in all three meningeal layers 40 , was broadly detected in the mesenchyme, but the signal was not seen or at much lower levels just underneath the epidermis at E11.5 (Fig. 6a, arrowheads). Importantly, expression domains of Foxc1 and Dermo1, molecular markers for the meninges and the dermis 41,42 , respectively, were mutually exclusive (Fig. 6a,b). Control mice showed no expression of Sox9 in the EMM (Fig. 6c), but Runx2 was expressed in the Dermo1 expressing mesenchyme as a thin layer (Fig. 6b,d). Msx1 expression was found in the whole head mesenchyme (Fig. 6e). Msx2 expression domain was localized to the outer layer of the EMM, including a part of the meninges and the dermis (Fig. 6f, compared to 6a,b).
In the NCC Dlx5 , the Foxc1 expression domain appeared to contain the ec primordium marked by Sox9 expression (Fig. 6g,i), which is consistent with the phenotype in which the ec is surrounded by the dura mater (Fig. 4j). Remarkably, Dermo1 expression was highly upregulated in the NCC Dlx5 compared to the control (Fig. 6h), indicating that Dlx5-augmentation enhanced the dermis formation. Runx2 expression of the NCC Dlx5 was more evident compared to the control (Fig. 6d,j). Sox9 expression domain was included in the Runx2 domain (Fig. 6i,j). Because the hb developed outside of the ec (Fig. 4f), Runx2 expressing cells outside of Sox9-positive layer were thought to differentiate into osteoblasts. Importantly, these osteoblasts also expressed Dermo1 (Fig. 6h,j), suggesting that the hb was derived from the dermal layer. In contrast, the ec shown by Sox9 expressing domain seemed not to show Dermo1 expression (Fig. 6h,i).
Moreover, Msx1 expression was present in the arachnoid and the pia mater, and was not expressed in other parts of the EMM (Fig. 6k). Msx2 was downregulated at some areas of head mesenchyme, however, expressed in the ec and hb (Fig. 6l). Since Lmx1b loss-of-function induced hb formation 27 , we also examined Lmx1b expression. Lmx1b was expressed in both the ec and hb ( Supplementary Fig. S4), suggesting that since Dlx5 is a downstream of Lmx1b 27 , Dlx5-overexpression does not affect Lmx1b expression.
PDGFRα, WNT/β-catenin and Bmp2 signals are upregulated in the NCC Dlx5 . Platelet-derived growth factor receptor Alpha (Pdgfra) augmented in NCCs generated ec at the coronal suture, which was similar to the ec of the NCC Dlx543 . We performed double imunnofluorescent staining for PDGFRα and SOX9 at E11.5 (n = 3), PDGFRα was present in the outer portion of the EMM in the control, while SOX9 signal was not detected (Fig. 7a-d). In the NCC Dlx5 , PDGFRα expression levels were more intensive in the dermal and SOX9positive layers (Fig. 7e-h). Semi-quantitative analysis on immunofluorescent staining showed that PDGFRα signal was intensified by Dlx5-augmentation (Fig. 7u). WNT/β-catenin signalling induces osteoblast differentiation in intramembranous ossification 44,45 . The conditional β-catenin loss-of-function in the dermis using Dermo1-Cre or Engrailed1-Cre driver resulted in the loss of dermis and cranial bones. Instead, cartilages were induced between the epidermis and the thinner meninges 42 . Reversely, the NCC Dlx5 had thickened dermis (Fig. 6h) and hb (Fig. 4f). We conducted double immunofluorescent staining for β-catenin and RUNX2 signals at E11.5 (n = 3). In the control, β-catenin and RUNX2 were sparsely expressed below the epidermis (Fig. 7i-l). In the NCC Dlx5 , RUNX2 signal illustrated the hb (Fig. 7m,n), β-catenin signal in the dermis and hb forming area were clearly upregulated (Fig. 7m-p). Semi-quantitative analysis for β-catenin showed that the signal was significantly upregulated by Dlx5-augmentation (Fig. 7v).
Bone morphogenetic protein (Bmp) is involved in hb formation in the interfrontal suture 32,46 and it regulates Runx2 expression through DLX5 47 . Our data showed that Bmp2 was not expressed in the control head mesenchyme at E11.5 (Fig. 7q,r). However, it was ectopically induced in hb forming area in the NCC Dlx5 (n = 3) (Fig. 7s,t).

Discussion
Dlx5 expression in the NCC is involved in jaw patterning. Aside from that, Dlx5 is expressed in the several NCC-derived head components and is related to their differentiation 1 , thus, we further examined predisposition of the NCC affected by Dlx5-augmentation. Investigations of Snai1 expression indicated that the migration and distribution of NCCs were unaffected by Dlx5-overexpression (Fig. 1). There was little effect on trigeminal ganglion development, but the number of pigment cells was increased in the NCC Dlx5 (Fig. 2), which possibly corresponds to enhanced dermal cell proliferation in the NCC Dlx5 (Figs. 5h, 6h). Despite the extra skeletogenesis in the NCC Dlx5 , non-skeletal NCC-derivatives such as trigeminal ganglion, dermis, and pigment cells were not attenuated, suggesting that NCCs did not fluctuate between non-skeletogenic and skeletogenic fates. Previous reports showed that modifications in molecular cascades in mouse head mesenchyme resulted in either ec or hb formation in the skull vault. In particular, Msx1/2 cko/cko and Lmx1b LOF HM caused hb formation at the posterior of the frontal bone similar to the NCC Dlx5 , but ec formation was not reported in those mutants 27,32 . By contrast, Pdgfra upregulation in NCCs generated ec in the coronal suture, meanwhile, the frontal bone appeared unchanged 43 . Some mutants demonstrated that cartilages replaced calvarial bones in mice, such as β-catenin knock-out 42,45 and fibroblast growth factor 8 (Fgf8) gain-of-function 48 . Therefore, chondrogenesis could be upregulated at the expense of osteogenesis. In this study, we showed that chondrogenesis and osteogenesis were promoted simultaneously in the NCC Dlx5 calvaria (Fig. 2). Ec and hb formation in the NCC Dlx5 occurred in the meningeal and dermal layers of the EMM, respectively (Figs. 4, 6). We also found that, in normal development, the EMM seems to be committed to dermal and meningeal layers by E11.5 (Fig. 6). The distinct cell differentiation of apical head mesenchyme in response to Dlx5-augmentation in NCCs strongly support the idea that there are different cell populations in the EMM by this stage.
The development of the ec in dura mater layer highly suggests that head mesenchyme which was originally destined to be meningeal precursor cells could be turned into chondrocytes. The ec formation within the dura mater was previously reported; residual cartilages are occasionally formed above the trigeminal ganglia in mammals 49 , and the pila antotica near the ala temporalis in therians develops inside the dura mater as atavistic relics 50 . In the clinical aspect, some meningeal chondrosarcomas, which are tumours containing cartilaginous islands, in the dura mater are reported 51 . Additionally, when the dura mater is transplanted to the trunk in contact with mesodermal elements, the transplant sometimes develops cartilage 52 . In addition, explants of mouse mesenchymal cells from the vertex area of the head (corresponding to the EMM) at E12.5-14.0 show the potential for bone and cartilage formations 39 . These reports support our idea of dura-to-cartilage transformation.
In the NCC Dlx5 , the ec develops in the frontal bone area in proximity to the coronal suture, which is the NCC-MES boundary, and some small cartilages in the interfrontal suture. Loss of β-catenin is one of the causes of ec induction during calvarial development 42 . However, β-catenin was detected in the ec forming area in the NCC Dlx5 (Fig. 7f,o), suggesting that formation of ec was not attributed to change in WNT/β-catenin signalling. Pdgfra upregulation in the NCC in mice exhibits ec formation in the coronal and interfrontal suture, which is similar to the ec in the NCC Dlx543 . We found that PDGFRα signal was increased in the NCC Dlx5 . Although the layer of ec formation was not studied in the Pdgfra mutant 43 , it is suggested that the interfrontal area, the NCC-MES boundary show potential for cartilage differentiation when stimulated by PDGFRα (Fig. 8b).
Our gene expression analyses suggest that, in the control, Runx2-expressing layer just below the epidermis in the EMM does not associate with the meninges at E11.5. The Runx2-expressing layer is present within the  (Fig. 8a). Furthermore, the Runx2 expression in mesenchymal cells suggests the intrinsic osteogenic potential of the dermis. Anti-osteogenic functions of Msx2 and Lmx1b are likely to suppress the osteogenic potential in the EMM 27,32 .
Msx2 can act as either osteogenic inhibitor or activator (Fig. 8, Msx2 (in), Msx2 (ac)) 53 . MSX2 inhibits Runx2 transcriptional activity 54,55 , and competes with RUNX2 in binding to a regulatory sequence of Osteocalcin (Ocn), an osteogenic induction gene 56 (Msx2 (in)). The anti-osteogenic activity of Msx1 and Msx2 in the EMM explained the hb in the Msx1/2 cko/cko at the early stage of calvarial development (~ E12.5) 32 . In the Lmx1b LOF HM , Msx2 expression was downregulated in the EMM, and the phenotype could be explained similarly to the Msx1/2 cko/cko27 . We thus propose that Msx2 inhibits Runx2 osteogenic induction in the dermis in normal situation (Fig. 8a), and possibly Msx1 is also related to this function to some extent.
Contrary to the anti-osteogenic function above mentioned, Msx2 has been shown to promote both proliferation (undifferentiated condition) and differentiation of osteoblast lineage cells 57,58 (Msx2 (ac)). Despite the similar hb formation to the Msx1/2 cko/cko , Msx2 was expressed in the hb forming region in the NCC Dlx5 (Fig. 6). Dlx5 is an important factor to antagonize Msx2 anti-osteogenic function 55,59,60 , it therefore seems that Dlx5overexpression suppressed the anti-osteogenic function of Msx2 in the NCC Dlx5 (Fig. 8b). By contrast, Dlx5 appears to promote the pro-osteogenic function of Msx2 in concert with other osteoblast activators such as Bmp2 and β-catenin (Fig. 8b, Msx2(ac)). Dlx5 is a downstream target of Bmp2 47 , and we observed increased Bmp2 expression in head mesenchyme of the NCC Dlx5 . These results suggest that Dlx5 activates Bmp2 through positive feedback. Since Msx2 is also a main downstream of Bmp2 61 , maintenance of Msx2 expression in the NCC Dlx5 could be caused by Bmp2 induction. WNT/β-catenin promotes intramembranous bone formation and dermal layer differentiation 42 . We demonstrated that β-catenin levels were increased in the forming hb and probably associated with the enhanced dermis differentiation (Fig. 7m-p). Our data are consistent with previous reports that Bmp2 and β-catenin synergistically induce Msx2 62,63 . Bmp2 upregulation in the NCC Dlx5 is consistent with Msx1/Msx2 cko/cko and Lmx1b LOF HM mutants 27,32 . Altogether, the hb formation in the NCC Dlx5 was caused by enhanced osteogenic induction of Bmp2 and β-catenin signalling pathways that involve Msx2 (Fig. 8b). It will be intriguing to clarify in more detail about the molecular mechanism for the phenotype, involved in the dual function of Msx2 in the future.
It was reported that Lmx1b prevents ossification of EMM from E9.5 27 . In this study, we found that expression patterns of Dermo1 and Foxc1 are mutually complementary at E11.5 in the control (Fig. 6a,b). Taken together, we propose that after non-skeletogenic commitment, the EMM is divided into two populations, dermal and meningeal layers by E11.5 in normal development (Fig. 8a). However, this commitment is not irreversible because the cell fate can be altered to cartilage and bone by responding to pro-skeletogenic signals such as augmentation of Dlx5. Given that the EMM has differentiation potentials to both cartilage and bone, it should be carefully evaluated which mesenchyme (EMM or SOM) contributes to ectopic and heterotopic skeletogenesis in the calvarium when mutant mice are examined.
Apical-basal patterning in cranial development by the interaction between the EMM and SOM has yet to be fully elucidated. Although the molecular basis of EMM differentiation in its early development is still unknown, our findings provide a more detailed picture of the EMM sublayers together with their potentials, shedding light on developmental mechanisms of cranial development.   Alizarin red and alcian blue skeletal staining. Post-natal day 0 (P0) mice were skinned, followed by fixation in 96% ethanol for one week. Skeletal staining was performed in a mixture of 0.02% alcian blue (Sigma, 05500-10G), 0.005% alizarin red (Wako, 013-25452), 5% acetic acid in 70% ethanol for three days with rocking www.nature.com/scientificreports/ at room temperature. Samples were then washed by distilled water and optically cleared by glycerol in 0.5% KOH until the bone and cartilage were visible.
μCT was taken by inspeXio SMX100CT (Shimadzu, Japan). The data were analyzed by Avizo 6.3. μCT scans were uploaded to Avizo 6.3 as DICOM files and visualized using Isosurface in Avizo 6.3. The NCC-derived bone length was measured in three-dimension at the midline, tracing the top of the calvaria (Fig. 3g-j, n = 3). The bone volume was calculated by Avizo 6.3, including all the bone components within the frontal bone forming area.
Histological analysis. Heads of fetuses at E15.5 were fixed in Bouin's fixative solution for 48 h. The samples were washed by 70% ethanol, then dehydrated in a gradient of ethanol until 100%, followed by xylene treatment and embedded in paraffin. Sections were cut at 5 μm thickness (Leica RM2235, Germany), then stained by 1% alcian blue (Sigma, 05500-10G) in 3% acetic acid, followed by Mayer's Hematoxylin and 1% Eosin Y solution.
Comparisons were made among at least three independent littermates.
Transmission electron microscopy (TEM). Heads of E15.5 fetuses were trimmed to collect the targeted tissues, then fixed in 2.5% glutaraldehyde/0.1 M phosphate buffer (PB) for 72 h. After washing in PB overnight at 4 °C, samples were postfixed with Osmium tetroxide (OsO4) for 2 h. Samples were then dehydrated in ethanol, followed by infiltration of epon resin and propylene oxide catalyst, then embedded in epon resin. Semi-thin sections at 1 μm and toluidine blue staining were utilized to examine the samples. Ultrathin sections at 80 nm were collected and double-stained with uranyl acetate and lead citrate on carbon-coated copper grids. Sections were observed by transmission electron microscopy (Hitachi H-7100, Japan) (n = 3).