The Ihh^+/− and Gli3^+/Xt−J mice maintenance, breeding and PCR genotyping has been described previously (Koziel et al., 2005). Wild-type (WT) littermates NMRI mice were used as controls. All animal experiments were approved by the University of Helsinki, Helsinki University Hospital and the Southern Finland Council Animal Welfare and Ethics committees.

E13.5 and E15.5 tissues were fixed in a copious volume of freshly prepared 4% paraformaldehyde overnight at 4°C. They were then washed in PBS before dehydrating in an increasing strength ethanol or methanol series before in situ hybridization on either tissue sections or whole mounts, respectively.

Ihh, Ptch1, Gli1-3, Runx2, Osx, Bglap, and Ibsp riboprobes were prepared, and in situ hybridization was performed as described previously and briefly described below (Rice et al., 2006; Tanimoto et al., 2012).

Sections (7μm) were deparaffinised, rehydrated and permeabilized with proteinase K. Tissues were hybridized overnight at 52°C with ³⁵S-UTP labeled riboprobes. Hybridization was followed by high stringency washes at 50°C and at 65°C. Slides were then washed in NTE at 37° and treated with ribonuclease A to remove non-specifically bound and excess probe. The slides were coated with autoradiography liquid emulsion and exposed in a dark box for 10–18 days at 4°C. The slides were developed, fixed and then counterstained with hematoxylin.

Samples were rehydrated, bleached with H₂O₂ and permeabilized with proteinase K. Tissues were prehybridized in PBST. Tissues were hybridized overnight at 64°C with denatured digoxigenin-labeled probes followed by high stringency washes. Next tissues were washed MABT then pre-blocked at room temperature for 3 h prior to overnight incubation with anti-dig-antibody coupled to alkaline phosphatase 4°C. Following extensive MABT and NTMT washes the color reaction was performed with NBT/BCIP.

Immunohistochemical staining was performed as described previously (Rice et al., 2016). Briefly tissue sections were permeabilized and blocked, then incubated overnight at 4°C with anti-IHH (AF1705, R&D systems), anti-GLI1 (L42B10, Cell Signaling), anti-GLI3 (AF3690, R&D systems) or anti-PTCH1(G-19, Santa Cruz Biotechnology). Signal visualization was performed using Enzmet HRP detection (Nanoprobes). Sections were counterstained with nuclear fast red.

Heads of mutant mice and WT littermates aged E16.5 and E18.5 were fixed in 95% ethanol overnight and stained with Alcian blue and Alizarin red and then cleared in 1% KOH and transferred to glycerol. For calculations of calvarial bone and suture size, images were captured using AnalySIS software (Soft Imaging System) and Olympus BX41 microscope and analyzed in Adobe® Photoshop CS4.

Calvaria were dissected from E15.5 WT embryos and tissue samples taken from osteogenic front of the frontal bone and from interfrontal midsutural mesenchyme. Samples were pooled from three calvaria of the same litter and this was repeated 4 times. Age-matched brain tissue was used as control. Western blotting was carried out as described previously (Tanimoto et al., 2012). Briefly, 10 μg of each sample was probed for GLI3 antibody (AF3690, R&D Systems), GLI1 antibody (2643, Cell Signaling Technology), α-tubulin (T6199, Sigma-Aldrich). 20 μg of total protein from siRNA treated calvarial cells was probed for RUNX2 antibody (8486, Cell Signaling Technology), IHH antibody (AF1705, R&D Systems), and normalized against signal by Actin antibody (A2066 Sigma), detected by secondary antibodies goat anti-mouse IRDye 800CW (926-32210, LI-COR) and goat anti-rabbit 680RD (926-68071, LI-COR). Blots were analyzed with an Odyssey CLx infrared imager (model 9120) (Li-Cor) and Image-J (NIH) software. Statistical values were calculated using Student's t-test, with P-value below 0.05 indicated as significant.

Mouse E15.5 wild type (WT) primary calvaria derived cells (CDC) were isolated by trysinization and cultured for several days in DMEM containing high glucose and supplemented with 10% FBS. Cells were pooled from at least 2 calvaria. After 1st passage cells were seeded onto 6-well plates for experiments (100,000 cells per well). After 48 h of culture, cells were treated with recombinant human/mouse IHH protein (1705-HH-025 R&D systems): 50, 100, 250 ng/ml, and control cells with BSA (bovine serum albumin) for 1, 2, 6, and 24 h. RNA isolated from control and IHH treated CDC cells was quantified. 300 ng of RNA was used for cDNA synthesis using reverse transcriptase enzyme and random hexamer primers. Thereafter, RT-qPCR was performed using the primers Gli1 (FP: CAGCATGGGAACAGAAGGACT, RP: CTCTGGCTGCTCCATAACCC), Gli2 (FP: AACTTTTGTCTCCTCGGGTCC, RP: CTGCTGTCCTCCAAGAGACC) and Gli3 (FP: AAGCCCATGACATCTCAGCC, RP: CTCGAGCCCACTGTTGGAAT). SYBR fast (Kapa Biosystems) qPCR master mix was used for real time quantitative PCR using Quant studio 3 from Thermo scientific. Mouse 18S rRNA gene was used as the reference gene and negative control with no cDNA was used. The experiment conducted 3 times, each time with different biological samples.

E15.5 mouse calvarial cells were isolated by four sequential trypsin-treatments of whole calvaria separated from the skin and meninges. The first set of cells after 15 min 0.25% trypsin incubation were discarded, and the cells from the following trypsin treatments were pooled and cultured in growth medium (DMEM containing 4,5 g/L glucose supplemented with 1 mM Na-pyruvate, 4 mM L-glutamine, 1% penicillin-streptomycine, and 10% FBS, (Lonza and Gibco).

Analysis of Wt and Gli3^{Xt−J/Xt−J} calvarial cells: At passage 2 the cells were transferred into new growth medium until confluent from which cell lysates were derived. Western blot analyses were carried out with 20 μg of total protein.

siRNA experiments: At passage 2 the cells were transfected either with negative control siRNA (Ambion Silencer Select control #1 siRNA, 4390843, Thermo Fisher) or Runx2 siRNAs (Ambion 4390771 Runx2 Silencer Select Pre-designed siRNA, Thermo Fisher) using Lipofectamine RNAiMAX (Thermo Fisher). Cells were grown on six well plates with the siRNA transfection complex for 3 days, after which the medium was replaced by osteogenic medium (DMEM, 10 mM β-glycerophosphate, (Sigma), 50 μg/ml ascorbic acid, Sigma, and 100 ng/ml BMP2 (R&D Systems) for 24 h from which cell lysates were derived. Western blot analyses were carried out with 20 μg of total protein. siRNA experiments were carried out 3 times.

E15 mouse calvarias were dissected for Trowell type of organ culture (Rice et al., 2000). Affi-gel agarose beads (Bio-Rad) were washed with PBS, and circa 200 beads were soaked with premixed siRNA—RNAiMAX reagent (Thermo Fisher) at 37°C for 30 min. Several beads were placed on the calvarias that were further cultured for 3 days in osteogenic medium. Efforts were made to place beads onto comparable areas of the calvaria that had endogenous expression of either Runx2 full length or Ihh (Rice et al., 2010). The areas of Ihh expression in WT E15.5 calvaria were further identified (Figure 1B) and beads impregnated with either negative control or anti-RUNX2 SiRNAs were placed on the parietal bone osteogenic fronts for IHH experiments, which can be identified in bright field microscopy (Kim et al., 1998). Beads placed on the parietal bone or osteogenic fronts for RUNX2 experiments.

To determine the specific transcript (Figures 1A–C) and protein (Figure 1D) expression domains and protein activity of HH pathway members in the developing calvaria in situ hybridization and immunohistochemistry were performed. Detailed in situ hybridization was performed at different developmental stages of calvarial and suture maturation. In an attempt to get comparable data, at both E13.5 and E15.5 analysis was done on consecutive E15.5 calvarial tissue sections. Also, at E15.5 whole calvarial tissue was analyzed to determine whether all sutures have similar expression patterns.

Ihh mRNA and protein were detected in the osteogenic fronts of all the calvarial bones, in a very restricted group of osteoprogenitors (Figures 1C,D). In addition there was weak protein expression across the suture. The cell surface HH receptor Ptch1 and Gli1 and Gli2 expression domains were overlapping with that of Ihh. Ptch1 and Gli1 are induced by HH ligand and therefore indicate the limit of Hedgehog activation (Lee et al., 2016). In contrast, Gli3 was diffusely expressed across the sutural mesenchyme.

In general, the calvarial osteogenic fronts (E15.5) have similar expression patterns to the osteogenic condensations (E13.5). Indeed, the osteogenic fronts can be considered analogous to osteogenic condensations with similar steps and regulation of osteoblastic differentiation and development. We also observe a differential expression of Gli3 in different sutures which may, in part, explain the suture specific phenotype seen in mice and patients with GLI3 mutations.

To establish if GLI3R is the predominant protein isoform in the calvarial suture, we performed western blotting on tissue samples dissected from either the midsuture or the osteogenic front from WT calvaria (Figure 1E). The total amount of GLI3 protein was higher in the midsutural samples. Both isoforms were observed in both samples, but the level of GLI3R was higher in the suture compared to the osteogenic front, and the opposite was true for the GLI3FL (Full length) (Figures 1F,G). The ratio of GLI3R/GLI3FL was significantly higher in the suture mesenchyme indicating that GLI3R is predominate in the midsutural tissue.

To verify whether a sample was midsutural or from the osteogenic front, we blotted the samples against GLI1. GLI1 also acts as a read-out for HH signaling, and based on our in situ hybridization results, HH signaling is active only in the osteogenic front. The level of GLI1 protein was significantly higher in the osteogenic front samples compared to the samples isolated from the suture (Figures 1F,G). Taken together, HH signaling is activated in the osteogenic fronts and GLI3R is predominantly localized in the midsutural mesenchyme.

RUNX2-I and RUNX2-II isoforms differ in their N-termini due to alternative promoter usage which results in dissimilar functions during bone development (Fujiwara et al., 1999; Zhang et al., 2009). Across developing sutures, Runx2 isoforms are differentially expressed with Runx2-I primarily expressed in relatively undifferentiated osteoprogenitors and Runx2-II in cells further along in osteoblastic differentiation (Park et al., 2001). We have previously shown that Runx2 is ectopically expressed across the Gli3^{Xt−J/Xt−J} suture (Rice et al., 2010). Here we analyzed their protein levels in osteogenic cells isolated from Gli3^{Xt−J/Xt−J} calvaria. We found that RUNX2-II was upregulated and RUNX2-I downregulated, compared to wild type controls indicating that in Gli3^{Xt−J/Xt−J} mice the osteoprogenitors within the suture were more differentiated than their WT littermates (Figures 2B,C). These results indicate that GLI3 maintains osteoprogenitors in an undifferentiated state by participating in the regulation of RUNX2. These data help explain the location specific differences seen in mRNA expression in Gli3^{Xt−J/Xt−J} calvaria and the craniosynostosis observed.

As Runx2 expression is altered in Gli3^{Xt−J/Xt−J} calvaria (Figure 2) (Rice et al., 2010), and RUNX2 is known to regulate limb growth through the induction of Ihh (Yoshida et al., 2004), we analyzed IHH protein levels in Gli3^{Xt−J/Xt−J} cells. We found IHH to be upregulated (Figure 2D). This indicates that the lack of GLI3 has allowed either increased IHH expression or a broader IHH protein expression domain.

In mesenchymal cells of the developing limb bud, Sonic hedgehog is known to up-regulate Gli transcription, while down-regulating Gli3 expression (Marigo et al., 1996). To test whether IHH has a direct effect on the expression of Gli3 in WT primary calvaria derived cells (CDC) we added recombinant mouse IHH, of varying concentration, to mouse E15.5 calvarial cells and investigated Gli1, −2 and −3 expression levels RT-qPCR at different time points.

Exogenous IHH did not alter the mRNA levels of Gli3 or Gli2 (Supplementary Figure 1). At 24 h, IHH exposure resulted in a gradual up-regulation of Gli1 expression in a dose-dependent manner (not statistically significant) (Supplementary Figure 1D). As Gli1 is a known target of hedgehog signaling activation this response was used as a positive control. This is consistent with previous work in a chondrogenesis model using human bone marrow stromal/stem cells (Handorf et al., 2015).

GLI transcription factors are effectors of HH signaling, with GLI3 acting as a strong repressor of IHH signals in chondrocyte differentiation with the developmental defects in endochondral ossification observed in Ihh^−/− mice being partially normalized by the loss of Gli3 (Ihh^−/−;Gli3^{Xt−J/Xt−J} mice) (Koziel et al., 2005). We considered an analogous role during osteoblast development in the calvaria, and compared the calvarial phenotypes of Ihh^−/−, Gli3^{Xt−J/Xt−J} and Ihh^−/−;Gli3^{Xt−J/Xt−J} mutant mice, E16.5–18.5, to those of WT littermates (Figure 3).

Calvarial bones of Ihh^−/− mice were smaller than WT. The size of the frontal and the interparietal bones was most affected, the difference being the greatest at E16.5 [frontal bone 43% smaller (p < 0.05), interparietal bone 33% smaller compared to WT]. Concomitant with the reduced bone size was a widening of all of the calvarial sutures [E16.5: posterior interfrontal suture 32% wider (p < 0.05), sagittal suture 43% wider, lambdoid suture 45% wider. E18.5: lambdoid suture comparable to WT, posterior interfrontal suture 47% wider (p < 0.05), sagittal suture 71% wider (p < 0.05)] (Figures 3E–H,Q–T). The posterior part of the frontal bone was reduced in height (in apical direction) and the interfrontal suture was the widest posteriorly. At E18.5, the overall length of Ihh^−/− skull (nasal bones to the occiput) was slightly reduced (7%, not significant). This was in part due to the short cranial base (12% shorter, p < 0.05) (Figures 3E,G).

Gli3^{Xt−J/Xt−J} mutant mice showed very specific stage and location differences in bone size and suture width (Figures 3I–L,Q–T). Already at E16.5 25% of Gli3^{Xt−J/Xt−J} mice showed synostosis of the lambdoid suture. This was primarily due to an increase in the size of the interparietal bone which was enlarged from E16.5 onwards (Figure 3T) as opposed to a generalized increase in the size of the parietal bone. At E16.5 the width of the posterior interfrontal suture in Gli3^{Xt−J/Xt−J} calvaria was significantly wider (26%, p < 0.05). However, the surface area of the frontal bones was only 15% smaller. This mismatch of the widened interfrontal suture in between the near normally sized frontal bones, indicated that frontal bone development was affected from an early stage and that extrinsic factors, for example brain shape and size and may affect frontal bone and interfrontal suture morphology. Later, the size of the frontal bones relative to WT increased. This was in part due to strong heterotopic ossification in the interfrontal suture, which culminated in partial fusion of the suture at E18.5. (Figure 3K). At every stage examined the width of sagittal suture was significantly wider (p < 0.05) (51% wider compared to WT at E16.5, 87% wider at E18.5), but the size of parietal bones was not significantly smaller compared to WT (15% smaller at E16.5, 12% smaller at E18.5). This sagittal suture widening may be caused also by extrinsic factors. At E16.5 and E18.5 overall length of Gli3^{Xt−J/Xt−J} mutant skull was comparable to WT samples (Figures 3I–L).

In conclusion, Gli3^{Xt−J/Xt−J} calvaria exhibited enlargement of the interparietal bone and heterotopic interfrontal bones with consequent interfrontal and lambdoid suture fusion. In contrast, the same sutures were wider and the bones smaller in Ihh^−/− mice. These opposite differences led us to test whether Gli3^{Xt−J/Xt−J} phenotype could be rescued by deleting Ihh. The craniosynostosis phenotype across the interfrontal and lambdoid sutures in Gli3^{Xt−J/Xt−J} mice was unaffected by the loss of Ihh (Figures 3M–P), despite there being a small normalization of interparietal size at E16.5 (not significant) (Figure 3T). Thus, even in the absence of IHH ligand, Gli3 deletion is sufficient to drive precocious ossification of the suture mesenchyme.

Ihh^−/− mice exhibit a severe disruption of endochondral osteogenesis with a complete lack osteoblasts. In contrast Ihh^−/− mice do form calvarial bones (St-Jacques et al., 1999). Using Runx2, Osterix (Osx, Sp7), Integrin binding sialoprotein (Ibsp), and Bglap (Osteocalcin) as markers of increasing osteoblastic maturity, we analyzed osteoblastogenesis in Ihh^−/− embryonic calvaria (Figure 4). We found that osteoblastogenesis progressed normally even though the calvarial bones were small and thin, and the sutures wide with marker expression reflecting this phenotype (Figures 4E–H).

In Ihh^−/− endochondral bones, we have previously shown that, de-repression of GLI3R (Ihh^−/−;Gli3^{Xt−J/Xt−J} mice) is sufficient to restore the RUNX2-positive osteoprogenitor population (Koziel et al., 2005) but the additional action of GLI2A is required to restore the progression of osteoblastogenesis to the RUNX2-positive/Osterix-positive cell stage (Hilton et al., 2005). In Ihh^−/−;Gli3^{Xt−J/Xt−J} calvaria the osteoblastic expression pattern was similar to that in Gli3^{Xt−J/Xt−J} mice although the heterotopic ossifications were less extensive (Figures 4M–P). Interestingly Gli3^{Xt−J/Xt−J} mice, which exhibit heterotopic ossifications in the interfrontal suture, showed expression of immature osteoblasts (Runx2 and Osx) extending across the full width of the suture and more mature osteoblastic expression markers (Ibsp and Bglap) were limited to isolated ossification islands (Figures 4I-L).

Ptch1 and Gli1 are direct transcriptional targets of HH signaling and both are upregulated by IHH (Lee et al., 2016). In Gli3^{Xt−J/Xt−J} mice, Ptch1 and Gli1 are expressed in the developing calvarial bones and in the abnormal craniosynostotic sutures. In Ihh^−/− calvaria however, Ptch1 and Gli1 were absent, indicating that no other Hh ligands are functional to compensate for the lack of Ihh (Figures 5A–F).

Endochondral ossification defects observed in Ihh^−/− mice can be partially normalized by deletion of Gli3 (Ihh^−/−;Gli3^{Xt−J/Xt−J} mice). This rescue is accompanied by a concomitant normalization of Ptch1 expression (Koziel et al., 2005). We therefore decided to test whether Ptch1 and Gli1 are normalized in the developing calvaria of Ihh^−/−;Gli3^{Xt−J/Xt−J} mice. However, similar to Ihh^−/− calvaria, Ptch1 and Gli1 expression were absent in Ihh^−/−;Gli3^{Xt−J/Xt−J} mice (Figures 5G,H). This suggests that the mechanism of Ptch1 and HH signaling activation in the calvarial osteoprogenitors differs compared to that in chondrocytes in the developing growth plate.

In the developing growth plate, it has been shown that Runx2 directly regulates Ihh expression in chondrocytes (Yoshida et al., 2004). To test an analogous interaction during intramembranous osteogenesis, we silenced Runx2 in E15.5 primary calvarial osteoblasts by selective Runx2 siRNA. Silencing was verified by western blot analysis showing that Runx2 protein was downregulated (Figures 6A,B). Ihh protein levels were significantly (p < 0.001) downregulated with Runx2 siRNA, when compared to negative controls (Figures 6A,B). The regulation of Ihh by Runx2 was then tested in embryonic calvarial organ culture with reduced expression of Ihh observed adjacent to beads coated in Runx2 siRNA, assayed by whole mount in situ hybridization (Figure 6C).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.