Loss of Iroquois homeobox transcription factors 3 and 5 in osteoblasts disrupts cranial mineralization

Cranial malformations are a significant cause of perinatal morbidity and mortality. Iroquois homeobox transcription factors (IRX) are expressed early in bone tissue formation and facilitate patterning and mineralization of the skeleton. Mice lacking Irx5 appear grossly normal, suggesting that redundancy within the Iroquois family. However, global loss of both Irx3 and Irx5 in mice leads to significant skeletal malformations and embryonic lethality from cardiac defects. Here, we study the bone-specific functions of Irx3 and Irx5 using Osx-Cre to drive osteoblast lineage–specific deletion of Irx3 in Irx5−/− mice. Although we found that the Osx-Cre transgene alone could also affect craniofacial mineralization, newborn Irx3flox/flox/Irx5−/−/Osx-Cre+ mice displayed additional mineralization defects in parietal, interparietal, and frontal bones with enlarged sutures and reduced calvarial expression of osteogenic genes. Newborn endochondral long bones were largely unaffected, but we observed marked reductions in 3–4-week old bone mineral content of Irx3flox/flox/Irx5−/−/Osx-Cre+ mice. Our findings indicate that IRX3 and IRX5 can work together to regulate mineralization of specific cranial bones. Our results also provide insight into the causes of the skeletal changes and mineralization defects seen in Hamamy syndrome patients carrying mutations in IRX5.


Introduction
Craniofacial development requires tight coordination of cell migration, proliferation, and mineralization of osteogenic lineages (Wilkie & Morriss-Kay, 2001;Franz-Odendaal, 2011). Osteoblast dysfunction is thought to be a major contributor to diseases that affect craniofacial bones and mineralization (Rice, 2008). The complex genetic and spatial interactions that occur during craniofacial development pose major challenges to understanding mineralization during the development of the skull.
Iroquois homeobox domain transcription factors (IRX) are highly conserved proteins that regulate neural, cardiac, and bone development (Kerner et al., 2009;Cavodeassi et al., 2001;Kim et al., 2012;Li et al., 2014). IRX proteins all contain two highly conserved domains. The homeodomain is postulated to regulate interactions between transcriptional regulators by binding to genomic regions to regulate target gene expression, and the IRO Box is involved in protein-protein binding (Cavodeassi et al., 2001;Hiroi et al., 2001). Six IRX transcription factors have been identified (IRX1-IRX6), of which Irx1, Irx2, and Irx4 cluster to chromosome 5 in humans (chromosome 13 in mice) and Irx3, Irx5, and Irx6 cluster to chromosome 16 (chromosome 8 in mice) (Gaborit et al., 2012;Houweling et al., 2001). Irx3 and Irx5 expression is strikingly similar in developing mouse tissues (Houweling et al., 2001). IRX proteins are required for the formation of limbs and skeletal tissues. Irx1 is important for the specification of individual placodes through BMP signaling (Glavic et al., 2004). IRX3 has been shown to bind to the Bmp10 promoter, which is important for ventricular septation, while IRX5 can bind to GATA3 and TRPS1 to regulate CXCL12 during bone progenitor migration in Xenopus embryos (Gaborit et al., 2012;Bonnard et al., 2012). Irx1 and Irx2 have been shown to regulate vertebrate digit formation, while Irx3 and Irx5 mediate early mouse limb bud specification by regulating Gli3 expression (Li et al., 2014;Becker et al., 2001;McDonald et al., 2010). Surprisingly, loss of Irx5 alone leads to a grossly normal mouse (Gaborit et al., 2012), while loss of both Irx3 and Irx5 together result in an embryonic lethal phenotype from cardiac defects and skeletal malformations (Li et al., 2014;Gaborit et al., 2012). Unfortunately, the early embryonic lethality of Irx3 −/− /Irx5 −/− mice contributes to our incomplete understanding of the role of Irx3 and Irx5 in osteoblast function (Li et al., 2014;Gaborit et al., 2012).
Two mutations in human IRX5, Ala150Pro and Asn166Lys occur in patients with Hamamy syndrome (OMIM MIM611174; Bonnard et al., 2012), who present with craniofacial dysmorphism, osteopenia, tooth eruption defects, and hip dysplasia, along with cardiac defects and microcytic hypochromic anemia (Bonnard et al., 2012;Hamamy et al., 2007aHamamy et al., , 2007b. The function of IRX5 seems to differ in mice and humans as the human phenotype is not observed in Irx5 −/− mice (Costantini et al., 2005). Iroquois homeodomains helix two and helix three are completely conserved in IRX1-IRX6 (Bonnard et al., 2012). The Hamamy mutations Ala150Pro and Asn166Lys occur in the IRX5 helix two or three respectively, indicating the importance of these residues in IRX homeodomain function. Irx5 was previously shown to form homodimers and heterodimers with Irx3 and Irx4 in transfected 10T1/2 cells (He et al., 2009). Previous meta-analysis of IRX3, IRX5, GATA3, and TRPS1 also indicated these proteins were involved in a regulatory network (Bonnard et al., 2012). Interestingly, patients with a 3.2 Mb deletion at 16q12.2-13, which includes IRX3, IRX5, and IRX6, show similar craniofacial features to those of Hamamy patients (Bonnard et al., 2012;Chang et al., 2010).

Body size and craniofacial bone mineralization are reduced in newborn
We focused our studies on newborn Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice ( Supplementary Fig. 1A) to understand the role of Irx3 and Irx5 in early bone mineralization. The allelic separation of Irx3 and Irx5 was a rare event (approximately 0.5-2%, Supplementary Fig. 1B) and thus single allele mice were excluded from further analysis. Newborn Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice appeared grossly normal but slightly smaller than control littermates (Fig. 1A). 43% of Irx3 flox/flox /Irx5 −/− / Osx-Cre + were viable at birth, which appeared to be an effect of the Osx-Cre allele as Irx3 +/+ /Irx5 +/+ /Osx-Cre + also showed similar decreased neonatal viability ( Supplementary Fig. 1C). However, Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice rarely survived to 4 weeks of age. This early lethality was not observed in 3-4-week-old Irx3 flox/flox / Irx5 −/− /Osx-Cre − mice ( Supplementary Fig. 1D) suggesting that the loss of Irx3 and Irx5 contributed to this decrease in survival in the neonatal stage. We observed small but statistically significant reductions in newborn total body weight and length in mice with a loss of Irx5 independent of Osx-Cre expression (Fig. 1B, C and Supplementary Fig. 2A and B). Lower limbs of Irx3 flox/flox /Irx5 −/− /Osx-Cre + assessed by Dual energy X-ray absorptiometry (DEXA) showed no differences in bone mineral density ( Fig. 1D and Supplementary Fig. 2C). These data indicate that global deletion of Irx5 alone was sufficient to affect body size, but that additional deletion of Irx3 in osteoblastic cells did not have further effects on body size.
2.2. Irx3 flox/flox /Irx5 −/− /Osx-Cre + skulls have reduced osteoblastic mineralization We next used hematoxylin and eosin staining to identify alterations to the bone architecture and mineralization in Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice. We found that bone accumulation was less in Irx3 flox/flox / Irx5 −/− /Osx-Cre + mice in the parietal and interparietal bones than in Irx3 flox/flox /Irx5 −/− /Osx-Cre − mice (Fig. 4A). Closer inspection of the frontal, parietal, and interparietal bones revealed a reduction in the thickness of mineralized bone in Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice with no major change in cuboidal osteoblasts adjacent to bony surfaces ( Fig. 4B, green arrows). Occipital bones showed no significant differences among the genotypes (data not shown). These histological analyses suggest that there is not a major change in osteoblasts morphology and numbers at the bone surfaces, but decreased osteoblast mineralization caused by loss of Irx3 and Irx5.
In studies using Xenopus laevis embryos, IRX5 interacted with GATA3 and TRPS1, forming a complex that down regulated CXCL12 production (Bonnard et al., 2012), although IRX3 and IRX5 did not directly influence Gata3 transcription (data not shown). We examined the expression of Cxcl12 and Trps1 in order to determine if the reduced bone mineralization in Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice was through downstream mediators of Gata3, Irx3, and Irx5. Cxcl12 expression was not significantly altered in Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice (Fig. 5K). Interestingly, there was a significant reduction in Trps1 in Irx3 flox/flox /Irx5 −/− /Osx-Cre + calvaria (Fig. 5L), a gene that can reduce Bglap expression in vitro and is required for proper osteoblast mineralization (Piscopo et al., 2009;Kuzynski et al., 2014). These findings suggest a role for Irx3 and Irx5 in the regulation osteoblast mineralization gene expression and suggest that in mineralization, Irx3 and Irx5 may function through a pathway that is distinct from Gata3.

Bone density is reduced in Hamamy syndrome patients
Hamamy patient mutations in IRX5 result in craniofacial dysmorphisms and mineralization defects while loss of Irx5 in mice results in no detectable bone abnormalities (Li et al., 2014). Since the global loss of both Irx3 and Irx5 leads to cardiac phenotypes similar to those seen in Hamamy patients, we wanted to see if the decreased mineralization we found in Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice also reflected the clinical presentation of Hamamy syndrome patients. Patients with Hamamy syndrome at 8 and 9 years of age displayed reduced bone mineral density with spine lumbar Z-scores of −3.7 and − 1.5 (Table 1). Bone mineral density improved with age in these patients to −2.7 and −1.4 at ages 19 and 20 years of age, respectively ( Table 1). The femoral Z-score was determined at 9 years of age for one Hamamy patient (femoral Z-score of −2.2), but both patients femoral Z-scores remained above −1.0 at 19 and 20 years of age (Table 1). We noted dramatic reductions in bone mineral content in 3-4 week old Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice with spontaneous fractures in 3 out of 10 mice, indicating that 3-4 week old Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice have similarities to Hamamy patient bone mineralization. Furthermore, the spontaneous fractures observed in 3-4 week old Irx3 flox/flox /Irx5 −/− / Osx-Cre + mice resembles the bone fragility reported in femora and other long bones of 8-10 year old Hamamy patients (Hamamy et al., 2007a).

Discussion
Proper craniofacial development requires control of bone mineralization by osteoblastic cell lineages (Mackie et al., 2008;Percival & Richtsmeier, 2013). Our studies show that osteoblast-specific loss of Irx3 and Irx5 leads to impaired mineralization in a very specific subset of cranial bones, possibly by blocking their expression of mature osteoblast mineralization genes (Fig. 7).
During the course of our study, we unexpectedly discovered that Osx-Cre mice have a newborn mineralization defect independent of the Irx3 and Irx5 mutation status. Our studies are consistent with recent reports that Osx-Cre mice alone have a newborn bone mineralization defect, specifically in intramembranous bones (Wang et al., 2015;Huang & Olsen, 2015). Interestingly, the absence of both Irx3 and Irx5 in osteoblastic cells caused an even more dramatic defect in intramembranous mineralization. Furthermore, Osx-Cre mice can survive past weaning and later stages of bone development occur normally, whereas Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice experience premature lethality around 3.5-4 weeks of age with bone fragility and spontaneous fractures. This indicates that the absence of both Irx3 and Irx5 in osteoblastic cells can influence neonatal survival at later stages of development. Our findings also emphasize the importance of using Osx-Cre + littermates as controls for studies involving skeletal development. Furthermore, our results suggest that other Cre drivers, such as Runx2-Cre or Bglap-Cre mice, may be useful for future studies to confirm early skeletal mineralization phenotypes (Elefteriou & Yang, 2011).
Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice that survive to 3-4 weeks of age have smaller femora and tibiae and appeared to have signs of bone fragility, which is consistent with reports of Hamamy syndrome patients developing bone fragility and long bone fractures later in life (Hamamy et al., 2007a). Hamamy syndrome patients also had reduced BMD that was not observed in either newborn or 3-4 week old Irx3 flox/ flox /Irx5 −/− /Osx-Cre + mice. BMD measurements in mice are not particularly sensitive and more detailed analysis of Irx3 flox/flox /Irx5 −/− /Osx-Cre + bones may be warranted. Our data demonstrate that IRX3 and IRX5 are important for both early osteoblast mineralization function and later skeletal mineralization, which also will help in understanding the bone fragility that occurs in Hamamy patients.
While it is clear that IRX3 and IRX5 can regulate cranial bone mineralization, the mechanism for how IRX3 and IRX5 control bone mineralization remains unclear. IRX5 might bind to GATA3 and TRPS1 proteins in a complex that regulates cranial neural crest cell migration (Bonnard et al., 2012). However, our results did not show differential expression of Cxcl12 (Gaborit et al., 2012), a target of IRX3 and IRX5 which is thought to be regulated by GATA3. In addition, the affected bones in Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice were largely derived from the mesenchyme, rather than neural crest derivatives. Furthermore, Trps1 levels were significantly reduced in Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice suggesting that loss of mineralization is occurring through a Trps1 specific pathway (Piscopo et al., 2009). Indeed, previous studies have shown that murine Irx5 co-immunoprecipitated with Gata3 and Trps1. When co-immunoprecipitation was done with these proteins with the Irx5 Asn166Lys mutation, there was markedly less binding to Trps1 and Gata3, demonstrating that Irx5 binding is reduced by Hamamy mutations and that Trps1 binding to Irx3 and Irx5 is likely affected in Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice (Bonnard et al., 2012).
Gata3 −/− mice are embryonically lethal from noradrenaline deficiency (Lim et al., 2000) and Gata3 −/− rescued embryos display cranial bone development defects, but Gata3 +/− mice appear to have no bone developmental abnormalities, which suggests that Gata3 is important for the development of skeletal tissues, but may not be involved in the regulation of bone mineralization gene expression (Lim et al., 2000;Pandolfi et al., 1995). Unfortunately, the expression of Trps1 has not been reported in Gata3 +/− and Gata3 −/− mice (Lim et al., 2000). Future studies to understand how Trps1 is regulated at specific stages of osteoblast mineralization, and if this defect is associated with the early lethality prior to puberty, will help us to determine the role of Trps1 in the reduced mineralization of Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice.
Finally, why mice require loss of both IRX3 and IRX5 to develop reductions in mineralization similar to Hamamy patients who carry a nonsense mutation in the IRX5 gene remains unclear (Gaborit et al., 2012). One notion to explain the similar reductions in bone mineralization between Irx3 flox/flox /Irx5 −/− /Osx-Cre + mice and Hamamy patients is that deletion of both IRX3 and IRX5 in osteoblasts removes the ability to compensate for the loss of these proteins, whereas Hamamy IRX5 nonsense mutations influences IRX3 and IRX5 heterodimer formation and downstream mineralization function (Gaborit et al., 2012;He et al., 2009). More detailed analysis of Hamamy syndrome patients cells using a human induced pluripotent stem cell model may help demonstrate the dynamics of IRX3 and IRX5 interactions in Hamamy patient cells.
In conclusion, we identified a novel role for IRX3 and IRX5 in early cranial mineralization of osteoblastic cells. Furthermore, Irx3 flox/flox /Irx5 −/− / Osx-Cre + mice displayed reduced bone mineralization without affecting early osteogenic gene expression. Finally, our finding that Irx3 flox/flox / Irx5 −/− /Osx-Cre + mice have reduced osteoblastic mineralization indicates that IRX3 to IRX5 binding maintains an important role in Hamamy syndrome and understanding the role of IRX3 and IRX5 together will help provide insight into the roles of IRX proteins in other organs.

X-ray analysis of Jordanian and Turkish patients with Hamamy Syndrome
The Jordanian patients were originally described by Hanan Hamamy at Jordan University Hospital (Hamamy et al., 2007a). The Turkish family was diagnosed by Hülya Kayserili at the Medical Genetics Department of the Istanbul Medical Faculty (Bonnard et al., 2012;Hamamy et al., 2007a). Both sets of patients provided informed consent for radiographs to be published, and all studies have been approved by the local ethic commissions as described (Bonnard et al., 2012). Radiograph analysis was done on IRX5 Asn166Lys and IRX5 Ala150Pro patients and compared to control to assess osteopenia and craniofacial dysmorphisms. DEXA was used to measure area, BMD, and BMC of the lumbar spine and femoral neck, and to calculate Z-score. DEXA were performed just before puberty age (8-9 years old) and at adult age (19-20 years old).

Alizarin red and alcian blue staining of skeletons
Newborn mice and 3-4 week old mice of both sexes were euthanized and prepped for alizarin red and alcian blue skeletal staining (Ovchinnikov, 2009) by fixing in 100% ethanol for 24 h. Samples were then switched to acetone (Sigma-Aldrich) for an additional 24 h. Once fixed, samples were stained with final concentration of 5% glacial acetic acid, 0.5% alizarin red S (Sigma-Aldrich), 0.9% alcian blue 8GX (Sigma-Aldrich) in ethanol for 3 h at 37°C and then at room temperature for 24 h. Samples were then placed in 1% KOH (Amresco) for 3 h and replaced with fresh KOH until non-bone tissue was transparent. Samples were then replaced with increasing concentrations of glycerol and photographed with a Leica MZFLIII dissection microscope with Diagnostic Instruments 14.2 Color Mosaic camera for newborn samples. 3-4 week old samples were photographed with a Nikon E5200 without a microscope.

Histology
Newborn skulls were skinned and fixed in neutral buffered formalin for at least 48 h and then replaced with 70% ethanol for at least 24 h. Skull tissues were paraffin embedded and sectioned. Skulls were then cut at the midline and then stained with hematoxylin & eosin, using standard protocols (J. David Gladstone Institutes Histology Core).

Bone densitometry and microCT imaging
DEXA was used to measure mouse whole-body BMD and BMC. Mice were anesthetized with inhaled isofluorane (1.5% to 2% in oxygen) and scanned on a GE Lunar Piximus2 (Piximus). Newborn mice that underwent whole-mouse microCT scans were sacrificed and stored in 70% ethanol before scanning. Ex vivo images were obtained on a Scanco vivaCT-40 microCT scanner (SCANCO) at an X-ray energy of 55 kV, with sigma 0.8/support 1/threshold 120 (103.7 mg HA/cm 3 ), a voxel size of 76 μm, and integration times of 200 ms for whole-body images.
4.6. RNA isolation, cDNA synthesis, and qPCR Whole calvaria or dissected calvarial tissues were placed in Trizol (Invitrogen) and homogenized using a Powergen 125 homogenizer (Fisher). RNA was isolated using chloroform extraction for whole calvaria or by Picopure RNA isolation columns for dissected calvaria (Life Technologies). Purified mRNA was then used as a template to synthesize cDNA with oligo dT primers with the Superscript III (Invitrogen) kit as described (Cain & Manilay, 2013). qPCR expression analysis was performed using TaqMan primers for qPCR reactions (Supplementary  Table 1) on a Viia7 real-time thermocycler (Applied Biosystems) run in 5 μl sample volumes in triplicate or preamplified using Fludigm preamplication qPCR mix and assayed using Fluidigm dynamic array IFC qPCR plates (Fludigm). All expression values were normalized to Gapdh levels.

Statistics
Differences between the means of biological replicates for all analyses were calculated using two tailed Student's T-test (GraphPad Prism. La Jolla, CA). Analyses were considered statistically significant if p ≤ 0.05.

Conflict of interest statement
Edward Hsiao receives funding from Clementia Pharmaceuticals for an unrelated clinical trial.