The regenerative capacity of the adult skeleton depends on the re-induction of cellular and molecular cues that occur during embryonic development.1 Fracture repair of long bones involves an inflammatory step followed by the formation of a cartilaginous callus. This callus is then replaced with bone, and excess bone callus is finally remodeled to achieve bony union.2 In the developing limb, mesenchymal condensations undergo a series of physiological changes to form cartilage and bone, whereas postnatally the formation of cartilaginous callus during fracture repair requires the recruitment of periosteal progenitor cells and their differentiation into chondrocytes and osteoblasts.3–5 Failure to form a proper cartilaginous callus results in impaired fracture healing.6–8 Understanding the molecular mechanisms of progenitor cell differentiation into chondrocytes during fracture healing is important for developing novel therapeutic approaches to prevent delayed healing and reverse non-unions.
The Runx proteins (Runx1, Runx2, and Runx3) are the mammalian homologues of the Drosophila runt and lozenge transcription factors.9, 10 The Runx proteins are expressed in a variety of tissues, and play key functions ranging from organogenesis to pathogenesis. Runx2 is essential for bone development11, 12 and Runx3 is important during nerve cell differentiation.13, 14 Genetic evidence indicates Runx1 is required for definitive hematopoiesis, neuronal differentiation and survival, T cell development, and formation of the sternum.15–20 Similarly, conditional deletion of Runx1 in mouse bone marrow causes impairments in megakaryocyte and lymphocyte development leading to thrombocytopenia and lymphopenia in mice.21 In humans, haploinsufficiency of Runx1 results in an autosomal dominant syndrome called familial platelet disorder with propensity to acute myeloid leukemia.22
The overlapping expression of Runx proteins in skeletal elements suggested multiple, perhaps overlapping roles for all three Runx proteins during bone and cartilage development. For example, Runx1 is expressed in immature chondrocytes, perichondrium, periosteum, rib, vertebrae, limb, mandible, and maxilla.23–26 Furthermore, our in vitro studies suggest that Runx1 mediates mesenchymal stem cell commitment to the chondrogenic lineage.26 However, the role of Runx1 during skeletal development in vivo is not fully understood, and its potential role in skeletal repair remains unknown. Thus, the early embryonic lethality of the Runx1 global knockout mice15, 16 warranted the development of other strategies to explore Runx1's potential role during cartilage development and fracture repair. Here we describe the generation of a hypomorphic Runx1 allele, which we used to generate in vivo models with incremental loss of Runx1 function. We also executed a conditional deletion of Runx1 in mesenchymal chondro/osteoprogenitor cells using a floxed allele21 in conjunction with Prx1-Cre.27 Using these animal models, we determined the in vivo function of Runx1 during skeletal development and repair.
Subjects and Methods
We generated the Runx1L148A allele by homologous recombination as described for other mutations in exon 4 sequences.28 Runx1+/+, Runx1L148A/+, Runx1L148A/L148A, Runx1+/− (Runx1− = Runx1tm1Spe),16 Runx1L148A/−, and Runx1L148A/R174Q (Runx1R174Q = Runx1tm5Spe)28 were generated by intercrossing the appropriate mouse strains. Runx1F/+, Prx1-Cre;Runx1F/+ and Prx1-Cre;Runx1F/F were generated from the breeding of Runx1F/F (Runx1F = Runx1tm3Spe)21 and Prx1-Cre mice.27 Mice were housed at the University of Connecticut Health Center (UCHC) and Dartmouth Medical School animal facilities according to state and federal law requirements.
Yeast one- and modified one-hybrid assays
We conducted the yeast one- and modified one-hybrid assays as described.28–30
Equilibrium binding constants
We purified the Runt domain from bacteria, and we performed quantitative electrophoretic mobility shift assays as described,28, 30, 31 using the Core site from the SL3–3 murine leukemia virus enhancer (GGATATCTGTGGTTAAGCA) as the DNA probe 1.
Clinical blood cell counts
We collected retro-orbital blood from 6- to 10-week-old mice, transferred it to tubes coated with EDTA (Sarstedt Inc., Newtown, NC, USA), and counted it on a Drew Scientific Hemavet 850.
Osteocalcin promoter luciferase activities
We isolated bone marrow cells from the long bones of 8-week-old mice. Cells were maintained in α minimum essential media (α-MEM) containing 10% fetal calf serum and antibiotics in the presence of 30 ng/mL of macrophage colony-stimulating factor (M-CSF). Cells were co-transfected with 400 ng of the osteocalcin promoter plasmid and 100 ng of renilla luciferase vector using the FuGENE HD transfection reagent (Roche, Indianapolis, IN, USA) and incubated for 30 hours. Luciferase activity was measured using a dual-luciferase assay kit (Promega, Madison, WI, USA), and luminescence was detected with a luminometer (Lumat LB 9507; Berthold Technologies, Oak Ridge, TN, USA).
Femoral fracture procedure
We anesthetized the animals and gave them analgesic prior to the surgical procedure following the principles and procedures approved by the Animal Care Committee at UCHC. Femoral fractures were performed using an Einhorn device32 on the left limb of 7- to 9-week-old female mice as described.7, 33, 34 Fractures were radiographed by a Faxitron X-ray system (Faxitron X-ray, Wheeling, IL, USA) under anesthesia to monitor initial pin fixation as well as ongoing callus formation at days 7, 14, and 21 postfracture. Mice were euthanized on days 7, 14, and 21 and fractured legs were collected for micro–computed tomographic (µCT), histologic, and histomorphometric analyses. Fracture callus were also harvested at day 7 for total RNA extraction and immunohistochemistry.
After euthanasia, we disarticulated the fractured femurs at the hip and knee, dissected them of soft tissue, and fixed them in 10% neutral buffered formation overnight at room temperature. Callus morphometry was measured using cone-beam microfocus X-ray computed tomography (µCT40; Scanco Medical AG, Brüttisellen, Switzerland) as described.7 Standard algorithms describing trabecular morphometry were applied to obtain direct measures of both total and mineralized callus.
Histological and histomorphometric analyses
We harvested tibias and fractured femurs for histological assessments as described.7, 34 For histomorphometric analyses, sections were stained with hematoxylin and eosin (H&E) and the lengths of the proliferating and hypertrophic zones were determined as described.35 The perimeters of stromal, cartilaginous, and bony tissues were outlined in the fracture callus and these regions were quantified using NIS-Elements BR 3.0 (Nikon, Melville, NY, USA).
We deparaffinized sections from paraffin-embedded tibias and fractured femurs and sequentially incubated them with either 4N HCl (anti-Sox9 and anti-Runx3) or citrate buffer, pH 6.0 (anti-Runx2) for antigen retrieval (anti-Sox9 [ab59265], anti-Runx2 [ab54868] and anti-Runx3 [ab68938] from AbCam, Cambridge, MA, USA). Incubation and detection was performed following our published protocol.33, 36 Image acquisition was performed using a Q-Imaging Retig 200R camera connected to a Nikon Eclipse 50i microscope and image analyses were performed using an NIS-Elements BR 3.0 software (Nikon).
Gene expression analysis
We extracted total RNA from fracture calluses for gene expression analyses via Q-RT-PCR as described.7, 34 Specific primers were used for murine Runx1, Runx2, Runx3, Sox9, Sox5, Sox6, Type II collagen, and Type X collagen (see Table 1). Gene expression was normalized to that of β-actin. PCR products were analyzed and quantified using the StepOne Software (Applied Biosystems, Foster City, CA, USA).
Table 1. Mouse Primers Used for Real-Time RT-PCR Experiments
CGTTCT TCACCGACT TCCTC
Type II collagen
Type X collagen
Limb bud micromass cultures
We harvested forelimbs and hindlimbs from Runx1F/F embryos at embryonic day 12.5 (E12.5) and isolated limb bud cells as described.26, 37 Cell suspensions (2 × 107/mL) were transduced with either Cre adenovirus or the green fluorescent protein (GFP) control virus at 100 multiplicity of infection (MOI) in growth media. Infected cells seeded at a density of 2 × 105 cells per 10 µL for 2 hours to allow adherence. Micromass cultures were then submerged in media supplemented with 25 µg/mL ascorbic acid and 54 µg/mL beta-glycerol phosphate. Media were changed every 2 to 3 days and cells were harvested at days 4, 7, and 14 for Alcian blue staining, Western blotting, and chromatin immunoprecipitation (ChIP) assays.
We subjected 10 µg of nuclear proteins harvested from micromass cultures to SDS-PAGE as previously published.36, 37 Western blotting was accomplished using antibodies for Runx1 (Calbiochem, La Jolla, CA, USA), Runx2, Runx3, Sox9 (anti-Sox9 [ab59265], anti-Runx2 [ab54868], and anti-Runx3 [ab49117] from AbCam, Cambridge, MA, USA), or actin (A2668, Sigma, St. Louis, MO, USA) and corresponding secondary antibodies (anti-mouse immunoglobulin G [IgG] 7076, anti-rabbit IgG 7074, Cell Signaling Technology, Danvers, MA, USA). Chemiluminescent detection was performed using Pierce Femto reagent (Pierce, Rockford, IL, USA). Density of each band was analyzed using the NIH Image J analysis software.
We prepared limb bud micromasses as described above in Limb bud micromass cultures. Cells were harvested at days 4, 7, 10, and 14, and chromatin was isolated using the ChIP-IT kit (Active Motif, Carlsbad, CA, USA) for ChIP assay following the manufacturer's instructions. Ten micrograms (10 µg) of sheared fragments were immunoprecipitated with 5 µg anti-RUNX1 antibody (ab23980; Abcam) or the same amount of IgG antibody as negative control. Immunoprecipitated and input DNA were analyzed by real-time PCR using primers that embrace Runx1-responsive elements on Sox9, Sox5, and Sox6 promoter regions (Table 2).
Table 2. Mouse Primers That Embrace Runx1-Responsive Elements of the Sox9, Sox5, and Sox6 Promoters
Creation of a hypomorphic Runx1 allele
Runx1 interacts with DNA through a DNA-binding sequence called Runt domain, which also interacts with Runx1 partner protein, CBFβ.30 To generate a hypomorphic Runx1 allele, previously generated Runt domain mutants targeting exon 4-encoded sequences (amino acids 141–177) were analyzed to identify mutations that result in disruption of DNA binding.28, 30 Semiquantitative yeast one-hybrid analyses using a liquid β-galactosidase assay was employed to measure the ability of Runt domain proteins, containing one of five different mutations: K144A, L148A, V152A, P156A, or A160G29, 30 to bind DNA. All of these mutations decreased DNA binding in the range of fourfold to 13-fold (Fig. 1A). A modified yeast one-hybrid assay was also used to measure the binding of the Runt domain:CBFβ heterodimer to DNA.28, 30 The mutations decreased DNA binding of the heterodimer by twofold to sixfold (Fig. 1B). Interestingly, several of the mutations (L148A, P156A, and A160G) negatively impacted DNA binding of the isolated Runt domain more extensively than they impaired DNA binding by the Runt domain:CBFβ heterodimer complex. All of the mutations were predicted to perturb the fold of the Runt domain (or in the case of K144A a salt bond).30 CBFβ stabilizes some of the mutant Runt domains (L148A, P156A, and A160G) to a greater extent than others (K144A and V152A), ameliorating the effects of the mutations on DNA binding activity.
Because the K114A and L148A mutants most negatively impacted DNA and CBFβ, we specifically focused on these mutations. Neither of these residues directly contacts DNA or CBFβ. K144 is located in the βE′-F loop in close proximity to the DNA contacting residue R142.29, 38 The K144A mutation decreased binding in the yeast one-hybrid assay by 12.8-fold (Fig. 1A), and decreased DNA binding by the heterodimer (modified one-hybrid) by sixfold (Fig. 1B). The L148 side chain is in the βF strand, buried in the hydrophobic core of the Runt domain. The L148A mutation impaired DNA binding of the isolated Runt domain by 10.1-fold, and of the heterodimer by 2.1-fold. Differences in equilibrium dissociation constants for the Runt domain–DNA complex measured by quantitative electrophoretic mobility shift assays28, 30 were consistent with those obtained in the yeast one-hybrid analyses (Fig. 1C).
The K144A and L148A mutations were targeted into exon 4 of the Runx1 locus using the strategy described28 (Fig. 1D). Mice heterozygous for the K144A and L148A mutations appeared grossly normal. Multiple litters from Runx1K144A/+ intercrosses yielded only one live Runx1K144A/K144A pup out of a total of 51, indicating that homozygosity for this mutation was lethal in most progeny. Runx1L148A/L148A mice, on the other hand, were viable, and indistinguishable from their wild-type littermates. However, when the Runx1L148A allele was carried over a nonfunctional (Runx1−) or weakly dominant negative Runx1 allele (Runx1R174Q)28 the frequency of viable offspring was reduced (Supplementary Table 1). Clinical blood counts from mice homozygous for the Runx1L148A allele, or Runx1L148A/− mice revealed significantly lower numbers of red blood cells, hemoglobin, hematocrit levels, lymphocytes, platelets, and neutrophils compared to wild-type mice, whereas most of these parameters (with the exception of platelet numbers) were equivalent in wild-type and Runx1+/− mice (Supplementary Table 2). Altogether the data indicate that Runx1L148A was a hypomorphic Runx1 allele.
Increased loss of Runx1 function results in growth abnormalities
A hypomorphic allelic series was generated consisting of wild type, Runx1L148A/+, Runx1+/−, and Runx1L148A/– mice to study the role of Runx1 in skeletal development and fracture repair. Examination of the gross phenotypes during the exponential growth phase (4 weeks postnatal) showed that mice with the highest loss of Runx1 function (Runx1L148A/–) were runted (Fig. 2A). This is further evidenced by X-ray analyses showing shorter femurs in Runx1L148A/– mice compared to their Runx1+/+, Runx1L148A/+, and Runx1+/− littermates (Fig. 2A). This runted phenotype correlated with lower body mass of Runx1L148A/– mutants compared to other genotypes for up to 11 weeks of age (data not shown). Marrow cells were isolated from each genotype and transfected with the osteocalcin promoter reporter, which contains three Runx binding sites to measure Runx1 activity. Although promoter activity was reduced by 50% in Runx1+/− mice, it was further decreased by 70% in the Runx1L148A/– mutants (Fig. 2B).
Because of the impaired skeletal growth in Runx1L148A/– mice, histological examination of growth plates was performed at 4 weeks of age. Figure 2C shows that although the incremental loss of Runx1 function does not impair the organization of growth plate, the lengths of the proliferating and hypertrophic zones were enhanced with increasing loss of Runx1 function. Histomorphometric analyses showed that the ratio of the length of the proliferating zone divided by the length of the Runx1L148A/– tibia was increased by 9%, 21%, and 53% in Runx1+/−, Runx1L148A/+, and Runx1+/+ mice, respectively. Runx1L148A/– and Runx1+/− mice also displayed significantly longer hypertrophic zones compared to Runx1+/+ (37% and 24%, respectively) (Fig. 2D). These changes were not observed during embryonic development when we examined the gross morphology of long bones at E16.5, indicating that these effects are most likely postnatal (data not shown).
Decreased expression of chondrogenic markers in growth plate chondrocytes of mice with decreased Runx1 activity
We examined protein expression of Sox9, Runx2, and Runx3 in the proliferating zones of the growth plates of all genotypes by immunohistochemistry (Supplementary Figure 1). As expected, although Runx2 was expressed in hypertrophic chondrocytes, Sox9 and Runx3 were excluded from these populations in all genotypes (data not shown). We quantified the number of Sox9-, Runx2-, and Runx3-expressing cells in the proliferating zone as percentages of immunoreactive cells over total number of cells per fixed surface area. We found that when Runx1 activity was abrogated by 50% or more, the number of Sox9, Runx2, and Runx3 cells was significantly decreased in the proliferating zone.
Inhibited Runx1 activity results in delayed fracture healing
To examine whether Runx1 plays a role during bone repair, we assessed the ability of Runx1+/+, Runx1L148A/+, Runx1+/−, and Runx1L148A/– mice to heal their bones following a femoral fracture. Radiographic evaluation of the mid-diaphyseal fractures over a time course of 21 days postfracture showed callus formation in fractured bones from all genotypes. At day 14, while radiolucency persisted in calluses of Runx1L148A/– mice, the other genotypes showed fully mineralized bridging. By day 21, all genotypes had achieved union (Supplementary Figure 2A).
µCT imaging showed decreased callus size with an unmineralized void at the center of calluses in the Runx1L148A/– mice compared to mice of other genotypes at 14 days postfracture (Supplementary Figure 2B). By day 21 all mineralized calluses were smaller than at day 14, indicating the onset of fracture callus remodeling (Supplementary Figure 2B). Quantitative analyses of these images confirmed that total callus size was significantly smaller in Runx1L148A/– mice at day 14 compared to the wild-type control (Supplementary Figure 2C), and mineralized callus size was also reduced (Supplementary Figure 2D). Total callus size was significantly decreased in all genotypes at day 21 compared to day 14 (Supplementary Figure 2C). However, no significant differences were observed between day 14 and day 21 in mineralized callus size of Runx1L148A/– mice (Supplementary Figure 2D). The data indicate that global loss of Runx1 activity is concomitant with delayed fracture repair of long bones.
Incremental loss of Runx1 function results in altered cartilaginous callus formation during bone fracture healing
We complemented our imaging results with histological and histomorphometric analyses of safranin-O/fast green-stained fractured calluses of all genotypes at days 7, 14, and 21 postoperation. Total callus size was smaller in Runx1+/− and Runx1L148A/– compared to Runx1+/+ and Runx1L148A/+ mice at day 7 postfracture (Fig. 3A). At 14 days postfracture, while the majority of the cartilaginous callus from Runx1+/+, Runx1L148A/+, and Runx1+/− fractured bone was replaced with mineralized bone, residual cartilage remained in calluses from Runx1L148A/– mice. Although callus sizes were similar among all genotypes at day 21 postfracture, residual cartilage persisted in the mineralized callus of Runx1L148A/– mice.
Histomorphometric analyses showed that cartilage content in the early (day 7) fracture calluses was significantly decreased in Runx1+/− and Runx1L148A/– mice compared to that of Runx1+/+ and Runx1L148/+ mice (Fig. 3B). Conversely, a significant increase in cartilage content of the calluses from Runx1L148A/– mice was seen at days 14 and 21 compared to all other genotypes. Stromal tissue contents were also significantly decreased in Runx1L148A/– mice at day 7. The bony callus content remained unchanged between all genotypes at days 7 and 14. However, at day 21 postfracture, significant amounts of residual bony callus persisted in Runx1L148A/– mice compared to Runx1+/+ mice. Together, the data indicate that loss of Runx1 function results in delayed cartilaginous callus formation and remodeling, leading to impaired fracture healing.
Because of the changes in the soft callus composition in Runx1L148A/–, we performed immunohistochemical analyses to examine the cellular representation of other chondrogenic transcription factors using specific antibodies for Sox9, Runx2, and Runx3. Images of immunostained calluses of all genotypes are presented in Supplementary Figure 3. Immunohistochemical data was quantified by counting the number of Sox9-, Runx2-, and Runx3-positive cells per fixed surface area in both the stroma and cartilage of fracture calluses. No significant differences were observed in the percentages of Sox9, Runx2, and Runx3 in the stroma of calluses between Runx1+/+ and Runx1L148A/+ calluses (Fig. 4). The percentages of Sox9-, Runx2-, and Runx3-positive cells were significantly decreased in the stroma of Runx1+/− compared to Runx1+/+ calluses by 39%, 38%, and 51%, respectively. This decrease was further accentuated in the stroma of Runx1L148A/– calluses by 81%, 92%, and 93% in comparison to Runx1+/+, respectively. The number of Sox9-positive cells was slightly but significantly decreased in the cartilaginous calluses of Runx1L148A/+ compared to Runx1+/+. No significant changes in the percentages of Runx2- or Runx3-positive cells were observed between Runx1L148A/+ and Runx1+/+ calluses. The percentages of Sox9-, Runx2-, and Runx3-positive cells were all significantly decreased in cartilaginous calluses of Runx1+/− compared to Runx1+/+ mice by 29%, 36%, and 47%, respectively, and that of Runx1L148A/– compared to Runx1+/+ mice by 25%, 39%, and 38%, respectively. Together, the data suggest that Runx1 enhanced the differentiation of mesenchymal progenitor cell into chondrocytes in the fracture callus, possibly through modulation of Sox9, Runx2, and Runx3.
We also performed Q-PCR analyses of Runx1, Runx2, and Runx3, and cartilage phenotypic marker genes Sox9, type II collagen, and type X collagen using RNA extracted from fracture calluses at day 7 postfracture. Runx1 transcripts were reduced in Runx1+/− and Runx1L148A/– mice. Neither Runx2 nor Runx3 expression was affected by the loss of Runx1 activities. However, gene expression of the chondrocyte phenotypic markers Sox9 and type II collagen was inhibited in the calluses of Runx1+/− mice, and to a greater degree in Runx1L148/–mice. Type X collagen expression was only significantly decreased in Runx1L148/–mice (data not shown).
Impaired cartilaginous callus formation following tissue-specific deletion of Runx1
The periosteum contains Prx1-positive mesenchymal stem cells that can differentiate into chondrocytes during cartilaginous callus formation.5 Runx1F/F mice were crossed with Prx1-Cre transgenic mice and femoral fractures performed in the resulting Prx1-Cre;Runx1F/F mutants. Mice with 100% Runx1 activity (Runx1F/+) and partial deletion of Runx1 function (Prx1-Cre;Runx1F/+) were used as controls. Figure 5A shows that the area of the cartilaginous tissue and bony tissue within the fracture calluses were comparable between Prx1-Cre;Runx1F/+ and Runx1F./+. Conversely, the fracture callus of Prx1-Cre;Runx1F/F mice was much smaller than those of Prx1-Cre;Runx1F/+ and Runx1F/+ littermate controls, with limited cartilage formation. Histomorphometric analyses showed that Prx1-Cre;Runx1F/F calluses contained significantly reduced amounts of both cartilaginous and bony tissues compared to Prx1-Cre;Runx1F/+ and Runx1F/+ (Fig. 5B). We further extracted RNA from fracture calluses of Prx1-Cre;Runx1F/F and Runx1F/+ at day 7 postfracture and examined mRNA levels of Runx1 as well as Runx2, Runx3, Sox9, Sox5, and Sox6 (Fig. 5C). We observed a significant reduction of Runx1 expression by 81% in fracture callus of Prx1-Cre;Runx1F/F mice, and significant decreases in Sox9, Runx2, and Runx3 expression by 42%, 29%, and 39%, respectively. Levels of Sox5 and Sox6 expression were also decreased by 23% and 43%, respectively (data not shown). These results suggest that Prx1-mediated deletion of Runx1 resulted in decreased cartilaginous callus formation in the early stages of fracture healing.
Runx1 regulates mesenchymal progenitor cell differentiation into chondrocytes
To examine whether Runx1 is required for mesenchymal cell differentiation into the cartilaginous phenotype, we performed in vitro deletion of Runx1 from Runx1F/F derived micromass cultures using adeno-Cre viruses or adeno-GFP controls. Runx1 protein expression was efficiently inhibited in the adeno-Cre–infected cultures at all time points compared to GFP-infected controls (Fig. 6A). Expression of Runx2 and Sox9 proteins was significantly inhibited at day 4 following Runx1 deletion (by 59% and 63%, respectively), then it was significantly upregulated at 14 days (by 3.7-fold and 7.2-fold, respectively) compared to controls. Runx3 protein levels were only significantly inhibited in micromass lacking Runx1 after 4 days in culture (by 44%). Loss of Runx1 function also resulted in decreased chondrogenesis and chondrocyte differentiation compared to GFP-infected controls as evidenced by decreased Alcian blue staining at day 4 and fewer nodules at later stages (Fig. 6B). The data indicate that Runx1 regulates chondrogenesis of a mesenchymal progenitor cell population in vitro, independently from its potential role in hematopoiesis.
Runx1 directly binds regulatory sequences on the Sox5, Sox6, and Sox9 promoters
Given the effects observed on chondrogenesis, we examined the direct occupancy of the promoter regions of the three major chondrogenic factors Sox9, Sox5, and Sox6.39 We first identified two 100% matching Runx responsive elements between nucleotides –1705 and –1700 and between nucleotides –1047 and –1042 on the Sox9 promoter sequence. Similarly, the Sox5 and Sox6 promoters harbor Runx1 response elements (between nucleotides –1824 and –1818 for Sox5, and between nucleotides –2923 and –2918 as well as between nucleotides –1123 and –1119 for Sox6) (Fig. 7A).
Figure 7B shows that Runx1 binds directly to all response elements on Sox9, Sox5, and Sox6 promoters as evidenced by ChIP assay analyses using chromatin from limb-bud micromass cultured for 4, 7, 10, and 14 days. A single PCR product for each of the Sox9, Sox5, and Sox6 promoter regions was obtained in input regardless of the stage of chondrocyte differentiation (data not shown). Conversely, no or minimal PCR amplification was observed in chromatin extracts that were immunoprecipitated with a negative IgG control antibody (data not shown). Quantification of these bindings (Fig. 7B) showed that binding of endogenous Runx1 to both sites on the Sox9 promoter was increased with chondrocyte differentiation between days 4 and 10 then decreased by day 14. Similarly, all Runx binding sites on the Sox5 and Sox6 promoters were gradually occupied with Runx1 protein between days 4 and 14 with a maximum of residence in 10-day micromass cultures. The data indicate that Runx1 directly bound regulatory sequences of the promoters of Sox9, Sox5, and Sox6 genes.
The observation that Runx1 is expressed in skeletal elements of developing mice23–26 and the periosteum of adult mice24 prompted us to examine its potential roles in skeletal tissue. Additionally, whereas the function of Runx1 during skeletal development remains poorly understood, its role during skeletal repair was completely unknown. Thus, to circumvent the embryonic lethality of the Runx1 global knockout mice,16 we generated surviving mice with global incremental loss of Runx1 function and found that mutants with the lowest level of Runx1 activity (Runx1L148A/–) were runted. This runted phenotype was observed in Runx1L148A/– mice but not in mice haploinsufficient for Runx1, indicating that more than 50% loss of Runx1 activity was required to observe a growth plate phenotype. These findings are consistent with findings by Yoshida and colleagues,40 showing that Runx2−/− mice and Runx1+/−; Runx2−/− mice exhibited a similar degree of delayed chondrocyte differentiation, indicating that haploinsufficiency for Runx1 may not be sufficient to impair chondrogenesis; even when Runx2 is also deleted. Our data clearly indicate that a threshold of Runx1 is necessary to mediate proper growth.
The Runx1L148A/– mice displayed extended proliferating and hypertrophic zones within their growth plates. Similar phenotypes were previously observed in neonatal pups lacking the bone-related isoform of Runx2, which exhibited elongation of both the proliferating and hypertrophic zones.41 Subsequent delay in chondrocyte differentiation and terminal maturation also occurred as a consequence of absence or decrease of Runx3.36, 40 Mice haploinsufficient for Sox9 exhibited extension of their hypertrophic zone whereas their proliferating zone was not affected.42 Because we demonstrated that incremental loss of Runx1 function resulted in decreased number of Sox9-, Runx2-, and Runx3-positive cells in the proliferating zone of the Runx1L148A/– growth plates, we speculate that the delayed endochondral ossification in the Runx1L148A/– mice was likely due to the inhibition of chondrocyte differentiation and subsequently that of terminal maturation.
We and others have generated tissue-specific Runx1 deletions in chondrocytes using the collagen type II-Cre transgenic mice19 (data not shown). No skeletal phenotype was observed in these conditional knockouts, indicating that Runx1 may be important at a stage of mesenchymal cell differentiation into the chondrogenic phenotype that is prior to type II collagen expression. Interestingly, when Runx1 was specifically deleted in limb mesenchyme using Prx1-Cre, only a slight and transient inhibition of sternal and xiphoidal mineralization was observed.19 Limb-specific deletion of Runx1 and global loss of Runx2 double mutants (DKO: Prx1-Cre;Runx1F/F;Runx2−/−) also displayed a defective rib cage due to the absence of sternal structures, but no limb phenotype was observed.19 When Liakhovitskaia and colleagues20 generated mice that express Runx1 only in endothelial and hematopoietic cells they found that sternebrae, the mineralization centers in the sternum, were less well developed at E17.5 to E18.5 (the mice died at birth), but noted no limb defects. Conditional deletion of Runx1 in all blood cells using Vav1-Cre also failed to produce a growth phenotype.43 It is possible that the difference in limb phenotypes between the Prx1-Cre;Runx1F/F conditional knockouts and the Runx1L148A/– mice could only be reconciled if the mesenchymal and hematopoietic cell lineages both require Runx1 for normal limb development. Because of the heterogeneity of the cell populations that contribute to limb development, we speculate that both Prx1-positive and Prx1-negative cells contribute to the overall development of the limb in vivo, which could explain why we did not see a limb phenotype in the Runx1 conditional knockouts. The 90% loss of Runx1 function in our micromass model further supports a role for Runx1 in regulating mesenchymal chondroprogenitor cell differentiation in vitro.
When we challenged adult Runx1L148A/– and Prx1-Cre;Runx1F/F mice for femoral fractures, both genetic models displayed impaired cartilaginous callus formation during bone repair. This delay in fracture healing may be due to inhibited recruitment of mesenchymal progenitor cells at the fracture site and a subsequent delay in chondrocyte differentiation. Interestingly, the absence of a limb phenotype in the fetus of Prx1-Cre;Runx1F/F contrasts with the obvious defect in fracture callus formation observed in adult mice. This is not the first report in which a phenotype is revealed by injury or stress. For example, mice with deletion of the bone morphogenetic protein BMP-2 by Prx1-Cre did not display severe limb abnormalities, yet when fractured, cartilaginous callus formation was abrogated.8 One of the reservoirs of Prx1-positive cells in adult mice is the periosteum, and these Prx1-positive periosteal progenitors are recruited to the fracture site for cartilaginous callus formation during fracture repair.5 Thus, our data imply that unlike embryonic skeletal development, intact Prx1-positive periosteal progenitor cells are necessary for cartilaginous callus formation during fracture healing. We speculate that loss of Runx1 function in Prx1-positive periosteal cells may lead to decreased commitment of these progenitor cells to the chondrogenic lineage.
Runx1L148A/– mice exhibited significantly decreased expression of Sox9-, Runx2-, and Runx3-positive stromal cells and chondrocytes within the fracture calluses. Fracture calluses of Prx1-Cre;Runx1F/F mutants also showed significantly reduced mRNA expression of Sox9, Sox5, and Sox6, which is reflective of decreased mesenchymal progenitor cell commitment to the early chondrogenic phenotype. Furthermore, Runx2 and Runx3 transcripts were inhibited in these calluses. This finding could explain the decreased cartilaginous callus formation observed. Our in vitro limb bud cultures in which Runx1 was abrogated demonstrated that Sox9, Runx2, and Runx3 gene expression was only inhibited in the early stages of chondrocyte differentiation. These results further suggest that Runx1 function in cartilage is primarily associated with the early stages of chondrogenesis. Kimura and colleagues19 showed that dose-dependent abrogation of Runx1 and Runx2 resulted in a strong although highly localized cartilage phenotype in the rib cage, and concluded that Runx1 is essential for the differentiation of mesenchymal cells into chondrocytes through its regulation of Sox5 and Sox6. They reported that Runx1 did not regulate Sox9. Our ChIP data, however, indicate that the chondrogenic transcription factors Sox9, Sox5, and Sox6 are all direct targets of Runx1 during chondrogenesis in vitro. We show that whereas relatively lower occupancy of Sox5, Sox6, and Sox9 promoters by Runx1 is observed at day 4, Sox9 protein levels were relatively higher at the same time point. Based on the data, we speculate that since Runx binding sites on the promoter of Sox9 are potential targets for Runx2 and Runx3 as well as Runx1, endogenous Runx2 and Runx3 may competitively bind the Runx binding sites on these promoters. Future studies will delineate the temporal regulation of Sox genes by Runx1 and potentially the other members of the Runx family of transcription factors.
Based on our results, we conclude that while Runx1 alone in the Prx1-positive mesenchyme may not be required for the normal development of the limb, during fracture healing Runx1 controls early chondrocyte differentiation, possibly through temporal regulation of the Sox genes as well as Runx2 and Runx3.
All authors state that they have no conflicts of interest
We acknowledge NIAMS and NCI for the support to conduct these experiments (NIH-RO1: AR052674-01 to HD and CA89419 to NAS). The University of Connecticut Health Center financed a part of this work. We are grateful to Robert Kosher and Caroline Dealy for sharing the Prx1-Cre transgenic mice.
Authors' roles: DYS, NAS, and HD participated in the study design. DYS, LT, CJM, and MES performed the experiments. DYS, LT, CJM, and MES collected the data. All authors participated in data interpretation and analyses as well as manuscript preparation. All the authors approved the final version of the manuscript.